From 226884b8f583c59087852523f53dd1ed4b73cc3f Mon Sep 17 00:00:00 2001 From: Nico de Poel Date: Fri, 15 Dec 2023 16:26:51 +0100 Subject: [PATCH] Removed old FSR2 shader code --- Assets/Resources/FSR2.meta | 8 - .../FSR2/ffx_fsr2_accumulate_pass.compute | 41 - .../ffx_fsr2_accumulate_pass.compute.meta | 8 - .../ffx_fsr2_autogen_reactive_pass.compute | 32 - ...fx_fsr2_autogen_reactive_pass.compute.meta | 8 - ...sr2_compute_luminance_pyramid_pass.compute | 42 - ...ompute_luminance_pyramid_pass.compute.meta | 8 - .../FSR2/ffx_fsr2_depth_clip_pass.compute | 32 - .../ffx_fsr2_depth_clip_pass.compute.meta | 8 - .../Resources/FSR2/ffx_fsr2_lock_pass.compute | 30 - .../FSR2/ffx_fsr2_lock_pass.compute.meta | 8 - .../Resources/FSR2/ffx_fsr2_rcas_pass.compute | 29 - .../FSR2/ffx_fsr2_rcas_pass.compute.meta | 8 - ...r2_reconstruct_previous_depth_pass.compute | 33 - ...construct_previous_depth_pass.compute.meta | 8 - .../FSR2/ffx_fsr2_tcr_autogen_pass.compute | 32 - .../ffx_fsr2_tcr_autogen_pass.compute.meta | 8 - .../FSR2/ffx_fsr2_unity_common.cginc | 81 - .../FSR2/ffx_fsr2_unity_common.cginc.meta | 27 - Assets/Resources/FSR2/shaders.meta | 8 - .../Resources/FSR2/shaders/ffx_common_types.h | 525 --- .../FSR2/shaders/ffx_common_types.h.meta | 60 - Assets/Resources/FSR2/shaders/ffx_core.h | 52 - Assets/Resources/FSR2/shaders/ffx_core.h.meta | 60 - Assets/Resources/FSR2/shaders/ffx_core_cpu.h | 332 -- .../FSR2/shaders/ffx_core_cpu.h.meta | 60 - .../FSR2/shaders/ffx_core_gpu_common.h | 2784 --------------- .../FSR2/shaders/ffx_core_gpu_common.h.meta | 60 - .../FSR2/shaders/ffx_core_gpu_common_half.h | 2978 ----------------- .../shaders/ffx_core_gpu_common_half.h.meta | 60 - Assets/Resources/FSR2/shaders/ffx_core_hlsl.h | 1502 --------- .../FSR2/shaders/ffx_core_hlsl.h.meta | 60 - .../FSR2/shaders/ffx_core_portability.h | 50 - .../FSR2/shaders/ffx_core_portability.h.meta | 60 - Assets/Resources/FSR2/shaders/ffx_fsr1.h | 1250 ------- Assets/Resources/FSR2/shaders/ffx_fsr1.h.meta | 60 - .../FSR2/shaders/ffx_fsr2_accumulate.h | 295 -- .../FSR2/shaders/ffx_fsr2_accumulate.h.meta | 60 - .../shaders/ffx_fsr2_accumulate_pass.hlsl | 78 - .../ffx_fsr2_accumulate_pass.hlsl.meta | 7 - .../ffx_fsr2_autogen_reactive_pass.hlsl | 85 - .../ffx_fsr2_autogen_reactive_pass.hlsl.meta | 7 - .../FSR2/shaders/ffx_fsr2_callbacks_hlsl.h | 817 ----- .../shaders/ffx_fsr2_callbacks_hlsl.h.meta | 60 - .../Resources/FSR2/shaders/ffx_fsr2_common.h | 565 ---- .../FSR2/shaders/ffx_fsr2_common.h.meta | 60 - .../ffx_fsr2_compute_luminance_pyramid.h | 189 -- .../ffx_fsr2_compute_luminance_pyramid.h.meta | 60 - ...x_fsr2_compute_luminance_pyramid_pass.hlsl | 131 - ...2_compute_luminance_pyramid_pass.hlsl.meta | 7 - .../FSR2/shaders/ffx_fsr2_depth_clip.h | 258 -- .../FSR2/shaders/ffx_fsr2_depth_clip.h.meta | 60 - .../shaders/ffx_fsr2_depth_clip_pass.hlsl | 66 - .../ffx_fsr2_depth_clip_pass.hlsl.meta | 7 - Assets/Resources/FSR2/shaders/ffx_fsr2_lock.h | 115 - .../FSR2/shaders/ffx_fsr2_lock.h.meta | 60 - .../FSR2/shaders/ffx_fsr2_lock_pass.hlsl | 53 - .../FSR2/shaders/ffx_fsr2_lock_pass.hlsl.meta | 7 - .../ffx_fsr2_postprocess_lock_status.h | 106 - .../ffx_fsr2_postprocess_lock_status.h.meta | 60 - Assets/Resources/FSR2/shaders/ffx_fsr2_rcas.h | 67 - .../FSR2/shaders/ffx_fsr2_rcas.h.meta | 60 - .../FSR2/shaders/ffx_fsr2_rcas_pass.hlsl | 75 - .../FSR2/shaders/ffx_fsr2_rcas_pass.hlsl.meta | 7 - ...ruct_dilated_velocity_and_previous_depth.h | 145 - ...dilated_velocity_and_previous_depth.h.meta | 60 - ..._fsr2_reconstruct_previous_depth_pass.hlsl | 63 - ..._reconstruct_previous_depth_pass.hlsl.meta | 7 - .../FSR2/shaders/ffx_fsr2_reproject.h | 136 - .../FSR2/shaders/ffx_fsr2_reproject.h.meta | 60 - .../FSR2/shaders/ffx_fsr2_resources.h | 105 - .../FSR2/shaders/ffx_fsr2_resources.h.meta | 60 - .../Resources/FSR2/shaders/ffx_fsr2_sample.h | 605 ---- .../FSR2/shaders/ffx_fsr2_sample.h.meta | 60 - .../FSR2/shaders/ffx_fsr2_tcr_autogen.h | 250 -- .../FSR2/shaders/ffx_fsr2_tcr_autogen.h.meta | 60 - .../shaders/ffx_fsr2_tcr_autogen_pass.hlsl | 114 - .../ffx_fsr2_tcr_autogen_pass.hlsl.meta | 7 - .../FSR2/shaders/ffx_fsr2_upsample.h | 194 -- .../FSR2/shaders/ffx_fsr2_upsample.h.meta | 60 - Assets/Resources/FSR2/shaders/ffx_spd.h | 929 ----- Assets/Resources/FSR2/shaders/ffx_spd.h.meta | 60 - 82 files changed, 16809 deletions(-) delete mode 100644 Assets/Resources/FSR2.meta delete mode 100644 Assets/Resources/FSR2/ffx_fsr2_accumulate_pass.compute delete mode 100644 Assets/Resources/FSR2/ffx_fsr2_accumulate_pass.compute.meta delete mode 100644 Assets/Resources/FSR2/ffx_fsr2_autogen_reactive_pass.compute delete mode 100644 Assets/Resources/FSR2/ffx_fsr2_autogen_reactive_pass.compute.meta delete mode 100644 Assets/Resources/FSR2/ffx_fsr2_compute_luminance_pyramid_pass.compute delete mode 100644 Assets/Resources/FSR2/ffx_fsr2_compute_luminance_pyramid_pass.compute.meta delete mode 100644 Assets/Resources/FSR2/ffx_fsr2_depth_clip_pass.compute delete mode 100644 Assets/Resources/FSR2/ffx_fsr2_depth_clip_pass.compute.meta delete mode 100644 Assets/Resources/FSR2/ffx_fsr2_lock_pass.compute delete mode 100644 Assets/Resources/FSR2/ffx_fsr2_lock_pass.compute.meta delete mode 100644 Assets/Resources/FSR2/ffx_fsr2_rcas_pass.compute delete mode 100644 Assets/Resources/FSR2/ffx_fsr2_rcas_pass.compute.meta delete mode 100644 Assets/Resources/FSR2/ffx_fsr2_reconstruct_previous_depth_pass.compute delete mode 100644 Assets/Resources/FSR2/ffx_fsr2_reconstruct_previous_depth_pass.compute.meta delete mode 100644 Assets/Resources/FSR2/ffx_fsr2_tcr_autogen_pass.compute delete mode 100644 Assets/Resources/FSR2/ffx_fsr2_tcr_autogen_pass.compute.meta delete mode 100644 Assets/Resources/FSR2/ffx_fsr2_unity_common.cginc delete mode 100644 Assets/Resources/FSR2/ffx_fsr2_unity_common.cginc.meta delete mode 100644 Assets/Resources/FSR2/shaders.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_common_types.h delete mode 100644 Assets/Resources/FSR2/shaders/ffx_common_types.h.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_core.h delete mode 100644 Assets/Resources/FSR2/shaders/ffx_core.h.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_core_cpu.h delete mode 100644 Assets/Resources/FSR2/shaders/ffx_core_cpu.h.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_core_gpu_common.h delete mode 100644 Assets/Resources/FSR2/shaders/ffx_core_gpu_common.h.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_core_gpu_common_half.h delete mode 100644 Assets/Resources/FSR2/shaders/ffx_core_gpu_common_half.h.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_core_hlsl.h delete mode 100644 Assets/Resources/FSR2/shaders/ffx_core_hlsl.h.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_core_portability.h delete mode 100644 Assets/Resources/FSR2/shaders/ffx_core_portability.h.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr1.h delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr1.h.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_accumulate.h delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_accumulate.h.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_accumulate_pass.hlsl delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_accumulate_pass.hlsl.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_autogen_reactive_pass.hlsl delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_autogen_reactive_pass.hlsl.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_callbacks_hlsl.h delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_callbacks_hlsl.h.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_common.h delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_common.h.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_compute_luminance_pyramid.h delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_compute_luminance_pyramid.h.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_compute_luminance_pyramid_pass.hlsl delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_compute_luminance_pyramid_pass.hlsl.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_depth_clip.h delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_depth_clip.h.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_depth_clip_pass.hlsl delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_depth_clip_pass.hlsl.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_lock.h delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_lock.h.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_lock_pass.hlsl delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_lock_pass.hlsl.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_postprocess_lock_status.h delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_postprocess_lock_status.h.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_rcas.h delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_rcas.h.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_rcas_pass.hlsl delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_rcas_pass.hlsl.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_reconstruct_dilated_velocity_and_previous_depth.h delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_reconstruct_dilated_velocity_and_previous_depth.h.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_reconstruct_previous_depth_pass.hlsl delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_reconstruct_previous_depth_pass.hlsl.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_reproject.h delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_reproject.h.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_resources.h delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_resources.h.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_sample.h delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_sample.h.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_tcr_autogen.h delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_tcr_autogen.h.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_tcr_autogen_pass.hlsl delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_tcr_autogen_pass.hlsl.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_upsample.h delete mode 100644 Assets/Resources/FSR2/shaders/ffx_fsr2_upsample.h.meta delete mode 100644 Assets/Resources/FSR2/shaders/ffx_spd.h delete mode 100644 Assets/Resources/FSR2/shaders/ffx_spd.h.meta diff --git a/Assets/Resources/FSR2.meta b/Assets/Resources/FSR2.meta deleted file mode 100644 index 6d47067..0000000 --- a/Assets/Resources/FSR2.meta +++ /dev/null @@ -1,8 +0,0 @@ -fileFormatVersion: 2 -guid: cad7d53fa2166a0449bec7a9b4f17d69 -folderAsset: yes -DefaultImporter: - externalObjects: {} - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/ffx_fsr2_accumulate_pass.compute b/Assets/Resources/FSR2/ffx_fsr2_accumulate_pass.compute deleted file mode 100644 index a359485..0000000 --- a/Assets/Resources/FSR2/ffx_fsr2_accumulate_pass.compute +++ /dev/null @@ -1,41 +0,0 @@ -// Copyright (c) 2023 Nico de Poel -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// -// The above copyright notice and this permission notice shall be included in all -// copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#pragma kernel CS - -#pragma multi_compile_local __ FFX_HALF -#pragma multi_compile_local __ FFX_FSR2_OPTION_REPROJECT_USE_LANCZOS_TYPE -#pragma multi_compile_local __ FFX_FSR2_OPTION_HDR_COLOR_INPUT -#pragma multi_compile_local __ FFX_FSR2_OPTION_LOW_RESOLUTION_MOTION_VECTORS -#pragma multi_compile_local __ FFX_FSR2_OPTION_JITTERED_MOTION_VECTORS -#pragma multi_compile_local __ FFX_FSR2_OPTION_INVERTED_DEPTH -#pragma multi_compile_local __ FFX_FSR2_OPTION_APPLY_SHARPENING - -#pragma multi_compile_local __ UNITY_FSR2_HDRP - -#include "ffx_fsr2_unity_common.cginc" - -// Ensure the correct value is defined for this keyword, as it is used to select one of multiple sampler functions -#ifdef FFX_FSR2_OPTION_REPROJECT_USE_LANCZOS_TYPE -#undef FFX_FSR2_OPTION_REPROJECT_USE_LANCZOS_TYPE -#define FFX_FSR2_OPTION_REPROJECT_USE_LANCZOS_TYPE 1 -#endif - -#include "shaders/ffx_fsr2_accumulate_pass.hlsl" diff --git a/Assets/Resources/FSR2/ffx_fsr2_accumulate_pass.compute.meta b/Assets/Resources/FSR2/ffx_fsr2_accumulate_pass.compute.meta deleted file mode 100644 index 3864379..0000000 --- a/Assets/Resources/FSR2/ffx_fsr2_accumulate_pass.compute.meta +++ /dev/null @@ -1,8 +0,0 @@ -fileFormatVersion: 2 -guid: 702560780e923c84394c9b22ba460a9c -ComputeShaderImporter: - externalObjects: {} - preprocessorOverride: 0 - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/ffx_fsr2_autogen_reactive_pass.compute b/Assets/Resources/FSR2/ffx_fsr2_autogen_reactive_pass.compute deleted file mode 100644 index d27a02d..0000000 --- a/Assets/Resources/FSR2/ffx_fsr2_autogen_reactive_pass.compute +++ /dev/null @@ -1,32 +0,0 @@ -// Copyright (c) 2023 Nico de Poel -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// -// The above copyright notice and this permission notice shall be included in all -// copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#pragma kernel CS - -#pragma multi_compile_local __ FFX_HALF -#pragma multi_compile_local __ FFX_FSR2_OPTION_LOW_RESOLUTION_MOTION_VECTORS -#pragma multi_compile_local __ FFX_FSR2_OPTION_JITTERED_MOTION_VECTORS -#pragma multi_compile_local __ FFX_FSR2_OPTION_INVERTED_DEPTH - -#pragma multi_compile_local __ UNITY_FSR2_HDRP - -#include "ffx_fsr2_unity_common.cginc" - -#include "shaders/ffx_fsr2_autogen_reactive_pass.hlsl" diff --git a/Assets/Resources/FSR2/ffx_fsr2_autogen_reactive_pass.compute.meta b/Assets/Resources/FSR2/ffx_fsr2_autogen_reactive_pass.compute.meta deleted file mode 100644 index 19d1df3..0000000 --- a/Assets/Resources/FSR2/ffx_fsr2_autogen_reactive_pass.compute.meta +++ /dev/null @@ -1,8 +0,0 @@ -fileFormatVersion: 2 -guid: d18fb8811ca4753469c439784546104e -ComputeShaderImporter: - externalObjects: {} - preprocessorOverride: 0 - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/ffx_fsr2_compute_luminance_pyramid_pass.compute b/Assets/Resources/FSR2/ffx_fsr2_compute_luminance_pyramid_pass.compute deleted file mode 100644 index 385dfe4..0000000 --- a/Assets/Resources/FSR2/ffx_fsr2_compute_luminance_pyramid_pass.compute +++ /dev/null @@ -1,42 +0,0 @@ -// Copyright (c) 2023 Nico de Poel -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// -// The above copyright notice and this permission notice shall be included in all -// copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#pragma kernel CS - -//#pragma multi_compile_local __ FFX_HALF // causes a hard-coded error message from the shader include ¯\_(ツ)_/¯ -#pragma multi_compile_local __ FFX_FSR2_OPTION_LOW_RESOLUTION_MOTION_VECTORS -#pragma multi_compile_local __ FFX_FSR2_OPTION_JITTERED_MOTION_VECTORS -#pragma multi_compile_local __ FFX_FSR2_OPTION_INVERTED_DEPTH - -#pragma multi_compile_local __ UNITY_FSR2_HDRP - -#include "ffx_fsr2_unity_common.cginc" - -// Wave operations require shader model 6.0; this can only be enabled when using DXC on D3D12 -// These pragmas are commented out by default as Unity will sometimes ignore the #if's and try to enable these features anyway. -// Uncomment the below lines if you intend to try wave operations on DX12 with the DXC compiler. -//#if defined(UNITY_COMPILER_DXC) && defined(SHADER_API_D3D12) -//#pragma require WaveBasic // Required for WaveGetLaneIndex -//#pragma require WaveBallot // Required for WaveReadLaneAt -//#else -#define SPD_NO_WAVE_OPERATIONS -//#endif - -#include "shaders/ffx_fsr2_compute_luminance_pyramid_pass.hlsl" diff --git a/Assets/Resources/FSR2/ffx_fsr2_compute_luminance_pyramid_pass.compute.meta b/Assets/Resources/FSR2/ffx_fsr2_compute_luminance_pyramid_pass.compute.meta deleted file mode 100644 index 49033cb..0000000 --- a/Assets/Resources/FSR2/ffx_fsr2_compute_luminance_pyramid_pass.compute.meta +++ /dev/null @@ -1,8 +0,0 @@ -fileFormatVersion: 2 -guid: 0894ebeefb6aa7d4680c71dffbda3fee -ComputeShaderImporter: - externalObjects: {} - preprocessorOverride: 0 - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/ffx_fsr2_depth_clip_pass.compute b/Assets/Resources/FSR2/ffx_fsr2_depth_clip_pass.compute deleted file mode 100644 index 4e41933..0000000 --- a/Assets/Resources/FSR2/ffx_fsr2_depth_clip_pass.compute +++ /dev/null @@ -1,32 +0,0 @@ -// Copyright (c) 2023 Nico de Poel -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// -// The above copyright notice and this permission notice shall be included in all -// copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#pragma kernel CS - -#pragma multi_compile_local __ FFX_HALF -#pragma multi_compile_local __ FFX_FSR2_OPTION_LOW_RESOLUTION_MOTION_VECTORS -#pragma multi_compile_local __ FFX_FSR2_OPTION_JITTERED_MOTION_VECTORS -#pragma multi_compile_local __ FFX_FSR2_OPTION_INVERTED_DEPTH - -#pragma multi_compile_local __ UNITY_FSR2_HDRP - -#include "ffx_fsr2_unity_common.cginc" - -#include "shaders/ffx_fsr2_depth_clip_pass.hlsl" diff --git a/Assets/Resources/FSR2/ffx_fsr2_depth_clip_pass.compute.meta b/Assets/Resources/FSR2/ffx_fsr2_depth_clip_pass.compute.meta deleted file mode 100644 index 26fb9ba..0000000 --- a/Assets/Resources/FSR2/ffx_fsr2_depth_clip_pass.compute.meta +++ /dev/null @@ -1,8 +0,0 @@ -fileFormatVersion: 2 -guid: 8026d9d8542eab6499c32d5d46beb2b6 -ComputeShaderImporter: - externalObjects: {} - preprocessorOverride: 0 - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/ffx_fsr2_lock_pass.compute b/Assets/Resources/FSR2/ffx_fsr2_lock_pass.compute deleted file mode 100644 index 5f8c13d..0000000 --- a/Assets/Resources/FSR2/ffx_fsr2_lock_pass.compute +++ /dev/null @@ -1,30 +0,0 @@ -// Copyright (c) 2023 Nico de Poel -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// -// The above copyright notice and this permission notice shall be included in all -// copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#pragma kernel CS - -#pragma multi_compile_local __ FFX_HALF -#pragma multi_compile_local __ FFX_FSR2_OPTION_LOW_RESOLUTION_MOTION_VECTORS -#pragma multi_compile_local __ FFX_FSR2_OPTION_JITTERED_MOTION_VECTORS -#pragma multi_compile_local __ FFX_FSR2_OPTION_INVERTED_DEPTH - -#include "ffx_fsr2_unity_common.cginc" - -#include "shaders/ffx_fsr2_lock_pass.hlsl" diff --git a/Assets/Resources/FSR2/ffx_fsr2_lock_pass.compute.meta b/Assets/Resources/FSR2/ffx_fsr2_lock_pass.compute.meta deleted file mode 100644 index a197b49..0000000 --- a/Assets/Resources/FSR2/ffx_fsr2_lock_pass.compute.meta +++ /dev/null @@ -1,8 +0,0 @@ -fileFormatVersion: 2 -guid: 3c96d72b39a840c428c901628dab92c0 -ComputeShaderImporter: - externalObjects: {} - preprocessorOverride: 0 - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/ffx_fsr2_rcas_pass.compute b/Assets/Resources/FSR2/ffx_fsr2_rcas_pass.compute deleted file mode 100644 index 6d97414..0000000 --- a/Assets/Resources/FSR2/ffx_fsr2_rcas_pass.compute +++ /dev/null @@ -1,29 +0,0 @@ -// Copyright (c) 2023 Nico de Poel -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// -// The above copyright notice and this permission notice shall be included in all -// copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#pragma kernel CS - -#pragma multi_compile_local __ FFX_FSR2_OPTION_LOW_RESOLUTION_MOTION_VECTORS -#pragma multi_compile_local __ FFX_FSR2_OPTION_JITTERED_MOTION_VECTORS -#pragma multi_compile_local __ FFX_FSR2_OPTION_INVERTED_DEPTH - -#include "ffx_fsr2_unity_common.cginc" - -#include "shaders/ffx_fsr2_rcas_pass.hlsl" diff --git a/Assets/Resources/FSR2/ffx_fsr2_rcas_pass.compute.meta b/Assets/Resources/FSR2/ffx_fsr2_rcas_pass.compute.meta deleted file mode 100644 index ea6c978..0000000 --- a/Assets/Resources/FSR2/ffx_fsr2_rcas_pass.compute.meta +++ /dev/null @@ -1,8 +0,0 @@ -fileFormatVersion: 2 -guid: 5a82801f160ff6a4eb47db567216e592 -ComputeShaderImporter: - externalObjects: {} - preprocessorOverride: 0 - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/ffx_fsr2_reconstruct_previous_depth_pass.compute b/Assets/Resources/FSR2/ffx_fsr2_reconstruct_previous_depth_pass.compute deleted file mode 100644 index c6cc99f..0000000 --- a/Assets/Resources/FSR2/ffx_fsr2_reconstruct_previous_depth_pass.compute +++ /dev/null @@ -1,33 +0,0 @@ -// Copyright (c) 2023 Nico de Poel -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// -// The above copyright notice and this permission notice shall be included in all -// copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#pragma kernel CS - -#pragma multi_compile_local __ FFX_HALF -#pragma multi_compile_local __ FFX_FSR2_OPTION_HDR_COLOR_INPUT -#pragma multi_compile_local __ FFX_FSR2_OPTION_LOW_RESOLUTION_MOTION_VECTORS -#pragma multi_compile_local __ FFX_FSR2_OPTION_JITTERED_MOTION_VECTORS -#pragma multi_compile_local __ FFX_FSR2_OPTION_INVERTED_DEPTH - -#pragma multi_compile_local __ UNITY_FSR2_HDRP - -#include "ffx_fsr2_unity_common.cginc" - -#include "shaders/ffx_fsr2_reconstruct_previous_depth_pass.hlsl" diff --git a/Assets/Resources/FSR2/ffx_fsr2_reconstruct_previous_depth_pass.compute.meta b/Assets/Resources/FSR2/ffx_fsr2_reconstruct_previous_depth_pass.compute.meta deleted file mode 100644 index 6396b86..0000000 --- a/Assets/Resources/FSR2/ffx_fsr2_reconstruct_previous_depth_pass.compute.meta +++ /dev/null @@ -1,8 +0,0 @@ -fileFormatVersion: 2 -guid: 6ef1b4c25874e334dad5ba3ac6345e32 -ComputeShaderImporter: - externalObjects: {} - preprocessorOverride: 0 - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/ffx_fsr2_tcr_autogen_pass.compute b/Assets/Resources/FSR2/ffx_fsr2_tcr_autogen_pass.compute deleted file mode 100644 index 62d386b..0000000 --- a/Assets/Resources/FSR2/ffx_fsr2_tcr_autogen_pass.compute +++ /dev/null @@ -1,32 +0,0 @@ -// Copyright (c) 2023 Nico de Poel -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// -// The above copyright notice and this permission notice shall be included in all -// copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#pragma kernel CS - -#pragma multi_compile_local __ FFX_HALF -#pragma multi_compile_local __ FFX_FSR2_OPTION_LOW_RESOLUTION_MOTION_VECTORS -#pragma multi_compile_local __ FFX_FSR2_OPTION_JITTERED_MOTION_VECTORS -#pragma multi_compile_local __ FFX_FSR2_OPTION_INVERTED_DEPTH - -#pragma multi_compile_local __ UNITY_FSR2_HDRP - -#include "ffx_fsr2_unity_common.cginc" - -#include "shaders/ffx_fsr2_tcr_autogen_pass.hlsl" diff --git a/Assets/Resources/FSR2/ffx_fsr2_tcr_autogen_pass.compute.meta b/Assets/Resources/FSR2/ffx_fsr2_tcr_autogen_pass.compute.meta deleted file mode 100644 index 0002090..0000000 --- a/Assets/Resources/FSR2/ffx_fsr2_tcr_autogen_pass.compute.meta +++ /dev/null @@ -1,8 +0,0 @@ -fileFormatVersion: 2 -guid: b478fba0a6a87b44b8be7c35deb5f0dc -ComputeShaderImporter: - externalObjects: {} - preprocessorOverride: 0 - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/ffx_fsr2_unity_common.cginc b/Assets/Resources/FSR2/ffx_fsr2_unity_common.cginc deleted file mode 100644 index 98c5b17..0000000 --- a/Assets/Resources/FSR2/ffx_fsr2_unity_common.cginc +++ /dev/null @@ -1,81 +0,0 @@ -// Copyright (c) 2023 Nico de Poel -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// -// The above copyright notice and this permission notice shall be included in all -// copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -// Suppress a few warnings produced by FFX's HLSL code -#pragma warning(disable: 3078) // Loop control variable conflicts -#pragma warning(disable: 3203) // Signed/unsigned mismatch - -#define FFX_GPU // Compiling for GPU -#define FFX_HLSL // Compile for plain HLSL - -// Use the DXC shader compiler on modern graphics APIs to enable a few advanced features -// The DXC-related pragmas are disabled by default, as DXC doesn't support all platforms yet and will break on some platforms when enabled. -// Consider this to be an experimental feature. If you want to benefit from 16-bit floating point and wave operations, and don't care about supporting older graphics APIs, then it's worth a try. -//#if defined(SHADER_API_D3D12) || defined(SHADER_API_VULKAN) || defined(SHADER_API_METAL) -//#pragma use_dxc // Using DXC will currently break DX11 support since DX11 and DX12 share the same shader bytecode in Unity. -//#endif - -// Enable half precision data types on platforms that support it -//#if defined(UNITY_COMPILER_DXC) && defined(FFX_HALF) -//#pragma require Native16Bit -//#endif - -// Hack to work around the lack of texture atomics on Metal -#if defined(SHADER_API_METAL) -#define InterlockedAdd(dest, val, orig) { (orig) = (dest); (dest) += (val); } -#define InterlockedMin(dest, val) { (dest) = min((dest), (val)); } -#define InterlockedMax(dest, val) { (dest) = max((dest), (val)); } -#endif - -// Workaround for HDRP using texture arrays for its camera buffers on some platforms -// The below defines are copied from: Packages/com.unity.render-pipelines.high-definition/Runtime/ShaderLibrary/TextureXR.hlsl -#if defined(UNITY_FSR2_HDRP) - // Must be in sync with C# with property useTexArray in TextureXR.cs - #if ((defined(SHADER_API_D3D11) || defined(SHADER_API_D3D12)) && !defined(SHADER_API_XBOXONE) && !defined(SHADER_API_GAMECORE)) || defined(SHADER_API_PSSL) || defined(SHADER_API_VULKAN) - #define UNITY_TEXTURE2D_X_ARRAY_SUPPORTED - #endif - - // Control if TEXTURE2D_X macros will expand to texture arrays - #if defined(UNITY_TEXTURE2D_X_ARRAY_SUPPORTED) && !defined(DISABLE_TEXTURE2D_X_ARRAY) - #define USE_TEXTURE2D_X_AS_ARRAY - #endif - - // Early defines for single-pass instancing - #if defined(STEREO_INSTANCING_ON) && defined(UNITY_TEXTURE2D_X_ARRAY_SUPPORTED) - #define UNITY_STEREO_INSTANCING_ENABLED - #endif - - // Helper macros to handle XR single-pass with Texture2DArray - #if defined(USE_TEXTURE2D_X_AS_ARRAY) - - // Only single-pass stereo instancing used array indexing - #if defined(UNITY_STEREO_INSTANCING_ENABLED) - #define SLICE_ARRAY_INDEX unity_StereoEyeIndex - #else - #define SLICE_ARRAY_INDEX 0 - #endif - - // Declare and sample camera buffers as texture arrays - #define UNITY_FSR2_TEX2D(type) Texture2DArray - #define UNITY_FSR2_POS(pxPos) FfxUInt32x3(pxPos, SLICE_ARRAY_INDEX) - #define UNITY_FSR2_UV(uv) FfxFloat32x3(uv, SLICE_ARRAY_INDEX) - - #endif -#endif diff --git a/Assets/Resources/FSR2/ffx_fsr2_unity_common.cginc.meta b/Assets/Resources/FSR2/ffx_fsr2_unity_common.cginc.meta deleted file mode 100644 index 8393144..0000000 --- a/Assets/Resources/FSR2/ffx_fsr2_unity_common.cginc.meta +++ /dev/null @@ -1,27 +0,0 @@ -fileFormatVersion: 2 -guid: bb2d4d4671c448698877526c29f2fc99 -PluginImporter: - externalObjects: {} - serializedVersion: 2 - iconMap: {} - executionOrder: {} - defineConstraints: [] - isPreloaded: 0 - isOverridable: 0 - isExplicitlyReferenced: 0 - validateReferences: 1 - platformData: - - first: - Any: - second: - enabled: 1 - settings: {} - - first: - Editor: Editor - second: - enabled: 0 - settings: - DefaultValueInitialized: true - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders.meta b/Assets/Resources/FSR2/shaders.meta deleted file mode 100644 index 676ff9a..0000000 --- a/Assets/Resources/FSR2/shaders.meta +++ /dev/null @@ -1,8 +0,0 @@ -fileFormatVersion: 2 -guid: 4a24e63edc822264a871f58397325d51 -folderAsset: yes -DefaultImporter: - externalObjects: {} - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_common_types.h b/Assets/Resources/FSR2/shaders/ffx_common_types.h deleted file mode 100644 index 4594dc1..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_common_types.h +++ /dev/null @@ -1,525 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. -#ifndef FFX_COMMON_TYPES_H -#define FFX_COMMON_TYPES_H - -#if defined(FFX_CPU) -#define FFX_PARAMETER_IN -#define FFX_PARAMETER_OUT -#define FFX_PARAMETER_INOUT -#elif defined(FFX_HLSL) -#define FFX_PARAMETER_IN in -#define FFX_PARAMETER_OUT out -#define FFX_PARAMETER_INOUT inout -#elif defined(FFX_GLSL) -#define FFX_PARAMETER_IN in -#define FFX_PARAMETER_OUT out -#define FFX_PARAMETER_INOUT inout -#endif // #if defined(FFX_CPU) - -#if defined(FFX_CPU) -/// A typedef for a boolean value. -/// -/// @ingroup CPU -typedef bool FfxBoolean; - -/// A typedef for a unsigned 8bit integer. -/// -/// @ingroup CPU -typedef uint8_t FfxUInt8; - -/// A typedef for a unsigned 16bit integer. -/// -/// @ingroup CPU -typedef uint16_t FfxUInt16; - -/// A typedef for a unsigned 32bit integer. -/// -/// @ingroup CPU -typedef uint32_t FfxUInt32; - -/// A typedef for a unsigned 64bit integer. -/// -/// @ingroup CPU -typedef uint64_t FfxUInt64; - -/// A typedef for a signed 8bit integer. -/// -/// @ingroup CPU -typedef int8_t FfxInt8; - -/// A typedef for a signed 16bit integer. -/// -/// @ingroup CPU -typedef int16_t FfxInt16; - -/// A typedef for a signed 32bit integer. -/// -/// @ingroup CPU -typedef int32_t FfxInt32; - -/// A typedef for a signed 64bit integer. -/// -/// @ingroup CPU -typedef int64_t FfxInt64; - -/// A typedef for a floating point value. -/// -/// @ingroup CPU -typedef float FfxFloat32; - -/// A typedef for a 2-dimensional floating point value. -/// -/// @ingroup CPU -typedef float FfxFloat32x2[2]; - -/// A typedef for a 3-dimensional floating point value. -/// -/// @ingroup CPU -typedef float FfxFloat32x3[3]; - -/// A typedef for a 4-dimensional floating point value. -/// -/// @ingroup CPU -typedef float FfxFloat32x4[4]; - -/// A typedef for a 2-dimensional 32bit unsigned integer. -/// -/// @ingroup CPU -typedef uint32_t FfxUInt32x2[2]; - -/// A typedef for a 3-dimensional 32bit unsigned integer. -/// -/// @ingroup CPU -typedef uint32_t FfxUInt32x3[3]; - -/// A typedef for a 4-dimensional 32bit unsigned integer. -/// -/// @ingroup CPU -typedef uint32_t FfxUInt32x4[4]; -#endif // #if defined(FFX_CPU) - -#if defined(FFX_HLSL) -/// A typedef for a boolean value. -/// -/// @ingroup GPU -typedef bool FfxBoolean; - -#if FFX_HLSL_6_2 -typedef float32_t FfxFloat32; -typedef float32_t2 FfxFloat32x2; -typedef float32_t3 FfxFloat32x3; -typedef float32_t4 FfxFloat32x4; - -/// A typedef for a unsigned 32bit integer. -/// -/// @ingroup GPU -typedef uint32_t FfxUInt32; -typedef uint32_t2 FfxUInt32x2; -typedef uint32_t3 FfxUInt32x3; -typedef uint32_t4 FfxUInt32x4; -typedef int32_t FfxInt32; -typedef int32_t2 FfxInt32x2; -typedef int32_t3 FfxInt32x3; -typedef int32_t4 FfxInt32x4; -#else -#define FfxFloat32 float -#define FfxFloat32x2 float2 -#define FfxFloat32x3 float3 -#define FfxFloat32x4 float4 - -/// A typedef for a unsigned 32bit integer. -/// -/// @ingroup GPU -typedef uint FfxUInt32; -typedef uint2 FfxUInt32x2; -typedef uint3 FfxUInt32x3; -typedef uint4 FfxUInt32x4; -typedef int FfxInt32; -typedef int2 FfxInt32x2; -typedef int3 FfxInt32x3; -typedef int4 FfxInt32x4; -#endif // #if defined(FFX_HLSL_6_2) - -#if FFX_HALF -#if FFX_HLSL_6_2 -typedef float16_t FfxFloat16; -typedef float16_t2 FfxFloat16x2; -typedef float16_t3 FfxFloat16x3; -typedef float16_t4 FfxFloat16x4; - -/// A typedef for an unsigned 16bit integer. -/// -/// @ingroup GPU -typedef uint16_t FfxUInt16; -typedef uint16_t2 FfxUInt16x2; -typedef uint16_t3 FfxUInt16x3; -typedef uint16_t4 FfxUInt16x4; - -/// A typedef for a signed 16bit integer. -/// -/// @ingroup GPU -typedef int16_t FfxInt16; -typedef int16_t2 FfxInt16x2; -typedef int16_t3 FfxInt16x3; -typedef int16_t4 FfxInt16x4; -#elif SHADER_API_PSSL -#pragma argument(realtypes) // Enable true 16-bit types - -typedef half FfxFloat16; -typedef half2 FfxFloat16x2; -typedef half3 FfxFloat16x3; -typedef half4 FfxFloat16x4; - -/// A typedef for an unsigned 16bit integer. -/// -/// @ingroup GPU -typedef ushort FfxUInt16; -typedef ushort2 FfxUInt16x2; -typedef ushort3 FfxUInt16x3; -typedef ushort4 FfxUInt16x4; - -/// A typedef for a signed 16bit integer. -/// -/// @ingroup GPU -typedef short FfxInt16; -typedef short2 FfxInt16x2; -typedef short3 FfxInt16x3; -typedef short4 FfxInt16x4; -#else -typedef min16float FfxFloat16; -typedef min16float2 FfxFloat16x2; -typedef min16float3 FfxFloat16x3; -typedef min16float4 FfxFloat16x4; - -/// A typedef for an unsigned 16bit integer. -/// -/// @ingroup GPU -typedef min16uint FfxUInt16; -typedef min16uint2 FfxUInt16x2; -typedef min16uint3 FfxUInt16x3; -typedef min16uint4 FfxUInt16x4; - -/// A typedef for a signed 16bit integer. -/// -/// @ingroup GPU -typedef min16int FfxInt16; -typedef min16int2 FfxInt16x2; -typedef min16int3 FfxInt16x3; -typedef min16int4 FfxInt16x4; -#endif // FFX_HLSL_6_2 -#endif // FFX_HALF -#endif // #if defined(FFX_HLSL) - -#if defined(FFX_GLSL) -/// A typedef for a boolean value. -/// -/// @ingroup GPU -#define FfxBoolean bool -#define FfxFloat32 float -#define FfxFloat32x2 vec2 -#define FfxFloat32x3 vec3 -#define FfxFloat32x4 vec4 -#define FfxUInt32 uint -#define FfxUInt32x2 uvec2 -#define FfxUInt32x3 uvec3 -#define FfxUInt32x4 uvec4 -#define FfxInt32 int -#define FfxInt32x2 ivec2 -#define FfxInt32x3 ivec3 -#define FfxInt32x4 ivec4 -#if FFX_HALF -#define FfxFloat16 float16_t -#define FfxFloat16x2 f16vec2 -#define FfxFloat16x3 f16vec3 -#define FfxFloat16x4 f16vec4 -#define FfxUInt16 uint16_t -#define FfxUInt16x2 u16vec2 -#define FfxUInt16x3 u16vec3 -#define FfxUInt16x4 u16vec4 -#define FfxInt16 int16_t -#define FfxInt16x2 i16vec2 -#define FfxInt16x3 i16vec3 -#define FfxInt16x4 i16vec4 -#endif // FFX_HALF -#endif // #if defined(FFX_GLSL) - -// Global toggles: -// #define FFX_HALF (1) -// #define FFX_HLSL_6_2 (1) - -#if defined(FFX_HALF) && !defined(SHADER_API_PSSL) - -#if FFX_HLSL_6_2 - -#define FFX_MIN16_SCALAR( TypeName, BaseComponentType ) typedef BaseComponentType##16_t TypeName; -#define FFX_MIN16_VECTOR( TypeName, BaseComponentType, COL ) typedef vector TypeName; -#define FFX_MIN16_MATRIX( TypeName, BaseComponentType, ROW, COL ) typedef matrix TypeName; - -#define FFX_16BIT_SCALAR( TypeName, BaseComponentType ) typedef BaseComponentType##16_t TypeName; -#define FFX_16BIT_VECTOR( TypeName, BaseComponentType, COL ) typedef vector TypeName; -#define FFX_16BIT_MATRIX( TypeName, BaseComponentType, ROW, COL ) typedef matrix TypeName; - -#else //FFX_HLSL_6_2 - -#define FFX_MIN16_SCALAR( TypeName, BaseComponentType ) typedef min16##BaseComponentType TypeName; -#define FFX_MIN16_VECTOR( TypeName, BaseComponentType, COL ) typedef vector TypeName; -#define FFX_MIN16_MATRIX( TypeName, BaseComponentType, ROW, COL ) typedef matrix TypeName; - -#define FFX_16BIT_SCALAR( TypeName, BaseComponentType ) FFX_MIN16_SCALAR( TypeName, BaseComponentType ); -#define FFX_16BIT_VECTOR( TypeName, BaseComponentType, COL ) FFX_MIN16_VECTOR( TypeName, BaseComponentType, COL ); -#define FFX_16BIT_MATRIX( TypeName, BaseComponentType, ROW, COL ) FFX_MIN16_MATRIX( TypeName, BaseComponentType, ROW, COL ); - -#endif //FFX_HLSL_6_2 - -#else //FFX_HALF - -#define FFX_MIN16_SCALAR( TypeName, BaseComponentType ) typedef BaseComponentType TypeName; -#define FFX_MIN16_VECTOR( TypeName, BaseComponentType, COL ) typedef vector TypeName; -#define FFX_MIN16_MATRIX( TypeName, BaseComponentType, ROW, COL ) typedef matrix TypeName; - -#define FFX_16BIT_SCALAR( TypeName, BaseComponentType ) typedef BaseComponentType TypeName; -#define FFX_16BIT_VECTOR( TypeName, BaseComponentType, COL ) typedef vector TypeName; -#define FFX_16BIT_MATRIX( TypeName, BaseComponentType, ROW, COL ) typedef matrix TypeName; - -#endif //FFX_HALF - -#if defined(FFX_GPU) -// Common typedefs: -#if defined(FFX_HLSL) && !defined(SHADER_API_PSSL) -FFX_MIN16_SCALAR( FFX_MIN16_F , float ); -FFX_MIN16_VECTOR( FFX_MIN16_F2, float, 2 ); -FFX_MIN16_VECTOR( FFX_MIN16_F3, float, 3 ); -FFX_MIN16_VECTOR( FFX_MIN16_F4, float, 4 ); - -FFX_MIN16_SCALAR( FFX_MIN16_I, int ); -FFX_MIN16_VECTOR( FFX_MIN16_I2, int, 2 ); -FFX_MIN16_VECTOR( FFX_MIN16_I3, int, 3 ); -FFX_MIN16_VECTOR( FFX_MIN16_I4, int, 4 ); - -FFX_MIN16_SCALAR( FFX_MIN16_U, uint ); -FFX_MIN16_VECTOR( FFX_MIN16_U2, uint, 2 ); -FFX_MIN16_VECTOR( FFX_MIN16_U3, uint, 3 ); -FFX_MIN16_VECTOR( FFX_MIN16_U4, uint, 4 ); - -FFX_16BIT_SCALAR( FFX_F16_t , float ); -FFX_16BIT_VECTOR( FFX_F16_t2, float, 2 ); -FFX_16BIT_VECTOR( FFX_F16_t3, float, 3 ); -FFX_16BIT_VECTOR( FFX_F16_t4, float, 4 ); - -FFX_16BIT_SCALAR( FFX_I16_t, int ); -FFX_16BIT_VECTOR( FFX_I16_t2, int, 2 ); -FFX_16BIT_VECTOR( FFX_I16_t3, int, 3 ); -FFX_16BIT_VECTOR( FFX_I16_t4, int, 4 ); - -FFX_16BIT_SCALAR( FFX_U16_t, uint ); -FFX_16BIT_VECTOR( FFX_U16_t2, uint, 2 ); -FFX_16BIT_VECTOR( FFX_U16_t3, uint, 3 ); -FFX_16BIT_VECTOR( FFX_U16_t4, uint, 4 ); - -#define TYPEDEF_MIN16_TYPES(Prefix) \ -typedef FFX_MIN16_F Prefix##_F; \ -typedef FFX_MIN16_F2 Prefix##_F2; \ -typedef FFX_MIN16_F3 Prefix##_F3; \ -typedef FFX_MIN16_F4 Prefix##_F4; \ -typedef FFX_MIN16_I Prefix##_I; \ -typedef FFX_MIN16_I2 Prefix##_I2; \ -typedef FFX_MIN16_I3 Prefix##_I3; \ -typedef FFX_MIN16_I4 Prefix##_I4; \ -typedef FFX_MIN16_U Prefix##_U; \ -typedef FFX_MIN16_U2 Prefix##_U2; \ -typedef FFX_MIN16_U3 Prefix##_U3; \ -typedef FFX_MIN16_U4 Prefix##_U4; - -#define TYPEDEF_16BIT_TYPES(Prefix) \ -typedef FFX_16BIT_F Prefix##_F; \ -typedef FFX_16BIT_F2 Prefix##_F2; \ -typedef FFX_16BIT_F3 Prefix##_F3; \ -typedef FFX_16BIT_F4 Prefix##_F4; \ -typedef FFX_16BIT_I Prefix##_I; \ -typedef FFX_16BIT_I2 Prefix##_I2; \ -typedef FFX_16BIT_I3 Prefix##_I3; \ -typedef FFX_16BIT_I4 Prefix##_I4; \ -typedef FFX_16BIT_U Prefix##_U; \ -typedef FFX_16BIT_U2 Prefix##_U2; \ -typedef FFX_16BIT_U3 Prefix##_U3; \ -typedef FFX_16BIT_U4 Prefix##_U4; - -#define TYPEDEF_FULL_PRECISION_TYPES(Prefix) \ -typedef FfxFloat32 Prefix##_F; \ -typedef FfxFloat32x2 Prefix##_F2; \ -typedef FfxFloat32x3 Prefix##_F3; \ -typedef FfxFloat32x4 Prefix##_F4; \ -typedef FfxInt32 Prefix##_I; \ -typedef FfxInt32x2 Prefix##_I2; \ -typedef FfxInt32x3 Prefix##_I3; \ -typedef FfxInt32x4 Prefix##_I4; \ -typedef FfxUInt32 Prefix##_U; \ -typedef FfxUInt32x2 Prefix##_U2; \ -typedef FfxUInt32x3 Prefix##_U3; \ -typedef FfxUInt32x4 Prefix##_U4; -#endif // #if defined(FFX_HLSL) - -#if defined(SHADER_API_PSSL) - -#define unorm -#define globallycoherent - -#if FFX_HALF - -#define FFX_MIN16_F half -#define FFX_MIN16_F2 half2 -#define FFX_MIN16_F3 half3 -#define FFX_MIN16_F4 half4 - -#define FFX_MIN16_I short -#define FFX_MIN16_I2 short2 -#define FFX_MIN16_I3 short3 -#define FFX_MIN16_I4 short4 - -#define FFX_MIN16_U ushort -#define FFX_MIN16_U2 ushort2 -#define FFX_MIN16_U3 ushort3 -#define FFX_MIN16_U4 ushort4 - -#define FFX_16BIT_F half -#define FFX_16BIT_F2 half2 -#define FFX_16BIT_F3 half3 -#define FFX_16BIT_F4 half4 - -#define FFX_16BIT_I short -#define FFX_16BIT_I2 short2 -#define FFX_16BIT_I3 short3 -#define FFX_16BIT_I4 short4 - -#define FFX_16BIT_U ushort -#define FFX_16BIT_U2 ushort2 -#define FFX_16BIT_U3 ushort3 -#define FFX_16BIT_U4 ushort4 - -#else // FFX_HALF - -#define FFX_MIN16_F float -#define FFX_MIN16_F2 float2 -#define FFX_MIN16_F3 float3 -#define FFX_MIN16_F4 float4 - -#define FFX_MIN16_I int -#define FFX_MIN16_I2 int2 -#define FFX_MIN16_I3 int3 -#define FFX_MIN16_I4 int4 - -#define FFX_MIN16_U uint -#define FFX_MIN16_U2 uint2 -#define FFX_MIN16_U3 uint3 -#define FFX_MIN16_U4 uint4 - -#define FFX_16BIT_F float -#define FFX_16BIT_F2 float2 -#define FFX_16BIT_F3 float3 -#define FFX_16BIT_F4 float4 - -#define FFX_16BIT_I int -#define FFX_16BIT_I2 int2 -#define FFX_16BIT_I3 int3 -#define FFX_16BIT_I4 int4 - -#define FFX_16BIT_U uint -#define FFX_16BIT_U2 uint2 -#define FFX_16BIT_U3 uint3 -#define FFX_16BIT_U4 uint4 - -#endif // FFX_HALF - -#endif // #if defined(SHADER_API_PSSL) - -#if defined(FFX_GLSL) - -#if FFX_HALF - -#define FFX_MIN16_F float16_t -#define FFX_MIN16_F2 f16vec2 -#define FFX_MIN16_F3 f16vec3 -#define FFX_MIN16_F4 f16vec4 - -#define FFX_MIN16_I int16_t -#define FFX_MIN16_I2 i16vec2 -#define FFX_MIN16_I3 i16vec3 -#define FFX_MIN16_I4 i16vec4 - -#define FFX_MIN16_U uint16_t -#define FFX_MIN16_U2 u16vec2 -#define FFX_MIN16_U3 u16vec3 -#define FFX_MIN16_U4 u16vec4 - -#define FFX_16BIT_F float16_t -#define FFX_16BIT_F2 f16vec2 -#define FFX_16BIT_F3 f16vec3 -#define FFX_16BIT_F4 f16vec4 - -#define FFX_16BIT_I int16_t -#define FFX_16BIT_I2 i16vec2 -#define FFX_16BIT_I3 i16vec3 -#define FFX_16BIT_I4 i16vec4 - -#define FFX_16BIT_U uint16_t -#define FFX_16BIT_U2 u16vec2 -#define FFX_16BIT_U3 u16vec3 -#define FFX_16BIT_U4 u16vec4 - -#else // FFX_HALF - -#define FFX_MIN16_F float -#define FFX_MIN16_F2 vec2 -#define FFX_MIN16_F3 vec3 -#define FFX_MIN16_F4 vec4 - -#define FFX_MIN16_I int -#define FFX_MIN16_I2 ivec2 -#define FFX_MIN16_I3 ivec3 -#define FFX_MIN16_I4 ivec4 - -#define FFX_MIN16_U uint -#define FFX_MIN16_U2 uvec2 -#define FFX_MIN16_U3 uvec3 -#define FFX_MIN16_U4 uvec4 - -#define FFX_16BIT_F float -#define FFX_16BIT_F2 vec2 -#define FFX_16BIT_F3 vec3 -#define FFX_16BIT_F4 vec4 - -#define FFX_16BIT_I int -#define FFX_16BIT_I2 ivec2 -#define FFX_16BIT_I3 ivec3 -#define FFX_16BIT_I4 ivec4 - -#define FFX_16BIT_U uint -#define FFX_16BIT_U2 uvec2 -#define FFX_16BIT_U3 uvec3 -#define FFX_16BIT_U4 uvec4 - -#endif // FFX_HALF - -#endif // #if defined(FFX_GLSL) - -#endif // #if defined(FFX_GPU) -#endif // #ifndef FFX_COMMON_TYPES_H diff --git a/Assets/Resources/FSR2/shaders/ffx_common_types.h.meta b/Assets/Resources/FSR2/shaders/ffx_common_types.h.meta deleted file mode 100644 index f2cc805..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_common_types.h.meta +++ /dev/null @@ -1,60 +0,0 @@ -fileFormatVersion: 2 -guid: c6da07d8aae05f04f87e4db20f84c73e -PluginImporter: - externalObjects: {} - serializedVersion: 2 - iconMap: {} - executionOrder: {} - defineConstraints: [] - isPreloaded: 0 - isOverridable: 0 - isExplicitlyReferenced: 0 - validateReferences: 1 - platformData: - - first: - : Any - second: - enabled: 0 - settings: - Exclude Editor: 1 - Exclude GameCoreScarlett: 1 - Exclude GameCoreXboxOne: 1 - Exclude Linux64: 1 - Exclude OSXUniversal: 1 - Exclude PS4: 1 - Exclude PS5: 1 - Exclude WebGL: 1 - Exclude Win: 1 - Exclude Win64: 1 - - first: - Any: - second: - enabled: 0 - settings: {} - - first: - Editor: Editor - second: - enabled: 0 - settings: - DefaultValueInitialized: true - - first: - Standalone: Linux64 - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win64 - second: - enabled: 0 - settings: - CPU: None - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_core.h b/Assets/Resources/FSR2/shaders/ffx_core.h deleted file mode 100644 index 4e687d6..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_core.h +++ /dev/null @@ -1,52 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -/// @defgroup Core -/// @defgroup HLSL -/// @defgroup GLSL -/// @defgroup GPU -/// @defgroup CPU -/// @defgroup CAS -/// @defgroup FSR1 - -#if !defined(FFX_CORE_H) -#define FFX_CORE_H - -#include "ffx_common_types.h" - -#if defined(FFX_CPU) - #include "ffx_core_cpu.h" -#endif // #if defined(FFX_CPU) - -#if defined(FFX_GLSL) && defined(FFX_GPU) - #include "ffx_core_glsl.h" -#endif // #if defined(FFX_GLSL) && defined(FFX_GPU) - -#if defined(FFX_HLSL) && defined(FFX_GPU) - #include "ffx_core_hlsl.h" -#endif // #if defined(FFX_HLSL) && defined(FFX_GPU) - -#if defined(FFX_GPU) - #include "ffx_core_gpu_common.h" - #include "ffx_core_gpu_common_half.h" - #include "ffx_core_portability.h" -#endif // #if defined(FFX_GPU) -#endif // #if !defined(FFX_CORE_H) \ No newline at end of file diff --git a/Assets/Resources/FSR2/shaders/ffx_core.h.meta b/Assets/Resources/FSR2/shaders/ffx_core.h.meta deleted file mode 100644 index a5f80a7..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_core.h.meta +++ /dev/null @@ -1,60 +0,0 @@ -fileFormatVersion: 2 -guid: b37eb663a0ae01b469b0b5a54365b301 -PluginImporter: - externalObjects: {} - serializedVersion: 2 - iconMap: {} - executionOrder: {} - defineConstraints: [] - isPreloaded: 0 - isOverridable: 0 - isExplicitlyReferenced: 0 - validateReferences: 1 - platformData: - - first: - : Any - second: - enabled: 0 - settings: - Exclude Editor: 1 - Exclude GameCoreScarlett: 1 - Exclude GameCoreXboxOne: 1 - Exclude Linux64: 1 - Exclude OSXUniversal: 1 - Exclude PS4: 1 - Exclude PS5: 1 - Exclude WebGL: 1 - Exclude Win: 1 - Exclude Win64: 1 - - first: - Any: - second: - enabled: 0 - settings: {} - - first: - Editor: Editor - second: - enabled: 0 - settings: - DefaultValueInitialized: true - - first: - Standalone: Linux64 - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win64 - second: - enabled: 0 - settings: - CPU: None - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_core_cpu.h b/Assets/Resources/FSR2/shaders/ffx_core_cpu.h deleted file mode 100644 index 3bf0295..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_core_cpu.h +++ /dev/null @@ -1,332 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -/// A define for a true value in a boolean expression. -/// -/// @ingroup CPU -#define FFX_TRUE (1) - -/// A define for a false value in a boolean expression. -/// -/// @ingroup CPU -#define FFX_FALSE (0) - -#if !defined(FFX_STATIC) -/// A define to abstract declaration of static variables and functions. -/// -/// @ingroup CPU -#define FFX_STATIC static -#endif // #if !defined(FFX_STATIC) - -#ifdef __clang__ -#pragma clang diagnostic ignored "-Wunused-variable" -#endif - -/// Interpret the bit layout of an IEEE-754 floating point value as an unsigned integer. -/// -/// @param [in] x A 32bit floating value. -/// -/// @returns -/// An unsigned 32bit integer value containing the bit pattern of x. -/// -/// @ingroup CPU -FFX_STATIC FfxUInt32 ffxAsUInt32(FfxFloat32 x) -{ - union - { - FfxFloat32 f; - FfxUInt32 u; - } bits; - - bits.f = x; - return bits.u; -} - -FFX_STATIC FfxFloat32 ffxDot2(FfxFloat32x2 a, FfxFloat32x2 b) -{ - return a[0] * b[0] + a[1] * b[1]; -} - -FFX_STATIC FfxFloat32 ffxDot3(FfxFloat32x3 a, FfxFloat32x3 b) -{ - return a[0] * b[0] + a[1] * b[1] + a[2] * b[2]; -} - -FFX_STATIC FfxFloat32 ffxDot4(FfxFloat32x4 a, FfxFloat32x4 b) -{ - return a[0] * b[0] + a[1] * b[1] + a[2] * b[2] + a[3] * b[3]; -} - -/// Compute the linear interopation between two values. -/// -/// Implemented by calling the GLSL mix instrinsic function. Implements the -/// following math: -/// -/// (1 - t) * x + t * y -/// -/// @param [in] x The first value to lerp between. -/// @param [in] y The second value to lerp between. -/// @param [in] t The value to determine how much of x and how much of y. -/// -/// @returns -/// A linearly interpolated value between x and y according to t. -/// -/// @ingroup CPU -FFX_STATIC FfxFloat32 ffxLerp(FfxFloat32 x, FfxFloat32 y, FfxFloat32 t) -{ - return y * t + (-x * t + x); -} - -/// Compute the reciprocal of a value. -/// -/// @param [in] x The value to compute the reciprocal for. -/// -/// @returns -/// The reciprocal value of x. -/// -/// @ingroup CPU -FFX_STATIC FfxFloat32 ffxReciprocal(FfxFloat32 a) -{ - return 1.0f / a; -} - -/// Compute the square root of a value. -/// -/// @param [in] x The first value to compute the min of. -/// -/// @returns -/// The the square root of x. -/// -/// @ingroup CPU -FFX_STATIC FfxFloat32 ffxSqrt(FfxFloat32 x) -{ - return sqrt(x); -} - -FFX_STATIC FfxUInt32 AShrSU1(FfxUInt32 a, FfxUInt32 b) -{ - return FfxUInt32(FfxInt32(a) >> FfxInt32(b)); -} - -/// Compute the factional part of a decimal value. -/// -/// This function calculates x - floor(x). -/// -/// @param [in] x The value to compute the fractional part from. -/// -/// @returns -/// The fractional part of x. -/// -/// @ingroup CPU -FFX_STATIC FfxFloat32 ffxFract(FfxFloat32 a) -{ - return a - floor(a); -} - -/// Compute the reciprocal square root of a value. -/// -/// @param [in] x The value to compute the reciprocal for. -/// -/// @returns -/// The reciprocal square root value of x. -/// -/// @ingroup CPU -FFX_STATIC FfxFloat32 rsqrt(FfxFloat32 a) -{ - return ffxReciprocal(ffxSqrt(a)); -} - -FFX_STATIC FfxFloat32 ffxMin(FfxFloat32 x, FfxFloat32 y) -{ - return x < y ? x : y; -} - -FFX_STATIC FfxUInt32 ffxMin(FfxUInt32 x, FfxUInt32 y) -{ - return x < y ? x : y; -} - -FFX_STATIC FfxFloat32 ffxMax(FfxFloat32 x, FfxFloat32 y) -{ - return x > y ? x : y; -} - -FFX_STATIC FfxUInt32 ffxMax(FfxUInt32 x, FfxUInt32 y) -{ - return x > y ? x : y; -} - -/// Clamp a value to a [0..1] range. -/// -/// @param [in] x The value to clamp to [0..1] range. -/// -/// @returns -/// The clamped version of x. -/// -/// @ingroup CPU -FFX_STATIC FfxFloat32 ffxSaturate(FfxFloat32 a) -{ - return ffxMin(1.0f, ffxMax(0.0f, a)); -} - -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// - -FFX_STATIC void opAAddOneF3(FfxFloat32x3 d, FfxFloat32x3 a, FfxFloat32 b) -{ - d[0] = a[0] + b; - d[1] = a[1] + b; - d[2] = a[2] + b; - return; -} - -FFX_STATIC void opACpyF3(FfxFloat32x3 d, FfxFloat32x3 a) -{ - d[0] = a[0]; - d[1] = a[1]; - d[2] = a[2]; - return; -} - -FFX_STATIC void opAMulF3(FfxFloat32x3 d, FfxFloat32x3 a, FfxFloat32x3 b) -{ - d[0] = a[0] * b[0]; - d[1] = a[1] * b[1]; - d[2] = a[2] * b[2]; - return; -} - -FFX_STATIC void opAMulOneF3(FfxFloat32x3 d, FfxFloat32x3 a, FfxFloat32 b) -{ - d[0] = a[0] * b; - d[1] = a[1] * b; - d[2] = a[2] * b; - return; -} - -FFX_STATIC void opARcpF3(FfxFloat32x3 d, FfxFloat32x3 a) -{ - d[0] = ffxReciprocal(a[0]); - d[1] = ffxReciprocal(a[1]); - d[2] = ffxReciprocal(a[2]); - return; -} - -/// Convert FfxFloat32 to half (in lower 16-bits of output). -/// -/// This function implements the same fast technique that is documented here: ftp://ftp.fox-toolkit.org/pub/fasthalffloatconversion.pdf -/// -/// The function supports denormals. -/// -/// Some conversion rules are to make computations possibly "safer" on the GPU, -/// -INF & -NaN -> -65504 -/// +INF & +NaN -> +65504 -/// -/// @param [in] f The 32bit floating point value to convert. -/// -/// @returns -/// The closest 16bit floating point value to f. -/// -/// @ingroup CPU -FFX_STATIC FfxUInt32 f32tof16(FfxFloat32 f) -{ - static FfxUInt16 base[512] = { - 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, - 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, - 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, - 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, - 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, - 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0001, 0x0002, 0x0004, 0x0008, 0x0010, 0x0020, 0x0040, 0x0080, 0x0100, 0x0200, 0x0400, - 0x0800, 0x0c00, 0x1000, 0x1400, 0x1800, 0x1c00, 0x2000, 0x2400, 0x2800, 0x2c00, 0x3000, 0x3400, 0x3800, 0x3c00, 0x4000, 0x4400, 0x4800, 0x4c00, 0x5000, - 0x5400, 0x5800, 0x5c00, 0x6000, 0x6400, 0x6800, 0x6c00, 0x7000, 0x7400, 0x7800, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, - 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, - 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, - 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, - 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, - 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, - 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, - 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, - 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, - 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, - 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, - 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8001, 0x8002, - 0x8004, 0x8008, 0x8010, 0x8020, 0x8040, 0x8080, 0x8100, 0x8200, 0x8400, 0x8800, 0x8c00, 0x9000, 0x9400, 0x9800, 0x9c00, 0xa000, 0xa400, 0xa800, 0xac00, - 0xb000, 0xb400, 0xb800, 0xbc00, 0xc000, 0xc400, 0xc800, 0xcc00, 0xd000, 0xd400, 0xd800, 0xdc00, 0xe000, 0xe400, 0xe800, 0xec00, 0xf000, 0xf400, 0xf800, - 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, - 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, - 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, - 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, - 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, - 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff - }; - - static FfxUInt8 shift[512] = { - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x17, 0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0f, 0x0e, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, - 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x17, 0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0f, 0x0e, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, - 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, - 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18 - }; - - union - { - FfxFloat32 f; - FfxUInt32 u; - } bits; - - bits.f = f; - FfxUInt32 u = bits.u; - FfxUInt32 i = u >> 23; - return (FfxUInt32)(base[i]) + ((u & 0x7fffff) >> shift[i]); -} - -/// Pack 2x32-bit floating point values in a single 32bit value. -/// -/// This function first converts each component of value into their nearest 16-bit floating -/// point representation, and then stores the X and Y components in the lower and upper 16 bits of the -/// 32bit unsigned integer respectively. -/// -/// @param [in] value A 2-dimensional floating point value to convert and pack. -/// -/// @returns -/// A packed 32bit value containing 2 16bit floating point values. -/// -/// @ingroup CPU -FFX_STATIC FfxUInt32 packHalf2x16(FfxFloat32x2 a) -{ - return f32tof16(a[0]) + (f32tof16(a[1]) << 16); -} diff --git a/Assets/Resources/FSR2/shaders/ffx_core_cpu.h.meta b/Assets/Resources/FSR2/shaders/ffx_core_cpu.h.meta deleted file mode 100644 index 7edf590..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_core_cpu.h.meta +++ /dev/null @@ -1,60 +0,0 @@ -fileFormatVersion: 2 -guid: 9d1d6ed5c9da0c64b882f3ebc2bac307 -PluginImporter: - externalObjects: {} - serializedVersion: 2 - iconMap: {} - executionOrder: {} - defineConstraints: [] - isPreloaded: 0 - isOverridable: 0 - isExplicitlyReferenced: 0 - validateReferences: 1 - platformData: - - first: - : Any - second: - enabled: 0 - settings: - Exclude Editor: 1 - Exclude GameCoreScarlett: 1 - Exclude GameCoreXboxOne: 1 - Exclude Linux64: 1 - Exclude OSXUniversal: 1 - Exclude PS4: 1 - Exclude PS5: 1 - Exclude WebGL: 1 - Exclude Win: 1 - Exclude Win64: 1 - - first: - Any: - second: - enabled: 0 - settings: {} - - first: - Editor: Editor - second: - enabled: 0 - settings: - DefaultValueInitialized: true - - first: - Standalone: Linux64 - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win64 - second: - enabled: 0 - settings: - CPU: None - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_core_gpu_common.h b/Assets/Resources/FSR2/shaders/ffx_core_gpu_common.h deleted file mode 100644 index ae07642..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_core_gpu_common.h +++ /dev/null @@ -1,2784 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -/// A define for a true value in a boolean expression. -/// -/// @ingroup GPU -#define FFX_TRUE (true) - -/// A define for a false value in a boolean expression. -/// -/// @ingroup GPU -#define FFX_FALSE (false) - -/// A define value for positive infinity. -/// -/// @ingroup GPU -#define FFX_POSITIVE_INFINITY_FLOAT ffxAsFloat(0x7f800000u) - -/// A define value for negative infinity. -/// -/// @ingroup GPU -#define FFX_NEGATIVE_INFINITY_FLOAT ffxAsFloat(0xff800000u) - -/// A define value for PI. -/// -/// @ingroup GPU -#define FFX_PI (3.14159) - - -/// Compute the reciprocal of value. -/// -/// @param [in] value The value to compute the reciprocal of. -/// -/// @returns -/// The 1 / value. -/// -/// @ingroup GPU -FfxFloat32 ffxReciprocal(FfxFloat32 value) -{ - return rcp(value); -} - -/// Compute the reciprocal of value. -/// -/// @param [in] value The value to compute the reciprocal of. -/// -/// @returns -/// The 1 / value. -/// -/// @ingroup GPU -FfxFloat32x2 ffxReciprocal(FfxFloat32x2 value) -{ - return rcp(value); -} - -/// Compute the reciprocal of value. -/// -/// @param [in] value The value to compute the reciprocal of. -/// -/// @returns -/// The 1 / value. -/// -/// @ingroup GPU -FfxFloat32x3 ffxReciprocal(FfxFloat32x3 value) -{ - return rcp(value); -} - -/// Compute the reciprocal of value. -/// -/// @param [in] value The value to compute the reciprocal of. -/// -/// @returns -/// The 1 / value. -/// -/// @ingroup GPU -FfxFloat32x4 ffxReciprocal(FfxFloat32x4 value) -{ - return rcp(value); -} - -/// Compute the min of two values. -/// -/// @param [in] x The first value to compute the min of. -/// @param [in] y The second value to compute the min of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxFloat32 ffxMin(FfxFloat32 x, FfxFloat32 y) -{ - return min(x, y); -} - -/// Compute the min of two values. -/// -/// @param [in] x The first value to compute the min of. -/// @param [in] y The second value to compute the min of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxFloat32x2 ffxMin(FfxFloat32x2 x, FfxFloat32x2 y) -{ - return min(x, y); -} - -/// Compute the min of two values. -/// -/// @param [in] x The first value to compute the min of. -/// @param [in] y The second value to compute the min of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxFloat32x3 ffxMin(FfxFloat32x3 x, FfxFloat32x3 y) -{ - return min(x, y); -} - -/// Compute the min of two values. -/// -/// @param [in] x The first value to compute the min of. -/// @param [in] y The second value to compute the min of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxFloat32x4 ffxMin(FfxFloat32x4 x, FfxFloat32x4 y) -{ - return min(x, y); -} - -/// Compute the min of two values. -/// -/// @param [in] x The first value to compute the min of. -/// @param [in] y The second value to compute the min of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxInt32 ffxMin(FfxInt32 x, FfxInt32 y) -{ - return min(x, y); -} - -/// Compute the min of two values. -/// -/// @param [in] x The first value to compute the min of. -/// @param [in] y The second value to compute the min of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxInt32x2 ffxMin(FfxInt32x2 x, FfxInt32x2 y) -{ - return min(x, y); -} - -/// Compute the min of two values. -/// -/// @param [in] x The first value to compute the min of. -/// @param [in] y The second value to compute the min of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxInt32x3 ffxMin(FfxInt32x3 x, FfxInt32x3 y) -{ - return min(x, y); -} - -/// Compute the min of two values. -/// -/// @param [in] x The first value to compute the min of. -/// @param [in] y The second value to compute the min of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxInt32x4 ffxMin(FfxInt32x4 x, FfxInt32x4 y) -{ - return min(x, y); -} - -/// Compute the min of two values. -/// -/// @param [in] x The first value to compute the min of. -/// @param [in] y The second value to compute the min of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxUInt32 ffxMin(FfxUInt32 x, FfxUInt32 y) -{ - return min(x, y); -} - -/// Compute the min of two values. -/// -/// @param [in] x The first value to compute the min of. -/// @param [in] y The second value to compute the min of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxUInt32x2 ffxMin(FfxUInt32x2 x, FfxUInt32x2 y) -{ - return min(x, y); -} - -/// Compute the min of two values. -/// -/// @param [in] x The first value to compute the min of. -/// @param [in] y The second value to compute the min of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxUInt32x3 ffxMin(FfxUInt32x3 x, FfxUInt32x3 y) -{ - return min(x, y); -} - -/// Compute the min of two values. -/// -/// @param [in] x The first value to compute the min of. -/// @param [in] y The second value to compute the min of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxUInt32x4 ffxMin(FfxUInt32x4 x, FfxUInt32x4 y) -{ - return min(x, y); -} - -/// Compute the max of two values. -/// -/// @param [in] x The first value to compute the max of. -/// @param [in] y The second value to compute the max of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxFloat32 ffxMax(FfxFloat32 x, FfxFloat32 y) -{ - return max(x, y); -} - -/// Compute the max of two values. -/// -/// @param [in] x The first value to compute the max of. -/// @param [in] y The second value to compute the max of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxFloat32x2 ffxMax(FfxFloat32x2 x, FfxFloat32x2 y) -{ - return max(x, y); -} - -/// Compute the max of two values. -/// -/// @param [in] x The first value to compute the max of. -/// @param [in] y The second value to compute the max of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxFloat32x3 ffxMax(FfxFloat32x3 x, FfxFloat32x3 y) -{ - return max(x, y); -} - -/// Compute the max of two values. -/// -/// @param [in] x The first value to compute the max of. -/// @param [in] y The second value to compute the max of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxFloat32x4 ffxMax(FfxFloat32x4 x, FfxFloat32x4 y) -{ - return max(x, y); -} - -/// Compute the max of two values. -/// -/// @param [in] x The first value to compute the max of. -/// @param [in] y The second value to compute the max of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxInt32 ffxMax(FfxInt32 x, FfxInt32 y) -{ - return max(x, y); -} - -/// Compute the max of two values. -/// -/// @param [in] x The first value to compute the max of. -/// @param [in] y The second value to compute the max of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxInt32x2 ffxMax(FfxInt32x2 x, FfxInt32x2 y) -{ - return max(x, y); -} - -/// Compute the max of two values. -/// -/// @param [in] x The first value to compute the max of. -/// @param [in] y The second value to compute the max of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxInt32x3 ffxMax(FfxInt32x3 x, FfxInt32x3 y) -{ - return max(x, y); -} - -/// Compute the max of two values. -/// -/// @param [in] x The first value to compute the max of. -/// @param [in] y The second value to compute the max of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxInt32x4 ffxMax(FfxInt32x4 x, FfxInt32x4 y) -{ - return max(x, y); -} - -/// Compute the max of two values. -/// -/// @param [in] x The first value to compute the max of. -/// @param [in] y The second value to compute the max of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxUInt32 ffxMax(FfxUInt32 x, FfxUInt32 y) -{ - return max(x, y); -} - -/// Compute the max of two values. -/// -/// @param [in] x The first value to compute the max of. -/// @param [in] y The second value to compute the max of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxUInt32x2 ffxMax(FfxUInt32x2 x, FfxUInt32x2 y) -{ - return max(x, y); -} - -/// Compute the max of two values. -/// -/// @param [in] x The first value to compute the max of. -/// @param [in] y The second value to compute the max of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxUInt32x3 ffxMax(FfxUInt32x3 x, FfxUInt32x3 y) -{ - return max(x, y); -} - -/// Compute the max of two values. -/// -/// @param [in] x The first value to compute the max of. -/// @param [in] y The second value to compute the max of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxUInt32x4 ffxMax(FfxUInt32x4 x, FfxUInt32x4 y) -{ - return max(x, y); -} - -/// Compute the value of the first parameter raised to the power of the second. -/// -/// @param [in] x The value to raise to the power y. -/// @param [in] y The power to which to raise x. -/// -/// @returns -/// The value of the first parameter raised to the power of the second. -/// -/// @ingroup GPU -FfxFloat32 ffxPow(FfxFloat32 x, FfxFloat32 y) -{ - return pow(x, y); -} - -/// Compute the value of the first parameter raised to the power of the second. -/// -/// @param [in] x The value to raise to the power y. -/// @param [in] y The power to which to raise x. -/// -/// @returns -/// The value of the first parameter raised to the power of the second. -/// -/// @ingroup GPU -FfxFloat32x2 ffxPow(FfxFloat32x2 x, FfxFloat32x2 y) -{ - return pow(x, y); -} - -/// Compute the value of the first parameter raised to the power of the second. -/// -/// @param [in] x The value to raise to the power y. -/// @param [in] y The power to which to raise x. -/// -/// @returns -/// The value of the first parameter raised to the power of the second. -/// -/// @ingroup GPU -FfxFloat32x3 ffxPow(FfxFloat32x3 x, FfxFloat32x3 y) -{ - return pow(x, y); -} - -/// Compute the value of the first parameter raised to the power of the second. -/// -/// @param [in] x The value to raise to the power y. -/// @param [in] y The power to which to raise x. -/// -/// @returns -/// The value of the first parameter raised to the power of the second. -/// -/// @ingroup GPU -FfxFloat32x4 ffxPow(FfxFloat32x4 x, FfxFloat32x4 y) -{ - return pow(x, y); -} - -/// Compute the square root of a value. -/// -/// @param [in] x The first value to compute the min of. -/// -/// @returns -/// The the square root of x. -/// -/// @ingroup GPU -FfxFloat32 ffxSqrt(FfxFloat32 x) -{ - return sqrt(x); -} - -/// Compute the square root of a value. -/// -/// @param [in] x The first value to compute the min of. -/// -/// @returns -/// The the square root of x. -/// -/// @ingroup GPU -FfxFloat32x2 ffxSqrt(FfxFloat32x2 x) -{ - return sqrt(x); -} - -/// Compute the square root of a value. -/// -/// @param [in] x The first value to compute the min of. -/// -/// @returns -/// The the square root of x. -/// -/// @ingroup GPU -FfxFloat32x3 ffxSqrt(FfxFloat32x3 x) -{ - return sqrt(x); -} - -/// Compute the square root of a value. -/// -/// @param [in] x The first value to compute the min of. -/// -/// @returns -/// The the square root of x. -/// -/// @ingroup GPU -FfxFloat32x4 ffxSqrt(FfxFloat32x4 x) -{ - return sqrt(x); -} - -/// Copy the sign bit from 's' to positive 'd'. -/// -/// @param [in] d The value to copy the sign bit into. -/// @param [in] s The value to copy the sign bit from. -/// -/// @returns -/// The value of d with the sign bit from s. -/// -/// @ingroup GPU -FfxFloat32 ffxCopySignBit(FfxFloat32 d, FfxFloat32 s) -{ - return ffxAsFloat(ffxAsUInt32(d) | (ffxAsUInt32(s) & FfxUInt32(0x80000000u))); -} - -/// Copy the sign bit from 's' to positive 'd'. -/// -/// @param [in] d The value to copy the sign bit into. -/// @param [in] s The value to copy the sign bit from. -/// -/// @returns -/// The value of d with the sign bit from s. -/// -/// @ingroup GPU -FfxFloat32x2 ffxCopySignBit(FfxFloat32x2 d, FfxFloat32x2 s) -{ - return ffxAsFloat(ffxAsUInt32(d) | (ffxAsUInt32(s) & ffxBroadcast2(0x80000000u))); -} - -/// Copy the sign bit from 's' to positive 'd'. -/// -/// @param [in] d The value to copy the sign bit into. -/// @param [in] s The value to copy the sign bit from. -/// -/// @returns -/// The value of d with the sign bit from s. -/// -/// @ingroup GPU -FfxFloat32x3 ffxCopySignBit(FfxFloat32x3 d, FfxFloat32x3 s) -{ - return ffxAsFloat(ffxAsUInt32(d) | (ffxAsUInt32(s) & ffxBroadcast3(0x80000000u))); -} - -/// Copy the sign bit from 's' to positive 'd'. -/// -/// @param [in] d The value to copy the sign bit into. -/// @param [in] s The value to copy the sign bit from. -/// -/// @returns -/// The value of d with the sign bit from s. -/// -/// @ingroup GPU -FfxFloat32x4 ffxCopySignBit(FfxFloat32x4 d, FfxFloat32x4 s) -{ - return ffxAsFloat(ffxAsUInt32(d) | (ffxAsUInt32(s) & ffxBroadcast4(0x80000000u))); -} - -/// A single operation to return the following: -/// m = NaN := 0 -/// m >= 0 := 0 -/// m < 0 := 1 -/// -/// Uses the following useful floating point logic, -/// saturate(+a*(-INF)==-INF) := 0 -/// saturate( 0*(-INF)== NaN) := 0 -/// saturate(-a*(-INF)==+INF) := 1 -/// -/// This function is useful when creating masks for branch-free logic. -/// -/// @param [in] m The value to test against 0. -/// -/// @returns -/// 1.0 when the value is negative, or 0.0 when the value is 0 or position. -/// -/// @ingroup GPU -FfxFloat32 ffxIsSigned(FfxFloat32 m) -{ - return ffxSaturate(m * FfxFloat32(FFX_NEGATIVE_INFINITY_FLOAT)); -} - -/// A single operation to return the following: -/// m = NaN := 0 -/// m >= 0 := 0 -/// m < 0 := 1 -/// -/// Uses the following useful floating point logic, -/// saturate(+a*(-INF)==-INF) := 0 -/// saturate( 0*(-INF)== NaN) := 0 -/// saturate(-a*(-INF)==+INF) := 1 -/// -/// This function is useful when creating masks for branch-free logic. -/// -/// @param [in] m The value to test against 0. -/// -/// @returns -/// 1.0 when the value is negative, or 0.0 when the value is 0 or position. -/// -/// @ingroup GPU -FfxFloat32x2 ffxIsSigned(FfxFloat32x2 m) -{ - return ffxSaturate(m * ffxBroadcast2(FFX_NEGATIVE_INFINITY_FLOAT)); -} - -/// A single operation to return the following: -/// m = NaN := 0 -/// m >= 0 := 0 -/// m < 0 := 1 -/// -/// Uses the following useful floating point logic, -/// saturate(+a*(-INF)==-INF) := 0 -/// saturate( 0*(-INF)== NaN) := 0 -/// saturate(-a*(-INF)==+INF) := 1 -/// -/// This function is useful when creating masks for branch-free logic. -/// -/// @param [in] m The value to test against 0. -/// -/// @returns -/// 1.0 when the value is negative, or 0.0 when the value is 0 or position. -/// -/// @ingroup GPU -FfxFloat32x3 ffxIsSigned(FfxFloat32x3 m) -{ - return ffxSaturate(m * ffxBroadcast3(FFX_NEGATIVE_INFINITY_FLOAT)); -} - -/// A single operation to return the following: -/// m = NaN := 0 -/// m >= 0 := 0 -/// m < 0 := 1 -/// -/// Uses the following useful floating point logic, -/// saturate(+a*(-INF)==-INF) := 0 -/// saturate( 0*(-INF)== NaN) := 0 -/// saturate(-a*(-INF)==+INF) := 1 -/// -/// This function is useful when creating masks for branch-free logic. -/// -/// @param [in] m The value to test against for have the sign set. -/// -/// @returns -/// 1.0 when the value is negative, or 0.0 when the value is 0 or positive. -/// -/// @ingroup GPU -FfxFloat32x4 ffxIsSigned(FfxFloat32x4 m) -{ - return ffxSaturate(m * ffxBroadcast4(FFX_NEGATIVE_INFINITY_FLOAT)); -} - -/// A single operation to return the following: -/// m = NaN := 1 -/// m > 0 := 0 -/// m <= 0 := 1 -/// -/// This function is useful when creating masks for branch-free logic. -/// -/// @param [in] m The value to test against zero. -/// -/// @returns -/// 1.0 when the value is position, or 0.0 when the value is 0 or negative. -/// -/// @ingroup GPU -FfxFloat32 ffxIsGreaterThanZero(FfxFloat32 m) -{ - return ffxSaturate(m * FfxFloat32(FFX_POSITIVE_INFINITY_FLOAT)); -} - -/// A single operation to return the following: -/// m = NaN := 1 -/// m > 0 := 0 -/// m <= 0 := 1 -/// -/// This function is useful when creating masks for branch-free logic. -/// -/// @param [in] m The value to test against zero. -/// -/// @returns -/// 1.0 when the value is position, or 0.0 when the value is 0 or negative. -/// -/// @ingroup GPU -FfxFloat32x2 ffxIsGreaterThanZero(FfxFloat32x2 m) -{ - return ffxSaturate(m * ffxBroadcast2(FFX_POSITIVE_INFINITY_FLOAT)); -} - -/// A single operation to return the following: -/// m = NaN := 1 -/// m > 0 := 0 -/// m <= 0 := 1 -/// -/// This function is useful when creating masks for branch-free logic. -/// -/// @param [in] m The value to test against zero. -/// -/// @returns -/// 1.0 when the value is position, or 0.0 when the value is 0 or negative. -/// -/// @ingroup GPU -FfxFloat32x3 ffxIsGreaterThanZero(FfxFloat32x3 m) -{ - return ffxSaturate(m * ffxBroadcast3(FFX_POSITIVE_INFINITY_FLOAT)); -} - -/// A single operation to return the following: -/// m = NaN := 1 -/// m > 0 := 0 -/// m <= 0 := 1 -/// -/// This function is useful when creating masks for branch-free logic. -/// -/// @param [in] m The value to test against zero. -/// -/// @returns -/// 1.0 when the value is position, or 0.0 when the value is 0 or negative. -/// -/// @ingroup GPU -FfxFloat32x4 ffxIsGreaterThanZero(FfxFloat32x4 m) -{ - return ffxSaturate(m * ffxBroadcast4(FFX_POSITIVE_INFINITY_FLOAT)); -} - -/// Convert a 32bit floating point value to sortable integer. -/// -/// - If sign bit=0, flip the sign bit (positives). -/// - If sign bit=1, flip all bits (negatives). -/// -/// The function has the side effects that: -/// - Larger integers are more positive values. -/// - Float zero is mapped to center of integers (so clear to integer zero is a nice default for atomic max usage). -/// -/// @param [in] value The floating point value to make sortable. -/// -/// @returns -/// The sortable integer value. -/// -/// @ingroup GPU -FfxUInt32 ffxFloatToSortableInteger(FfxUInt32 value) -{ - return value ^ ((AShrSU1(value, FfxUInt32(31))) | FfxUInt32(0x80000000)); -} - -/// Convert a sortable integer to a 32bit floating point value. -/// -/// The function has the side effects that: -/// - If sign bit=1, flip the sign bit (positives). -/// - If sign bit=0, flip all bits (negatives). -/// -/// @param [in] value The floating point value to make sortable. -/// -/// @returns -/// The sortable integer value. -/// -/// @ingroup GPU -FfxUInt32 ffxSortableIntegerToFloat(FfxUInt32 value) -{ - return value ^ ((~AShrSU1(value, FfxUInt32(31))) | FfxUInt32(0x80000000)); -} - -/// Calculate a low-quality approximation for the square root of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] value The value to calculate an approximate to the square root for. -/// -/// @returns -/// An approximation of the square root, estimated to low quality. -/// -/// @ingroup GPU -FfxFloat32 ffxApproximateSqrt(FfxFloat32 a) -{ - return ffxAsFloat((ffxAsUInt32(a) >> FfxUInt32(1)) + FfxUInt32(0x1fbc4639)); -} - -/// Calculate a low-quality approximation for the reciprocal of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] value The value to calculate an approximate to the reciprocal for. -/// -/// @returns -/// An approximation of the reciprocal, estimated to low quality. -/// -/// @ingroup GPU -FfxFloat32 ffxApproximateReciprocal(FfxFloat32 a) -{ - return ffxAsFloat(FfxUInt32(0x7ef07ebb) - ffxAsUInt32(a)); -} - -/// Calculate a medium-quality approximation for the reciprocal of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] value The value to calculate an approximate to the reciprocal for. -/// -/// @returns -/// An approximation of the reciprocal, estimated to medium quality. -/// -/// @ingroup GPU -FfxFloat32 ffxApproximateReciprocalMedium(FfxFloat32 value) -{ - FfxFloat32 b = ffxAsFloat(FfxUInt32(0x7ef19fff) - ffxAsUInt32(value)); - return b * (-b * value + FfxFloat32(2.0)); -} - -/// Calculate a low-quality approximation for the reciprocal of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] value The value to calculate an approximate to the reciprocal square root for. -/// -/// @returns -/// An approximation of the reciprocal square root, estimated to low quality. -/// -/// @ingroup GPU -FfxFloat32 ffxApproximateReciprocalSquareRoot(FfxFloat32 a) -{ - return ffxAsFloat(FfxUInt32(0x5f347d74) - (ffxAsUInt32(a) >> FfxUInt32(1))); -} - -/// Calculate a low-quality approximation for the square root of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] value The value to calculate an approximate to the square root for. -/// -/// @returns -/// An approximation of the square root, estimated to low quality. -/// -/// @ingroup GPU -FfxFloat32x2 ffxApproximateSqrt(FfxFloat32x2 a) -{ - return ffxAsFloat((ffxAsUInt32(a) >> ffxBroadcast2(1u)) + ffxBroadcast2(0x1fbc4639u)); -} - -/// Calculate a low-quality approximation for the reciprocal of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] value The value to calculate an approximate to the reciprocal for. -/// -/// @returns -/// An approximation of the reciprocal, estimated to low quality. -/// -/// @ingroup GPU -FfxFloat32x2 ffxApproximateReciprocal(FfxFloat32x2 a) -{ - return ffxAsFloat(ffxBroadcast2(0x7ef07ebbu) - ffxAsUInt32(a)); -} - -/// Calculate a medium-quality approximation for the reciprocal of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] value The value to calculate an approximate to the reciprocal for. -/// -/// @returns -/// An approximation of the reciprocal, estimated to medium quality. -/// -/// @ingroup GPU -FfxFloat32x2 ffxApproximateReciprocalMedium(FfxFloat32x2 a) -{ - FfxFloat32x2 b = ffxAsFloat(ffxBroadcast2(0x7ef19fffu) - ffxAsUInt32(a)); - return b * (-b * a + ffxBroadcast2(2.0f)); -} - -/// Calculate a low-quality approximation for the square root of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] value The value to calculate an approximate to the square root for. -/// -/// @returns -/// An approximation of the square root, estimated to low quality. -/// -/// @ingroup GPU -FfxFloat32x2 ffxApproximateReciprocalSquareRoot(FfxFloat32x2 a) -{ - return ffxAsFloat(ffxBroadcast2(0x5f347d74u) - (ffxAsUInt32(a) >> ffxBroadcast2(1u))); -} - -/// Calculate a low-quality approximation for the square root of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] value The value to calculate an approximate to the square root for. -/// -/// @returns -/// An approximation of the square root, estimated to low quality. -/// -/// @ingroup GPU -FfxFloat32x3 ffxApproximateSqrt(FfxFloat32x3 a) -{ - return ffxAsFloat((ffxAsUInt32(a) >> ffxBroadcast3(1u)) + ffxBroadcast3(0x1fbc4639u)); -} - -/// Calculate a low-quality approximation for the reciprocal of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] value The value to calculate an approximate to the reciprocal for. -/// -/// @returns -/// An approximation of the reciprocal, estimated to low quality. -/// -/// @ingroup GPU -FfxFloat32x3 ffxApproximateReciprocal(FfxFloat32x3 a) -{ - return ffxAsFloat(ffxBroadcast3(0x7ef07ebbu) - ffxAsUInt32(a)); -} - -/// Calculate a medium-quality approximation for the reciprocal of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] value The value to calculate an approximate to the reciprocal for. -/// -/// @returns -/// An approximation of the reciprocal, estimated to medium quality. -/// -/// @ingroup GPU -FfxFloat32x3 ffxApproximateReciprocalMedium(FfxFloat32x3 a) -{ - FfxFloat32x3 b = ffxAsFloat(ffxBroadcast3(0x7ef19fffu) - ffxAsUInt32(a)); - return b * (-b * a + ffxBroadcast3(2.0f)); -} - -/// Calculate a low-quality approximation for the square root of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] value The value to calculate an approximate to the square root for. -/// -/// @returns -/// An approximation of the square root, estimated to low quality. -/// -/// @ingroup GPU -FfxFloat32x3 ffxApproximateReciprocalSquareRoot(FfxFloat32x3 a) -{ - return ffxAsFloat(ffxBroadcast3(0x5f347d74u) - (ffxAsUInt32(a) >> ffxBroadcast3(1u))); -} - -/// Calculate a low-quality approximation for the square root of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] value The value to calculate an approximate to the square root for. -/// -/// @returns -/// An approximation of the square root, estimated to low quality. -/// -/// @ingroup GPU -FfxFloat32x4 ffxApproximateSqrt(FfxFloat32x4 a) -{ - return ffxAsFloat((ffxAsUInt32(a) >> ffxBroadcast4(1u)) + ffxBroadcast4(0x1fbc4639u)); -} - -/// Calculate a low-quality approximation for the reciprocal of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] value The value to calculate an approximate to the reciprocal for. -/// -/// @returns -/// An approximation of the reciprocal, estimated to low quality. -/// -/// @ingroup GPU -FfxFloat32x4 ffxApproximateReciprocal(FfxFloat32x4 a) -{ - return ffxAsFloat(ffxBroadcast4(0x7ef07ebbu) - ffxAsUInt32(a)); -} - -/// Calculate a medium-quality approximation for the reciprocal of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] value The value to calculate an approximate to the reciprocal for. -/// -/// @returns -/// An approximation of the reciprocal, estimated to medium quality. -/// -/// @ingroup GPU -FfxFloat32x4 ffxApproximateReciprocalMedium(FfxFloat32x4 a) -{ - FfxFloat32x4 b = ffxAsFloat(ffxBroadcast4(0x7ef19fffu) - ffxAsUInt32(a)); - return b * (-b * a + ffxBroadcast4(2.0f)); -} - -/// Calculate a low-quality approximation for the square root of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] value The value to calculate an approximate to the square root for. -/// -/// @returns -/// An approximation of the square root, estimated to low quality. -/// -/// @ingroup GPU -FfxFloat32x4 ffxApproximateReciprocalSquareRoot(FfxFloat32x4 a) -{ - return ffxAsFloat(ffxBroadcast4(0x5f347d74u) - (ffxAsUInt32(a) >> ffxBroadcast4(1u))); -} - -/// Calculate dot product of 'a' and 'b'. -/// -/// @param [in] a First vector input. -/// @param [in] b Second vector input. -/// -/// @returns -/// The value of a dot b. -/// -/// @ingroup GPU -FfxFloat32 ffxDot2(FfxFloat32x2 a, FfxFloat32x2 b) -{ - return dot(a, b); -} - -/// Calculate dot product of 'a' and 'b'. -/// -/// @param [in] a First vector input. -/// @param [in] b Second vector input. -/// -/// @returns -/// The value of a dot b. -/// -/// @ingroup GPU -FfxFloat32 ffxDot3(FfxFloat32x3 a, FfxFloat32x3 b) -{ - return dot(a, b); -} - -/// Calculate dot product of 'a' and 'b'. -/// -/// @param [in] a First vector input. -/// @param [in] b Second vector input. -/// -/// @returns -/// The value of a dot b. -/// -/// @ingroup GPU -FfxFloat32 ffxDot4(FfxFloat32x4 a, FfxFloat32x4 b) -{ - return dot(a, b); -} - - -/// Compute an approximate conversion from PQ to Gamma2 space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between PQ and Gamma2. -/// -/// @returns -/// The value a converted into Gamma2. -/// -/// @ingroup GPU -FfxFloat32 ffxApproximatePQToGamma2Medium(FfxFloat32 a) -{ - return a * a * a * a; -} - -/// Compute an approximate conversion from PQ to linear space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between PQ and linear. -/// -/// @returns -/// The value a converted into linear. -/// -/// @ingroup GPU -FfxFloat32 ffxApproximatePQToLinear(FfxFloat32 a) -{ - return a * a * a * a * a * a * a * a; -} - -/// Compute an approximate conversion from gamma2 to PQ space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between gamma2 and PQ. -/// -/// @returns -/// The value a converted into PQ. -/// -/// @ingroup GPU -FfxFloat32 ffxApproximateGamma2ToPQ(FfxFloat32 a) -{ - return ffxAsFloat((ffxAsUInt32(a) >> FfxUInt32(2)) + FfxUInt32(0x2F9A4E46)); -} - -/// Compute a more accurate approximate conversion from gamma2 to PQ space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between gamma2 and PQ. -/// -/// @returns -/// The value a converted into PQ. -/// -/// @ingroup GPU -FfxFloat32 ffxApproximateGamma2ToPQMedium(FfxFloat32 a) -{ - FfxFloat32 b = ffxAsFloat((ffxAsUInt32(a) >> FfxUInt32(2)) + FfxUInt32(0x2F9A4E46)); - FfxFloat32 b4 = b * b * b * b; - return b - b * (b4 - a) / (FfxFloat32(4.0) * b4); -} - -/// Compute a high accuracy approximate conversion from gamma2 to PQ space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between gamma2 and PQ. -/// -/// @returns -/// The value a converted into PQ. -/// -/// @ingroup GPU -FfxFloat32 ffxApproximateGamma2ToPQHigh(FfxFloat32 a) -{ - return ffxSqrt(ffxSqrt(a)); -} - -/// Compute an approximate conversion from linear to PQ space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between linear and PQ. -/// -/// @returns -/// The value a converted into PQ. -/// -/// @ingroup GPU -FfxFloat32 ffxApproximateLinearToPQ(FfxFloat32 a) -{ - return ffxAsFloat((ffxAsUInt32(a) >> FfxUInt32(3)) + FfxUInt32(0x378D8723)); -} - -/// Compute a more accurate approximate conversion from linear to PQ space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between linear and PQ. -/// -/// @returns -/// The value a converted into PQ. -/// -/// @ingroup GPU -FfxFloat32 ffxApproximateLinearToPQMedium(FfxFloat32 a) -{ - FfxFloat32 b = ffxAsFloat((ffxAsUInt32(a) >> FfxUInt32(3)) + FfxUInt32(0x378D8723)); - FfxFloat32 b8 = b * b * b * b * b * b * b * b; - return b - b * (b8 - a) / (FfxFloat32(8.0) * b8); -} - -/// Compute a very accurate approximate conversion from linear to PQ space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between linear and PQ. -/// -/// @returns -/// The value a converted into PQ. -/// -/// @ingroup GPU -FfxFloat32 ffxApproximateLinearToPQHigh(FfxFloat32 a) -{ - return ffxSqrt(ffxSqrt(ffxSqrt(a))); -} - -/// Compute an approximate conversion from PQ to Gamma2 space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between PQ and Gamma2. -/// -/// @returns -/// The value a converted into Gamma2. -/// -/// @ingroup GPU -FfxFloat32x2 ffxApproximatePQToGamma2Medium(FfxFloat32x2 a) -{ - return a * a * a * a; -} - -/// Compute an approximate conversion from PQ to linear space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between PQ and linear. -/// -/// @returns -/// The value a converted into linear. -/// -/// @ingroup GPU -FfxFloat32x2 ffxApproximatePQToLinear(FfxFloat32x2 a) -{ - return a * a * a * a * a * a * a * a; -} - -/// Compute an approximate conversion from gamma2 to PQ space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between gamma2 and PQ. -/// -/// @returns -/// The value a converted into PQ. -/// -/// @ingroup GPU -FfxFloat32x2 ffxApproximateGamma2ToPQ(FfxFloat32x2 a) -{ - return ffxAsFloat((ffxAsUInt32(a) >> ffxBroadcast2(2u)) + ffxBroadcast2(0x2F9A4E46u)); -} - -/// Compute a more accurate approximate conversion from gamma2 to PQ space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between gamma2 and PQ. -/// -/// @returns -/// The value a converted into PQ. -/// -/// @ingroup GPU -FfxFloat32x2 ffxApproximateGamma2ToPQMedium(FfxFloat32x2 a) -{ - FfxFloat32x2 b = ffxAsFloat((ffxAsUInt32(a) >> ffxBroadcast2(2u)) + ffxBroadcast2(0x2F9A4E46u)); - FfxFloat32x2 b4 = b * b * b * b; - return b - b * (b4 - a) / (FfxFloat32(4.0) * b4); -} - -/// Compute a high accuracy approximate conversion from gamma2 to PQ space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between gamma2 and PQ. -/// -/// @returns -/// The value a converted into PQ. -/// -/// @ingroup GPU -FfxFloat32x2 ffxApproximateGamma2ToPQHigh(FfxFloat32x2 a) -{ - return ffxSqrt(ffxSqrt(a)); -} - -/// Compute an approximate conversion from linear to PQ space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between linear and PQ. -/// -/// @returns -/// The value a converted into PQ. -/// -/// @ingroup GPU -FfxFloat32x2 ffxApproximateLinearToPQ(FfxFloat32x2 a) -{ - return ffxAsFloat((ffxAsUInt32(a) >> ffxBroadcast2(3u)) + ffxBroadcast2(0x378D8723u)); -} - -/// Compute a more accurate approximate conversion from linear to PQ space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between linear and PQ. -/// -/// @returns -/// The value a converted into PQ. -/// -/// @ingroup GPU -FfxFloat32x2 ffxApproximateLinearToPQMedium(FfxFloat32x2 a) -{ - FfxFloat32x2 b = ffxAsFloat((ffxAsUInt32(a) >> ffxBroadcast2(3u)) + ffxBroadcast2(0x378D8723u)); - FfxFloat32x2 b8 = b * b * b * b * b * b * b * b; - return b - b * (b8 - a) / (FfxFloat32(8.0) * b8); -} - -/// Compute a very accurate approximate conversion from linear to PQ space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between linear and PQ. -/// -/// @returns -/// The value a converted into PQ. -/// -/// @ingroup GPU -FfxFloat32x2 ffxApproximateLinearToPQHigh(FfxFloat32x2 a) -{ - return ffxSqrt(ffxSqrt(ffxSqrt(a))); -} - -/// Compute an approximate conversion from PQ to Gamma2 space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between PQ and Gamma2. -/// -/// @returns -/// The value a converted into Gamma2. -/// -/// @ingroup GPU -FfxFloat32x3 ffxApproximatePQToGamma2Medium(FfxFloat32x3 a) -{ - return a * a * a * a; -} - -/// Compute an approximate conversion from PQ to linear space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between PQ and linear. -/// -/// @returns -/// The value a converted into linear. -/// -/// @ingroup GPU -FfxFloat32x3 ffxApproximatePQToLinear(FfxFloat32x3 a) -{ - return a * a * a * a * a * a * a * a; -} - -/// Compute an approximate conversion from gamma2 to PQ space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between gamma2 and PQ. -/// -/// @returns -/// The value a converted into PQ. -/// -/// @ingroup GPU -FfxFloat32x3 ffxApproximateGamma2ToPQ(FfxFloat32x3 a) -{ - return ffxAsFloat((ffxAsUInt32(a) >> ffxBroadcast3(2u)) + ffxBroadcast3(0x2F9A4E46u)); -} - -/// Compute a more accurate approximate conversion from gamma2 to PQ space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between gamma2 and PQ. -/// -/// @returns -/// The value a converted into PQ. -/// -/// @ingroup GPU -FfxFloat32x3 ffxApproximateGamma2ToPQMedium(FfxFloat32x3 a) -{ - FfxFloat32x3 b = ffxAsFloat((ffxAsUInt32(a) >> ffxBroadcast3(2u)) + ffxBroadcast3(0x2F9A4E46u)); - FfxFloat32x3 b4 = b * b * b * b; - return b - b * (b4 - a) / (FfxFloat32(4.0) * b4); -} - -/// Compute a high accuracy approximate conversion from gamma2 to PQ space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between gamma2 and PQ. -/// -/// @returns -/// The value a converted into PQ. -/// -/// @ingroup GPU -FfxFloat32x3 ffxApproximateGamma2ToPQHigh(FfxFloat32x3 a) -{ - return ffxSqrt(ffxSqrt(a)); -} - -/// Compute an approximate conversion from linear to PQ space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between linear and PQ. -/// -/// @returns -/// The value a converted into PQ. -/// -/// @ingroup GPU -FfxFloat32x3 ffxApproximateLinearToPQ(FfxFloat32x3 a) -{ - return ffxAsFloat((ffxAsUInt32(a) >> ffxBroadcast3(3u)) + ffxBroadcast3(0x378D8723u)); -} - -/// Compute a more accurate approximate conversion from linear to PQ space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between linear and PQ. -/// -/// @returns -/// The value a converted into PQ. -/// -/// @ingroup GPU -FfxFloat32x3 ffxApproximateLinearToPQMedium(FfxFloat32x3 a) -{ - FfxFloat32x3 b = ffxAsFloat((ffxAsUInt32(a) >> ffxBroadcast3(3u)) + ffxBroadcast3(0x378D8723u)); - FfxFloat32x3 b8 = b * b * b * b * b * b * b * b; - return b - b * (b8 - a) / (FfxFloat32(8.0) * b8); -} - -/// Compute a very accurate approximate conversion from linear to PQ space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between linear and PQ. -/// -/// @returns -/// The value a converted into PQ. -/// -/// @ingroup GPU -FfxFloat32x3 ffxApproximateLinearToPQHigh(FfxFloat32x3 a) -{ - return ffxSqrt(ffxSqrt(ffxSqrt(a))); -} - -/// Compute an approximate conversion from PQ to Gamma2 space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between PQ and Gamma2. -/// -/// @returns -/// The value a converted into Gamma2. -/// -/// @ingroup GPU -FfxFloat32x4 ffxApproximatePQToGamma2Medium(FfxFloat32x4 a) -{ - return a * a * a * a; -} - -/// Compute an approximate conversion from PQ to linear space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between PQ and linear. -/// -/// @returns -/// The value a converted into linear. -/// -/// @ingroup GPU -FfxFloat32x4 ffxApproximatePQToLinear(FfxFloat32x4 a) -{ - return a * a * a * a * a * a * a * a; -} - -/// Compute an approximate conversion from gamma2 to PQ space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between gamma2 and PQ. -/// -/// @returns -/// The value a converted into PQ. -/// -/// @ingroup GPU -FfxFloat32x4 ffxApproximateGamma2ToPQ(FfxFloat32x4 a) -{ - return ffxAsFloat((ffxAsUInt32(a) >> ffxBroadcast4(2u)) + ffxBroadcast4(0x2F9A4E46u)); -} - -/// Compute a more accurate approximate conversion from gamma2 to PQ space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between gamma2 and PQ. -/// -/// @returns -/// The value a converted into PQ. -/// -/// @ingroup GPU -FfxFloat32x4 ffxApproximateGamma2ToPQMedium(FfxFloat32x4 a) -{ - FfxFloat32x4 b = ffxAsFloat((ffxAsUInt32(a) >> ffxBroadcast4(2u)) + ffxBroadcast4(0x2F9A4E46u)); - FfxFloat32x4 b4 = b * b * b * b * b * b * b * b; - return b - b * (b4 - a) / (FfxFloat32(4.0) * b4); -} - -/// Compute a high accuracy approximate conversion from gamma2 to PQ space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between gamma2 and PQ. -/// -/// @returns -/// The value a converted into PQ. -/// -/// @ingroup GPU -FfxFloat32x4 ffxApproximateGamma2ToPQHigh(FfxFloat32x4 a) -{ - return ffxSqrt(ffxSqrt(a)); -} - -/// Compute an approximate conversion from linear to PQ space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between linear and PQ. -/// -/// @returns -/// The value a converted into PQ. -/// -/// @ingroup GPU -FfxFloat32x4 ffxApproximateLinearToPQ(FfxFloat32x4 a) -{ - return ffxAsFloat((ffxAsUInt32(a) >> ffxBroadcast4(3u)) + ffxBroadcast4(0x378D8723u)); -} - -/// Compute a more accurate approximate conversion from linear to PQ space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between linear and PQ. -/// -/// @returns -/// The value a converted into PQ. -/// -/// @ingroup GPU -FfxFloat32x4 ffxApproximateLinearToPQMedium(FfxFloat32x4 a) -{ - FfxFloat32x4 b = ffxAsFloat((ffxAsUInt32(a) >> ffxBroadcast4(3u)) + ffxBroadcast4(0x378D8723u)); - FfxFloat32x4 b8 = b * b * b * b * b * b * b * b; - return b - b * (b8 - a) / (FfxFloat32(8.0) * b8); -} - -/// Compute a very accurate approximate conversion from linear to PQ space. -/// -/// PQ is very close to x^(1/8). The functions below Use the fast FfxFloat32 approximation method to do -/// PQ conversions to and from Gamma2 (4th power and fast 4th root), and PQ to and from Linear -/// (8th power and fast 8th root). The maximum error is approximately 0.2%. -/// -/// @param a The value to convert between linear and PQ. -/// -/// @returns -/// The value a converted into PQ. -/// -/// @ingroup GPU -FfxFloat32x4 ffxApproximateLinearToPQHigh(FfxFloat32x4 a) -{ - return ffxSqrt(ffxSqrt(ffxSqrt(a))); -} - -// An approximation of sine. -// -// Valid input range is {-1 to 1} representing {0 to 2 pi}, and the output range -// is {-1/4 to 1/4} representing {-1 to 1}. -// -// @param [in] value The value to calculate approximate sine for. -// -// @returns -// The approximate sine of value. -FfxFloat32 ffxParabolicSin(FfxFloat32 value) -{ - return value * abs(value) - value; -} - -// An approximation of sine. -// -// Valid input range is {-1 to 1} representing {0 to 2 pi}, and the output range -// is {-1/4 to 1/4} representing {-1 to 1}. -// -// @param [in] value The value to calculate approximate sine for. -// -// @returns -// The approximate sine of value. -FfxFloat32x2 ffxParabolicSin(FfxFloat32x2 x) -{ - return x * abs(x) - x; -} - -// An approximation of cosine. -// -// Valid input range is {-1 to 1} representing {0 to 2 pi}, and the output range -// is {-1/4 to 1/4} representing {-1 to 1}. -// -// @param [in] value The value to calculate approximate cosine for. -// -// @returns -// The approximate cosine of value. -FfxFloat32 ffxParabolicCos(FfxFloat32 x) -{ - x = ffxFract(x * FfxFloat32(0.5) + FfxFloat32(0.75)); - x = x * FfxFloat32(2.0) - FfxFloat32(1.0); - return ffxParabolicSin(x); -} - -// An approximation of cosine. -// -// Valid input range is {-1 to 1} representing {0 to 2 pi}, and the output range -// is {-1/4 to 1/4} representing {-1 to 1}. -// -// @param [in] value The value to calculate approximate cosine for. -// -// @returns -// The approximate cosine of value. -FfxFloat32x2 ffxParabolicCos(FfxFloat32x2 x) -{ - x = ffxFract(x * ffxBroadcast2(0.5f) + ffxBroadcast2(0.75f)); - x = x * ffxBroadcast2(2.0f) - ffxBroadcast2(1.0f); - return ffxParabolicSin(x); -} - -// An approximation of both sine and cosine. -// -// Valid input range is {-1 to 1} representing {0 to 2 pi}, and the output range -// is {-1/4 to 1/4} representing {-1 to 1}. -// -// @param [in] value The value to calculate approximate cosine for. -// -// @returns -// A FfxFloat32x2 containing approximations of both sine and cosine of value. -FfxFloat32x2 ffxParabolicSinCos(FfxFloat32 x) -{ - FfxFloat32 y = ffxFract(x * FfxFloat32(0.5) + FfxFloat32(0.75)); - y = y * FfxFloat32(2.0) - FfxFloat32(1.0); - return ffxParabolicSin(FfxFloat32x2(x, y)); -} - -/// Conditional free logic AND operation using values. -/// -/// @param [in] x The first value to be fed into the AND operator. -/// @param [in] y The second value to be fed into the AND operator. -/// -/// @returns -/// Result of the AND operation. -/// -/// @ingroup GPU -FfxUInt32 ffxZeroOneAnd(FfxUInt32 x, FfxUInt32 y) -{ - return min(x, y); -} - -/// Conditional free logic AND operation using two values. -/// -/// @param [in] x The first value to be fed into the AND operator. -/// @param [in] y The second value to be fed into the AND operator. -/// -/// @returns -/// Result of the AND operation. -/// -/// @ingroup GPU -FfxUInt32x2 ffxZeroOneAnd(FfxUInt32x2 x, FfxUInt32x2 y) -{ - return min(x, y); -} - -/// Conditional free logic AND operation using two values. -/// -/// @param [in] x The first value to be fed into the AND operator. -/// @param [in] y The second value to be fed into the AND operator. -/// -/// @returns -/// Result of the AND operation. -/// -/// @ingroup GPU -FfxUInt32x3 ffxZeroOneAnd(FfxUInt32x3 x, FfxUInt32x3 y) -{ - return min(x, y); -} - -/// Conditional free logic AND operation using two values. -/// -/// @param [in] x The first value to be fed into the AND operator. -/// @param [in] y The second value to be fed into the AND operator. -/// -/// @returns -/// Result of the AND operation. -/// -/// @ingroup GPU -FfxUInt32x4 ffxZeroOneAnd(FfxUInt32x4 x, FfxUInt32x4 y) -{ - return min(x, y); -} - -/// Conditional free logic NOT operation using two values. -/// -/// @param [in] x The first value to be fed into the NOT operator. -/// @param [in] y The second value to be fed into the NOT operator. -/// -/// @returns -/// Result of the NOT operation. -/// -/// @ingroup GPU -FfxUInt32 ffxZeroOneAnd(FfxUInt32 x) -{ - return x ^ FfxUInt32(1); -} - -/// Conditional free logic NOT operation using two values. -/// -/// @param [in] x The first value to be fed into the NOT operator. -/// @param [in] y The second value to be fed into the NOT operator. -/// -/// @returns -/// Result of the NOT operation. -/// -/// @ingroup GPU -FfxUInt32x2 ffxZeroOneAnd(FfxUInt32x2 x) -{ - return x ^ ffxBroadcast2(1u); -} - -/// Conditional free logic NOT operation using two values. -/// -/// @param [in] x The first value to be fed into the NOT operator. -/// @param [in] y The second value to be fed into the NOT operator. -/// -/// @returns -/// Result of the NOT operation. -/// -/// @ingroup GPU -FfxUInt32x3 ffxZeroOneAnd(FfxUInt32x3 x) -{ - return x ^ ffxBroadcast3(1u); -} - -/// Conditional free logic NOT operation using two values. -/// -/// @param [in] x The first value to be fed into the NOT operator. -/// @param [in] y The second value to be fed into the NOT operator. -/// -/// @returns -/// Result of the NOT operation. -/// -/// @ingroup GPU -FfxUInt32x4 ffxZeroOneAnd(FfxUInt32x4 x) -{ - return x ^ ffxBroadcast4(1u); -} - -/// Conditional free logic OR operation using two values. -/// -/// @param [in] x The first value to be fed into the OR operator. -/// @param [in] y The second value to be fed into the OR operator. -/// -/// @returns -/// Result of the OR operation. -/// -/// @ingroup GPU -FfxUInt32 ffxZeroOneOr(FfxUInt32 x, FfxUInt32 y) -{ - return max(x, y); -} - -/// Conditional free logic OR operation using two values. -/// -/// @param [in] x The first value to be fed into the OR operator. -/// @param [in] y The second value to be fed into the OR operator. -/// -/// @returns -/// Result of the OR operation. -/// -/// @ingroup GPU -FfxUInt32x2 ffxZeroOneOr(FfxUInt32x2 x, FfxUInt32x2 y) -{ - return max(x, y); -} - -/// Conditional free logic OR operation using two values. -/// -/// @param [in] x The first value to be fed into the OR operator. -/// @param [in] y The second value to be fed into the OR operator. -/// -/// @returns -/// Result of the OR operation. -/// -/// @ingroup GPU -FfxUInt32x3 ffxZeroOneOr(FfxUInt32x3 x, FfxUInt32x3 y) -{ - return max(x, y); -} - -/// Conditional free logic OR operation using two values. -/// -/// @param [in] x The first value to be fed into the OR operator. -/// @param [in] y The second value to be fed into the OR operator. -/// -/// @returns -/// Result of the OR operation. -/// -/// @ingroup GPU -FfxUInt32x4 ffxZeroOneOr(FfxUInt32x4 x, FfxUInt32x4 y) -{ - return max(x, y); -} - -/// Conditional free logic signed NOT operation using two half-precision FfxFloat32 values. -/// -/// @param [in] x The first value to be fed into the AND OR operator. -/// -/// @returns -/// Result of the AND OR operation. -/// -/// @ingroup GPU -FfxUInt32 ffxZeroOneAndToU1(FfxFloat32 x) -{ - return FfxUInt32(FfxFloat32(1.0) - x); -} - -/// Conditional free logic signed NOT operation using two half-precision FfxFloat32 values. -/// -/// @param [in] x The first value to be fed into the AND OR operator. -/// -/// @returns -/// Result of the AND OR operation. -/// -/// @ingroup GPU -FfxUInt32x2 ffxZeroOneAndToU2(FfxFloat32x2 x) -{ - return FfxUInt32x2(ffxBroadcast2(1.0) - x); -} - -/// Conditional free logic signed NOT operation using two half-precision FfxFloat32 values. -/// -/// @param [in] x The first value to be fed into the AND OR operator. -/// -/// @returns -/// Result of the AND OR operation. -/// -/// @ingroup GPU -FfxUInt32x3 ffxZeroOneAndToU3(FfxFloat32x3 x) -{ - return FfxUInt32x3(ffxBroadcast3(1.0) - x); -} - -/// Conditional free logic signed NOT operation using two half-precision FfxFloat32 values. -/// -/// @param [in] x The first value to be fed into the AND OR operator. -/// -/// @returns -/// Result of the AND OR operation. -/// -/// @ingroup GPU -FfxUInt32x4 ffxZeroOneAndToU4(FfxFloat32x4 x) -{ - return FfxUInt32x4(ffxBroadcast4(1.0) - x); -} - -/// Conditional free logic AND operation using two values followed by a NOT operation -/// using the resulting value and a third value. -/// -/// @param [in] x The first value to be fed into the AND operator. -/// @param [in] y The second value to be fed into the AND operator. -/// @param [in] z The second value to be fed into the OR operator. -/// -/// @returns -/// Result of the AND OR operation. -/// -/// @ingroup GPU -FfxFloat32 ffxZeroOneAndOr(FfxFloat32 x, FfxFloat32 y, FfxFloat32 z) -{ - return ffxSaturate(x * y + z); -} - -/// Conditional free logic AND operation using two values followed by a NOT operation -/// using the resulting value and a third value. -/// -/// @param [in] x The first value to be fed into the AND operator. -/// @param [in] y The second value to be fed into the AND operator. -/// @param [in] z The second value to be fed into the OR operator. -/// -/// @returns -/// Result of the AND OR operation. -/// -/// @ingroup GPU -FfxFloat32x2 ffxZeroOneAndOr(FfxFloat32x2 x, FfxFloat32x2 y, FfxFloat32x2 z) -{ - return ffxSaturate(x * y + z); -} - -/// Conditional free logic AND operation using two values followed by a NOT operation -/// using the resulting value and a third value. -/// -/// @param [in] x The first value to be fed into the AND operator. -/// @param [in] y The second value to be fed into the AND operator. -/// @param [in] z The second value to be fed into the OR operator. -/// -/// @returns -/// Result of the AND OR operation. -/// -/// @ingroup GPU -FfxFloat32x3 ffxZeroOneAndOr(FfxFloat32x3 x, FfxFloat32x3 y, FfxFloat32x3 z) -{ - return ffxSaturate(x * y + z); -} - -/// Conditional free logic AND operation using two values followed by a NOT operation -/// using the resulting value and a third value. -/// -/// @param [in] x The first value to be fed into the AND operator. -/// @param [in] y The second value to be fed into the AND operator. -/// @param [in] z The second value to be fed into the OR operator. -/// -/// @returns -/// Result of the AND OR operation. -/// -/// @ingroup GPU -FfxFloat32x4 ffxZeroOneAndOr(FfxFloat32x4 x, FfxFloat32x4 y, FfxFloat32x4 z) -{ - return ffxSaturate(x * y + z); -} - -/// Given a value, returns 1.0 if greater than zero and 0.0 if not. -/// -/// @param [in] x The value to be compared. -/// -/// @returns -/// Result of the greater than zero comparison. -/// -/// @ingroup GPU -FfxFloat32 ffxZeroOneIsGreaterThanZero(FfxFloat32 x) -{ - return ffxSaturate(x * FfxFloat32(FFX_POSITIVE_INFINITY_FLOAT)); -} - -/// Given a value, returns 1.0 if greater than zero and 0.0 if not. -/// -/// @param [in] x The value to be compared. -/// -/// @returns -/// Result of the greater than zero comparison. -/// -/// @ingroup GPU -FfxFloat32x2 ffxZeroOneIsGreaterThanZero(FfxFloat32x2 x) -{ - return ffxSaturate(x * ffxBroadcast2(FFX_POSITIVE_INFINITY_FLOAT)); -} - -/// Given a value, returns 1.0 if greater than zero and 0.0 if not. -/// -/// @param [in] x The value to be compared. -/// -/// @returns -/// Result of the greater than zero comparison. -/// -/// @ingroup GPU -FfxFloat32x3 ffxZeroOneIsGreaterThanZero(FfxFloat32x3 x) -{ - return ffxSaturate(x * ffxBroadcast3(FFX_POSITIVE_INFINITY_FLOAT)); -} - -/// Given a value, returns 1.0 if greater than zero and 0.0 if not. -/// -/// @param [in] x The value to be compared. -/// -/// @returns -/// Result of the greater than zero comparison. -/// -/// @ingroup GPU -FfxFloat32x4 ffxZeroOneIsGreaterThanZero(FfxFloat32x4 x) -{ - return ffxSaturate(x * ffxBroadcast4(FFX_POSITIVE_INFINITY_FLOAT)); -} - -/// Conditional free logic signed NOT operation using two FfxFloat32 values. -/// -/// @param [in] x The first value to be fed into the AND OR operator. -/// -/// @returns -/// Result of the AND OR operation. -/// -/// @ingroup GPU -FfxFloat32 ffxZeroOneAnd(FfxFloat32 x) -{ - return FfxFloat32(1.0) - x; -} - -/// Conditional free logic signed NOT operation using two FfxFloat32 values. -/// -/// @param [in] x The first value to be fed into the AND OR operator. -/// -/// @returns -/// Result of the AND OR operation. -/// -/// @ingroup GPU -FfxFloat32x2 ffxZeroOneAnd(FfxFloat32x2 x) -{ - return ffxBroadcast2(1.0) - x; -} - -/// Conditional free logic signed NOT operation using two FfxFloat32 values. -/// -/// @param [in] x The first value to be fed into the AND OR operator. -/// -/// @returns -/// Result of the AND OR operation. -/// -/// @ingroup GPU -FfxFloat32x3 ffxZeroOneAnd(FfxFloat32x3 x) -{ - return ffxBroadcast3(1.0) - x; -} - -/// Conditional free logic signed NOT operation using two FfxFloat32 values. -/// -/// @param [in] x The first value to be fed into the AND OR operator. -/// -/// @returns -/// Result of the AND OR operation. -/// -/// @ingroup GPU -FfxFloat32x4 ffxZeroOneAnd(FfxFloat32x4 x) -{ - return ffxBroadcast4(1.0) - x; -} - -/// Conditional free logic OR operation using two FfxFloat32 values. -/// -/// @param [in] x The first value to be fed into the OR operator. -/// @param [in] y The second value to be fed into the OR operator. -/// -/// @returns -/// Result of the OR operation. -/// -/// @ingroup GPU -FfxFloat32 ffxZeroOneOr(FfxFloat32 x, FfxFloat32 y) -{ - return max(x, y); -} - -/// Conditional free logic OR operation using two FfxFloat32 values. -/// -/// @param [in] x The first value to be fed into the OR operator. -/// @param [in] y The second value to be fed into the OR operator. -/// -/// @returns -/// Result of the OR operation. -/// -/// @ingroup GPU -FfxFloat32x2 ffxZeroOneOr(FfxFloat32x2 x, FfxFloat32x2 y) -{ - return max(x, y); -} - -/// Conditional free logic OR operation using two FfxFloat32 values. -/// -/// @param [in] x The first value to be fed into the OR operator. -/// @param [in] y The second value to be fed into the OR operator. -/// -/// @returns -/// Result of the OR operation. -/// -/// @ingroup GPU -FfxFloat32x3 ffxZeroOneOr(FfxFloat32x3 x, FfxFloat32x3 y) -{ - return max(x, y); -} - -/// Conditional free logic OR operation using two FfxFloat32 values. -/// -/// @param [in] x The first value to be fed into the OR operator. -/// @param [in] y The second value to be fed into the OR operator. -/// -/// @returns -/// Result of the OR operation. -/// -/// @ingroup GPU -FfxFloat32x4 ffxZeroOneOr(FfxFloat32x4 x, FfxFloat32x4 y) -{ - return max(x, y); -} - -/// Choose between two FfxFloat32 values if the first paramter is greater than zero. -/// -/// @param [in] x The value to compare against zero. -/// @param [in] y The value to return if the comparision is greater than zero. -/// @param [in] z The value to return if the comparision is less than or equal to zero. -/// -/// @returns -/// The selected value. -/// -/// @ingroup GPU -FfxFloat32 ffxZeroOneSelect(FfxFloat32 x, FfxFloat32 y, FfxFloat32 z) -{ - FfxFloat32 r = (-x) * z + z; - return x * y + r; -} - -/// Choose between two FfxFloat32 values if the first paramter is greater than zero. -/// -/// @param [in] x The value to compare against zero. -/// @param [in] y The value to return if the comparision is greater than zero. -/// @param [in] z The value to return if the comparision is less than or equal to zero. -/// -/// @returns -/// The selected value. -/// -/// @ingroup GPU -FfxFloat32x2 ffxZeroOneSelect(FfxFloat32x2 x, FfxFloat32x2 y, FfxFloat32x2 z) -{ - FfxFloat32x2 r = (-x) * z + z; - return x * y + r; -} - -/// Choose between two FfxFloat32 values if the first paramter is greater than zero. -/// -/// @param [in] x The value to compare against zero. -/// @param [in] y The value to return if the comparision is greater than zero. -/// @param [in] z The value to return if the comparision is less than or equal to zero. -/// -/// @returns -/// The selected value. -/// -/// @ingroup GPU -FfxFloat32x3 ffxZeroOneSelect(FfxFloat32x3 x, FfxFloat32x3 y, FfxFloat32x3 z) -{ - FfxFloat32x3 r = (-x) * z + z; - return x * y + r; -} - -/// Choose between two FfxFloat32 values if the first paramter is greater than zero. -/// -/// @param [in] x The value to compare against zero. -/// @param [in] y The value to return if the comparision is greater than zero. -/// @param [in] z The value to return if the comparision is less than or equal to zero. -/// -/// @returns -/// The selected value. -/// -/// @ingroup GPU -FfxFloat32x4 ffxZeroOneSelect(FfxFloat32x4 x, FfxFloat32x4 y, FfxFloat32x4 z) -{ - FfxFloat32x4 r = (-x) * z + z; - return x * y + r; -} - -/// Given a value, returns 1.0 if less than zero and 0.0 if not. -/// -/// @param [in] x The value to be compared. -/// -/// @returns -/// Result of the sign value. -/// -/// @ingroup GPU -FfxFloat32 ffxZeroOneIsSigned(FfxFloat32 x) -{ - return ffxSaturate(x * FfxFloat32(FFX_NEGATIVE_INFINITY_FLOAT)); -} - -/// Given a value, returns 1.0 if less than zero and 0.0 if not. -/// -/// @param [in] x The value to be compared. -/// -/// @returns -/// Result of the sign value. -/// -/// @ingroup GPU -FfxFloat32x2 ffxZeroOneIsSigned(FfxFloat32x2 x) -{ - return ffxSaturate(x * ffxBroadcast2(FFX_NEGATIVE_INFINITY_FLOAT)); -} - -/// Given a value, returns 1.0 if less than zero and 0.0 if not. -/// -/// @param [in] x The value to be compared. -/// -/// @returns -/// Result of the sign value. -/// -/// @ingroup GPU -FfxFloat32x3 ffxZeroOneIsSigned(FfxFloat32x3 x) -{ - return ffxSaturate(x * ffxBroadcast3(FFX_NEGATIVE_INFINITY_FLOAT)); -} - -/// Given a value, returns 1.0 if less than zero and 0.0 if not. -/// -/// @param [in] x The value to be compared. -/// -/// @returns -/// Result of the sign value. -/// -/// @ingroup GPU -FfxFloat32x4 ffxZeroOneIsSigned(FfxFloat32x4 x) -{ - return ffxSaturate(x * ffxBroadcast4(FFX_NEGATIVE_INFINITY_FLOAT)); -} - -/// Compute a Rec.709 color space. -/// -/// Rec.709 is used for some HDTVs. -/// -/// Both Rec.709 and sRGB have a linear segment which as spec'ed would intersect the curved segment 2 times. -/// (a.) For 8-bit sRGB, steps {0 to 10.3} are in the linear region (4% of the encoding range). -/// (b.) For 8-bit 709, steps {0 to 20.7} are in the linear region (8% of the encoding range). -/// -/// @param [in] color The color to convert to Rec. 709. -/// -/// @returns -/// The color in linear space. -/// -/// @ingroup GPU -FfxFloat32 ffxRec709FromLinear(FfxFloat32 color) -{ - FfxFloat32x3 j = FfxFloat32x3(0.018 * 4.5, 4.5, 0.45); - FfxFloat32x2 k = FfxFloat32x2(1.099, -0.099); - return clamp(j.x, color * j.y, pow(color, j.z) * k.x + k.y); -} - -/// Compute a Rec.709 color space. -/// -/// Rec.709 is used for some HDTVs. -/// -/// Both Rec.709 and sRGB have a linear segment which as spec'ed would intersect the curved segment 2 times. -/// (a.) For 8-bit sRGB, steps {0 to 10.3} are in the linear region (4% of the encoding range). -/// (b.) For 8-bit 709, steps {0 to 20.7} are in the linear region (8% of the encoding range). -/// -/// @param [in] color The color to convert to Rec. 709. -/// -/// @returns -/// The color in linear space. -/// -/// @ingroup GPU -FfxFloat32x2 ffxRec709FromLinear(FfxFloat32x2 color) -{ - FfxFloat32x3 j = FfxFloat32x3(0.018 * 4.5, 4.5, 0.45); - FfxFloat32x2 k = FfxFloat32x2(1.099, -0.099); - return clamp(j.xx, color * j.yy, pow(color, j.zz) * k.xx + k.yy); -} - -/// Compute a Rec.709 color space. -/// -/// Rec.709 is used for some HDTVs. -/// -/// Both Rec.709 and sRGB have a linear segment which as spec'ed would intersect the curved segment 2 times. -/// (a.) For 8-bit sRGB, steps {0 to 10.3} are in the linear region (4% of the encoding range). -/// (b.) For 8-bit 709, steps {0 to 20.7} are in the linear region (8% of the encoding range). -/// -/// @param [in] color The color to convert to Rec. 709. -/// -/// @returns -/// The color in linear space. -/// -/// @ingroup GPU -FfxFloat32x3 ffxRec709FromLinear(FfxFloat32x3 color) -{ - FfxFloat32x3 j = FfxFloat32x3(0.018 * 4.5, 4.5, 0.45); - FfxFloat32x2 k = FfxFloat32x2(1.099, -0.099); - return clamp(j.xxx, color * j.yyy, pow(color, j.zzz) * k.xxx + k.yyy); -} - -/// Compute a gamma value from a linear value. -/// -/// Typically 2.2 for some PC displays, or 2.4-2.5 for CRTs, or 2.2 FreeSync2 native. -/// -/// Note: 'rcpX' is '1/x', where the 'x' is what would be used in ffxLinearFromGamma. -/// -/// @param [in] value The value to convert to gamma space from linear. -/// @param [in] power The reciprocal of power value used for the gamma curve. -/// -/// @returns -/// A value in gamma space. -/// -/// @ingroup GPU -FfxFloat32 ffxGammaFromLinear(FfxFloat32 color, FfxFloat32 rcpX) -{ - return pow(color, FfxFloat32(rcpX)); -} - -/// Compute a gamma value from a linear value. -/// -/// Typically 2.2 for some PC displays, or 2.4-2.5 for CRTs, or 2.2 FreeSync2 native. -/// -/// Note: 'rcpX' is '1/x', where the 'x' is what would be used in ffxLinearFromGamma. -/// -/// @param [in] value The value to convert to gamma space from linear. -/// @param [in] power The reciprocal of power value used for the gamma curve. -/// -/// @returns -/// A value in gamma space. -/// -/// @ingroup GPU -FfxFloat32x2 ffxGammaFromLinear(FfxFloat32x2 color, FfxFloat32 rcpX) -{ - return pow(color, ffxBroadcast2(rcpX)); -} - -/// Compute a gamma value from a linear value. -/// -/// Typically 2.2 for some PC displays, or 2.4-2.5 for CRTs, or 2.2 FreeSync2 native. -/// -/// Note: 'rcpX' is '1/x', where the 'x' is what would be used in ffxLinearFromGamma. -/// -/// @param [in] value The value to convert to gamma space from linear. -/// @param [in] power The reciprocal of power value used for the gamma curve. -/// -/// @returns -/// A value in gamma space. -/// -/// @ingroup GPU -FfxFloat32x3 ffxGammaFromLinear(FfxFloat32x3 color, FfxFloat32 rcpX) -{ - return pow(color, ffxBroadcast3(rcpX)); -} - -/// Compute a PQ value from a linear value. -/// -/// @param [in] value The value to convert to PQ from linear. -/// -/// @returns -/// A value in linear space. -/// -/// @ingroup GPU -FfxFloat32 ffxPQToLinear(FfxFloat32 x) -{ - FfxFloat32 p = pow(x, FfxFloat32(0.159302)); - return pow((FfxFloat32(0.835938) + FfxFloat32(18.8516) * p) / (FfxFloat32(1.0) + FfxFloat32(18.6875) * p), FfxFloat32(78.8438)); -} - -/// Compute a PQ value from a linear value. -/// -/// @param [in] value The value to convert to PQ from linear. -/// -/// @returns -/// A value in linear space. -/// -/// @ingroup GPU -FfxFloat32x2 ffxPQToLinear(FfxFloat32x2 x) -{ - FfxFloat32x2 p = pow(x, ffxBroadcast2(0.159302)); - return pow((ffxBroadcast2(0.835938) + ffxBroadcast2(18.8516) * p) / (ffxBroadcast2(1.0) + ffxBroadcast2(18.6875) * p), ffxBroadcast2(78.8438)); -} - -/// Compute a PQ value from a linear value. -/// -/// @param [in] value The value to convert to PQ from linear. -/// -/// @returns -/// A value in linear space. -/// -/// @ingroup GPU -FfxFloat32x3 ffxPQToLinear(FfxFloat32x3 x) -{ - FfxFloat32x3 p = pow(x, ffxBroadcast3(0.159302)); - return pow((ffxBroadcast3(0.835938) + ffxBroadcast3(18.8516) * p) / (ffxBroadcast3(1.0) + ffxBroadcast3(18.6875) * p), ffxBroadcast3(78.8438)); -} - -/// Compute a linear value from a SRGB value. -/// -/// @param [in] value The value to convert to linear from SRGB. -/// -/// @returns -/// A value in SRGB space. -/// -/// @ingroup GPU -FfxFloat32 ffxSrgbToLinear(FfxFloat32 color) -{ - FfxFloat32x3 j = FfxFloat32x3(0.0031308 * 12.92, 12.92, 1.0 / 2.4); - FfxFloat32x2 k = FfxFloat32x2(1.055, -0.055); - return clamp(j.x, color * j.y, pow(color, j.z) * k.x + k.y); -} - -/// Compute a linear value from a SRGB value. -/// -/// @param [in] value The value to convert to linear from SRGB. -/// -/// @returns -/// A value in SRGB space. -/// -/// @ingroup GPU -FfxFloat32x2 ffxSrgbToLinear(FfxFloat32x2 color) -{ - FfxFloat32x3 j = FfxFloat32x3(0.0031308 * 12.92, 12.92, 1.0 / 2.4); - FfxFloat32x2 k = FfxFloat32x2(1.055, -0.055); - return clamp(j.xx, color * j.yy, pow(color, j.zz) * k.xx + k.yy); -} - -/// Compute a linear value from a SRGB value. -/// -/// @param [in] value The value to convert to linear from SRGB. -/// -/// @returns -/// A value in SRGB space. -/// -/// @ingroup GPU -FfxFloat32x3 ffxSrgbToLinear(FfxFloat32x3 color) -{ - FfxFloat32x3 j = FfxFloat32x3(0.0031308 * 12.92, 12.92, 1.0 / 2.4); - FfxFloat32x2 k = FfxFloat32x2(1.055, -0.055); - return clamp(j.xxx, color * j.yyy, pow(color, j.zzz) * k.xxx + k.yyy); -} - -/// Compute a linear value from a REC.709 value. -/// -/// @param [in] color The value to convert to linear from REC.709. -/// -/// @returns -/// A value in linear space. -/// -/// @ingroup GPU -FfxFloat32 ffxLinearFromRec709(FfxFloat32 color) -{ - FfxFloat32x3 j = FfxFloat32x3(0.081 / 4.5, 1.0 / 4.5, 1.0 / 0.45); - FfxFloat32x2 k = FfxFloat32x2(1.0 / 1.099, 0.099 / 1.099); - return ffxZeroOneSelect(ffxZeroOneIsSigned(color - j.x), color * j.y, pow(color * k.x + k.y, j.z)); -} - -/// Compute a linear value from a REC.709 value. -/// -/// @param [in] color The value to convert to linear from REC.709. -/// -/// @returns -/// A value in linear space. -/// -/// @ingroup GPU -FfxFloat32x2 ffxLinearFromRec709(FfxFloat32x2 color) -{ - FfxFloat32x3 j = FfxFloat32x3(0.081 / 4.5, 1.0 / 4.5, 1.0 / 0.45); - FfxFloat32x2 k = FfxFloat32x2(1.0 / 1.099, 0.099 / 1.099); - return ffxZeroOneSelect(ffxZeroOneIsSigned(color - j.xx), color * j.yy, pow(color * k.xx + k.yy, j.zz)); -} - -/// Compute a linear value from a REC.709 value. -/// -/// @param [in] color The value to convert to linear from REC.709. -/// -/// @returns -/// A value in linear space. -/// -/// @ingroup GPU -FfxFloat32x3 ffxLinearFromRec709(FfxFloat32x3 color) -{ - FfxFloat32x3 j = FfxFloat32x3(0.081 / 4.5, 1.0 / 4.5, 1.0 / 0.45); - FfxFloat32x2 k = FfxFloat32x2(1.0 / 1.099, 0.099 / 1.099); - return ffxZeroOneSelect(ffxZeroOneIsSigned(color - j.xxx), color * j.yyy, pow(color * k.xxx + k.yyy, j.zzz)); -} - -/// Compute a linear value from a value in a gamma space. -/// -/// Typically 2.2 for some PC displays, or 2.4-2.5 for CRTs, or 2.2 FreeSync2 native. -/// -/// @param [in] color The value to convert to linear in gamma space. -/// @param [in] power The power value used for the gamma curve. -/// -/// @returns -/// A value in linear space. -/// -/// @ingroup GPU -FfxFloat32 ffxLinearFromGamma(FfxFloat32 color, FfxFloat32 power) -{ - return pow(color, FfxFloat32(power)); -} - -/// Compute a linear value from a value in a gamma space. -/// -/// Typically 2.2 for some PC displays, or 2.4-2.5 for CRTs, or 2.2 FreeSync2 native. -/// -/// @param [in] color The value to convert to linear in gamma space. -/// @param [in] power The power value used for the gamma curve. -/// -/// @returns -/// A value in linear space. -/// -/// @ingroup GPU -FfxFloat32x2 ffxLinearFromGamma(FfxFloat32x2 color, FfxFloat32 power) -{ - return pow(color, ffxBroadcast2(power)); -} - -/// Compute a linear value from a value in a gamma space. -/// -/// Typically 2.2 for some PC displays, or 2.4-2.5 for CRTs, or 2.2 FreeSync2 native. -/// -/// @param [in] color The value to convert to linear in gamma space. -/// @param [in] power The power value used for the gamma curve. -/// -/// @returns -/// A value in linear space. -/// -/// @ingroup GPU -FfxFloat32x3 ffxLinearFromGamma(FfxFloat32x3 color, FfxFloat32 power) -{ - return pow(color, ffxBroadcast3(power)); -} - -/// Compute a linear value from a value in a PQ space. -/// -/// Typically 2.2 for some PC displays, or 2.4-2.5 for CRTs, or 2.2 FreeSync2 native. -/// -/// @param [in] value The value to convert to linear in PQ space. -/// -/// @returns -/// A value in linear space. -/// -/// @ingroup GPU -FfxFloat32 ffxLinearFromPQ(FfxFloat32 x) -{ - FfxFloat32 p = pow(x, FfxFloat32(0.0126833)); - return pow(ffxSaturate(p - FfxFloat32(0.835938)) / (FfxFloat32(18.8516) - FfxFloat32(18.6875) * p), FfxFloat32(6.27739)); -} - -/// Compute a linear value from a value in a PQ space. -/// -/// Typically 2.2 for some PC displays, or 2.4-2.5 for CRTs, or 2.2 FreeSync2 native. -/// -/// @param [in] value The value to convert to linear in PQ space. -/// -/// @returns -/// A value in linear space. -/// -/// @ingroup GPU -FfxFloat32x2 ffxLinearFromPQ(FfxFloat32x2 x) -{ - FfxFloat32x2 p = pow(x, ffxBroadcast2(0.0126833)); - return pow(ffxSaturate(p - ffxBroadcast2(0.835938)) / (ffxBroadcast2(18.8516) - ffxBroadcast2(18.6875) * p), ffxBroadcast2(6.27739)); -} - -/// Compute a linear value from a value in a PQ space. -/// -/// Typically 2.2 for some PC displays, or 2.4-2.5 for CRTs, or 2.2 FreeSync2 native. -/// -/// @param [in] value The value to convert to linear in PQ space. -/// -/// @returns -/// A value in linear space. -/// -/// @ingroup GPU -FfxFloat32x3 ffxLinearFromPQ(FfxFloat32x3 x) -{ - FfxFloat32x3 p = pow(x, ffxBroadcast3(0.0126833)); - return pow(ffxSaturate(p - ffxBroadcast3(0.835938)) / (ffxBroadcast3(18.8516) - ffxBroadcast3(18.6875) * p), ffxBroadcast3(6.27739)); -} - -/// Compute a linear value from a value in a SRGB space. -/// -/// Typically 2.2 for some PC displays, or 2.4-2.5 for CRTs, or 2.2 FreeSync2 native. -/// -/// @param [in] value The value to convert to linear in SRGB space. -/// -/// @returns -/// A value in linear space. -/// -/// @ingroup GPU -FfxFloat32 ffxLinearFromSrgb(FfxFloat32 color) -{ - FfxFloat32x3 j = FfxFloat32x3(0.04045 / 12.92, 1.0 / 12.92, 2.4); - FfxFloat32x2 k = FfxFloat32x2(1.0 / 1.055, 0.055 / 1.055); - return ffxZeroOneSelect(ffxZeroOneIsSigned(color - j.x), color * j.y, pow(color * k.x + k.y, j.z)); -} - -/// Compute a linear value from a value in a SRGB space. -/// -/// Typically 2.2 for some PC displays, or 2.4-2.5 for CRTs, or 2.2 FreeSync2 native. -/// -/// @param [in] value The value to convert to linear in SRGB space. -/// -/// @returns -/// A value in linear space. -/// -/// @ingroup GPU -FfxFloat32x2 ffxLinearFromSrgb(FfxFloat32x2 color) -{ - FfxFloat32x3 j = FfxFloat32x3(0.04045 / 12.92, 1.0 / 12.92, 2.4); - FfxFloat32x2 k = FfxFloat32x2(1.0 / 1.055, 0.055 / 1.055); - return ffxZeroOneSelect(ffxZeroOneIsSigned(color - j.xx), color * j.yy, pow(color * k.xx + k.yy, j.zz)); -} - -/// Compute a linear value from a value in a SRGB space. -/// -/// Typically 2.2 for some PC displays, or 2.4-2.5 for CRTs, or 2.2 FreeSync2 native. -/// -/// @param [in] value The value to convert to linear in SRGB space. -/// -/// @returns -/// A value in linear space. -/// -/// @ingroup GPU -FfxFloat32x3 ffxLinearFromSrgb(FfxFloat32x3 color) -{ - FfxFloat32x3 j = FfxFloat32x3(0.04045 / 12.92, 1.0 / 12.92, 2.4); - FfxFloat32x2 k = FfxFloat32x2(1.0 / 1.055, 0.055 / 1.055); - return ffxZeroOneSelect(ffxZeroOneIsSigned(color - j.xxx), color * j.yyy, pow(color * k.xxx + k.yyy, j.zzz)); -} - -/// A remapping of 64x1 to 8x8 imposing rotated 2x2 pixel quads in quad linear. -/// -/// 543210 -/// ====== -/// ..xxx. -/// yy...y -/// -/// @param [in] a The input 1D coordinates to remap. -/// -/// @returns -/// The remapped 2D coordinates. -/// -/// @ingroup GPU -FfxUInt32x2 ffxRemapForQuad(FfxUInt32 a) -{ - return FfxUInt32x2(bitfieldExtract(a, 1u, 3u), bitfieldInsertMask(bitfieldExtract(a, 3u, 3u), a, 1u)); -} - -/// A helper function performing a remap 64x1 to 8x8 remapping which is necessary for 2D wave reductions. -/// -/// The 64-wide lane indices to 8x8 remapping is performed as follows: -/// -/// 00 01 08 09 10 11 18 19 -/// 02 03 0a 0b 12 13 1a 1b -/// 04 05 0c 0d 14 15 1c 1d -/// 06 07 0e 0f 16 17 1e 1f -/// 20 21 28 29 30 31 38 39 -/// 22 23 2a 2b 32 33 3a 3b -/// 24 25 2c 2d 34 35 3c 3d -/// 26 27 2e 2f 36 37 3e 3f -/// -/// @param [in] a The input 1D coordinate to remap. -/// -/// @returns -/// The remapped 2D coordinates. -/// -/// @ingroup GPU -FfxUInt32x2 ffxRemapForWaveReduction(FfxUInt32 a) -{ - return FfxUInt32x2(bitfieldInsertMask(bitfieldExtract(a, 2u, 3u), a, 1u), bitfieldInsertMask(bitfieldExtract(a, 3u, 3u), bitfieldExtract(a, 1u, 2u), 2u)); -} diff --git a/Assets/Resources/FSR2/shaders/ffx_core_gpu_common.h.meta b/Assets/Resources/FSR2/shaders/ffx_core_gpu_common.h.meta deleted file mode 100644 index 8f9711f..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_core_gpu_common.h.meta +++ /dev/null @@ -1,60 +0,0 @@ -fileFormatVersion: 2 -guid: 402c509393f5bf647b41a962a48ed8e2 -PluginImporter: - externalObjects: {} - serializedVersion: 2 - iconMap: {} - executionOrder: {} - defineConstraints: [] - isPreloaded: 0 - isOverridable: 0 - isExplicitlyReferenced: 0 - validateReferences: 1 - platformData: - - first: - : Any - second: - enabled: 0 - settings: - Exclude Editor: 1 - Exclude GameCoreScarlett: 1 - Exclude GameCoreXboxOne: 1 - Exclude Linux64: 1 - Exclude OSXUniversal: 1 - Exclude PS4: 1 - Exclude PS5: 1 - Exclude WebGL: 1 - Exclude Win: 1 - Exclude Win64: 1 - - first: - Any: - second: - enabled: 0 - settings: {} - - first: - Editor: Editor - second: - enabled: 0 - settings: - DefaultValueInitialized: true - - first: - Standalone: Linux64 - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win64 - second: - enabled: 0 - settings: - CPU: None - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_core_gpu_common_half.h b/Assets/Resources/FSR2/shaders/ffx_core_gpu_common_half.h deleted file mode 100644 index c46ccb3..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_core_gpu_common_half.h +++ /dev/null @@ -1,2978 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#if FFX_HALF -#if FFX_HLSL_6_2 -/// A define value for 16bit positive infinity. -/// -/// @ingroup GPU -#define FFX_POSITIVE_INFINITY_HALF FFX_TO_FLOAT16((uint16_t)0x7c00u) - -/// A define value for 16bit negative infinity. -/// -/// @ingroup GPU -#define FFX_NEGATIVE_INFINITY_HALF FFX_TO_FLOAT16((uint16_t)0xfc00u) -#else -/// A define value for 16bit positive infinity. -/// -/// @ingroup GPU -#define FFX_POSITIVE_INFINITY_HALF FFX_TO_FLOAT16(0x7c00u) - -/// A define value for 16bit negative infinity. -/// -/// @ingroup GPU -#define FFX_NEGATIVE_INFINITY_HALF FFX_TO_FLOAT16(0xfc00u) -#endif // FFX_HLSL_6_2 - -/// Compute the min of two values. -/// -/// @param [in] x The first value to compute the min of. -/// @param [in] y The second value to compute the min of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxFloat16 ffxMin(FfxFloat16 x, FfxFloat16 y) -{ - return min(x, y); -} - -/// Compute the min of two values. -/// -/// @param [in] x The first value to compute the min of. -/// @param [in] y The second value to compute the min of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxFloat16x2 ffxMin(FfxFloat16x2 x, FfxFloat16x2 y) -{ - return min(x, y); -} - -/// Compute the min of two values. -/// -/// @param [in] x The first value to compute the min of. -/// @param [in] y The second value to compute the min of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxFloat16x3 ffxMin(FfxFloat16x3 x, FfxFloat16x3 y) -{ - return min(x, y); -} - -/// Compute the min of two values. -/// -/// @param [in] x The first value to compute the min of. -/// @param [in] y The second value to compute the min of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxFloat16x4 ffxMin(FfxFloat16x4 x, FfxFloat16x4 y) -{ - return min(x, y); -} - -/// Compute the min of two values. -/// -/// @param [in] x The first value to compute the min of. -/// @param [in] y The second value to compute the min of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxInt16 ffxMin(FfxInt16 x, FfxInt16 y) -{ - return min(x, y); -} - -/// Compute the min of two values. -/// -/// @param [in] x The first value to compute the min of. -/// @param [in] y The second value to compute the min of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxInt16x2 ffxMin(FfxInt16x2 x, FfxInt16x2 y) -{ - return min(x, y); -} - -/// Compute the min of two values. -/// -/// @param [in] x The first value to compute the min of. -/// @param [in] y The second value to compute the min of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxInt16x3 ffxMin(FfxInt16x3 x, FfxInt16x3 y) -{ - return min(x, y); -} - -/// Compute the min of two values. -/// -/// @param [in] x The first value to compute the min of. -/// @param [in] y The second value to compute the min of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxInt16x4 ffxMin(FfxInt16x4 x, FfxInt16x4 y) -{ - return min(x, y); -} - -/// Compute the min of two values. -/// -/// @param [in] x The first value to compute the min of. -/// @param [in] y The second value to compute the min of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxUInt16 ffxMin(FfxUInt16 x, FfxUInt16 y) -{ - return min(x, y); -} - -/// Compute the min of two values. -/// -/// @param [in] x The first value to compute the min of. -/// @param [in] y The second value to compute the min of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxUInt16x2 ffxMin(FfxUInt16x2 x, FfxUInt16x2 y) -{ - return min(x, y); -} - -/// Compute the min of two values. -/// -/// @param [in] x The first value to compute the min of. -/// @param [in] y The second value to compute the min of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxUInt16x3 ffxMin(FfxUInt16x3 x, FfxUInt16x3 y) -{ - return min(x, y); -} - -/// Compute the min of two values. -/// -/// @param [in] x The first value to compute the min of. -/// @param [in] y The second value to compute the min of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxUInt16x4 ffxMin(FfxUInt16x4 x, FfxUInt16x4 y) -{ - return min(x, y); -} - -/// Compute the max of two values. -/// -/// @param [in] x The first value to compute the max of. -/// @param [in] y The second value to compute the max of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxFloat16 ffxMax(FfxFloat16 x, FfxFloat16 y) -{ - return max(x, y); -} - -/// Compute the max of two values. -/// -/// @param [in] x The first value to compute the max of. -/// @param [in] y The second value to compute the max of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxFloat16x2 ffxMax(FfxFloat16x2 x, FfxFloat16x2 y) -{ - return max(x, y); -} - -/// Compute the max of two values. -/// -/// @param [in] x The first value to compute the max of. -/// @param [in] y The second value to compute the max of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxFloat16x3 ffxMax(FfxFloat16x3 x, FfxFloat16x3 y) -{ - return max(x, y); -} - -/// Compute the max of two values. -/// -/// @param [in] x The first value to compute the max of. -/// @param [in] y The second value to compute the max of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxFloat16x4 ffxMax(FfxFloat16x4 x, FfxFloat16x4 y) -{ - return max(x, y); -} - -/// Compute the max of two values. -/// -/// @param [in] x The first value to compute the max of. -/// @param [in] y The second value to compute the max of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxInt16 ffxMax(FfxInt16 x, FfxInt16 y) -{ - return max(x, y); -} - -/// Compute the max of two values. -/// -/// @param [in] x The first value to compute the max of. -/// @param [in] y The second value to compute the max of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxInt16x2 ffxMax(FfxInt16x2 x, FfxInt16x2 y) -{ - return max(x, y); -} - -/// Compute the max of two values. -/// -/// @param [in] x The first value to compute the max of. -/// @param [in] y The second value to compute the max of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxInt16x3 ffxMax(FfxInt16x3 x, FfxInt16x3 y) -{ - return max(x, y); -} - -/// Compute the max of two values. -/// -/// @param [in] x The first value to compute the max of. -/// @param [in] y The second value to compute the max of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxInt16x4 ffxMax(FfxInt16x4 x, FfxInt16x4 y) -{ - return max(x, y); -} - -/// Compute the max of two values. -/// -/// @param [in] x The first value to compute the max of. -/// @param [in] y The second value to compute the max of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxUInt16 ffxMax(FfxUInt16 x, FfxUInt16 y) -{ - return max(x, y); -} - -/// Compute the max of two values. -/// -/// @param [in] x The first value to compute the max of. -/// @param [in] y The second value to compute the max of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxUInt16x2 ffxMax(FfxUInt16x2 x, FfxUInt16x2 y) -{ - return max(x, y); -} - -/// Compute the max of two values. -/// -/// @param [in] x The first value to compute the max of. -/// @param [in] y The second value to compute the max of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxUInt16x3 ffxMax(FfxUInt16x3 x, FfxUInt16x3 y) -{ - return max(x, y); -} - -/// Compute the max of two values. -/// -/// @param [in] x The first value to compute the max of. -/// @param [in] y The second value to compute the max of. -/// -/// @returns -/// The the lowest of two values. -/// -/// @ingroup GPU -FfxUInt16x4 ffxMax(FfxUInt16x4 x, FfxUInt16x4 y) -{ - return max(x, y); -} - -/// Compute the value of the first parameter raised to the power of the second. -/// -/// @param [in] x The value to raise to the power y. -/// @param [in] y The power to which to raise x. -/// -/// @returns -/// The value of the first parameter raised to the power of the second. -/// -/// @ingroup GPU -FfxFloat16 ffxPow(FfxFloat16 x, FfxFloat16 y) -{ - return pow(x, y); -} - -/// Compute the value of the first parameter raised to the power of the second. -/// -/// @param [in] x The value to raise to the power y. -/// @param [in] y The power to which to raise x. -/// -/// @returns -/// The value of the first parameter raised to the power of the second. -/// -/// @ingroup GPU -FfxFloat16x2 ffxPow(FfxFloat16x2 x, FfxFloat16x2 y) -{ - return pow(x, y); -} - -/// Compute the value of the first parameter raised to the power of the second. -/// -/// @param [in] x The value to raise to the power y. -/// @param [in] y The power to which to raise x. -/// -/// @returns -/// The value of the first parameter raised to the power of the second. -/// -/// @ingroup GPU -FfxFloat16x3 ffxPow(FfxFloat16x3 x, FfxFloat16x3 y) -{ - return pow(x, y); -} - -/// Compute the value of the first parameter raised to the power of the second. -/// -/// @param [in] x The value to raise to the power y. -/// @param [in] y The power to which to raise x. -/// -/// @returns -/// The value of the first parameter raised to the power of the second. -/// -/// @ingroup GPU -FfxFloat16x4 ffxPow(FfxFloat16x4 x, FfxFloat16x4 y) -{ - return pow(x, y); -} - -/// Compute the square root of a value. -/// -/// @param [in] x The first value to compute the min of. -/// -/// @returns -/// The the square root of x. -/// -/// @ingroup GPU -FfxFloat16 ffxSqrt(FfxFloat16 x) -{ - return sqrt(x); -} - -/// Compute the square root of a value. -/// -/// @param [in] x The first value to compute the min of. -/// -/// @returns -/// The the square root of x. -/// -/// @ingroup GPU -FfxFloat16x2 ffxSqrt(FfxFloat16x2 x) -{ - return sqrt(x); -} - -/// Compute the square root of a value. -/// -/// @param [in] x The first value to compute the min of. -/// -/// @returns -/// The the square root of x. -/// -/// @ingroup GPU -FfxFloat16x3 ffxSqrt(FfxFloat16x3 x) -{ - return sqrt(x); -} - -/// Compute the square root of a value. -/// -/// @param [in] x The first value to compute the min of. -/// -/// @returns -/// The the square root of x. -/// -/// @ingroup GPU -FfxFloat16x4 ffxSqrt(FfxFloat16x4 x) -{ - return sqrt(x); -} - -/// Copy the sign bit from 's' to positive 'd'. -/// -/// @param [in] d The value to copy the sign bit into. -/// @param [in] s The value to copy the sign bit from. -/// -/// @returns -/// The value of d with the sign bit from s. -/// -/// @ingroup GPU -FfxFloat16 ffxCopySignBitHalf(FfxFloat16 d, FfxFloat16 s) -{ - return FFX_TO_FLOAT16(FFX_TO_UINT16(d) | (FFX_TO_UINT16(s) & FFX_BROADCAST_UINT16(0x8000u))); -} - -/// Copy the sign bit from 's' to positive 'd'. -/// -/// @param [in] d The value to copy the sign bit into. -/// @param [in] s The value to copy the sign bit from. -/// -/// @returns -/// The value of d with the sign bit from s. -/// -/// @ingroup GPU -FfxFloat16x2 ffxCopySignBitHalf(FfxFloat16x2 d, FfxFloat16x2 s) -{ - return FFX_TO_FLOAT16X2(FFX_TO_UINT16X2(d) | (FFX_TO_UINT16X2(s) & FFX_BROADCAST_UINT16X2(0x8000u))); -} - -/// Copy the sign bit from 's' to positive 'd'. -/// -/// @param [in] d The value to copy the sign bit into. -/// @param [in] s The value to copy the sign bit from. -/// -/// @returns -/// The value of d with the sign bit from s. -/// -/// @ingroup GPU -FfxFloat16x3 ffxCopySignBitHalf(FfxFloat16x3 d, FfxFloat16x3 s) -{ - return FFX_TO_FLOAT16X3(FFX_TO_UINT16X3(d) | (FFX_TO_UINT16X3(s) & FFX_BROADCAST_UINT16X3(0x8000u))); -} - -/// Copy the sign bit from 's' to positive 'd'. -/// -/// @param [in] d The value to copy the sign bit into. -/// @param [in] s The value to copy the sign bit from. -/// -/// @returns -/// The value of d with the sign bit from s. -/// -/// @ingroup GPU -FfxFloat16x4 ffxCopySignBitHalf(FfxFloat16x4 d, FfxFloat16x4 s) -{ - return FFX_TO_FLOAT16X4(FFX_TO_UINT16X4(d) | (FFX_TO_UINT16X4(s) & FFX_BROADCAST_UINT16X4(0x8000u))); -} - -/// A single operation to return the following: -/// m = NaN := 0 -/// m >= 0 := 0 -/// m < 0 := 1 -/// -/// Uses the following useful floating point logic, -/// saturate(+a*(-INF)==-INF) := 0 -/// saturate( 0*(-INF)== NaN) := 0 -/// saturate(-a*(-INF)==+INF) := 1 -/// -/// This function is useful when creating masks for branch-free logic. -/// -/// @param [in] m The value to test against 0. -/// -/// @returns -/// 1.0 when the value is negative, or 0.0 when the value is 0 or position. -/// -/// @ingroup GPU -FfxFloat16 ffxIsSignedHalf(FfxFloat16 m) -{ - return ffxSaturate(m * FFX_BROADCAST_FLOAT16(FFX_NEGATIVE_INFINITY_HALF)); -} - -/// A single operation to return the following: -/// m = NaN := 0 -/// m >= 0 := 0 -/// m < 0 := 1 -/// -/// Uses the following useful floating point logic, -/// saturate(+a*(-INF)==-INF) := 0 -/// saturate( 0*(-INF)== NaN) := 0 -/// saturate(-a*(-INF)==+INF) := 1 -/// -/// This function is useful when creating masks for branch-free logic. -/// -/// @param [in] m The value to test against 0. -/// -/// @returns -/// 1.0 when the value is negative, or 0.0 when the value is 0 or position. -/// -/// @ingroup GPU -FfxFloat16x2 ffxIsSignedHalf(FfxFloat16x2 m) -{ - return ffxSaturate(m * FFX_BROADCAST_FLOAT16X2(FFX_NEGATIVE_INFINITY_HALF)); -} - -/// A single operation to return the following: -/// m = NaN := 0 -/// m >= 0 := 0 -/// m < 0 := 1 -/// -/// Uses the following useful floating point logic, -/// saturate(+a*(-INF)==-INF) := 0 -/// saturate( 0*(-INF)== NaN) := 0 -/// saturate(-a*(-INF)==+INF) := 1 -/// -/// This function is useful when creating masks for branch-free logic. -/// -/// @param [in] m The value to test against 0. -/// -/// @returns -/// 1.0 when the value is negative, or 0.0 when the value is 0 or position. -/// -/// @ingroup GPU -FfxFloat16x3 ffxIsSignedHalf(FfxFloat16x3 m) -{ - return ffxSaturate(m * FFX_BROADCAST_FLOAT16X3(FFX_NEGATIVE_INFINITY_HALF)); -} - -/// A single operation to return the following: -/// m = NaN := 0 -/// m >= 0 := 0 -/// m < 0 := 1 -/// -/// Uses the following useful floating point logic, -/// saturate(+a*(-INF)==-INF) := 0 -/// saturate( 0*(-INF)== NaN) := 0 -/// saturate(-a*(-INF)==+INF) := 1 -/// -/// This function is useful when creating masks for branch-free logic. -/// -/// @param [in] m The value to test against 0. -/// -/// @returns -/// 1.0 when the value is negative, or 0.0 when the value is 0 or position. -/// -/// @ingroup GPU -FfxFloat16x4 ffxIsSignedHalf(FfxFloat16x4 m) -{ - return ffxSaturate(m * FFX_BROADCAST_FLOAT16X4(FFX_NEGATIVE_INFINITY_HALF)); -} - -/// A single operation to return the following: -/// m = NaN := 1 -/// m > 0 := 0 -/// m <= 0 := 1 -/// -/// This function is useful when creating masks for branch-free logic. -/// -/// @param [in] m The value to test against zero. -/// -/// @returns -/// 1.0 when the value is position, or 0.0 when the value is 0 or negative. -/// -/// @ingroup GPU -FfxFloat16 ffxIsGreaterThanZeroHalf(FfxFloat16 m) -{ - return ffxSaturate(m * FFX_BROADCAST_FLOAT16(FFX_POSITIVE_INFINITY_HALF)); -} - -/// A single operation to return the following: -/// m = NaN := 1 -/// m > 0 := 0 -/// m <= 0 := 1 -/// -/// This function is useful when creating masks for branch-free logic. -/// -/// @param [in] m The value to test against zero. -/// -/// @returns -/// 1.0 when the value is position, or 0.0 when the value is 0 or negative. -/// -/// @ingroup GPU -FfxFloat16x2 ffxIsGreaterThanZeroHalf(FfxFloat16x2 m) -{ - return ffxSaturate(m * FFX_BROADCAST_FLOAT16X2(FFX_POSITIVE_INFINITY_HALF)); -} - -/// A single operation to return the following: -/// m = NaN := 1 -/// m > 0 := 0 -/// m <= 0 := 1 -/// -/// This function is useful when creating masks for branch-free logic. -/// -/// @param [in] m The value to test against zero. -/// -/// @returns -/// 1.0 when the value is position, or 0.0 when the value is 0 or negative. -/// -/// @ingroup GPU -FfxFloat16x3 ffxIsGreaterThanZeroHalf(FfxFloat16x3 m) -{ - return ffxSaturate(m * FFX_BROADCAST_FLOAT16X3(FFX_POSITIVE_INFINITY_HALF)); -} - -/// A single operation to return the following: -/// m = NaN := 1 -/// m > 0 := 0 -/// m <= 0 := 1 -/// -/// This function is useful when creating masks for branch-free logic. -/// -/// @param [in] m The value to test against zero. -/// -/// @returns -/// 1.0 when the value is position, or 0.0 when the value is 0 or negative. -/// -/// @ingroup GPU -FfxFloat16x4 ffxIsGreaterThanZeroHalf(FfxFloat16x4 m) -{ - return ffxSaturate(m * FFX_BROADCAST_FLOAT16X4(FFX_POSITIVE_INFINITY_HALF)); -} - -/// Convert a 16bit floating point value to sortable integer. -/// -/// - If sign bit=0, flip the sign bit (positives). -/// - If sign bit=1, flip all bits (negatives). -/// -/// The function has the side effects that: -/// - Larger integers are more positive values. -/// - Float zero is mapped to center of integers (so clear to integer zero is a nice default for atomic max usage). -/// -/// @param [in] x The floating point value to make sortable. -/// -/// @returns -/// The sortable integer value. -/// -/// @ingroup GPU -FfxUInt16 ffxFloatToSortableIntegerHalf(FfxUInt16 x) -{ - return x ^ ((ffxBitShiftRightHalf(x, FFX_BROADCAST_UINT16(15))) | FFX_BROADCAST_UINT16(0x8000)); -} - -/// Convert a sortable integer to a 16bit floating point value. -/// -/// The function has the side effects that: -/// - If sign bit=1, flip the sign bit (positives). -/// - If sign bit=0, flip all bits (negatives). -/// -/// @param [in] x The sortable integer value to make floating point. -/// -/// @returns -/// The floating point value. -/// -/// @ingroup GPU -FfxUInt16 ffxSortableIntegerToFloatHalf(FfxUInt16 x) -{ - return x ^ ((~ffxBitShiftRightHalf(x, FFX_BROADCAST_UINT16(15))) | FFX_BROADCAST_UINT16(0x8000)); -} - -/// Convert a pair of 16bit floating point values to a pair of sortable integers. -/// -/// - If sign bit=0, flip the sign bit (positives). -/// - If sign bit=1, flip all bits (negatives). -/// -/// The function has the side effects that: -/// - Larger integers are more positive values. -/// - Float zero is mapped to center of integers (so clear to integer zero is a nice default for atomic max usage). -/// -/// @param [in] x The floating point values to make sortable. -/// -/// @returns -/// The sortable integer values. -/// -/// @ingroup GPU -FfxUInt16x2 ffxFloatToSortableIntegerHalf(FfxUInt16x2 x) -{ - return x ^ ((ffxBitShiftRightHalf(x, FFX_BROADCAST_UINT16X2(15))) | FFX_BROADCAST_UINT16X2(0x8000)); -} - -/// Convert a pair of sortable integers to a pair of 16bit floating point values. -/// -/// The function has the side effects that: -/// - If sign bit=1, flip the sign bit (positives). -/// - If sign bit=0, flip all bits (negatives). -/// -/// @param [in] x The sortable integer values to make floating point. -/// -/// @returns -/// The floating point values. -/// -/// @ingroup GPU -FfxUInt16x2 ffxSortableIntegerToFloatHalf(FfxUInt16x2 x) -{ - return x ^ ((~ffxBitShiftRightHalf(x, FFX_BROADCAST_UINT16X2(15))) | FFX_BROADCAST_UINT16X2(0x8000)); -} - -/// Packs the bytes from the X and Y components of a FfxUInt32x2 into a single 32-bit integer. -/// -/// The resulting integer will contain bytes in the following order, from most to least significant: -/// [Zero] Y0 [Zero] X0 -/// -/// @param [in] i The integer pair to pack. -/// -/// @returns -/// The packed integer value. -/// -/// @ingroup GPU -FfxUInt32 ffxPackBytesZeroY0ZeroX0(FfxUInt32x2 i) -{ - return ((i.x) & 0xffu) | ((i.y << 16) & 0xff0000u); -} - -/// Packs the bytes from the X and Y components of a FfxUInt32x2 into a single 32-bit integer. -/// -/// The resulting integer will contain bytes in the following order, from most to least significant: -/// [Zero] Y1 [Zero] X1 -/// -/// @param [in] i The integer pair to pack. -/// -/// @returns -/// The packed integer value. -/// -/// @ingroup GPU -FfxUInt32 ffxPackBytesZeroY1ZeroX1(FfxUInt32x2 i) -{ - return ((i.x >> 8) & 0xffu) | ((i.y << 8) & 0xff0000u); -} - -/// Packs the bytes from the X and Y components of a FfxUInt32x2 into a single 32-bit integer. -/// -/// The resulting integer will contain bytes in the following order, from most to least significant: -/// [Zero] Y2 [Zero] X2 -/// -/// @param [in] i The integer pair to pack. -/// -/// @returns -/// The packed integer value. -/// -/// @ingroup GPU -FfxUInt32 ffxPackBytesZeroY2ZeroX2(FfxUInt32x2 i) -{ - return ((i.x >> 16) & 0xffu) | ((i.y) & 0xff0000u); -} - -/// Packs the bytes from the X and Y components of a FfxUInt32x2 into a single 32-bit integer. -/// -/// The resulting integer will contain bytes in the following order, from most to least significant: -/// [Zero] Y3 [Zero] X3 -/// -/// @param [in] i The integer pair to pack. -/// -/// @returns -/// The packed integer value. -/// -/// @ingroup GPU -FfxUInt32 ffxPackBytesZeroY3ZeroX3(FfxUInt32x2 i) -{ - return ((i.x >> 24) & 0xffu) | ((i.y >> 8) & 0xff0000u); -} - -/// Packs the bytes from the X and Y components of a FfxUInt32x2 into a single 32-bit integer. -/// -/// The resulting integer will contain bytes in the following order, from most to least significant: -/// Y3 Y2 Y1 X0 -/// -/// @param [in] i The integer pair to pack. -/// -/// @returns -/// The packed integer value. -/// -/// @ingroup GPU -FfxUInt32 ffxPackBytesY3Y2Y1X0(FfxUInt32x2 i) -{ - return ((i.x) & 0x000000ffu) | (i.y & 0xffffff00u); -} - -/// Packs the bytes from the X and Y components of a FfxUInt32x2 into a single 32-bit integer. -/// -/// The resulting integer will contain bytes in the following order, from most to least significant: -/// Y3 Y2 Y1 X2 -/// -/// @param [in] i The integer pair to pack. -/// -/// @returns -/// The packed integer value. -/// -/// @ingroup GPU -FfxUInt32 ffxPackBytesY3Y2Y1X2(FfxUInt32x2 i) -{ - return ((i.x >> 16) & 0x000000ffu) | (i.y & 0xffffff00u); -} - -/// Packs the bytes from the X and Y components of a FfxUInt32x2 into a single 32-bit integer. -/// -/// The resulting integer will contain bytes in the following order, from most to least significant: -/// Y3 Y2 X0 Y0 -/// -/// @param [in] i The integer pair to pack. -/// -/// @returns -/// The packed integer value. -/// -/// @ingroup GPU -FfxUInt32 ffxPackBytesY3Y2X0Y0(FfxUInt32x2 i) -{ - return ((i.x << 8) & 0x0000ff00u) | (i.y & 0xffff00ffu); -} - -/// Packs the bytes from the X and Y components of a FfxUInt32x2 into a single 32-bit integer. -/// -/// The resulting integer will contain bytes in the following order, from most to least significant: -/// Y3 Y2 X2 Y0 -/// -/// @param [in] i The integer pair to pack. -/// -/// @returns -/// The packed integer value. -/// -/// @ingroup GPU -FfxUInt32 ffxPackBytesY3Y2X2Y0(FfxUInt32x2 i) -{ - return ((i.x >> 8) & 0x0000ff00u) | (i.y & 0xffff00ffu); -} - -/// Packs the bytes from the X and Y components of a FfxUInt32x2 into a single 32-bit integer. -/// -/// The resulting integer will contain bytes in the following order, from most to least significant: -/// Y3 X0 Y1 Y0 -/// -/// @param [in] i The integer pair to pack. -/// -/// @returns -/// The packed integer value. -/// -/// @ingroup GPU -FfxUInt32 ffxPackBytesY3X0Y1Y0(FfxUInt32x2 i) -{ - return ((i.x << 16) & 0x00ff0000u) | (i.y & 0xff00ffffu); -} - -/// Packs the bytes from the X and Y components of a FfxUInt32x2 into a single 32-bit integer. -/// -/// The resulting integer will contain bytes in the following order, from most to least significant: -/// Y3 X2 Y1 Y0 -/// -/// @param [in] i The integer pair to pack. -/// -/// @returns -/// The packed integer value. -/// -/// @ingroup GPU -FfxUInt32 ffxPackBytesY3X2Y1Y0(FfxUInt32x2 i) -{ - return ((i.x) & 0x00ff0000u) | (i.y & 0xff00ffffu); -} - -/// Packs the bytes from the X and Y components of a FfxUInt32x2 into a single 32-bit integer. -/// -/// The resulting integer will contain bytes in the following order, from most to least significant: -/// X0 Y2 Y1 Y0 -/// -/// @param [in] i The integer pair to pack. -/// -/// @returns -/// The packed integer value. -/// -/// @ingroup GPU -FfxUInt32 ffxPackBytesX0Y2Y1Y0(FfxUInt32x2 i) -{ - return ((i.x << 24) & 0xff000000u) | (i.y & 0x00ffffffu); -} - -/// Packs the bytes from the X and Y components of a FfxUInt32x2 into a single 32-bit integer. -/// -/// The resulting integer will contain bytes in the following order, from most to least significant: -/// X2 Y2 Y1 Y0 -/// -/// @param [in] i The integer pair to pack. -/// -/// @returns -/// The packed integer value. -/// -/// @ingroup GPU -FfxUInt32 ffxPackBytesX2Y2Y1Y0(FfxUInt32x2 i) -{ - return ((i.x << 8) & 0xff000000u) | (i.y & 0x00ffffffu); -} - -/// Packs the bytes from the X and Y components of a FfxUInt32x2 into a single 32-bit integer. -/// -/// The resulting integer will contain bytes in the following order, from most to least significant: -/// Y2 X2 Y0 X0 -/// -/// @param [in] i The integer pair to pack. -/// -/// @returns -/// The packed integer value. -/// -/// @ingroup GPU -FfxUInt32 ffxPackBytesY2X2Y0X0(FfxUInt32x2 i) -{ - return ((i.x) & 0x00ff00ffu) | ((i.y << 8) & 0xff00ff00u); -} - -/// Packs the bytes from the X and Y components of a FfxUInt32x2 into a single 32-bit integer. -/// -/// The resulting integer will contain bytes in the following order, from most to least significant: -/// Y2 Y0 X2 X0 -/// -/// @param [in] i The integer pair to pack. -/// -/// @returns -/// The packed integer value. -/// -/// @ingroup GPU -FfxUInt32 ffxPackBytesY2Y0X2X0(FfxUInt32x2 i) -{ - return (((i.x) & 0xffu) | ((i.x >> 8) & 0xff00u) | ((i.y << 16) & 0xff0000u) | ((i.y << 8) & 0xff000000u)); -} - -/// Takes two Float16x2 values x and y, normalizes them and builds a single Uint16x2 value in the format {{x0,y0},{x1,y1}}. -/// -/// @param [in] x The first float16x2 value to pack. -/// @param [in] y The second float16x2 value to pack. -/// -/// @returns -/// The packed FfxUInt32x2 value. -/// -/// @ingroup GPU -FfxUInt16x2 ffxPackX0Y0X1Y1UnsignedToUint16x2(FfxFloat16x2 x, FfxFloat16x2 y) -{ - x *= FFX_BROADCAST_FLOAT16X2(1.0 / 32768.0); - y *= FFX_BROADCAST_FLOAT16X2(1.0 / 32768.0); - return FFX_UINT32_TO_UINT16X2(ffxPackBytesY2X2Y0X0(FfxUInt32x2(FFX_UINT16X2_TO_UINT32(FFX_TO_UINT16X2(x)), FFX_UINT16X2_TO_UINT32(FFX_TO_UINT16X2(y))))); -} - -/// Given a FfxUInt32x2 value d, Float16x2 value i and a resulting FfxUInt32x2 value r, this function packs d.x[0:7] into r.x[0:7], -/// d.y[0:7] into r.y[0:7], i.x[8:15] into r.x[8:15], r.y[8:15] and i.y[0:15] into r.x[16:31], r.y[16:31] using 3 ops. -/// -/// r=ffxPermuteUByte0Float16x2ToUint2(d,i) -/// Where 'k0' is an SGPR with {1.0/32768.0} packed into the lower 16-bits -/// Where 'k1' is an SGPR with 0x???? -/// Where 'k2' is an SGPR with 0x???? -/// V_PK_FMA_F16 i,i,k0.x,0 -/// V_PERM_B32 r.x,i,i,k1 -/// V_PERM_B32 r.y,i,i,k2 -/// -/// @param [in] d The FfxUInt32x2 value to be packed. -/// @param [in] i The FfxFloat16x2 value to be packed. -/// -/// @returns -/// The packed FfxUInt32x2 value. -/// -/// @ingroup GPU -FfxUInt32x2 ffxPermuteUByte0Float16x2ToUint2(FfxUInt32x2 d, FfxFloat16x2 i) -{ - FfxUInt32 b = FFX_UINT16X2_TO_UINT32(FFX_TO_UINT16X2(i * FFX_BROADCAST_FLOAT16X2(1.0 / 32768.0))); - return FfxUInt32x2(ffxPackBytesY3Y2Y1X0(FfxUInt32x2(d.x, b)), ffxPackBytesY3Y2Y1X2(FfxUInt32x2(d.y, b))); -} - -/// Given a FfxUInt32x2 value d, Float16x2 value i and a resulting FfxUInt32x2 value r, this function packs d.x[0:7] into r.x[8:15], -/// d.y[0:7] into r.y[8:15], i.x[0:7] into r.x[0:7], r.y[0:7] and i.y[0:15] into r.x[16:31], r.y[16:31] using 3 ops. -/// -/// r=ffxPermuteUByte1Float16x2ToUint2(d,i) -/// Where 'k0' is an SGPR with {1.0/32768.0} packed into the lower 16-bits -/// Where 'k1' is an SGPR with 0x???? -/// Where 'k2' is an SGPR with 0x???? -/// V_PK_FMA_F16 i,i,k0.x,0 -/// V_PERM_B32 r.x,i,i,k1 -/// V_PERM_B32 r.y,i,i,k2 -/// -/// @param [in] d The FfxUInt32x2 value to be packed. -/// @param [in] i The FfxFloat16x2 value to be packed. -/// -/// @returns -/// The packed FfxUInt32x2 value. -/// -/// @ingroup GPU -FfxUInt32x2 ffxPermuteUByte1Float16x2ToUint2(FfxUInt32x2 d, FfxFloat16x2 i) -{ - FfxUInt32 b = FFX_UINT16X2_TO_UINT32(FFX_TO_UINT16X2(i * FFX_BROADCAST_FLOAT16X2(1.0 / 32768.0))); - return FfxUInt32x2(ffxPackBytesY3Y2X0Y0(FfxUInt32x2(d.x, b)), ffxPackBytesY3Y2X2Y0(FfxUInt32x2(d.y, b))); -} - -/// Given a FfxUInt32x2 value d, Float16x2 value i and a resulting FfxUInt32x2 value r, this function packs d.x[0:7] into r.x[16:23], -/// d.y[0:7] into r.y[16:23], i.x[0:15] into r.x[0:15], r.y[0:15] and i.y[8:15] into r.x[24:31], r.y[24:31] using 3 ops. -/// -/// r=ffxPermuteUByte2Float16x2ToUint2(d,i) -/// Where 'k0' is an SGPR with {1.0/32768.0} packed into the lower 16-bits -/// Where 'k1' is an SGPR with 0x???? -/// Where 'k2' is an SGPR with 0x???? -/// V_PK_FMA_F16 i,i,k0.x,0 -/// V_PERM_B32 r.x,i,i,k1 -/// V_PERM_B32 r.y,i,i,k2 -/// -/// @param [in] d The FfxUInt32x2 value to be packed. -/// @param [in] i The FfxFloat16x2 value to be packed. -/// -/// @returns -/// The packed FfxUInt32x2 value. -/// -/// @ingroup GPU -FfxUInt32x2 ffxPermuteUByte2Float16x2ToUint2(FfxUInt32x2 d, FfxFloat16x2 i) -{ - FfxUInt32 b = FFX_UINT16X2_TO_UINT32(FFX_TO_UINT16X2(i * FFX_BROADCAST_FLOAT16X2(1.0 / 32768.0))); - return FfxUInt32x2(ffxPackBytesY3X0Y1Y0(FfxUInt32x2(d.x, b)), ffxPackBytesY3X2Y1Y0(FfxUInt32x2(d.y, b))); -} - -/// Given a FfxUInt32x2 value d, Float16x2 value i and a resulting FfxUInt32x2 value r, this function packs d.x[0:7] into r.x[24:31], -/// d.y[0:7] into r.y[24:31], i.x[0:15] into r.x[0:15], r.y[0:15] and i.y[0:7] into r.x[16:23], r.y[16:23] using 3 ops. -/// -/// r=ffxPermuteUByte3Float16x2ToUint2(d,i) -/// Where 'k0' is an SGPR with {1.0/32768.0} packed into the lower 16-bits -/// Where 'k1' is an SGPR with 0x???? -/// Where 'k2' is an SGPR with 0x???? -/// V_PK_FMA_F16 i,i,k0.x,0 -/// V_PERM_B32 r.x,i,i,k1 -/// V_PERM_B32 r.y,i,i,k2 -/// -/// @param [in] d The FfxUInt32x2 value to be packed. -/// @param [in] i The FfxFloat16x2 value to be packed. -/// -/// @returns -/// The packed FfxUInt32x2 value. -/// -/// @ingroup GPU -FfxUInt32x2 ffxPermuteUByte3Float16x2ToUint2(FfxUInt32x2 d, FfxFloat16x2 i) -{ - FfxUInt32 b = FFX_UINT16X2_TO_UINT32(FFX_TO_UINT16X2(i * FFX_BROADCAST_FLOAT16X2(1.0 / 32768.0))); - return FfxUInt32x2(ffxPackBytesX0Y2Y1Y0(FfxUInt32x2(d.x, b)), ffxPackBytesX2Y2Y1Y0(FfxUInt32x2(d.y, b))); -} - -/// Given a FfxUInt32x2 value i and a resulting Float16x2 value r, this function packs i.x[0:7] into r.x[0:7] and i.y[0:7] into r.y[0:7] using 2 ops. -/// -/// @param [in] i The FfxUInt32x2 value to be unpacked. -/// -/// @returns -/// The unpacked FfxFloat16x2. -/// -/// @ingroup GPU -FfxFloat16x2 ffxPermuteUByte0Uint2ToFloat16x2(FfxUInt32x2 i) -{ - return FFX_TO_FLOAT16X2(FFX_UINT32_TO_UINT16X2(ffxPackBytesZeroY0ZeroX0(i))) * FFX_BROADCAST_FLOAT16X2(32768.0); -} - -/// Given a FfxUInt32x2 value i and a resulting Float16x2 value r, this function packs i.x[8:15] into r.x[0:7] and i.y[8:15] into r.y[0:7] using 2 ops. -/// -/// @param [in] i The FfxUInt32x2 value to be unpacked. -/// -/// @returns -/// The unpacked FfxFloat16x2. -/// -/// @ingroup GPU -FfxFloat16x2 ffxPermuteUByte1Uint2ToFloat16x2(FfxUInt32x2 i) -{ - return FFX_TO_FLOAT16X2(FFX_UINT32_TO_UINT16X2(ffxPackBytesZeroY1ZeroX1(i))) * FFX_BROADCAST_FLOAT16X2(32768.0); -} - -/// Given a FfxUInt32x2 value i and a resulting Float16x2 value r, this function packs i.x[16:23] into r.x[0:7] and i.y[16:23] into r.y[0:7] using 2 ops. -/// -/// @param [in] i The FfxUInt32x2 value to be unpacked. -/// -/// @returns -/// The unpacked FfxFloat16x2. -/// -/// @ingroup GPU -FfxFloat16x2 ffxPermuteUByte2Uint2ToFloat16x2(FfxUInt32x2 i) -{ - return FFX_TO_FLOAT16X2(FFX_UINT32_TO_UINT16X2(ffxPackBytesZeroY2ZeroX2(i))) * FFX_BROADCAST_FLOAT16X2(32768.0); -} - -/// Given a FfxUInt32x2 value i and a resulting Float16x2 value r, this function packs i.x[24:31] into r.x[0:7] and i.y[24:31] into r.y[0:7] using 2 ops. -/// -/// @param [in] i The FfxUInt32x2 value to be unpacked. -/// -/// @returns -/// The unpacked FfxFloat16x2. -/// -/// @ingroup GPU -FfxFloat16x2 ffxPermuteUByte3Uint2ToFloat16x2(FfxUInt32x2 i) -{ - return FFX_TO_FLOAT16X2(FFX_UINT32_TO_UINT16X2(ffxPackBytesZeroY3ZeroX3(i))) * FFX_BROADCAST_FLOAT16X2(32768.0); -} - -/// Takes two Float16x2 values x and y, normalizes them and builds a single Uint16x2 value in the format {{x0,y0},{x1,y1}}. -/// -/// @param [in] x The first float16x2 value to pack. -/// @param [in] y The second float16x2 value to pack. -/// -/// @returns -/// The packed FfxUInt32x2 value. -/// -/// @ingroup GPU -FfxUInt16x2 ffxPackX0Y0X1Y1SignedToUint16x2(FfxFloat16x2 x, FfxFloat16x2 y) -{ - x = x * FFX_BROADCAST_FLOAT16X2(1.0 / 32768.0) + FFX_BROADCAST_FLOAT16X2(0.25 / 32768.0); - y = y * FFX_BROADCAST_FLOAT16X2(1.0 / 32768.0) + FFX_BROADCAST_FLOAT16X2(0.25 / 32768.0); - return FFX_UINT32_TO_UINT16X2(ffxPackBytesY2X2Y0X0(FfxUInt32x2(FFX_UINT16X2_TO_UINT32(FFX_TO_UINT16X2(x)), FFX_UINT16X2_TO_UINT32(FFX_TO_UINT16X2(y))))); -} - -/// Given a FfxUInt32x2 value d, Float16x2 value i and a resulting FfxUInt32x2 value r, this function packs d.x[0:7] into r.x[0:7], -/// d.y[0:7] into r.y[0:7], i.x[8:15] into r.x[8:15], r.y[8:15] and i.y[0:15] into r.x[16:31], r.y[16:31] using 3 ops. -/// -/// Handles signed byte values. -/// -/// @param [in] d The FfxUInt32x2 value to be packed. -/// @param [in] i The FfxFloat16x2 value to be packed. -/// -/// @returns -/// The packed FfxUInt32x2 value. -/// -/// @ingroup GPU -FfxUInt32x2 ffxPermuteSByte0Float16x2ToUint2(FfxUInt32x2 d, FfxFloat16x2 i) -{ - FfxUInt32 b = FFX_UINT16X2_TO_UINT32(FFX_TO_UINT16X2(i * FFX_BROADCAST_FLOAT16X2(1.0 / 32768.0) + FFX_BROADCAST_FLOAT16X2(0.25 / 32768.0))); - return FfxUInt32x2(ffxPackBytesY3Y2Y1X0(FfxUInt32x2(d.x, b)), ffxPackBytesY3Y2Y1X2(FfxUInt32x2(d.y, b))); -} - -/// Given a FfxUInt32x2 value d, Float16x2 value i and a resulting FfxUInt32x2 value r, this function packs d.x[0:7] into r.x[8:15], -/// d.y[0:7] into r.y[8:15], i.x[0:7] into r.x[0:7], r.y[0:7] and i.y[0:15] into r.x[16:31], r.y[16:31] using 3 ops. -/// -/// Handles signed byte values. -/// -/// @param [in] d The FfxUInt32x2 value to be packed. -/// @param [in] i The FfxFloat16x2 value to be packed. -/// -/// @returns -/// The packed FfxUInt32x2 value. -/// -/// @ingroup GPU -FfxUInt32x2 ffxPermuteSByte1Float16x2ToUint2(FfxUInt32x2 d, FfxFloat16x2 i) -{ - FfxUInt32 b = FFX_UINT16X2_TO_UINT32(FFX_TO_UINT16X2(i * FFX_BROADCAST_FLOAT16X2(1.0 / 32768.0) + FFX_BROADCAST_FLOAT16X2(0.25 / 32768.0))); - return FfxUInt32x2(ffxPackBytesY3Y2X0Y0(FfxUInt32x2(d.x, b)), ffxPackBytesY3Y2X2Y0(FfxUInt32x2(d.y, b))); -} - -/// Given a FfxUInt32x2 value d, Float16x2 value i and a resulting FfxUInt32x2 value r, this function packs d.x[0:7] into r.x[16:23], -/// d.y[0:7] into r.y[16:23], i.x[0:15] into r.x[0:15], r.y[0:15] and i.y[8:15] into r.x[24:31], r.y[24:31] using 3 ops. -/// -/// Handles signed byte values. -/// -/// @param [in] d The FfxUInt32x2 value to be packed. -/// @param [in] i The FfxFloat16x2 value to be packed. -/// -/// @returns -/// The packed FfxUInt32x2 value. -/// -/// @ingroup GPU -FfxUInt32x2 ffxPermuteSByte2Float16x2ToUint2(FfxUInt32x2 d, FfxFloat16x2 i) -{ - FfxUInt32 b = FFX_UINT16X2_TO_UINT32(FFX_TO_UINT16X2(i * FFX_BROADCAST_FLOAT16X2(1.0 / 32768.0) + FFX_BROADCAST_FLOAT16X2(0.25 / 32768.0))); - return FfxUInt32x2(ffxPackBytesY3X0Y1Y0(FfxUInt32x2(d.x, b)), ffxPackBytesY3X2Y1Y0(FfxUInt32x2(d.y, b))); -} - -/// Given a FfxUInt32x2 value d, Float16x2 value i and a resulting FfxUInt32x2 value r, this function packs d.x[0:7] into r.x[24:31], -/// d.y[0:7] into r.y[24:31], i.x[0:15] into r.x[0:15], r.y[0:15] and i.y[0:7] into r.x[16:23], r.y[16:23] using 3 ops. -/// -/// Handles signed byte values. -/// -/// @param [in] d The FfxUInt32x2 value to be packed. -/// @param [in] i The FfxFloat16x2 value to be packed. -/// -/// @returns -/// The packed FfxUInt32x2 value. -/// -/// @ingroup GPU -FfxUInt32x2 ffxPermuteSByte3Float16x2ToUint2(FfxUInt32x2 d, FfxFloat16x2 i) -{ - FfxUInt32 b = FFX_UINT16X2_TO_UINT32(FFX_TO_UINT16X2(i * FFX_BROADCAST_FLOAT16X2(1.0 / 32768.0) + FFX_BROADCAST_FLOAT16X2(0.25 / 32768.0))); - return FfxUInt32x2(ffxPackBytesX0Y2Y1Y0(FfxUInt32x2(d.x, b)), ffxPackBytesX2Y2Y1Y0(FfxUInt32x2(d.y, b))); -} - -/// Given a FfxUInt32x2 value d, Float16x2 value i and a resulting FfxUInt32x2 value r, this function packs d.x[0:7] into r.x[0:7], -/// d.y[0:7] into r.y[0:7], i.x[8:15] into r.x[8:15], r.y[8:15] and i.y[0:15] into r.x[16:31], r.y[16:31] using 3 ops. -/// -/// Zero-based flips the MSB bit of the byte (making 128 "exact zero" actually zero). -/// This is useful if there is a desire for cleared values to decode as zero. -/// -/// Handles signed byte values. -/// -/// @param [in] d The FfxUInt32x2 value to be packed. -/// @param [in] i The FfxFloat16x2 value to be packed. -/// -/// @returns -/// The packed FfxUInt32x2 value. -/// -/// @ingroup GPU -FfxUInt32x2 ffxPermuteZeroBasedSByte0Float16x2ToUint2(FfxUInt32x2 d, FfxFloat16x2 i) -{ - FfxUInt32 b = FFX_UINT16X2_TO_UINT32(FFX_TO_UINT16X2(i * FFX_BROADCAST_FLOAT16X2(1.0 / 32768.0) + FFX_BROADCAST_FLOAT16X2(0.25 / 32768.0))) ^ 0x00800080u; - return FfxUInt32x2(ffxPackBytesY3Y2Y1X0(FfxUInt32x2(d.x, b)), ffxPackBytesY3Y2Y1X2(FfxUInt32x2(d.y, b))); -} - -/// Given a FfxUInt32x2 value d, Float16x2 value i and a resulting FfxUInt32x2 value r, this function packs d.x[0:7] into r.x[8:15], -/// d.y[0:7] into r.y[8:15], i.x[0:7] into r.x[0:7], r.y[0:7] and i.y[0:15] into r.x[16:31], r.y[16:31] using 3 ops. -/// -/// Zero-based flips the MSB bit of the byte (making 128 "exact zero" actually zero). -/// This is useful if there is a desire for cleared values to decode as zero. -/// -/// Handles signed byte values. -/// -/// @param [in] d The FfxUInt32x2 value to be packed. -/// @param [in] i The FfxFloat16x2 value to be packed. -/// -/// @returns -/// The packed FfxUInt32x2 value. -/// -/// @ingroup GPU -FfxUInt32x2 ffxPermuteZeroBasedSByte1Float16x2ToUint2(FfxUInt32x2 d, FfxFloat16x2 i) -{ - FfxUInt32 b = FFX_UINT16X2_TO_UINT32(FFX_TO_UINT16X2(i * FFX_BROADCAST_FLOAT16X2(1.0 / 32768.0) + FFX_BROADCAST_FLOAT16X2(0.25 / 32768.0))) ^ 0x00800080u; - return FfxUInt32x2(ffxPackBytesY3Y2X0Y0(FfxUInt32x2(d.x, b)), ffxPackBytesY3Y2X2Y0(FfxUInt32x2(d.y, b))); -} - -/// Given a FfxUInt32x2 value d, Float16x2 value i and a resulting FfxUInt32x2 value r, this function packs d.x[0:7] into r.x[16:23], -/// d.y[0:7] into r.y[16:23], i.x[0:15] into r.x[0:15], r.y[0:15] and i.y[8:15] into r.x[24:31], r.y[24:31] using 3 ops. -/// -/// Zero-based flips the MSB bit of the byte (making 128 "exact zero" actually zero). -/// This is useful if there is a desire for cleared values to decode as zero. -/// -/// Handles signed byte values. -/// -/// @param [in] d The FfxUInt32x2 value to be packed. -/// @param [in] i The FfxFloat16x2 value to be packed. -/// -/// @returns -/// The packed FfxUInt32x2 value. -/// -/// @ingroup GPU -FfxUInt32x2 ffxPermuteZeroBasedSByte2Float16x2ToUint2(FfxUInt32x2 d, FfxFloat16x2 i) -{ - FfxUInt32 b = FFX_UINT16X2_TO_UINT32(FFX_TO_UINT16X2(i * FFX_BROADCAST_FLOAT16X2(1.0 / 32768.0) + FFX_BROADCAST_FLOAT16X2(0.25 / 32768.0))) ^ 0x00800080u; - return FfxUInt32x2(ffxPackBytesY3X0Y1Y0(FfxUInt32x2(d.x, b)), ffxPackBytesY3X2Y1Y0(FfxUInt32x2(d.y, b))); -} - -/// Given a FfxUInt32x2 value d, Float16x2 value i and a resulting FfxUInt32x2 value r, this function packs d.x[0:7] into r.x[24:31], -/// d.y[0:7] into r.y[24:31], i.x[0:15] into r.x[0:15], r.y[0:15] and i.y[0:7] into r.x[16:23], r.y[16:23] using 3 ops. -/// -/// Zero-based flips the MSB bit of the byte (making 128 "exact zero" actually zero). -/// This is useful if there is a desire for cleared values to decode as zero. -/// -/// Handles signed byte values. -/// -/// @param [in] d The FfxUInt32x2 value to be packed. -/// @param [in] i The FfxFloat16x2 value to be packed. -/// -/// @returns -/// The packed FfxUInt32x2 value. -/// -/// @ingroup GPU -FfxUInt32x2 ffxPermuteZeroBasedSByte3Float16x2ToUint2(FfxUInt32x2 d, FfxFloat16x2 i) -{ - FfxUInt32 b = FFX_UINT16X2_TO_UINT32(FFX_TO_UINT16X2(i * FFX_BROADCAST_FLOAT16X2(1.0 / 32768.0) + FFX_BROADCAST_FLOAT16X2(0.25 / 32768.0))) ^ 0x00800080u; - return FfxUInt32x2(ffxPackBytesX0Y2Y1Y0(FfxUInt32x2(d.x, b)), ffxPackBytesX2Y2Y1Y0(FfxUInt32x2(d.y, b))); -} - -/// Given a FfxUInt32x2 value i and a resulting Float16x2 value r, this function packs i.x[0:7] into r.x[0:7] and i.y[0:7] into r.y[0:7] using 2 ops. -/// -/// Handles signed byte values. -/// -/// @param [in] i The FfxUInt32x2 value to be unpacked. -/// -/// @returns -/// The unpacked FfxFloat16x2. -/// -/// @ingroup GPU -FfxFloat16x2 ffxPermuteSByte0Uint2ToFloat16x2(FfxUInt32x2 i) -{ - return FFX_TO_FLOAT16X2(FFX_UINT32_TO_UINT16X2(ffxPackBytesZeroY0ZeroX0(i))) * FFX_BROADCAST_FLOAT16X2(32768.0) - FFX_BROADCAST_FLOAT16X2(0.25); -} - -/// Given a FfxUInt32x2 value i and a resulting Float16x2 value r, this function packs i.x[8:15] into r.x[0:7] and i.y[8:15] into r.y[0:7] using 2 ops. -/// -/// Handles signed byte values. -/// -/// @param [in] i The FfxUInt32x2 value to be unpacked. -/// -/// @returns -/// The unpacked FfxFloat16x2. -/// -/// @ingroup GPU -FfxFloat16x2 ffxPermuteSByte1Uint2ToFloat16x2(FfxUInt32x2 i) -{ - return FFX_TO_FLOAT16X2(FFX_UINT32_TO_UINT16X2(ffxPackBytesZeroY1ZeroX1(i))) * FFX_BROADCAST_FLOAT16X2(32768.0) - FFX_BROADCAST_FLOAT16X2(0.25); -} - -/// Given a FfxUInt32x2 value i and a resulting Float16x2 value r, this function packs i.x[16:23] into r.x[0:7] and i.y[16:23] into r.y[0:7] using 2 ops. -/// -/// Handles signed byte values. -/// -/// @param [in] i The FfxUInt32x2 value to be unpacked. -/// -/// @returns -/// The unpacked FfxFloat16x2. -/// -/// @ingroup GPU -FfxFloat16x2 ffxPermuteSByte2Uint2ToFloat16x2(FfxUInt32x2 i) -{ - return FFX_TO_FLOAT16X2(FFX_UINT32_TO_UINT16X2(ffxPackBytesZeroY2ZeroX2(i))) * FFX_BROADCAST_FLOAT16X2(32768.0) - FFX_BROADCAST_FLOAT16X2(0.25); -} - -/// Given a FfxUInt32x2 value i and a resulting Float16x2 value r, this function packs i.x[24:31] into r.x[0:7] and i.y[24:31] into r.y[0:7] using 2 ops. -/// -/// Handles signed byte values. -/// -/// @param [in] i The FfxUInt32x2 value to be unpacked. -/// -/// @returns -/// The unpacked FfxFloat16x2. -/// -/// @ingroup GPU -FfxFloat16x2 ffxPermuteSByte3Uint2ToFloat16x2(FfxUInt32x2 i) -{ - return FFX_TO_FLOAT16X2(FFX_UINT32_TO_UINT16X2(ffxPackBytesZeroY3ZeroX3(i))) * FFX_BROADCAST_FLOAT16X2(32768.0) - FFX_BROADCAST_FLOAT16X2(0.25); -} - -/// Given a FfxUInt32x2 value i and a resulting Float16x2 value r, this function packs i.x[0:7] into r.x[0:7] and i.y[0:7] into r.y[0:7] using 2 ops. -/// -/// Handles signed byte values. -/// -/// @param [in] i The FfxUInt32x2 value to be unpacked. -/// -/// @returns -/// The unpacked FfxFloat16x2. -/// -/// @ingroup GPU -FfxFloat16x2 ffxPermuteZeroBasedSByte0Uint2ToFloat16x2(FfxUInt32x2 i) -{ - return FFX_TO_FLOAT16X2(FFX_UINT32_TO_UINT16X2(ffxPackBytesZeroY0ZeroX0(i) ^ 0x00800080u)) * FFX_BROADCAST_FLOAT16X2(32768.0) - FFX_BROADCAST_FLOAT16X2(0.25); -} - -/// Given a FfxUInt32x2 value i and a resulting Float16x2 value r, this function packs i.x[8:15] into r.x[0:7] and i.y[8:15] into r.y[0:7] using 2 ops. -/// -/// Handles signed byte values. -/// -/// @param [in] i The FfxUInt32x2 value to be unpacked. -/// -/// @returns -/// The unpacked FfxFloat16x2. -/// -/// @ingroup GPU -FfxFloat16x2 ffxPermuteZeroBasedSByte1Uint2ToFloat16x2(FfxUInt32x2 i) -{ - return FFX_TO_FLOAT16X2(FFX_UINT32_TO_UINT16X2(ffxPackBytesZeroY1ZeroX1(i) ^ 0x00800080u)) * FFX_BROADCAST_FLOAT16X2(32768.0) - FFX_BROADCAST_FLOAT16X2(0.25); -} - -/// Given a FfxUInt32x2 value i and a resulting Float16x2 value r, this function packs i.x[16:23] into r.x[0:7] and i.y[16:23] into r.y[0:7] using 2 ops. -/// -/// Handles signed byte values. -/// -/// @param [in] i The FfxUInt32x2 value to be unpacked. -/// -/// @returns -/// The unpacked FfxFloat16x2. -/// -/// @ingroup GPU -FfxFloat16x2 ffxPermuteZeroBasedSByte2Uint2ToFloat16x2(FfxUInt32x2 i) -{ - return FFX_TO_FLOAT16X2(FFX_UINT32_TO_UINT16X2(ffxPackBytesZeroY2ZeroX2(i) ^ 0x00800080u)) * FFX_BROADCAST_FLOAT16X2(32768.0) - FFX_BROADCAST_FLOAT16X2(0.25); -} - -/// Given a FfxUInt32x2 value i and a resulting Float16x2 value r, this function packs i.x[24:31] into r.x[0:7] and i.y[24:31] into r.y[0:7] using 2 ops. -/// -/// Handles signed byte values. -/// -/// @param [in] i The FfxUInt32x2 value to be unpacked. -/// -/// @returns -/// The unpacked FfxFloat16x2. -/// -/// @ingroup GPU -FfxFloat16x2 ffxPermuteZeroBasedSByte3Uint2ToFloat16x2(FfxUInt32x2 i) -{ - return FFX_TO_FLOAT16X2(FFX_UINT32_TO_UINT16X2(ffxPackBytesZeroY3ZeroX3(i) ^ 0x00800080u)) * FFX_BROADCAST_FLOAT16X2(32768.0) - FFX_BROADCAST_FLOAT16X2(0.25); -} - -/// Calculate a half-precision low-quality approximation for the square root of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] a The value to calculate an approximate to the square root for. -/// -/// @returns -/// An approximation of the square root, estimated to low quality. -/// -/// @ingroup GPU -FfxFloat16 ffxApproximateSqrtHalf(FfxFloat16 a) -{ - return FFX_TO_FLOAT16((FFX_TO_UINT16(a) >> FFX_BROADCAST_UINT16(1)) + FFX_BROADCAST_UINT16(0x1de2)); -} - -/// Calculate a half-precision low-quality approximation for the square root of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] a The value to calculate an approximate to the square root for. -/// -/// @returns -/// An approximation of the square root, estimated to low quality. -/// -/// @ingroup GPU -FfxFloat16x2 ffxApproximateSqrtHalf(FfxFloat16x2 a) -{ - return FFX_TO_FLOAT16X2((FFX_TO_UINT16X2(a) >> FFX_BROADCAST_UINT16X2(1)) + FFX_BROADCAST_UINT16X2(0x1de2)); -} - -/// Calculate a half-precision low-quality approximation for the square root of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] a The value to calculate an approximate to the square root for. -/// -/// @returns -/// An approximation of the square root, estimated to low quality. -/// -/// @ingroup GPU -FfxFloat16x3 ffxApproximateSqrtHalf(FfxFloat16x3 a) -{ - return FFX_TO_FLOAT16X3((FFX_TO_UINT16X3(a) >> FFX_BROADCAST_UINT16X3(1)) + FFX_BROADCAST_UINT16X3(0x1de2)); -} - -/// Calculate a half-precision low-quality approximation for the reciprocal of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] a The value to calculate an approximate to the reciprocal for. -/// -/// @returns -/// An approximation of the reciprocal, estimated to low quality. -/// -/// @ingroup GPU -FfxFloat16 ffxApproximateReciprocalHalf(FfxFloat16 a) -{ - return FFX_TO_FLOAT16(FFX_BROADCAST_UINT16(0x7784) - FFX_TO_UINT16(a)); -} - -/// Calculate a half-precision low-quality approximation for the reciprocal of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] a The value to calculate an approximate to the reciprocal for. -/// -/// @returns -/// An approximation of the reciprocal, estimated to low quality. -/// -/// @ingroup GPU -FfxFloat16x2 ffxApproximateReciprocalHalf(FfxFloat16x2 a) -{ - return FFX_TO_FLOAT16X2(FFX_BROADCAST_UINT16X2(0x7784) - FFX_TO_UINT16X2(a)); -} - -/// Calculate a half-precision low-quality approximation for the reciprocal of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] a The value to calculate an approximate to the reciprocal for. -/// -/// @returns -/// An approximation of the reciprocal, estimated to low quality. -/// -/// @ingroup GPU -FfxFloat16x3 ffxApproximateReciprocalHalf(FfxFloat16x3 a) -{ - return FFX_TO_FLOAT16X3(FFX_BROADCAST_UINT16X3(0x7784) - FFX_TO_UINT16X3(a)); -} - -/// Calculate a half-precision low-quality approximation for the reciprocal of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] a The value to calculate an approximate to the reciprocal for. -/// -/// @returns -/// An approximation of the reciprocal, estimated to low quality. -/// -/// @ingroup GPU -FfxFloat16x4 ffxApproximateReciprocalHalf(FfxFloat16x4 a) -{ - return FFX_TO_FLOAT16X4(FFX_BROADCAST_UINT16X4(0x7784) - FFX_TO_UINT16X4(a)); -} - -/// Calculate a half-precision medium-quality approximation for the reciprocal of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] a The value to calculate an approximate to the reciprocal for. -/// -/// @returns -/// An approximation of the reciprocal, estimated to medium quality. -/// -/// @ingroup GPU -FfxFloat16 ffxApproximateReciprocalMediumHalf(FfxFloat16 a) -{ - FfxFloat16 b = FFX_TO_FLOAT16(FFX_BROADCAST_UINT16(0x778d) - FFX_TO_UINT16(a)); - return b * (-b * a + FFX_BROADCAST_FLOAT16(2.0)); -} - -/// Calculate a half-precision medium-quality approximation for the reciprocal of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] a The value to calculate an approximate to the reciprocal for. -/// -/// @returns -/// An approximation of the reciprocal, estimated to medium quality. -/// -/// @ingroup GPU -FfxFloat16x2 ffxApproximateReciprocalMediumHalf(FfxFloat16x2 a) -{ - FfxFloat16x2 b = FFX_TO_FLOAT16X2(FFX_BROADCAST_UINT16X2(0x778d) - FFX_TO_UINT16X2(a)); - return b * (-b * a + FFX_BROADCAST_FLOAT16X2(2.0)); -} - -/// Calculate a half-precision medium-quality approximation for the reciprocal of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] a The value to calculate an approximate to the reciprocal for. -/// -/// @returns -/// An approximation of the reciprocal, estimated to medium quality. -/// -/// @ingroup GPU -FfxFloat16x3 ffxApproximateReciprocalMediumHalf(FfxFloat16x3 a) -{ - FfxFloat16x3 b = FFX_TO_FLOAT16X3(FFX_BROADCAST_UINT16X3(0x778d) - FFX_TO_UINT16X3(a)); - return b * (-b * a + FFX_BROADCAST_FLOAT16X3(2.0)); -} - -/// Calculate a half-precision medium-quality approximation for the reciprocal of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] a The value to calculate an approximate to the reciprocal for. -/// -/// @returns -/// An approximation of the reciprocal, estimated to medium quality. -/// -/// @ingroup GPU -FfxFloat16x4 ffxApproximateReciprocalMediumHalf(FfxFloat16x4 a) -{ - FfxFloat16x4 b = FFX_TO_FLOAT16X4(FFX_BROADCAST_UINT16X4(0x778d) - FFX_TO_UINT16X4(a)); - return b * (-b * a + FFX_BROADCAST_FLOAT16X4(2.0)); -} - -/// Calculate a half-precision low-quality approximation for the reciprocal of the square root of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] a The value to calculate an approximate to the reciprocal of the square root for. -/// -/// @returns -/// An approximation of the reciprocal of the square root, estimated to low quality. -/// -/// @ingroup GPU -FfxFloat16 ffxApproximateReciprocalSquareRootHalf(FfxFloat16 a) -{ - return FFX_TO_FLOAT16(FFX_BROADCAST_UINT16(0x59a3) - (FFX_TO_UINT16(a) >> FFX_BROADCAST_UINT16(1))); -} - -/// Calculate a half-precision low-quality approximation for the reciprocal of the square root of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] a The value to calculate an approximate to the reciprocal of the square root for. -/// -/// @returns -/// An approximation of the reciprocal of the square root, estimated to low quality. -/// -/// @ingroup GPU -FfxFloat16x2 ffxApproximateReciprocalSquareRootHalf(FfxFloat16x2 a) -{ - return FFX_TO_FLOAT16X2(FFX_BROADCAST_UINT16X2(0x59a3) - (FFX_TO_UINT16X2(a) >> FFX_BROADCAST_UINT16X2(1))); -} - -/// Calculate a half-precision low-quality approximation for the reciprocal of the square root of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] a The value to calculate an approximate to the reciprocal of the square root for. -/// -/// @returns -/// An approximation of the reciprocal of the square root, estimated to low quality. -/// -/// @ingroup GPU -FfxFloat16x3 ffxApproximateReciprocalSquareRootHalf(FfxFloat16x3 a) -{ - return FFX_TO_FLOAT16X3(FFX_BROADCAST_UINT16X3(0x59a3) - (FFX_TO_UINT16X3(a) >> FFX_BROADCAST_UINT16X3(1))); -} - -/// Calculate a half-precision low-quality approximation for the reciprocal of the square root of a value. -/// -/// For additional information on the approximation family of functions, you can refer to Michal Drobot's excellent -/// presentation materials: -/// -/// - https://michaldrobot.files.wordpress.com/2014/05/gcn_alu_opt_digitaldragons2014.pdf -/// - https://github.com/michaldrobot/ShaderFastLibs/blob/master/ShaderFastMathLib.h -/// -/// @param [in] a The value to calculate an approximate to the reciprocal of the square root for. -/// -/// @returns -/// An approximation of the reciprocal of the square root, estimated to low quality. -/// -/// @ingroup GPU -FfxFloat16x4 ffxApproximateReciprocalSquareRootHalf(FfxFloat16x4 a) -{ - return FFX_TO_FLOAT16X4(FFX_BROADCAST_UINT16X4(0x59a3) - (FFX_TO_UINT16X4(a) >> FFX_BROADCAST_UINT16X4(1))); -} - -/// An approximation of sine. -/// -/// Valid input range is {-1 to 1} representing {0 to 2 pi}, and the output range -/// is {-1/4 to 1/4} representing {-1 to 1}. -/// -/// @param [in] x The value to calculate approximate sine for. -/// -/// @returns -/// The approximate sine of value. -FfxFloat16 ffxParabolicSinHalf(FfxFloat16 x) -{ - return x * abs(x) - x; -} - -/// An approximation of sine. -/// -/// Valid input range is {-1 to 1} representing {0 to 2 pi}, and the output range -/// is {-1/4 to 1/4} representing {-1 to 1}. -/// -/// @param [in] x The value to calculate approximate sine for. -/// -/// @returns -/// The approximate sine of value. -FfxFloat16x2 ffxParabolicSinHalf(FfxFloat16x2 x) -{ - return x * abs(x) - x; -} - -/// An approximation of cosine. -/// -/// Valid input range is {-1 to 1} representing {0 to 2 pi}, and the output range -/// is {-1/4 to 1/4} representing {-1 to 1}. -/// -/// @param [in] x The value to calculate approximate cosine for. -/// -/// @returns -/// The approximate cosine of value. -FfxFloat16 ffxParabolicCosHalf(FfxFloat16 x) -{ - x = ffxFract(x * FFX_BROADCAST_FLOAT16(0.5) + FFX_BROADCAST_FLOAT16(0.75)); - x = x * FFX_BROADCAST_FLOAT16(2.0) - FFX_BROADCAST_FLOAT16(1.0); - return ffxParabolicSinHalf(x); -} - -/// An approximation of cosine. -/// -/// Valid input range is {-1 to 1} representing {0 to 2 pi}, and the output range -/// is {-1/4 to 1/4} representing {-1 to 1}. -/// -/// @param [in] x The value to calculate approximate cosine for. -/// -/// @returns -/// The approximate cosine of value. -FfxFloat16x2 ffxParabolicCosHalf(FfxFloat16x2 x) -{ - x = ffxFract(x * FFX_BROADCAST_FLOAT16X2(0.5) + FFX_BROADCAST_FLOAT16X2(0.75)); - x = x * FFX_BROADCAST_FLOAT16X2(2.0) - FFX_BROADCAST_FLOAT16X2(1.0); - return ffxParabolicSinHalf(x); -} - -/// An approximation of both sine and cosine. -/// -/// Valid input range is {-1 to 1} representing {0 to 2 pi}, and the output range -/// is {-1/4 to 1/4} representing {-1 to 1}. -/// -/// @param [in] x The value to calculate approximate cosine for. -/// -/// @returns -/// A FfxFloat32x2 containing approximations of both sine and cosine of value. -FfxFloat16x2 ffxParabolicSinCosHalf(FfxFloat16 x) -{ - FfxFloat16 y = ffxFract(x * FFX_BROADCAST_FLOAT16(0.5) + FFX_BROADCAST_FLOAT16(0.75)); - y = y * FFX_BROADCAST_FLOAT16(2.0) - FFX_BROADCAST_FLOAT16(1.0); - return ffxParabolicSinHalf(FfxFloat16x2(x, y)); -} - -/// Conditional free logic AND operation using two half-precision values. -/// -/// @param [in] x The first value to be fed into the AND operator. -/// @param [in] y The second value to be fed into the AND operator. -/// -/// @returns -/// Result of the AND operation. -/// -/// @ingroup GPU -FfxUInt16 ffxZeroOneAndHalf(FfxUInt16 x, FfxUInt16 y) -{ - return min(x, y); -} - -/// Conditional free logic AND operation using two half-precision values. -/// -/// @param [in] x The first value to be fed into the AND operator. -/// @param [in] y The second value to be fed into the AND operator. -/// -/// @returns -/// Result of the AND operation. -/// -/// @ingroup GPU -FfxUInt16x2 ffxZeroOneAndHalf(FfxUInt16x2 x, FfxUInt16x2 y) -{ - return min(x, y); -} - -/// Conditional free logic AND operation using two half-precision values. -/// -/// @param [in] x The first value to be fed into the AND operator. -/// @param [in] y The second value to be fed into the AND operator. -/// -/// @returns -/// Result of the AND operation. -/// -/// @ingroup GPU -FfxUInt16x3 ffxZeroOneAndHalf(FfxUInt16x3 x, FfxUInt16x3 y) -{ - return min(x, y); -} - -/// Conditional free logic AND operation using two half-precision values. -/// -/// @param [in] x The first value to be fed into the AND operator. -/// @param [in] y The second value to be fed into the AND operator. -/// -/// @returns -/// Result of the AND operation. -/// -/// @ingroup GPU -FfxUInt16x4 ffxZeroOneAndHalf(FfxUInt16x4 x, FfxUInt16x4 y) -{ - return min(x, y); -} - -/// Conditional free logic NOT operation using two half-precision values. -/// -/// @param [in] x The first value to be fed into the NOT operator. -/// @param [in] y The second value to be fed into the NOT operator. -/// -/// @returns -/// Result of the NOT operation. -/// -/// @ingroup GPU -FfxUInt16 ffxZeroOneNotHalf(FfxUInt16 x) -{ - return x ^ FFX_BROADCAST_UINT16(1); -} - -/// Conditional free logic NOT operation using two half-precision values. -/// -/// @param [in] x The first value to be fed into the NOT operator. -/// @param [in] y The second value to be fed into the NOT operator. -/// -/// @returns -/// Result of the NOT operation. -/// -/// @ingroup GPU -FfxUInt16x2 ffxZeroOneNotHalf(FfxUInt16x2 x) -{ - return x ^ FFX_BROADCAST_UINT16X2(1); -} - -/// Conditional free logic NOT operation using two half-precision values. -/// -/// @param [in] x The first value to be fed into the NOT operator. -/// @param [in] y The second value to be fed into the NOT operator. -/// -/// @returns -/// Result of the NOT operation. -/// -/// @ingroup GPU -FfxUInt16x3 ffxZeroOneNotHalf(FfxUInt16x3 x) -{ - return x ^ FFX_BROADCAST_UINT16X3(1); -} - -/// Conditional free logic NOT operation using two half-precision values. -/// -/// @param [in] x The first value to be fed into the NOT operator. -/// @param [in] y The second value to be fed into the NOT operator. -/// -/// @returns -/// Result of the NOT operation. -/// -/// @ingroup GPU -FfxUInt16x4 ffxZeroOneNotHalf(FfxUInt16x4 x) -{ - return x ^ FFX_BROADCAST_UINT16X4(1); -} - -/// Conditional free logic OR operation using two half-precision values. -/// -/// @param [in] x The first value to be fed into the OR operator. -/// @param [in] y The second value to be fed into the OR operator. -/// -/// @returns -/// Result of the OR operation. -/// -/// @ingroup GPU -FfxUInt16 ffxZeroOneOrHalf(FfxUInt16 x, FfxUInt16 y) -{ - return max(x, y); -} - -/// Conditional free logic OR operation using two half-precision values. -/// -/// @param [in] x The first value to be fed into the OR operator. -/// @param [in] y The second value to be fed into the OR operator. -/// -/// @returns -/// Result of the OR operation. -/// -/// @ingroup GPU -FfxUInt16x2 ffxZeroOneOrHalf(FfxUInt16x2 x, FfxUInt16x2 y) -{ - return max(x, y); -} - -/// Conditional free logic OR operation using two half-precision values. -/// -/// @param [in] x The first value to be fed into the OR operator. -/// @param [in] y The second value to be fed into the OR operator. -/// -/// @returns -/// Result of the OR operation. -/// -/// @ingroup GPU -FfxUInt16x3 ffxZeroOneOrHalf(FfxUInt16x3 x, FfxUInt16x3 y) -{ - return max(x, y); -} - -/// Conditional free logic OR operation using two half-precision values. -/// -/// @param [in] x The first value to be fed into the OR operator. -/// @param [in] y The second value to be fed into the OR operator. -/// -/// @returns -/// Result of the OR operation. -/// -/// @ingroup GPU -FfxUInt16x4 ffxZeroOneOrHalf(FfxUInt16x4 x, FfxUInt16x4 y) -{ - return max(x, y); -} - -/// Convert a half-precision FfxFloat32 value between 0.0f and 1.0f to a half-precision Uint. -/// -/// @param [in] x The value to converted to a Uint. -/// -/// @returns -/// The converted Uint value. -/// -/// @ingroup GPU -FfxUInt16 ffxZeroOneFloat16ToUint16(FfxFloat16 x) -{ - return FFX_TO_UINT16(x * FFX_TO_FLOAT16(FFX_TO_UINT16(1))); -} - -/// Convert a half-precision FfxFloat32 value between 0.0f and 1.0f to a half-precision Uint. -/// -/// @param [in] x The value to converted to a Uint. -/// -/// @returns -/// The converted Uint value. -/// -/// @ingroup GPU -FfxUInt16x2 ffxZeroOneFloat16x2ToUint16x2(FfxFloat16x2 x) -{ - return FFX_TO_UINT16X2(x * FFX_TO_FLOAT16X2(FfxUInt16x2(1, 1))); -} - -/// Convert a half-precision FfxFloat32 value between 0.0f and 1.0f to a half-precision Uint. -/// -/// @param [in] x The value to converted to a Uint. -/// -/// @returns -/// The converted Uint value. -/// -/// @ingroup GPU -FfxUInt16x3 ffxZeroOneFloat16x3ToUint16x3(FfxFloat16x3 x) -{ - return FFX_TO_UINT16X3(x * FFX_TO_FLOAT16X3(FfxUInt16x3(1, 1, 1))); -} - -/// Convert a half-precision FfxFloat32 value between 0.0f and 1.0f to a half-precision Uint. -/// -/// @param [in] x The value to converted to a Uint. -/// -/// @returns -/// The converted Uint value. -/// -/// @ingroup GPU -FfxUInt16x4 ffxZeroOneFloat16x4ToUint16x4(FfxFloat16x4 x) -{ - return FFX_TO_UINT16X4(x * FFX_TO_FLOAT16X4(FfxUInt16x4(1, 1, 1, 1))); -} - -/// Convert a half-precision FfxUInt32 value between 0 and 1 to a half-precision FfxFloat32. -/// -/// @param [in] x The value to converted to a half-precision FfxFloat32. -/// -/// @returns -/// The converted half-precision FfxFloat32 value. -/// -/// @ingroup GPU -FfxFloat16 ffxZeroOneUint16ToFloat16(FfxUInt16 x) -{ - return FFX_TO_FLOAT16(x * FFX_TO_UINT16(FFX_TO_FLOAT16(1.0))); -} - -/// Convert a half-precision FfxUInt32 value between 0 and 1 to a half-precision FfxFloat32. -/// -/// @param [in] x The value to converted to a half-precision FfxFloat32. -/// -/// @returns -/// The converted half-precision FfxFloat32 value. -/// -/// @ingroup GPU -FfxFloat16x2 ffxZeroOneUint16x2ToFloat16x2(FfxUInt16x2 x) -{ - return FFX_TO_FLOAT16X2(x * FFX_TO_UINT16X2(FfxUInt16x2(FFX_TO_FLOAT16(1.0), FFX_TO_FLOAT16(1.0)))); -} - -/// Convert a half-precision FfxUInt32 value between 0 and 1 to a half-precision FfxFloat32. -/// -/// @param [in] x The value to converted to a half-precision FfxFloat32. -/// -/// @returns -/// The converted half-precision FfxFloat32 value. -/// -/// @ingroup GPU -FfxFloat16x3 ffxZeroOneUint16x3ToFloat16x3(FfxUInt16x3 x) -{ - return FFX_TO_FLOAT16X3(x * FFX_TO_UINT16X3(FfxUInt16x3(FFX_TO_FLOAT16(1.0), FFX_TO_FLOAT16(1.0), FFX_TO_FLOAT16(1.0)))); -} - -/// Convert a half-precision FfxUInt32 value between 0 and 1 to a half-precision FfxFloat32. -/// -/// @param [in] x The value to converted to a half-precision FfxFloat32. -/// -/// @returns -/// The converted half-precision FfxFloat32 value. -/// -/// @ingroup GPU -FfxFloat16x4 ffxZeroOneUint16x4ToFloat16x4(FfxUInt16x4 x) -{ - return FFX_TO_FLOAT16X4(x * FFX_TO_UINT16X4(FfxUInt16x4(FFX_TO_FLOAT16(1.0), FFX_TO_FLOAT16(1.0), FFX_TO_FLOAT16(1.0), FFX_TO_FLOAT16(1.0)))); -} - -/// Conditional free logic AND operation using two half-precision values. -/// -/// @param [in] x The first value to be fed into the AND operator. -/// @param [in] y The second value to be fed into the AND operator. -/// -/// @returns -/// Result of the AND operation. -/// -/// @ingroup GPU -FfxFloat16 ffxZeroOneAndHalf(FfxFloat16 x, FfxFloat16 y) -{ - return min(x, y); -} - -/// Conditional free logic AND operation using two half-precision values. -/// -/// @param [in] x The first value to be fed into the AND operator. -/// @param [in] y The second value to be fed into the AND operator. -/// -/// @returns -/// Result of the AND operation. -/// -/// @ingroup GPU -FfxFloat16x2 ffxZeroOneAndHalf(FfxFloat16x2 x, FfxFloat16x2 y) -{ - return min(x, y); -} - -/// Conditional free logic AND operation using two half-precision values. -/// -/// @param [in] x The first value to be fed into the AND operator. -/// @param [in] y The second value to be fed into the AND operator. -/// -/// @returns -/// Result of the AND operation. -/// -/// @ingroup GPU -FfxFloat16x3 ffxZeroOneAndHalf(FfxFloat16x3 x, FfxFloat16x3 y) -{ - return min(x, y); -} - -/// Conditional free logic AND operation using two half-precision values. -/// -/// @param [in] x The first value to be fed into the AND operator. -/// @param [in] y The second value to be fed into the AND operator. -/// -/// @returns -/// Result of the AND operation. -/// -/// @ingroup GPU -FfxFloat16x4 ffxZeroOneAndHalf(FfxFloat16x4 x, FfxFloat16x4 y) -{ - return min(x, y); -} - -/// Conditional free logic AND NOT operation using two half-precision values. -/// -/// @param [in] x The first value to be fed into the AND NOT operator. -/// @param [in] y The second value to be fed into the AND NOT operator. -/// -/// @returns -/// Result of the AND NOT operation. -/// -/// @ingroup GPU -FfxFloat16 ffxSignedZeroOneAndOrHalf(FfxFloat16 x, FfxFloat16 y) -{ - return (-x) * y + FFX_BROADCAST_FLOAT16(1.0); -} - -/// Conditional free logic AND NOT operation using two half-precision values. -/// -/// @param [in] x The first value to be fed into the AND NOT operator. -/// @param [in] y The second value to be fed into the AND NOT operator. -/// -/// @returns -/// Result of the AND NOT operation. -/// -/// @ingroup GPU -FfxFloat16x2 ffxSignedZeroOneAndOrHalf(FfxFloat16x2 x, FfxFloat16x2 y) -{ - return (-x) * y + FFX_BROADCAST_FLOAT16X2(1.0); -} - -/// Conditional free logic AND NOT operation using two half-precision values. -/// -/// @param [in] x The first value to be fed into the AND NOT operator. -/// @param [in] y The second value to be fed into the AND NOT operator. -/// -/// @returns -/// Result of the AND NOT operation. -/// -/// @ingroup GPU -FfxFloat16x3 ffxSignedZeroOneAndOrHalf(FfxFloat16x3 x, FfxFloat16x3 y) -{ - return (-x) * y + FFX_BROADCAST_FLOAT16X3(1.0); -} - -/// Conditional free logic AND NOT operation using two half-precision values. -/// -/// @param [in] x The first value to be fed into the AND NOT operator. -/// @param [in] y The second value to be fed into the AND NOT operator. -/// -/// @returns -/// Result of the AND NOT operation. -/// -/// @ingroup GPU -FfxFloat16x4 ffxSignedZeroOneAndOrHalf(FfxFloat16x4 x, FfxFloat16x4 y) -{ - return (-x) * y + FFX_BROADCAST_FLOAT16X4(1.0); -} - -/// Conditional free logic AND operation using two half-precision values followed by -/// a NOT operation using the resulting value and a third half-precision value. -/// -/// @param [in] x The first value to be fed into the AND operator. -/// @param [in] y The second value to be fed into the AND operator. -/// @param [in] z The second value to be fed into the OR operator. -/// -/// @returns -/// Result of the AND OR operation. -/// -/// @ingroup GPU -FfxFloat16 ffxZeroOneAndOrHalf(FfxFloat16 x, FfxFloat16 y, FfxFloat16 z) -{ - return ffxSaturate(x * y + z); -} - -/// Conditional free logic AND operation using two half-precision values followed by -/// a NOT operation using the resulting value and a third half-precision value. -/// -/// @param [in] x The first value to be fed into the AND operator. -/// @param [in] y The second value to be fed into the AND operator. -/// @param [in] z The second value to be fed into the OR operator. -/// -/// @returns -/// Result of the AND OR operation. -/// -/// @ingroup GPU -FfxFloat16x2 ffxZeroOneAndOrHalf(FfxFloat16x2 x, FfxFloat16x2 y, FfxFloat16x2 z) -{ - return ffxSaturate(x * y + z); -} - -/// Conditional free logic AND operation using two half-precision values followed by -/// a NOT operation using the resulting value and a third half-precision value. -/// -/// @param [in] x The first value to be fed into the AND operator. -/// @param [in] y The second value to be fed into the AND operator. -/// @param [in] z The second value to be fed into the OR operator. -/// -/// @returns -/// Result of the AND OR operation. -/// -/// @ingroup GPU -FfxFloat16x3 ffxZeroOneAndOrHalf(FfxFloat16x3 x, FfxFloat16x3 y, FfxFloat16x3 z) -{ - return ffxSaturate(x * y + z); -} - -/// Conditional free logic AND operation using two half-precision values followed by -/// a NOT operation using the resulting value and a third half-precision value. -/// -/// @param [in] x The first value to be fed into the AND operator. -/// @param [in] y The second value to be fed into the AND operator. -/// @param [in] z The second value to be fed into the OR operator. -/// -/// @returns -/// Result of the AND OR operation. -/// -/// @ingroup GPU -FfxFloat16x4 ffxZeroOneAndOrHalf(FfxFloat16x4 x, FfxFloat16x4 y, FfxFloat16x4 z) -{ - return ffxSaturate(x * y + z); -} - -/// Given a half-precision value, returns 1.0 if greater than zero and 0.0 if not. -/// -/// @param [in] x The value to be compared. -/// -/// @returns -/// Result of the greater than zero comparison. -/// -/// @ingroup GPU -FfxFloat16 ffxZeroOneIsGreaterThanZeroHalf(FfxFloat16 x) -{ - return ffxSaturate(x * FFX_BROADCAST_FLOAT16(FFX_POSITIVE_INFINITY_HALF)); -} - -/// Given a half-precision value, returns 1.0 if greater than zero and 0.0 if not. -/// -/// @param [in] x The value to be compared. -/// -/// @returns -/// Result of the greater than zero comparison. -/// -/// @ingroup GPU -FfxFloat16x2 ffxZeroOneIsGreaterThanZeroHalf(FfxFloat16x2 x) -{ - return ffxSaturate(x * FFX_BROADCAST_FLOAT16X2(FFX_POSITIVE_INFINITY_HALF)); -} - -/// Given a half-precision value, returns 1.0 if greater than zero and 0.0 if not. -/// -/// @param [in] x The value to be compared. -/// -/// @returns -/// Result of the greater than zero comparison. -/// -/// @ingroup GPU -FfxFloat16x3 ffxZeroOneIsGreaterThanZeroHalf(FfxFloat16x3 x) -{ - return ffxSaturate(x * FFX_BROADCAST_FLOAT16X3(FFX_POSITIVE_INFINITY_HALF)); -} - -/// Given a half-precision value, returns 1.0 if greater than zero and 0.0 if not. -/// -/// @param [in] x The value to be compared. -/// -/// @returns -/// Result of the greater than zero comparison. -/// -/// @ingroup GPU -FfxFloat16x4 ffxZeroOneIsGreaterThanZeroHalf(FfxFloat16x4 x) -{ - return ffxSaturate(x * FFX_BROADCAST_FLOAT16X4(FFX_POSITIVE_INFINITY_HALF)); -} - -/// Conditional free logic signed NOT operation using two half-precision FfxFloat32 values. -/// -/// @param [in] x The first value to be fed into the AND OR operator. -/// -/// @returns -/// Result of the AND OR operation. -/// -/// @ingroup GPU -FfxFloat16 ffxZeroOneNotHalf(FfxFloat16 x) -{ - return FFX_BROADCAST_FLOAT16(1.0) - x; -} - -/// Conditional free logic signed NOT operation using two half-precision FfxFloat32 values. -/// -/// @param [in] x The first value to be fed into the AND OR operator. -/// -/// @returns -/// Result of the AND OR operation. -/// -/// @ingroup GPU -FfxFloat16x2 ffxZeroOneNotHalf(FfxFloat16x2 x) -{ - return FFX_BROADCAST_FLOAT16X2(1.0) - x; -} - -/// Conditional free logic signed NOT operation using two half-precision FfxFloat32 values. -/// -/// @param [in] x The first value to be fed into the AND OR operator. -/// -/// @returns -/// Result of the AND OR operation. -/// -/// @ingroup GPU -FfxFloat16x3 ffxZeroOneNotHalf(FfxFloat16x3 x) -{ - return FFX_BROADCAST_FLOAT16X3(1.0) - x; -} - -/// Conditional free logic signed NOT operation using two half-precision FfxFloat32 values. -/// -/// @param [in] x The first value to be fed into the AND OR operator. -/// -/// @returns -/// Result of the AND OR operation. -/// -/// @ingroup GPU -FfxFloat16x4 ffxZeroOneNotHalf(FfxFloat16x4 x) -{ - return FFX_BROADCAST_FLOAT16X4(1.0) - x; -} - -/// Conditional free logic OR operation using two half-precision FfxFloat32 values. -/// -/// @param [in] x The first value to be fed into the OR operator. -/// @param [in] y The second value to be fed into the OR operator. -/// -/// @returns -/// Result of the OR operation. -/// -/// @ingroup GPU -FfxFloat16 ffxZeroOneOrHalf(FfxFloat16 x, FfxFloat16 y) -{ - return max(x, y); -} - -/// Conditional free logic OR operation using two half-precision FfxFloat32 values. -/// -/// @param [in] x The first value to be fed into the OR operator. -/// @param [in] y The second value to be fed into the OR operator. -/// -/// @returns -/// Result of the OR operation. -/// -/// @ingroup GPU -FfxFloat16x2 ffxZeroOneOrHalf(FfxFloat16x2 x, FfxFloat16x2 y) -{ - return max(x, y); -} - -/// Conditional free logic OR operation using two half-precision FfxFloat32 values. -/// -/// @param [in] x The first value to be fed into the OR operator. -/// @param [in] y The second value to be fed into the OR operator. -/// -/// @returns -/// Result of the OR operation. -/// -/// @ingroup GPU -FfxFloat16x3 ffxZeroOneOrHalf(FfxFloat16x3 x, FfxFloat16x3 y) -{ - return max(x, y); -} - -/// Conditional free logic OR operation using two half-precision FfxFloat32 values. -/// -/// @param [in] x The first value to be fed into the OR operator. -/// @param [in] y The second value to be fed into the OR operator. -/// -/// @returns -/// Result of the OR operation. -/// -/// @ingroup GPU -FfxFloat16x4 ffxZeroOneOrHalf(FfxFloat16x4 x, FfxFloat16x4 y) -{ - return max(x, y); -} - -/// Choose between two half-precision FfxFloat32 values if the first paramter is greater than zero. -/// -/// @param [in] x The value to compare against zero. -/// @param [in] y The value to return if the comparision is greater than zero. -/// @param [in] z The value to return if the comparision is less than or equal to zero. -/// -/// @returns -/// The selected value. -/// -/// @ingroup GPU -FfxFloat16 ffxZeroOneSelectHalf(FfxFloat16 x, FfxFloat16 y, FfxFloat16 z) -{ - FfxFloat16 r = (-x) * z + z; - return x * y + r; -} - -/// Choose between two half-precision FfxFloat32 values if the first paramter is greater than zero. -/// -/// @param [in] x The value to compare against zero. -/// @param [in] y The value to return if the comparision is greater than zero. -/// @param [in] z The value to return if the comparision is less than or equal to zero. -/// -/// @returns -/// The selected value. -/// -/// @ingroup GPU -FfxFloat16x2 ffxZeroOneSelectHalf(FfxFloat16x2 x, FfxFloat16x2 y, FfxFloat16x2 z) -{ - FfxFloat16x2 r = (-x) * z + z; - return x * y + r; -} - -/// Choose between two half-precision FfxFloat32 values if the first paramter is greater than zero. -/// -/// @param [in] x The value to compare against zero. -/// @param [in] y The value to return if the comparision is greater than zero. -/// @param [in] z The value to return if the comparision is less than or equal to zero. -/// -/// @returns -/// The selected value. -/// -/// @ingroup GPU -FfxFloat16x3 ffxZeroOneSelectHalf(FfxFloat16x3 x, FfxFloat16x3 y, FfxFloat16x3 z) -{ - FfxFloat16x3 r = (-x) * z + z; - return x * y + r; -} - -/// Choose between two half-precision FfxFloat32 values if the first paramter is greater than zero. -/// -/// @param [in] x The value to compare against zero. -/// @param [in] y The value to return if the comparision is greater than zero. -/// @param [in] z The value to return if the comparision is less than or equal to zero. -/// -/// @returns -/// The selected value. -/// -/// @ingroup GPU -FfxFloat16x4 ffxZeroOneSelectHalf(FfxFloat16x4 x, FfxFloat16x4 y, FfxFloat16x4 z) -{ - FfxFloat16x4 r = (-x) * z + z; - return x * y + r; -} - -/// Given a half-precision value, returns 1.0 if less than zero and 0.0 if not. -/// -/// @param [in] x The value to be compared. -/// -/// @returns -/// Result of the sign value. -/// -/// @ingroup GPU -FfxFloat16 ffxZeroOneIsSignedHalf(FfxFloat16 x) -{ - return ffxSaturate(x * FFX_BROADCAST_FLOAT16(FFX_NEGATIVE_INFINITY_HALF)); -} - -/// Given a half-precision value, returns 1.0 if less than zero and 0.0 if not. -/// -/// @param [in] x The value to be compared. -/// -/// @returns -/// Result of the sign value. -/// -/// @ingroup GPU -FfxFloat16x2 ffxZeroOneIsSignedHalf(FfxFloat16x2 x) -{ - return ffxSaturate(x * FFX_BROADCAST_FLOAT16X2(FFX_NEGATIVE_INFINITY_HALF)); -} - -/// Given a half-precision value, returns 1.0 if less than zero and 0.0 if not. -/// -/// @param [in] x The value to be compared. -/// -/// @returns -/// Result of the sign value. -/// -/// @ingroup GPU -FfxFloat16x3 ffxZeroOneIsSignedHalf(FfxFloat16x3 x) -{ - return ffxSaturate(x * FFX_BROADCAST_FLOAT16X3(FFX_NEGATIVE_INFINITY_HALF)); -} - -/// Given a half-precision value, returns 1.0 if less than zero and 0.0 if not. -/// -/// @param [in] x The value to be compared. -/// -/// @returns -/// Result of the sign value. -/// -/// @ingroup GPU -FfxFloat16x4 ffxZeroOneIsSignedHalf(FfxFloat16x4 x) -{ - return ffxSaturate(x * FFX_BROADCAST_FLOAT16X4(FFX_NEGATIVE_INFINITY_HALF)); -} - -/// Compute a Rec.709 color space. -/// -/// Rec.709 is used for some HDTVs. -/// -/// Both Rec.709 and sRGB have a linear segment which as spec'ed would intersect the curved segment 2 times. -/// (a.) For 8-bit sRGB, steps {0 to 10.3} are in the linear region (4% of the encoding range). -/// (b.) For 8-bit 709, steps {0 to 20.7} are in the linear region (8% of the encoding range). -/// -/// @param [in] c The color to convert to Rec. 709. -/// -/// @returns -/// The color in Rec.709 space. -/// -/// @ingroup GPU -FfxFloat16 ffxRec709FromLinearHalf(FfxFloat16 c) -{ - FfxFloat16x3 j = FfxFloat16x3(0.018 * 4.5, 4.5, 0.45); - FfxFloat16x2 k = FfxFloat16x2(1.099, -0.099); - return clamp(j.x, c * j.y, pow(c, j.z) * k.x + k.y); -} - -/// Compute a Rec.709 color space. -/// -/// Rec.709 is used for some HDTVs. -/// -/// Both Rec.709 and sRGB have a linear segment which as spec'ed would intersect the curved segment 2 times. -/// (a.) For 8-bit sRGB, steps {0 to 10.3} are in the linear region (4% of the encoding range). -/// (b.) For 8-bit 709, steps {0 to 20.7} are in the linear region (8% of the encoding range). -/// -/// @param [in] c The color to convert to Rec. 709. -/// -/// @returns -/// The color in Rec.709 space. -/// -/// @ingroup GPU -FfxFloat16x2 ffxRec709FromLinearHalf(FfxFloat16x2 c) -{ - FfxFloat16x3 j = FfxFloat16x3(0.018 * 4.5, 4.5, 0.45); - FfxFloat16x2 k = FfxFloat16x2(1.099, -0.099); - return clamp(j.xx, c * j.yy, pow(c, j.zz) * k.xx + k.yy); -} - -/// Compute a Rec.709 color space. -/// -/// Rec.709 is used for some HDTVs. -/// -/// Both Rec.709 and sRGB have a linear segment which as spec'ed would intersect the curved segment 2 times. -/// (a.) For 8-bit sRGB, steps {0 to 10.3} are in the linear region (4% of the encoding range). -/// (b.) For 8-bit 709, steps {0 to 20.7} are in the linear region (8% of the encoding range). -/// -/// @param [in] c The color to convert to Rec. 709. -/// -/// @returns -/// The color in Rec.709 space. -/// -/// @ingroup GPU -FfxFloat16x3 ffxRec709FromLinearHalf(FfxFloat16x3 c) -{ - FfxFloat16x3 j = FfxFloat16x3(0.018 * 4.5, 4.5, 0.45); - FfxFloat16x2 k = FfxFloat16x2(1.099, -0.099); - return clamp(j.xxx, c * j.yyy, pow(c, j.zzz) * k.xxx + k.yyy); -} - -/// Compute a gamma value from a linear value. -/// -/// Typically 2.2 for some PC displays, or 2.4-2.5 for CRTs, or 2.2 FreeSync2 native. -/// -/// Note: 'rcpX' is '1/x', where the 'x' is what would be used in ffxLinearFromGammaHalf. -/// -/// @param [in] c The value to convert to gamma space from linear. -/// @param [in] rcpX The reciprocal of power value used for the gamma curve. -/// -/// @returns -/// A value in gamma space. -/// -/// @ingroup GPU -FfxFloat16 ffxGammaFromLinearHalf(FfxFloat16 c, FfxFloat16 rcpX) -{ - return pow(c, FFX_BROADCAST_FLOAT16(rcpX)); -} - -/// Compute a gamma value from a linear value. -/// -/// Typically 2.2 for some PC displays, or 2.4-2.5 for CRTs, or 2.2 FreeSync2 native. -/// -/// Note: 'rcpX' is '1/x', where the 'x' is what would be used in ffxLinearFromGammaHalf. -/// -/// @param [in] c The value to convert to gamma space from linear. -/// @param [in] rcpX The reciprocal of power value used for the gamma curve. -/// -/// @returns -/// A value in gamma space. -/// -/// @ingroup GPU -FfxFloat16x2 ffxGammaFromLinearHalf(FfxFloat16x2 c, FfxFloat16 rcpX) -{ - return pow(c, FFX_BROADCAST_FLOAT16X2(rcpX)); -} - -/// Compute a gamma value from a linear value. -/// -/// Typically 2.2 for some PC displays, or 2.4-2.5 for CRTs, or 2.2 FreeSync2 native. -/// -/// Note: 'rcpX' is '1/x', where the 'x' is what would be used in ffxLinearFromGammaHalf. -/// -/// @param [in] c The value to convert to gamma space from linear. -/// @param [in] rcpX The reciprocal of power value used for the gamma curve. -/// -/// @returns -/// A value in gamma space. -/// -/// @ingroup GPU -FfxFloat16x3 ffxGammaFromLinearHalf(FfxFloat16x3 c, FfxFloat16 rcpX) -{ - return pow(c, FFX_BROADCAST_FLOAT16X3(rcpX)); -} - -/// Compute an SRGB value from a linear value. -/// -/// @param [in] c The value to convert to SRGB from linear. -/// -/// @returns -/// A value in SRGB space. -/// -/// @ingroup GPU -FfxFloat16 ffxSrgbFromLinearHalf(FfxFloat16 c) -{ - FfxFloat16x3 j = FfxFloat16x3(0.0031308 * 12.92, 12.92, 1.0 / 2.4); - FfxFloat16x2 k = FfxFloat16x2(1.055, -0.055); - return clamp(j.x, c * j.y, pow(c, j.z) * k.x + k.y); -} - -/// Compute an SRGB value from a linear value. -/// -/// @param [in] c The value to convert to SRGB from linear. -/// -/// @returns -/// A value in SRGB space. -/// -/// @ingroup GPU -FfxFloat16x2 ffxSrgbFromLinearHalf(FfxFloat16x2 c) -{ - FfxFloat16x3 j = FfxFloat16x3(0.0031308 * 12.92, 12.92, 1.0 / 2.4); - FfxFloat16x2 k = FfxFloat16x2(1.055, -0.055); - return clamp(j.xx, c * j.yy, pow(c, j.zz) * k.xx + k.yy); -} - -/// Compute an SRGB value from a linear value. -/// -/// @param [in] c The value to convert to SRGB from linear. -/// -/// @returns -/// A value in SRGB space. -/// -/// @ingroup GPU -FfxFloat16x3 ffxSrgbFromLinearHalf(FfxFloat16x3 c) -{ - FfxFloat16x3 j = FfxFloat16x3(0.0031308 * 12.92, 12.92, 1.0 / 2.4); - FfxFloat16x2 k = FfxFloat16x2(1.055, -0.055); - return clamp(j.xxx, c * j.yyy, pow(c, j.zzz) * k.xxx + k.yyy); -} - -/// Compute the square root of a value. -/// -/// @param [in] c The value to compute the square root for. -/// -/// @returns -/// A square root of the input value. -/// -/// @ingroup GPU -FfxFloat16 ffxSquareRootHalf(FfxFloat16 c) -{ - return sqrt(c); -} - -/// Compute the square root of a value. -/// -/// @param [in] c The value to compute the square root for. -/// -/// @returns -/// A square root of the input value. -/// -/// @ingroup GPU -FfxFloat16x2 ffxSquareRootHalf(FfxFloat16x2 c) -{ - return sqrt(c); -} - -/// Compute the square root of a value. -/// -/// @param [in] c The value to compute the square root for. -/// -/// @returns -/// A square root of the input value. -/// -/// @ingroup GPU -FfxFloat16x3 ffxSquareRootHalf(FfxFloat16x3 c) -{ - return sqrt(c); -} - -/// Compute the cube root of a value. -/// -/// @param [in] c The value to compute the cube root for. -/// -/// @returns -/// A cube root of the input value. -/// -/// @ingroup GPU -FfxFloat16 ffxCubeRootHalf(FfxFloat16 c) -{ - return pow(c, FFX_BROADCAST_FLOAT16(1.0 / 3.0)); -} - -/// Compute the cube root of a value. -/// -/// @param [in] c The value to compute the cube root for. -/// -/// @returns -/// A cube root of the input value. -/// -/// @ingroup GPU -FfxFloat16x2 ffxCubeRootHalf(FfxFloat16x2 c) -{ - return pow(c, FFX_BROADCAST_FLOAT16X2(1.0 / 3.0)); -} - -/// Compute the cube root of a value. -/// -/// @param [in] c The value to compute the cube root for. -/// -/// @returns -/// A cube root of the input value. -/// -/// @ingroup GPU -FfxFloat16x3 ffxCubeRootHalf(FfxFloat16x3 c) -{ - return pow(c, FFX_BROADCAST_FLOAT16X3(1.0 / 3.0)); -} - -/// Compute a linear value from a REC.709 value. -/// -/// @param [in] c The value to convert to linear from REC.709. -/// -/// @returns -/// A value in linear space. -/// -/// @ingroup GPU -FfxFloat16 ffxLinearFromRec709Half(FfxFloat16 c) -{ - FfxFloat16x3 j = FfxFloat16x3(0.081 / 4.5, 1.0 / 4.5, 1.0 / 0.45); - FfxFloat16x2 k = FfxFloat16x2(1.0 / 1.099, 0.099 / 1.099); - return ffxZeroOneSelectHalf(ffxZeroOneIsSignedHalf(c - j.x), c * j.y, pow(c * k.x + k.y, j.z)); -} - -/// Compute a linear value from a REC.709 value. -/// -/// @param [in] c The value to convert to linear from REC.709. -/// -/// @returns -/// A value in linear space. -/// -/// @ingroup GPU -FfxFloat16x2 ffxLinearFromRec709Half(FfxFloat16x2 c) -{ - FfxFloat16x3 j = FfxFloat16x3(0.081 / 4.5, 1.0 / 4.5, 1.0 / 0.45); - FfxFloat16x2 k = FfxFloat16x2(1.0 / 1.099, 0.099 / 1.099); - return ffxZeroOneSelectHalf(ffxZeroOneIsSignedHalf(c - j.xx), c * j.yy, pow(c * k.xx + k.yy, j.zz)); -} - -/// Compute a linear value from a REC.709 value. -/// -/// @param [in] c The value to convert to linear from REC.709. -/// -/// @returns -/// A value in linear space. -/// -/// @ingroup GPU -FfxFloat16x3 ffxLinearFromRec709Half(FfxFloat16x3 c) -{ - FfxFloat16x3 j = FfxFloat16x3(0.081 / 4.5, 1.0 / 4.5, 1.0 / 0.45); - FfxFloat16x2 k = FfxFloat16x2(1.0 / 1.099, 0.099 / 1.099); - return ffxZeroOneSelectHalf(ffxZeroOneIsSignedHalf(c - j.xxx), c * j.yyy, pow(c * k.xxx + k.yyy, j.zzz)); -} - -/// Compute a linear value from a value in a gamma space. -/// -/// Typically 2.2 for some PC displays, or 2.4-2.5 for CRTs, or 2.2 FreeSync2 native. -/// -/// @param [in] c The value to convert to linear in gamma space. -/// @param [in] x The power value used for the gamma curve. -/// -/// @returns -/// A value in linear space. -/// -/// @ingroup GPU -FfxFloat16 ffxLinearFromGammaHalf(FfxFloat16 c, FfxFloat16 x) -{ - return pow(c, FFX_BROADCAST_FLOAT16(x)); -} - -/// Compute a linear value from a value in a gamma space. -/// -/// Typically 2.2 for some PC displays, or 2.4-2.5 for CRTs, or 2.2 FreeSync2 native. -/// -/// @param [in] c The value to convert to linear in gamma space. -/// @param [in] x The power value used for the gamma curve. -/// -/// @returns -/// A value in linear space. -/// -/// @ingroup GPU -FfxFloat16x2 ffxLinearFromGammaHalf(FfxFloat16x2 c, FfxFloat16 x) -{ - return pow(c, FFX_BROADCAST_FLOAT16X2(x)); -} - -/// Compute a linear value from a value in a gamma space. -/// -/// Typically 2.2 for some PC displays, or 2.4-2.5 for CRTs, or 2.2 FreeSync2 native. -/// -/// @param [in] c The value to convert to linear in gamma space. -/// @param [in] x The power value used for the gamma curve. -/// -/// @returns -/// A value in linear space. -/// -/// @ingroup GPU -FfxFloat16x3 ffxLinearFromGammaHalf(FfxFloat16x3 c, FfxFloat16 x) -{ - return pow(c, FFX_BROADCAST_FLOAT16X3(x)); -} - -/// Compute a linear value from a value in a SRGB space. -/// -/// Typically 2.2 for some PC displays, or 2.4-2.5 for CRTs, or 2.2 FreeSync2 native. -/// -/// @param [in] c The value to convert to linear in SRGB space. -/// -/// @returns -/// A value in linear space. -/// -/// @ingroup GPU -FfxFloat16 ffxLinearFromSrgbHalf(FfxFloat16 c) -{ - FfxFloat16x3 j = FfxFloat16x3(0.04045 / 12.92, 1.0 / 12.92, 2.4); - FfxFloat16x2 k = FfxFloat16x2(1.0 / 1.055, 0.055 / 1.055); - return ffxZeroOneSelectHalf(ffxZeroOneIsSignedHalf(c - j.x), c * j.y, pow(c * k.x + k.y, j.z)); -} - -/// Compute a linear value from a value in a SRGB space. -/// -/// Typically 2.2 for some PC displays, or 2.4-2.5 for CRTs, or 2.2 FreeSync2 native. -/// -/// @param [in] c The value to convert to linear in SRGB space. -/// -/// @returns -/// A value in linear space. -/// -/// @ingroup GPU -FfxFloat16x2 ffxLinearFromSrgbHalf(FfxFloat16x2 c) -{ - FfxFloat16x3 j = FfxFloat16x3(0.04045 / 12.92, 1.0 / 12.92, 2.4); - FfxFloat16x2 k = FfxFloat16x2(1.0 / 1.055, 0.055 / 1.055); - return ffxZeroOneSelectHalf(ffxZeroOneIsSignedHalf(c - j.xx), c * j.yy, pow(c * k.xx + k.yy, j.zz)); -} - -/// Compute a linear value from a value in a SRGB space. -/// -/// Typically 2.2 for some PC displays, or 2.4-2.5 for CRTs, or 2.2 FreeSync2 native. -/// -/// @param [in] c The value to convert to linear in SRGB space. -/// -/// @returns -/// A value in linear space. -/// -/// @ingroup GPU -FfxFloat16x3 ffxLinearFromSrgbHalf(FfxFloat16x3 c) -{ - FfxFloat16x3 j = FfxFloat16x3(0.04045 / 12.92, 1.0 / 12.92, 2.4); - FfxFloat16x2 k = FfxFloat16x2(1.0 / 1.055, 0.055 / 1.055); - return ffxZeroOneSelectHalf(ffxZeroOneIsSignedHalf(c - j.xxx), c * j.yyy, pow(c * k.xxx + k.yyy, j.zzz)); -} - -/// A remapping of 64x1 to 8x8 imposing rotated 2x2 pixel quads in quad linear. -/// -/// 543210 -/// ====== -/// ..xxx. -/// yy...y -/// -/// @param [in] a The input 1D coordinates to remap. -/// -/// @returns -/// The remapped 2D coordinates. -/// -/// @ingroup GPU -FfxUInt16x2 ffxRemapForQuadHalf(FfxUInt32 a) -{ - return FfxUInt16x2(bitfieldExtract(a, 1u, 3u), bitfieldInsertMask(bitfieldExtract(a, 3u, 3u), a, 1u)); -} - -/// A helper function performing a remap 64x1 to 8x8 remapping which is necessary for 2D wave reductions. -/// -/// The 64-wide lane indices to 8x8 remapping is performed as follows: -/// -/// 00 01 08 09 10 11 18 19 -/// 02 03 0a 0b 12 13 1a 1b -/// 04 05 0c 0d 14 15 1c 1d -/// 06 07 0e 0f 16 17 1e 1f -/// 20 21 28 29 30 31 38 39 -/// 22 23 2a 2b 32 33 3a 3b -/// 24 25 2c 2d 34 35 3c 3d -/// 26 27 2e 2f 36 37 3e 3f -/// -/// @param [in] a The input 1D coordinate to remap. -/// -/// @returns -/// The remapped 2D coordinates. -/// -/// @ingroup GPU -FfxUInt16x2 ffxRemapForWaveReductionHalf(FfxUInt32 a) -{ - return FfxUInt16x2(bitfieldInsertMask(bitfieldExtract(a, 2u, 3u), a, 1u), bitfieldInsertMask(bitfieldExtract(a, 3u, 3u), bitfieldExtract(a, 1u, 2u), 2u)); -} - -#endif // FFX_HALF diff --git a/Assets/Resources/FSR2/shaders/ffx_core_gpu_common_half.h.meta b/Assets/Resources/FSR2/shaders/ffx_core_gpu_common_half.h.meta deleted file mode 100644 index 1d826e4..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_core_gpu_common_half.h.meta +++ /dev/null @@ -1,60 +0,0 @@ -fileFormatVersion: 2 -guid: 142bf3947ada43541a0f31a328fdec07 -PluginImporter: - externalObjects: {} - serializedVersion: 2 - iconMap: {} - executionOrder: {} - defineConstraints: [] - isPreloaded: 0 - isOverridable: 0 - isExplicitlyReferenced: 0 - validateReferences: 1 - platformData: - - first: - : Any - second: - enabled: 0 - settings: - Exclude Editor: 1 - Exclude GameCoreScarlett: 1 - Exclude GameCoreXboxOne: 1 - Exclude Linux64: 1 - Exclude OSXUniversal: 1 - Exclude PS4: 1 - Exclude PS5: 1 - Exclude WebGL: 1 - Exclude Win: 1 - Exclude Win64: 1 - - first: - Any: - second: - enabled: 0 - settings: {} - - first: - Editor: Editor - second: - enabled: 0 - settings: - DefaultValueInitialized: true - - first: - Standalone: Linux64 - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win64 - second: - enabled: 0 - settings: - CPU: None - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_core_hlsl.h b/Assets/Resources/FSR2/shaders/ffx_core_hlsl.h deleted file mode 100644 index ad4ff65..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_core_hlsl.h +++ /dev/null @@ -1,1502 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -/// A define for abstracting shared memory between shading languages. -/// -/// @ingroup GPU -#define FFX_GROUPSHARED groupshared - -/// A define for abstracting compute memory barriers between shading languages. -/// -/// @ingroup GPU -#define FFX_GROUP_MEMORY_BARRIER GroupMemoryBarrierWithGroupSync - -/// A define added to accept static markup on functions to aid CPU/GPU portability of code. -/// -/// @ingroup GPU -#define FFX_STATIC static - -/// A define for abstracting loop unrolling between shading languages. -/// -/// @ingroup GPU -#define FFX_UNROLL [unroll] - -/// A define for abstracting a 'greater than' comparison operator between two types. -/// -/// @ingroup GPU -#define FFX_GREATER_THAN(x, y) x > y - -/// A define for abstracting a 'greater than or equal' comparison operator between two types. -/// -/// @ingroup GPU -#define FFX_GREATER_THAN_EQUAL(x, y) x >= y - -/// A define for abstracting a 'less than' comparison operator between two types. -/// -/// @ingroup GPU -#define FFX_LESS_THAN(x, y) x < y - -/// A define for abstracting a 'less than or equal' comparison operator between two types. -/// -/// @ingroup GPU -#define FFX_LESS_THAN_EQUAL(x, y) x <= y - -/// A define for abstracting an 'equal' comparison operator between two types. -/// -/// @ingroup GPU -#define FFX_EQUAL(x, y) x == y - -/// A define for abstracting a 'not equal' comparison operator between two types. -/// -/// @ingroup GPU -#define FFX_NOT_EQUAL(x, y) x != y - -/// Broadcast a scalar value to a 1-dimensional floating point vector. -/// -/// @ingroup GPU -#define FFX_BROADCAST_FLOAT32(x) FfxFloat32(x) - -/// Broadcast a scalar value to a 2-dimensional floating point vector. -/// -/// @ingroup GPU -#define FFX_BROADCAST_FLOAT32X2(x) FfxFloat32(x) - -/// Broadcast a scalar value to a 3-dimensional floating point vector. -/// -/// @ingroup GPU -#define FFX_BROADCAST_FLOAT32X3(x) FfxFloat32(x) - -/// Broadcast a scalar value to a 4-dimensional floating point vector. -/// -/// @ingroup GPU -#define FFX_BROADCAST_FLOAT32X4(x) FfxFloat32(x) - -/// Broadcast a scalar value to a 1-dimensional unsigned integer vector. -/// -/// @ingroup GPU -#define FFX_BROADCAST_UINT32(x) FfxUInt32(x) - -/// Broadcast a scalar value to a 2-dimensional unsigned integer vector. -/// -/// @ingroup GPU -#define FFX_BROADCAST_UINT32X2(x) FfxUInt32(x) - -/// Broadcast a scalar value to a 4-dimensional unsigned integer vector. -/// -/// @ingroup GPU -#define FFX_BROADCAST_UINT32X3(x) FfxUInt32(x) - -/// Broadcast a scalar value to a 4-dimensional unsigned integer vector. -/// -/// @ingroup GPU -#define FFX_BROADCAST_UINT32X4(x) FfxUInt32(x) - -/// Broadcast a scalar value to a 1-dimensional signed integer vector. -/// -/// @ingroup GPU -#define FFX_BROADCAST_INT32(x) FfxInt32(x) - -/// Broadcast a scalar value to a 2-dimensional signed integer vector. -/// -/// @ingroup GPU -#define FFX_BROADCAST_INT32X2(x) FfxInt32(x) - -/// Broadcast a scalar value to a 3-dimensional signed integer vector. -/// -/// @ingroup GPU -#define FFX_BROADCAST_INT32X3(x) FfxInt32(x) - -/// Broadcast a scalar value to a 4-dimensional signed integer vector. -/// -/// @ingroup GPU -#define FFX_BROADCAST_INT32X4(x) FfxInt32(x) - -/// Broadcast a scalar value to a 1-dimensional half-precision floating point vector. -/// -/// @ingroup GPU -#define FFX_BROADCAST_MIN_FLOAT16(a) FFX_MIN16_F(a) - -/// Broadcast a scalar value to a 2-dimensional half-precision floating point vector. -/// -/// @ingroup GPU -#define FFX_BROADCAST_MIN_FLOAT16X2(a) FFX_MIN16_F(a) - -/// Broadcast a scalar value to a 3-dimensional half-precision floating point vector. -/// -/// @ingroup GPU -#define FFX_BROADCAST_MIN_FLOAT16X3(a) FFX_MIN16_F(a) - -/// Broadcast a scalar value to a 4-dimensional half-precision floating point vector. -/// -/// @ingroup GPU -#define FFX_BROADCAST_MIN_FLOAT16X4(a) FFX_MIN16_F(a) - -/// Broadcast a scalar value to a 1-dimensional half-precision unsigned integer vector. -/// -/// @ingroup GPU -#define FFX_BROADCAST_MIN_UINT16(a) FFX_MIN16_U(a) - -/// Broadcast a scalar value to a 2-dimensional half-precision unsigned integer vector. -/// -/// @ingroup GPU -#define FFX_BROADCAST_MIN_UINT16X2(a) FFX_MIN16_U(a) - -/// Broadcast a scalar value to a 3-dimensional half-precision unsigned integer vector. -/// -/// @ingroup GPU -#define FFX_BROADCAST_MIN_UINT16X3(a) FFX_MIN16_U(a) - -/// Broadcast a scalar value to a 4-dimensional half-precision unsigned integer vector. -/// -/// @ingroup GPU -#define FFX_BROADCAST_MIN_UINT16X4(a) FFX_MIN16_U(a) - -/// Broadcast a scalar value to a 1-dimensional half-precision signed integer vector. -/// -/// @ingroup GPU -#define FFX_BROADCAST_MIN_INT16(a) FFX_MIN16_I(a) - -/// Broadcast a scalar value to a 2-dimensional half-precision signed integer vector. -/// -/// @ingroup GPU -#define FFX_BROADCAST_MIN_INT16X2(a) FFX_MIN16_I(a) - -/// Broadcast a scalar value to a 3-dimensional half-precision signed integer vector. -/// -/// @ingroup GPU -#define FFX_BROADCAST_MIN_INT16X3(a) FFX_MIN16_I(a) - -/// Broadcast a scalar value to a 4-dimensional half-precision signed integer vector. -/// -/// @ingroup GPU -#define FFX_BROADCAST_MIN_INT16X4(a) FFX_MIN16_I(a) - -/// Pack 2x32-bit floating point values in a single 32bit value. -/// -/// This function first converts each component of value into their nearest 16-bit floating -/// point representation, and then stores the X and Y components in the lower and upper 16 bits of the -/// 32bit unsigned integer respectively. -/// -/// @param [in] value A 2-dimensional floating point value to convert and pack. -/// -/// @returns -/// A packed 32bit value containing 2 16bit floating point values. -/// -/// @ingroup HLSL -FfxUInt32 packHalf2x16(FfxFloat32x2 value) -{ - return f32tof16(value.x) | (f32tof16(value.y) << 16); -} - -/// Broadcast a scalar value to a 2-dimensional floating point vector. -/// -/// @param [in] value The value to to broadcast. -/// -/// @returns -/// A 2-dimensional floating point vector with value in each component. -/// -/// @ingroup HLSL -FfxFloat32x2 ffxBroadcast2(FfxFloat32 value) -{ - return FfxFloat32x2(value, value); -} - -/// Broadcast a scalar value to a 3-dimensional floating point vector. -/// -/// @param [in] value The value to to broadcast. -/// -/// @returns -/// A 3-dimensional floating point vector with value in each component. -/// -/// @ingroup HLSL -FfxFloat32x3 ffxBroadcast3(FfxFloat32 value) -{ - return FfxFloat32x3(value, value, value); -} - -/// Broadcast a scalar value to a 4-dimensional floating point vector. -/// -/// @param [in] value The value to to broadcast. -/// -/// @returns -/// A 4-dimensional floating point vector with value in each component. -/// -/// @ingroup HLSL -FfxFloat32x4 ffxBroadcast4(FfxFloat32 value) -{ - return FfxFloat32x4(value, value, value, value); -} - -/// Broadcast a scalar value to a 2-dimensional signed integer vector. -/// -/// @param [in] value The value to to broadcast. -/// -/// @returns -/// A 2-dimensional signed integer vector with value in each component. -/// -/// @ingroup HLSL -FfxInt32x2 ffxBroadcast2(FfxInt32 value) -{ - return FfxInt32x2(value, value); -} - -/// Broadcast a scalar value to a 3-dimensional signed integer vector. -/// -/// @param [in] value The value to to broadcast. -/// -/// @returns -/// A 3-dimensional signed integer vector with value in each component. -/// -/// @ingroup HLSL -FfxUInt32x3 ffxBroadcast3(FfxInt32 value) -{ - return FfxUInt32x3(value, value, value); -} - -/// Broadcast a scalar value to a 4-dimensional signed integer vector. -/// -/// @param [in] value The value to to broadcast. -/// -/// @returns -/// A 4-dimensional signed integer vector with value in each component. -/// -/// @ingroup HLSL -FfxInt32x4 ffxBroadcast4(FfxInt32 value) -{ - return FfxInt32x4(value, value, value, value); -} - -/// Broadcast a scalar value to a 2-dimensional unsigned integer vector. -/// -/// @param [in] value The value to to broadcast. -/// -/// @returns -/// A 2-dimensional unsigned integer vector with value in each component. -/// -/// @ingroup HLSL -FfxUInt32x2 ffxBroadcast2(FfxUInt32 value) -{ - return FfxUInt32x2(value, value); -} - -/// Broadcast a scalar value to a 3-dimensional unsigned integer vector. -/// -/// @param [in] value The value to to broadcast. -/// -/// @returns -/// A 3-dimensional unsigned integer vector with value in each component. -/// -/// @ingroup HLSL -FfxUInt32x3 ffxBroadcast3(FfxUInt32 value) -{ - return FfxUInt32x3(value, value, value); -} - -/// Broadcast a scalar value to a 4-dimensional unsigned integer vector. -/// -/// @param [in] value The value to to broadcast. -/// -/// @returns -/// A 4-dimensional unsigned integer vector with value in each component. -/// -/// @ingroup HLSL -FfxUInt32x4 ffxBroadcast4(FfxUInt32 value) -{ - return FfxUInt32x4(value, value, value, value); -} - -FfxUInt32 bitfieldExtract(FfxUInt32 src, FfxUInt32 off, FfxUInt32 bits) -{ - FfxUInt32 mask = (1u << bits) - 1; - return (src >> off) & mask; -} - -FfxUInt32 bitfieldInsert(FfxUInt32 src, FfxUInt32 ins, FfxUInt32 mask) -{ - return (ins & mask) | (src & (~mask)); -} - -FfxUInt32 bitfieldInsertMask(FfxUInt32 src, FfxUInt32 ins, FfxUInt32 bits) -{ - FfxUInt32 mask = (1u << bits) - 1; - return (ins & mask) | (src & (~mask)); -} - -/// Interprets the bit pattern of x as an unsigned integer. -/// -/// @param [in] value The input value. -/// -/// @returns -/// The input interpreted as an unsigned integer. -/// -/// @ingroup HLSL -FfxUInt32 ffxAsUInt32(FfxFloat32 x) -{ - return asuint(x); -} - -/// Interprets the bit pattern of x as an unsigned integer. -/// -/// @param [in] value The input value. -/// -/// @returns -/// The input interpreted as an unsigned integer. -/// -/// @ingroup HLSL -FfxUInt32x2 ffxAsUInt32(FfxFloat32x2 x) -{ - return asuint(x); -} - -/// Interprets the bit pattern of x as an unsigned integer. -/// -/// @param [in] value The input value. -/// -/// @returns -/// The input interpreted as an unsigned integer. -/// -/// @ingroup HLSL -FfxUInt32x3 ffxAsUInt32(FfxFloat32x3 x) -{ - return asuint(x); -} - -/// Interprets the bit pattern of x as an unsigned integer. -/// -/// @param [in] value The input value. -/// -/// @returns -/// The input interpreted as an unsigned integer. -/// -/// @ingroup HLSL -FfxUInt32x4 ffxAsUInt32(FfxFloat32x4 x) -{ - return asuint(x); -} - -/// Interprets the bit pattern of x as a floating-point number. -/// -/// @param [in] value The input value. -/// -/// @returns -/// The input interpreted as a floating-point number. -/// -/// @ingroup HLSL -FfxFloat32 ffxAsFloat(FfxUInt32 x) -{ - return asfloat(x); -} - -/// Interprets the bit pattern of x as a floating-point number. -/// -/// @param [in] value The input value. -/// -/// @returns -/// The input interpreted as a floating-point number. -/// -/// @ingroup HLSL -FfxFloat32x2 ffxAsFloat(FfxUInt32x2 x) -{ - return asfloat(x); -} - -/// Interprets the bit pattern of x as a floating-point number. -/// -/// @param [in] value The input value. -/// -/// @returns -/// The input interpreted as a floating-point number. -/// -/// @ingroup HLSL -FfxFloat32x3 ffxAsFloat(FfxUInt32x3 x) -{ - return asfloat(x); -} - -/// Interprets the bit pattern of x as a floating-point number. -/// -/// @param [in] value The input value. -/// -/// @returns -/// The input interpreted as a floating-point number. -/// -/// @ingroup HLSL -FfxFloat32x4 ffxAsFloat(FfxUInt32x4 x) -{ - return asfloat(x); -} - -/// Compute the linear interopation between two values. -/// -/// Implemented by calling the HLSL mix instrinsic function. Implements the -/// following math: -/// -/// (1 - t) * x + t * y -/// -/// @param [in] x The first value to lerp between. -/// @param [in] y The second value to lerp between. -/// @param [in] t The value to determine how much of x and how much of y. -/// -/// @returns -/// A linearly interpolated value between x and y according to t. -/// -/// @ingroup HLSL -FfxFloat32 ffxLerp(FfxFloat32 x, FfxFloat32 y, FfxFloat32 t) -{ - return lerp(x, y, t); -} - -/// Compute the linear interopation between two values. -/// -/// Implemented by calling the HLSL mix instrinsic function. Implements the -/// following math: -/// -/// (1 - t) * x + t * y -/// -/// @param [in] x The first value to lerp between. -/// @param [in] y The second value to lerp between. -/// @param [in] t The value to determine how much of x and how much of y. -/// -/// @returns -/// A linearly interpolated value between x and y according to t. -/// -/// @ingroup HLSL -FfxFloat32x2 ffxLerp(FfxFloat32x2 x, FfxFloat32x2 y, FfxFloat32 t) -{ - return lerp(x, y, t); -} - -/// Compute the linear interopation between two values. -/// -/// Implemented by calling the HLSL mix instrinsic function. Implements the -/// following math: -/// -/// (1 - t) * x + t * y -/// -/// @param [in] x The first value to lerp between. -/// @param [in] y The second value to lerp between. -/// @param [in] t The value to determine how much of x and how much of y. -/// -/// @returns -/// A linearly interpolated value between x and y according to t. -/// -/// @ingroup HLSL -FfxFloat32x2 ffxLerp(FfxFloat32x2 x, FfxFloat32x2 y, FfxFloat32x2 t) -{ - return lerp(x, y, t); -} - -/// Compute the linear interopation between two values. -/// -/// Implemented by calling the HLSL mix instrinsic function. Implements the -/// following math: -/// -/// (1 - t) * x + t * y -/// -/// @param [in] x The first value to lerp between. -/// @param [in] y The second value to lerp between. -/// @param [in] t The value to determine how much of x and how much of y. -/// -/// @returns -/// A linearly interpolated value between x and y according to t. -/// -/// @ingroup HLSL -FfxFloat32x3 ffxLerp(FfxFloat32x3 x, FfxFloat32x3 y, FfxFloat32 t) -{ - return lerp(x, y, t); -} - -/// Compute the linear interopation between two values. -/// -/// Implemented by calling the HLSL mix instrinsic function. Implements the -/// following math: -/// -/// (1 - t) * x + t * y -/// -/// @param [in] x The first value to lerp between. -/// @param [in] y The second value to lerp between. -/// @param [in] t The value to determine how much of x and how much of y. -/// -/// @returns -/// A linearly interpolated value between x and y according to t. -/// -/// @ingroup HLSL -FfxFloat32x3 ffxLerp(FfxFloat32x3 x, FfxFloat32x3 y, FfxFloat32x3 t) -{ - return lerp(x, y, t); -} - -/// Compute the linear interopation between two values. -/// -/// Implemented by calling the HLSL mix instrinsic function. Implements the -/// following math: -/// -/// (1 - t) * x + t * y -/// -/// @param [in] x The first value to lerp between. -/// @param [in] y The second value to lerp between. -/// @param [in] t The value to determine how much of x and how much of y. -/// -/// @returns -/// A linearly interpolated value between x and y according to t. -/// -/// @ingroup HLSL -FfxFloat32x4 ffxLerp(FfxFloat32x4 x, FfxFloat32x4 y, FfxFloat32 t) -{ - return lerp(x, y, t); -} - -/// Compute the linear interopation between two values. -/// -/// Implemented by calling the HLSL mix instrinsic function. Implements the -/// following math: -/// -/// (1 - t) * x + t * y -/// -/// @param [in] x The first value to lerp between. -/// @param [in] y The second value to lerp between. -/// @param [in] t The value to determine how much of x and how much of y. -/// -/// @returns -/// A linearly interpolated value between x and y according to t. -/// -/// @ingroup HLSL -FfxFloat32x4 ffxLerp(FfxFloat32x4 x, FfxFloat32x4 y, FfxFloat32x4 t) -{ - return lerp(x, y, t); -} - -/// Clamp a value to a [0..1] range. -/// -/// @param [in] x The value to clamp to [0..1] range. -/// -/// @returns -/// The clamped version of x. -/// -/// @ingroup HLSL -FfxFloat32 ffxSaturate(FfxFloat32 x) -{ - return saturate(x); -} - -/// Clamp a value to a [0..1] range. -/// -/// @param [in] x The value to clamp to [0..1] range. -/// -/// @returns -/// The clamped version of x. -/// -/// @ingroup HLSL -FfxFloat32x2 ffxSaturate(FfxFloat32x2 x) -{ - return saturate(x); -} - -/// Clamp a value to a [0..1] range. -/// -/// @param [in] x The value to clamp to [0..1] range. -/// -/// @returns -/// The clamped version of x. -/// -/// @ingroup HLSL -FfxFloat32x3 ffxSaturate(FfxFloat32x3 x) -{ - return saturate(x); -} - -/// Clamp a value to a [0..1] range. -/// -/// @param [in] x The value to clamp to [0..1] range. -/// -/// @returns -/// The clamped version of x. -/// -/// @ingroup HLSL -FfxFloat32x4 ffxSaturate(FfxFloat32x4 x) -{ - return saturate(x); -} - -/// Compute the factional part of a decimal value. -/// -/// This function calculates x - floor(x). Where floor is the intrinsic HLSL function. -/// -/// NOTE: This function should compile down to a single V_MAX3_F32 operation on GCN/RDNA hardware. It is -/// worth further noting that this function is intentionally distinct from the HLSL frac intrinsic -/// function. -/// -/// @param [in] x The value to compute the fractional part from. -/// -/// @returns -/// The fractional part of x. -/// -/// @ingroup HLSL -FfxFloat32 ffxFract(FfxFloat32 x) -{ - return x - floor(x); -} - -/// Compute the factional part of a decimal value. -/// -/// This function calculates x - floor(x). Where floor is the intrinsic HLSL function. -/// -/// NOTE: This function should compile down to a single V_MAX3_F32 operation on GCN/RDNA hardware. It is -/// worth further noting that this function is intentionally distinct from the HLSL frac intrinsic -/// function. -/// -/// @param [in] x The value to compute the fractional part from. -/// -/// @returns -/// The fractional part of x. -/// -/// @ingroup HLSL -FfxFloat32x2 ffxFract(FfxFloat32x2 x) -{ - return x - floor(x); -} - -/// Compute the factional part of a decimal value. -/// -/// This function calculates x - floor(x). Where floor is the intrinsic HLSL function. -/// -/// NOTE: This function should compile down to a single V_MAX3_F32 operation on GCN/RDNA hardware. It is -/// worth further noting that this function is intentionally distinct from the HLSL frac intrinsic -/// function. -/// -/// @param [in] x The value to compute the fractional part from. -/// -/// @returns -/// The fractional part of x. -/// -/// @ingroup HLSL -FfxFloat32x3 ffxFract(FfxFloat32x3 x) -{ - return x - floor(x); -} - -/// Compute the factional part of a decimal value. -/// -/// This function calculates x - floor(x). Where floor is the intrinsic HLSL function. -/// -/// NOTE: This function should compile down to a single V_MAX3_F32 operation on GCN/RDNA hardware. It is -/// worth further noting that this function is intentionally distinct from the HLSL frac intrinsic -/// function. -/// -/// @param [in] x The value to compute the fractional part from. -/// -/// @returns -/// The fractional part of x. -/// -/// @ingroup HLSL -FfxFloat32x4 ffxFract(FfxFloat32x4 x) -{ - return x - floor(x); -} - -/// Compute the maximum of three values. -/// -/// NOTE: This function should compile down to a single V_MAX3_F32 operation on GCN/RDNA hardware. -/// -/// @param [in] x The first value to include in the max calculation. -/// @param [in] y The second value to include in the max calcuation. -/// @param [in] z The third value to include in the max calcuation. -/// -/// @returns -/// The maximum value of x, y, and z. -/// -/// @ingroup HLSL -FfxFloat32 ffxMax3(FfxFloat32 x, FfxFloat32 y, FfxFloat32 z) -{ - return max(x, max(y, z)); -} - -/// Compute the maximum of three values. -/// -/// NOTE: This function should compile down to a single V_MAX3_F32 operation on GCN/RDNA hardware. -/// -/// @param [in] x The first value to include in the max calculation. -/// @param [in] y The second value to include in the max calcuation. -/// @param [in] z The third value to include in the max calcuation. -/// -/// @returns -/// The maximum value of x, y, and z. -/// -/// @ingroup HLSL -FfxFloat32x2 ffxMax3(FfxFloat32x2 x, FfxFloat32x2 y, FfxFloat32x2 z) -{ - return max(x, max(y, z)); -} - -/// Compute the maximum of three values. -/// -/// NOTE: This function should compile down to a single V_MAX3_F32 operation on GCN/RDNA hardware. -/// -/// @param [in] x The first value to include in the max calculation. -/// @param [in] y The second value to include in the max calcuation. -/// @param [in] z The third value to include in the max calcuation. -/// -/// @returns -/// The maximum value of x, y, and z. -/// -/// @ingroup HLSL -FfxFloat32x3 ffxMax3(FfxFloat32x3 x, FfxFloat32x3 y, FfxFloat32x3 z) -{ - return max(x, max(y, z)); -} - -/// Compute the maximum of three values. -/// -/// NOTE: This function should compile down to a single V_MAX3_F32 operation on GCN/RDNA hardware. -/// -/// @param [in] x The first value to include in the max calculation. -/// @param [in] y The second value to include in the max calcuation. -/// @param [in] z The third value to include in the max calcuation. -/// -/// @returns -/// The maximum value of x, y, and z. -/// -/// @ingroup HLSL -FfxFloat32x4 ffxMax3(FfxFloat32x4 x, FfxFloat32x4 y, FfxFloat32x4 z) -{ - return max(x, max(y, z)); -} - -/// Compute the maximum of three values. -/// -/// NOTE: This function should compile down to a single V_MAX3_F32 operation on GCN/RDNA hardware. -/// -/// @param [in] x The first value to include in the max calculation. -/// @param [in] y The second value to include in the max calcuation. -/// @param [in] z The third value to include in the max calcuation. -/// -/// @returns -/// The maximum value of x, y, and z. -/// -/// @ingroup HLSL -FfxUInt32 ffxMax3(FfxUInt32 x, FfxUInt32 y, FfxUInt32 z) -{ - return max(x, max(y, z)); -} - -/// Compute the maximum of three values. -/// -/// NOTE: This function should compile down to a single V_MAX3_F32 operation on GCN/RDNA hardware. -/// -/// @param [in] x The first value to include in the max calculation. -/// @param [in] y The second value to include in the max calcuation. -/// @param [in] z The third value to include in the max calcuation. -/// -/// @returns -/// The maximum value of x, y, and z. -/// -/// @ingroup HLSL -FfxUInt32x2 ffxMax3(FfxUInt32x2 x, FfxUInt32x2 y, FfxUInt32x2 z) -{ - return max(x, max(y, z)); -} - -/// Compute the maximum of three values. -/// -/// NOTE: This function should compile down to a single V_MAX3_F32 operation on GCN/RDNA hardware. -/// -/// @param [in] x The first value to include in the max calculation. -/// @param [in] y The second value to include in the max calcuation. -/// @param [in] z The third value to include in the max calcuation. -/// -/// @returns -/// The maximum value of x, y, and z. -/// -/// @ingroup HLSL -FfxUInt32x3 ffxMax3(FfxUInt32x3 x, FfxUInt32x3 y, FfxUInt32x3 z) -{ - return max(x, max(y, z)); -} - -/// Compute the maximum of three values. -/// -/// NOTE: This function should compile down to a single V_MAX3_F32 operation on GCN/RDNA hardware. -/// -/// @param [in] x The first value to include in the max calculation. -/// @param [in] y The second value to include in the max calcuation. -/// @param [in] z The third value to include in the max calcuation. -/// -/// @returns -/// The maximum value of x, y, and z. -/// -/// @ingroup HLSL -FfxUInt32x4 ffxMax3(FfxUInt32x4 x, FfxUInt32x4 y, FfxUInt32x4 z) -{ - return max(x, max(y, z)); -} - -/// Compute the median of three values. -/// -/// NOTE: This function should compile down to a single V_MED3_F32 operation on GCN/RDNA hardware. -/// -/// @param [in] x The first value to include in the median calculation. -/// @param [in] y The second value to include in the median calcuation. -/// @param [in] z The third value to include in the median calcuation. -/// -/// @returns -/// The median value of x, y, and z. -/// -/// @ingroup HLSL -FfxFloat32 ffxMed3(FfxFloat32 x, FfxFloat32 y, FfxFloat32 z) -{ - return max(min(x, y), min(max(x, y), z)); -} - -/// Compute the median of three values. -/// -/// NOTE: This function should compile down to a single V_MED3_F32 operation on GCN/RDNA hardware. -/// -/// @param [in] x The first value to include in the median calculation. -/// @param [in] y The second value to include in the median calcuation. -/// @param [in] z The third value to include in the median calcuation. -/// -/// @returns -/// The median value of x, y, and z. -/// -/// @ingroup HLSL -FfxFloat32x2 ffxMed3(FfxFloat32x2 x, FfxFloat32x2 y, FfxFloat32x2 z) -{ - return max(min(x, y), min(max(x, y), z)); -} - -/// Compute the median of three values. -/// -/// NOTE: This function should compile down to a single V_MED3_F32 operation on GCN/RDNA hardware. -/// -/// @param [in] x The first value to include in the median calculation. -/// @param [in] y The second value to include in the median calcuation. -/// @param [in] z The third value to include in the median calcuation. -/// -/// @returns -/// The median value of x, y, and z. -/// -/// @ingroup HLSL -FfxFloat32x3 ffxMed3(FfxFloat32x3 x, FfxFloat32x3 y, FfxFloat32x3 z) -{ - return max(min(x, y), min(max(x, y), z)); -} - -/// Compute the median of three values. -/// -/// NOTE: This function should compile down to a single V_MED3_F32 operation on GCN/RDNA hardware. -/// -/// @param [in] x The first value to include in the median calculation. -/// @param [in] y The second value to include in the median calcuation. -/// @param [in] z The third value to include in the median calcuation. -/// -/// @returns -/// The median value of x, y, and z. -/// -/// @ingroup HLSL -FfxFloat32x4 ffxMed3(FfxFloat32x4 x, FfxFloat32x4 y, FfxFloat32x4 z) -{ - return max(min(x, y), min(max(x, y), z)); -} - -/// Compute the median of three values. -/// -/// NOTE: This function should compile down to a single V_MED3_F32 operation on GCN/RDNA hardware. -/// -/// @param [in] x The first value to include in the median calculation. -/// @param [in] y The second value to include in the median calcuation. -/// @param [in] z The third value to include in the median calcuation. -/// -/// @returns -/// The median value of x, y, and z. -/// -/// @ingroup HLSL -FfxInt32 ffxMed3(FfxInt32 x, FfxInt32 y, FfxInt32 z) -{ - return max(min(x, y), min(max(x, y), z)); - // return min(max(min(y, z), x), max(y, z)); - // return max(max(x, y), z) == x ? max(y, z) : (max(max(x, y), z) == y ? max(x, z) : max(x, y)); -} - -/// Compute the median of three values. -/// -/// NOTE: This function should compile down to a single V_MED3_F32 operation on GCN/RDNA hardware. -/// -/// @param [in] x The first value to include in the median calculation. -/// @param [in] y The second value to include in the median calcuation. -/// @param [in] z The third value to include in the median calcuation. -/// -/// @returns -/// The median value of x, y, and z. -/// -/// @ingroup HLSL -FfxInt32x2 ffxMed3(FfxInt32x2 x, FfxInt32x2 y, FfxInt32x2 z) -{ - return max(min(x, y), min(max(x, y), z)); - // return min(max(min(y, z), x), max(y, z)); - // return max(max(x, y), z) == x ? max(y, z) : (max(max(x, y), z) == y ? max(x, z) : max(x, y)); -} - -/// Compute the median of three values. -/// -/// NOTE: This function should compile down to a single V_MED3_F32 operation on GCN/RDNA hardware. -/// -/// @param [in] x The first value to include in the median calculation. -/// @param [in] y The second value to include in the median calcuation. -/// @param [in] z The third value to include in the median calcuation. -/// -/// @returns -/// The median value of x, y, and z. -/// -/// @ingroup HLSL -FfxInt32x3 ffxMed3(FfxInt32x3 x, FfxInt32x3 y, FfxInt32x3 z) -{ - return max(min(x, y), min(max(x, y), z)); -} - -/// Compute the median of three values. -/// -/// NOTE: This function should compile down to a single V_MED3_I32 operation on GCN/RDNA hardware. -/// -/// @param [in] x The first value to include in the median calculation. -/// @param [in] y The second value to include in the median calcuation. -/// @param [in] z The third value to include in the median calcuation. -/// -/// @returns -/// The median value of x, y, and z. -/// -/// @ingroup HLSL -FfxInt32x4 ffxMed3(FfxInt32x4 x, FfxInt32x4 y, FfxInt32x4 z) -{ - return max(min(x, y), min(max(x, y), z)); -} - -/// Compute the minimum of three values. -/// -/// NOTE: This function should compile down to a single V_MIN3_I32 operation on GCN/RDNA hardware. -/// -/// @param [in] x The first value to include in the min calculation. -/// @param [in] y The second value to include in the min calcuation. -/// @param [in] z The third value to include in the min calcuation. -/// -/// @returns -/// The minimum value of x, y, and z. -/// -/// @ingroup HLSL -FfxFloat32 ffxMin3(FfxFloat32 x, FfxFloat32 y, FfxFloat32 z) -{ - return min(x, min(y, z)); -} - -/// Compute the minimum of three values. -/// -/// NOTE: This function should compile down to a single V_MIN3_I32 operation on GCN/RDNA hardware. -/// -/// @param [in] x The first value to include in the min calculation. -/// @param [in] y The second value to include in the min calcuation. -/// @param [in] z The third value to include in the min calcuation. -/// -/// @returns -/// The minimum value of x, y, and z. -/// -/// @ingroup HLSL -FfxFloat32x2 ffxMin3(FfxFloat32x2 x, FfxFloat32x2 y, FfxFloat32x2 z) -{ - return min(x, min(y, z)); -} - -/// Compute the minimum of three values. -/// -/// NOTE: This function should compile down to a single V_MIN3_I32 operation on GCN/RDNA hardware. -/// -/// @param [in] x The first value to include in the min calculation. -/// @param [in] y The second value to include in the min calcuation. -/// @param [in] z The third value to include in the min calcuation. -/// -/// @returns -/// The minimum value of x, y, and z. -/// -/// @ingroup HLSL -FfxFloat32x3 ffxMin3(FfxFloat32x3 x, FfxFloat32x3 y, FfxFloat32x3 z) -{ - return min(x, min(y, z)); -} - -/// Compute the minimum of three values. -/// -/// NOTE: This function should compile down to a single V_MIN3_F32 operation on GCN/RDNA hardware. -/// -/// @param [in] x The first value to include in the min calculation. -/// @param [in] y The second value to include in the min calcuation. -/// @param [in] z The third value to include in the min calcuation. -/// -/// @returns -/// The minimum value of x, y, and z. -/// -/// @ingroup HLSL -FfxFloat32x4 ffxMin3(FfxFloat32x4 x, FfxFloat32x4 y, FfxFloat32x4 z) -{ - return min(x, min(y, z)); -} - -/// Compute the minimum of three values. -/// -/// NOTE: This function should compile down to a single V_MIN3_F32 operation on GCN/RDNA hardware. -/// -/// @param [in] x The first value to include in the min calculation. -/// @param [in] y The second value to include in the min calcuation. -/// @param [in] z The third value to include in the min calcuation. -/// -/// @returns -/// The minimum value of x, y, and z. -/// -/// @ingroup HLSL -FfxUInt32 ffxMin3(FfxUInt32 x, FfxUInt32 y, FfxUInt32 z) -{ - return min(x, min(y, z)); -} - -/// Compute the minimum of three values. -/// -/// NOTE: This function should compile down to a single V_MIN3_F32 operation on GCN/RDNA hardware. -/// -/// @param [in] x The first value to include in the min calculation. -/// @param [in] y The second value to include in the min calcuation. -/// @param [in] z The third value to include in the min calcuation. -/// -/// @returns -/// The minimum value of x, y, and z. -/// -/// @ingroup HLSL -FfxUInt32x2 ffxMin3(FfxUInt32x2 x, FfxUInt32x2 y, FfxUInt32x2 z) -{ - return min(x, min(y, z)); -} - -/// Compute the minimum of three values. -/// -/// NOTE: This function should compile down to a single V_MIN3_F32 operation on GCN/RDNA hardware. -/// -/// @param [in] x The first value to include in the min calculation. -/// @param [in] y The second value to include in the min calcuation. -/// @param [in] z The third value to include in the min calcuation. -/// -/// @returns -/// The minimum value of x, y, and z. -/// -/// @ingroup HLSL -FfxUInt32x3 ffxMin3(FfxUInt32x3 x, FfxUInt32x3 y, FfxUInt32x3 z) -{ - return min(x, min(y, z)); -} - -/// Compute the minimum of three values. -/// -/// NOTE: This function should compile down to a single V_MIN3_F32 operation on GCN/RDNA hardware. -/// -/// @param [in] x The first value to include in the min calculation. -/// @param [in] y The second value to include in the min calcuation. -/// @param [in] z The third value to include in the min calcuation. -/// -/// @returns -/// The minimum value of x, y, and z. -/// -/// @ingroup HLSL -FfxUInt32x4 ffxMin3(FfxUInt32x4 x, FfxUInt32x4 y, FfxUInt32x4 z) -{ - return min(x, min(y, z)); -} - - -FfxUInt32 AShrSU1(FfxUInt32 a, FfxUInt32 b) -{ - return FfxUInt32(FfxInt32(a) >> FfxInt32(b)); -} - -//============================================================================================================================== -// HLSL HALF -//============================================================================================================================== -#if FFX_HALF - -//============================================================================================================================== -// Need to use manual unpack to get optimal execution (don't use packed types in buffers directly). -// Unpack requires this pattern: https://gpuopen.com/first-steps-implementing-fp16/ -FFX_MIN16_F2 ffxUint32ToFloat16x2(FfxUInt32 x) -{ - FfxFloat32x2 t = f16tof32(FfxUInt32x2(x & 0xFFFF, x >> 16)); - return FFX_MIN16_F2(t); -} -FFX_MIN16_F4 ffxUint32x2ToFloat16x4(FfxUInt32x2 x) -{ - return FFX_MIN16_F4(ffxUint32ToFloat16x2(x.x), ffxUint32ToFloat16x2(x.y)); -} -FFX_MIN16_U2 ffxUint32ToUint16x2(FfxUInt32 x) -{ - FfxUInt32x2 t = FfxUInt32x2(x & 0xFFFF, x >> 16); - return FFX_MIN16_U2(t); -} -FFX_MIN16_U4 ffxUint32x2ToUint16x4(FfxUInt32x2 x) -{ - return FFX_MIN16_U4(ffxUint32ToUint16x2(x.x), ffxUint32ToUint16x2(x.y)); -} -#define FFX_UINT32_TO_FLOAT16X2(x) ffxUint32ToFloat16x2(FfxUInt32(x)) -#define FFX_UINT32X2_TO_FLOAT16X4(x) ffxUint32x2ToFloat16x4(FfxUInt32x2(x)) -#define FFX_UINT32_TO_UINT16X2(x) ffxUint32ToUint16x2(FfxUInt32(x)) -#define FFX_UINT32X2_TO_UINT16X4(x) ffxUint32x2ToUint16x4(FfxUInt32x2(x)) -//------------------------------------------------------------------------------------------------------------------------------ -FfxUInt32 FFX_MIN16_F2ToUint32(FFX_MIN16_F2 x) -{ - return f32tof16(x.x) + (f32tof16(x.y) << 16); -} -FfxUInt32x2 FFX_MIN16_F4ToUint32x2(FFX_MIN16_F4 x) -{ - return FfxUInt32x2(FFX_MIN16_F2ToUint32(x.xy), FFX_MIN16_F2ToUint32(x.zw)); -} -FfxUInt32 FFX_MIN16_U2ToUint32(FFX_MIN16_U2 x) -{ - return FfxUInt32(x.x) + (FfxUInt32(x.y) << 16); -} -FfxUInt32x2 FFX_MIN16_U4ToUint32x2(FFX_MIN16_U4 x) -{ - return FfxUInt32x2(FFX_MIN16_U2ToUint32(x.xy), FFX_MIN16_U2ToUint32(x.zw)); -} -#define FFX_FLOAT16X2_TO_UINT32(x) FFX_MIN16_F2ToUint32(FFX_MIN16_F2(x)) -#define FFX_FLOAT16X4_TO_UINT32X2(x) FFX_MIN16_F4ToUint32x2(FFX_MIN16_F4(x)) -#define FFX_UINT16X2_TO_UINT32(x) FFX_MIN16_U2ToUint32(FFX_MIN16_U2(x)) -#define FFX_UINT16X4_TO_UINT32X2(x) FFX_MIN16_U4ToUint32x2(FFX_MIN16_U4(x)) - -#if defined(FFX_HLSL_6_2) && !defined(FFX_NO_16_BIT_CAST) -#define FFX_TO_UINT16(x) asuint16(x) -#define FFX_TO_UINT16X2(x) asuint16(x) -#define FFX_TO_UINT16X3(x) asuint16(x) -#define FFX_TO_UINT16X4(x) asuint16(x) -#else -#define FFX_TO_UINT16(a) FFX_MIN16_U(f32tof16(FfxFloat32(a))) -#define FFX_TO_UINT16X2(a) FFX_MIN16_U2(FFX_TO_UINT16((a).x), FFX_TO_UINT16((a).y)) -#define FFX_TO_UINT16X3(a) FFX_MIN16_U3(FFX_TO_UINT16((a).x), FFX_TO_UINT16((a).y), FFX_TO_UINT16((a).z)) -#define FFX_TO_UINT16X4(a) FFX_MIN16_U4(FFX_TO_UINT16((a).x), FFX_TO_UINT16((a).y), FFX_TO_UINT16((a).z), FFX_TO_UINT16((a).w)) -#endif // #if defined(FFX_HLSL_6_2) && !defined(FFX_NO_16_BIT_CAST) - -#if defined(FFX_HLSL_6_2) && !defined(FFX_NO_16_BIT_CAST) -#define FFX_TO_FLOAT16(x) asfloat16(x) -#define FFX_TO_FLOAT16X2(x) asfloat16(x) -#define FFX_TO_FLOAT16X3(x) asfloat16(x) -#define FFX_TO_FLOAT16X4(x) asfloat16(x) -#else -#define FFX_TO_FLOAT16(a) FFX_MIN16_F(f16tof32(FfxUInt32(a))) -#define FFX_TO_FLOAT16X2(a) FFX_MIN16_F2(FFX_TO_FLOAT16((a).x), FFX_TO_FLOAT16((a).y)) -#define FFX_TO_FLOAT16X3(a) FFX_MIN16_F3(FFX_TO_FLOAT16((a).x), FFX_TO_FLOAT16((a).y), FFX_TO_FLOAT16((a).z)) -#define FFX_TO_FLOAT16X4(a) FFX_MIN16_F4(FFX_TO_FLOAT16((a).x), FFX_TO_FLOAT16((a).y), FFX_TO_FLOAT16((a).z), FFX_TO_FLOAT16((a).w)) -#endif // #if defined(FFX_HLSL_6_2) && !defined(FFX_NO_16_BIT_CAST) - -//============================================================================================================================== -#define FFX_BROADCAST_FLOAT16(a) FFX_MIN16_F(a) -#define FFX_BROADCAST_FLOAT16X2(a) FFX_MIN16_F(a) -#define FFX_BROADCAST_FLOAT16X3(a) FFX_MIN16_F(a) -#define FFX_BROADCAST_FLOAT16X4(a) FFX_MIN16_F(a) - -//------------------------------------------------------------------------------------------------------------------------------ -#define FFX_BROADCAST_INT16(a) FFX_MIN16_I(a) -#define FFX_BROADCAST_INT16X2(a) FFX_MIN16_I(a) -#define FFX_BROADCAST_INT16X3(a) FFX_MIN16_I(a) -#define FFX_BROADCAST_INT16X4(a) FFX_MIN16_I(a) - -//------------------------------------------------------------------------------------------------------------------------------ -#define FFX_BROADCAST_UINT16(a) FFX_MIN16_U(a) -#define FFX_BROADCAST_UINT16X2(a) FFX_MIN16_U(a) -#define FFX_BROADCAST_UINT16X3(a) FFX_MIN16_U(a) -#define FFX_BROADCAST_UINT16X4(a) FFX_MIN16_U(a) - -//============================================================================================================================== -FFX_MIN16_U ffxAbsHalf(FFX_MIN16_U a) -{ - return FFX_MIN16_U(abs(FFX_MIN16_I(a))); -} -FFX_MIN16_U2 ffxAbsHalf(FFX_MIN16_U2 a) -{ - return FFX_MIN16_U2(abs(FFX_MIN16_I2(a))); -} -FFX_MIN16_U3 ffxAbsHalf(FFX_MIN16_U3 a) -{ - return FFX_MIN16_U3(abs(FFX_MIN16_I3(a))); -} -FFX_MIN16_U4 ffxAbsHalf(FFX_MIN16_U4 a) -{ - return FFX_MIN16_U4(abs(FFX_MIN16_I4(a))); -} -//------------------------------------------------------------------------------------------------------------------------------ -FFX_MIN16_F ffxClampHalf(FFX_MIN16_F x, FFX_MIN16_F n, FFX_MIN16_F m) -{ - return max(n, min(x, m)); -} -FFX_MIN16_F2 ffxClampHalf(FFX_MIN16_F2 x, FFX_MIN16_F2 n, FFX_MIN16_F2 m) -{ - return max(n, min(x, m)); -} -FFX_MIN16_F3 ffxClampHalf(FFX_MIN16_F3 x, FFX_MIN16_F3 n, FFX_MIN16_F3 m) -{ - return max(n, min(x, m)); -} -FFX_MIN16_F4 ffxClampHalf(FFX_MIN16_F4 x, FFX_MIN16_F4 n, FFX_MIN16_F4 m) -{ - return max(n, min(x, m)); -} -//------------------------------------------------------------------------------------------------------------------------------ -// V_FRACT_F16 (note DX frac() is different). -FFX_MIN16_F ffxFract(FFX_MIN16_F x) -{ - return x - floor(x); -} -FFX_MIN16_F2 ffxFract(FFX_MIN16_F2 x) -{ - return x - floor(x); -} -FFX_MIN16_F3 ffxFract(FFX_MIN16_F3 x) -{ - return x - floor(x); -} -FFX_MIN16_F4 ffxFract(FFX_MIN16_F4 x) -{ - return x - floor(x); -} -//------------------------------------------------------------------------------------------------------------------------------ -FFX_MIN16_F ffxLerp(FFX_MIN16_F x, FFX_MIN16_F y, FFX_MIN16_F a) -{ - return lerp(x, y, a); -} -FFX_MIN16_F2 ffxLerp(FFX_MIN16_F2 x, FFX_MIN16_F2 y, FFX_MIN16_F a) -{ - return lerp(x, y, a); -} -FFX_MIN16_F2 ffxLerp(FFX_MIN16_F2 x, FFX_MIN16_F2 y, FFX_MIN16_F2 a) -{ - return lerp(x, y, a); -} -FFX_MIN16_F3 ffxLerp(FFX_MIN16_F3 x, FFX_MIN16_F3 y, FFX_MIN16_F a) -{ - return lerp(x, y, a); -} -FFX_MIN16_F3 ffxLerp(FFX_MIN16_F3 x, FFX_MIN16_F3 y, FFX_MIN16_F3 a) -{ - return lerp(x, y, a); -} -FFX_MIN16_F4 ffxLerp(FFX_MIN16_F4 x, FFX_MIN16_F4 y, FFX_MIN16_F a) -{ - return lerp(x, y, a); -} -FFX_MIN16_F4 ffxLerp(FFX_MIN16_F4 x, FFX_MIN16_F4 y, FFX_MIN16_F4 a) -{ - return lerp(x, y, a); -} -//------------------------------------------------------------------------------------------------------------------------------ -FFX_MIN16_F ffxMax3Half(FFX_MIN16_F x, FFX_MIN16_F y, FFX_MIN16_F z) -{ - return max(x, max(y, z)); -} -FFX_MIN16_F2 ffxMax3Half(FFX_MIN16_F2 x, FFX_MIN16_F2 y, FFX_MIN16_F2 z) -{ - return max(x, max(y, z)); -} -FFX_MIN16_F3 ffxMax3Half(FFX_MIN16_F3 x, FFX_MIN16_F3 y, FFX_MIN16_F3 z) -{ - return max(x, max(y, z)); -} -FFX_MIN16_F4 ffxMax3Half(FFX_MIN16_F4 x, FFX_MIN16_F4 y, FFX_MIN16_F4 z) -{ - return max(x, max(y, z)); -} -//------------------------------------------------------------------------------------------------------------------------------ -FFX_MIN16_F ffxMin3Half(FFX_MIN16_F x, FFX_MIN16_F y, FFX_MIN16_F z) -{ - return min(x, min(y, z)); -} -FFX_MIN16_F2 ffxMin3Half(FFX_MIN16_F2 x, FFX_MIN16_F2 y, FFX_MIN16_F2 z) -{ - return min(x, min(y, z)); -} -FFX_MIN16_F3 ffxMin3Half(FFX_MIN16_F3 x, FFX_MIN16_F3 y, FFX_MIN16_F3 z) -{ - return min(x, min(y, z)); -} -FFX_MIN16_F4 ffxMin3Half(FFX_MIN16_F4 x, FFX_MIN16_F4 y, FFX_MIN16_F4 z) -{ - return min(x, min(y, z)); -} -//------------------------------------------------------------------------------------------------------------------------------ -FFX_MIN16_F ffxMed3Half(FFX_MIN16_F x, FFX_MIN16_F y, FFX_MIN16_F z) -{ - return max(min(x, y), min(max(x, y), z)); -} -FFX_MIN16_F2 ffxMed3Half(FFX_MIN16_F2 x, FFX_MIN16_F2 y, FFX_MIN16_F2 z) -{ - return max(min(x, y), min(max(x, y), z)); -} -FFX_MIN16_F3 ffxMed3Half(FFX_MIN16_F3 x, FFX_MIN16_F3 y, FFX_MIN16_F3 z) -{ - return max(min(x, y), min(max(x, y), z)); -} -FFX_MIN16_F4 ffxMed3Half(FFX_MIN16_F4 x, FFX_MIN16_F4 y, FFX_MIN16_F4 z) -{ - return max(min(x, y), min(max(x, y), z)); -} -//------------------------------------------------------------------------------------------------------------------------------ -FFX_MIN16_I ffxMed3Half(FFX_MIN16_I x, FFX_MIN16_I y, FFX_MIN16_I z) -{ - return max(min(x, y), min(max(x, y), z)); -} -FFX_MIN16_I2 ffxMed3Half(FFX_MIN16_I2 x, FFX_MIN16_I2 y, FFX_MIN16_I2 z) -{ - return max(min(x, y), min(max(x, y), z)); -} -FFX_MIN16_I3 ffxMed3Half(FFX_MIN16_I3 x, FFX_MIN16_I3 y, FFX_MIN16_I3 z) -{ - return max(min(x, y), min(max(x, y), z)); -} -FFX_MIN16_I4 ffxMed3Half(FFX_MIN16_I4 x, FFX_MIN16_I4 y, FFX_MIN16_I4 z) -{ - return max(min(x, y), min(max(x, y), z)); -} -//------------------------------------------------------------------------------------------------------------------------------ -FFX_MIN16_F ffxReciprocalHalf(FFX_MIN16_F x) -{ - return rcp(x); -} -FFX_MIN16_F2 ffxReciprocalHalf(FFX_MIN16_F2 x) -{ - return rcp(x); -} -FFX_MIN16_F3 ffxReciprocalHalf(FFX_MIN16_F3 x) -{ - return rcp(x); -} -FFX_MIN16_F4 ffxReciprocalHalf(FFX_MIN16_F4 x) -{ - return rcp(x); -} -//------------------------------------------------------------------------------------------------------------------------------ -FFX_MIN16_F ffxReciprocalSquareRootHalf(FFX_MIN16_F x) -{ - return rsqrt(x); -} -FFX_MIN16_F2 ffxReciprocalSquareRootHalf(FFX_MIN16_F2 x) -{ - return rsqrt(x); -} -FFX_MIN16_F3 ffxReciprocalSquareRootHalf(FFX_MIN16_F3 x) -{ - return rsqrt(x); -} -FFX_MIN16_F4 ffxReciprocalSquareRootHalf(FFX_MIN16_F4 x) -{ - return rsqrt(x); -} -//------------------------------------------------------------------------------------------------------------------------------ -FFX_MIN16_F ffxSaturate(FFX_MIN16_F x) -{ - return saturate(x); -} -FFX_MIN16_F2 ffxSaturate(FFX_MIN16_F2 x) -{ - return saturate(x); -} -FFX_MIN16_F3 ffxSaturate(FFX_MIN16_F3 x) -{ - return saturate(x); -} -FFX_MIN16_F4 ffxSaturate(FFX_MIN16_F4 x) -{ - return saturate(x); -} -//------------------------------------------------------------------------------------------------------------------------------ -FFX_MIN16_U ffxBitShiftRightHalf(FFX_MIN16_U a, FFX_MIN16_U b) -{ - return FFX_MIN16_U(FFX_MIN16_I(a) >> FFX_MIN16_I(b)); -} -FFX_MIN16_U2 ffxBitShiftRightHalf(FFX_MIN16_U2 a, FFX_MIN16_U2 b) -{ - return FFX_MIN16_U2(FFX_MIN16_I2(a) >> FFX_MIN16_I2(b)); -} -FFX_MIN16_U3 ffxBitShiftRightHalf(FFX_MIN16_U3 a, FFX_MIN16_U3 b) -{ - return FFX_MIN16_U3(FFX_MIN16_I3(a) >> FFX_MIN16_I3(b)); -} -FFX_MIN16_U4 ffxBitShiftRightHalf(FFX_MIN16_U4 a, FFX_MIN16_U4 b) -{ - return FFX_MIN16_U4(FFX_MIN16_I4(a) >> FFX_MIN16_I4(b)); -} -#endif // FFX_HALF - -//============================================================================================================================== -// HLSL WAVE -//============================================================================================================================== -#if defined(FFX_WAVE) -// Where 'x' must be a compile time literal. -FfxFloat32 AWaveXorF1(FfxFloat32 v, FfxUInt32 x) -{ - return WaveReadLaneAt(v, WaveGetLaneIndex() ^ x); -} -FfxFloat32x2 AWaveXorF2(FfxFloat32x2 v, FfxUInt32 x) -{ - return WaveReadLaneAt(v, WaveGetLaneIndex() ^ x); -} -FfxFloat32x3 AWaveXorF3(FfxFloat32x3 v, FfxUInt32 x) -{ - return WaveReadLaneAt(v, WaveGetLaneIndex() ^ x); -} -FfxFloat32x4 AWaveXorF4(FfxFloat32x4 v, FfxUInt32 x) -{ - return WaveReadLaneAt(v, WaveGetLaneIndex() ^ x); -} -FfxUInt32 AWaveXorU1(FfxUInt32 v, FfxUInt32 x) -{ - return WaveReadLaneAt(v, WaveGetLaneIndex() ^ x); -} -FfxUInt32x2 AWaveXorU1(FfxUInt32x2 v, FfxUInt32 x) -{ - return WaveReadLaneAt(v, WaveGetLaneIndex() ^ x); -} -FfxUInt32x3 AWaveXorU1(FfxUInt32x3 v, FfxUInt32 x) -{ - return WaveReadLaneAt(v, WaveGetLaneIndex() ^ x); -} -FfxUInt32x4 AWaveXorU1(FfxUInt32x4 v, FfxUInt32 x) -{ - return WaveReadLaneAt(v, WaveGetLaneIndex() ^ x); -} - -#if FFX_HALF -FfxFloat16x2 ffxWaveXorFloat16x2(FfxFloat16x2 v, FfxUInt32 x) -{ - return FFX_UINT32_TO_FLOAT16X2(WaveReadLaneAt(FFX_FLOAT16X2_TO_UINT32(v), WaveGetLaneIndex() ^ x)); -} -FfxFloat16x4 ffxWaveXorFloat16x4(FfxFloat16x4 v, FfxUInt32 x) -{ - return FFX_UINT32X2_TO_FLOAT16X4(WaveReadLaneAt(FFX_FLOAT16X4_TO_UINT32X2(v), WaveGetLaneIndex() ^ x)); -} -FfxUInt16x2 ffxWaveXorUint16x2(FfxUInt16x2 v, FfxUInt32 x) -{ - return FFX_UINT32_TO_UINT16X2(WaveReadLaneAt(FFX_UINT16X2_TO_UINT32(v), WaveGetLaneIndex() ^ x)); -} -FfxUInt16x4 ffxWaveXorUint16x4(FfxUInt16x4 v, FfxUInt32 x) -{ - return AW4_FFX_UINT32(WaveReadLaneAt(FFX_UINT32_AW4(v), WaveGetLaneIndex() ^ x)); -} -#endif // FFX_HALF -#endif // #if defined(FFX_WAVE) diff --git a/Assets/Resources/FSR2/shaders/ffx_core_hlsl.h.meta b/Assets/Resources/FSR2/shaders/ffx_core_hlsl.h.meta deleted file mode 100644 index 5a91c43..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_core_hlsl.h.meta +++ /dev/null @@ -1,60 +0,0 @@ -fileFormatVersion: 2 -guid: 89d6e02f97594f64ca2da4c8124df6cf -PluginImporter: - externalObjects: {} - serializedVersion: 2 - iconMap: {} - executionOrder: {} - defineConstraints: [] - isPreloaded: 0 - isOverridable: 0 - isExplicitlyReferenced: 0 - validateReferences: 1 - platformData: - - first: - : Any - second: - enabled: 0 - settings: - Exclude Editor: 1 - Exclude GameCoreScarlett: 1 - Exclude GameCoreXboxOne: 1 - Exclude Linux64: 1 - Exclude OSXUniversal: 1 - Exclude PS4: 1 - Exclude PS5: 1 - Exclude WebGL: 1 - Exclude Win: 1 - Exclude Win64: 1 - - first: - Any: - second: - enabled: 0 - settings: {} - - first: - Editor: Editor - second: - enabled: 0 - settings: - DefaultValueInitialized: true - - first: - Standalone: Linux64 - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win64 - second: - enabled: 0 - settings: - CPU: None - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_core_portability.h b/Assets/Resources/FSR2/shaders/ffx_core_portability.h deleted file mode 100644 index 45be059..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_core_portability.h +++ /dev/null @@ -1,50 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -FfxFloat32x3 opAAddOneF3(FfxFloat32x3 d, FfxFloat32x3 a, FfxFloat32 b) -{ - d = a + ffxBroadcast3(b); - return d; -} - -FfxFloat32x3 opACpyF3(FfxFloat32x3 d, FfxFloat32x3 a) -{ - d = a; - return d; -} - -FfxFloat32x3 opAMulF3(FfxFloat32x3 d, FfxFloat32x3 a, FfxFloat32x3 b) -{ - d = a * b; - return d; -} - -FfxFloat32x3 opAMulOneF3(FfxFloat32x3 d, FfxFloat32x3 a, FfxFloat32 b) -{ - d = a * ffxBroadcast3(b); - return d; -} - -FfxFloat32x3 opARcpF3(FfxFloat32x3 d, FfxFloat32x3 a) -{ - d = rcp(a); - return d; -} diff --git a/Assets/Resources/FSR2/shaders/ffx_core_portability.h.meta b/Assets/Resources/FSR2/shaders/ffx_core_portability.h.meta deleted file mode 100644 index aad361f..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_core_portability.h.meta +++ /dev/null @@ -1,60 +0,0 @@ -fileFormatVersion: 2 -guid: 5d6e692075988194382122bac7819f02 -PluginImporter: - externalObjects: {} - serializedVersion: 2 - iconMap: {} - executionOrder: {} - defineConstraints: [] - isPreloaded: 0 - isOverridable: 0 - isExplicitlyReferenced: 0 - validateReferences: 1 - platformData: - - first: - : Any - second: - enabled: 0 - settings: - Exclude Editor: 1 - Exclude GameCoreScarlett: 1 - Exclude GameCoreXboxOne: 1 - Exclude Linux64: 1 - Exclude OSXUniversal: 1 - Exclude PS4: 1 - Exclude PS5: 1 - Exclude WebGL: 1 - Exclude Win: 1 - Exclude Win64: 1 - - first: - Any: - second: - enabled: 0 - settings: {} - - first: - Editor: Editor - second: - enabled: 0 - settings: - DefaultValueInitialized: true - - first: - Standalone: Linux64 - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win64 - second: - enabled: 0 - settings: - CPU: None - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr1.h b/Assets/Resources/FSR2/shaders/ffx_fsr1.h deleted file mode 100644 index 1ac23cf..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr1.h +++ /dev/null @@ -1,1250 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#ifdef __clang__ -#pragma clang diagnostic ignored "-Wunused-variable" -#endif - -/// Setup required constant values for EASU (works on CPU or GPU). -/// -/// @param [out] con0 -/// @param [out] con1 -/// @param [out] con2 -/// @param [out] con3 -/// @param [in] inputViewportInPixelsX The rendered image resolution being upscaled in X dimension. -/// @param [in] inputViewportInPixelsY The rendered image resolution being upscaled in Y dimension. -/// @param [in] inputSizeInPixelsX The resolution of the resource containing the input image (useful for dynamic resolution) in X dimension. -/// @param [in] inputSizeInPixelsY The resolution of the resource containing the input image (useful for dynamic resolution) in Y dimension. -/// @param [in] outputSizeInPixelsX The display resolution which the input image gets upscaled to in X dimension. -/// @param [in] outputSizeInPixelsY The display resolution which the input image gets upscaled to in Y dimension. -/// -/// @ingroup FSR1 -FFX_STATIC void ffxFsrPopulateEasuConstants( - FFX_PARAMETER_INOUT FfxUInt32x4 con0, - FFX_PARAMETER_INOUT FfxUInt32x4 con1, - FFX_PARAMETER_INOUT FfxUInt32x4 con2, - FFX_PARAMETER_INOUT FfxUInt32x4 con3, - FFX_PARAMETER_IN FfxFloat32 inputViewportInPixelsX, - FFX_PARAMETER_IN FfxFloat32 inputViewportInPixelsY, - FFX_PARAMETER_IN FfxFloat32 inputSizeInPixelsX, - FFX_PARAMETER_IN FfxFloat32 inputSizeInPixelsY, - FFX_PARAMETER_IN FfxFloat32 outputSizeInPixelsX, - FFX_PARAMETER_IN FfxFloat32 outputSizeInPixelsY) -{ - // Output integer position to a pixel position in viewport. - con0[0] = ffxAsUInt32(inputViewportInPixelsX * ffxReciprocal(outputSizeInPixelsX)); - con0[1] = ffxAsUInt32(inputViewportInPixelsY * ffxReciprocal(outputSizeInPixelsY)); - con0[2] = ffxAsUInt32(FfxFloat32(0.5) * inputViewportInPixelsX * ffxReciprocal(outputSizeInPixelsX) - FfxFloat32(0.5)); - con0[3] = ffxAsUInt32(FfxFloat32(0.5) * inputViewportInPixelsY * ffxReciprocal(outputSizeInPixelsY) - FfxFloat32(0.5)); - - // Viewport pixel position to normalized image space. - // This is used to get upper-left of 'F' tap. - con1[0] = ffxAsUInt32(ffxReciprocal(inputSizeInPixelsX)); - con1[1] = ffxAsUInt32(ffxReciprocal(inputSizeInPixelsY)); - - // Centers of gather4, first offset from upper-left of 'F'. - // +---+---+ - // | | | - // +--(0)--+ - // | b | c | - // +---F---+---+---+ - // | e | f | g | h | - // +--(1)--+--(2)--+ - // | i | j | k | l | - // +---+---+---+---+ - // | n | o | - // +--(3)--+ - // | | | - // +---+---+ - con1[2] = ffxAsUInt32(FfxFloat32(1.0) * ffxReciprocal(inputSizeInPixelsX)); - con1[3] = ffxAsUInt32(FfxFloat32(-1.0) * ffxReciprocal(inputSizeInPixelsY)); - - // These are from (0) instead of 'F'. - con2[0] = ffxAsUInt32(FfxFloat32(-1.0) * ffxReciprocal(inputSizeInPixelsX)); - con2[1] = ffxAsUInt32(FfxFloat32(2.0) * ffxReciprocal(inputSizeInPixelsY)); - con2[2] = ffxAsUInt32(FfxFloat32(1.0) * ffxReciprocal(inputSizeInPixelsX)); - con2[3] = ffxAsUInt32(FfxFloat32(2.0) * ffxReciprocal(inputSizeInPixelsY)); - con3[0] = ffxAsUInt32(FfxFloat32(0.0) * ffxReciprocal(inputSizeInPixelsX)); - con3[1] = ffxAsUInt32(FfxFloat32(4.0) * ffxReciprocal(inputSizeInPixelsY)); - con3[2] = con3[3] = 0; -} - -/// Setup required constant values for EASU (works on CPU or GPU). -/// -/// @param [out] con0 -/// @param [out] con1 -/// @param [out] con2 -/// @param [out] con3 -/// @param [in] inputViewportInPixelsX The resolution of the input in the X dimension. -/// @param [in] inputViewportInPixelsY The resolution of the input in the Y dimension. -/// @param [in] inputSizeInPixelsX The input size in pixels in the X dimension. -/// @param [in] inputSizeInPixelsY The input size in pixels in the Y dimension. -/// @param [in] outputSizeInPixelsX The output size in pixels in the X dimension. -/// @param [in] outputSizeInPixelsY The output size in pixels in the Y dimension. -/// @param [in] inputOffsetInPixelsX The input image offset in the X dimension into the resource containing it (useful for dynamic resolution). -/// @param [in] inputOffsetInPixelsY The input image offset in the Y dimension into the resource containing it (useful for dynamic resolution). -/// -/// @ingroup FSR1 -FFX_STATIC void ffxFsrPopulateEasuConstantsOffset( - FFX_PARAMETER_INOUT FfxUInt32x4 con0, - FFX_PARAMETER_INOUT FfxUInt32x4 con1, - FFX_PARAMETER_INOUT FfxUInt32x4 con2, - FFX_PARAMETER_INOUT FfxUInt32x4 con3, - FFX_PARAMETER_IN FfxFloat32 inputViewportInPixelsX, - FFX_PARAMETER_IN FfxFloat32 inputViewportInPixelsY, - FFX_PARAMETER_IN FfxFloat32 inputSizeInPixelsX, - FFX_PARAMETER_IN FfxFloat32 inputSizeInPixelsY, - FFX_PARAMETER_IN FfxFloat32 outputSizeInPixelsX, - FFX_PARAMETER_IN FfxFloat32 outputSizeInPixelsY, - FFX_PARAMETER_IN FfxFloat32 inputOffsetInPixelsX, - FFX_PARAMETER_IN FfxFloat32 inputOffsetInPixelsY) -{ - ffxFsrPopulateEasuConstants( - con0, - con1, - con2, - con3, - inputViewportInPixelsX, - inputViewportInPixelsY, - inputSizeInPixelsX, - inputSizeInPixelsY, - outputSizeInPixelsX, - outputSizeInPixelsY); - - // override - con0[2] = ffxAsUInt32(FfxFloat32(0.5) * inputViewportInPixelsX * ffxReciprocal(outputSizeInPixelsX) - FfxFloat32(0.5) + inputOffsetInPixelsX); - con0[3] = ffxAsUInt32(FfxFloat32(0.5) * inputViewportInPixelsY * ffxReciprocal(outputSizeInPixelsY) - FfxFloat32(0.5) + inputOffsetInPixelsY); -} - -#if defined(FFX_GPU) && defined(FFX_FSR_EASU_FLOAT) -// Input callback prototypes, need to be implemented by calling shader -FfxFloat32x4 FsrEasuRF(FfxFloat32x2 p); -FfxFloat32x4 FsrEasuGF(FfxFloat32x2 p); -FfxFloat32x4 FsrEasuBF(FfxFloat32x2 p); - -// Filtering for a given tap for the scalar. -void fsrEasuTapFloat( - FFX_PARAMETER_INOUT FfxFloat32x3 accumulatedColor, // Accumulated color, with negative lobe. - FFX_PARAMETER_INOUT FfxFloat32 accumulatedWeight, // Accumulated weight. - FFX_PARAMETER_IN FfxFloat32x2 pixelOffset, // Pixel offset from resolve position to tap. - FFX_PARAMETER_IN FfxFloat32x2 gradientDirection, // Gradient direction. - FFX_PARAMETER_IN FfxFloat32x2 length, // Length. - FFX_PARAMETER_IN FfxFloat32 negativeLobeStrength, // Negative lobe strength. - FFX_PARAMETER_IN FfxFloat32 clippingPoint, // Clipping point. - FFX_PARAMETER_IN FfxFloat32x3 color) // Tap color. -{ - // Rotate offset by direction. - FfxFloat32x2 rotatedOffset; - rotatedOffset.x = (pixelOffset.x * (gradientDirection.x)) + (pixelOffset.y * gradientDirection.y); - rotatedOffset.y = (pixelOffset.x * (-gradientDirection.y)) + (pixelOffset.y * gradientDirection.x); - - // Anisotropy. - rotatedOffset *= length; - - // Compute distance^2. - FfxFloat32 distanceSquared = rotatedOffset.x * rotatedOffset.x + rotatedOffset.y * rotatedOffset.y; - - // Limit to the window as at corner, 2 taps can easily be outside. - distanceSquared = ffxMin(distanceSquared, clippingPoint); - - // Approximation of lancos2 without sin() or rcp(), or sqrt() to get x. - // (25/16 * (2/5 * x^2 - 1)^2 - (25/16 - 1)) * (1/4 * x^2 - 1)^2 - // |_______________________________________| |_______________| - // base window - // The general form of the 'base' is, - // (a*(b*x^2-1)^2-(a-1)) - // Where 'a=1/(2*b-b^2)' and 'b' moves around the negative lobe. - FfxFloat32 weightB = FfxFloat32(2.0 / 5.0) * distanceSquared + FfxFloat32(-1.0); - FfxFloat32 weightA = negativeLobeStrength * distanceSquared + FfxFloat32(-1.0); - weightB *= weightB; - weightA *= weightA; - weightB = FfxFloat32(25.0 / 16.0) * weightB + FfxFloat32(-(25.0 / 16.0 - 1.0)); - FfxFloat32 weight = weightB * weightA; - - // Do weighted average. - accumulatedColor += color * weight; - accumulatedWeight += weight; -} - -// Accumulate direction and length. -void fsrEasuSetFloat( - FFX_PARAMETER_INOUT FfxFloat32x2 direction, - FFX_PARAMETER_INOUT FfxFloat32 length, - FFX_PARAMETER_IN FfxFloat32x2 pp, - FFX_PARAMETER_IN FfxBoolean biS, - FFX_PARAMETER_IN FfxBoolean biT, - FFX_PARAMETER_IN FfxBoolean biU, - FFX_PARAMETER_IN FfxBoolean biV, - FFX_PARAMETER_IN FfxFloat32 lA, - FFX_PARAMETER_IN FfxFloat32 lB, - FFX_PARAMETER_IN FfxFloat32 lC, - FFX_PARAMETER_IN FfxFloat32 lD, - FFX_PARAMETER_IN FfxFloat32 lE) -{ - // Compute bilinear weight, branches factor out as predicates are compiler time immediates. - // s t - // u v - FfxFloat32 weight = FfxFloat32(0.0); - if (biS) - weight = (FfxFloat32(1.0) - pp.x) * (FfxFloat32(1.0) - pp.y); - if (biT) - weight = pp.x * (FfxFloat32(1.0) - pp.y); - if (biU) - weight = (FfxFloat32(1.0) - pp.x) * pp.y; - if (biV) - weight = pp.x * pp.y; - - // Direction is the '+' diff. - // a - // b c d - // e - // Then takes magnitude from abs average of both sides of 'c'. - // Length converts gradient reversal to 0, smoothly to non-reversal at 1, shaped, then adding horz and vert terms. - FfxFloat32 dc = lD - lC; - FfxFloat32 cb = lC - lB; - FfxFloat32 lengthX = max(abs(dc), abs(cb)); - lengthX = ffxApproximateReciprocal(lengthX); - FfxFloat32 directionX = lD - lB; - direction.x += directionX * weight; - lengthX = ffxSaturate(abs(directionX) * lengthX); - lengthX *= lengthX; - length += lengthX * weight; - - // Repeat for the y axis. - FfxFloat32 ec = lE - lC; - FfxFloat32 ca = lC - lA; - FfxFloat32 lengthY = max(abs(ec), abs(ca)); - lengthY = ffxApproximateReciprocal(lengthY); - FfxFloat32 directionY = lE - lA; - direction.y += directionY * weight; - lengthY = ffxSaturate(abs(directionY) * lengthY); - lengthY *= lengthY; - length += lengthY * weight; -} - -/// Apply edge-aware spatial upsampling using 32bit floating point precision calculations. -/// -/// @param [out] outPixel The computed color of a pixel. -/// @param [in] integerPosition Integer pixel position within the output. -/// @param [in] con0 The first constant value generated by ffxFsrPopulateEasuConstants. -/// @param [in] con1 The second constant value generated by ffxFsrPopulateEasuConstants. -/// @param [in] con2 The third constant value generated by ffxFsrPopulateEasuConstants. -/// @param [in] con3 The fourth constant value generated by ffxFsrPopulateEasuConstants. -/// -/// @ingroup FSR -void ffxFsrEasuFloat( - FFX_PARAMETER_OUT FfxFloat32x3 pix, - FFX_PARAMETER_IN FfxUInt32x2 ip, - FFX_PARAMETER_IN FfxUInt32x4 con0, - FFX_PARAMETER_IN FfxUInt32x4 con1, - FFX_PARAMETER_IN FfxUInt32x4 con2, - FFX_PARAMETER_IN FfxUInt32x4 con3) -{ - // Get position of 'f'. - FfxFloat32x2 pp = FfxFloat32x2(ip) * ffxAsFloat(con0.xy) + ffxAsFloat(con0.zw); - FfxFloat32x2 fp = floor(pp); - pp -= fp; - - // 12-tap kernel. - // b c - // e f g h - // i j k l - // n o - // Gather 4 ordering. - // a b - // r g - // For packed FP16, need either {rg} or {ab} so using the following setup for gather in all versions, - // a b <- unused (z) - // r g - // a b a b - // r g r g - // a b - // r g <- unused (z) - // Allowing dead-code removal to remove the 'z's. - FfxFloat32x2 p0 = fp * ffxAsFloat(con1.xy) + ffxAsFloat(con1.zw); - - // These are from p0 to avoid pulling two constants on pre-Navi hardware. - FfxFloat32x2 p1 = p0 + ffxAsFloat(con2.xy); - FfxFloat32x2 p2 = p0 + ffxAsFloat(con2.zw); - FfxFloat32x2 p3 = p0 + ffxAsFloat(con3.xy); - FfxFloat32x4 bczzR = FsrEasuRF(p0); - FfxFloat32x4 bczzG = FsrEasuGF(p0); - FfxFloat32x4 bczzB = FsrEasuBF(p0); - FfxFloat32x4 ijfeR = FsrEasuRF(p1); - FfxFloat32x4 ijfeG = FsrEasuGF(p1); - FfxFloat32x4 ijfeB = FsrEasuBF(p1); - FfxFloat32x4 klhgR = FsrEasuRF(p2); - FfxFloat32x4 klhgG = FsrEasuGF(p2); - FfxFloat32x4 klhgB = FsrEasuBF(p2); - FfxFloat32x4 zzonR = FsrEasuRF(p3); - FfxFloat32x4 zzonG = FsrEasuGF(p3); - FfxFloat32x4 zzonB = FsrEasuBF(p3); - - // Simplest multi-channel approximate luma possible (luma times 2, in 2 FMA/MAD). - FfxFloat32x4 bczzL = bczzB * ffxBroadcast4(0.5) + (bczzR * ffxBroadcast4(0.5) + bczzG); - FfxFloat32x4 ijfeL = ijfeB * ffxBroadcast4(0.5) + (ijfeR * ffxBroadcast4(0.5) + ijfeG); - FfxFloat32x4 klhgL = klhgB * ffxBroadcast4(0.5) + (klhgR * ffxBroadcast4(0.5) + klhgG); - FfxFloat32x4 zzonL = zzonB * ffxBroadcast4(0.5) + (zzonR * ffxBroadcast4(0.5) + zzonG); - - // Rename. - FfxFloat32 bL = bczzL.x; - FfxFloat32 cL = bczzL.y; - FfxFloat32 iL = ijfeL.x; - FfxFloat32 jL = ijfeL.y; - FfxFloat32 fL = ijfeL.z; - FfxFloat32 eL = ijfeL.w; - FfxFloat32 kL = klhgL.x; - FfxFloat32 lL = klhgL.y; - FfxFloat32 hL = klhgL.z; - FfxFloat32 gL = klhgL.w; - FfxFloat32 oL = zzonL.z; - FfxFloat32 nL = zzonL.w; - - // Accumulate for bilinear interpolation. - FfxFloat32x2 dir = ffxBroadcast2(0.0); - FfxFloat32 len = FfxFloat32(0.0); - fsrEasuSetFloat(dir, len, pp, FFX_TRUE, FFX_FALSE, FFX_FALSE, FFX_FALSE, bL, eL, fL, gL, jL); - fsrEasuSetFloat(dir, len, pp, FFX_FALSE, FFX_TRUE, FFX_FALSE, FFX_FALSE, cL, fL, gL, hL, kL); - fsrEasuSetFloat(dir, len, pp, FFX_FALSE, FFX_FALSE, FFX_TRUE, FFX_FALSE, fL, iL, jL, kL, nL); - fsrEasuSetFloat(dir, len, pp, FFX_FALSE, FFX_FALSE, FFX_FALSE, FFX_TRUE, gL, jL, kL, lL, oL); - - // Normalize with approximation, and cleanup close to zero. - FfxFloat32x2 dir2 = dir * dir; - FfxFloat32 dirR = dir2.x + dir2.y; - FfxUInt32 zro = dirR < FfxFloat32(1.0 / 32768.0); - dirR = ffxApproximateReciprocalSquareRoot(dirR); - dirR = zro ? FfxFloat32(1.0) : dirR; - dir.x = zro ? FfxFloat32(1.0) : dir.x; - dir *= ffxBroadcast2(dirR); - - // Transform from {0 to 2} to {0 to 1} range, and shape with square. - len = len * FfxFloat32(0.5); - len *= len; - - // Stretch kernel {1.0 vert|horz, to sqrt(2.0) on diagonal}. - FfxFloat32 stretch = (dir.x * dir.x + dir.y * dir.y) * ffxApproximateReciprocal(max(abs(dir.x), abs(dir.y))); - - // Anisotropic length after rotation, - // x := 1.0 lerp to 'stretch' on edges - // y := 1.0 lerp to 2x on edges - FfxFloat32x2 len2 = FfxFloat32x2(FfxFloat32(1.0) + (stretch - FfxFloat32(1.0)) * len, FfxFloat32(1.0) + FfxFloat32(-0.5) * len); - - // Based on the amount of 'edge', - // the window shifts from +/-{sqrt(2.0) to slightly beyond 2.0}. - FfxFloat32 lob = FfxFloat32(0.5) + FfxFloat32((1.0 / 4.0 - 0.04) - 0.5) * len; - - // Set distance^2 clipping point to the end of the adjustable window. - FfxFloat32 clp = ffxApproximateReciprocal(lob); - - // Accumulation mixed with min/max of 4 nearest. - // b c - // e f g h - // i j k l - // n o - FfxFloat32x3 min4 = - ffxMin(ffxMin3(FfxFloat32x3(ijfeR.z, ijfeG.z, ijfeB.z), FfxFloat32x3(klhgR.w, klhgG.w, klhgB.w), FfxFloat32x3(ijfeR.y, ijfeG.y, ijfeB.y)), - FfxFloat32x3(klhgR.x, klhgG.x, klhgB.x)); - FfxFloat32x3 max4 = - max(ffxMax3(FfxFloat32x3(ijfeR.z, ijfeG.z, ijfeB.z), FfxFloat32x3(klhgR.w, klhgG.w, klhgB.w), FfxFloat32x3(ijfeR.y, ijfeG.y, ijfeB.y)), FfxFloat32x3(klhgR.x, klhgG.x, klhgB.x)); - - // Accumulation. - FfxFloat32x3 aC = ffxBroadcast3(0.0); - FfxFloat32 aW = FfxFloat32(0.0); - fsrEasuTapFloat(aC, aW, FfxFloat32x2(0.0, -1.0) - pp, dir, len2, lob, clp, FfxFloat32x3(bczzR.x, bczzG.x, bczzB.x)); // b - fsrEasuTapFloat(aC, aW, FfxFloat32x2(1.0, -1.0) - pp, dir, len2, lob, clp, FfxFloat32x3(bczzR.y, bczzG.y, bczzB.y)); // c - fsrEasuTapFloat(aC, aW, FfxFloat32x2(-1.0, 1.0) - pp, dir, len2, lob, clp, FfxFloat32x3(ijfeR.x, ijfeG.x, ijfeB.x)); // i - fsrEasuTapFloat(aC, aW, FfxFloat32x2(0.0, 1.0) - pp, dir, len2, lob, clp, FfxFloat32x3(ijfeR.y, ijfeG.y, ijfeB.y)); // j - fsrEasuTapFloat(aC, aW, FfxFloat32x2(0.0, 0.0) - pp, dir, len2, lob, clp, FfxFloat32x3(ijfeR.z, ijfeG.z, ijfeB.z)); // f - fsrEasuTapFloat(aC, aW, FfxFloat32x2(-1.0, 0.0) - pp, dir, len2, lob, clp, FfxFloat32x3(ijfeR.w, ijfeG.w, ijfeB.w)); // e - fsrEasuTapFloat(aC, aW, FfxFloat32x2(1.0, 1.0) - pp, dir, len2, lob, clp, FfxFloat32x3(klhgR.x, klhgG.x, klhgB.x)); // k - fsrEasuTapFloat(aC, aW, FfxFloat32x2(2.0, 1.0) - pp, dir, len2, lob, clp, FfxFloat32x3(klhgR.y, klhgG.y, klhgB.y)); // l - fsrEasuTapFloat(aC, aW, FfxFloat32x2(2.0, 0.0) - pp, dir, len2, lob, clp, FfxFloat32x3(klhgR.z, klhgG.z, klhgB.z)); // h - fsrEasuTapFloat(aC, aW, FfxFloat32x2(1.0, 0.0) - pp, dir, len2, lob, clp, FfxFloat32x3(klhgR.w, klhgG.w, klhgB.w)); // g - fsrEasuTapFloat(aC, aW, FfxFloat32x2(1.0, 2.0) - pp, dir, len2, lob, clp, FfxFloat32x3(zzonR.z, zzonG.z, zzonB.z)); // o - fsrEasuTapFloat(aC, aW, FfxFloat32x2(0.0, 2.0) - pp, dir, len2, lob, clp, FfxFloat32x3(zzonR.w, zzonG.w, zzonB.w)); // n - - // Normalize and dering. - pix = ffxMin(max4, max(min4, aC * ffxBroadcast3(rcp(aW)))); -} -#endif // #if defined(FFX_GPU) && defined(FFX_FSR_EASU_FLOAT) - -#if defined(FFX_GPU) && FFX_HALF == 1 && defined(FFX_FSR_EASU_HALF) -// Input callback prototypes, need to be implemented by calling shader -FfxFloat16x4 FsrEasuRH(FfxFloat32x2 p); -FfxFloat16x4 FsrEasuGH(FfxFloat32x2 p); -FfxFloat16x4 FsrEasuBH(FfxFloat32x2 p); - -// This runs 2 taps in parallel. -void FsrEasuTapH( - FFX_PARAMETER_INOUT FfxFloat16x2 aCR, - FFX_PARAMETER_INOUT FfxFloat16x2 aCG, - FFX_PARAMETER_INOUT FfxFloat16x2 aCB, - FFX_PARAMETER_INOUT FfxFloat16x2 aW, - FFX_PARAMETER_IN FfxFloat16x2 offX, - FFX_PARAMETER_IN FfxFloat16x2 offY, - FFX_PARAMETER_IN FfxFloat16x2 dir, - FFX_PARAMETER_IN FfxFloat16x2 len, - FFX_PARAMETER_IN FfxFloat16 lob, - FFX_PARAMETER_IN FfxFloat16 clp, - FFX_PARAMETER_IN FfxFloat16x2 cR, - FFX_PARAMETER_IN FfxFloat16x2 cG, - FFX_PARAMETER_IN FfxFloat16x2 cB) -{ - FfxFloat16x2 vX, vY; - vX = offX * dir.xx + offY * dir.yy; - vY = offX * (-dir.yy) + offY * dir.xx; - vX *= len.x; - vY *= len.y; - FfxFloat16x2 d2 = vX * vX + vY * vY; - d2 = min(d2, FFX_BROADCAST_FLOAT16X2(clp)); - FfxFloat16x2 wB = FFX_BROADCAST_FLOAT16X2(2.0 / 5.0) * d2 + FFX_BROADCAST_FLOAT16X2(-1.0); - FfxFloat16x2 wA = FFX_BROADCAST_FLOAT16X2(lob) * d2 + FFX_BROADCAST_FLOAT16X2(-1.0); - wB *= wB; - wA *= wA; - wB = FFX_BROADCAST_FLOAT16X2(25.0 / 16.0) * wB + FFX_BROADCAST_FLOAT16X2(-(25.0 / 16.0 - 1.0)); - FfxFloat16x2 w = wB * wA; - aCR += cR * w; - aCG += cG * w; - aCB += cB * w; - aW += w; -} - -// This runs 2 taps in parallel. -void FsrEasuSetH( - FFX_PARAMETER_INOUT FfxFloat16x2 dirPX, - FFX_PARAMETER_INOUT FfxFloat16x2 dirPY, - FFX_PARAMETER_INOUT FfxFloat16x2 lenP, - FFX_PARAMETER_IN FfxFloat16x2 pp, - FFX_PARAMETER_IN FfxBoolean biST, - FFX_PARAMETER_IN FfxBoolean biUV, - FFX_PARAMETER_IN FfxFloat16x2 lA, - FFX_PARAMETER_IN FfxFloat16x2 lB, - FFX_PARAMETER_IN FfxFloat16x2 lC, - FFX_PARAMETER_IN FfxFloat16x2 lD, - FFX_PARAMETER_IN FfxFloat16x2 lE) -{ - FfxFloat16x2 w = FFX_BROADCAST_FLOAT16X2(0.0); - - if (biST) - w = (FfxFloat16x2(1.0, 0.0) + FfxFloat16x2(-pp.x, pp.x)) * FFX_BROADCAST_FLOAT16X2(FFX_BROADCAST_FLOAT16(1.0) - pp.y); - - if (biUV) - w = (FfxFloat16x2(1.0, 0.0) + FfxFloat16x2(-pp.x, pp.x)) * FFX_BROADCAST_FLOAT16X2(pp.y); - - // ABS is not free in the packed FP16 path. - FfxFloat16x2 dc = lD - lC; - FfxFloat16x2 cb = lC - lB; - FfxFloat16x2 lenX = max(abs(dc), abs(cb)); - lenX = ffxReciprocalHalf(lenX); - - FfxFloat16x2 dirX = lD - lB; - dirPX += dirX * w; - lenX = ffxSaturate(abs(dirX) * lenX); - lenX *= lenX; - lenP += lenX * w; - FfxFloat16x2 ec = lE - lC; - FfxFloat16x2 ca = lC - lA; - FfxFloat16x2 lenY = max(abs(ec), abs(ca)); - lenY = ffxReciprocalHalf(lenY); - FfxFloat16x2 dirY = lE - lA; - dirPY += dirY * w; - lenY = ffxSaturate(abs(dirY) * lenY); - lenY *= lenY; - lenP += lenY * w; -} - -void FsrEasuH( - FFX_PARAMETER_OUT FfxFloat16x3 pix, - FFX_PARAMETER_IN FfxUInt32x2 ip, - FFX_PARAMETER_IN FfxUInt32x4 con0, - FFX_PARAMETER_IN FfxUInt32x4 con1, - FFX_PARAMETER_IN FfxUInt32x4 con2, - FFX_PARAMETER_IN FfxUInt32x4 con3) -{ - FfxFloat32x2 pp = FfxFloat32x2(ip) * ffxAsFloat(con0.xy) + ffxAsFloat(con0.zw); - FfxFloat32x2 fp = floor(pp); - pp -= fp; - FfxFloat16x2 ppp = FfxFloat16x2(pp); - - FfxFloat32x2 p0 = fp * ffxAsFloat(con1.xy) + ffxAsFloat(con1.zw); - FfxFloat32x2 p1 = p0 + ffxAsFloat(con2.xy); - FfxFloat32x2 p2 = p0 + ffxAsFloat(con2.zw); - FfxFloat32x2 p3 = p0 + ffxAsFloat(con3.xy); - FfxFloat16x4 bczzR = FsrEasuRH(p0); - FfxFloat16x4 bczzG = FsrEasuGH(p0); - FfxFloat16x4 bczzB = FsrEasuBH(p0); - FfxFloat16x4 ijfeR = FsrEasuRH(p1); - FfxFloat16x4 ijfeG = FsrEasuGH(p1); - FfxFloat16x4 ijfeB = FsrEasuBH(p1); - FfxFloat16x4 klhgR = FsrEasuRH(p2); - FfxFloat16x4 klhgG = FsrEasuGH(p2); - FfxFloat16x4 klhgB = FsrEasuBH(p2); - FfxFloat16x4 zzonR = FsrEasuRH(p3); - FfxFloat16x4 zzonG = FsrEasuGH(p3); - FfxFloat16x4 zzonB = FsrEasuBH(p3); - - FfxFloat16x4 bczzL = bczzB * FFX_BROADCAST_FLOAT16X4(0.5) + (bczzR * FFX_BROADCAST_FLOAT16X4(0.5) + bczzG); - FfxFloat16x4 ijfeL = ijfeB * FFX_BROADCAST_FLOAT16X4(0.5) + (ijfeR * FFX_BROADCAST_FLOAT16X4(0.5) + ijfeG); - FfxFloat16x4 klhgL = klhgB * FFX_BROADCAST_FLOAT16X4(0.5) + (klhgR * FFX_BROADCAST_FLOAT16X4(0.5) + klhgG); - FfxFloat16x4 zzonL = zzonB * FFX_BROADCAST_FLOAT16X4(0.5) + (zzonR * FFX_BROADCAST_FLOAT16X4(0.5) + zzonG); - FfxFloat16 bL = bczzL.x; - FfxFloat16 cL = bczzL.y; - FfxFloat16 iL = ijfeL.x; - FfxFloat16 jL = ijfeL.y; - FfxFloat16 fL = ijfeL.z; - FfxFloat16 eL = ijfeL.w; - FfxFloat16 kL = klhgL.x; - FfxFloat16 lL = klhgL.y; - FfxFloat16 hL = klhgL.z; - FfxFloat16 gL = klhgL.w; - FfxFloat16 oL = zzonL.z; - FfxFloat16 nL = zzonL.w; - - // This part is different, accumulating 2 taps in parallel. - FfxFloat16x2 dirPX = FFX_BROADCAST_FLOAT16X2(0.0); - FfxFloat16x2 dirPY = FFX_BROADCAST_FLOAT16X2(0.0); - FfxFloat16x2 lenP = FFX_BROADCAST_FLOAT16X2(0.0); - FsrEasuSetH(dirPX, - dirPY, - lenP, - ppp, - FfxUInt32(true), - FfxUInt32(false), - FfxFloat16x2(bL, cL), - FfxFloat16x2(eL, fL), - FfxFloat16x2(fL, gL), - FfxFloat16x2(gL, hL), - FfxFloat16x2(jL, kL)); - FsrEasuSetH(dirPX, - dirPY, - lenP, - ppp, - FfxUInt32(false), - FfxUInt32(true), - FfxFloat16x2(fL, gL), - FfxFloat16x2(iL, jL), - FfxFloat16x2(jL, kL), - FfxFloat16x2(kL, lL), - FfxFloat16x2(nL, oL)); - FfxFloat16x2 dir = FfxFloat16x2(dirPX.r + dirPX.g, dirPY.r + dirPY.g); - FfxFloat16 len = lenP.r + lenP.g; - - FfxFloat16x2 dir2 = dir * dir; - FfxFloat16 dirR = dir2.x + dir2.y; - FfxBoolean zro = FfxBoolean(dirR < FFX_BROADCAST_FLOAT16(1.0 / 32768.0)); - dirR = ffxApproximateReciprocalSquareRootHalf(dirR); - dirR = (zro > 0) ? FFX_BROADCAST_FLOAT16(1.0) : dirR; - dir.x = (zro > 0) ? FFX_BROADCAST_FLOAT16(1.0) : dir.x; - dir *= FFX_BROADCAST_FLOAT16X2(dirR); - len = len * FFX_BROADCAST_FLOAT16(0.5); - len *= len; - FfxFloat16 stretch = (dir.x * dir.x + dir.y * dir.y) * ffxApproximateReciprocalHalf(max(abs(dir.x), abs(dir.y))); - FfxFloat16x2 len2 = - FfxFloat16x2(FFX_BROADCAST_FLOAT16(1.0) + (stretch - FFX_BROADCAST_FLOAT16(1.0)) * len, FFX_BROADCAST_FLOAT16(1.0) + FFX_BROADCAST_FLOAT16(-0.5) * len); - FfxFloat16 lob = FFX_BROADCAST_FLOAT16(0.5) + FFX_BROADCAST_FLOAT16((1.0 / 4.0 - 0.04) - 0.5) * len; - FfxFloat16 clp = ffxApproximateReciprocalHalf(lob); - - // FP16 is different, using packed trick to do min and max in same operation. - FfxFloat16x2 bothR = - max(max(FfxFloat16x2(-ijfeR.z, ijfeR.z), FfxFloat16x2(-klhgR.w, klhgR.w)), max(FfxFloat16x2(-ijfeR.y, ijfeR.y), FfxFloat16x2(-klhgR.x, klhgR.x))); - FfxFloat16x2 bothG = - max(max(FfxFloat16x2(-ijfeG.z, ijfeG.z), FfxFloat16x2(-klhgG.w, klhgG.w)), max(FfxFloat16x2(-ijfeG.y, ijfeG.y), FfxFloat16x2(-klhgG.x, klhgG.x))); - FfxFloat16x2 bothB = - max(max(FfxFloat16x2(-ijfeB.z, ijfeB.z), FfxFloat16x2(-klhgB.w, klhgB.w)), max(FfxFloat16x2(-ijfeB.y, ijfeB.y), FfxFloat16x2(-klhgB.x, klhgB.x))); - - // This part is different for FP16, working pairs of taps at a time. - FfxFloat16x2 pR = FFX_BROADCAST_FLOAT16X2(0.0); - FfxFloat16x2 pG = FFX_BROADCAST_FLOAT16X2(0.0); - FfxFloat16x2 pB = FFX_BROADCAST_FLOAT16X2(0.0); - FfxFloat16x2 pW = FFX_BROADCAST_FLOAT16X2(0.0); - FsrEasuTapH(pR, pG, pB, pW, FfxFloat16x2(0.0, 1.0) - ppp.xx, FfxFloat16x2(-1.0, -1.0) - ppp.yy, dir, len2, lob, clp, bczzR.xy, bczzG.xy, bczzB.xy); - FsrEasuTapH(pR, pG, pB, pW, FfxFloat16x2(-1.0, 0.0) - ppp.xx, FfxFloat16x2(1.0, 1.0) - ppp.yy, dir, len2, lob, clp, ijfeR.xy, ijfeG.xy, ijfeB.xy); - FsrEasuTapH(pR, pG, pB, pW, FfxFloat16x2(0.0, -1.0) - ppp.xx, FfxFloat16x2(0.0, 0.0) - ppp.yy, dir, len2, lob, clp, ijfeR.zw, ijfeG.zw, ijfeB.zw); - FsrEasuTapH(pR, pG, pB, pW, FfxFloat16x2(1.0, 2.0) - ppp.xx, FfxFloat16x2(1.0, 1.0) - ppp.yy, dir, len2, lob, clp, klhgR.xy, klhgG.xy, klhgB.xy); - FsrEasuTapH(pR, pG, pB, pW, FfxFloat16x2(2.0, 1.0) - ppp.xx, FfxFloat16x2(0.0, 0.0) - ppp.yy, dir, len2, lob, clp, klhgR.zw, klhgG.zw, klhgB.zw); - FsrEasuTapH(pR, pG, pB, pW, FfxFloat16x2(1.0, 0.0) - ppp.xx, FfxFloat16x2(2.0, 2.0) - ppp.yy, dir, len2, lob, clp, zzonR.zw, zzonG.zw, zzonB.zw); - FfxFloat16x3 aC = FfxFloat16x3(pR.x + pR.y, pG.x + pG.y, pB.x + pB.y); - FfxFloat16 aW = pW.x + pW.y; - - // Slightly different for FP16 version due to combined min and max. - pix = min(FfxFloat16x3(bothR.y, bothG.y, bothB.y), max(-FfxFloat16x3(bothR.x, bothG.x, bothB.x), aC * FFX_BROADCAST_FLOAT16X3(ffxReciprocalHalf(aW)))); -} -#endif // #if defined(FFX_GPU) && defined(FFX_HALF) && defined(FFX_FSR_EASU_HALF) - -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// -//_____________________________________________________________/\_______________________________________________________________ -//============================================================================================================================== -// -// FSR - [RCAS] ROBUST CONTRAST ADAPTIVE SHARPENING -// -//------------------------------------------------------------------------------------------------------------------------------ -// CAS uses a simplified mechanism to convert local contrast into a variable amount of sharpness. -// RCAS uses a more exact mechanism, solving for the maximum local sharpness possible before clipping. -// RCAS also has a built in process to limit sharpening of what it detects as possible noise. -// RCAS sharper does not support scaling, as it should be applied after EASU scaling. -// Pass EASU output straight into RCAS, no color conversions necessary. -//------------------------------------------------------------------------------------------------------------------------------ -// RCAS is based on the following logic. -// RCAS uses a 5 tap filter in a cross pattern (same as CAS), -// w n -// w 1 w for taps w m e -// w s -// Where 'w' is the negative lobe weight. -// output = (w*(n+e+w+s)+m)/(4*w+1) -// RCAS solves for 'w' by seeing where the signal might clip out of the {0 to 1} input range, -// 0 == (w*(n+e+w+s)+m)/(4*w+1) -> w = -m/(n+e+w+s) -// 1 == (w*(n+e+w+s)+m)/(4*w+1) -> w = (1-m)/(n+e+w+s-4*1) -// Then chooses the 'w' which results in no clipping, limits 'w', and multiplies by the 'sharp' amount. -// This solution above has issues with MSAA input as the steps along the gradient cause edge detection issues. -// So RCAS uses 4x the maximum and 4x the minimum (depending on equation)in place of the individual taps. -// As well as switching from 'm' to either the minimum or maximum (depending on side), to help in energy conservation. -// This stabilizes RCAS. -// RCAS does a simple highpass which is normalized against the local contrast then shaped, -// 0.25 -// 0.25 -1 0.25 -// 0.25 -// This is used as a noise detection filter, to reduce the effect of RCAS on grain, and focus on real edges. -// -// GLSL example for the required callbacks : -// -// FfxFloat16x4 FsrRcasLoadH(FfxInt16x2 p){return FfxFloat16x4(imageLoad(imgSrc,FfxInt32x2(p)));} -// void FsrRcasInputH(inout FfxFloat16 r,inout FfxFloat16 g,inout FfxFloat16 b) -// { -// //do any simple input color conversions here or leave empty if none needed -// } -// -// FsrRcasCon need to be called from the CPU or GPU to set up constants. -// Including a GPU example here, the 'con' value would be stored out to a constant buffer. -// -// FfxUInt32x4 con; -// FsrRcasCon(con, -// 0.0); // The scale is {0.0 := maximum sharpness, to N>0, where N is the number of stops (halving) of the reduction of sharpness}. -// --------------- -// RCAS sharpening supports a CAS-like pass-through alpha via, -// #define FSR_RCAS_PASSTHROUGH_ALPHA 1 -// RCAS also supports a define to enable a more expensive path to avoid some sharpening of noise. -// Would suggest it is better to apply film grain after RCAS sharpening (and after scaling) instead of using this define, -// #define FSR_RCAS_DENOISE 1 -//============================================================================================================================== -// This is set at the limit of providing unnatural results for sharpening. -#define FSR_RCAS_LIMIT (0.25-(1.0/16.0)) -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// -//_____________________________________________________________/\_______________________________________________________________ -//============================================================================================================================== -// CONSTANT SETUP -//============================================================================================================================== -// Call to setup required constant values (works on CPU or GPU). - FFX_STATIC void FsrRcasCon(FfxUInt32x4 con, - // The scale is {0.0 := maximum, to N>0, where N is the number of stops (halving) of the reduction of sharpness}. - FfxFloat32 sharpness) - { - // Transform from stops to linear value. - sharpness = exp2(-sharpness); - FfxFloat32x2 hSharp = {sharpness, sharpness}; - con[0] = ffxAsUInt32(sharpness); - con[1] = packHalf2x16(hSharp); - con[2] = 0; - con[3] = 0; - } - //////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// -//_____________________________________________________________/\_______________________________________________________________ -//============================================================================================================================== -// NON-PACKED 32-BIT VERSION -//============================================================================================================================== -#if defined(FFX_GPU)&&defined(FSR_RCAS_F) - // Input callback prototypes that need to be implemented by calling shader - FfxFloat32x4 FsrRcasLoadF(FfxInt32x2 p); - void FsrRcasInputF(inout FfxFloat32 r,inout FfxFloat32 g,inout FfxFloat32 b); -//------------------------------------------------------------------------------------------------------------------------------ - void FsrRcasF(out FfxFloat32 pixR, // Output values, non-vector so port between RcasFilter() and RcasFilterH() is easy. - out FfxFloat32 pixG, - out FfxFloat32 pixB, -#ifdef FSR_RCAS_PASSTHROUGH_ALPHA - out FfxFloat32 pixA, -#endif - FfxUInt32x2 ip, // Integer pixel position in output. - FfxUInt32x4 con) - { // Constant generated by RcasSetup(). - // Algorithm uses minimal 3x3 pixel neighborhood. - // b - // d e f - // h - FfxInt32x2 sp = FfxInt32x2(ip); - FfxFloat32x3 b = FsrRcasLoadF(sp + FfxInt32x2(0, -1)).rgb; - FfxFloat32x3 d = FsrRcasLoadF(sp + FfxInt32x2(-1, 0)).rgb; -#ifdef FSR_RCAS_PASSTHROUGH_ALPHA - FfxFloat32x4 ee = FsrRcasLoadF(sp); - FfxFloat32x3 e = ee.rgb; - pixA = ee.a; -#else - FfxFloat32x3 e = FsrRcasLoadF(sp).rgb; -#endif - FfxFloat32x3 f = FsrRcasLoadF(sp + FfxInt32x2(1, 0)).rgb; - FfxFloat32x3 h = FsrRcasLoadF(sp + FfxInt32x2(0, 1)).rgb; - // Rename (32-bit) or regroup (16-bit). - FfxFloat32 bR = b.r; - FfxFloat32 bG = b.g; - FfxFloat32 bB = b.b; - FfxFloat32 dR = d.r; - FfxFloat32 dG = d.g; - FfxFloat32 dB = d.b; - FfxFloat32 eR = e.r; - FfxFloat32 eG = e.g; - FfxFloat32 eB = e.b; - FfxFloat32 fR = f.r; - FfxFloat32 fG = f.g; - FfxFloat32 fB = f.b; - FfxFloat32 hR = h.r; - FfxFloat32 hG = h.g; - FfxFloat32 hB = h.b; - // Run optional input transform. - FsrRcasInputF(bR, bG, bB); - FsrRcasInputF(dR, dG, dB); - FsrRcasInputF(eR, eG, eB); - FsrRcasInputF(fR, fG, fB); - FsrRcasInputF(hR, hG, hB); - // Luma times 2. - FfxFloat32 bL = bB * FfxFloat32(0.5) + (bR * FfxFloat32(0.5) + bG); - FfxFloat32 dL = dB * FfxFloat32(0.5) + (dR * FfxFloat32(0.5) + dG); - FfxFloat32 eL = eB * FfxFloat32(0.5) + (eR * FfxFloat32(0.5) + eG); - FfxFloat32 fL = fB * FfxFloat32(0.5) + (fR * FfxFloat32(0.5) + fG); - FfxFloat32 hL = hB * FfxFloat32(0.5) + (hR * FfxFloat32(0.5) + hG); - // Noise detection. - FfxFloat32 nz = FfxFloat32(0.25) * bL + FfxFloat32(0.25) * dL + FfxFloat32(0.25) * fL + FfxFloat32(0.25) * hL - eL; - nz = ffxSaturate(abs(nz) * ffxApproximateReciprocalMedium(ffxMax3(ffxMax3(bL, dL, eL), fL, hL) - ffxMin3(ffxMin3(bL, dL, eL), fL, hL))); - nz = FfxFloat32(-0.5) * nz + FfxFloat32(1.0); - // Min and max of ring. - FfxFloat32 mn4R = ffxMin(ffxMin3(bR, dR, fR), hR); - FfxFloat32 mn4G = ffxMin(ffxMin3(bG, dG, fG), hG); - FfxFloat32 mn4B = ffxMin(ffxMin3(bB, dB, fB), hB); - FfxFloat32 mx4R = max(ffxMax3(bR, dR, fR), hR); - FfxFloat32 mx4G = max(ffxMax3(bG, dG, fG), hG); - FfxFloat32 mx4B = max(ffxMax3(bB, dB, fB), hB); - // Immediate constants for peak range. - FfxFloat32x2 peakC = FfxFloat32x2(1.0, -1.0 * 4.0); - // Limiters, these need to be high precision RCPs. - FfxFloat32 hitMinR = mn4R * rcp(FfxFloat32(4.0) * mx4R); - FfxFloat32 hitMinG = mn4G * rcp(FfxFloat32(4.0) * mx4G); - FfxFloat32 hitMinB = mn4B * rcp(FfxFloat32(4.0) * mx4B); - FfxFloat32 hitMaxR = (peakC.x - mx4R) * rcp(FfxFloat32(4.0) * mn4R + peakC.y); - FfxFloat32 hitMaxG = (peakC.x - mx4G) * rcp(FfxFloat32(4.0) * mn4G + peakC.y); - FfxFloat32 hitMaxB = (peakC.x - mx4B) * rcp(FfxFloat32(4.0) * mn4B + peakC.y); - FfxFloat32 lobeR = max(-hitMinR, hitMaxR); - FfxFloat32 lobeG = max(-hitMinG, hitMaxG); - FfxFloat32 lobeB = max(-hitMinB, hitMaxB); - FfxFloat32 lobe = max(FfxFloat32(-FSR_RCAS_LIMIT), ffxMin(ffxMax3(lobeR, lobeG, lobeB), FfxFloat32(0.0))) * ffxAsFloat - (con.x); - // Apply noise removal. -#ifdef FSR_RCAS_DENOISE - lobe *= nz; -#endif - // Resolve, which needs the medium precision rcp approximation to avoid visible tonality changes. - FfxFloat32 rcpL = ffxApproximateReciprocalMedium(FfxFloat32(4.0) * lobe + FfxFloat32(1.0)); - pixR = (lobe * bR + lobe * dR + lobe * hR + lobe * fR + eR) * rcpL; - pixG = (lobe * bG + lobe * dG + lobe * hG + lobe * fG + eG) * rcpL; - pixB = (lobe * bB + lobe * dB + lobe * hB + lobe * fB + eB) * rcpL; - return; - } -#endif -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// -//_____________________________________________________________/\_______________________________________________________________ -//============================================================================================================================== -// NON-PACKED 16-BIT VERSION -//============================================================================================================================== -#if defined(FFX_GPU) && FFX_HALF == 1 && defined(FSR_RCAS_H) - // Input callback prototypes that need to be implemented by calling shader - FfxFloat16x4 FsrRcasLoadH(FfxInt16x2 p); - void FsrRcasInputH(inout FfxFloat16 r,inout FfxFloat16 g,inout FfxFloat16 b); -//------------------------------------------------------------------------------------------------------------------------------ - void FsrRcasH( - out FfxFloat16 pixR, // Output values, non-vector so port between RcasFilter() and RcasFilterH() is easy. - out FfxFloat16 pixG, - out FfxFloat16 pixB, - #ifdef FSR_RCAS_PASSTHROUGH_ALPHA - out FfxFloat16 pixA, - #endif - FfxUInt32x2 ip, // Integer pixel position in output. - FfxUInt32x4 con){ // Constant generated by RcasSetup(). - // Sharpening algorithm uses minimal 3x3 pixel neighborhood. - // b - // d e f - // h - FfxInt16x2 sp=FfxInt16x2(ip); - FfxFloat16x3 b=FsrRcasLoadH(sp+FfxInt16x2( 0,-1)).rgb; - FfxFloat16x3 d=FsrRcasLoadH(sp+FfxInt16x2(-1, 0)).rgb; - #ifdef FSR_RCAS_PASSTHROUGH_ALPHA - FfxFloat16x4 ee=FsrRcasLoadH(sp); - FfxFloat16x3 e=ee.rgb;pixA=ee.a; - #else - FfxFloat16x3 e=FsrRcasLoadH(sp).rgb; - #endif - FfxFloat16x3 f=FsrRcasLoadH(sp+FfxInt16x2( 1, 0)).rgb; - FfxFloat16x3 h=FsrRcasLoadH(sp+FfxInt16x2( 0, 1)).rgb; - // Rename (32-bit) or regroup (16-bit). - FfxFloat16 bR=b.r; - FfxFloat16 bG=b.g; - FfxFloat16 bB=b.b; - FfxFloat16 dR=d.r; - FfxFloat16 dG=d.g; - FfxFloat16 dB=d.b; - FfxFloat16 eR=e.r; - FfxFloat16 eG=e.g; - FfxFloat16 eB=e.b; - FfxFloat16 fR=f.r; - FfxFloat16 fG=f.g; - FfxFloat16 fB=f.b; - FfxFloat16 hR=h.r; - FfxFloat16 hG=h.g; - FfxFloat16 hB=h.b; - // Run optional input transform. - FsrRcasInputH(bR,bG,bB); - FsrRcasInputH(dR,dG,dB); - FsrRcasInputH(eR,eG,eB); - FsrRcasInputH(fR,fG,fB); - FsrRcasInputH(hR,hG,hB); - // Luma times 2. - FfxFloat16 bL=bB*FFX_BROADCAST_FLOAT16(0.5)+(bR*FFX_BROADCAST_FLOAT16(0.5)+bG); - FfxFloat16 dL=dB*FFX_BROADCAST_FLOAT16(0.5)+(dR*FFX_BROADCAST_FLOAT16(0.5)+dG); - FfxFloat16 eL=eB*FFX_BROADCAST_FLOAT16(0.5)+(eR*FFX_BROADCAST_FLOAT16(0.5)+eG); - FfxFloat16 fL=fB*FFX_BROADCAST_FLOAT16(0.5)+(fR*FFX_BROADCAST_FLOAT16(0.5)+fG); - FfxFloat16 hL=hB*FFX_BROADCAST_FLOAT16(0.5)+(hR*FFX_BROADCAST_FLOAT16(0.5)+hG); - // Noise detection. - FfxFloat16 nz=FFX_BROADCAST_FLOAT16(0.25)*bL+FFX_BROADCAST_FLOAT16(0.25)*dL+FFX_BROADCAST_FLOAT16(0.25)*fL+FFX_BROADCAST_FLOAT16(0.25)*hL-eL; - nz=ffxSaturate(abs(nz)*ffxApproximateReciprocalMediumHalf(ffxMax3Half(ffxMax3Half(bL,dL,eL),fL,hL)-ffxMin3Half(ffxMin3Half(bL,dL,eL),fL,hL))); - nz=FFX_BROADCAST_FLOAT16(-0.5)*nz+FFX_BROADCAST_FLOAT16(1.0); - // Min and max of ring. - FfxFloat16 mn4R=min(ffxMin3Half(bR,dR,fR),hR); - FfxFloat16 mn4G=min(ffxMin3Half(bG,dG,fG),hG); - FfxFloat16 mn4B=min(ffxMin3Half(bB,dB,fB),hB); - FfxFloat16 mx4R=max(ffxMax3Half(bR,dR,fR),hR); - FfxFloat16 mx4G=max(ffxMax3Half(bG,dG,fG),hG); - FfxFloat16 mx4B=max(ffxMax3Half(bB,dB,fB),hB); - // Immediate constants for peak range. - FfxFloat16x2 peakC=FfxFloat16x2(1.0,-1.0*4.0); - // Limiters, these need to be high precision RCPs. - FfxFloat16 hitMinR=mn4R*ffxReciprocalHalf(FFX_BROADCAST_FLOAT16(4.0)*mx4R); - FfxFloat16 hitMinG=mn4G*ffxReciprocalHalf(FFX_BROADCAST_FLOAT16(4.0)*mx4G); - FfxFloat16 hitMinB=mn4B*ffxReciprocalHalf(FFX_BROADCAST_FLOAT16(4.0)*mx4B); - FfxFloat16 hitMaxR=(peakC.x-mx4R)*ffxReciprocalHalf(FFX_BROADCAST_FLOAT16(4.0)*mn4R+peakC.y); - FfxFloat16 hitMaxG=(peakC.x-mx4G)*ffxReciprocalHalf(FFX_BROADCAST_FLOAT16(4.0)*mn4G+peakC.y); - FfxFloat16 hitMaxB=(peakC.x-mx4B)*ffxReciprocalHalf(FFX_BROADCAST_FLOAT16(4.0)*mn4B+peakC.y); - FfxFloat16 lobeR=max(-hitMinR,hitMaxR); - FfxFloat16 lobeG=max(-hitMinG,hitMaxG); - FfxFloat16 lobeB=max(-hitMinB,hitMaxB); - FfxFloat16 lobe=max(FFX_BROADCAST_FLOAT16(-FSR_RCAS_LIMIT),min(ffxMax3Half(lobeR,lobeG,lobeB),FFX_BROADCAST_FLOAT16(0.0)))*FFX_UINT32_TO_FLOAT16X2(con.y).x; - // Apply noise removal. - #ifdef FSR_RCAS_DENOISE - lobe*=nz; - #endif - // Resolve, which needs the medium precision rcp approximation to avoid visible tonality changes. - FfxFloat16 rcpL=ffxApproximateReciprocalMediumHalf(FFX_BROADCAST_FLOAT16(4.0)*lobe+FFX_BROADCAST_FLOAT16(1.0)); - pixR=(lobe*bR+lobe*dR+lobe*hR+lobe*fR+eR)*rcpL; - pixG=(lobe*bG+lobe*dG+lobe*hG+lobe*fG+eG)*rcpL; - pixB=(lobe*bB+lobe*dB+lobe*hB+lobe*fB+eB)*rcpL; -} -#endif -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// -//_____________________________________________________________/\_______________________________________________________________ -//============================================================================================================================== -// PACKED 16-BIT VERSION -//============================================================================================================================== -#if defined(FFX_GPU)&& FFX_HALF == 1 && defined(FSR_RCAS_HX2) - // Input callback prototypes that need to be implemented by the calling shader - FfxFloat16x4 FsrRcasLoadHx2(FfxInt16x2 p); - void FsrRcasInputHx2(inout FfxFloat16x2 r,inout FfxFloat16x2 g,inout FfxFloat16x2 b); -//------------------------------------------------------------------------------------------------------------------------------ - // Can be used to convert from packed Structures of Arrays to Arrays of Structures for store. - void FsrRcasDepackHx2(out FfxFloat16x4 pix0,out FfxFloat16x4 pix1,FfxFloat16x2 pixR,FfxFloat16x2 pixG,FfxFloat16x2 pixB){ - #ifdef FFX_HLSL - // Invoke a slower path for DX only, since it won't allow uninitialized values. - pix0.a=pix1.a=0.0; - #endif - pix0.rgb=FfxFloat16x3(pixR.x,pixG.x,pixB.x); - pix1.rgb=FfxFloat16x3(pixR.y,pixG.y,pixB.y);} -//------------------------------------------------------------------------------------------------------------------------------ - void FsrRcasHx2( - // Output values are for 2 8x8 tiles in a 16x8 region. - // pix.x = left 8x8 tile - // pix.y = right 8x8 tile - // This enables later processing to easily be packed as well. - out FfxFloat16x2 pixR, - out FfxFloat16x2 pixG, - out FfxFloat16x2 pixB, - #ifdef FSR_RCAS_PASSTHROUGH_ALPHA - out FfxFloat16x2 pixA, - #endif - FfxUInt32x2 ip, // Integer pixel position in output. - FfxUInt32x4 con){ // Constant generated by RcasSetup(). - // No scaling algorithm uses minimal 3x3 pixel neighborhood. - FfxInt16x2 sp0=FfxInt16x2(ip); - FfxFloat16x3 b0=FsrRcasLoadHx2(sp0+FfxInt16x2( 0,-1)).rgb; - FfxFloat16x3 d0=FsrRcasLoadHx2(sp0+FfxInt16x2(-1, 0)).rgb; - #ifdef FSR_RCAS_PASSTHROUGH_ALPHA - FfxFloat16x4 ee0=FsrRcasLoadHx2(sp0); - FfxFloat16x3 e0=ee0.rgb;pixA.r=ee0.a; - #else - FfxFloat16x3 e0=FsrRcasLoadHx2(sp0).rgb; - #endif - FfxFloat16x3 f0=FsrRcasLoadHx2(sp0+FfxInt16x2( 1, 0)).rgb; - FfxFloat16x3 h0=FsrRcasLoadHx2(sp0+FfxInt16x2( 0, 1)).rgb; - FfxInt16x2 sp1=sp0+FfxInt16x2(8,0); - FfxFloat16x3 b1=FsrRcasLoadHx2(sp1+FfxInt16x2( 0,-1)).rgb; - FfxFloat16x3 d1=FsrRcasLoadHx2(sp1+FfxInt16x2(-1, 0)).rgb; - #ifdef FSR_RCAS_PASSTHROUGH_ALPHA - FfxFloat16x4 ee1=FsrRcasLoadHx2(sp1); - FfxFloat16x3 e1=ee1.rgb;pixA.g=ee1.a; - #else - FfxFloat16x3 e1=FsrRcasLoadHx2(sp1).rgb; - #endif - FfxFloat16x3 f1=FsrRcasLoadHx2(sp1+FfxInt16x2( 1, 0)).rgb; - FfxFloat16x3 h1=FsrRcasLoadHx2(sp1+FfxInt16x2( 0, 1)).rgb; - // Arrays of Structures to Structures of Arrays conversion. - FfxFloat16x2 bR=FfxFloat16x2(b0.r,b1.r); - FfxFloat16x2 bG=FfxFloat16x2(b0.g,b1.g); - FfxFloat16x2 bB=FfxFloat16x2(b0.b,b1.b); - FfxFloat16x2 dR=FfxFloat16x2(d0.r,d1.r); - FfxFloat16x2 dG=FfxFloat16x2(d0.g,d1.g); - FfxFloat16x2 dB=FfxFloat16x2(d0.b,d1.b); - FfxFloat16x2 eR=FfxFloat16x2(e0.r,e1.r); - FfxFloat16x2 eG=FfxFloat16x2(e0.g,e1.g); - FfxFloat16x2 eB=FfxFloat16x2(e0.b,e1.b); - FfxFloat16x2 fR=FfxFloat16x2(f0.r,f1.r); - FfxFloat16x2 fG=FfxFloat16x2(f0.g,f1.g); - FfxFloat16x2 fB=FfxFloat16x2(f0.b,f1.b); - FfxFloat16x2 hR=FfxFloat16x2(h0.r,h1.r); - FfxFloat16x2 hG=FfxFloat16x2(h0.g,h1.g); - FfxFloat16x2 hB=FfxFloat16x2(h0.b,h1.b); - // Run optional input transform. - FsrRcasInputHx2(bR,bG,bB); - FsrRcasInputHx2(dR,dG,dB); - FsrRcasInputHx2(eR,eG,eB); - FsrRcasInputHx2(fR,fG,fB); - FsrRcasInputHx2(hR,hG,hB); - // Luma times 2. - FfxFloat16x2 bL=bB*FFX_BROADCAST_FLOAT16X2(0.5)+(bR*FFX_BROADCAST_FLOAT16X2(0.5)+bG); - FfxFloat16x2 dL=dB*FFX_BROADCAST_FLOAT16X2(0.5)+(dR*FFX_BROADCAST_FLOAT16X2(0.5)+dG); - FfxFloat16x2 eL=eB*FFX_BROADCAST_FLOAT16X2(0.5)+(eR*FFX_BROADCAST_FLOAT16X2(0.5)+eG); - FfxFloat16x2 fL=fB*FFX_BROADCAST_FLOAT16X2(0.5)+(fR*FFX_BROADCAST_FLOAT16X2(0.5)+fG); - FfxFloat16x2 hL=hB*FFX_BROADCAST_FLOAT16X2(0.5)+(hR*FFX_BROADCAST_FLOAT16X2(0.5)+hG); - // Noise detection. - FfxFloat16x2 nz=FFX_BROADCAST_FLOAT16X2(0.25)*bL+FFX_BROADCAST_FLOAT16X2(0.25)*dL+FFX_BROADCAST_FLOAT16X2(0.25)*fL+FFX_BROADCAST_FLOAT16X2(0.25)*hL-eL; - nz=ffxSaturate(abs(nz)*ffxApproximateReciprocalMediumHalf(ffxMax3Half(ffxMax3Half(bL,dL,eL),fL,hL)-ffxMin3Half(ffxMin3Half(bL,dL,eL),fL,hL))); - nz=FFX_BROADCAST_FLOAT16X2(-0.5)*nz+FFX_BROADCAST_FLOAT16X2(1.0); - // Min and max of ring. - FfxFloat16x2 mn4R=min(ffxMin3Half(bR,dR,fR),hR); - FfxFloat16x2 mn4G=min(ffxMin3Half(bG,dG,fG),hG); - FfxFloat16x2 mn4B=min(ffxMin3Half(bB,dB,fB),hB); - FfxFloat16x2 mx4R=max(ffxMax3Half(bR,dR,fR),hR); - FfxFloat16x2 mx4G=max(ffxMax3Half(bG,dG,fG),hG); - FfxFloat16x2 mx4B=max(ffxMax3Half(bB,dB,fB),hB); - // Immediate constants for peak range. - FfxFloat16x2 peakC=FfxFloat16x2(1.0,-1.0*4.0); - // Limiters, these need to be high precision RCPs. - FfxFloat16x2 hitMinR=mn4R*ffxReciprocalHalf(FFX_BROADCAST_FLOAT16X2(4.0)*mx4R); - FfxFloat16x2 hitMinG=mn4G*ffxReciprocalHalf(FFX_BROADCAST_FLOAT16X2(4.0)*mx4G); - FfxFloat16x2 hitMinB=mn4B*ffxReciprocalHalf(FFX_BROADCAST_FLOAT16X2(4.0)*mx4B); - FfxFloat16x2 hitMaxR=(peakC.x-mx4R)*ffxReciprocalHalf(FFX_BROADCAST_FLOAT16X2(4.0)*mn4R+peakC.y); - FfxFloat16x2 hitMaxG=(peakC.x-mx4G)*ffxReciprocalHalf(FFX_BROADCAST_FLOAT16X2(4.0)*mn4G+peakC.y); - FfxFloat16x2 hitMaxB=(peakC.x-mx4B)*ffxReciprocalHalf(FFX_BROADCAST_FLOAT16X2(4.0)*mn4B+peakC.y); - FfxFloat16x2 lobeR=max(-hitMinR,hitMaxR); - FfxFloat16x2 lobeG=max(-hitMinG,hitMaxG); - FfxFloat16x2 lobeB=max(-hitMinB,hitMaxB); - FfxFloat16x2 lobe=max(FFX_BROADCAST_FLOAT16X2(-FSR_RCAS_LIMIT),min(ffxMax3Half(lobeR,lobeG,lobeB),FFX_BROADCAST_FLOAT16X2(0.0)))*FFX_BROADCAST_FLOAT16X2(FFX_UINT32_TO_FLOAT16X2(con.y).x); - // Apply noise removal. - #ifdef FSR_RCAS_DENOISE - lobe*=nz; - #endif - // Resolve, which needs the medium precision rcp approximation to avoid visible tonality changes. - FfxFloat16x2 rcpL=ffxApproximateReciprocalMediumHalf(FFX_BROADCAST_FLOAT16X2(4.0)*lobe+FFX_BROADCAST_FLOAT16X2(1.0)); - pixR=(lobe*bR+lobe*dR+lobe*hR+lobe*fR+eR)*rcpL; - pixG=(lobe*bG+lobe*dG+lobe*hG+lobe*fG+eG)*rcpL; - pixB=(lobe*bB+lobe*dB+lobe*hB+lobe*fB+eB)*rcpL;} -#endif -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// -//_____________________________________________________________/\_______________________________________________________________ -//============================================================================================================================== -// -// FSR - [LFGA] LINEAR FILM GRAIN APPLICATOR -// -//------------------------------------------------------------------------------------------------------------------------------ -// Adding output-resolution film grain after scaling is a good way to mask both rendering and scaling artifacts. -// Suggest using tiled blue noise as film grain input, with peak noise frequency set for a specific look and feel. -// The 'Lfga*()' functions provide a convenient way to introduce grain. -// These functions limit grain based on distance to signal limits. -// This is done so that the grain is temporally energy preserving, and thus won't modify image tonality. -// Grain application should be done in a linear colorspace. -// The grain should be temporally changing, but have a temporal sum per pixel that adds to zero (non-biased). -//------------------------------------------------------------------------------------------------------------------------------ -// Usage, -// FsrLfga*( -// color, // In/out linear colorspace color {0 to 1} ranged. -// grain, // Per pixel grain texture value {-0.5 to 0.5} ranged, input is 3-channel to support colored grain. -// amount); // Amount of grain (0 to 1} ranged. -//------------------------------------------------------------------------------------------------------------------------------ -// Example if grain texture is monochrome: 'FsrLfgaF(color,ffxBroadcast3(grain),amount)' -//============================================================================================================================== -#if defined(FFX_GPU) - // Maximum grain is the minimum distance to the signal limit. - void FsrLfgaF(inout FfxFloat32x3 c, FfxFloat32x3 t, FfxFloat32 a) - { - c += (t * ffxBroadcast3(a)) * ffxMin(ffxBroadcast3(1.0) - c, c); - } -#endif -//============================================================================================================================== -#if defined(FFX_GPU)&& FFX_HALF == 1 - // Half precision version (slower). - void FsrLfgaH(inout FfxFloat16x3 c, FfxFloat16x3 t, FfxFloat16 a) - { - c += (t * FFX_BROADCAST_FLOAT16X3(a)) * min(FFX_BROADCAST_FLOAT16X3(1.0) - c, c); - } - //------------------------------------------------------------------------------------------------------------------------------ - // Packed half precision version (faster). - void FsrLfgaHx2(inout FfxFloat16x2 cR,inout FfxFloat16x2 cG,inout FfxFloat16x2 cB,FfxFloat16x2 tR,FfxFloat16x2 tG,FfxFloat16x2 tB,FfxFloat16 a){ - cR+=(tR*FFX_BROADCAST_FLOAT16X2(a))*min(FFX_BROADCAST_FLOAT16X2(1.0)-cR,cR);cG+=(tG*FFX_BROADCAST_FLOAT16X2(a))*min(FFX_BROADCAST_FLOAT16X2(1.0)-cG,cG);cB+=(tB*FFX_BROADCAST_FLOAT16X2(a))*min(FFX_BROADCAST_FLOAT16X2(1.0)-cB,cB);} -#endif -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// -//_____________________________________________________________/\_______________________________________________________________ -//============================================================================================================================== -// -// FSR - [SRTM] SIMPLE REVERSIBLE TONE-MAPPER -// -//------------------------------------------------------------------------------------------------------------------------------ -// This provides a way to take linear HDR color {0 to FP16_MAX} and convert it into a temporary {0 to 1} ranged post-tonemapped linear. -// The tonemapper preserves RGB ratio, which helps maintain HDR color bleed during filtering. -//------------------------------------------------------------------------------------------------------------------------------ -// Reversible tonemapper usage, -// FsrSrtm*(color); // {0 to FP16_MAX} converted to {0 to 1}. -// FsrSrtmInv*(color); // {0 to 1} converted into {0 to 32768, output peak safe for FP16}. -//============================================================================================================================== -#if defined(FFX_GPU) - void FsrSrtmF(inout FfxFloat32x3 c) - { - c *= ffxBroadcast3(rcp(ffxMax3(c.r, c.g, c.b) + FfxFloat32(1.0))); - } - // The extra max solves the c=1.0 case (which is a /0). - void FsrSrtmInvF(inout FfxFloat32x3 c){c*=ffxBroadcast3(rcp(max(FfxFloat32(1.0/32768.0),FfxFloat32(1.0)-ffxMax3(c.r,c.g,c.b))));} -#endif -//============================================================================================================================== -#if defined(FFX_GPU )&& FFX_HALF == 1 - void FsrSrtmH(inout FfxFloat16x3 c) - { - c *= FFX_BROADCAST_FLOAT16X3(ffxReciprocalHalf(ffxMax3Half(c.r, c.g, c.b) + FFX_BROADCAST_FLOAT16(1.0))); - } - void FsrSrtmInvH(inout FfxFloat16x3 c) - { - c *= FFX_BROADCAST_FLOAT16X3(ffxReciprocalHalf(max(FFX_BROADCAST_FLOAT16(1.0 / 32768.0), FFX_BROADCAST_FLOAT16(1.0) - ffxMax3Half(c.r, c.g, c.b)))); - } - //------------------------------------------------------------------------------------------------------------------------------ - void FsrSrtmHx2(inout FfxFloat16x2 cR, inout FfxFloat16x2 cG, inout FfxFloat16x2 cB) - { - FfxFloat16x2 rcp = ffxReciprocalHalf(ffxMax3Half(cR, cG, cB) + FFX_BROADCAST_FLOAT16X2(1.0)); - cR *= rcp; - cG *= rcp; - cB *= rcp; - } - void FsrSrtmInvHx2(inout FfxFloat16x2 cR,inout FfxFloat16x2 cG,inout FfxFloat16x2 cB) - { - FfxFloat16x2 rcp=ffxReciprocalHalf(max(FFX_BROADCAST_FLOAT16X2(1.0/32768.0),FFX_BROADCAST_FLOAT16X2(1.0)-ffxMax3Half(cR,cG,cB))); - cR*=rcp; - cG*=rcp; - cB*=rcp; - } -#endif -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// -//_____________________________________________________________/\_______________________________________________________________ -//============================================================================================================================== -// -// FSR - [TEPD] TEMPORAL ENERGY PRESERVING DITHER -// -//------------------------------------------------------------------------------------------------------------------------------ -// Temporally energy preserving dithered {0 to 1} linear to gamma 2.0 conversion. -// Gamma 2.0 is used so that the conversion back to linear is just to square the color. -// The conversion comes in 8-bit and 10-bit modes, designed for output to 8-bit UNORM or 10:10:10:2 respectively. -// Given good non-biased temporal blue noise as dither input, -// the output dither will temporally conserve energy. -// This is done by choosing the linear nearest step point instead of perceptual nearest. -// See code below for details. -//------------------------------------------------------------------------------------------------------------------------------ -// DX SPEC RULES FOR FLOAT->UNORM 8-BIT CONVERSION -// =============================================== -// - Output is 'FfxUInt32(floor(saturate(n)*255.0+0.5))'. -// - Thus rounding is to nearest. -// - NaN gets converted to zero. -// - INF is clamped to {0.0 to 1.0}. -//============================================================================================================================== -#if defined(FFX_GPU) - // Hand tuned integer position to dither value, with more values than simple checkerboard. - // Only 32-bit has enough precision for this compddation. - // Output is {0 to <1}. - FfxFloat32 FsrTepdDitF(FfxUInt32x2 p, FfxUInt32 f) - { - FfxFloat32 x = FfxFloat32(p.x + f); - FfxFloat32 y = FfxFloat32(p.y); - // The 1.61803 golden ratio. - FfxFloat32 a = FfxFloat32((1.0 + ffxSqrt(5.0f)) / 2.0); - // Number designed to provide a good visual pattern. - FfxFloat32 b = FfxFloat32(1.0 / 3.69); - x = x * a + (y * b); - return ffxFract(x); - } - //------------------------------------------------------------------------------------------------------------------------------ - // This version is 8-bit gamma 2.0. - // The 'c' input is {0 to 1}. - // Output is {0 to 1} ready for image store. - void FsrTepdC8F(inout FfxFloat32x3 c, FfxFloat32 dit) - { - FfxFloat32x3 n = ffxSqrt(c); - n = floor(n * ffxBroadcast3(255.0)) * ffxBroadcast3(1.0 / 255.0); - FfxFloat32x3 a = n * n; - FfxFloat32x3 b = n + ffxBroadcast3(1.0 / 255.0); - b = b * b; - // Ratio of 'a' to 'b' required to produce 'c'. - // ffxApproximateReciprocal() won't work here (at least for very high dynamic ranges). - // ffxApproximateReciprocalMedium() is an IADD,FMA,MUL. - FfxFloat32x3 r = (c - b) * ffxApproximateReciprocalMedium(a - b); - // Use the ratio as a cutoff to choose 'a' or 'b'. - // ffxIsGreaterThanZero() is a MUL. - c = ffxSaturate(n + ffxIsGreaterThanZero(ffxBroadcast3(dit) - r) * ffxBroadcast3(1.0 / 255.0)); - } - //------------------------------------------------------------------------------------------------------------------------------ - // This version is 10-bit gamma 2.0. - // The 'c' input is {0 to 1}. - // Output is {0 to 1} ready for image store. - void FsrTepdC10F(inout FfxFloat32x3 c, FfxFloat32 dit) - { - FfxFloat32x3 n = ffxSqrt(c); - n = floor(n * ffxBroadcast3(1023.0)) * ffxBroadcast3(1.0 / 1023.0); - FfxFloat32x3 a = n * n; - FfxFloat32x3 b = n + ffxBroadcast3(1.0 / 1023.0); - b = b * b; - FfxFloat32x3 r = (c - b) * ffxApproximateReciprocalMedium(a - b); - c = ffxSaturate(n + ffxIsGreaterThanZero(ffxBroadcast3(dit) - r) * ffxBroadcast3(1.0 / 1023.0)); - } -#endif -//============================================================================================================================== -#if defined(FFX_GPU)&& FFX_HALF == 1 - FfxFloat16 FsrTepdDitH(FfxUInt32x2 p, FfxUInt32 f) - { - FfxFloat32 x = FfxFloat32(p.x + f); - FfxFloat32 y = FfxFloat32(p.y); - FfxFloat32 a = FfxFloat32((1.0 + ffxSqrt(5.0f)) / 2.0); - FfxFloat32 b = FfxFloat32(1.0 / 3.69); - x = x * a + (y * b); - return FfxFloat16(ffxFract(x)); - } - //------------------------------------------------------------------------------------------------------------------------------ - void FsrTepdC8H(inout FfxFloat16x3 c, FfxFloat16 dit) - { - FfxFloat16x3 n = sqrt(c); - n = floor(n * FFX_BROADCAST_FLOAT16X3(255.0)) * FFX_BROADCAST_FLOAT16X3(1.0 / 255.0); - FfxFloat16x3 a = n * n; - FfxFloat16x3 b = n + FFX_BROADCAST_FLOAT16X3(1.0 / 255.0); - b = b * b; - FfxFloat16x3 r = (c - b) * ffxApproximateReciprocalMediumHalf(a - b); - c = ffxSaturate(n + ffxIsGreaterThanZeroHalf(FFX_BROADCAST_FLOAT16X3(dit) - r) * FFX_BROADCAST_FLOAT16X3(1.0 / 255.0)); - } - //------------------------------------------------------------------------------------------------------------------------------ - void FsrTepdC10H(inout FfxFloat16x3 c, FfxFloat16 dit) - { - FfxFloat16x3 n = sqrt(c); - n = floor(n * FFX_BROADCAST_FLOAT16X3(1023.0)) * FFX_BROADCAST_FLOAT16X3(1.0 / 1023.0); - FfxFloat16x3 a = n * n; - FfxFloat16x3 b = n + FFX_BROADCAST_FLOAT16X3(1.0 / 1023.0); - b = b * b; - FfxFloat16x3 r = (c - b) * ffxApproximateReciprocalMediumHalf(a - b); - c = ffxSaturate(n + ffxIsGreaterThanZeroHalf(FFX_BROADCAST_FLOAT16X3(dit) - r) * FFX_BROADCAST_FLOAT16X3(1.0 / 1023.0)); - } - //============================================================================================================================== - // This computes dither for positions 'p' and 'p+{8,0}'. - FfxFloat16x2 FsrTepdDitHx2(FfxUInt32x2 p, FfxUInt32 f) - { - FfxFloat32x2 x; - x.x = FfxFloat32(p.x + f); - x.y = x.x + FfxFloat32(8.0); - FfxFloat32 y = FfxFloat32(p.y); - FfxFloat32 a = FfxFloat32((1.0 + ffxSqrt(5.0f)) / 2.0); - FfxFloat32 b = FfxFloat32(1.0 / 3.69); - x = x * ffxBroadcast2(a) + ffxBroadcast2(y * b); - return FfxFloat16x2(ffxFract(x)); - } - //------------------------------------------------------------------------------------------------------------------------------ - void FsrTepdC8Hx2(inout FfxFloat16x2 cR, inout FfxFloat16x2 cG, inout FfxFloat16x2 cB, FfxFloat16x2 dit) - { - FfxFloat16x2 nR = sqrt(cR); - FfxFloat16x2 nG = sqrt(cG); - FfxFloat16x2 nB = sqrt(cB); - nR = floor(nR * FFX_BROADCAST_FLOAT16X2(255.0)) * FFX_BROADCAST_FLOAT16X2(1.0 / 255.0); - nG = floor(nG * FFX_BROADCAST_FLOAT16X2(255.0)) * FFX_BROADCAST_FLOAT16X2(1.0 / 255.0); - nB = floor(nB * FFX_BROADCAST_FLOAT16X2(255.0)) * FFX_BROADCAST_FLOAT16X2(1.0 / 255.0); - FfxFloat16x2 aR = nR * nR; - FfxFloat16x2 aG = nG * nG; - FfxFloat16x2 aB = nB * nB; - FfxFloat16x2 bR = nR + FFX_BROADCAST_FLOAT16X2(1.0 / 255.0); - bR = bR * bR; - FfxFloat16x2 bG = nG + FFX_BROADCAST_FLOAT16X2(1.0 / 255.0); - bG = bG * bG; - FfxFloat16x2 bB = nB + FFX_BROADCAST_FLOAT16X2(1.0 / 255.0); - bB = bB * bB; - FfxFloat16x2 rR = (cR - bR) * ffxApproximateReciprocalMediumHalf(aR - bR); - FfxFloat16x2 rG = (cG - bG) * ffxApproximateReciprocalMediumHalf(aG - bG); - FfxFloat16x2 rB = (cB - bB) * ffxApproximateReciprocalMediumHalf(aB - bB); - cR = ffxSaturate(nR + ffxIsGreaterThanZeroHalf(dit - rR) * FFX_BROADCAST_FLOAT16X2(1.0 / 255.0)); - cG = ffxSaturate(nG + ffxIsGreaterThanZeroHalf(dit - rG) * FFX_BROADCAST_FLOAT16X2(1.0 / 255.0)); - cB = ffxSaturate(nB + ffxIsGreaterThanZeroHalf(dit - rB) * FFX_BROADCAST_FLOAT16X2(1.0 / 255.0)); - } - //------------------------------------------------------------------------------------------------------------------------------ - void FsrTepdC10Hx2(inout FfxFloat16x2 cR,inout FfxFloat16x2 cG,inout FfxFloat16x2 cB,FfxFloat16x2 dit){ - FfxFloat16x2 nR=sqrt(cR); - FfxFloat16x2 nG=sqrt(cG); - FfxFloat16x2 nB=sqrt(cB); - nR=floor(nR*FFX_BROADCAST_FLOAT16X2(1023.0))*FFX_BROADCAST_FLOAT16X2(1.0/1023.0); - nG=floor(nG*FFX_BROADCAST_FLOAT16X2(1023.0))*FFX_BROADCAST_FLOAT16X2(1.0/1023.0); - nB=floor(nB*FFX_BROADCAST_FLOAT16X2(1023.0))*FFX_BROADCAST_FLOAT16X2(1.0/1023.0); - FfxFloat16x2 aR=nR*nR; - FfxFloat16x2 aG=nG*nG; - FfxFloat16x2 aB=nB*nB; - FfxFloat16x2 bR=nR+FFX_BROADCAST_FLOAT16X2(1.0/1023.0);bR=bR*bR; - FfxFloat16x2 bG=nG+FFX_BROADCAST_FLOAT16X2(1.0/1023.0);bG=bG*bG; - FfxFloat16x2 bB=nB+FFX_BROADCAST_FLOAT16X2(1.0/1023.0);bB=bB*bB; - FfxFloat16x2 rR=(cR-bR)*ffxApproximateReciprocalMediumHalf(aR-bR); - FfxFloat16x2 rG=(cG-bG)*ffxApproximateReciprocalMediumHalf(aG-bG); - FfxFloat16x2 rB=(cB-bB)*ffxApproximateReciprocalMediumHalf(aB-bB); - cR=ffxSaturate(nR+ffxIsGreaterThanZeroHalf(dit-rR)*FFX_BROADCAST_FLOAT16X2(1.0/1023.0)); - cG=ffxSaturate(nG+ffxIsGreaterThanZeroHalf(dit-rG)*FFX_BROADCAST_FLOAT16X2(1.0/1023.0)); - cB = ffxSaturate(nB + ffxIsGreaterThanZeroHalf(dit - rB) * FFX_BROADCAST_FLOAT16X2(1.0 / 1023.0)); -} -#endif diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr1.h.meta b/Assets/Resources/FSR2/shaders/ffx_fsr1.h.meta deleted file mode 100644 index cf43a8b..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr1.h.meta +++ /dev/null @@ -1,60 +0,0 @@ -fileFormatVersion: 2 -guid: 7524e42f73f97f34bbeb414ea412a808 -PluginImporter: - externalObjects: {} - serializedVersion: 2 - iconMap: {} - executionOrder: {} - defineConstraints: [] - isPreloaded: 0 - isOverridable: 0 - isExplicitlyReferenced: 0 - validateReferences: 1 - platformData: - - first: - : Any - second: - enabled: 0 - settings: - Exclude Editor: 1 - Exclude GameCoreScarlett: 1 - Exclude GameCoreXboxOne: 1 - Exclude Linux64: 1 - Exclude OSXUniversal: 1 - Exclude PS4: 1 - Exclude PS5: 1 - Exclude WebGL: 1 - Exclude Win: 1 - Exclude Win64: 1 - - first: - Any: - second: - enabled: 0 - settings: {} - - first: - Editor: Editor - second: - enabled: 0 - settings: - DefaultValueInitialized: true - - first: - Standalone: Linux64 - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win64 - second: - enabled: 0 - settings: - CPU: None - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_accumulate.h b/Assets/Resources/FSR2/shaders/ffx_fsr2_accumulate.h deleted file mode 100644 index 7bd5892..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_accumulate.h +++ /dev/null @@ -1,295 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#ifndef FFX_FSR2_ACCUMULATE_H -#define FFX_FSR2_ACCUMULATE_H - -FfxFloat32 GetPxHrVelocity(FfxFloat32x2 fMotionVector) -{ - return length(fMotionVector * DisplaySize()); -} -#if FFX_HALF -FFX_MIN16_F GetPxHrVelocity(FFX_MIN16_F2 fMotionVector) -{ - return length(fMotionVector * FFX_MIN16_F2(DisplaySize())); -} -#endif - -void Accumulate(const AccumulationPassCommonParams params, FFX_PARAMETER_INOUT FfxFloat32x3 fHistoryColor, FfxFloat32x3 fAccumulation, FFX_PARAMETER_IN FfxFloat32x4 fUpsampledColorAndWeight) -{ - // Aviod invalid values when accumulation and upsampled weight is 0 - fAccumulation = ffxMax(FSR2_EPSILON.xxx, fAccumulation + fUpsampledColorAndWeight.www); - -#if FFX_FSR2_OPTION_HDR_COLOR_INPUT - //YCoCg -> RGB -> Tonemap -> YCoCg (Use RGB tonemapper to avoid color desaturation) - fUpsampledColorAndWeight.xyz = RGBToYCoCg(Tonemap(YCoCgToRGB(fUpsampledColorAndWeight.xyz))); - fHistoryColor = RGBToYCoCg(Tonemap(YCoCgToRGB(fHistoryColor))); -#endif - - const FfxFloat32x3 fAlpha = fUpsampledColorAndWeight.www / fAccumulation; - fHistoryColor = ffxLerp(fHistoryColor, fUpsampledColorAndWeight.xyz, fAlpha); - - fHistoryColor = YCoCgToRGB(fHistoryColor); - -#if FFX_FSR2_OPTION_HDR_COLOR_INPUT - fHistoryColor = InverseTonemap(fHistoryColor); -#endif -} - -void RectifyHistory( - const AccumulationPassCommonParams params, - RectificationBox clippingBox, - FFX_PARAMETER_INOUT FfxFloat32x3 fHistoryColor, - FFX_PARAMETER_INOUT FfxFloat32x3 fAccumulation, - FfxFloat32 fLockContributionThisFrame, - FfxFloat32 fTemporalReactiveFactor, - FfxFloat32 fLumaInstabilityFactor) -{ - FfxFloat32 fScaleFactorInfluence = ffxMin(20.0f, ffxPow(FfxFloat32(1.0f / length(DownscaleFactor().x * DownscaleFactor().y)), 3.0f)); - - const FfxFloat32 fVecolityFactor = ffxSaturate(params.fHrVelocity / 20.0f); - const FfxFloat32 fBoxScaleT = ffxMax(params.fDepthClipFactor, ffxMax(params.fAccumulationMask, fVecolityFactor)); - FfxFloat32 fBoxScale = ffxLerp(fScaleFactorInfluence, 1.0f, fBoxScaleT); - - FfxFloat32x3 fScaledBoxVec = clippingBox.boxVec * fBoxScale; - FfxFloat32x3 boxMin = clippingBox.boxCenter - fScaledBoxVec; - FfxFloat32x3 boxMax = clippingBox.boxCenter + fScaledBoxVec; - FfxFloat32x3 boxCenter = clippingBox.boxCenter; - FfxFloat32 boxVecSize = length(clippingBox.boxVec); - - boxMin = ffxMax(clippingBox.aabbMin, boxMin); - boxMax = ffxMin(clippingBox.aabbMax, boxMax); - - if (any(FFX_GREATER_THAN(boxMin, fHistoryColor)) || any(FFX_GREATER_THAN(fHistoryColor, boxMax))) { - - const FfxFloat32x3 fClampedHistoryColor = clamp(fHistoryColor, boxMin, boxMax); - - FfxFloat32x3 fHistoryContribution = ffxMax(fLumaInstabilityFactor, fLockContributionThisFrame).xxx; - - const FfxFloat32 fReactiveFactor = params.fDilatedReactiveFactor; - const FfxFloat32 fReactiveContribution = 1.0f - ffxPow(fReactiveFactor, 1.0f / 2.0f); - fHistoryContribution *= fReactiveContribution; - - // Scale history color using rectification info, also using accumulation mask to avoid potential invalid color protection - fHistoryColor = ffxLerp(fClampedHistoryColor, fHistoryColor, ffxSaturate(fHistoryContribution)); - - // Scale accumulation using rectification info - const FfxFloat32x3 fAccumulationMin = ffxMin(fAccumulation, FFX_BROADCAST_FLOAT32X3(0.1f)); - fAccumulation = ffxLerp(fAccumulationMin, fAccumulation, ffxSaturate(fHistoryContribution)); - } -} - -void WriteUpscaledOutput(FfxInt32x2 iPxHrPos, FfxFloat32x3 fUpscaledColor) -{ - StoreUpscaledOutput(iPxHrPos, fUpscaledColor); -} - -void FinalizeLockStatus(const AccumulationPassCommonParams params, FfxFloat32x2 fLockStatus, FfxFloat32 fUpsampledWeight) -{ - // we expect similar motion for next frame - // kill lock if that location is outside screen, avoid locks to be clamped to screen borders - FfxFloat32x2 fEstimatedUvNextFrame = params.fHrUv - params.fMotionVector; - if (IsUvInside(fEstimatedUvNextFrame) == false) { - KillLock(fLockStatus); - } - else { - // Decrease lock lifetime - const FfxFloat32 fLifetimeDecreaseLanczosMax = FfxFloat32(JitterSequenceLength()) * FfxFloat32(fAverageLanczosWeightPerFrame); - const FfxFloat32 fLifetimeDecrease = FfxFloat32(fUpsampledWeight / fLifetimeDecreaseLanczosMax); - fLockStatus[LOCK_LIFETIME_REMAINING] = ffxMax(FfxFloat32(0), fLockStatus[LOCK_LIFETIME_REMAINING] - fLifetimeDecrease); - } - - StoreLockStatus(params.iPxHrPos, fLockStatus); -} - - -FfxFloat32x3 ComputeBaseAccumulationWeight(const AccumulationPassCommonParams params, FfxFloat32 fThisFrameReactiveFactor, FfxBoolean bInMotionLastFrame, FfxFloat32 fUpsampledWeight, LockState lockState) -{ - // Always assume max accumulation was reached - FfxFloat32 fBaseAccumulation = fMaxAccumulationLanczosWeight * FfxFloat32(params.bIsExistingSample) * (1.0f - fThisFrameReactiveFactor) * (1.0f - params.fDepthClipFactor); - - fBaseAccumulation = ffxMin(fBaseAccumulation, ffxLerp(fBaseAccumulation, fUpsampledWeight * 10.0f, ffxMax(FfxFloat32(bInMotionLastFrame), ffxSaturate(params.fHrVelocity * FfxFloat32(10))))); - - fBaseAccumulation = ffxMin(fBaseAccumulation, ffxLerp(fBaseAccumulation, fUpsampledWeight, ffxSaturate(params.fHrVelocity / FfxFloat32(20)))); - - return fBaseAccumulation.xxx; -} - -FfxFloat32 ComputeLumaInstabilityFactor(const AccumulationPassCommonParams params, RectificationBox clippingBox, FfxFloat32 fThisFrameReactiveFactor, FfxFloat32 fLuminanceDiff) -{ - const FfxFloat32 fUnormThreshold = 1.0f / 255.0f; - const FfxInt32 N_MINUS_1 = 0; - const FfxInt32 N_MINUS_2 = 1; - const FfxInt32 N_MINUS_3 = 2; - const FfxInt32 N_MINUS_4 = 3; - - FfxFloat32 fCurrentFrameLuma = clippingBox.boxCenter.x; - -#if FFX_FSR2_OPTION_HDR_COLOR_INPUT - fCurrentFrameLuma = fCurrentFrameLuma / (1.0f + ffxMax(0.0f, fCurrentFrameLuma)); -#endif - - fCurrentFrameLuma = round(fCurrentFrameLuma * 255.0f) / 255.0f; - - const FfxBoolean bSampleLumaHistory = (ffxMax(ffxMax(params.fDepthClipFactor, params.fAccumulationMask), fLuminanceDiff) < 0.1f) && (params.bIsNewSample == false); - FfxFloat32x4 fCurrentFrameLumaHistory = bSampleLumaHistory ? SampleLumaHistory(params.fReprojectedHrUv) : FFX_BROADCAST_FLOAT32X4(0.0f); - - FfxFloat32 fLumaInstability = 0.0f; - FfxFloat32 fDiffs0 = (fCurrentFrameLuma - fCurrentFrameLumaHistory[N_MINUS_1]); - - FfxFloat32 fMin = abs(fDiffs0); - - if (fMin >= fUnormThreshold) - { - for (int i = N_MINUS_2; i <= N_MINUS_4; i++) { - FfxFloat32 fDiffs1 = (fCurrentFrameLuma - fCurrentFrameLumaHistory[i]); - - if (sign(fDiffs0) == sign(fDiffs1)) { - - // Scale difference to protect historically similar values - const FfxFloat32 fMinBias = 1.0f; - fMin = ffxMin(fMin, abs(fDiffs1) * fMinBias); - } - } - - const FfxFloat32 fBoxSize = clippingBox.boxVec.x; - const FfxFloat32 fBoxSizeFactor = ffxPow(ffxSaturate(fBoxSize / 0.1f), 6.0f); - - fLumaInstability = FfxFloat32(fMin != abs(fDiffs0)) * fBoxSizeFactor; - fLumaInstability = FfxFloat32(fLumaInstability > fUnormThreshold); - - fLumaInstability *= 1.0f - ffxMax(params.fAccumulationMask, ffxPow(fThisFrameReactiveFactor, 1.0f / 6.0f)); - } - - //shift history - fCurrentFrameLumaHistory[N_MINUS_4] = fCurrentFrameLumaHistory[N_MINUS_3]; - fCurrentFrameLumaHistory[N_MINUS_3] = fCurrentFrameLumaHistory[N_MINUS_2]; - fCurrentFrameLumaHistory[N_MINUS_2] = fCurrentFrameLumaHistory[N_MINUS_1]; - fCurrentFrameLumaHistory[N_MINUS_1] = fCurrentFrameLuma; - - StoreLumaHistory(params.iPxHrPos, fCurrentFrameLumaHistory); - - return fLumaInstability * FfxFloat32(fCurrentFrameLumaHistory[N_MINUS_4] != 0); -} - -FfxFloat32 ComputeTemporalReactiveFactor(const AccumulationPassCommonParams params, FfxFloat32 fTemporalReactiveFactor) -{ - FfxFloat32 fNewFactor = ffxMin(0.99f, fTemporalReactiveFactor); - - fNewFactor = ffxMax(fNewFactor, ffxLerp(fNewFactor, 0.4f, ffxSaturate(params.fHrVelocity))); - - fNewFactor = ffxMax(fNewFactor * fNewFactor, ffxMax(params.fDepthClipFactor * 0.1f, params.fDilatedReactiveFactor)); - - // Force reactive factor for new samples - fNewFactor = params.bIsNewSample ? 1.0f : fNewFactor; - - if (ffxSaturate(params.fHrVelocity * 10.0f) >= 1.0f) { - fNewFactor = ffxMax(FSR2_EPSILON, fNewFactor) * -1.0f; - } - - return fNewFactor; -} - -AccumulationPassCommonParams InitParams(FfxInt32x2 iPxHrPos) -{ - AccumulationPassCommonParams params; - - params.iPxHrPos = iPxHrPos; - const FfxFloat32x2 fHrUv = (iPxHrPos + 0.5f) / DisplaySize(); - params.fHrUv = fHrUv; - - const FfxFloat32x2 fLrUvJittered = fHrUv + Jitter() / RenderSize(); - params.fLrUv_HwSampler = ClampUv(fLrUvJittered, RenderSize(), MaxRenderSize()); - - params.fMotionVector = GetMotionVector(iPxHrPos, fHrUv); - params.fHrVelocity = GetPxHrVelocity(params.fMotionVector); - - ComputeReprojectedUVs(params, params.fReprojectedHrUv, params.bIsExistingSample); - - params.fDepthClipFactor = ffxSaturate(SampleDepthClip(params.fLrUv_HwSampler)); - - const FfxFloat32x2 fDilatedReactiveMasks = SampleDilatedReactiveMasks(params.fLrUv_HwSampler); - params.fDilatedReactiveFactor = fDilatedReactiveMasks.x; - params.fAccumulationMask = fDilatedReactiveMasks.y; - params.bIsResetFrame = (0 == FrameIndex()); - - params.bIsNewSample = (params.bIsExistingSample == false || params.bIsResetFrame); - - return params; -} - -void Accumulate(FfxInt32x2 iPxHrPos) -{ - const AccumulationPassCommonParams params = InitParams(iPxHrPos); - - FfxFloat32x3 fHistoryColor = FfxFloat32x3(0, 0, 0); - FfxFloat32x2 fLockStatus; - InitializeNewLockSample(fLockStatus); - - FfxFloat32 fTemporalReactiveFactor = 0.0f; - FfxBoolean bInMotionLastFrame = FFX_FALSE; - LockState lockState = { FFX_FALSE , FFX_FALSE }; - if (params.bIsExistingSample && !params.bIsResetFrame) { - ReprojectHistoryColor(params, fHistoryColor, fTemporalReactiveFactor, bInMotionLastFrame); - lockState = ReprojectHistoryLockStatus(params, fLockStatus); - } - - FfxFloat32 fThisFrameReactiveFactor = ffxMax(params.fDilatedReactiveFactor, fTemporalReactiveFactor); - - FfxFloat32 fLuminanceDiff = 0.0f; - FfxFloat32 fLockContributionThisFrame = 0.0f; - UpdateLockStatus(params, fThisFrameReactiveFactor, lockState, fLockStatus, fLockContributionThisFrame, fLuminanceDiff); - - // Load upsampled input color - RectificationBox clippingBox; - FfxFloat32x4 fUpsampledColorAndWeight = ComputeUpsampledColorAndWeight(params, clippingBox, fThisFrameReactiveFactor); - - const FfxFloat32 fLumaInstabilityFactor = ComputeLumaInstabilityFactor(params, clippingBox, fThisFrameReactiveFactor, fLuminanceDiff); - - - FfxFloat32x3 fAccumulation = ComputeBaseAccumulationWeight(params, fThisFrameReactiveFactor, bInMotionLastFrame, fUpsampledColorAndWeight.w, lockState); - - if (params.bIsNewSample) { - fHistoryColor = YCoCgToRGB(fUpsampledColorAndWeight.xyz); - } - else { - RectifyHistory(params, clippingBox, fHistoryColor, fAccumulation, fLockContributionThisFrame, fThisFrameReactiveFactor, fLumaInstabilityFactor); - - Accumulate(params, fHistoryColor, fAccumulation, fUpsampledColorAndWeight); - } - - fHistoryColor = UnprepareRgb(fHistoryColor, Exposure()); - - FinalizeLockStatus(params, fLockStatus, fUpsampledColorAndWeight.w); - - // Get new temporal reactive factor - fTemporalReactiveFactor = ComputeTemporalReactiveFactor(params, fThisFrameReactiveFactor); - - StoreInternalColorAndWeight(iPxHrPos, FfxFloat32x4(fHistoryColor, fTemporalReactiveFactor)); - - // Output final color when RCAS is disabled -#if FFX_FSR2_OPTION_APPLY_SHARPENING == 0 - WriteUpscaledOutput(iPxHrPos, fHistoryColor); -#endif - StoreNewLocks(iPxHrPos, 0); -} - -#endif // FFX_FSR2_ACCUMULATE_H diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_accumulate.h.meta b/Assets/Resources/FSR2/shaders/ffx_fsr2_accumulate.h.meta deleted file mode 100644 index 51859c3..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_accumulate.h.meta +++ /dev/null @@ -1,60 +0,0 @@ -fileFormatVersion: 2 -guid: a04cb2522aaff1045869a272ed129964 -PluginImporter: - externalObjects: {} - serializedVersion: 2 - iconMap: {} - executionOrder: {} - defineConstraints: [] - isPreloaded: 0 - isOverridable: 0 - isExplicitlyReferenced: 0 - validateReferences: 1 - platformData: - - first: - : Any - second: - enabled: 0 - settings: - Exclude Editor: 1 - Exclude GameCoreScarlett: 1 - Exclude GameCoreXboxOne: 1 - Exclude Linux64: 1 - Exclude OSXUniversal: 1 - Exclude PS4: 1 - Exclude PS5: 1 - Exclude WebGL: 1 - Exclude Win: 1 - Exclude Win64: 1 - - first: - Any: - second: - enabled: 0 - settings: {} - - first: - Editor: Editor - second: - enabled: 0 - settings: - DefaultValueInitialized: true - - first: - Standalone: Linux64 - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win64 - second: - enabled: 0 - settings: - CPU: None - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_accumulate_pass.hlsl b/Assets/Resources/FSR2/shaders/ffx_fsr2_accumulate_pass.hlsl deleted file mode 100644 index 747f380..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_accumulate_pass.hlsl +++ /dev/null @@ -1,78 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#define FSR2_BIND_SRV_INPUT_EXPOSURE 0 -#define FSR2_BIND_SRV_DILATED_REACTIVE_MASKS 1 -#if FFX_FSR2_OPTION_LOW_RESOLUTION_MOTION_VECTORS -#define FSR2_BIND_SRV_DILATED_MOTION_VECTORS 2 -#else -#define FSR2_BIND_SRV_INPUT_MOTION_VECTORS 2 -#endif -#define FSR2_BIND_SRV_INTERNAL_UPSCALED 3 -#define FSR2_BIND_SRV_LOCK_STATUS 4 -#define FSR2_BIND_SRV_PREPARED_INPUT_COLOR 5 -#define FSR2_BIND_SRV_LANCZOS_LUT 6 -#define FSR2_BIND_SRV_UPSCALE_MAXIMUM_BIAS_LUT 7 -#define FSR2_BIND_SRV_SCENE_LUMINANCE_MIPS 8 -#define FSR2_BIND_SRV_AUTO_EXPOSURE 9 -#define FSR2_BIND_SRV_LUMA_HISTORY 10 - -#define FSR2_BIND_UAV_INTERNAL_UPSCALED 0 -#define FSR2_BIND_UAV_LOCK_STATUS 1 -#define FSR2_BIND_UAV_UPSCALED_OUTPUT 2 -#define FSR2_BIND_UAV_NEW_LOCKS 3 -#define FSR2_BIND_UAV_LUMA_HISTORY 4 - -#define FSR2_BIND_CB_FSR2 0 - -#include "ffx_fsr2_callbacks_hlsl.h" -#include "ffx_fsr2_common.h" -#include "ffx_fsr2_sample.h" -#include "ffx_fsr2_upsample.h" -#include "ffx_fsr2_postprocess_lock_status.h" -#include "ffx_fsr2_reproject.h" -#include "ffx_fsr2_accumulate.h" - -#ifndef FFX_FSR2_THREAD_GROUP_WIDTH -#define FFX_FSR2_THREAD_GROUP_WIDTH 8 -#endif // #ifndef FFX_FSR2_THREAD_GROUP_WIDTH -#ifndef FFX_FSR2_THREAD_GROUP_HEIGHT -#define FFX_FSR2_THREAD_GROUP_HEIGHT 8 -#endif // FFX_FSR2_THREAD_GROUP_HEIGHT -#ifndef FFX_FSR2_THREAD_GROUP_DEPTH -#define FFX_FSR2_THREAD_GROUP_DEPTH 1 -#endif // #ifndef FFX_FSR2_THREAD_GROUP_DEPTH -#ifndef FFX_FSR2_NUM_THREADS -#define FFX_FSR2_NUM_THREADS [numthreads(FFX_FSR2_THREAD_GROUP_WIDTH, FFX_FSR2_THREAD_GROUP_HEIGHT, FFX_FSR2_THREAD_GROUP_DEPTH)] -#endif // #ifndef FFX_FSR2_NUM_THREADS - -FFX_FSR2_PREFER_WAVE64 -FFX_FSR2_NUM_THREADS -FFX_FSR2_EMBED_ROOTSIG_CONTENT -void CS(uint2 uGroupId : SV_GroupID, uint2 uGroupThreadId : SV_GroupThreadID) -{ - const uint GroupRows = (uint(DisplaySize().y) + FFX_FSR2_THREAD_GROUP_HEIGHT - 1) / FFX_FSR2_THREAD_GROUP_HEIGHT; - uGroupId.y = GroupRows - uGroupId.y - 1; - - uint2 uDispatchThreadId = uGroupId * uint2(FFX_FSR2_THREAD_GROUP_WIDTH, FFX_FSR2_THREAD_GROUP_HEIGHT) + uGroupThreadId; - - Accumulate(uDispatchThreadId); -} diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_accumulate_pass.hlsl.meta b/Assets/Resources/FSR2/shaders/ffx_fsr2_accumulate_pass.hlsl.meta deleted file mode 100644 index cad5022..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_accumulate_pass.hlsl.meta +++ /dev/null @@ -1,7 +0,0 @@ -fileFormatVersion: 2 -guid: 356ec46d3f01672428b5a7a0de727548 -ShaderIncludeImporter: - externalObjects: {} - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_autogen_reactive_pass.hlsl b/Assets/Resources/FSR2/shaders/ffx_fsr2_autogen_reactive_pass.hlsl deleted file mode 100644 index a78a8e3..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_autogen_reactive_pass.hlsl +++ /dev/null @@ -1,85 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#define FSR2_BIND_SRV_INPUT_OPAQUE_ONLY 0 -#define FSR2_BIND_SRV_INPUT_COLOR 1 -#define FSR2_BIND_UAV_AUTOREACTIVE 0 - -#define FSR2_BIND_CB_FSR2 0 -#define FSR2_BIND_CB_REACTIVE 1 - -#include "ffx_fsr2_callbacks_hlsl.h" -#include "ffx_fsr2_common.h" - -#ifndef FFX_FSR2_THREAD_GROUP_WIDTH -#define FFX_FSR2_THREAD_GROUP_WIDTH 8 -#endif // #ifndef FFX_FSR2_THREAD_GROUP_WIDTH -#ifndef FFX_FSR2_THREAD_GROUP_HEIGHT -#define FFX_FSR2_THREAD_GROUP_HEIGHT 8 -#endif // FFX_FSR2_THREAD_GROUP_HEIGHT -#ifndef FFX_FSR2_THREAD_GROUP_DEPTH -#define FFX_FSR2_THREAD_GROUP_DEPTH 1 -#endif // #ifndef FFX_FSR2_THREAD_GROUP_DEPTH -#ifndef FFX_FSR2_NUM_THREADS -#define FFX_FSR2_NUM_THREADS [numthreads(FFX_FSR2_THREAD_GROUP_WIDTH, FFX_FSR2_THREAD_GROUP_HEIGHT, FFX_FSR2_THREAD_GROUP_DEPTH)] -#endif // #ifndef FFX_FSR2_NUM_THREADS - -#if defined(FSR2_BIND_CB_REACTIVE) -cbuffer cbGenerateReactive : FFX_FSR2_DECLARE_CB(FSR2_BIND_CB_REACTIVE) -{ - float scale; - float threshold; - float binaryValue; - uint flags; -}; -#endif - -FFX_FSR2_NUM_THREADS -FFX_FSR2_EMBED_ROOTSIG_CONTENT -void CS(uint2 uGroupId : SV_GroupID, uint2 uGroupThreadId : SV_GroupThreadID) -{ - uint2 uDispatchThreadId = uGroupId * uint2(FFX_FSR2_THREAD_GROUP_WIDTH, FFX_FSR2_THREAD_GROUP_HEIGHT) + uGroupThreadId; - - float3 ColorPreAlpha = LoadOpaqueOnly( FFX_MIN16_I2(uDispatchThreadId) ).rgb; - float3 ColorPostAlpha = LoadInputColor(uDispatchThreadId).rgb; - - if (flags & FFX_FSR2_AUTOREACTIVEFLAGS_APPLY_TONEMAP) - { - ColorPreAlpha = Tonemap(ColorPreAlpha); - ColorPostAlpha = Tonemap(ColorPostAlpha); - } - - if (flags & FFX_FSR2_AUTOREACTIVEFLAGS_APPLY_INVERSETONEMAP) - { - ColorPreAlpha = InverseTonemap(ColorPreAlpha); - ColorPostAlpha = InverseTonemap(ColorPostAlpha); - } - - float out_reactive_value = 0.f; - float3 delta = abs(ColorPostAlpha - ColorPreAlpha); - - out_reactive_value = (flags & FFX_FSR2_AUTOREACTIVEFLAGS_USE_COMPONENTS_MAX) ? max(delta.x, max(delta.y, delta.z)) : length(delta); - out_reactive_value *= scale; - - out_reactive_value = (flags & FFX_FSR2_AUTOREACTIVEFLAGS_APPLY_THRESHOLD) ? (out_reactive_value < threshold ? 0 : binaryValue) : out_reactive_value; - - rw_output_autoreactive[uDispatchThreadId] = out_reactive_value; -} diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_autogen_reactive_pass.hlsl.meta b/Assets/Resources/FSR2/shaders/ffx_fsr2_autogen_reactive_pass.hlsl.meta deleted file mode 100644 index 2e93435..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_autogen_reactive_pass.hlsl.meta +++ /dev/null @@ -1,7 +0,0 @@ -fileFormatVersion: 2 -guid: cc76bd6f46792f3418a56b79eb5c959b -ShaderIncludeImporter: - externalObjects: {} - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_callbacks_hlsl.h b/Assets/Resources/FSR2/shaders/ffx_fsr2_callbacks_hlsl.h deleted file mode 100644 index de79c82..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_callbacks_hlsl.h +++ /dev/null @@ -1,817 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#include "ffx_fsr2_resources.h" - -#if defined(FFX_GPU) -#ifdef __hlsl_dx_compiler -#pragma dxc diagnostic push -#pragma dxc diagnostic ignored "-Wambig-lit-shift" -#endif //__hlsl_dx_compiler -#include "ffx_core.h" -#ifdef __hlsl_dx_compiler -#pragma dxc diagnostic pop -#endif //__hlsl_dx_compiler -#endif // #if defined(FFX_GPU) - -#if defined(FFX_GPU) -#ifndef FFX_FSR2_PREFER_WAVE64 -#define FFX_FSR2_PREFER_WAVE64 -#endif // #if defined(FFX_GPU) - -#if defined(FFX_GPU) -#pragma warning(disable: 3205) // conversion from larger type to smaller -#endif // #if defined(FFX_GPU) - -#define DECLARE_SRV_REGISTER(regIndex) t##regIndex -#define DECLARE_UAV_REGISTER(regIndex) u##regIndex -#define DECLARE_CB_REGISTER(regIndex) b##regIndex -#define FFX_FSR2_DECLARE_SRV(regIndex) register(DECLARE_SRV_REGISTER(regIndex)) -#define FFX_FSR2_DECLARE_UAV(regIndex) register(DECLARE_UAV_REGISTER(regIndex)) -#define FFX_FSR2_DECLARE_CB(regIndex) register(DECLARE_CB_REGISTER(regIndex)) - -#if defined(FSR2_BIND_CB_FSR2) || defined(FFX_INTERNAL) - cbuffer cbFSR2 : FFX_FSR2_DECLARE_CB(FSR2_BIND_CB_FSR2) - { - FfxInt32x2 iRenderSize; - FfxInt32x2 iMaxRenderSize; - FfxInt32x2 iDisplaySize; - FfxInt32x2 iInputColorResourceDimensions; - FfxInt32x2 iLumaMipDimensions; - FfxInt32 iLumaMipLevelToUse; - FfxInt32 iFrameIndex; - - FfxFloat32x4 fDeviceToViewDepth; - FfxFloat32x2 fJitter; - FfxFloat32x2 fMotionVectorScale; - FfxFloat32x2 fDownscaleFactor; - FfxFloat32x2 fMotionVectorJitterCancellation; - FfxFloat32 fPreExposure; - FfxFloat32 fPreviousFramePreExposure; - FfxFloat32 fTanHalfFOV; - FfxFloat32 fJitterSequenceLength; - FfxFloat32 fDeltaTime; - FfxFloat32 fDynamicResChangeFactor; - FfxFloat32 fViewSpaceToMetersFactor; - - FfxInt32 iDummy; - }; - -#define FFX_FSR2_CONSTANT_BUFFER_1_SIZE (sizeof(cbFSR2) / 4) // Number of 32-bit values. This must be kept in sync with the cbFSR2 size. -#endif - -#if defined(FFX_GPU) -#define FFX_FSR2_ROOTSIG_STRINGIFY(p) FFX_FSR2_ROOTSIG_STR(p) -#define FFX_FSR2_ROOTSIG_STR(p) #p -#define FFX_FSR2_ROOTSIG [RootSignature( "DescriptorTable(UAV(u0, numDescriptors = " FFX_FSR2_ROOTSIG_STRINGIFY(FFX_FSR2_RESOURCE_IDENTIFIER_COUNT) ")), " \ - "DescriptorTable(SRV(t0, numDescriptors = " FFX_FSR2_ROOTSIG_STRINGIFY(FFX_FSR2_RESOURCE_IDENTIFIER_COUNT) ")), " \ - "RootConstants(num32BitConstants=" FFX_FSR2_ROOTSIG_STRINGIFY(FFX_FSR2_CONSTANT_BUFFER_1_SIZE) ", b0), " \ - "StaticSampler(s0, filter = FILTER_MIN_MAG_MIP_POINT, " \ - "addressU = TEXTURE_ADDRESS_CLAMP, " \ - "addressV = TEXTURE_ADDRESS_CLAMP, " \ - "addressW = TEXTURE_ADDRESS_CLAMP, " \ - "comparisonFunc = COMPARISON_NEVER, " \ - "borderColor = STATIC_BORDER_COLOR_TRANSPARENT_BLACK), " \ - "StaticSampler(s1, filter = FILTER_MIN_MAG_MIP_LINEAR, " \ - "addressU = TEXTURE_ADDRESS_CLAMP, " \ - "addressV = TEXTURE_ADDRESS_CLAMP, " \ - "addressW = TEXTURE_ADDRESS_CLAMP, " \ - "comparisonFunc = COMPARISON_NEVER, " \ - "borderColor = STATIC_BORDER_COLOR_TRANSPARENT_BLACK)" )] - -#define FFX_FSR2_CONSTANT_BUFFER_2_SIZE 6 // Number of 32-bit values. This must be kept in sync with max( cbRCAS , cbSPD) size. - -#define FFX_FSR2_CB2_ROOTSIG [RootSignature( "DescriptorTable(UAV(u0, numDescriptors = " FFX_FSR2_ROOTSIG_STRINGIFY(FFX_FSR2_RESOURCE_IDENTIFIER_COUNT) ")), " \ - "DescriptorTable(SRV(t0, numDescriptors = " FFX_FSR2_ROOTSIG_STRINGIFY(FFX_FSR2_RESOURCE_IDENTIFIER_COUNT) ")), " \ - "RootConstants(num32BitConstants=" FFX_FSR2_ROOTSIG_STRINGIFY(FFX_FSR2_CONSTANT_BUFFER_1_SIZE) ", b0), " \ - "RootConstants(num32BitConstants=" FFX_FSR2_ROOTSIG_STRINGIFY(FFX_FSR2_CONSTANT_BUFFER_2_SIZE) ", b1), " \ - "StaticSampler(s0, filter = FILTER_MIN_MAG_MIP_POINT, " \ - "addressU = TEXTURE_ADDRESS_CLAMP, " \ - "addressV = TEXTURE_ADDRESS_CLAMP, " \ - "addressW = TEXTURE_ADDRESS_CLAMP, " \ - "comparisonFunc = COMPARISON_NEVER, " \ - "borderColor = STATIC_BORDER_COLOR_TRANSPARENT_BLACK), " \ - "StaticSampler(s1, filter = FILTER_MIN_MAG_MIP_LINEAR, " \ - "addressU = TEXTURE_ADDRESS_CLAMP, " \ - "addressV = TEXTURE_ADDRESS_CLAMP, " \ - "addressW = TEXTURE_ADDRESS_CLAMP, " \ - "comparisonFunc = COMPARISON_NEVER, " \ - "borderColor = STATIC_BORDER_COLOR_TRANSPARENT_BLACK)" )] -#if defined(FFX_FSR2_EMBED_ROOTSIG) -#define FFX_FSR2_EMBED_ROOTSIG_CONTENT FFX_FSR2_ROOTSIG -#define FFX_FSR2_EMBED_CB2_ROOTSIG_CONTENT FFX_FSR2_CB2_ROOTSIG -#else -#define FFX_FSR2_EMBED_ROOTSIG_CONTENT -#define FFX_FSR2_EMBED_CB2_ROOTSIG_CONTENT -#endif // #if FFX_FSR2_EMBED_ROOTSIG -#endif // #if defined(FFX_GPU) - -// Declare and sample camera buffers as regular textures, unless overridden -#if !defined(UNITY_FSR2_TEX2D) -#define UNITY_FSR2_TEX2D(type) Texture2D -#endif -#if !defined(UNITY_FSR2_POS) -#define UNITY_FSR2_POS(pxPos) (pxPos) -#endif -#if !defined(UNITY_FSR2_UV) -#define UNITY_FSR2_UV(uv) (uv) -#endif - -/* Define getter functions in the order they are defined in the CB! */ -FfxInt32x2 RenderSize() -{ - return iRenderSize; -} - -FfxInt32x2 MaxRenderSize() -{ - return iMaxRenderSize; -} - -FfxInt32x2 DisplaySize() -{ - return iDisplaySize; -} - -FfxInt32x2 InputColorResourceDimensions() -{ - return iInputColorResourceDimensions; -} - -FfxInt32x2 LumaMipDimensions() -{ - return iLumaMipDimensions; -} - -FfxInt32 LumaMipLevelToUse() -{ - return iLumaMipLevelToUse; -} - -FfxInt32 FrameIndex() -{ - return iFrameIndex; -} - -FfxFloat32x2 Jitter() -{ - return fJitter; -} - -FfxFloat32x4 DeviceToViewSpaceTransformFactors() -{ - return fDeviceToViewDepth; -} - -FfxFloat32x2 MotionVectorScale() -{ - return fMotionVectorScale; -} - -FfxFloat32x2 DownscaleFactor() -{ - return fDownscaleFactor; -} - -FfxFloat32x2 MotionVectorJitterCancellation() -{ - return fMotionVectorJitterCancellation; -} - -FfxFloat32 PreExposure() -{ - return fPreExposure; -} - -FfxFloat32 PreviousFramePreExposure() -{ - return fPreviousFramePreExposure; -} - -FfxFloat32 TanHalfFoV() -{ - return fTanHalfFOV; -} - -FfxFloat32 JitterSequenceLength() -{ - return fJitterSequenceLength; -} - -FfxFloat32 DeltaTime() -{ - return fDeltaTime; -} - -FfxFloat32 DynamicResChangeFactor() -{ - return fDynamicResChangeFactor; -} - -FfxFloat32 ViewSpaceToMetersFactor() -{ - return fViewSpaceToMetersFactor; -} - - -SamplerState s_PointClamp : register(s0); -SamplerState s_LinearClamp : register(s1); - -// SRVs -#if defined(FFX_INTERNAL) - UNITY_FSR2_TEX2D(FfxFloat32x4) r_input_opaque_only : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_INPUT_OPAQUE_ONLY); - UNITY_FSR2_TEX2D(FfxFloat32x4) r_input_color_jittered : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_INPUT_COLOR); - UNITY_FSR2_TEX2D(FfxFloat32x4) r_input_motion_vectors : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_INPUT_MOTION_VECTORS); - UNITY_FSR2_TEX2D(FfxFloat32) r_input_depth : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_INPUT_DEPTH); - Texture2D r_input_exposure : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_INPUT_EXPOSURE); - Texture2D r_auto_exposure : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_AUTO_EXPOSURE); - Texture2D r_reactive_mask : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_INPUT_REACTIVE_MASK); - Texture2D r_transparency_and_composition_mask : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_INPUT_TRANSPARENCY_AND_COMPOSITION_MASK); - Texture2D r_reconstructed_previous_nearest_depth : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_RECONSTRUCTED_PREVIOUS_NEAREST_DEPTH); - Texture2D r_dilated_motion_vectors : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_DILATED_MOTION_VECTORS); - Texture2D r_previous_dilated_motion_vectors : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_PREVIOUS_DILATED_MOTION_VECTORS); - Texture2D r_dilatedDepth : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_DILATED_DEPTH); - Texture2D r_internal_upscaled_color : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_INTERNAL_UPSCALED_COLOR); - Texture2D r_lock_status : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_LOCK_STATUS); - Texture2D r_lock_input_luma : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_LOCK_INPUT_LUMA); - Texture2D r_new_locks : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_NEW_LOCKS); - Texture2D r_prepared_input_color : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_PREPARED_INPUT_COLOR); - Texture2D r_luma_history : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_LUMA_HISTORY); - Texture2D r_rcas_input : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_RCAS_INPUT); - Texture2D r_lanczos_lut : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_LANCZOS_LUT); - Texture2D r_imgMips : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE); - Texture2D r_upsample_maximum_bias_lut : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTITIER_UPSAMPLE_MAXIMUM_BIAS_LUT); - Texture2D r_dilated_reactive_masks : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_DILATED_REACTIVE_MASKS); - Texture2D r_input_prev_color_pre_alpha : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_PREV_PRE_ALPHA_COLOR); - Texture2D r_input_prev_color_post_alpha : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_PREV_POST_ALPHA_COLOR); - - Texture2D r_debug_out : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_DEBUG_OUTPUT); - - // UAV declarations - RWTexture2D rw_reconstructed_previous_nearest_depth : FFX_FSR2_DECLARE_UAV(FFX_FSR2_RESOURCE_IDENTIFIER_RECONSTRUCTED_PREVIOUS_NEAREST_DEPTH); - RWTexture2D rw_dilated_motion_vectors : FFX_FSR2_DECLARE_UAV(FFX_FSR2_RESOURCE_IDENTIFIER_DILATED_MOTION_VECTORS); - RWTexture2D rw_dilatedDepth : FFX_FSR2_DECLARE_UAV(FFX_FSR2_RESOURCE_IDENTIFIER_DILATED_DEPTH); - RWTexture2D rw_internal_upscaled_color : FFX_FSR2_DECLARE_UAV(FFX_FSR2_RESOURCE_IDENTIFIER_INTERNAL_UPSCALED_COLOR); - RWTexture2D rw_lock_status : FFX_FSR2_DECLARE_UAV(FFX_FSR2_RESOURCE_IDENTIFIER_LOCK_STATUS); - RWTexture2D rw_lock_input_luma : FFX_FSR2_DECLARE_UAV(FFX_FSR2_RESOURCE_IDENTIFIER_LOCK_INPUT_LUMA); - RWTexture2D rw_new_locks : FFX_FSR2_DECLARE_UAV(FFX_FSR2_RESOURCE_IDENTIFIER_NEW_LOCKS); - RWTexture2D rw_prepared_input_color : FFX_FSR2_DECLARE_UAV(FFX_FSR2_RESOURCE_IDENTIFIER_PREPARED_INPUT_COLOR); - RWTexture2D rw_luma_history : FFX_FSR2_DECLARE_UAV(FFX_FSR2_RESOURCE_IDENTIFIER_LUMA_HISTORY); - RWTexture2D rw_upscaled_output : FFX_FSR2_DECLARE_UAV(FFX_FSR2_RESOURCE_IDENTIFIER_UPSCALED_OUTPUT); - - globallycoherent RWTexture2D rw_img_mip_shading_change : FFX_FSR2_DECLARE_UAV(FFX_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_SHADING_CHANGE); - globallycoherent RWTexture2D rw_img_mip_5 : FFX_FSR2_DECLARE_UAV(FFX_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_5); - RWTexture2D rw_dilated_reactive_masks : FFX_FSR2_DECLARE_UAV(FFX_FSR2_RESOURCE_IDENTIFIER_DILATED_REACTIVE_MASKS); - RWTexture2D rw_auto_exposure : FFX_FSR2_DECLARE_UAV(FFX_FSR2_RESOURCE_IDENTIFIER_AUTO_EXPOSURE); - globallycoherent RWTexture2D rw_spd_global_atomic : FFX_FSR2_DECLARE_UAV(FFX_FSR2_RESOURCE_IDENTIFIER_SPD_ATOMIC_COUNT); - RWTexture2D rw_debug_out : FFX_FSR2_DECLARE_UAV(FFX_FSR2_RESOURCE_IDENTIFIER_DEBUG_OUTPUT); - - RWTexture2D rw_output_autoreactive : FFX_FSR2_DECLARE_UAV(FFX_FSR2_RESOURCE_IDENTIFIER_AUTOREACTIVE); - RWTexture2D rw_output_autocomposition : FFX_FSR2_DECLARE_UAV(FFX_FSR2_RESOURCE_IDENTIFIER_AUTOCOMPOSITION); - RWTexture2D rw_output_prev_color_pre_alpha : FFX_FSR2_DECLARE_UAV(FFX_FSR2_RESOURCE_IDENTIFIER_PREV_PRE_ALPHA_COLOR); - RWTexture2D rw_output_prev_color_post_alpha : FFX_FSR2_DECLARE_UAV(FFX_FSR2_RESOURCE_IDENTIFIER_PREV_POST_ALPHA_COLOR); - -#else // #if defined(FFX_INTERNAL) - #if defined FSR2_BIND_SRV_INPUT_COLOR - UNITY_FSR2_TEX2D(FfxFloat32x4) r_input_color_jittered : FFX_FSR2_DECLARE_SRV(FSR2_BIND_SRV_INPUT_COLOR); - #endif - #if defined FSR2_BIND_SRV_INPUT_OPAQUE_ONLY - UNITY_FSR2_TEX2D(FfxFloat32x4) r_input_opaque_only : FFX_FSR2_DECLARE_SRV(FSR2_BIND_SRV_INPUT_OPAQUE_ONLY); - #endif - #if defined FSR2_BIND_SRV_INPUT_MOTION_VECTORS - UNITY_FSR2_TEX2D(FfxFloat32x4) r_input_motion_vectors : FFX_FSR2_DECLARE_SRV(FSR2_BIND_SRV_INPUT_MOTION_VECTORS); - #endif - #if defined FSR2_BIND_SRV_INPUT_DEPTH - UNITY_FSR2_TEX2D(FfxFloat32) r_input_depth : FFX_FSR2_DECLARE_SRV(FSR2_BIND_SRV_INPUT_DEPTH); - #endif - #if defined FSR2_BIND_SRV_INPUT_EXPOSURE - Texture2D r_input_exposure : FFX_FSR2_DECLARE_SRV(FSR2_BIND_SRV_INPUT_EXPOSURE); - #endif - #if defined FSR2_BIND_SRV_AUTO_EXPOSURE - Texture2D r_auto_exposure : FFX_FSR2_DECLARE_SRV(FSR2_BIND_SRV_AUTO_EXPOSURE); - #endif - #if defined FSR2_BIND_SRV_REACTIVE_MASK - Texture2D r_reactive_mask : FFX_FSR2_DECLARE_SRV(FSR2_BIND_SRV_REACTIVE_MASK); - #endif - #if defined FSR2_BIND_SRV_TRANSPARENCY_AND_COMPOSITION_MASK - Texture2D r_transparency_and_composition_mask : FFX_FSR2_DECLARE_SRV(FSR2_BIND_SRV_TRANSPARENCY_AND_COMPOSITION_MASK); - #endif - #if defined FSR2_BIND_SRV_RECONSTRUCTED_PREV_NEAREST_DEPTH - Texture2D r_reconstructed_previous_nearest_depth : FFX_FSR2_DECLARE_SRV(FSR2_BIND_SRV_RECONSTRUCTED_PREV_NEAREST_DEPTH); - #endif - #if defined FSR2_BIND_SRV_DILATED_MOTION_VECTORS - Texture2D r_dilated_motion_vectors : FFX_FSR2_DECLARE_SRV(FSR2_BIND_SRV_DILATED_MOTION_VECTORS); - #endif - #if defined FSR2_BIND_SRV_PREVIOUS_DILATED_MOTION_VECTORS - Texture2D r_previous_dilated_motion_vectors : FFX_FSR2_DECLARE_SRV(FSR2_BIND_SRV_PREVIOUS_DILATED_MOTION_VECTORS); - #endif - #if defined FSR2_BIND_SRV_DILATED_DEPTH - Texture2D r_dilatedDepth : FFX_FSR2_DECLARE_SRV(FSR2_BIND_SRV_DILATED_DEPTH); - #endif - #if defined FSR2_BIND_SRV_INTERNAL_UPSCALED - Texture2D r_internal_upscaled_color : FFX_FSR2_DECLARE_SRV(FSR2_BIND_SRV_INTERNAL_UPSCALED); - #endif - #if defined FSR2_BIND_SRV_LOCK_STATUS - Texture2D r_lock_status : FFX_FSR2_DECLARE_SRV(FSR2_BIND_SRV_LOCK_STATUS); - #endif - #if defined FSR2_BIND_SRV_LOCK_INPUT_LUMA - Texture2D r_lock_input_luma : FFX_FSR2_DECLARE_SRV(FSR2_BIND_SRV_LOCK_INPUT_LUMA); - #endif - #if defined FSR2_BIND_SRV_NEW_LOCKS - Texture2D r_new_locks : FFX_FSR2_DECLARE_SRV(FSR2_BIND_SRV_NEW_LOCKS); - #endif - #if defined FSR2_BIND_SRV_PREPARED_INPUT_COLOR - Texture2D r_prepared_input_color : FFX_FSR2_DECLARE_SRV(FSR2_BIND_SRV_PREPARED_INPUT_COLOR); - #endif - #if defined FSR2_BIND_SRV_LUMA_HISTORY - Texture2D r_luma_history : FFX_FSR2_DECLARE_SRV(FSR2_BIND_SRV_LUMA_HISTORY); - #endif - #if defined FSR2_BIND_SRV_RCAS_INPUT - Texture2D r_rcas_input : FFX_FSR2_DECLARE_SRV(FSR2_BIND_SRV_RCAS_INPUT); - #endif - #if defined FSR2_BIND_SRV_LANCZOS_LUT - Texture2D r_lanczos_lut : FFX_FSR2_DECLARE_SRV(FSR2_BIND_SRV_LANCZOS_LUT); - #endif - #if defined FSR2_BIND_SRV_SCENE_LUMINANCE_MIPS - Texture2D r_imgMips : FFX_FSR2_DECLARE_SRV(FSR2_BIND_SRV_SCENE_LUMINANCE_MIPS); - #endif - #if defined FSR2_BIND_SRV_UPSCALE_MAXIMUM_BIAS_LUT - Texture2D r_upsample_maximum_bias_lut : FFX_FSR2_DECLARE_SRV(FSR2_BIND_SRV_UPSCALE_MAXIMUM_BIAS_LUT); - #endif - #if defined FSR2_BIND_SRV_DILATED_REACTIVE_MASKS - Texture2D r_dilated_reactive_masks : FFX_FSR2_DECLARE_SRV(FSR2_BIND_SRV_DILATED_REACTIVE_MASKS); - #endif - - #if defined FSR2_BIND_SRV_PREV_PRE_ALPHA_COLOR - Texture2D r_input_prev_color_pre_alpha : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_PREV_PRE_ALPHA_COLOR); - #endif - #if defined FSR2_BIND_SRV_PREV_POST_ALPHA_COLOR - Texture2D r_input_prev_color_post_alpha : FFX_FSR2_DECLARE_SRV(FFX_FSR2_RESOURCE_IDENTIFIER_PREV_POST_ALPHA_COLOR); - #endif - - // UAV declarations - #if defined FSR2_BIND_UAV_RECONSTRUCTED_PREV_NEAREST_DEPTH - RWTexture2D rw_reconstructed_previous_nearest_depth : FFX_FSR2_DECLARE_UAV(FSR2_BIND_UAV_RECONSTRUCTED_PREV_NEAREST_DEPTH); - #endif - #if defined FSR2_BIND_UAV_DILATED_MOTION_VECTORS - RWTexture2D rw_dilated_motion_vectors : FFX_FSR2_DECLARE_UAV(FSR2_BIND_UAV_DILATED_MOTION_VECTORS); - #endif - #if defined FSR2_BIND_UAV_DILATED_DEPTH - RWTexture2D rw_dilatedDepth : FFX_FSR2_DECLARE_UAV(FSR2_BIND_UAV_DILATED_DEPTH); - #endif - #if defined FSR2_BIND_UAV_INTERNAL_UPSCALED - RWTexture2D rw_internal_upscaled_color : FFX_FSR2_DECLARE_UAV(FSR2_BIND_UAV_INTERNAL_UPSCALED); - #endif - #if defined FSR2_BIND_UAV_LOCK_STATUS - RWTexture2D rw_lock_status : FFX_FSR2_DECLARE_UAV(FSR2_BIND_UAV_LOCK_STATUS); - #endif - #if defined FSR2_BIND_UAV_LOCK_INPUT_LUMA - RWTexture2D rw_lock_input_luma : FFX_FSR2_DECLARE_UAV(FSR2_BIND_UAV_LOCK_INPUT_LUMA); - #endif - #if defined FSR2_BIND_UAV_NEW_LOCKS - RWTexture2D rw_new_locks : FFX_FSR2_DECLARE_UAV(FSR2_BIND_UAV_NEW_LOCKS); - #endif - #if defined FSR2_BIND_UAV_PREPARED_INPUT_COLOR - RWTexture2D rw_prepared_input_color : FFX_FSR2_DECLARE_UAV(FSR2_BIND_UAV_PREPARED_INPUT_COLOR); - #endif - #if defined FSR2_BIND_UAV_LUMA_HISTORY - RWTexture2D rw_luma_history : FFX_FSR2_DECLARE_UAV(FSR2_BIND_UAV_LUMA_HISTORY); - #endif - #if defined FSR2_BIND_UAV_UPSCALED_OUTPUT - RWTexture2D rw_upscaled_output : FFX_FSR2_DECLARE_UAV(FSR2_BIND_UAV_UPSCALED_OUTPUT); - #endif - #if defined FSR2_BIND_UAV_EXPOSURE_MIP_LUMA_CHANGE - globallycoherent RWTexture2D rw_img_mip_shading_change : FFX_FSR2_DECLARE_UAV(FSR2_BIND_UAV_EXPOSURE_MIP_LUMA_CHANGE); - #endif - #if defined FSR2_BIND_UAV_EXPOSURE_MIP_5 - globallycoherent RWTexture2D rw_img_mip_5 : FFX_FSR2_DECLARE_UAV(FSR2_BIND_UAV_EXPOSURE_MIP_5); - #endif - #if defined FSR2_BIND_UAV_DILATED_REACTIVE_MASKS - RWTexture2D rw_dilated_reactive_masks : FFX_FSR2_DECLARE_UAV(FSR2_BIND_UAV_DILATED_REACTIVE_MASKS); - #endif - #if defined FSR2_BIND_UAV_EXPOSURE - RWTexture2D rw_exposure : FFX_FSR2_DECLARE_UAV(FSR2_BIND_UAV_EXPOSURE); - #endif - #if defined FSR2_BIND_UAV_AUTO_EXPOSURE - RWTexture2D rw_auto_exposure : FFX_FSR2_DECLARE_UAV(FSR2_BIND_UAV_AUTO_EXPOSURE); - #endif - #if defined FSR2_BIND_UAV_SPD_GLOBAL_ATOMIC - globallycoherent RWTexture2D rw_spd_global_atomic : FFX_FSR2_DECLARE_UAV(FSR2_BIND_UAV_SPD_GLOBAL_ATOMIC); - #endif - - #if defined FSR2_BIND_UAV_AUTOREACTIVE - RWTexture2D rw_output_autoreactive : FFX_FSR2_DECLARE_UAV(FSR2_BIND_UAV_AUTOREACTIVE); - #endif - #if defined FSR2_BIND_UAV_AUTOCOMPOSITION - RWTexture2D rw_output_autocomposition : FFX_FSR2_DECLARE_UAV(FSR2_BIND_UAV_AUTOCOMPOSITION); - #endif - #if defined FSR2_BIND_UAV_PREV_PRE_ALPHA_COLOR - RWTexture2D rw_output_prev_color_pre_alpha : FFX_FSR2_DECLARE_UAV(FSR2_BIND_UAV_PREV_PRE_ALPHA_COLOR); - #endif - #if defined FSR2_BIND_UAV_PREV_POST_ALPHA_COLOR - RWTexture2D rw_output_prev_color_post_alpha : FFX_FSR2_DECLARE_UAV(FSR2_BIND_UAV_PREV_POST_ALPHA_COLOR); - #endif -#endif // #if defined(FFX_INTERNAL) - -#if defined(FSR2_BIND_SRV_SCENE_LUMINANCE_MIPS) || defined(FFX_INTERNAL) -FfxFloat32 LoadMipLuma(FfxUInt32x2 iPxPos, FfxUInt32 mipLevel) -{ - return r_imgMips.mips[mipLevel][iPxPos]; -} -#endif - -#if defined(FSR2_BIND_SRV_SCENE_LUMINANCE_MIPS) || defined(FFX_INTERNAL) -FfxFloat32 SampleMipLuma(FfxFloat32x2 fUV, FfxUInt32 mipLevel) -{ - return r_imgMips.SampleLevel(s_LinearClamp, fUV, mipLevel); -} -#endif - -#if defined(FSR2_BIND_SRV_INPUT_DEPTH) || defined(FFX_INTERNAL) -FfxFloat32 LoadInputDepth(FfxUInt32x2 iPxPos) -{ - return r_input_depth[UNITY_FSR2_POS(iPxPos)]; -} -#endif - -#if defined(FSR2_BIND_SRV_INPUT_DEPTH) || defined(FFX_INTERNAL) -FfxFloat32 SampleInputDepth(FfxFloat32x2 fUV) -{ - return r_input_depth.SampleLevel(s_LinearClamp, UNITY_FSR2_UV(fUV), 0).x; -} -#endif - -#if defined(FSR2_BIND_SRV_REACTIVE_MASK) || defined(FFX_INTERNAL) -FfxFloat32 LoadReactiveMask(FfxUInt32x2 iPxPos) -{ - return r_reactive_mask[iPxPos]; -} -#endif - -#if defined(FSR2_BIND_SRV_TRANSPARENCY_AND_COMPOSITION_MASK) || defined(FFX_INTERNAL) -FfxFloat32 LoadTransparencyAndCompositionMask(FfxUInt32x2 iPxPos) -{ - return r_transparency_and_composition_mask[iPxPos]; -} -#endif - -#if defined(FSR2_BIND_SRV_INPUT_COLOR) || defined(FFX_INTERNAL) -FfxFloat32x3 LoadInputColor(FfxUInt32x2 iPxPos) -{ - return r_input_color_jittered[UNITY_FSR2_POS(iPxPos)].rgb; -} -#endif - -#if defined(FSR2_BIND_SRV_INPUT_COLOR) || defined(FFX_INTERNAL) -FfxFloat32x3 SampleInputColor(FfxFloat32x2 fUV) -{ - return r_input_color_jittered.SampleLevel(s_LinearClamp, UNITY_FSR2_UV(fUV), 0).rgb; -} -#endif - -#if defined(FSR2_BIND_SRV_PREPARED_INPUT_COLOR) || defined(FFX_INTERNAL) -FfxFloat32x3 LoadPreparedInputColor(FfxUInt32x2 iPxPos) -{ - return r_prepared_input_color[iPxPos].xyz; -} -#endif - -#if defined(FSR2_BIND_SRV_INPUT_MOTION_VECTORS) || defined(FFX_INTERNAL) -FfxFloat32x2 LoadInputMotionVector(FfxUInt32x2 iPxDilatedMotionVectorPos) -{ - FfxFloat32x2 fSrcMotionVector = r_input_motion_vectors[UNITY_FSR2_POS(iPxDilatedMotionVectorPos)].xy; - - FfxFloat32x2 fUvMotionVector = fSrcMotionVector * MotionVectorScale(); - -#if FFX_FSR2_OPTION_JITTERED_MOTION_VECTORS - fUvMotionVector -= MotionVectorJitterCancellation(); -#endif - - return fUvMotionVector; -} -#endif - -#if defined(FSR2_BIND_SRV_INTERNAL_UPSCALED) || defined(FFX_INTERNAL) -FfxFloat32x4 LoadHistory(FfxUInt32x2 iPxHistory) -{ - return r_internal_upscaled_color[iPxHistory]; -} -#endif - -#if defined(FSR2_BIND_UAV_LUMA_HISTORY) || defined(FFX_INTERNAL) -void StoreLumaHistory(FfxUInt32x2 iPxPos, FfxFloat32x4 fLumaHistory) -{ - rw_luma_history[iPxPos] = fLumaHistory; -} - -FfxFloat32x4 LoadRwLumaHistory(FFX_MIN16_I2 iPxPos) -{ - return rw_luma_history[iPxPos]; -} -#endif - -#if defined(FSR2_BIND_SRV_LUMA_HISTORY) || defined(FFX_INTERNAL) -FfxFloat32x4 SampleLumaHistory(FfxFloat32x2 fUV) -{ - return r_luma_history.SampleLevel(s_LinearClamp, fUV, 0); -} -#endif - -#if defined(FFX_INTERNAL) -FfxFloat32x4 SampleDebug(FfxFloat32x2 fUV) -{ - return r_debug_out.SampleLevel(s_LinearClamp, fUV, 0).w; -} -#endif - -#if defined(FSR2_BIND_UAV_INTERNAL_UPSCALED) || defined(FFX_INTERNAL) -void StoreReprojectedHistory(FfxUInt32x2 iPxHistory, FfxFloat32x4 fHistory) -{ - rw_internal_upscaled_color[iPxHistory] = fHistory; -} -#endif - -#if defined(FSR2_BIND_UAV_INTERNAL_UPSCALED) || defined(FFX_INTERNAL) -void StoreInternalColorAndWeight(FfxUInt32x2 iPxPos, FfxFloat32x4 fColorAndWeight) -{ - rw_internal_upscaled_color[iPxPos] = fColorAndWeight; -} -#endif - -#if defined(FSR2_BIND_UAV_UPSCALED_OUTPUT) || defined(FFX_INTERNAL) -void StoreUpscaledOutput(FfxUInt32x2 iPxPos, FfxFloat32x3 fColor) -{ - rw_upscaled_output[iPxPos] = FfxFloat32x4(fColor, 1.f); -} -#endif - -//LOCK_LIFETIME_REMAINING == 0 -//Should make LockInitialLifetime() return a const 1.0f later -#if defined(FSR2_BIND_SRV_LOCK_STATUS) || defined(FFX_INTERNAL) -FfxFloat32x2 LoadLockStatus(FfxUInt32x2 iPxPos) -{ - return r_lock_status[iPxPos]; -} -#endif - -#if defined(FSR2_BIND_UAV_LOCK_STATUS) || defined(FFX_INTERNAL) -void StoreLockStatus(FfxUInt32x2 iPxPos, FfxFloat32x2 fLockStatus) -{ - rw_lock_status[iPxPos] = fLockStatus; -} -#endif - -#if defined(FSR2_BIND_SRV_LOCK_INPUT_LUMA) || defined(FFX_INTERNAL) -FfxFloat32 LoadLockInputLuma(FfxUInt32x2 iPxPos) -{ - return r_lock_input_luma[iPxPos]; -} -#endif - -#if defined(FSR2_BIND_UAV_LOCK_INPUT_LUMA) || defined(FFX_INTERNAL) -void StoreLockInputLuma(FfxUInt32x2 iPxPos, FfxFloat32 fLuma) -{ - rw_lock_input_luma[iPxPos] = fLuma; -} -#endif - -#if defined(FSR2_BIND_SRV_NEW_LOCKS) || defined(FFX_INTERNAL) -FfxFloat32 LoadNewLocks(FfxUInt32x2 iPxPos) -{ - return r_new_locks[iPxPos]; -} -#endif - -#if defined(FSR2_BIND_UAV_NEW_LOCKS) || defined(FFX_INTERNAL) -FfxFloat32 LoadRwNewLocks(FfxUInt32x2 iPxPos) -{ - return rw_new_locks[iPxPos]; -} -#endif - -#if defined(FSR2_BIND_UAV_NEW_LOCKS) || defined(FFX_INTERNAL) -void StoreNewLocks(FfxUInt32x2 iPxPos, FfxFloat32 newLock) -{ - rw_new_locks[iPxPos] = newLock; -} -#endif - -#if defined(FSR2_BIND_UAV_PREPARED_INPUT_COLOR) || defined(FFX_INTERNAL) -void StorePreparedInputColor(FFX_PARAMETER_IN FfxUInt32x2 iPxPos, FFX_PARAMETER_IN FfxFloat32x4 fTonemapped) -{ - rw_prepared_input_color[iPxPos] = fTonemapped; -} -#endif - -#if defined(FSR2_BIND_SRV_PREPARED_INPUT_COLOR) || defined(FFX_INTERNAL) -FfxFloat32 SampleDepthClip(FfxFloat32x2 fUV) -{ - return r_prepared_input_color.SampleLevel(s_LinearClamp, fUV, 0).w; -} -#endif - -#if defined(FSR2_BIND_SRV_LOCK_STATUS) || defined(FFX_INTERNAL) -FfxFloat32x2 SampleLockStatus(FfxFloat32x2 fUV) -{ - FfxFloat32x2 fLockStatus = r_lock_status.SampleLevel(s_LinearClamp, fUV, 0); - return fLockStatus; -} -#endif - -#if defined(FSR2_BIND_SRV_RECONSTRUCTED_PREV_NEAREST_DEPTH) || defined(FFX_INTERNAL) -FfxFloat32 LoadReconstructedPrevDepth(FfxUInt32x2 iPxPos) -{ - return asfloat(r_reconstructed_previous_nearest_depth[iPxPos]); -} -#endif - -#if defined(FSR2_BIND_UAV_RECONSTRUCTED_PREV_NEAREST_DEPTH) || defined(FFX_INTERNAL) -void StoreReconstructedDepth(FfxUInt32x2 iPxSample, FfxFloat32 fDepth) -{ - FfxUInt32 uDepth = asuint(fDepth); - - #if FFX_FSR2_OPTION_INVERTED_DEPTH - InterlockedMax(rw_reconstructed_previous_nearest_depth[iPxSample], uDepth); - #else - InterlockedMin(rw_reconstructed_previous_nearest_depth[iPxSample], uDepth); // min for standard, max for inverted depth - #endif -} -#endif - -#if defined(FSR2_BIND_UAV_RECONSTRUCTED_PREV_NEAREST_DEPTH) || defined(FFX_INTERNAL) -void SetReconstructedDepth(FfxUInt32x2 iPxSample, const FfxUInt32 uValue) -{ - rw_reconstructed_previous_nearest_depth[iPxSample] = uValue; -} -#endif - -#if defined(FSR2_BIND_UAV_DILATED_DEPTH) || defined(FFX_INTERNAL) -void StoreDilatedDepth(FFX_PARAMETER_IN FfxUInt32x2 iPxPos, FFX_PARAMETER_IN FfxFloat32 fDepth) -{ - rw_dilatedDepth[iPxPos] = fDepth; -} -#endif - -#if defined(FSR2_BIND_UAV_DILATED_MOTION_VECTORS) || defined(FFX_INTERNAL) -void StoreDilatedMotionVector(FFX_PARAMETER_IN FfxUInt32x2 iPxPos, FFX_PARAMETER_IN FfxFloat32x2 fMotionVector) -{ - rw_dilated_motion_vectors[iPxPos] = fMotionVector; -} -#endif - -#if defined(FSR2_BIND_SRV_DILATED_MOTION_VECTORS) || defined(FFX_INTERNAL) -FfxFloat32x2 LoadDilatedMotionVector(FfxUInt32x2 iPxInput) -{ - return r_dilated_motion_vectors[iPxInput].xy; -} -#endif - -#if defined(FSR2_BIND_SRV_PREVIOUS_DILATED_MOTION_VECTORS) || defined(FFX_INTERNAL) -FfxFloat32x2 LoadPreviousDilatedMotionVector(FfxUInt32x2 iPxInput) -{ - return r_previous_dilated_motion_vectors[iPxInput].xy; -} - -FfxFloat32x2 SamplePreviousDilatedMotionVector(FfxFloat32x2 uv) -{ - return r_previous_dilated_motion_vectors.SampleLevel(s_LinearClamp, uv, 0).xy; -} -#endif - -#if defined(FSR2_BIND_SRV_DILATED_DEPTH) || defined(FFX_INTERNAL) -FfxFloat32 LoadDilatedDepth(FfxUInt32x2 iPxInput) -{ - return r_dilatedDepth[iPxInput]; -} -#endif - -#if defined(FSR2_BIND_SRV_INPUT_EXPOSURE) || defined(FFX_INTERNAL) -FfxFloat32 Exposure() -{ - FfxFloat32 exposure = r_input_exposure[FfxUInt32x2(0, 0)].x; - - if (exposure == 0.0f) { - exposure = 1.0f; - } - - return exposure; -} -#endif - -#if defined(FSR2_BIND_SRV_AUTO_EXPOSURE) || defined(FFX_INTERNAL) -FfxFloat32 AutoExposure() -{ - FfxFloat32 exposure = r_auto_exposure[FfxUInt32x2(0, 0)].x; - - if (exposure == 0.0f) { - exposure = 1.0f; - } - - return exposure; -} -#endif - -FfxFloat32 SampleLanczos2Weight(FfxFloat32 x) -{ -#if defined(FSR2_BIND_SRV_LANCZOS_LUT) || defined(FFX_INTERNAL) - return r_lanczos_lut.SampleLevel(s_LinearClamp, FfxFloat32x2(x / 2, 0.5f), 0); -#else - return 0.f; -#endif -} - -#if defined(FSR2_BIND_SRV_UPSCALE_MAXIMUM_BIAS_LUT) || defined(FFX_INTERNAL) -FfxFloat32 SampleUpsampleMaximumBias(FfxFloat32x2 uv) -{ - // Stored as a SNORM, so make sure to multiply by 2 to retrieve the actual expected range. - return FfxFloat32(2.0) * r_upsample_maximum_bias_lut.SampleLevel(s_LinearClamp, abs(uv) * 2.0, 0); -} -#endif - -#if defined(FSR2_BIND_SRV_DILATED_REACTIVE_MASKS) || defined(FFX_INTERNAL) -FfxFloat32x2 SampleDilatedReactiveMasks(FfxFloat32x2 fUV) -{ - return r_dilated_reactive_masks.SampleLevel(s_LinearClamp, fUV, 0); -} -#endif - -#if defined(FSR2_BIND_SRV_DILATED_REACTIVE_MASKS) || defined(FFX_INTERNAL) -FfxFloat32x2 LoadDilatedReactiveMasks(FFX_PARAMETER_IN FfxUInt32x2 iPxPos) -{ - return r_dilated_reactive_masks[iPxPos]; -} -#endif - -#if defined(FSR2_BIND_UAV_DILATED_REACTIVE_MASKS) || defined(FFX_INTERNAL) -void StoreDilatedReactiveMasks(FFX_PARAMETER_IN FfxUInt32x2 iPxPos, FFX_PARAMETER_IN FfxFloat32x2 fDilatedReactiveMasks) -{ - rw_dilated_reactive_masks[iPxPos] = fDilatedReactiveMasks; -} -#endif - -#if defined(FSR2_BIND_SRV_INPUT_OPAQUE_ONLY) || defined(FFX_INTERNAL) -FfxFloat32x3 LoadOpaqueOnly(FFX_PARAMETER_IN FFX_MIN16_I2 iPxPos) -{ - return r_input_opaque_only[UNITY_FSR2_POS(iPxPos)].xyz; -} -#endif - -#if defined(FSR2_BIND_SRV_PREV_PRE_ALPHA_COLOR) || defined(FFX_INTERNAL) -FfxFloat32x3 LoadPrevPreAlpha(FFX_PARAMETER_IN FFX_MIN16_I2 iPxPos) -{ - return r_input_prev_color_pre_alpha[iPxPos]; -} -#endif - -#if defined(FSR2_BIND_SRV_PREV_POST_ALPHA_COLOR) || defined(FFX_INTERNAL) -FfxFloat32x3 LoadPrevPostAlpha(FFX_PARAMETER_IN FFX_MIN16_I2 iPxPos) -{ - return r_input_prev_color_post_alpha[iPxPos]; -} -#endif - -#if defined(FSR2_BIND_UAV_AUTOREACTIVE) || defined(FFX_INTERNAL) -#if defined(FSR2_BIND_UAV_AUTOCOMPOSITION) || defined(FFX_INTERNAL) -void StoreAutoReactive(FFX_PARAMETER_IN FFX_MIN16_I2 iPxPos, FFX_PARAMETER_IN FFX_MIN16_F2 fReactive) -{ - rw_output_autoreactive[iPxPos] = fReactive.x; - - rw_output_autocomposition[iPxPos] = fReactive.y; -} -#endif -#endif - -#if defined(FSR2_BIND_UAV_PREV_PRE_ALPHA_COLOR) || defined(FFX_INTERNAL) -void StorePrevPreAlpha(FFX_PARAMETER_IN FFX_MIN16_I2 iPxPos, FFX_PARAMETER_IN FFX_MIN16_F3 color) -{ - rw_output_prev_color_pre_alpha[iPxPos] = color; - -} -#endif - -#if defined(FSR2_BIND_UAV_PREV_POST_ALPHA_COLOR) || defined(FFX_INTERNAL) -void StorePrevPostAlpha(FFX_PARAMETER_IN FFX_MIN16_I2 iPxPos, FFX_PARAMETER_IN FFX_MIN16_F3 color) -{ - rw_output_prev_color_post_alpha[iPxPos] = color; -} -#endif - -#endif // #if defined(FFX_GPU) diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_callbacks_hlsl.h.meta b/Assets/Resources/FSR2/shaders/ffx_fsr2_callbacks_hlsl.h.meta deleted file mode 100644 index 97858ae..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_callbacks_hlsl.h.meta +++ /dev/null @@ -1,60 +0,0 @@ -fileFormatVersion: 2 -guid: eb121968296f9ba44b35d7e18d2b79df -PluginImporter: - externalObjects: {} - serializedVersion: 2 - iconMap: {} - executionOrder: {} - defineConstraints: [] - isPreloaded: 0 - isOverridable: 0 - isExplicitlyReferenced: 0 - validateReferences: 1 - platformData: - - first: - : Any - second: - enabled: 0 - settings: - Exclude Editor: 1 - Exclude GameCoreScarlett: 1 - Exclude GameCoreXboxOne: 1 - Exclude Linux64: 1 - Exclude OSXUniversal: 1 - Exclude PS4: 1 - Exclude PS5: 1 - Exclude WebGL: 1 - Exclude Win: 1 - Exclude Win64: 1 - - first: - Any: - second: - enabled: 0 - settings: {} - - first: - Editor: Editor - second: - enabled: 0 - settings: - DefaultValueInitialized: true - - first: - Standalone: Linux64 - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win64 - second: - enabled: 0 - settings: - CPU: None - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_common.h b/Assets/Resources/FSR2/shaders/ffx_fsr2_common.h deleted file mode 100644 index 0c72aa8..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_common.h +++ /dev/null @@ -1,565 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#if !defined(FFX_FSR2_COMMON_H) -#define FFX_FSR2_COMMON_H - -#if defined(FFX_CPU) || defined(FFX_GPU) -//Locks -#define LOCK_LIFETIME_REMAINING 0 -#define LOCK_TEMPORAL_LUMA 1 -#endif // #if defined(FFX_CPU) || defined(FFX_GPU) - -#if defined(FFX_GPU) -FFX_STATIC const FfxFloat32 FSR2_FP16_MIN = 6.10e-05f; -FFX_STATIC const FfxFloat32 FSR2_FP16_MAX = 65504.0f; -FFX_STATIC const FfxFloat32 FSR2_EPSILON = 1e-03f; -FFX_STATIC const FfxFloat32 FSR2_TONEMAP_EPSILON = 1.0f / FSR2_FP16_MAX; -FFX_STATIC const FfxFloat32 FSR2_FLT_MAX = 3.402823466e+38f; -FFX_STATIC const FfxFloat32 FSR2_FLT_MIN = 1.175494351e-38f; - -// treat vector truncation warnings as errors -#pragma warning(error: 3206) - -// suppress warnings -#pragma warning(disable: 3205) // conversion from larger type to smaller -#pragma warning(disable: 3571) // in ffxPow(f, e), f could be negative - -// Reconstructed depth usage -FFX_STATIC const FfxFloat32 fReconstructedDepthBilinearWeightThreshold = 0.01f; - -// Accumulation -FFX_STATIC const FfxFloat32 fUpsampleLanczosWeightScale = 1.0f / 12.0f; -FFX_STATIC const FfxFloat32 fMaxAccumulationLanczosWeight = 1.0f; -FFX_STATIC const FfxFloat32 fAverageLanczosWeightPerFrame = 0.74f * fUpsampleLanczosWeightScale; // Average lanczos weight for jitter accumulated samples -FFX_STATIC const FfxFloat32 fAccumulationMaxOnMotion = 3.0f * fUpsampleLanczosWeightScale; - -// Auto exposure -FFX_STATIC const FfxFloat32 resetAutoExposureAverageSmoothing = 1e8f; - -struct AccumulationPassCommonParams -{ - FfxInt32x2 iPxHrPos; - FfxFloat32x2 fHrUv; - FfxFloat32x2 fLrUv_HwSampler; - FfxFloat32x2 fMotionVector; - FfxFloat32x2 fReprojectedHrUv; - FfxFloat32 fHrVelocity; - FfxFloat32 fDepthClipFactor; - FfxFloat32 fDilatedReactiveFactor; - FfxFloat32 fAccumulationMask; - - FfxBoolean bIsResetFrame; - FfxBoolean bIsExistingSample; - FfxBoolean bIsNewSample; -}; - -struct LockState -{ - FfxBoolean NewLock; //Set for both unique new and re-locked new - FfxBoolean WasLockedPrevFrame; //Set to identify if the pixel was already locked (relock) -}; - -void InitializeNewLockSample(FFX_PARAMETER_OUT FfxFloat32x2 fLockStatus) -{ - fLockStatus = FfxFloat32x2(0, 0); -} - -#if FFX_HALF -void InitializeNewLockSample(FFX_PARAMETER_OUT FFX_MIN16_F2 fLockStatus) -{ - fLockStatus = FFX_MIN16_F2(0, 0); -} -#endif - - -void KillLock(FFX_PARAMETER_INOUT FfxFloat32x2 fLockStatus) -{ - fLockStatus[LOCK_LIFETIME_REMAINING] = 0; -} - -#if FFX_HALF -void KillLock(FFX_PARAMETER_INOUT FFX_MIN16_F2 fLockStatus) -{ - fLockStatus[LOCK_LIFETIME_REMAINING] = FFX_MIN16_F(0); -} -#endif - -struct RectificationBox -{ - FfxFloat32x3 boxCenter; - FfxFloat32x3 boxVec; - FfxFloat32x3 aabbMin; - FfxFloat32x3 aabbMax; - FfxFloat32 fBoxCenterWeight; -}; -#if FFX_HALF -struct RectificationBoxMin16 -{ - FFX_MIN16_F3 boxCenter; - FFX_MIN16_F3 boxVec; - FFX_MIN16_F3 aabbMin; - FFX_MIN16_F3 aabbMax; - FFX_MIN16_F fBoxCenterWeight; -}; -#endif - -void RectificationBoxReset(FFX_PARAMETER_INOUT RectificationBox rectificationBox) -{ - rectificationBox.fBoxCenterWeight = FfxFloat32(0); - - rectificationBox.boxCenter = FfxFloat32x3(0, 0, 0); - rectificationBox.boxVec = FfxFloat32x3(0, 0, 0); - rectificationBox.aabbMin = FfxFloat32x3(FSR2_FLT_MAX, FSR2_FLT_MAX, FSR2_FLT_MAX); - rectificationBox.aabbMax = -FfxFloat32x3(FSR2_FLT_MAX, FSR2_FLT_MAX, FSR2_FLT_MAX); -} -#if FFX_HALF -void RectificationBoxReset(FFX_PARAMETER_INOUT RectificationBoxMin16 rectificationBox) -{ - rectificationBox.fBoxCenterWeight = FFX_MIN16_F(0); - - rectificationBox.boxCenter = FFX_MIN16_F3(0, 0, 0); - rectificationBox.boxVec = FFX_MIN16_F3(0, 0, 0); - rectificationBox.aabbMin = FFX_MIN16_F3(FSR2_FP16_MAX, FSR2_FP16_MAX, FSR2_FP16_MAX); - rectificationBox.aabbMax = -FFX_MIN16_F3(FSR2_FP16_MAX, FSR2_FP16_MAX, FSR2_FP16_MAX); -} -#endif - -void RectificationBoxAddInitialSample(FFX_PARAMETER_INOUT RectificationBox rectificationBox, const FfxFloat32x3 colorSample, const FfxFloat32 fSampleWeight) -{ - rectificationBox.aabbMin = colorSample; - rectificationBox.aabbMax = colorSample; - - FfxFloat32x3 weightedSample = colorSample * fSampleWeight; - rectificationBox.boxCenter = weightedSample; - rectificationBox.boxVec = colorSample * weightedSample; - rectificationBox.fBoxCenterWeight = fSampleWeight; -} - -void RectificationBoxAddSample(FfxBoolean bInitialSample, FFX_PARAMETER_INOUT RectificationBox rectificationBox, const FfxFloat32x3 colorSample, const FfxFloat32 fSampleWeight) -{ - if (bInitialSample) { - RectificationBoxAddInitialSample(rectificationBox, colorSample, fSampleWeight); - } else { - rectificationBox.aabbMin = ffxMin(rectificationBox.aabbMin, colorSample); - rectificationBox.aabbMax = ffxMax(rectificationBox.aabbMax, colorSample); - - FfxFloat32x3 weightedSample = colorSample * fSampleWeight; - rectificationBox.boxCenter += weightedSample; - rectificationBox.boxVec += colorSample * weightedSample; - rectificationBox.fBoxCenterWeight += fSampleWeight; - } -} -#if FFX_HALF -void RectificationBoxAddInitialSample(FFX_PARAMETER_INOUT RectificationBoxMin16 rectificationBox, const FFX_MIN16_F3 colorSample, const FFX_MIN16_F fSampleWeight) -{ - rectificationBox.aabbMin = colorSample; - rectificationBox.aabbMax = colorSample; - - FFX_MIN16_F3 weightedSample = colorSample * fSampleWeight; - rectificationBox.boxCenter = weightedSample; - rectificationBox.boxVec = colorSample * weightedSample; - rectificationBox.fBoxCenterWeight = fSampleWeight; -} - -void RectificationBoxAddSample(FfxBoolean bInitialSample, FFX_PARAMETER_INOUT RectificationBoxMin16 rectificationBox, const FFX_MIN16_F3 colorSample, const FFX_MIN16_F fSampleWeight) -{ - if (bInitialSample) { - RectificationBoxAddInitialSample(rectificationBox, colorSample, fSampleWeight); - } else { - rectificationBox.aabbMin = ffxMin(rectificationBox.aabbMin, colorSample); - rectificationBox.aabbMax = ffxMax(rectificationBox.aabbMax, colorSample); - - FFX_MIN16_F3 weightedSample = colorSample * fSampleWeight; - rectificationBox.boxCenter += weightedSample; - rectificationBox.boxVec += colorSample * weightedSample; - rectificationBox.fBoxCenterWeight += fSampleWeight; - } -} -#endif - -void RectificationBoxComputeVarianceBoxData(FFX_PARAMETER_INOUT RectificationBox rectificationBox) -{ - rectificationBox.fBoxCenterWeight = (abs(rectificationBox.fBoxCenterWeight) > FfxFloat32(FSR2_EPSILON) ? rectificationBox.fBoxCenterWeight : FfxFloat32(1.f)); - rectificationBox.boxCenter /= rectificationBox.fBoxCenterWeight; - rectificationBox.boxVec /= rectificationBox.fBoxCenterWeight; - FfxFloat32x3 stdDev = sqrt(abs(rectificationBox.boxVec - rectificationBox.boxCenter * rectificationBox.boxCenter)); - rectificationBox.boxVec = stdDev; -} -#if FFX_HALF -void RectificationBoxComputeVarianceBoxData(FFX_PARAMETER_INOUT RectificationBoxMin16 rectificationBox) -{ - rectificationBox.fBoxCenterWeight = (abs(rectificationBox.fBoxCenterWeight) > FFX_MIN16_F(FSR2_EPSILON) ? rectificationBox.fBoxCenterWeight : FFX_MIN16_F(1.f)); - rectificationBox.boxCenter /= rectificationBox.fBoxCenterWeight; - rectificationBox.boxVec /= rectificationBox.fBoxCenterWeight; - FFX_MIN16_F3 stdDev = sqrt(abs(rectificationBox.boxVec - rectificationBox.boxCenter * rectificationBox.boxCenter)); - rectificationBox.boxVec = stdDev; -} -#endif - -FfxFloat32x3 SafeRcp3(FfxFloat32x3 v) -{ - return (all(FFX_NOT_EQUAL(v, FfxFloat32x3(0, 0, 0)))) ? (FfxFloat32x3(1, 1, 1) / v) : FfxFloat32x3(0, 0, 0); -} -#if FFX_HALF -FFX_MIN16_F3 SafeRcp3(FFX_MIN16_F3 v) -{ - return (all(FFX_NOT_EQUAL(v, FFX_MIN16_F3(0, 0, 0)))) ? (FFX_MIN16_F3(1, 1, 1) / v) : FFX_MIN16_F3(0, 0, 0); -} -#endif - -FfxFloat32 MinDividedByMax(const FfxFloat32 v0, const FfxFloat32 v1) -{ - const FfxFloat32 m = ffxMax(v0, v1); - return m != 0 ? ffxMin(v0, v1) / m : 0; -} - -#if FFX_HALF -FFX_MIN16_F MinDividedByMax(const FFX_MIN16_F v0, const FFX_MIN16_F v1) -{ - const FFX_MIN16_F m = ffxMax(v0, v1); - return m != FFX_MIN16_F(0) ? ffxMin(v0, v1) / m : FFX_MIN16_F(0); -} -#endif - -FfxFloat32x3 YCoCgToRGB(FfxFloat32x3 fYCoCg) -{ - FfxFloat32x3 fRgb; - - fRgb = FfxFloat32x3( - fYCoCg.x + fYCoCg.y - fYCoCg.z, - fYCoCg.x + fYCoCg.z, - fYCoCg.x - fYCoCg.y - fYCoCg.z); - - return fRgb; -} -#if FFX_HALF -FFX_MIN16_F3 YCoCgToRGB(FFX_MIN16_F3 fYCoCg) -{ - FFX_MIN16_F3 fRgb; - - fRgb = FFX_MIN16_F3( - fYCoCg.x + fYCoCg.y - fYCoCg.z, - fYCoCg.x + fYCoCg.z, - fYCoCg.x - fYCoCg.y - fYCoCg.z); - - return fRgb; -} -#endif - -FfxFloat32x3 RGBToYCoCg(FfxFloat32x3 fRgb) -{ - FfxFloat32x3 fYCoCg; - - fYCoCg = FfxFloat32x3( - 0.25f * fRgb.r + 0.5f * fRgb.g + 0.25f * fRgb.b, - 0.5f * fRgb.r - 0.5f * fRgb.b, - -0.25f * fRgb.r + 0.5f * fRgb.g - 0.25f * fRgb.b); - - return fYCoCg; -} -#if FFX_HALF -FFX_MIN16_F3 RGBToYCoCg(FFX_MIN16_F3 fRgb) -{ - FFX_MIN16_F3 fYCoCg; - - fYCoCg = FFX_MIN16_F3( - 0.25 * fRgb.r + 0.5 * fRgb.g + 0.25 * fRgb.b, - 0.5 * fRgb.r - 0.5 * fRgb.b, - -0.25 * fRgb.r + 0.5 * fRgb.g - 0.25 * fRgb.b); - - return fYCoCg; -} -#endif - -FfxFloat32 RGBToLuma(FfxFloat32x3 fLinearRgb) -{ - return dot(fLinearRgb, FfxFloat32x3(0.2126f, 0.7152f, 0.0722f)); -} -#if FFX_HALF -FFX_MIN16_F RGBToLuma(FFX_MIN16_F3 fLinearRgb) -{ - return dot(fLinearRgb, FFX_MIN16_F3(0.2126f, 0.7152f, 0.0722f)); -} -#endif - -FfxFloat32 RGBToPerceivedLuma(FfxFloat32x3 fLinearRgb) -{ - FfxFloat32 fLuminance = RGBToLuma(fLinearRgb); - - FfxFloat32 fPercievedLuminance = 0; - if (fLuminance <= 216.0f / 24389.0f) { - fPercievedLuminance = fLuminance * (24389.0f / 27.0f); - } - else { - fPercievedLuminance = ffxPow(fLuminance, 1.0f / 3.0f) * 116.0f - 16.0f; - } - - return fPercievedLuminance * 0.01f; -} -#if FFX_HALF -FFX_MIN16_F RGBToPerceivedLuma(FFX_MIN16_F3 fLinearRgb) -{ - FFX_MIN16_F fLuminance = RGBToLuma(fLinearRgb); - - FFX_MIN16_F fPercievedLuminance = FFX_MIN16_F(0); - if (fLuminance <= FFX_MIN16_F(216.0f / 24389.0f)) { - fPercievedLuminance = fLuminance * FFX_MIN16_F(24389.0f / 27.0f); - } - else { - fPercievedLuminance = ffxPow(fLuminance, FFX_MIN16_F(1.0f / 3.0f)) * FFX_MIN16_F(116.0f) - FFX_MIN16_F(16.0f); - } - - return fPercievedLuminance * FFX_MIN16_F(0.01f); -} -#endif - -FfxFloat32x3 Tonemap(FfxFloat32x3 fRgb) -{ - return fRgb / (ffxMax(ffxMax(0.f, fRgb.r), ffxMax(fRgb.g, fRgb.b)) + 1.f).xxx; -} - -FfxFloat32x3 InverseTonemap(FfxFloat32x3 fRgb) -{ - return fRgb / ffxMax(FSR2_TONEMAP_EPSILON, 1.f - ffxMax(fRgb.r, ffxMax(fRgb.g, fRgb.b))).xxx; -} - -#if FFX_HALF -FFX_MIN16_F3 Tonemap(FFX_MIN16_F3 fRgb) -{ - return fRgb / (ffxMax(ffxMax(FFX_MIN16_F(0.f), fRgb.r), ffxMax(fRgb.g, fRgb.b)) + FFX_MIN16_F(1.f)).xxx; -} - -FFX_MIN16_F3 InverseTonemap(FFX_MIN16_F3 fRgb) -{ - return fRgb / ffxMax(FFX_MIN16_F(FSR2_TONEMAP_EPSILON), FFX_MIN16_F(1.f) - ffxMax(fRgb.r, ffxMax(fRgb.g, fRgb.b))).xxx; -} -#endif - -FfxInt32x2 ClampLoad(FfxInt32x2 iPxSample, FfxInt32x2 iPxOffset, FfxInt32x2 iTextureSize) -{ - FfxInt32x2 result = iPxSample + iPxOffset; - result.x = (iPxOffset.x < 0) ? ffxMax(result.x, 0) : result.x; - result.x = (iPxOffset.x > 0) ? ffxMin(result.x, iTextureSize.x - 1) : result.x; - result.y = (iPxOffset.y < 0) ? ffxMax(result.y, 0) : result.y; - result.y = (iPxOffset.y > 0) ? ffxMin(result.y, iTextureSize.y - 1) : result.y; - return result; - - // return ffxMed3(iPxSample + iPxOffset, FfxInt32x2(0, 0), iTextureSize - FfxInt32x2(1, 1)); -} -#if FFX_HALF -FFX_MIN16_I2 ClampLoad(FFX_MIN16_I2 iPxSample, FFX_MIN16_I2 iPxOffset, FFX_MIN16_I2 iTextureSize) -{ - FFX_MIN16_I2 result = iPxSample + iPxOffset; - result.x = (iPxOffset.x < 0) ? ffxMax(result.x, FFX_MIN16_I(0)) : result.x; - result.x = (iPxOffset.x > 0) ? ffxMin(result.x, iTextureSize.x - FFX_MIN16_I(1)) : result.x; - result.y = (iPxOffset.y < 0) ? ffxMax(result.y, FFX_MIN16_I(0)) : result.y; - result.y = (iPxOffset.y > 0) ? ffxMin(result.y, iTextureSize.y - FFX_MIN16_I(1)) : result.y; - return result; - - // return ffxMed3Half(iPxSample + iPxOffset, FFX_MIN16_I2(0, 0), iTextureSize - FFX_MIN16_I2(1, 1)); -} -#endif - -FfxFloat32x2 ClampUv(FfxFloat32x2 fUv, FfxInt32x2 iTextureSize, FfxInt32x2 iResourceSize) -{ - const FfxFloat32x2 fSampleLocation = fUv * iTextureSize; - const FfxFloat32x2 fClampedLocation = ffxMax(FfxFloat32x2(0.5f, 0.5f), ffxMin(fSampleLocation, FfxFloat32x2(iTextureSize) - FfxFloat32x2(0.5f, 0.5f))); - const FfxFloat32x2 fClampedUv = fClampedLocation / FfxFloat32x2(iResourceSize); - - return fClampedUv; -} - -FfxBoolean IsOnScreen(FfxInt32x2 pos, FfxInt32x2 size) -{ - return all(FFX_LESS_THAN(FfxUInt32x2(pos), FfxUInt32x2(size))); -} -#if FFX_HALF -FfxBoolean IsOnScreen(FFX_MIN16_I2 pos, FFX_MIN16_I2 size) -{ - return all(FFX_LESS_THAN(FFX_MIN16_U2(pos), FFX_MIN16_U2(size))); -} -#endif - -FfxFloat32 ComputeAutoExposureFromLavg(FfxFloat32 Lavg) -{ - Lavg = exp(Lavg); - - const FfxFloat32 S = 100.0f; //ISO arithmetic speed - const FfxFloat32 K = 12.5f; - FfxFloat32 ExposureISO100 = log2((Lavg * S) / K); - - const FfxFloat32 q = 0.65f; - FfxFloat32 Lmax = (78.0f / (q * S)) * ffxPow(2.0f, ExposureISO100); - - return 1 / Lmax; -} -#if FFX_HALF -FFX_MIN16_F ComputeAutoExposureFromLavg(FFX_MIN16_F Lavg) -{ - Lavg = exp(Lavg); - - const FFX_MIN16_F S = FFX_MIN16_F(100.0f); //ISO arithmetic speed - const FFX_MIN16_F K = FFX_MIN16_F(12.5f); - const FFX_MIN16_F ExposureISO100 = log2((Lavg * S) / K); - - const FFX_MIN16_F q = FFX_MIN16_F(0.65f); - const FFX_MIN16_F Lmax = (FFX_MIN16_F(78.0f) / (q * S)) * ffxPow(FFX_MIN16_F(2.0f), ExposureISO100); - - return FFX_MIN16_F(1) / Lmax; -} -#endif - -FfxInt32x2 ComputeHrPosFromLrPos(FfxInt32x2 iPxLrPos) -{ - FfxFloat32x2 fSrcJitteredPos = FfxFloat32x2(iPxLrPos) + 0.5f - Jitter(); - FfxFloat32x2 fLrPosInHr = (fSrcJitteredPos / RenderSize()) * DisplaySize(); - FfxInt32x2 iPxHrPos = FfxInt32x2(floor(fLrPosInHr)); - return iPxHrPos; -} -#if FFX_HALF -FFX_MIN16_I2 ComputeHrPosFromLrPos(FFX_MIN16_I2 iPxLrPos) -{ - FFX_MIN16_F2 fSrcJitteredPos = FFX_MIN16_F2(iPxLrPos) + FFX_MIN16_F(0.5f) - FFX_MIN16_F2(Jitter()); - FFX_MIN16_F2 fLrPosInHr = (fSrcJitteredPos / FFX_MIN16_F2(RenderSize())) * FFX_MIN16_F2(DisplaySize()); - FFX_MIN16_I2 iPxHrPos = FFX_MIN16_I2(floor(fLrPosInHr)); - return iPxHrPos; -} -#endif - -FfxFloat32x2 ComputeNdc(FfxFloat32x2 fPxPos, FfxInt32x2 iSize) -{ - return fPxPos / FfxFloat32x2(iSize) * FfxFloat32x2(2.0f, -2.0f) + FfxFloat32x2(-1.0f, 1.0f); -} - -FfxFloat32 GetViewSpaceDepth(FfxFloat32 fDeviceDepth) -{ - const FfxFloat32x4 fDeviceToViewDepth = DeviceToViewSpaceTransformFactors(); - - // fDeviceToViewDepth details found in ffx_fsr2.cpp - return (fDeviceToViewDepth[1] / (fDeviceDepth - fDeviceToViewDepth[0])); -} - -FfxFloat32 GetViewSpaceDepthInMeters(FfxFloat32 fDeviceDepth) -{ - return GetViewSpaceDepth(fDeviceDepth) * ViewSpaceToMetersFactor(); -} - -FfxFloat32x3 GetViewSpacePosition(FfxInt32x2 iViewportPos, FfxInt32x2 iViewportSize, FfxFloat32 fDeviceDepth) -{ - const FfxFloat32x4 fDeviceToViewDepth = DeviceToViewSpaceTransformFactors(); - - const FfxFloat32 Z = GetViewSpaceDepth(fDeviceDepth); - - const FfxFloat32x2 fNdcPos = ComputeNdc(iViewportPos, iViewportSize); - const FfxFloat32 X = fDeviceToViewDepth[2] * fNdcPos.x * Z; - const FfxFloat32 Y = fDeviceToViewDepth[3] * fNdcPos.y * Z; - - return FfxFloat32x3(X, Y, Z); -} - -FfxFloat32x3 GetViewSpacePositionInMeters(FfxInt32x2 iViewportPos, FfxInt32x2 iViewportSize, FfxFloat32 fDeviceDepth) -{ - return GetViewSpacePosition(iViewportPos, iViewportSize, fDeviceDepth) * ViewSpaceToMetersFactor(); -} - -FfxFloat32 GetMaxDistanceInMeters() -{ -#if FFX_FSR2_OPTION_INVERTED_DEPTH - return GetViewSpaceDepth(0.0f) * ViewSpaceToMetersFactor(); -#else - return GetViewSpaceDepth(1.0f) * ViewSpaceToMetersFactor(); -#endif -} - -FfxFloat32x3 PrepareRgb(FfxFloat32x3 fRgb, FfxFloat32 fExposure, FfxFloat32 fPreExposure) -{ - fRgb /= fPreExposure; - fRgb *= fExposure; - - fRgb = clamp(fRgb, 0.0f, FSR2_FP16_MAX); - - return fRgb; -} - -FfxFloat32x3 UnprepareRgb(FfxFloat32x3 fRgb, FfxFloat32 fExposure) -{ - fRgb /= fExposure; - fRgb *= PreExposure(); - - return fRgb; -} - - -struct BilinearSamplingData -{ - FfxInt32x2 iOffsets[4]; - FfxFloat32 fWeights[4]; - FfxInt32x2 iBasePos; -}; - -BilinearSamplingData GetBilinearSamplingData(FfxFloat32x2 fUv, FfxInt32x2 iSize) -{ - BilinearSamplingData data; - - FfxFloat32x2 fPxSample = (fUv * iSize) - FfxFloat32x2(0.5f, 0.5f); - data.iBasePos = FfxInt32x2(floor(fPxSample)); - FfxFloat32x2 fPxFrac = ffxFract(fPxSample); - - data.iOffsets[0] = FfxInt32x2(0, 0); - data.iOffsets[1] = FfxInt32x2(1, 0); - data.iOffsets[2] = FfxInt32x2(0, 1); - data.iOffsets[3] = FfxInt32x2(1, 1); - - data.fWeights[0] = (1 - fPxFrac.x) * (1 - fPxFrac.y); - data.fWeights[1] = (fPxFrac.x) * (1 - fPxFrac.y); - data.fWeights[2] = (1 - fPxFrac.x) * (fPxFrac.y); - data.fWeights[3] = (fPxFrac.x) * (fPxFrac.y); - - return data; -} - -struct PlaneData -{ - FfxFloat32x3 fNormal; - FfxFloat32 fDistanceFromOrigin; -}; - -PlaneData GetPlaneFromPoints(FfxFloat32x3 fP0, FfxFloat32x3 fP1, FfxFloat32x3 fP2) -{ - PlaneData plane; - - FfxFloat32x3 v0 = fP0 - fP1; - FfxFloat32x3 v1 = fP0 - fP2; - plane.fNormal = normalize(cross(v0, v1)); - plane.fDistanceFromOrigin = -dot(fP0, plane.fNormal); - - return plane; -} - -FfxFloat32 PointToPlaneDistance(PlaneData plane, FfxFloat32x3 fPoint) -{ - return abs(dot(plane.fNormal, fPoint) + plane.fDistanceFromOrigin); -} - -#endif // #if defined(FFX_GPU) - -#endif //!defined(FFX_FSR2_COMMON_H) diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_common.h.meta b/Assets/Resources/FSR2/shaders/ffx_fsr2_common.h.meta deleted file mode 100644 index 0e3bb9c..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_common.h.meta +++ /dev/null @@ -1,60 +0,0 @@ -fileFormatVersion: 2 -guid: 2176dca22b6e9604da8329c79abae68d -PluginImporter: - externalObjects: {} - serializedVersion: 2 - iconMap: {} - executionOrder: {} - defineConstraints: [] - isPreloaded: 0 - isOverridable: 0 - isExplicitlyReferenced: 0 - validateReferences: 1 - platformData: - - first: - : Any - second: - enabled: 0 - settings: - Exclude Editor: 1 - Exclude GameCoreScarlett: 1 - Exclude GameCoreXboxOne: 1 - Exclude Linux64: 1 - Exclude OSXUniversal: 1 - Exclude PS4: 1 - Exclude PS5: 1 - Exclude WebGL: 1 - Exclude Win: 1 - Exclude Win64: 1 - - first: - Any: - second: - enabled: 0 - settings: {} - - first: - Editor: Editor - second: - enabled: 0 - settings: - DefaultValueInitialized: true - - first: - Standalone: Linux64 - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win64 - second: - enabled: 0 - settings: - CPU: None - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_compute_luminance_pyramid.h b/Assets/Resources/FSR2/shaders/ffx_fsr2_compute_luminance_pyramid.h deleted file mode 100644 index c63f182..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_compute_luminance_pyramid.h +++ /dev/null @@ -1,189 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -FFX_GROUPSHARED FfxUInt32 spdCounter; - -#ifndef SPD_PACKED_ONLY -FFX_GROUPSHARED FfxFloat32 spdIntermediateR[16][16]; -FFX_GROUPSHARED FfxFloat32 spdIntermediateG[16][16]; -FFX_GROUPSHARED FfxFloat32 spdIntermediateB[16][16]; -FFX_GROUPSHARED FfxFloat32 spdIntermediateA[16][16]; - -FfxFloat32x4 SpdLoadSourceImage(FfxFloat32x2 tex, FfxUInt32 slice) -{ - FfxFloat32x2 fUv = (tex + 0.5f + Jitter()) / RenderSize(); - fUv = ClampUv(fUv, RenderSize(), InputColorResourceDimensions()); - FfxFloat32x3 fRgb = SampleInputColor(fUv); - - fRgb /= PreExposure(); - - //compute log luma - const FfxFloat32 fLogLuma = log(ffxMax(FSR2_EPSILON, RGBToLuma(fRgb))); - - // Make sure out of screen pixels contribute no value to the end result - const FfxFloat32 result = all(FFX_LESS_THAN(tex, RenderSize())) ? fLogLuma : 0.0f; - - return FfxFloat32x4(result, 0, 0, 0); -} - -FfxFloat32x4 SpdLoad(FfxInt32x2 tex, FfxUInt32 slice) -{ - return SPD_LoadMipmap5(tex); -} - -void SpdStore(FfxInt32x2 pix, FfxFloat32x4 outValue, FfxUInt32 index, FfxUInt32 slice) -{ - if (index == LumaMipLevelToUse() || index == 5) - { - SPD_SetMipmap(pix, index, outValue.r); - } - - if (index == MipCount() - 1) { //accumulate on 1x1 level - - if (all(FFX_EQUAL(pix, FfxInt32x2(0, 0)))) - { - FfxFloat32 prev = SPD_LoadExposureBuffer().y; - FfxFloat32 result = outValue.r; - - if (prev < resetAutoExposureAverageSmoothing) // Compare Lavg, so small or negative values - { - FfxFloat32 rate = 1.0f; - result = prev + (result - prev) * (1 - exp(-DeltaTime() * rate)); - } - FfxFloat32x2 spdOutput = FfxFloat32x2(ComputeAutoExposureFromLavg(result), result); - SPD_SetExposureBuffer(spdOutput); - } - } -} - -void SpdIncreaseAtomicCounter(FfxUInt32 slice) -{ - SPD_IncreaseAtomicCounter(spdCounter); -} - -FfxUInt32 SpdGetAtomicCounter() -{ - return spdCounter; -} - -void SpdResetAtomicCounter(FfxUInt32 slice) -{ - SPD_ResetAtomicCounter(); -} - -FfxFloat32x4 SpdLoadIntermediate(FfxUInt32 x, FfxUInt32 y) -{ - return FfxFloat32x4( - spdIntermediateR[x][y], - spdIntermediateG[x][y], - spdIntermediateB[x][y], - spdIntermediateA[x][y]); -} -void SpdStoreIntermediate(FfxUInt32 x, FfxUInt32 y, FfxFloat32x4 value) -{ - spdIntermediateR[x][y] = value.x; - spdIntermediateG[x][y] = value.y; - spdIntermediateB[x][y] = value.z; - spdIntermediateA[x][y] = value.w; -} -FfxFloat32x4 SpdReduce4(FfxFloat32x4 v0, FfxFloat32x4 v1, FfxFloat32x4 v2, FfxFloat32x4 v3) -{ - return (v0 + v1 + v2 + v3) * 0.25f; -} -#endif - -// define fetch and store functions Packed -#if FFX_HALF -#error Callback must be implemented - -FFX_GROUPSHARED FfxFloat16x2 spdIntermediateRG[16][16]; -FFX_GROUPSHARED FfxFloat16x2 spdIntermediateBA[16][16]; - -FfxFloat16x4 SpdLoadSourceImageH(FfxFloat32x2 tex, FfxUInt32 slice) -{ - return FfxFloat16x4(imgDst[0][FfxFloat32x3(tex, slice)]); -} -FfxFloat16x4 SpdLoadH(FfxInt32x2 p, FfxUInt32 slice) -{ - return FfxFloat16x4(imgDst6[FfxUInt32x3(p, slice)]); -} -void SpdStoreH(FfxInt32x2 p, FfxFloat16x4 value, FfxUInt32 mip, FfxUInt32 slice) -{ - if (index == LumaMipLevelToUse() || index == 5) - { - imgDst6[FfxUInt32x3(p, slice)] = FfxFloat32x4(value); - return; - } - imgDst[mip + 1][FfxUInt32x3(p, slice)] = FfxFloat32x4(value); -} -void SpdIncreaseAtomicCounter(FfxUInt32 slice) -{ - InterlockedAdd(rw_spd_global_atomic[FfxInt16x2(0, 0)].counter[slice], 1, spdCounter); -} -FfxUInt32 SpdGetAtomicCounter() -{ - return spdCounter; -} -void SpdResetAtomicCounter(FfxUInt32 slice) -{ - rw_spd_global_atomic[FfxInt16x2(0, 0)].counter[slice] = 0; -} -FfxFloat16x4 SpdLoadIntermediateH(FfxUInt32 x, FfxUInt32 y) -{ - return FfxFloat16x4( - spdIntermediateRG[x][y].x, - spdIntermediateRG[x][y].y, - spdIntermediateBA[x][y].x, - spdIntermediateBA[x][y].y); -} -void SpdStoreIntermediateH(FfxUInt32 x, FfxUInt32 y, FfxFloat16x4 value) -{ - spdIntermediateRG[x][y] = value.xy; - spdIntermediateBA[x][y] = value.zw; -} -FfxFloat16x4 SpdReduce4H(FfxFloat16x4 v0, FfxFloat16x4 v1, FfxFloat16x4 v2, FfxFloat16x4 v3) -{ - return (v0 + v1 + v2 + v3) * FfxFloat16(0.25); -} -#endif - -#include "ffx_spd.h" - -void ComputeAutoExposure(FfxUInt32x3 WorkGroupId, FfxUInt32 LocalThreadIndex) -{ -#if FFX_HALF - SpdDownsampleH( - FfxUInt32x2(WorkGroupId.xy), - FfxUInt32(LocalThreadIndex), - FfxUInt32(MipCount()), - FfxUInt32(NumWorkGroups()), - FfxUInt32(WorkGroupId.z), - FfxUInt32x2(WorkGroupOffset())); -#else - SpdDownsample( - FfxUInt32x2(WorkGroupId.xy), - FfxUInt32(LocalThreadIndex), - FfxUInt32(MipCount()), - FfxUInt32(NumWorkGroups()), - FfxUInt32(WorkGroupId.z), - FfxUInt32x2(WorkGroupOffset())); -#endif -} \ No newline at end of file diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_compute_luminance_pyramid.h.meta b/Assets/Resources/FSR2/shaders/ffx_fsr2_compute_luminance_pyramid.h.meta deleted file mode 100644 index c8cb631..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_compute_luminance_pyramid.h.meta +++ /dev/null @@ -1,60 +0,0 @@ -fileFormatVersion: 2 -guid: dbcdb6dfb36311a49aa7b05bc5054280 -PluginImporter: - externalObjects: {} - serializedVersion: 2 - iconMap: {} - executionOrder: {} - defineConstraints: [] - isPreloaded: 0 - isOverridable: 0 - isExplicitlyReferenced: 0 - validateReferences: 1 - platformData: - - first: - : Any - second: - enabled: 0 - settings: - Exclude Editor: 1 - Exclude GameCoreScarlett: 1 - Exclude GameCoreXboxOne: 1 - Exclude Linux64: 1 - Exclude OSXUniversal: 1 - Exclude PS4: 1 - Exclude PS5: 1 - Exclude WebGL: 1 - Exclude Win: 1 - Exclude Win64: 1 - - first: - Any: - second: - enabled: 0 - settings: {} - - first: - Editor: Editor - second: - enabled: 0 - settings: - DefaultValueInitialized: true - - first: - Standalone: Linux64 - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win64 - second: - enabled: 0 - settings: - CPU: None - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_compute_luminance_pyramid_pass.hlsl b/Assets/Resources/FSR2/shaders/ffx_fsr2_compute_luminance_pyramid_pass.hlsl deleted file mode 100644 index 2b96636..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_compute_luminance_pyramid_pass.hlsl +++ /dev/null @@ -1,131 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#define FSR2_BIND_SRV_INPUT_COLOR 0 -#define FSR2_BIND_UAV_SPD_GLOBAL_ATOMIC 0 -#define FSR2_BIND_UAV_EXPOSURE_MIP_LUMA_CHANGE 1 -#define FSR2_BIND_UAV_EXPOSURE_MIP_5 2 -#define FSR2_BIND_UAV_AUTO_EXPOSURE 3 -#define FSR2_BIND_CB_FSR2 0 -#define FSR2_BIND_CB_SPD 1 - -#include "ffx_fsr2_callbacks_hlsl.h" -#include "ffx_fsr2_common.h" - -#if defined(FSR2_BIND_CB_SPD) - cbuffer cbSPD : FFX_FSR2_DECLARE_CB(FSR2_BIND_CB_SPD) { - - FfxUInt32 mips; - FfxUInt32 numWorkGroups; - FfxUInt32x2 workGroupOffset; - FfxUInt32x2 renderSize; - }; - - FfxUInt32 MipCount() - { - return mips; - } - - FfxUInt32 NumWorkGroups() - { - return numWorkGroups; - } - - FfxUInt32x2 WorkGroupOffset() - { - return workGroupOffset; - } - - FfxUInt32x2 SPD_RenderSize() - { - return renderSize; - } -#endif - - -FfxFloat32x2 SPD_LoadExposureBuffer() -{ - return rw_auto_exposure[FfxInt32x2(0,0)]; -} - -void SPD_SetExposureBuffer(FfxFloat32x2 value) -{ - rw_auto_exposure[FfxInt32x2(0,0)] = value; -} - -FfxFloat32x4 SPD_LoadMipmap5(FfxInt32x2 iPxPos) -{ - return FfxFloat32x4(rw_img_mip_5[iPxPos], 0, 0, 0); -} - -void SPD_SetMipmap(FfxInt32x2 iPxPos, FfxInt32 slice, FfxFloat32 value) -{ - switch (slice) - { - case FFX_FSR2_SHADING_CHANGE_MIP_LEVEL: - rw_img_mip_shading_change[iPxPos] = value; - break; - case 5: - rw_img_mip_5[iPxPos] = value; - break; - default: - - // avoid flattened side effect -#if defined(FSR2_BIND_UAV_EXPOSURE_MIP_LUMA_CHANGE) || defined(FFX_INTERNAL) - rw_img_mip_shading_change[iPxPos] = rw_img_mip_shading_change[iPxPos]; -#elif defined(FSR2_BIND_UAV_EXPOSURE_MIP_5) || defined(FFX_INTERNAL) - rw_img_mip_5[iPxPos] = rw_img_mip_5[iPxPos]; -#endif - break; - } -} - -void SPD_IncreaseAtomicCounter(inout FfxUInt32 spdCounter) -{ - InterlockedAdd(rw_spd_global_atomic[FfxInt32x2(0,0)], 1, spdCounter); -} - -void SPD_ResetAtomicCounter() -{ - rw_spd_global_atomic[FfxInt32x2(0,0)] = 0; -} - -#include "ffx_fsr2_compute_luminance_pyramid.h" - -#ifndef FFX_FSR2_THREAD_GROUP_WIDTH -#define FFX_FSR2_THREAD_GROUP_WIDTH 256 -#endif // #ifndef FFX_FSR2_THREAD_GROUP_WIDTH -#ifndef FFX_FSR2_THREAD_GROUP_HEIGHT -#define FFX_FSR2_THREAD_GROUP_HEIGHT 1 -#endif // #ifndef FFX_FSR2_THREAD_GROUP_HEIGHT -#ifndef FFX_FSR2_THREAD_GROUP_DEPTH -#define FFX_FSR2_THREAD_GROUP_DEPTH 1 -#endif // #ifndef FFX_FSR2_THREAD_GROUP_DEPTH -#ifndef FFX_FSR2_NUM_THREADS -#define FFX_FSR2_NUM_THREADS [numthreads(FFX_FSR2_THREAD_GROUP_WIDTH, FFX_FSR2_THREAD_GROUP_HEIGHT, FFX_FSR2_THREAD_GROUP_DEPTH)] -#endif // #ifndef FFX_FSR2_NUM_THREADS - -FFX_FSR2_NUM_THREADS -FFX_FSR2_EMBED_CB2_ROOTSIG_CONTENT -void CS(uint3 WorkGroupId : SV_GroupID, uint LocalThreadIndex : SV_GroupIndex) -{ - ComputeAutoExposure(WorkGroupId, LocalThreadIndex); -} diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_compute_luminance_pyramid_pass.hlsl.meta b/Assets/Resources/FSR2/shaders/ffx_fsr2_compute_luminance_pyramid_pass.hlsl.meta deleted file mode 100644 index ac39806..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_compute_luminance_pyramid_pass.hlsl.meta +++ /dev/null @@ -1,7 +0,0 @@ -fileFormatVersion: 2 -guid: 19dfb00afb70c3144b43ec2dc05ecdd9 -ShaderIncludeImporter: - externalObjects: {} - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_depth_clip.h b/Assets/Resources/FSR2/shaders/ffx_fsr2_depth_clip.h deleted file mode 100644 index fa4c975..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_depth_clip.h +++ /dev/null @@ -1,258 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#ifndef FFX_FSR2_DEPTH_CLIP_H -#define FFX_FSR2_DEPTH_CLIP_H - -FFX_STATIC const FfxFloat32 DepthClipBaseScale = 4.0f; - -FfxFloat32 ComputeDepthClip(FfxFloat32x2 fUvSample, FfxFloat32 fCurrentDepthSample) -{ - FfxFloat32 fCurrentDepthViewSpace = GetViewSpaceDepth(fCurrentDepthSample); - BilinearSamplingData bilinearInfo = GetBilinearSamplingData(fUvSample, RenderSize()); - - FfxFloat32 fDilatedSum = 0.0f; - FfxFloat32 fDepth = 0.0f; - FfxFloat32 fWeightSum = 0.0f; - for (FfxInt32 iSampleIndex = 0; iSampleIndex < 4; iSampleIndex++) { - - const FfxInt32x2 iOffset = bilinearInfo.iOffsets[iSampleIndex]; - const FfxInt32x2 iSamplePos = bilinearInfo.iBasePos + iOffset; - - if (IsOnScreen(iSamplePos, RenderSize())) { - const FfxFloat32 fWeight = bilinearInfo.fWeights[iSampleIndex]; - if (fWeight > fReconstructedDepthBilinearWeightThreshold) { - - const FfxFloat32 fPrevDepthSample = LoadReconstructedPrevDepth(iSamplePos); - const FfxFloat32 fPrevNearestDepthViewSpace = GetViewSpaceDepth(fPrevDepthSample); - - const FfxFloat32 fDepthDiff = fCurrentDepthViewSpace - fPrevNearestDepthViewSpace; - - if (fDepthDiff > 0.0f) { - -#if FFX_FSR2_OPTION_INVERTED_DEPTH - const FfxFloat32 fPlaneDepth = ffxMin(fPrevDepthSample, fCurrentDepthSample); -#else - const FfxFloat32 fPlaneDepth = ffxMax(fPrevDepthSample, fCurrentDepthSample); -#endif - - const FfxFloat32x3 fCenter = GetViewSpacePosition(FfxInt32x2(RenderSize() * 0.5f), RenderSize(), fPlaneDepth); - const FfxFloat32x3 fCorner = GetViewSpacePosition(FfxInt32x2(0, 0), RenderSize(), fPlaneDepth); - - const FfxFloat32 fHalfViewportWidth = length(FfxFloat32x2(RenderSize())); - const FfxFloat32 fDepthThreshold = ffxMax(fCurrentDepthViewSpace, fPrevNearestDepthViewSpace); - - const FfxFloat32 Ksep = 1.37e-05f; - const FfxFloat32 Kfov = length(fCorner) / length(fCenter); - const FfxFloat32 fRequiredDepthSeparation = Ksep * Kfov * fHalfViewportWidth * fDepthThreshold; - - const FfxFloat32 fResolutionFactor = ffxSaturate(length(FfxFloat32x2(RenderSize())) / length(FfxFloat32x2(1920.0f, 1080.0f))); - const FfxFloat32 fPower = ffxLerp(1.0f, 3.0f, fResolutionFactor); - fDepth += ffxPow(ffxSaturate(FfxFloat32(fRequiredDepthSeparation / fDepthDiff)), fPower) * fWeight; - fWeightSum += fWeight; - } - } - } - } - - return (fWeightSum > 0) ? ffxSaturate(1.0f - fDepth / fWeightSum) : 0.0f; -} - -FfxFloat32 ComputeMotionDivergence(FfxInt32x2 iPxPos, FfxInt32x2 iPxInputMotionVectorSize) -{ - FfxFloat32 minconvergence = 1.0f; - - FfxFloat32x2 fMotionVectorNucleus = LoadInputMotionVector(iPxPos); - FfxFloat32 fNucleusVelocityLr = length(fMotionVectorNucleus * RenderSize()); - FfxFloat32 fMaxVelocityUv = length(fMotionVectorNucleus); - - const FfxFloat32 MotionVectorVelocityEpsilon = 1e-02f; - - if (fNucleusVelocityLr > MotionVectorVelocityEpsilon) { - for (FfxInt32 y = -1; y <= 1; ++y) { - for (FfxInt32 x = -1; x <= 1; ++x) { - - FfxInt32x2 sp = ClampLoad(iPxPos, FfxInt32x2(x, y), iPxInputMotionVectorSize); - - FfxFloat32x2 fMotionVector = LoadInputMotionVector(sp); - FfxFloat32 fVelocityUv = length(fMotionVector); - - fMaxVelocityUv = ffxMax(fVelocityUv, fMaxVelocityUv); - fVelocityUv = ffxMax(fVelocityUv, fMaxVelocityUv); - minconvergence = ffxMin(minconvergence, dot(fMotionVector / fVelocityUv, fMotionVectorNucleus / fVelocityUv)); - } - } - } - - return ffxSaturate(1.0f - minconvergence) * ffxSaturate(fMaxVelocityUv / 0.01f); -} - -FfxFloat32 ComputeDepthDivergence(FfxInt32x2 iPxPos) -{ - const FfxFloat32 fMaxDistInMeters = GetMaxDistanceInMeters(); - FfxFloat32 fDepthMax = 0.0f; - FfxFloat32 fDepthMin = fMaxDistInMeters; - - FfxInt32 iMaxDistFound = 0; - - for (FfxInt32 y = -1; y < 2; y++) { - for (FfxInt32 x = -1; x < 2; x++) { - - const FfxInt32x2 iOffset = FfxInt32x2(x, y); - const FfxInt32x2 iSamplePos = iPxPos + iOffset; - - const FfxFloat32 fOnScreenFactor = IsOnScreen(iSamplePos, RenderSize()) ? 1.0f : 0.0f; - FfxFloat32 fDepth = GetViewSpaceDepthInMeters(LoadDilatedDepth(iSamplePos)) * fOnScreenFactor; - - iMaxDistFound |= FfxInt32(fMaxDistInMeters == fDepth); - - fDepthMin = ffxMin(fDepthMin, fDepth); - fDepthMax = ffxMax(fDepthMax, fDepth); - } - } - - return (1.0f - fDepthMin / fDepthMax) * (FfxBoolean(iMaxDistFound) ? 0.0f : 1.0f); -} - -FfxFloat32 ComputeTemporalMotionDivergence(FfxInt32x2 iPxPos) -{ - const FfxFloat32x2 fUv = FfxFloat32x2(iPxPos + 0.5f) / RenderSize(); - - FfxFloat32x2 fMotionVector = LoadDilatedMotionVector(iPxPos); - FfxFloat32x2 fReprojectedUv = fUv + fMotionVector; - fReprojectedUv = ClampUv(fReprojectedUv, RenderSize(), MaxRenderSize()); - FfxFloat32x2 fPrevMotionVector = SamplePreviousDilatedMotionVector(fReprojectedUv); - - float fPxDistance = length(fMotionVector * DisplaySize()); - return fPxDistance > 1.0f ? ffxLerp(0.0f, 1.0f - ffxSaturate(length(fPrevMotionVector) / length(fMotionVector)), ffxSaturate(ffxPow(fPxDistance / 20.0f, 3.0f))) : 0; -} - -void PreProcessReactiveMasks(FfxInt32x2 iPxLrPos, FfxFloat32 fMotionDivergence) -{ - // Compensate for bilinear sampling in accumulation pass - - FfxFloat32x3 fReferenceColor = LoadInputColor(iPxLrPos).xyz; - FfxFloat32x2 fReactiveFactor = FfxFloat32x2(0.0f, fMotionDivergence); - - float fMasksSum = 0.0f; - - FfxFloat32x3 fColorSamples[9]; - FfxFloat32 fReactiveSamples[9]; - FfxFloat32 fTransparencyAndCompositionSamples[9]; - - FFX_UNROLL - for (FfxInt32 y = -1; y < 2; y++) { - FFX_UNROLL - for (FfxInt32 x = -1; x < 2; x++) { - - const FfxInt32x2 sampleCoord = ClampLoad(iPxLrPos, FfxInt32x2(x, y), FfxInt32x2(RenderSize())); - - FfxInt32 sampleIdx = (y + 1) * 3 + x + 1; - - FfxFloat32x3 fColorSample = LoadInputColor(sampleCoord).xyz; - FfxFloat32 fReactiveSample = LoadReactiveMask(sampleCoord); - FfxFloat32 fTransparencyAndCompositionSample = LoadTransparencyAndCompositionMask(sampleCoord); - - fColorSamples[sampleIdx] = fColorSample; - fReactiveSamples[sampleIdx] = fReactiveSample; - fTransparencyAndCompositionSamples[sampleIdx] = fTransparencyAndCompositionSample; - - fMasksSum += (fReactiveSample + fTransparencyAndCompositionSample); - } - } - - if (fMasksSum > 0) - { - for (FfxInt32 sampleIdx = 0; sampleIdx < 9; sampleIdx++) - { - FfxFloat32x3 fColorSample = fColorSamples[sampleIdx]; - FfxFloat32 fReactiveSample = fReactiveSamples[sampleIdx]; - FfxFloat32 fTransparencyAndCompositionSample = fTransparencyAndCompositionSamples[sampleIdx]; - - const FfxFloat32 fMaxLenSq = ffxMax(dot(fReferenceColor, fReferenceColor), dot(fColorSample, fColorSample)); - const FfxFloat32 fSimilarity = dot(fReferenceColor, fColorSample) / fMaxLenSq; - - // Increase power for non-similar samples - const FfxFloat32 fPowerBiasMax = 6.0f; - const FfxFloat32 fSimilarityPower = 1.0f + (fPowerBiasMax - fSimilarity * fPowerBiasMax); - const FfxFloat32 fWeightedReactiveSample = ffxPow(fReactiveSample, fSimilarityPower); - const FfxFloat32 fWeightedTransparencyAndCompositionSample = ffxPow(fTransparencyAndCompositionSample, fSimilarityPower); - - fReactiveFactor = ffxMax(fReactiveFactor, FfxFloat32x2(fWeightedReactiveSample, fWeightedTransparencyAndCompositionSample)); - } - } - - StoreDilatedReactiveMasks(iPxLrPos, fReactiveFactor); -} - -FfxFloat32x3 ComputePreparedInputColor(FfxInt32x2 iPxLrPos) -{ - //We assume linear data. if non-linear input (sRGB, ...), - //then we should convert to linear first and back to sRGB on output. - FfxFloat32x3 fRgb = ffxMax(FfxFloat32x3(0, 0, 0), LoadInputColor(iPxLrPos)); - - fRgb = PrepareRgb(fRgb, Exposure(), PreExposure()); - - const FfxFloat32x3 fPreparedYCoCg = RGBToYCoCg(fRgb); - - return fPreparedYCoCg; -} - -FfxFloat32 EvaluateSurface(FfxInt32x2 iPxPos, FfxFloat32x2 fMotionVector) -{ - FfxFloat32 d0 = GetViewSpaceDepth(LoadReconstructedPrevDepth(iPxPos + FfxInt32x2(0, -1))); - FfxFloat32 d1 = GetViewSpaceDepth(LoadReconstructedPrevDepth(iPxPos + FfxInt32x2(0, 0))); - FfxFloat32 d2 = GetViewSpaceDepth(LoadReconstructedPrevDepth(iPxPos + FfxInt32x2(0, 1))); - - return 1.0f - FfxFloat32(((d0 - d1) > (d1 * 0.01f)) && ((d1 - d2) > (d2 * 0.01f))); -} - -void DepthClip(FfxInt32x2 iPxPos) -{ - FfxFloat32x2 fDepthUv = (iPxPos + 0.5f) / RenderSize(); - FfxFloat32x2 fMotionVector = LoadDilatedMotionVector(iPxPos); - - // Discard tiny mvs - fMotionVector *= FfxFloat32(length(fMotionVector * DisplaySize()) > 0.01f); - - const FfxFloat32x2 fDilatedUv = fDepthUv + fMotionVector; - const FfxFloat32 fDilatedDepth = LoadDilatedDepth(iPxPos); - const FfxFloat32 fCurrentDepthViewSpace = GetViewSpaceDepth(LoadInputDepth(iPxPos)); - - // Compute prepared input color and depth clip - FfxFloat32 fDepthClip = ComputeDepthClip(fDilatedUv, fDilatedDepth) * EvaluateSurface(iPxPos, fMotionVector); - FfxFloat32x3 fPreparedYCoCg = ComputePreparedInputColor(iPxPos); - StorePreparedInputColor(iPxPos, FfxFloat32x4(fPreparedYCoCg, fDepthClip)); - - // Compute dilated reactive mask -#if FFX_FSR2_OPTION_LOW_RESOLUTION_MOTION_VECTORS - FfxInt32x2 iSamplePos = iPxPos; -#else - FfxInt32x2 iSamplePos = ComputeHrPosFromLrPos(iPxPos); -#endif - - FfxFloat32 fMotionDivergence = ComputeMotionDivergence(iSamplePos, RenderSize()); - FfxFloat32 fTemporalMotionDifference = ffxSaturate(ComputeTemporalMotionDivergence(iPxPos) - ComputeDepthDivergence(iPxPos)); - - PreProcessReactiveMasks(iPxPos, ffxMax(fTemporalMotionDifference, fMotionDivergence)); -} - -#endif //!defined( FFX_FSR2_DEPTH_CLIPH ) \ No newline at end of file diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_depth_clip.h.meta b/Assets/Resources/FSR2/shaders/ffx_fsr2_depth_clip.h.meta deleted file mode 100644 index dbdf299..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_depth_clip.h.meta +++ /dev/null @@ -1,60 +0,0 @@ -fileFormatVersion: 2 -guid: f7c16477aeb3a9b4f94f4ef818d10d9b -PluginImporter: - externalObjects: {} - serializedVersion: 2 - iconMap: {} - executionOrder: {} - defineConstraints: [] - isPreloaded: 0 - isOverridable: 0 - isExplicitlyReferenced: 0 - validateReferences: 1 - platformData: - - first: - : Any - second: - enabled: 0 - settings: - Exclude Editor: 1 - Exclude GameCoreScarlett: 1 - Exclude GameCoreXboxOne: 1 - Exclude Linux64: 1 - Exclude OSXUniversal: 1 - Exclude PS4: 1 - Exclude PS5: 1 - Exclude WebGL: 1 - Exclude Win: 1 - Exclude Win64: 1 - - first: - Any: - second: - enabled: 0 - settings: {} - - first: - Editor: Editor - second: - enabled: 0 - settings: - DefaultValueInitialized: true - - first: - Standalone: Linux64 - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win64 - second: - enabled: 0 - settings: - CPU: None - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_depth_clip_pass.hlsl b/Assets/Resources/FSR2/shaders/ffx_fsr2_depth_clip_pass.hlsl deleted file mode 100644 index 3cf501c..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_depth_clip_pass.hlsl +++ /dev/null @@ -1,66 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#define FSR2_BIND_SRV_RECONSTRUCTED_PREV_NEAREST_DEPTH 0 -#define FSR2_BIND_SRV_DILATED_MOTION_VECTORS 1 -#define FSR2_BIND_SRV_DILATED_DEPTH 2 -#define FSR2_BIND_SRV_REACTIVE_MASK 3 -#define FSR2_BIND_SRV_TRANSPARENCY_AND_COMPOSITION_MASK 4 -#define FSR2_BIND_SRV_PREVIOUS_DILATED_MOTION_VECTORS 5 -#define FSR2_BIND_SRV_INPUT_MOTION_VECTORS 6 -#define FSR2_BIND_SRV_INPUT_COLOR 7 -#define FSR2_BIND_SRV_INPUT_DEPTH 8 -#define FSR2_BIND_SRV_INPUT_EXPOSURE 9 - -#define FSR2_BIND_UAV_DILATED_REACTIVE_MASKS 0 -#define FSR2_BIND_UAV_PREPARED_INPUT_COLOR 1 - -#define FSR2_BIND_CB_FSR2 0 - -#include "ffx_fsr2_callbacks_hlsl.h" -#include "ffx_fsr2_common.h" -#include "ffx_fsr2_sample.h" -#include "ffx_fsr2_depth_clip.h" - -#ifndef FFX_FSR2_THREAD_GROUP_WIDTH -#define FFX_FSR2_THREAD_GROUP_WIDTH 8 -#endif // #ifndef FFX_FSR2_THREAD_GROUP_WIDTH -#ifndef FFX_FSR2_THREAD_GROUP_HEIGHT -#define FFX_FSR2_THREAD_GROUP_HEIGHT 8 -#endif // #ifndef FFX_FSR2_THREAD_GROUP_HEIGHT -#ifndef FFX_FSR2_THREAD_GROUP_DEPTH -#define FFX_FSR2_THREAD_GROUP_DEPTH 1 -#endif // #ifndef FFX_FSR2_THREAD_GROUP_DEPTH -#ifndef FFX_FSR2_NUM_THREADS -#define FFX_FSR2_NUM_THREADS [numthreads(FFX_FSR2_THREAD_GROUP_WIDTH, FFX_FSR2_THREAD_GROUP_HEIGHT, FFX_FSR2_THREAD_GROUP_DEPTH)] -#endif // #ifndef FFX_FSR2_NUM_THREADS - -FFX_FSR2_PREFER_WAVE64 -FFX_FSR2_NUM_THREADS -FFX_FSR2_EMBED_ROOTSIG_CONTENT -void CS( - int2 iGroupId : SV_GroupID, - int2 iDispatchThreadId : SV_DispatchThreadID, - int2 iGroupThreadId : SV_GroupThreadID, - int iGroupIndex : SV_GroupIndex) -{ - DepthClip(iDispatchThreadId); -} diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_depth_clip_pass.hlsl.meta b/Assets/Resources/FSR2/shaders/ffx_fsr2_depth_clip_pass.hlsl.meta deleted file mode 100644 index d7ce66d..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_depth_clip_pass.hlsl.meta +++ /dev/null @@ -1,7 +0,0 @@ -fileFormatVersion: 2 -guid: 7981b48622ddaa944909ebf209284d83 -ShaderIncludeImporter: - externalObjects: {} - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_lock.h b/Assets/Resources/FSR2/shaders/ffx_fsr2_lock.h deleted file mode 100644 index 8347fa8..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_lock.h +++ /dev/null @@ -1,115 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#ifndef FFX_FSR2_LOCK_H -#define FFX_FSR2_LOCK_H - -void ClearResourcesForNextFrame(in FfxInt32x2 iPxHrPos) -{ - if (all(FFX_LESS_THAN(iPxHrPos, FfxInt32x2(RenderSize())))) - { -#if FFX_FSR2_OPTION_INVERTED_DEPTH - const FfxUInt32 farZ = 0x0; -#else - const FfxUInt32 farZ = 0x3f800000; -#endif - SetReconstructedDepth(iPxHrPos, farZ); - } -} - -FfxBoolean ComputeThinFeatureConfidence(FfxInt32x2 pos) -{ - const FfxInt32 RADIUS = 1; - - FfxFloat32 fNucleus = LoadLockInputLuma(pos); - - FfxFloat32 similar_threshold = 1.05f; - FfxFloat32 dissimilarLumaMin = FSR2_FLT_MAX; - FfxFloat32 dissimilarLumaMax = 0; - - /* - 0 1 2 - 3 4 5 - 6 7 8 - */ - - #define SETBIT(x) (1U << x) - - FfxUInt32 mask = SETBIT(4); //flag fNucleus as similar - - const FfxUInt32 uNumRejectionMasks = 4; - const FfxUInt32 uRejectionMasks[uNumRejectionMasks] = { - SETBIT(0) | SETBIT(1) | SETBIT(3) | SETBIT(4), //Upper left - SETBIT(1) | SETBIT(2) | SETBIT(4) | SETBIT(5), //Upper right - SETBIT(3) | SETBIT(4) | SETBIT(6) | SETBIT(7), //Lower left - SETBIT(4) | SETBIT(5) | SETBIT(7) | SETBIT(8), //Lower right - }; - - FfxInt32 idx = 0; - FFX_UNROLL - for (FfxInt32 y = -RADIUS; y <= RADIUS; y++) { - FFX_UNROLL - for (FfxInt32 x = -RADIUS; x <= RADIUS; x++, idx++) { - if (x == 0 && y == 0) continue; - - FfxInt32x2 samplePos = ClampLoad(pos, FfxInt32x2(x, y), FfxInt32x2(RenderSize())); - - FfxFloat32 sampleLuma = LoadLockInputLuma(samplePos); - FfxFloat32 difference = ffxMax(sampleLuma, fNucleus) / ffxMin(sampleLuma, fNucleus); - - if (difference > 0 && (difference < similar_threshold)) { - mask |= SETBIT(idx); - } else { - dissimilarLumaMin = ffxMin(dissimilarLumaMin, sampleLuma); - dissimilarLumaMax = ffxMax(dissimilarLumaMax, sampleLuma); - } - } - } - - FfxBoolean isRidge = fNucleus > dissimilarLumaMax || fNucleus < dissimilarLumaMin; - - if (FFX_FALSE == isRidge) { - - return false; - } - - FFX_UNROLL - for (FfxInt32 i = 0; i < 4; i++) { - - if ((mask & uRejectionMasks[i]) == uRejectionMasks[i]) { - return false; - } - } - - return true; -} - -void ComputeLock(FfxInt32x2 iPxLrPos) -{ - if (ComputeThinFeatureConfidence(iPxLrPos)) - { - StoreNewLocks(ComputeHrPosFromLrPos(iPxLrPos), 1.f); - } - - ClearResourcesForNextFrame(iPxLrPos); -} - -#endif // FFX_FSR2_LOCK_H diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_lock.h.meta b/Assets/Resources/FSR2/shaders/ffx_fsr2_lock.h.meta deleted file mode 100644 index 21f4814..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_lock.h.meta +++ /dev/null @@ -1,60 +0,0 @@ -fileFormatVersion: 2 -guid: 9b4cdc5f81194ac4fa946c31b86234ed -PluginImporter: - externalObjects: {} - serializedVersion: 2 - iconMap: {} - executionOrder: {} - defineConstraints: [] - isPreloaded: 0 - isOverridable: 0 - isExplicitlyReferenced: 0 - validateReferences: 1 - platformData: - - first: - : Any - second: - enabled: 0 - settings: - Exclude Editor: 1 - Exclude GameCoreScarlett: 1 - Exclude GameCoreXboxOne: 1 - Exclude Linux64: 1 - Exclude OSXUniversal: 1 - Exclude PS4: 1 - Exclude PS5: 1 - Exclude WebGL: 1 - Exclude Win: 1 - Exclude Win64: 1 - - first: - Any: - second: - enabled: 0 - settings: {} - - first: - Editor: Editor - second: - enabled: 0 - settings: - DefaultValueInitialized: true - - first: - Standalone: Linux64 - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win64 - second: - enabled: 0 - settings: - CPU: None - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_lock_pass.hlsl b/Assets/Resources/FSR2/shaders/ffx_fsr2_lock_pass.hlsl deleted file mode 100644 index 1409dce..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_lock_pass.hlsl +++ /dev/null @@ -1,53 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#define FSR2_BIND_SRV_LOCK_INPUT_LUMA 0 -#define FSR2_BIND_UAV_NEW_LOCKS 0 -#define FSR2_BIND_UAV_RECONSTRUCTED_PREV_NEAREST_DEPTH 1 -#define FSR2_BIND_CB_FSR2 0 - -#include "ffx_fsr2_callbacks_hlsl.h" -#include "ffx_fsr2_common.h" -#include "ffx_fsr2_sample.h" -#include "ffx_fsr2_lock.h" - -#ifndef FFX_FSR2_THREAD_GROUP_WIDTH -#define FFX_FSR2_THREAD_GROUP_WIDTH 8 -#endif // #ifndef FFX_FSR2_THREAD_GROUP_WIDTH -#ifndef FFX_FSR2_THREAD_GROUP_HEIGHT -#define FFX_FSR2_THREAD_GROUP_HEIGHT 8 -#endif // #ifndef FFX_FSR2_THREAD_GROUP_HEIGHT -#ifndef FFX_FSR2_THREAD_GROUP_DEPTH -#define FFX_FSR2_THREAD_GROUP_DEPTH 1 -#endif // #ifndef FFX_FSR2_THREAD_GROUP_DEPTH -#ifndef FFX_FSR2_NUM_THREADS -#define FFX_FSR2_NUM_THREADS [numthreads(FFX_FSR2_THREAD_GROUP_WIDTH, FFX_FSR2_THREAD_GROUP_HEIGHT, FFX_FSR2_THREAD_GROUP_DEPTH)] -#endif // #ifndef FFX_FSR2_NUM_THREADS - -FFX_FSR2_PREFER_WAVE64 -FFX_FSR2_NUM_THREADS -FFX_FSR2_EMBED_ROOTSIG_CONTENT -void CS(uint2 uGroupId : SV_GroupID, uint2 uGroupThreadId : SV_GroupThreadID) -{ - uint2 uDispatchThreadId = uGroupId * uint2(FFX_FSR2_THREAD_GROUP_WIDTH, FFX_FSR2_THREAD_GROUP_HEIGHT) + uGroupThreadId; - - ComputeLock(uDispatchThreadId); -} diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_lock_pass.hlsl.meta b/Assets/Resources/FSR2/shaders/ffx_fsr2_lock_pass.hlsl.meta deleted file mode 100644 index 4ded638..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_lock_pass.hlsl.meta +++ /dev/null @@ -1,7 +0,0 @@ -fileFormatVersion: 2 -guid: 471a3f7a033c72f4fa737d4f8238a9bd -ShaderIncludeImporter: - externalObjects: {} - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_postprocess_lock_status.h b/Assets/Resources/FSR2/shaders/ffx_fsr2_postprocess_lock_status.h deleted file mode 100644 index cee9e14..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_postprocess_lock_status.h +++ /dev/null @@ -1,106 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#ifndef FFX_FSR2_POSTPROCESS_LOCK_STATUS_H -#define FFX_FSR2_POSTPROCESS_LOCK_STATUS_H - -FfxFloat32x4 WrapShadingChangeLuma(FfxInt32x2 iPxSample) -{ - return FfxFloat32x4(LoadMipLuma(iPxSample, LumaMipLevelToUse()), 0, 0, 0); -} - -#if FFX_HALF -FFX_MIN16_F4 WrapShadingChangeLuma(FFX_MIN16_I2 iPxSample) -{ - return FFX_MIN16_F4(LoadMipLuma(iPxSample, LumaMipLevelToUse()), 0, 0, 0); -} -#endif - -#if FFX_FSR2_OPTION_POSTPROCESSLOCKSTATUS_SAMPLERS_USE_DATA_HALF && FFX_HALF -DeclareCustomFetchBilinearSamplesMin16(FetchShadingChangeLumaSamples, WrapShadingChangeLuma) -#else -DeclareCustomFetchBicubicSamples(FetchShadingChangeLumaSamples, WrapShadingChangeLuma) -#endif -DeclareCustomTextureSample(ShadingChangeLumaSample, Lanczos2, FetchShadingChangeLumaSamples) - -FfxFloat32 GetShadingChangeLuma(FfxInt32x2 iPxHrPos, FfxFloat32x2 fUvCoord) -{ - FfxFloat32 fShadingChangeLuma = 0; - -#if 0 - fShadingChangeLuma = Exposure() * exp(ShadingChangeLumaSample(fUvCoord, LumaMipDimensions()).x); -#else - - const FfxFloat32 fDiv = FfxFloat32(2 << LumaMipLevelToUse()); - FfxInt32x2 iMipRenderSize = FfxInt32x2(RenderSize() / fDiv); - - fUvCoord = ClampUv(fUvCoord, iMipRenderSize, LumaMipDimensions()); - fShadingChangeLuma = Exposure() * exp(FfxFloat32(SampleMipLuma(fUvCoord, LumaMipLevelToUse()))); -#endif - - fShadingChangeLuma = ffxPow(fShadingChangeLuma, 1.0f / 6.0f); - - return fShadingChangeLuma; -} - -void UpdateLockStatus(AccumulationPassCommonParams params, - FFX_PARAMETER_INOUT FfxFloat32 fReactiveFactor, LockState state, - FFX_PARAMETER_INOUT FfxFloat32x2 fLockStatus, - FFX_PARAMETER_OUT FfxFloat32 fLockContributionThisFrame, - FFX_PARAMETER_OUT FfxFloat32 fLuminanceDiff) { - - const FfxFloat32 fShadingChangeLuma = GetShadingChangeLuma(params.iPxHrPos, params.fHrUv); - - //init temporal shading change factor, init to -1 or so in reproject to know if "true new"? - fLockStatus[LOCK_TEMPORAL_LUMA] = (fLockStatus[LOCK_TEMPORAL_LUMA] == FfxFloat32(0.0f)) ? fShadingChangeLuma : fLockStatus[LOCK_TEMPORAL_LUMA]; - - FfxFloat32 fPreviousShadingChangeLuma = fLockStatus[LOCK_TEMPORAL_LUMA]; - - fLuminanceDiff = 1.0f - MinDividedByMax(fPreviousShadingChangeLuma, fShadingChangeLuma); - - if (state.NewLock) { - fLockStatus[LOCK_TEMPORAL_LUMA] = fShadingChangeLuma; - - fLockStatus[LOCK_LIFETIME_REMAINING] = (fLockStatus[LOCK_LIFETIME_REMAINING] != 0.0f) ? 2.0f : 1.0f; - } - else if(fLockStatus[LOCK_LIFETIME_REMAINING] <= 1.0f) { - fLockStatus[LOCK_TEMPORAL_LUMA] = ffxLerp(fLockStatus[LOCK_TEMPORAL_LUMA], FfxFloat32(fShadingChangeLuma), 0.5f); - } - else { - if (fLuminanceDiff > 0.1f) { - KillLock(fLockStatus); - } - } - - fReactiveFactor = ffxMax(fReactiveFactor, ffxSaturate((fLuminanceDiff - 0.1f) * 10.0f)); - fLockStatus[LOCK_LIFETIME_REMAINING] *= (1.0f - fReactiveFactor); - - fLockStatus[LOCK_LIFETIME_REMAINING] *= ffxSaturate(1.0f - params.fAccumulationMask); - fLockStatus[LOCK_LIFETIME_REMAINING] *= FfxFloat32(params.fDepthClipFactor < 0.1f); - - // Compute this frame lock contribution - const FfxFloat32 fLifetimeContribution = ffxSaturate(fLockStatus[LOCK_LIFETIME_REMAINING] - 1.0f); - const FfxFloat32 fShadingChangeContribution = ffxSaturate(MinDividedByMax(fLockStatus[LOCK_TEMPORAL_LUMA], fShadingChangeLuma)); - - fLockContributionThisFrame = ffxSaturate(ffxSaturate(fLifetimeContribution * 4.0f) * fShadingChangeContribution); -} - -#endif //!defined( FFX_FSR2_POSTPROCESS_LOCK_STATUS_H ) diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_postprocess_lock_status.h.meta b/Assets/Resources/FSR2/shaders/ffx_fsr2_postprocess_lock_status.h.meta deleted file mode 100644 index ddb6dda..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_postprocess_lock_status.h.meta +++ /dev/null @@ -1,60 +0,0 @@ -fileFormatVersion: 2 -guid: 3945c3cfd2cc1a64cb0513864d88d8ca -PluginImporter: - externalObjects: {} - serializedVersion: 2 - iconMap: {} - executionOrder: {} - defineConstraints: [] - isPreloaded: 0 - isOverridable: 0 - isExplicitlyReferenced: 0 - validateReferences: 1 - platformData: - - first: - : Any - second: - enabled: 0 - settings: - Exclude Editor: 1 - Exclude GameCoreScarlett: 1 - Exclude GameCoreXboxOne: 1 - Exclude Linux64: 1 - Exclude OSXUniversal: 1 - Exclude PS4: 1 - Exclude PS5: 1 - Exclude WebGL: 1 - Exclude Win: 1 - Exclude Win64: 1 - - first: - Any: - second: - enabled: 0 - settings: {} - - first: - Editor: Editor - second: - enabled: 0 - settings: - DefaultValueInitialized: true - - first: - Standalone: Linux64 - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win64 - second: - enabled: 0 - settings: - CPU: None - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_rcas.h b/Assets/Resources/FSR2/shaders/ffx_fsr2_rcas.h deleted file mode 100644 index d9006cd..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_rcas.h +++ /dev/null @@ -1,67 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#define GROUP_SIZE 8 - -#define FSR_RCAS_DENOISE 1 - -void WriteUpscaledOutput(FFX_MIN16_U2 iPxHrPos, FfxFloat32x3 fUpscaledColor) -{ - StoreUpscaledOutput(FFX_MIN16_I2(iPxHrPos), fUpscaledColor); -} - -#define FSR_RCAS_F -FfxFloat32x4 FsrRcasLoadF(FfxInt32x2 p) -{ - FfxFloat32x4 fColor = LoadRCAS_Input(p); - - fColor.rgb = PrepareRgb(fColor.rgb, Exposure(), PreExposure()); - - return fColor; -} - -void FsrRcasInputF(inout FfxFloat32 r, inout FfxFloat32 g, inout FfxFloat32 b) {} - -#include "ffx_fsr1.h" - - -void CurrFilter(FFX_MIN16_U2 pos) -{ - FfxFloat32x3 c; - FsrRcasF(c.r, c.g, c.b, pos, RCASConfig()); - - c = UnprepareRgb(c, Exposure()); - - WriteUpscaledOutput(pos, c); -} - -void RCAS(FfxUInt32x3 LocalThreadId, FfxUInt32x3 WorkGroupId, FfxUInt32x3 Dtid) -{ - // Do remapping of local xy in workgroup for a more PS-like swizzle pattern. - FfxUInt32x2 gxy = ffxRemapForQuad(LocalThreadId.x) + FfxUInt32x2(WorkGroupId.x << 4u, WorkGroupId.y << 4u); - CurrFilter(FFX_MIN16_U2(gxy)); - gxy.x += 8u; - CurrFilter(FFX_MIN16_U2(gxy)); - gxy.y += 8u; - CurrFilter(FFX_MIN16_U2(gxy)); - gxy.x -= 8u; - CurrFilter(FFX_MIN16_U2(gxy)); -} diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_rcas.h.meta b/Assets/Resources/FSR2/shaders/ffx_fsr2_rcas.h.meta deleted file mode 100644 index e962499..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_rcas.h.meta +++ /dev/null @@ -1,60 +0,0 @@ -fileFormatVersion: 2 -guid: 7bd7d4eb34c626342966cb9b3fe00363 -PluginImporter: - externalObjects: {} - serializedVersion: 2 - iconMap: {} - executionOrder: {} - defineConstraints: [] - isPreloaded: 0 - isOverridable: 0 - isExplicitlyReferenced: 0 - validateReferences: 1 - platformData: - - first: - : Any - second: - enabled: 0 - settings: - Exclude Editor: 1 - Exclude GameCoreScarlett: 1 - Exclude GameCoreXboxOne: 1 - Exclude Linux64: 1 - Exclude OSXUniversal: 1 - Exclude PS4: 1 - Exclude PS5: 1 - Exclude WebGL: 1 - Exclude Win: 1 - Exclude Win64: 1 - - first: - Any: - second: - enabled: 0 - settings: {} - - first: - Editor: Editor - second: - enabled: 0 - settings: - DefaultValueInitialized: true - - first: - Standalone: Linux64 - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win64 - second: - enabled: 0 - settings: - CPU: None - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_rcas_pass.hlsl b/Assets/Resources/FSR2/shaders/ffx_fsr2_rcas_pass.hlsl deleted file mode 100644 index f447b7e..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_rcas_pass.hlsl +++ /dev/null @@ -1,75 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#define FSR2_BIND_SRV_INPUT_EXPOSURE 0 -#define FSR2_BIND_SRV_RCAS_INPUT 1 -#define FSR2_BIND_UAV_UPSCALED_OUTPUT 0 -#define FSR2_BIND_CB_FSR2 0 -#define FSR2_BIND_CB_RCAS 1 - -#include "ffx_fsr2_callbacks_hlsl.h" -#include "ffx_fsr2_common.h" - -//Move to prototype shader! -#if defined(FSR2_BIND_CB_RCAS) - cbuffer cbRCAS : FFX_FSR2_DECLARE_CB(FSR2_BIND_CB_RCAS) - { - uint4 rcasConfig; - }; - - uint4 RCASConfig() - { - return rcasConfig; - } -#else - uint4 RCASConfig() - { - return 0; - } -#endif - - -float4 LoadRCAS_Input(FfxInt32x2 iPxPos) -{ - return r_rcas_input[iPxPos]; -} - -#include "ffx_fsr2_rcas.h" - -#ifndef FFX_FSR2_THREAD_GROUP_WIDTH -#define FFX_FSR2_THREAD_GROUP_WIDTH 64 -#endif // #ifndef FFX_FSR2_THREAD_GROUP_WIDTH -#ifndef FFX_FSR2_THREAD_GROUP_HEIGHT -#define FFX_FSR2_THREAD_GROUP_HEIGHT 1 -#endif // #ifndef FFX_FSR2_THREAD_GROUP_HEIGHT -#ifndef FFX_FSR2_THREAD_GROUP_DEPTH -#define FFX_FSR2_THREAD_GROUP_DEPTH 1 -#endif // #ifndef FFX_FSR2_THREAD_GROUP_DEPTH -#ifndef FFX_FSR2_NUM_THREADS -#define FFX_FSR2_NUM_THREADS [numthreads(FFX_FSR2_THREAD_GROUP_WIDTH, FFX_FSR2_THREAD_GROUP_HEIGHT, FFX_FSR2_THREAD_GROUP_DEPTH)] -#endif // #ifndef FFX_FSR2_NUM_THREADS - -FFX_FSR2_NUM_THREADS -FFX_FSR2_EMBED_CB2_ROOTSIG_CONTENT -void CS(uint3 LocalThreadId : SV_GroupThreadID, uint3 WorkGroupId : SV_GroupID, uint3 Dtid : SV_DispatchThreadID) -{ - RCAS(LocalThreadId, WorkGroupId, Dtid); -} diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_rcas_pass.hlsl.meta b/Assets/Resources/FSR2/shaders/ffx_fsr2_rcas_pass.hlsl.meta deleted file mode 100644 index b7e089b..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_rcas_pass.hlsl.meta +++ /dev/null @@ -1,7 +0,0 @@ -fileFormatVersion: 2 -guid: 871ca1938c701d64f94ef8ec00ef06f4 -ShaderIncludeImporter: - externalObjects: {} - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_reconstruct_dilated_velocity_and_previous_depth.h b/Assets/Resources/FSR2/shaders/ffx_fsr2_reconstruct_dilated_velocity_and_previous_depth.h deleted file mode 100644 index e9ccc4b..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_reconstruct_dilated_velocity_and_previous_depth.h +++ /dev/null @@ -1,145 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#ifndef FFX_FSR2_RECONSTRUCT_DILATED_VELOCITY_AND_PREVIOUS_DEPTH_H -#define FFX_FSR2_RECONSTRUCT_DILATED_VELOCITY_AND_PREVIOUS_DEPTH_H - -void ReconstructPrevDepth(FfxInt32x2 iPxPos, FfxFloat32 fDepth, FfxFloat32x2 fMotionVector, FfxInt32x2 iPxDepthSize) -{ - fMotionVector *= FfxFloat32(length(fMotionVector * DisplaySize()) > 0.1f); - - FfxFloat32x2 fUv = (iPxPos + FfxFloat32(0.5)) / iPxDepthSize; - FfxFloat32x2 fReprojectedUv = fUv + fMotionVector; - - BilinearSamplingData bilinearInfo = GetBilinearSamplingData(fReprojectedUv, RenderSize()); - - // Project current depth into previous frame locations. - // Push to all pixels having some contribution if reprojection is using bilinear logic. - for (FfxInt32 iSampleIndex = 0; iSampleIndex < 4; iSampleIndex++) { - - const FfxInt32x2 iOffset = bilinearInfo.iOffsets[iSampleIndex]; - FfxFloat32 fWeight = bilinearInfo.fWeights[iSampleIndex]; - - if (fWeight > fReconstructedDepthBilinearWeightThreshold) { - - FfxInt32x2 iStorePos = bilinearInfo.iBasePos + iOffset; - if (IsOnScreen(iStorePos, iPxDepthSize)) { - StoreReconstructedDepth(iStorePos, fDepth); - } - } - } -} - -void FindNearestDepth(FFX_PARAMETER_IN FfxInt32x2 iPxPos, FFX_PARAMETER_IN FfxInt32x2 iPxSize, FFX_PARAMETER_OUT FfxFloat32 fNearestDepth, FFX_PARAMETER_OUT FfxInt32x2 fNearestDepthCoord) -{ - const FfxInt32 iSampleCount = 9; - const FfxInt32x2 iSampleOffsets[iSampleCount] = { - FfxInt32x2(+0, +0), - FfxInt32x2(+1, +0), - FfxInt32x2(+0, +1), - FfxInt32x2(+0, -1), - FfxInt32x2(-1, +0), - FfxInt32x2(-1, +1), - FfxInt32x2(+1, +1), - FfxInt32x2(-1, -1), - FfxInt32x2(+1, -1), - }; - - // pull out the depth loads to allow SC to batch them - FfxFloat32 depth[9]; - FfxInt32 iSampleIndex = 0; - FFX_UNROLL - for (iSampleIndex = 0; iSampleIndex < iSampleCount; ++iSampleIndex) { - - FfxInt32x2 iPos = iPxPos + iSampleOffsets[iSampleIndex]; - depth[iSampleIndex] = LoadInputDepth(iPos); - } - - // find closest depth - fNearestDepthCoord = iPxPos; - fNearestDepth = depth[0]; - FFX_UNROLL - for (iSampleIndex = 1; iSampleIndex < iSampleCount; ++iSampleIndex) { - - FfxInt32x2 iPos = iPxPos + iSampleOffsets[iSampleIndex]; - if (IsOnScreen(iPos, iPxSize)) { - - FfxFloat32 fNdDepth = depth[iSampleIndex]; -#if FFX_FSR2_OPTION_INVERTED_DEPTH - if (fNdDepth > fNearestDepth) { -#else - if (fNdDepth < fNearestDepth) { -#endif - fNearestDepthCoord = iPos; - fNearestDepth = fNdDepth; - } - } - } -} - -FfxFloat32 ComputeLockInputLuma(FfxInt32x2 iPxLrPos) -{ - //We assume linear data. if non-linear input (sRGB, ...), - //then we should convert to linear first and back to sRGB on output. - FfxFloat32x3 fRgb = ffxMax(FfxFloat32x3(0, 0, 0), LoadInputColor(iPxLrPos)); - - // Use internal auto exposure for locking logic - fRgb /= PreExposure(); - fRgb *= Exposure(); - -#if FFX_FSR2_OPTION_HDR_COLOR_INPUT - fRgb = Tonemap(fRgb); -#endif - - //compute luma used to lock pixels, if used elsewhere the ffxPow must be moved! - const FfxFloat32 fLockInputLuma = ffxPow(RGBToPerceivedLuma(fRgb), FfxFloat32(1.0 / 6.0)); - - return fLockInputLuma; -} - -void ReconstructAndDilate(FfxInt32x2 iPxLrPos) -{ - FfxFloat32 fDilatedDepth; - FfxInt32x2 iNearestDepthCoord; - - FindNearestDepth(iPxLrPos, RenderSize(), fDilatedDepth, iNearestDepthCoord); - -#if FFX_FSR2_OPTION_LOW_RESOLUTION_MOTION_VECTORS - FfxInt32x2 iSamplePos = iPxLrPos; - FfxInt32x2 iMotionVectorPos = iNearestDepthCoord; -#else - FfxInt32x2 iSamplePos = ComputeHrPosFromLrPos(iPxLrPos); - FfxInt32x2 iMotionVectorPos = ComputeHrPosFromLrPos(iNearestDepthCoord); -#endif - - FfxFloat32x2 fDilatedMotionVector = LoadInputMotionVector(iMotionVectorPos); - - StoreDilatedDepth(iPxLrPos, fDilatedDepth); - StoreDilatedMotionVector(iPxLrPos, fDilatedMotionVector); - - ReconstructPrevDepth(iPxLrPos, fDilatedDepth, fDilatedMotionVector, RenderSize()); - - FfxFloat32 fLockInputLuma = ComputeLockInputLuma(iPxLrPos); - StoreLockInputLuma(iPxLrPos, fLockInputLuma); -} - - -#endif //!defined( FFX_FSR2_RECONSTRUCT_DILATED_VELOCITY_AND_PREVIOUS_DEPTH_H ) diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_reconstruct_dilated_velocity_and_previous_depth.h.meta b/Assets/Resources/FSR2/shaders/ffx_fsr2_reconstruct_dilated_velocity_and_previous_depth.h.meta deleted file mode 100644 index 9b82a1a..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_reconstruct_dilated_velocity_and_previous_depth.h.meta +++ /dev/null @@ -1,60 +0,0 @@ -fileFormatVersion: 2 -guid: 13cd33c3d34a317409049dfd939e64ef -PluginImporter: - externalObjects: {} - serializedVersion: 2 - iconMap: {} - executionOrder: {} - defineConstraints: [] - isPreloaded: 0 - isOverridable: 0 - isExplicitlyReferenced: 0 - validateReferences: 1 - platformData: - - first: - : Any - second: - enabled: 0 - settings: - Exclude Editor: 1 - Exclude GameCoreScarlett: 1 - Exclude GameCoreXboxOne: 1 - Exclude Linux64: 1 - Exclude OSXUniversal: 1 - Exclude PS4: 1 - Exclude PS5: 1 - Exclude WebGL: 1 - Exclude Win: 1 - Exclude Win64: 1 - - first: - Any: - second: - enabled: 0 - settings: {} - - first: - Editor: Editor - second: - enabled: 0 - settings: - DefaultValueInitialized: true - - first: - Standalone: Linux64 - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win64 - second: - enabled: 0 - settings: - CPU: None - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_reconstruct_previous_depth_pass.hlsl b/Assets/Resources/FSR2/shaders/ffx_fsr2_reconstruct_previous_depth_pass.hlsl deleted file mode 100644 index 33c044e..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_reconstruct_previous_depth_pass.hlsl +++ /dev/null @@ -1,63 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#define FSR2_BIND_SRV_INPUT_MOTION_VECTORS 0 -#define FSR2_BIND_SRV_INPUT_DEPTH 1 -#define FSR2_BIND_SRV_INPUT_COLOR 2 -#define FSR2_BIND_SRV_INPUT_EXPOSURE 3 - -#define FSR2_BIND_UAV_RECONSTRUCTED_PREV_NEAREST_DEPTH 0 -#define FSR2_BIND_UAV_DILATED_MOTION_VECTORS 1 -#define FSR2_BIND_UAV_DILATED_DEPTH 2 -#define FSR2_BIND_UAV_LOCK_INPUT_LUMA 3 - -#define FSR2_BIND_CB_FSR2 0 - -#include "ffx_fsr2_callbacks_hlsl.h" -#include "ffx_fsr2_common.h" -#include "ffx_fsr2_sample.h" -#include "ffx_fsr2_reconstruct_dilated_velocity_and_previous_depth.h" - -#ifndef FFX_FSR2_THREAD_GROUP_WIDTH -#define FFX_FSR2_THREAD_GROUP_WIDTH 8 -#endif // #ifndef FFX_FSR2_THREAD_GROUP_WIDTH -#ifndef FFX_FSR2_THREAD_GROUP_HEIGHT -#define FFX_FSR2_THREAD_GROUP_HEIGHT 8 -#endif // #ifndef FFX_FSR2_THREAD_GROUP_HEIGHT -#ifndef FFX_FSR2_THREAD_GROUP_DEPTH -#define FFX_FSR2_THREAD_GROUP_DEPTH 1 -#endif // #ifndef FFX_FSR2_THREAD_GROUP_DEPTH -#ifndef FFX_FSR2_NUM_THREADS -#define FFX_FSR2_NUM_THREADS [numthreads(FFX_FSR2_THREAD_GROUP_WIDTH, FFX_FSR2_THREAD_GROUP_HEIGHT, FFX_FSR2_THREAD_GROUP_DEPTH)] -#endif // #ifndef FFX_FSR2_NUM_THREADS - -FFX_FSR2_PREFER_WAVE64 -FFX_FSR2_NUM_THREADS -FFX_FSR2_EMBED_ROOTSIG_CONTENT -void CS( - int2 iGroupId : SV_GroupID, - int2 iDispatchThreadId : SV_DispatchThreadID, - int2 iGroupThreadId : SV_GroupThreadID, - int iGroupIndex : SV_GroupIndex -) -{ - ReconstructAndDilate(iDispatchThreadId); -} diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_reconstruct_previous_depth_pass.hlsl.meta b/Assets/Resources/FSR2/shaders/ffx_fsr2_reconstruct_previous_depth_pass.hlsl.meta deleted file mode 100644 index 4f5036c..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_reconstruct_previous_depth_pass.hlsl.meta +++ /dev/null @@ -1,7 +0,0 @@ -fileFormatVersion: 2 -guid: 9355c255c8505ae48ae89af286943747 -ShaderIncludeImporter: - externalObjects: {} - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_reproject.h b/Assets/Resources/FSR2/shaders/ffx_fsr2_reproject.h deleted file mode 100644 index f7f3961..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_reproject.h +++ /dev/null @@ -1,136 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#ifndef FFX_FSR2_REPROJECT_H -#define FFX_FSR2_REPROJECT_H - -#ifndef FFX_FSR2_OPTION_REPROJECT_USE_LANCZOS_TYPE -#define FFX_FSR2_OPTION_REPROJECT_USE_LANCZOS_TYPE 0 // Reference -#endif - -FfxFloat32x4 WrapHistory(FfxInt32x2 iPxSample) -{ - return LoadHistory(iPxSample); -} - -#if FFX_HALF -FFX_MIN16_F4 WrapHistory(FFX_MIN16_I2 iPxSample) -{ - return FFX_MIN16_F4(LoadHistory(iPxSample)); -} -#endif - - -#if FFX_FSR2_OPTION_REPROJECT_SAMPLERS_USE_DATA_HALF && FFX_HALF -DeclareCustomFetchBicubicSamplesMin16(FetchHistorySamples, WrapHistory) -DeclareCustomTextureSampleMin16(HistorySample, FFX_FSR2_GET_LANCZOS_SAMPLER1D(FFX_FSR2_OPTION_REPROJECT_USE_LANCZOS_TYPE), FetchHistorySamples) -#else -DeclareCustomFetchBicubicSamples(FetchHistorySamples, WrapHistory) -DeclareCustomTextureSample(HistorySample, FFX_FSR2_GET_LANCZOS_SAMPLER1D(FFX_FSR2_OPTION_REPROJECT_USE_LANCZOS_TYPE), FetchHistorySamples) -#endif - -FfxFloat32x4 WrapLockStatus(FfxInt32x2 iPxSample) -{ - FfxFloat32x4 fSample = FfxFloat32x4(LoadLockStatus(iPxSample), 0.0f, 0.0f); - return fSample; -} - -#if FFX_HALF -FFX_MIN16_F4 WrapLockStatus(FFX_MIN16_I2 iPxSample) -{ - FFX_MIN16_F4 fSample = FFX_MIN16_F4(LoadLockStatus(iPxSample), 0.0, 0.0); - - return fSample; -} -#endif - -#if 1 -#if FFX_FSR2_OPTION_REPROJECT_SAMPLERS_USE_DATA_HALF && FFX_HALF -DeclareCustomFetchBilinearSamplesMin16(FetchLockStatusSamples, WrapLockStatus) -DeclareCustomTextureSampleMin16(LockStatusSample, Bilinear, FetchLockStatusSamples) -#else -DeclareCustomFetchBilinearSamples(FetchLockStatusSamples, WrapLockStatus) -DeclareCustomTextureSample(LockStatusSample, Bilinear, FetchLockStatusSamples) -#endif -#else -#if FFX_FSR2_OPTION_REPROJECT_SAMPLERS_USE_DATA_HALF && FFX_HALF -DeclareCustomFetchBicubicSamplesMin16(FetchLockStatusSamples, WrapLockStatus) -DeclareCustomTextureSampleMin16(LockStatusSample, FFX_FSR2_GET_LANCZOS_SAMPLER1D(FFX_FSR2_OPTION_REPROJECT_USE_LANCZOS_TYPE), FetchLockStatusSamples) -#else -DeclareCustomFetchBicubicSamples(FetchLockStatusSamples, WrapLockStatus) -DeclareCustomTextureSample(LockStatusSample, FFX_FSR2_GET_LANCZOS_SAMPLER1D(FFX_FSR2_OPTION_REPROJECT_USE_LANCZOS_TYPE), FetchLockStatusSamples) -#endif -#endif - -FfxFloat32x2 GetMotionVector(FfxInt32x2 iPxHrPos, FfxFloat32x2 fHrUv) -{ -#if FFX_FSR2_OPTION_LOW_RESOLUTION_MOTION_VECTORS - FfxFloat32x2 fDilatedMotionVector = LoadDilatedMotionVector(FFX_MIN16_I2(fHrUv * RenderSize())); -#else - FfxFloat32x2 fDilatedMotionVector = LoadInputMotionVector(iPxHrPos); -#endif - - return fDilatedMotionVector; -} - -FfxBoolean IsUvInside(FfxFloat32x2 fUv) -{ - return (fUv.x >= 0.0f && fUv.x <= 1.0f) && (fUv.y >= 0.0f && fUv.y <= 1.0f); -} - -void ComputeReprojectedUVs(const AccumulationPassCommonParams params, FFX_PARAMETER_OUT FfxFloat32x2 fReprojectedHrUv, FFX_PARAMETER_OUT FfxBoolean bIsExistingSample) -{ - fReprojectedHrUv = params.fHrUv + params.fMotionVector; - - bIsExistingSample = IsUvInside(fReprojectedHrUv); -} - -void ReprojectHistoryColor(const AccumulationPassCommonParams params, FFX_PARAMETER_OUT FfxFloat32x3 fHistoryColor, FFX_PARAMETER_OUT FfxFloat32 fTemporalReactiveFactor, FFX_PARAMETER_OUT FfxBoolean bInMotionLastFrame) -{ - FfxFloat32x4 fHistory = HistorySample(params.fReprojectedHrUv, DisplaySize()); - - fHistoryColor = PrepareRgb(fHistory.rgb, Exposure(), PreviousFramePreExposure()); - - fHistoryColor = RGBToYCoCg(fHistoryColor); - - //Compute temporal reactivity info - fTemporalReactiveFactor = ffxSaturate(abs(fHistory.w)); - bInMotionLastFrame = (fHistory.w < 0.0f); -} - -LockState ReprojectHistoryLockStatus(const AccumulationPassCommonParams params, FFX_PARAMETER_OUT FfxFloat32x2 fReprojectedLockStatus) -{ - LockState state = { FFX_FALSE, FFX_FALSE }; - const FfxFloat32 fNewLockIntensity = LoadRwNewLocks(params.iPxHrPos); - state.NewLock = fNewLockIntensity > (127.0f / 255.0f); - - FfxFloat32 fInPlaceLockLifetime = state.NewLock ? fNewLockIntensity : 0; - - fReprojectedLockStatus = SampleLockStatus(params.fReprojectedHrUv); - - if (fReprojectedLockStatus[LOCK_LIFETIME_REMAINING] != FfxFloat32(0.0f)) { - state.WasLockedPrevFrame = true; - } - - return state; -} - -#endif //!defined( FFX_FSR2_REPROJECT_H ) diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_reproject.h.meta b/Assets/Resources/FSR2/shaders/ffx_fsr2_reproject.h.meta deleted file mode 100644 index ef8ce86..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_reproject.h.meta +++ /dev/null @@ -1,60 +0,0 @@ -fileFormatVersion: 2 -guid: 63fc917ca6895cb4aac237ea35edb838 -PluginImporter: - externalObjects: {} - serializedVersion: 2 - iconMap: {} - executionOrder: {} - defineConstraints: [] - isPreloaded: 0 - isOverridable: 0 - isExplicitlyReferenced: 0 - validateReferences: 1 - platformData: - - first: - : Any - second: - enabled: 0 - settings: - Exclude Editor: 1 - Exclude GameCoreScarlett: 1 - Exclude GameCoreXboxOne: 1 - Exclude Linux64: 1 - Exclude OSXUniversal: 1 - Exclude PS4: 1 - Exclude PS5: 1 - Exclude WebGL: 1 - Exclude Win: 1 - Exclude Win64: 1 - - first: - Any: - second: - enabled: 0 - settings: {} - - first: - Editor: Editor - second: - enabled: 0 - settings: - DefaultValueInitialized: true - - first: - Standalone: Linux64 - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win64 - second: - enabled: 0 - settings: - CPU: None - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_resources.h b/Assets/Resources/FSR2/shaders/ffx_fsr2_resources.h deleted file mode 100644 index 535dbc3..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_resources.h +++ /dev/null @@ -1,105 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#ifndef FFX_FSR2_RESOURCES_H -#define FFX_FSR2_RESOURCES_H - -#if defined(FFX_CPU) || defined(FFX_GPU) -#define FFX_FSR2_RESOURCE_IDENTIFIER_NULL 0 -#define FFX_FSR2_RESOURCE_IDENTIFIER_INPUT_OPAQUE_ONLY 1 -#define FFX_FSR2_RESOURCE_IDENTIFIER_INPUT_COLOR 2 -#define FFX_FSR2_RESOURCE_IDENTIFIER_INPUT_MOTION_VECTORS 3 -#define FFX_FSR2_RESOURCE_IDENTIFIER_INPUT_DEPTH 4 -#define FFX_FSR2_RESOURCE_IDENTIFIER_INPUT_EXPOSURE 5 -#define FFX_FSR2_RESOURCE_IDENTIFIER_INPUT_REACTIVE_MASK 6 -#define FFX_FSR2_RESOURCE_IDENTIFIER_INPUT_TRANSPARENCY_AND_COMPOSITION_MASK 7 -#define FFX_FSR2_RESOURCE_IDENTIFIER_RECONSTRUCTED_PREVIOUS_NEAREST_DEPTH 8 -#define FFX_FSR2_RESOURCE_IDENTIFIER_DILATED_MOTION_VECTORS 9 -#define FFX_FSR2_RESOURCE_IDENTIFIER_DILATED_DEPTH 10 -#define FFX_FSR2_RESOURCE_IDENTIFIER_INTERNAL_UPSCALED_COLOR 11 -#define FFX_FSR2_RESOURCE_IDENTIFIER_LOCK_STATUS 12 -#define FFX_FSR2_RESOURCE_IDENTIFIER_NEW_LOCKS 13 -#define FFX_FSR2_RESOURCE_IDENTIFIER_PREPARED_INPUT_COLOR 14 -#define FFX_FSR2_RESOURCE_IDENTIFIER_LUMA_HISTORY 15 -#define FFX_FSR2_RESOURCE_IDENTIFIER_DEBUG_OUTPUT 16 -#define FFX_FSR2_RESOURCE_IDENTIFIER_LANCZOS_LUT 17 -#define FFX_FSR2_RESOURCE_IDENTIFIER_SPD_ATOMIC_COUNT 18 -#define FFX_FSR2_RESOURCE_IDENTIFIER_UPSCALED_OUTPUT 19 -#define FFX_FSR2_RESOURCE_IDENTIFIER_RCAS_INPUT 20 -#define FFX_FSR2_RESOURCE_IDENTIFIER_LOCK_STATUS_1 21 -#define FFX_FSR2_RESOURCE_IDENTIFIER_LOCK_STATUS_2 22 -#define FFX_FSR2_RESOURCE_IDENTIFIER_INTERNAL_UPSCALED_COLOR_1 23 -#define FFX_FSR2_RESOURCE_IDENTIFIER_INTERNAL_UPSCALED_COLOR_2 24 -#define FFX_FSR2_RESOURCE_IDENTIFIER_INTERNAL_DEFAULT_REACTIVITY 25 -#define FFX_FSR2_RESOURCE_IDENTIFIER_INTERNAL_DEFAULT_TRANSPARENCY_AND_COMPOSITION 26 -#define FFX_FSR2_RESOURCE_IDENTITIER_UPSAMPLE_MAXIMUM_BIAS_LUT 27 -#define FFX_FSR2_RESOURCE_IDENTIFIER_DILATED_REACTIVE_MASKS 28 -#define FFX_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE 29 // same as FFX_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_0 -#define FFX_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_0 29 -#define FFX_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_1 30 -#define FFX_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_2 31 -#define FFX_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_3 32 -#define FFX_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_4 33 -#define FFX_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_5 34 -#define FFX_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_6 35 -#define FFX_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_7 36 -#define FFX_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_8 37 -#define FFX_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_9 38 -#define FFX_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_10 39 -#define FFX_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_11 40 -#define FFX_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_12 41 -#define FFX_FSR2_RESOURCE_IDENTIFIER_INTERNAL_DEFAULT_EXPOSURE 42 -#define FFX_FSR2_RESOURCE_IDENTIFIER_AUTO_EXPOSURE 43 -#define FFX_FSR2_RESOURCE_IDENTIFIER_AUTOREACTIVE 44 -#define FFX_FSR2_RESOURCE_IDENTIFIER_AUTOCOMPOSITION 45 - -#define FFX_FSR2_RESOURCE_IDENTIFIER_PREV_PRE_ALPHA_COLOR 46 -#define FFX_FSR2_RESOURCE_IDENTIFIER_PREV_POST_ALPHA_COLOR 47 -#define FFX_FSR2_RESOURCE_IDENTIFIER_PREV_PRE_ALPHA_COLOR_1 48 -#define FFX_FSR2_RESOURCE_IDENTIFIER_PREV_POST_ALPHA_COLOR_1 49 -#define FFX_FSR2_RESOURCE_IDENTIFIER_PREV_PRE_ALPHA_COLOR_2 50 -#define FFX_FSR2_RESOURCE_IDENTIFIER_PREV_POST_ALPHA_COLOR_2 51 -#define FFX_FSR2_RESOURCE_IDENTIFIER_PREVIOUS_DILATED_MOTION_VECTORS 52 -#define FFX_FSR2_RESOURCE_IDENTIFIER_INTERNAL_DILATED_MOTION_VECTORS_1 53 -#define FFX_FSR2_RESOURCE_IDENTIFIER_INTERNAL_DILATED_MOTION_VECTORS_2 54 -#define FFX_FSR2_RESOURCE_IDENTIFIER_LUMA_HISTORY_1 55 -#define FFX_FSR2_RESOURCE_IDENTIFIER_LUMA_HISTORY_2 56 -#define FFX_FSR2_RESOURCE_IDENTIFIER_LOCK_INPUT_LUMA 57 - -// Shading change detection mip level setting, value must be in the range [FFX_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_0, FFX_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_12] -#define FFX_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_SHADING_CHANGE FFX_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_4 -#define FFX_FSR2_SHADING_CHANGE_MIP_LEVEL (FFX_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_SHADING_CHANGE - FFX_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE) - -#define FFX_FSR2_RESOURCE_IDENTIFIER_COUNT 58 - -#define FFX_FSR2_CONSTANTBUFFER_IDENTIFIER_FSR2 0 -#define FFX_FSR2_CONSTANTBUFFER_IDENTIFIER_SPD 1 -#define FFX_FSR2_CONSTANTBUFFER_IDENTIFIER_RCAS 2 -#define FFX_FSR2_CONSTANTBUFFER_IDENTIFIER_GENREACTIVE 3 - -#define FFX_FSR2_AUTOREACTIVEFLAGS_APPLY_TONEMAP 1 -#define FFX_FSR2_AUTOREACTIVEFLAGS_APPLY_INVERSETONEMAP 2 -#define FFX_FSR2_AUTOREACTIVEFLAGS_APPLY_THRESHOLD 4 -#define FFX_FSR2_AUTOREACTIVEFLAGS_USE_COMPONENTS_MAX 8 - -#endif // #if defined(FFX_CPU) || defined(FFX_GPU) - -#endif //!defined( FFX_FSR2_RESOURCES_H ) diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_resources.h.meta b/Assets/Resources/FSR2/shaders/ffx_fsr2_resources.h.meta deleted file mode 100644 index 69e95a4..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_resources.h.meta +++ /dev/null @@ -1,60 +0,0 @@ -fileFormatVersion: 2 -guid: fd5bed2bf4ba07444ae815390168a15d -PluginImporter: - externalObjects: {} - serializedVersion: 2 - iconMap: {} - executionOrder: {} - defineConstraints: [] - isPreloaded: 0 - isOverridable: 0 - isExplicitlyReferenced: 0 - validateReferences: 1 - platformData: - - first: - : Any - second: - enabled: 0 - settings: - Exclude Editor: 1 - Exclude GameCoreScarlett: 1 - Exclude GameCoreXboxOne: 1 - Exclude Linux64: 1 - Exclude OSXUniversal: 1 - Exclude PS4: 1 - Exclude PS5: 1 - Exclude WebGL: 1 - Exclude Win: 1 - Exclude Win64: 1 - - first: - Any: - second: - enabled: 0 - settings: {} - - first: - Editor: Editor - second: - enabled: 0 - settings: - DefaultValueInitialized: true - - first: - Standalone: Linux64 - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win64 - second: - enabled: 0 - settings: - CPU: None - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_sample.h b/Assets/Resources/FSR2/shaders/ffx_fsr2_sample.h deleted file mode 100644 index f94f40a..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_sample.h +++ /dev/null @@ -1,605 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#ifndef FFX_FSR2_SAMPLE_H -#define FFX_FSR2_SAMPLE_H - -// suppress warnings -#ifdef FFX_HLSL -#pragma warning(disable: 4008) // potentially divide by zero -#endif //FFX_HLSL - -struct FetchedBilinearSamples { - - FfxFloat32x4 fColor00; - FfxFloat32x4 fColor10; - - FfxFloat32x4 fColor01; - FfxFloat32x4 fColor11; -}; - -struct FetchedBicubicSamples { - - FfxFloat32x4 fColor00; - FfxFloat32x4 fColor10; - FfxFloat32x4 fColor20; - FfxFloat32x4 fColor30; - - FfxFloat32x4 fColor01; - FfxFloat32x4 fColor11; - FfxFloat32x4 fColor21; - FfxFloat32x4 fColor31; - - FfxFloat32x4 fColor02; - FfxFloat32x4 fColor12; - FfxFloat32x4 fColor22; - FfxFloat32x4 fColor32; - - FfxFloat32x4 fColor03; - FfxFloat32x4 fColor13; - FfxFloat32x4 fColor23; - FfxFloat32x4 fColor33; -}; - -#if FFX_HALF -struct FetchedBilinearSamplesMin16 { - - FFX_MIN16_F4 fColor00; - FFX_MIN16_F4 fColor10; - - FFX_MIN16_F4 fColor01; - FFX_MIN16_F4 fColor11; -}; - -struct FetchedBicubicSamplesMin16 { - - FFX_MIN16_F4 fColor00; - FFX_MIN16_F4 fColor10; - FFX_MIN16_F4 fColor20; - FFX_MIN16_F4 fColor30; - - FFX_MIN16_F4 fColor01; - FFX_MIN16_F4 fColor11; - FFX_MIN16_F4 fColor21; - FFX_MIN16_F4 fColor31; - - FFX_MIN16_F4 fColor02; - FFX_MIN16_F4 fColor12; - FFX_MIN16_F4 fColor22; - FFX_MIN16_F4 fColor32; - - FFX_MIN16_F4 fColor03; - FFX_MIN16_F4 fColor13; - FFX_MIN16_F4 fColor23; - FFX_MIN16_F4 fColor33; -}; -#else //FFX_HALF -#define FetchedBicubicSamplesMin16 FetchedBicubicSamples -#define FetchedBilinearSamplesMin16 FetchedBilinearSamples -#endif //FFX_HALF - -FfxFloat32x4 Linear(FfxFloat32x4 A, FfxFloat32x4 B, FfxFloat32 t) -{ - return A + (B - A) * t; -} - -FfxFloat32x4 Bilinear(FetchedBilinearSamples BilinearSamples, FfxFloat32x2 fPxFrac) -{ - FfxFloat32x4 fColorX0 = Linear(BilinearSamples.fColor00, BilinearSamples.fColor10, fPxFrac.x); - FfxFloat32x4 fColorX1 = Linear(BilinearSamples.fColor01, BilinearSamples.fColor11, fPxFrac.x); - FfxFloat32x4 fColorXY = Linear(fColorX0, fColorX1, fPxFrac.y); - return fColorXY; -} - -#if FFX_HALF -FFX_MIN16_F4 Linear(FFX_MIN16_F4 A, FFX_MIN16_F4 B, FFX_MIN16_F t) -{ - return A + (B - A) * t; -} - -FFX_MIN16_F4 Bilinear(FetchedBilinearSamplesMin16 BilinearSamples, FFX_MIN16_F2 fPxFrac) -{ - FFX_MIN16_F4 fColorX0 = Linear(BilinearSamples.fColor00, BilinearSamples.fColor10, fPxFrac.x); - FFX_MIN16_F4 fColorX1 = Linear(BilinearSamples.fColor01, BilinearSamples.fColor11, fPxFrac.x); - FFX_MIN16_F4 fColorXY = Linear(fColorX0, fColorX1, fPxFrac.y); - return fColorXY; -} -#endif - -FfxFloat32 Lanczos2NoClamp(FfxFloat32 x) -{ - const FfxFloat32 PI = 3.141592653589793f; // TODO: share SDK constants - return abs(x) < FSR2_EPSILON ? 1.f : (sin(PI * x) / (PI * x)) * (sin(0.5f * PI * x) / (0.5f * PI * x)); -} - -FfxFloat32 Lanczos2(FfxFloat32 x) -{ - x = ffxMin(abs(x), 2.0f); - return Lanczos2NoClamp(x); -} - -#if FFX_HALF - -#if 0 -FFX_MIN16_F Lanczos2NoClamp(FFX_MIN16_F x) -{ - const FFX_MIN16_F PI = FFX_MIN16_F(3.141592653589793f); // TODO: share SDK constants - return abs(x) < FFX_MIN16_F(FSR2_EPSILON) ? FFX_MIN16_F(1.f) : (sin(PI * x) / (PI * x)) * (sin(FFX_MIN16_F(0.5f) * PI * x) / (FFX_MIN16_F(0.5f) * PI * x)); -} -#endif - -FFX_MIN16_F Lanczos2(FFX_MIN16_F x) -{ - x = ffxMin(abs(x), FFX_MIN16_F(2.0f)); - return FFX_MIN16_F(Lanczos2NoClamp(x)); -} -#endif //FFX_HALF - -// FSR1 lanczos approximation. Input is x*x and must be <= 4. -FfxFloat32 Lanczos2ApproxSqNoClamp(FfxFloat32 x2) -{ - FfxFloat32 a = (2.0f / 5.0f) * x2 - 1; - FfxFloat32 b = (1.0f / 4.0f) * x2 - 1; - return ((25.0f / 16.0f) * a * a - (25.0f / 16.0f - 1)) * (b * b); -} - -#if FFX_HALF -FFX_MIN16_F Lanczos2ApproxSqNoClamp(FFX_MIN16_F x2) -{ - FFX_MIN16_F a = FFX_MIN16_F(2.0f / 5.0f) * x2 - FFX_MIN16_F(1); - FFX_MIN16_F b = FFX_MIN16_F(1.0f / 4.0f) * x2 - FFX_MIN16_F(1); - return (FFX_MIN16_F(25.0f / 16.0f) * a * a - FFX_MIN16_F(25.0f / 16.0f - 1)) * (b * b); -} -#endif //FFX_HALF - -FfxFloat32 Lanczos2ApproxSq(FfxFloat32 x2) -{ - x2 = ffxMin(x2, 4.0f); - return Lanczos2ApproxSqNoClamp(x2); -} - -#if FFX_HALF -FFX_MIN16_F Lanczos2ApproxSq(FFX_MIN16_F x2) -{ - x2 = ffxMin(x2, FFX_MIN16_F(4.0f)); - return Lanczos2ApproxSqNoClamp(x2); -} -#endif //FFX_HALF - -FfxFloat32 Lanczos2ApproxNoClamp(FfxFloat32 x) -{ - return Lanczos2ApproxSqNoClamp(x * x); -} - -#if FFX_HALF -FFX_MIN16_F Lanczos2ApproxNoClamp(FFX_MIN16_F x) -{ - return Lanczos2ApproxSqNoClamp(x * x); -} -#endif //FFX_HALF - -FfxFloat32 Lanczos2Approx(FfxFloat32 x) -{ - return Lanczos2ApproxSq(x * x); -} - -#if FFX_HALF -FFX_MIN16_F Lanczos2Approx(FFX_MIN16_F x) -{ - return Lanczos2ApproxSq(x * x); -} -#endif //FFX_HALF - -FfxFloat32 Lanczos2_UseLUT(FfxFloat32 x) -{ - return SampleLanczos2Weight(abs(x)); -} - -#if FFX_HALF -FFX_MIN16_F Lanczos2_UseLUT(FFX_MIN16_F x) -{ - return FFX_MIN16_F(SampleLanczos2Weight(abs(x))); -} -#endif //FFX_HALF - -FfxFloat32x4 Lanczos2_UseLUT(FfxFloat32x4 fColor0, FfxFloat32x4 fColor1, FfxFloat32x4 fColor2, FfxFloat32x4 fColor3, FfxFloat32 t) -{ - FfxFloat32 fWeight0 = Lanczos2_UseLUT(-1.f - t); - FfxFloat32 fWeight1 = Lanczos2_UseLUT(-0.f - t); - FfxFloat32 fWeight2 = Lanczos2_UseLUT(+1.f - t); - FfxFloat32 fWeight3 = Lanczos2_UseLUT(+2.f - t); - return (fWeight0 * fColor0 + fWeight1 * fColor1 + fWeight2 * fColor2 + fWeight3 * fColor3) / (fWeight0 + fWeight1 + fWeight2 + fWeight3); -} -#if FFX_HALF -FFX_MIN16_F4 Lanczos2_UseLUT(FFX_MIN16_F4 fColor0, FFX_MIN16_F4 fColor1, FFX_MIN16_F4 fColor2, FFX_MIN16_F4 fColor3, FFX_MIN16_F t) -{ - FFX_MIN16_F fWeight0 = Lanczos2_UseLUT(FFX_MIN16_F(-1.f) - t); - FFX_MIN16_F fWeight1 = Lanczos2_UseLUT(FFX_MIN16_F(-0.f) - t); - FFX_MIN16_F fWeight2 = Lanczos2_UseLUT(FFX_MIN16_F(+1.f) - t); - FFX_MIN16_F fWeight3 = Lanczos2_UseLUT(FFX_MIN16_F(+2.f) - t); - return (fWeight0 * fColor0 + fWeight1 * fColor1 + fWeight2 * fColor2 + fWeight3 * fColor3) / (fWeight0 + fWeight1 + fWeight2 + fWeight3); -} -#endif - -FfxFloat32x4 Lanczos2(FfxFloat32x4 fColor0, FfxFloat32x4 fColor1, FfxFloat32x4 fColor2, FfxFloat32x4 fColor3, FfxFloat32 t) -{ - FfxFloat32 fWeight0 = Lanczos2(-1.f - t); - FfxFloat32 fWeight1 = Lanczos2(-0.f - t); - FfxFloat32 fWeight2 = Lanczos2(+1.f - t); - FfxFloat32 fWeight3 = Lanczos2(+2.f - t); - return (fWeight0 * fColor0 + fWeight1 * fColor1 + fWeight2 * fColor2 + fWeight3 * fColor3) / (fWeight0 + fWeight1 + fWeight2 + fWeight3); -} - -FfxFloat32x4 Lanczos2(FetchedBicubicSamples Samples, FfxFloat32x2 fPxFrac) -{ - FfxFloat32x4 fColorX0 = Lanczos2(Samples.fColor00, Samples.fColor10, Samples.fColor20, Samples.fColor30, fPxFrac.x); - FfxFloat32x4 fColorX1 = Lanczos2(Samples.fColor01, Samples.fColor11, Samples.fColor21, Samples.fColor31, fPxFrac.x); - FfxFloat32x4 fColorX2 = Lanczos2(Samples.fColor02, Samples.fColor12, Samples.fColor22, Samples.fColor32, fPxFrac.x); - FfxFloat32x4 fColorX3 = Lanczos2(Samples.fColor03, Samples.fColor13, Samples.fColor23, Samples.fColor33, fPxFrac.x); - FfxFloat32x4 fColorXY = Lanczos2(fColorX0, fColorX1, fColorX2, fColorX3, fPxFrac.y); - - // Deringing - - // TODO: only use 4 by checking jitter - const FfxInt32 iDeringingSampleCount = 4; - const FfxFloat32x4 fDeringingSamples[4] = { - Samples.fColor11, - Samples.fColor21, - Samples.fColor12, - Samples.fColor22, - }; - - FfxFloat32x4 fDeringingMin = fDeringingSamples[0]; - FfxFloat32x4 fDeringingMax = fDeringingSamples[0]; - - FFX_UNROLL - for (FfxInt32 iSampleIndex = 1; iSampleIndex < iDeringingSampleCount; ++iSampleIndex) { - - fDeringingMin = ffxMin(fDeringingMin, fDeringingSamples[iSampleIndex]); - fDeringingMax = ffxMax(fDeringingMax, fDeringingSamples[iSampleIndex]); - } - - fColorXY = clamp(fColorXY, fDeringingMin, fDeringingMax); - - return fColorXY; -} - -#if FFX_HALF -FFX_MIN16_F4 Lanczos2(FFX_MIN16_F4 fColor0, FFX_MIN16_F4 fColor1, FFX_MIN16_F4 fColor2, FFX_MIN16_F4 fColor3, FFX_MIN16_F t) -{ - FFX_MIN16_F fWeight0 = Lanczos2(FFX_MIN16_F(-1.f) - t); - FFX_MIN16_F fWeight1 = Lanczos2(FFX_MIN16_F(-0.f) - t); - FFX_MIN16_F fWeight2 = Lanczos2(FFX_MIN16_F(+1.f) - t); - FFX_MIN16_F fWeight3 = Lanczos2(FFX_MIN16_F(+2.f) - t); - return (fWeight0 * fColor0 + fWeight1 * fColor1 + fWeight2 * fColor2 + fWeight3 * fColor3) / (fWeight0 + fWeight1 + fWeight2 + fWeight3); -} - -FFX_MIN16_F4 Lanczos2(FetchedBicubicSamplesMin16 Samples, FFX_MIN16_F2 fPxFrac) -{ - FFX_MIN16_F4 fColorX0 = Lanczos2(Samples.fColor00, Samples.fColor10, Samples.fColor20, Samples.fColor30, fPxFrac.x); - FFX_MIN16_F4 fColorX1 = Lanczos2(Samples.fColor01, Samples.fColor11, Samples.fColor21, Samples.fColor31, fPxFrac.x); - FFX_MIN16_F4 fColorX2 = Lanczos2(Samples.fColor02, Samples.fColor12, Samples.fColor22, Samples.fColor32, fPxFrac.x); - FFX_MIN16_F4 fColorX3 = Lanczos2(Samples.fColor03, Samples.fColor13, Samples.fColor23, Samples.fColor33, fPxFrac.x); - FFX_MIN16_F4 fColorXY = Lanczos2(fColorX0, fColorX1, fColorX2, fColorX3, fPxFrac.y); - - // Deringing - - // TODO: only use 4 by checking jitter - const FfxInt32 iDeringingSampleCount = 4; - const FFX_MIN16_F4 fDeringingSamples[4] = { - Samples.fColor11, - Samples.fColor21, - Samples.fColor12, - Samples.fColor22, - }; - - FFX_MIN16_F4 fDeringingMin = fDeringingSamples[0]; - FFX_MIN16_F4 fDeringingMax = fDeringingSamples[0]; - - FFX_UNROLL - for (FfxInt32 iSampleIndex = 1; iSampleIndex < iDeringingSampleCount; ++iSampleIndex) - { - fDeringingMin = ffxMin(fDeringingMin, fDeringingSamples[iSampleIndex]); - fDeringingMax = ffxMax(fDeringingMax, fDeringingSamples[iSampleIndex]); - } - - fColorXY = clamp(fColorXY, fDeringingMin, fDeringingMax); - - return fColorXY; -} -#endif //FFX_HALF - - -FfxFloat32x4 Lanczos2LUT(FetchedBicubicSamples Samples, FfxFloat32x2 fPxFrac) -{ - FfxFloat32x4 fColorX0 = Lanczos2_UseLUT(Samples.fColor00, Samples.fColor10, Samples.fColor20, Samples.fColor30, fPxFrac.x); - FfxFloat32x4 fColorX1 = Lanczos2_UseLUT(Samples.fColor01, Samples.fColor11, Samples.fColor21, Samples.fColor31, fPxFrac.x); - FfxFloat32x4 fColorX2 = Lanczos2_UseLUT(Samples.fColor02, Samples.fColor12, Samples.fColor22, Samples.fColor32, fPxFrac.x); - FfxFloat32x4 fColorX3 = Lanczos2_UseLUT(Samples.fColor03, Samples.fColor13, Samples.fColor23, Samples.fColor33, fPxFrac.x); - FfxFloat32x4 fColorXY = Lanczos2_UseLUT(fColorX0, fColorX1, fColorX2, fColorX3, fPxFrac.y); - - // Deringing - - // TODO: only use 4 by checking jitter - const FfxInt32 iDeringingSampleCount = 4; - const FfxFloat32x4 fDeringingSamples[4] = { - Samples.fColor11, - Samples.fColor21, - Samples.fColor12, - Samples.fColor22, - }; - - FfxFloat32x4 fDeringingMin = fDeringingSamples[0]; - FfxFloat32x4 fDeringingMax = fDeringingSamples[0]; - - FFX_UNROLL - for (FfxInt32 iSampleIndex = 1; iSampleIndex < iDeringingSampleCount; ++iSampleIndex) { - - fDeringingMin = ffxMin(fDeringingMin, fDeringingSamples[iSampleIndex]); - fDeringingMax = ffxMax(fDeringingMax, fDeringingSamples[iSampleIndex]); - } - - fColorXY = clamp(fColorXY, fDeringingMin, fDeringingMax); - - return fColorXY; -} - -#if FFX_HALF -FFX_MIN16_F4 Lanczos2LUT(FetchedBicubicSamplesMin16 Samples, FFX_MIN16_F2 fPxFrac) -{ - FFX_MIN16_F4 fColorX0 = Lanczos2_UseLUT(Samples.fColor00, Samples.fColor10, Samples.fColor20, Samples.fColor30, fPxFrac.x); - FFX_MIN16_F4 fColorX1 = Lanczos2_UseLUT(Samples.fColor01, Samples.fColor11, Samples.fColor21, Samples.fColor31, fPxFrac.x); - FFX_MIN16_F4 fColorX2 = Lanczos2_UseLUT(Samples.fColor02, Samples.fColor12, Samples.fColor22, Samples.fColor32, fPxFrac.x); - FFX_MIN16_F4 fColorX3 = Lanczos2_UseLUT(Samples.fColor03, Samples.fColor13, Samples.fColor23, Samples.fColor33, fPxFrac.x); - FFX_MIN16_F4 fColorXY = Lanczos2_UseLUT(fColorX0, fColorX1, fColorX2, fColorX3, fPxFrac.y); - - // Deringing - - // TODO: only use 4 by checking jitter - const FfxInt32 iDeringingSampleCount = 4; - const FFX_MIN16_F4 fDeringingSamples[4] = { - Samples.fColor11, - Samples.fColor21, - Samples.fColor12, - Samples.fColor22, - }; - - FFX_MIN16_F4 fDeringingMin = fDeringingSamples[0]; - FFX_MIN16_F4 fDeringingMax = fDeringingSamples[0]; - - FFX_UNROLL - for (FfxInt32 iSampleIndex = 1; iSampleIndex < iDeringingSampleCount; ++iSampleIndex) - { - fDeringingMin = ffxMin(fDeringingMin, fDeringingSamples[iSampleIndex]); - fDeringingMax = ffxMax(fDeringingMax, fDeringingSamples[iSampleIndex]); - } - - fColorXY = clamp(fColorXY, fDeringingMin, fDeringingMax); - - return fColorXY; -} -#endif //FFX_HALF - - - -FfxFloat32x4 Lanczos2Approx(FfxFloat32x4 fColor0, FfxFloat32x4 fColor1, FfxFloat32x4 fColor2, FfxFloat32x4 fColor3, FfxFloat32 t) -{ - FfxFloat32 fWeight0 = Lanczos2ApproxNoClamp(-1.f - t); - FfxFloat32 fWeight1 = Lanczos2ApproxNoClamp(-0.f - t); - FfxFloat32 fWeight2 = Lanczos2ApproxNoClamp(+1.f - t); - FfxFloat32 fWeight3 = Lanczos2ApproxNoClamp(+2.f - t); - return (fWeight0 * fColor0 + fWeight1 * fColor1 + fWeight2 * fColor2 + fWeight3 * fColor3) / (fWeight0 + fWeight1 + fWeight2 + fWeight3); -} - -#if FFX_HALF -FFX_MIN16_F4 Lanczos2Approx(FFX_MIN16_F4 fColor0, FFX_MIN16_F4 fColor1, FFX_MIN16_F4 fColor2, FFX_MIN16_F4 fColor3, FFX_MIN16_F t) -{ - FFX_MIN16_F fWeight0 = Lanczos2ApproxNoClamp(FFX_MIN16_F(-1.f) - t); - FFX_MIN16_F fWeight1 = Lanczos2ApproxNoClamp(FFX_MIN16_F(-0.f) - t); - FFX_MIN16_F fWeight2 = Lanczos2ApproxNoClamp(FFX_MIN16_F(+1.f) - t); - FFX_MIN16_F fWeight3 = Lanczos2ApproxNoClamp(FFX_MIN16_F(+2.f) - t); - return (fWeight0 * fColor0 + fWeight1 * fColor1 + fWeight2 * fColor2 + fWeight3 * fColor3) / (fWeight0 + fWeight1 + fWeight2 + fWeight3); -} -#endif //FFX_HALF - -FfxFloat32x4 Lanczos2Approx(FetchedBicubicSamples Samples, FfxFloat32x2 fPxFrac) -{ - FfxFloat32x4 fColorX0 = Lanczos2Approx(Samples.fColor00, Samples.fColor10, Samples.fColor20, Samples.fColor30, fPxFrac.x); - FfxFloat32x4 fColorX1 = Lanczos2Approx(Samples.fColor01, Samples.fColor11, Samples.fColor21, Samples.fColor31, fPxFrac.x); - FfxFloat32x4 fColorX2 = Lanczos2Approx(Samples.fColor02, Samples.fColor12, Samples.fColor22, Samples.fColor32, fPxFrac.x); - FfxFloat32x4 fColorX3 = Lanczos2Approx(Samples.fColor03, Samples.fColor13, Samples.fColor23, Samples.fColor33, fPxFrac.x); - FfxFloat32x4 fColorXY = Lanczos2Approx(fColorX0, fColorX1, fColorX2, fColorX3, fPxFrac.y); - - // Deringing - - // TODO: only use 4 by checking jitter - const FfxInt32 iDeringingSampleCount = 4; - const FfxFloat32x4 fDeringingSamples[4] = { - Samples.fColor11, - Samples.fColor21, - Samples.fColor12, - Samples.fColor22, - }; - - FfxFloat32x4 fDeringingMin = fDeringingSamples[0]; - FfxFloat32x4 fDeringingMax = fDeringingSamples[0]; - - FFX_UNROLL - for (FfxInt32 iSampleIndex = 1; iSampleIndex < iDeringingSampleCount; ++iSampleIndex) - { - fDeringingMin = ffxMin(fDeringingMin, fDeringingSamples[iSampleIndex]); - fDeringingMax = ffxMax(fDeringingMax, fDeringingSamples[iSampleIndex]); - } - - fColorXY = clamp(fColorXY, fDeringingMin, fDeringingMax); - - return fColorXY; -} - -#if FFX_HALF -FFX_MIN16_F4 Lanczos2Approx(FetchedBicubicSamplesMin16 Samples, FFX_MIN16_F2 fPxFrac) -{ - FFX_MIN16_F4 fColorX0 = Lanczos2Approx(Samples.fColor00, Samples.fColor10, Samples.fColor20, Samples.fColor30, fPxFrac.x); - FFX_MIN16_F4 fColorX1 = Lanczos2Approx(Samples.fColor01, Samples.fColor11, Samples.fColor21, Samples.fColor31, fPxFrac.x); - FFX_MIN16_F4 fColorX2 = Lanczos2Approx(Samples.fColor02, Samples.fColor12, Samples.fColor22, Samples.fColor32, fPxFrac.x); - FFX_MIN16_F4 fColorX3 = Lanczos2Approx(Samples.fColor03, Samples.fColor13, Samples.fColor23, Samples.fColor33, fPxFrac.x); - FFX_MIN16_F4 fColorXY = Lanczos2Approx(fColorX0, fColorX1, fColorX2, fColorX3, fPxFrac.y); - - // Deringing - - // TODO: only use 4 by checking jitter - const FfxInt32 iDeringingSampleCount = 4; - const FFX_MIN16_F4 fDeringingSamples[4] = { - Samples.fColor11, - Samples.fColor21, - Samples.fColor12, - Samples.fColor22, - }; - - FFX_MIN16_F4 fDeringingMin = fDeringingSamples[0]; - FFX_MIN16_F4 fDeringingMax = fDeringingSamples[0]; - - FFX_UNROLL - for (FfxInt32 iSampleIndex = 1; iSampleIndex < iDeringingSampleCount; ++iSampleIndex) - { - fDeringingMin = ffxMin(fDeringingMin, fDeringingSamples[iSampleIndex]); - fDeringingMax = ffxMax(fDeringingMax, fDeringingSamples[iSampleIndex]); - } - - fColorXY = clamp(fColorXY, fDeringingMin, fDeringingMax); - - return fColorXY; -} -#endif - -// Clamp by offset direction. Assuming iPxSample is already in range and iPxOffset is compile time constant. -FfxInt32x2 ClampCoord(FfxInt32x2 iPxSample, FfxInt32x2 iPxOffset, FfxInt32x2 iTextureSize) -{ - FfxInt32x2 result = iPxSample + iPxOffset; - result.x = (iPxOffset.x < 0) ? ffxMax(result.x, 0) : result.x; - result.x = (iPxOffset.x > 0) ? ffxMin(result.x, iTextureSize.x - 1) : result.x; - result.y = (iPxOffset.y < 0) ? ffxMax(result.y, 0) : result.y; - result.y = (iPxOffset.y > 0) ? ffxMin(result.y, iTextureSize.y - 1) : result.y; - return result; -} -#if FFX_HALF -FFX_MIN16_I2 ClampCoord(FFX_MIN16_I2 iPxSample, FFX_MIN16_I2 iPxOffset, FFX_MIN16_I2 iTextureSize) -{ - FFX_MIN16_I2 result = iPxSample + iPxOffset; - result.x = (iPxOffset.x < FFX_MIN16_I(0)) ? ffxMax(result.x, FFX_MIN16_I(0)) : result.x; - result.x = (iPxOffset.x > FFX_MIN16_I(0)) ? ffxMin(result.x, iTextureSize.x - FFX_MIN16_I(1)) : result.x; - result.y = (iPxOffset.y < FFX_MIN16_I(0)) ? ffxMax(result.y, FFX_MIN16_I(0)) : result.y; - result.y = (iPxOffset.y > FFX_MIN16_I(0)) ? ffxMin(result.y, iTextureSize.y - FFX_MIN16_I(1)) : result.y; - return result; -} -#endif //FFX_HALF - - -#define DeclareCustomFetchBicubicSamplesWithType(SampleType, TextureType, AddrType, Name, LoadTexture) \ - SampleType Name(AddrType iPxSample, AddrType iTextureSize) \ - { \ - SampleType Samples; \ - \ - Samples.fColor00 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(-1, -1), iTextureSize))); \ - Samples.fColor10 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+0, -1), iTextureSize))); \ - Samples.fColor20 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+1, -1), iTextureSize))); \ - Samples.fColor30 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+2, -1), iTextureSize))); \ - \ - Samples.fColor01 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(-1, +0), iTextureSize))); \ - Samples.fColor11 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+0, +0), iTextureSize))); \ - Samples.fColor21 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+1, +0), iTextureSize))); \ - Samples.fColor31 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+2, +0), iTextureSize))); \ - \ - Samples.fColor02 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(-1, +1), iTextureSize))); \ - Samples.fColor12 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+0, +1), iTextureSize))); \ - Samples.fColor22 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+1, +1), iTextureSize))); \ - Samples.fColor32 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+2, +1), iTextureSize))); \ - \ - Samples.fColor03 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(-1, +2), iTextureSize))); \ - Samples.fColor13 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+0, +2), iTextureSize))); \ - Samples.fColor23 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+1, +2), iTextureSize))); \ - Samples.fColor33 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+2, +2), iTextureSize))); \ - \ - return Samples; \ - } - -#define DeclareCustomFetchBicubicSamples(Name, LoadTexture) \ - DeclareCustomFetchBicubicSamplesWithType(FetchedBicubicSamples, FfxFloat32x4, FfxInt32x2, Name, LoadTexture) - -#define DeclareCustomFetchBicubicSamplesMin16(Name, LoadTexture) \ - DeclareCustomFetchBicubicSamplesWithType(FetchedBicubicSamplesMin16, FFX_MIN16_F4, FfxInt32x2, Name, LoadTexture) - -#define DeclareCustomFetchBilinearSamplesWithType(SampleType, TextureType,AddrType, Name, LoadTexture) \ - SampleType Name(AddrType iPxSample, AddrType iTextureSize) \ - { \ - SampleType Samples; \ - Samples.fColor00 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+0, +0), iTextureSize))); \ - Samples.fColor10 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+1, +0), iTextureSize))); \ - Samples.fColor01 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+0, +1), iTextureSize))); \ - Samples.fColor11 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+1, +1), iTextureSize))); \ - return Samples; \ - } - -#define DeclareCustomFetchBilinearSamples(Name, LoadTexture) \ - DeclareCustomFetchBilinearSamplesWithType(FetchedBilinearSamples, FfxFloat32x4, FfxInt32x2, Name, LoadTexture) - -#define DeclareCustomFetchBilinearSamplesMin16(Name, LoadTexture) \ - DeclareCustomFetchBilinearSamplesWithType(FetchedBilinearSamplesMin16, FFX_MIN16_F4, FfxInt32x2, Name, LoadTexture) - -// BE CAREFUL: there is some precision issues and (3253, 125) leading to (3252.9989778, 125.001102) -// is common, so iPxSample can "jitter" -#define DeclareCustomTextureSample(Name, InterpolateSamples, FetchSamples) \ - FfxFloat32x4 Name(FfxFloat32x2 fUvSample, FfxInt32x2 iTextureSize) \ - { \ - FfxFloat32x2 fPxSample = (fUvSample * FfxFloat32x2(iTextureSize)) - FfxFloat32x2(0.5f, 0.5f); \ - /* Clamp base coords */ \ - fPxSample.x = ffxMax(0.0f, ffxMin(FfxFloat32(iTextureSize.x), fPxSample.x)); \ - fPxSample.y = ffxMax(0.0f, ffxMin(FfxFloat32(iTextureSize.y), fPxSample.y)); \ - /* */ \ - FfxInt32x2 iPxSample = FfxInt32x2(floor(fPxSample)); \ - FfxFloat32x2 fPxFrac = ffxFract(fPxSample); \ - FfxFloat32x4 fColorXY = FfxFloat32x4(InterpolateSamples(FetchSamples(iPxSample, iTextureSize), fPxFrac)); \ - return fColorXY; \ - } - -#define DeclareCustomTextureSampleMin16(Name, InterpolateSamples, FetchSamples) \ - FFX_MIN16_F4 Name(FfxFloat32x2 fUvSample, FfxInt32x2 iTextureSize) \ - { \ - FfxFloat32x2 fPxSample = (fUvSample * FfxFloat32x2(iTextureSize)) - FfxFloat32x2(0.5f, 0.5f); \ - /* Clamp base coords */ \ - fPxSample.x = ffxMax(0.0f, ffxMin(FfxFloat32(iTextureSize.x), fPxSample.x)); \ - fPxSample.y = ffxMax(0.0f, ffxMin(FfxFloat32(iTextureSize.y), fPxSample.y)); \ - /* */ \ - FfxInt32x2 iPxSample = FfxInt32x2(floor(fPxSample)); \ - FFX_MIN16_F2 fPxFrac = FFX_MIN16_F2(ffxFract(fPxSample)); \ - FFX_MIN16_F4 fColorXY = FFX_MIN16_F4(InterpolateSamples(FetchSamples(iPxSample, iTextureSize), fPxFrac)); \ - return fColorXY; \ - } - -#define FFX_FSR2_CONCAT_ID(x, y) x ## y -#define FFX_FSR2_CONCAT(x, y) FFX_FSR2_CONCAT_ID(x, y) -#define FFX_FSR2_SAMPLER_1D_0 Lanczos2 -#define FFX_FSR2_SAMPLER_1D_1 Lanczos2LUT -#define FFX_FSR2_SAMPLER_1D_2 Lanczos2Approx - -#define FFX_FSR2_GET_LANCZOS_SAMPLER1D(x) FFX_FSR2_CONCAT(FFX_FSR2_SAMPLER_1D_, x) - -#endif //!defined( FFX_FSR2_SAMPLE_H ) diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_sample.h.meta b/Assets/Resources/FSR2/shaders/ffx_fsr2_sample.h.meta deleted file mode 100644 index acc8bab..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_sample.h.meta +++ /dev/null @@ -1,60 +0,0 @@ -fileFormatVersion: 2 -guid: ba3ee190167b6c240aaeb1f8f4dbcb67 -PluginImporter: - externalObjects: {} - serializedVersion: 2 - iconMap: {} - executionOrder: {} - defineConstraints: [] - isPreloaded: 0 - isOverridable: 0 - isExplicitlyReferenced: 0 - validateReferences: 1 - platformData: - - first: - : Any - second: - enabled: 0 - settings: - Exclude Editor: 1 - Exclude GameCoreScarlett: 1 - Exclude GameCoreXboxOne: 1 - Exclude Linux64: 1 - Exclude OSXUniversal: 1 - Exclude PS4: 1 - Exclude PS5: 1 - Exclude WebGL: 1 - Exclude Win: 1 - Exclude Win64: 1 - - first: - Any: - second: - enabled: 0 - settings: {} - - first: - Editor: Editor - second: - enabled: 0 - settings: - DefaultValueInitialized: true - - first: - Standalone: Linux64 - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win64 - second: - enabled: 0 - settings: - CPU: None - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_tcr_autogen.h b/Assets/Resources/FSR2/shaders/ffx_fsr2_tcr_autogen.h deleted file mode 100644 index 101b75d..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_tcr_autogen.h +++ /dev/null @@ -1,250 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#define USE_YCOCG 1 - -#define fAutogenEpsilon 0.01f - -// EXPERIMENTAL - -FFX_MIN16_F ComputeAutoTC_01(FFX_MIN16_I2 uDispatchThreadId, FFX_MIN16_I2 iPrevIdx) -{ - FfxFloat32x3 colorPreAlpha = LoadOpaqueOnly(uDispatchThreadId); - FfxFloat32x3 colorPostAlpha = LoadInputColor(uDispatchThreadId); - FfxFloat32x3 colorPrevPreAlpha = LoadPrevPreAlpha(iPrevIdx); - FfxFloat32x3 colorPrevPostAlpha = LoadPrevPostAlpha(iPrevIdx); - -#if USE_YCOCG - colorPreAlpha = RGBToYCoCg(colorPreAlpha); - colorPostAlpha = RGBToYCoCg(colorPostAlpha); - colorPrevPreAlpha = RGBToYCoCg(colorPrevPreAlpha); - colorPrevPostAlpha = RGBToYCoCg(colorPrevPostAlpha); -#endif - - FfxFloat32x3 colorDeltaCurr = colorPostAlpha - colorPreAlpha; - FfxFloat32x3 colorDeltaPrev = colorPrevPostAlpha - colorPrevPreAlpha; - bool hasAlpha = any(FFX_GREATER_THAN(abs(colorDeltaCurr), FfxFloat32x3(fAutogenEpsilon, fAutogenEpsilon, fAutogenEpsilon))); - bool hadAlpha = any(FFX_GREATER_THAN(abs(colorDeltaPrev), FfxFloat32x3(fAutogenEpsilon, fAutogenEpsilon, fAutogenEpsilon))); - - FfxFloat32x3 X = colorPreAlpha; - FfxFloat32x3 Y = colorPostAlpha; - FfxFloat32x3 Z = colorPrevPreAlpha; - FfxFloat32x3 W = colorPrevPostAlpha; - - FFX_MIN16_F retVal = FFX_MIN16_F(ffxSaturate(dot(abs(abs(Y - X) - abs(W - Z)), FfxFloat32x3(1, 1, 1)))); - - // cleanup very small values - retVal = (retVal < getTcThreshold()) ? FFX_MIN16_F(0.0f) : FFX_MIN16_F(1.f); - - return retVal; -} - -// works ok: thin edges -FFX_MIN16_F ComputeAutoTC_02(FFX_MIN16_I2 uDispatchThreadId, FFX_MIN16_I2 iPrevIdx) -{ - FfxFloat32x3 colorPreAlpha = LoadOpaqueOnly(uDispatchThreadId); - FfxFloat32x3 colorPostAlpha = LoadInputColor(uDispatchThreadId); - FfxFloat32x3 colorPrevPreAlpha = LoadPrevPreAlpha(iPrevIdx); - FfxFloat32x3 colorPrevPostAlpha = LoadPrevPostAlpha(iPrevIdx); - -#if USE_YCOCG - colorPreAlpha = RGBToYCoCg(colorPreAlpha); - colorPostAlpha = RGBToYCoCg(colorPostAlpha); - colorPrevPreAlpha = RGBToYCoCg(colorPrevPreAlpha); - colorPrevPostAlpha = RGBToYCoCg(colorPrevPostAlpha); -#endif - - FfxFloat32x3 colorDelta = colorPostAlpha - colorPreAlpha; - FfxFloat32x3 colorPrevDelta = colorPrevPostAlpha - colorPrevPreAlpha; - bool hasAlpha = any(FFX_GREATER_THAN(abs(colorDelta), FfxFloat32x3(fAutogenEpsilon, fAutogenEpsilon, fAutogenEpsilon))); - bool hadAlpha = any(FFX_GREATER_THAN(abs(colorPrevDelta), FfxFloat32x3(fAutogenEpsilon, fAutogenEpsilon, fAutogenEpsilon))); - - FfxFloat32x3 delta = colorPostAlpha - colorPreAlpha; //prev+1*d = post => d = color, alpha = - FfxFloat32x3 deltaPrev = colorPrevPostAlpha - colorPrevPreAlpha; - - FfxFloat32x3 X = colorPrevPreAlpha; - FfxFloat32x3 N = colorPreAlpha - colorPrevPreAlpha; - FfxFloat32x3 YAminusXA = colorPrevPostAlpha - colorPrevPreAlpha; - FfxFloat32x3 NminusNA = colorPostAlpha - colorPrevPostAlpha; - - FfxFloat32x3 A = (hasAlpha || hadAlpha) ? NminusNA / max(FfxFloat32x3(fAutogenEpsilon, fAutogenEpsilon, fAutogenEpsilon), N) : FfxFloat32x3(0, 0, 0); - - FFX_MIN16_F retVal = FFX_MIN16_F( max(max(A.x, A.y), A.z) ); - - // only pixels that have significantly changed in color shuold be considered - retVal = ffxSaturate(retVal * FFX_MIN16_F(length(colorPostAlpha - colorPrevPostAlpha)) ); - - return retVal; -} - -// This function computes the TransparencyAndComposition mask: -// This mask indicates pixels that should discard locks and apply color clamping. -// -// Typically this is the case for translucent pixels (that don't write depth values) or pixels where the correctness of -// the MVs can not be guaranteed (e.g. procedutal movement or vegetation that does not have MVs to reduce the cost during rasterization) -// Also, large changes in color due to changed lighting should be marked to remove locks on pixels with "old" lighting. -// -// This function takes a opaque only and a final texture and uses internal copies of those textures from the last frame. -// The function tries to determine where the color changes between opaque only and final image to determine the pixels that use transparency. -// Also it uses the previous frames and detects where the use of transparency changed to mark those pixels. -// Additionally it marks pixels where the color changed significantly in the opaque only image, e.g. due to lighting or texture animation. -// -// In the final step it stores the current textures in internal textures for the next frame - -FFX_MIN16_F ComputeTransparencyAndComposition(FFX_MIN16_I2 uDispatchThreadId, FFX_MIN16_I2 iPrevIdx) -{ - FFX_MIN16_F retVal = ComputeAutoTC_02(uDispatchThreadId, iPrevIdx); - - // [branch] - if (retVal > FFX_MIN16_F(0.01f)) - { - retVal = ComputeAutoTC_01(uDispatchThreadId, iPrevIdx); - } - return retVal; -} - -float computeSolidEdge(FFX_MIN16_I2 curPos, FFX_MIN16_I2 prevPos) -{ - float lum[9]; - int i = 0; - for (int y = -1; y < 2; ++y) - { - for (int x = -1; x < 2; ++x) - { - FfxFloat32x3 curCol = LoadOpaqueOnly(curPos + FFX_MIN16_I2(x, y)).rgb; - FfxFloat32x3 prevCol = LoadPrevPreAlpha(prevPos + FFX_MIN16_I2(x, y)).rgb; - lum[i++] = length(curCol - prevCol); - } - } - - //float gradX = abs(lum[3] - lum[4]) + abs(lum[5] - lum[4]); - //float gradY = abs(lum[1] - lum[4]) + abs(lum[7] - lum[4]); - - //return sqrt(gradX * gradX + gradY * gradY); - - float gradX = abs(lum[3] - lum[4]) * abs(lum[5] - lum[4]); - float gradY = abs(lum[1] - lum[4]) * abs(lum[7] - lum[4]); - - return sqrt(sqrt(gradX * gradY)); -} - -float computeAlphaEdge(FFX_MIN16_I2 curPos, FFX_MIN16_I2 prevPos) -{ - float lum[9]; - int i = 0; - for (int y = -1; y < 2; ++y) - { - for (int x = -1; x < 2; ++x) - { - FfxFloat32x3 curCol = abs(LoadInputColor(curPos + FFX_MIN16_I2(x, y)).rgb - LoadOpaqueOnly(curPos + FFX_MIN16_I2(x, y)).rgb); - FfxFloat32x3 prevCol = abs(LoadPrevPostAlpha(prevPos + FFX_MIN16_I2(x, y)).rgb - LoadPrevPreAlpha(prevPos + FFX_MIN16_I2(x, y)).rgb); - lum[i++] = length(curCol - prevCol); - } - } - - //float gradX = abs(lum[3] - lum[4]) + abs(lum[5] - lum[4]); - //float gradY = abs(lum[1] - lum[4]) + abs(lum[7] - lum[4]); - - //return sqrt(gradX * gradX + gradY * gradY); - - float gradX = abs(lum[3] - lum[4]) * abs(lum[5] - lum[4]); - float gradY = abs(lum[1] - lum[4]) * abs(lum[7] - lum[4]); - - return sqrt(sqrt(gradX * gradY)); -} - -FFX_MIN16_F ComputeAabbOverlap(FFX_MIN16_I2 uDispatchThreadId, FFX_MIN16_I2 iPrevIdx) -{ - FFX_MIN16_F retVal = FFX_MIN16_F(0.f); - - FfxFloat32x2 fMotionVector = LoadInputMotionVector(uDispatchThreadId); - FfxFloat32x3 colorPreAlpha = LoadOpaqueOnly(uDispatchThreadId); - FfxFloat32x3 colorPostAlpha = LoadInputColor(uDispatchThreadId); - FfxFloat32x3 colorPrevPreAlpha = LoadPrevPreAlpha(iPrevIdx); - FfxFloat32x3 colorPrevPostAlpha = LoadPrevPostAlpha(iPrevIdx); - -#if USE_YCOCG - colorPreAlpha = RGBToYCoCg(colorPreAlpha); - colorPostAlpha = RGBToYCoCg(colorPostAlpha); - colorPrevPreAlpha = RGBToYCoCg(colorPrevPreAlpha); - colorPrevPostAlpha = RGBToYCoCg(colorPrevPostAlpha); -#endif - FfxFloat32x3 minPrev = FFX_MIN16_F3(+1000.f, +1000.f, +1000.f); - FfxFloat32x3 maxPrev = FFX_MIN16_F3(-1000.f, -1000.f, -1000.f); - for (int y = -1; y < 2; ++y) - { - for (int x = -1; x < 2; ++x) - { - FfxFloat32x3 W = LoadPrevPostAlpha(iPrevIdx + FFX_MIN16_I2(x, y)); - -#if USE_YCOCG - W = RGBToYCoCg(W); -#endif - minPrev = min(minPrev, W); - maxPrev = max(maxPrev, W); - } - } - // instead of computing the overlap: simply count how many samples are outside - // set reactive based on that - FFX_MIN16_F count = FFX_MIN16_F(0.f); - for (int y = -1; y < 2; ++y) - { - for (int x = -1; x < 2; ++x) - { - FfxFloat32x3 Y = LoadInputColor(uDispatchThreadId + FFX_MIN16_I2(x, y)); - -#if USE_YCOCG - Y = RGBToYCoCg(Y); -#endif - count += ((Y.x < minPrev.x) || (Y.x > maxPrev.x)) ? FFX_MIN16_F(1.f) : FFX_MIN16_F(0.f); - count += ((Y.y < minPrev.y) || (Y.y > maxPrev.y)) ? FFX_MIN16_F(1.f) : FFX_MIN16_F(0.f); - count += ((Y.z < minPrev.z) || (Y.z > maxPrev.z)) ? FFX_MIN16_F(1.f) : FFX_MIN16_F(0.f); - } - } - retVal = count / FFX_MIN16_F(27.f); - - return retVal; -} - - -// This function computes the Reactive mask: -// We want pixels marked where the alpha portion of the frame changes a lot between neighbours -// Those pixels are expected to change quickly between frames, too. (e.g. small particles, reflections on curved surfaces...) -// As a result history would not be trustworthy. -// On the other hand we don't want pixels marked where pre-alpha has a large differnce, since those would profit from accumulation -// For mirrors we may assume the pre-alpha is pretty uniform color. -// -// This works well generally, but also marks edge pixels -FFX_MIN16_F ComputeReactive(FFX_MIN16_I2 uDispatchThreadId, FFX_MIN16_I2 iPrevIdx) -{ - // we only get here if alpha has a significant contribution and has changed since last frame. - FFX_MIN16_F retVal = FFX_MIN16_F(0.f); - - // mark pixels with huge variance in alpha as reactive - FFX_MIN16_F alphaEdge = FFX_MIN16_F(computeAlphaEdge(uDispatchThreadId, iPrevIdx)); - FFX_MIN16_F opaqueEdge = FFX_MIN16_F(computeSolidEdge(uDispatchThreadId, iPrevIdx)); - retVal = ffxSaturate(alphaEdge - opaqueEdge); - - // the above also marks edge pixels due to jitter, so we need to cancel those out - - - return retVal; -} diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_tcr_autogen.h.meta b/Assets/Resources/FSR2/shaders/ffx_fsr2_tcr_autogen.h.meta deleted file mode 100644 index a2af34e..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_tcr_autogen.h.meta +++ /dev/null @@ -1,60 +0,0 @@ -fileFormatVersion: 2 -guid: 18b8590c99b171a4e9af68dfd2c3ff02 -PluginImporter: - externalObjects: {} - serializedVersion: 2 - iconMap: {} - executionOrder: {} - defineConstraints: [] - isPreloaded: 0 - isOverridable: 0 - isExplicitlyReferenced: 0 - validateReferences: 1 - platformData: - - first: - : Any - second: - enabled: 0 - settings: - Exclude Editor: 1 - Exclude GameCoreScarlett: 1 - Exclude GameCoreXboxOne: 1 - Exclude Linux64: 1 - Exclude OSXUniversal: 1 - Exclude PS4: 1 - Exclude PS5: 1 - Exclude WebGL: 1 - Exclude Win: 1 - Exclude Win64: 1 - - first: - Any: - second: - enabled: 0 - settings: {} - - first: - Editor: Editor - second: - enabled: 0 - settings: - DefaultValueInitialized: true - - first: - Standalone: Linux64 - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win64 - second: - enabled: 0 - settings: - CPU: None - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_tcr_autogen_pass.hlsl b/Assets/Resources/FSR2/shaders/ffx_fsr2_tcr_autogen_pass.hlsl deleted file mode 100644 index 8e635d1..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_tcr_autogen_pass.hlsl +++ /dev/null @@ -1,114 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#define FSR2_BIND_SRV_INPUT_OPAQUE_ONLY 0 -#define FSR2_BIND_SRV_INPUT_COLOR 1 -#define FSR2_BIND_SRV_INPUT_MOTION_VECTORS 2 -#define FSR2_BIND_SRV_PREV_PRE_ALPHA_COLOR 3 -#define FSR2_BIND_SRV_PREV_POST_ALPHA_COLOR 4 -#define FSR2_BIND_SRV_REACTIVE_MASK 4 -#define FSR2_BIND_SRV_TRANSPARENCY_AND_COMPOSITION_MASK 5 - -#define FSR2_BIND_UAV_AUTOREACTIVE 0 -#define FSR2_BIND_UAV_AUTOCOMPOSITION 1 -#define FSR2_BIND_UAV_PREV_PRE_ALPHA_COLOR 2 -#define FSR2_BIND_UAV_PREV_POST_ALPHA_COLOR 3 - -#define FSR2_BIND_CB_FSR2 0 -#define FSR2_BIND_CB_AUTOREACTIVE 1 - -#include "ffx_fsr2_callbacks_hlsl.h" -#include "ffx_fsr2_common.h" - -#if defined(FSR2_BIND_CB_AUTOREACTIVE) - cbuffer cbGenerateReactive : FFX_FSR2_DECLARE_CB(FSR2_BIND_CB_AUTOREACTIVE) - { - float fTcThreshold; // 0.1 is a good starting value, lower will result in more TC pixels - float fTcScale; - float fReactiveScale; - float fReactiveMax; - }; - float getTcThreshold() - { - return fTcThreshold; - } -#else - #define fTcThreshold 0.05f - #define fTcScale 1.00f - #define fReactiveScale 10.0f - #define fReactiveMax 0.90f - float getTcThreshold() - { - return fTcThreshold; - } -#endif - -#include "ffx_fsr2_tcr_autogen.h" - -#ifndef FFX_FSR2_THREAD_GROUP_WIDTH -#define FFX_FSR2_THREAD_GROUP_WIDTH 8 -#endif // #ifndef FFX_FSR2_THREAD_GROUP_WIDTH -#ifndef FFX_FSR2_THREAD_GROUP_HEIGHT -#define FFX_FSR2_THREAD_GROUP_HEIGHT 8 -#endif // FFX_FSR2_THREAD_GROUP_HEIGHT -#ifndef FFX_FSR2_THREAD_GROUP_DEPTH -#define FFX_FSR2_THREAD_GROUP_DEPTH 1 -#endif // #ifndef FFX_FSR2_THREAD_GROUP_DEPTH -#ifndef FFX_FSR2_NUM_THREADS -#define FFX_FSR2_NUM_THREADS [numthreads(FFX_FSR2_THREAD_GROUP_WIDTH, FFX_FSR2_THREAD_GROUP_HEIGHT, FFX_FSR2_THREAD_GROUP_DEPTH)] -#endif // #ifndef FFX_FSR2_NUM_THREADS - -FFX_FSR2_NUM_THREADS -FFX_FSR2_EMBED_ROOTSIG_CONTENT -void CS(uint2 uGroupId : SV_GroupID, uint2 uGroupThreadId : SV_GroupThreadID) -{ - FFX_MIN16_I2 uDispatchThreadId = FFX_MIN16_I2(uGroupId * uint2(FFX_FSR2_THREAD_GROUP_WIDTH, FFX_FSR2_THREAD_GROUP_HEIGHT) + uGroupThreadId); - - // ToDo: take into account jitter (i.e. add delta of previous jitter and current jitter to previous UV - // fetch pre- and post-alpha color values - FFX_MIN16_F2 fUv = ( FFX_MIN16_F2(uDispatchThreadId) + FFX_MIN16_F2(0.5f, 0.5f) ) / FFX_MIN16_F2( RenderSize() ); - FFX_MIN16_F2 fPrevUV = fUv + FFX_MIN16_F2( LoadInputMotionVector(uDispatchThreadId) ); - FFX_MIN16_I2 iPrevIdx = FFX_MIN16_I2(fPrevUV * FFX_MIN16_F2(RenderSize()) - 0.5f); - - FFX_MIN16_F3 colorPreAlpha = FFX_MIN16_F3( LoadOpaqueOnly( uDispatchThreadId ) ); - FFX_MIN16_F3 colorPostAlpha = FFX_MIN16_F3( LoadInputColor( uDispatchThreadId ) ); - - FFX_MIN16_F2 outReactiveMask = 0; - - outReactiveMask.y = ComputeTransparencyAndComposition(uDispatchThreadId, iPrevIdx); - - if (outReactiveMask.y > 0.5f) - { - outReactiveMask.x = ComputeReactive(uDispatchThreadId, iPrevIdx); - outReactiveMask.x *= FFX_MIN16_F(fReactiveScale); - outReactiveMask.x = outReactiveMask.x < fReactiveMax ? outReactiveMask.x : FFX_MIN16_F( fReactiveMax ); - } - - outReactiveMask.y *= FFX_MIN16_F(fTcScale ); - - outReactiveMask.x = max( outReactiveMask.x, FFX_MIN16_F( LoadReactiveMask(uDispatchThreadId) ) ); - outReactiveMask.y = max( outReactiveMask.y, FFX_MIN16_F( LoadTransparencyAndCompositionMask(uDispatchThreadId) ) ); - - StoreAutoReactive(uDispatchThreadId, outReactiveMask); - - StorePrevPreAlpha(uDispatchThreadId, colorPreAlpha); - StorePrevPostAlpha(uDispatchThreadId, colorPostAlpha); -} diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_tcr_autogen_pass.hlsl.meta b/Assets/Resources/FSR2/shaders/ffx_fsr2_tcr_autogen_pass.hlsl.meta deleted file mode 100644 index d133259..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_tcr_autogen_pass.hlsl.meta +++ /dev/null @@ -1,7 +0,0 @@ -fileFormatVersion: 2 -guid: 9feb1fa4d6cff5a4799298dc69b12a8e -ShaderIncludeImporter: - externalObjects: {} - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_upsample.h b/Assets/Resources/FSR2/shaders/ffx_fsr2_upsample.h deleted file mode 100644 index abdb888..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_upsample.h +++ /dev/null @@ -1,194 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#ifndef FFX_FSR2_UPSAMPLE_H -#define FFX_FSR2_UPSAMPLE_H - -FFX_STATIC const FfxUInt32 iLanczos2SampleCount = 16; - -void Deringing(RectificationBox clippingBox, FFX_PARAMETER_INOUT FfxFloat32x3 fColor) -{ - fColor = clamp(fColor, clippingBox.aabbMin, clippingBox.aabbMax); -} -#if FFX_HALF -void Deringing(RectificationBoxMin16 clippingBox, FFX_PARAMETER_INOUT FFX_MIN16_F3 fColor) -{ - fColor = clamp(fColor, clippingBox.aabbMin, clippingBox.aabbMax); -} -#endif - -#ifndef FFX_FSR2_OPTION_UPSAMPLE_USE_LANCZOS_TYPE -#define FFX_FSR2_OPTION_UPSAMPLE_USE_LANCZOS_TYPE 2 // Approximate -#endif - -FfxFloat32 GetUpsampleLanczosWeight(FfxFloat32x2 fSrcSampleOffset, FfxFloat32 fKernelWeight) -{ - FfxFloat32x2 fSrcSampleOffsetBiased = fSrcSampleOffset * fKernelWeight.xx; -#if FFX_FSR2_OPTION_UPSAMPLE_USE_LANCZOS_TYPE == 0 // LANCZOS_TYPE_REFERENCE - FfxFloat32 fSampleWeight = Lanczos2(length(fSrcSampleOffsetBiased)); -#elif FFX_FSR2_OPTION_UPSAMPLE_USE_LANCZOS_TYPE == 1 // LANCZOS_TYPE_LUT - FfxFloat32 fSampleWeight = Lanczos2_UseLUT(length(fSrcSampleOffsetBiased)); -#elif FFX_FSR2_OPTION_UPSAMPLE_USE_LANCZOS_TYPE == 2 // LANCZOS_TYPE_APPROXIMATE - FfxFloat32 fSampleWeight = Lanczos2ApproxSq(dot(fSrcSampleOffsetBiased, fSrcSampleOffsetBiased)); -#else -#error "Invalid Lanczos type" -#endif - return fSampleWeight; -} - -#if FFX_HALF -FFX_MIN16_F GetUpsampleLanczosWeight(FFX_MIN16_F2 fSrcSampleOffset, FFX_MIN16_F fKernelWeight) -{ - FFX_MIN16_F2 fSrcSampleOffsetBiased = fSrcSampleOffset * fKernelWeight.xx; -#if FFX_FSR2_OPTION_UPSAMPLE_USE_LANCZOS_TYPE == 0 // LANCZOS_TYPE_REFERENCE - FFX_MIN16_F fSampleWeight = Lanczos2(length(fSrcSampleOffsetBiased)); -#elif FFX_FSR2_OPTION_UPSAMPLE_USE_LANCZOS_TYPE == 1 // LANCZOS_TYPE_LUT - FFX_MIN16_F fSampleWeight = Lanczos2_UseLUT(length(fSrcSampleOffsetBiased)); -#elif FFX_FSR2_OPTION_UPSAMPLE_USE_LANCZOS_TYPE == 2 // LANCZOS_TYPE_APPROXIMATE - FFX_MIN16_F fSampleWeight = Lanczos2ApproxSq(dot(fSrcSampleOffsetBiased, fSrcSampleOffsetBiased)); - - // To Test: Save reciproqual sqrt compute - // FfxFloat32 fSampleWeight = Lanczos2Sq_UseLUT(dot(fSrcSampleOffsetBiased, fSrcSampleOffsetBiased)); -#else -#error "Invalid Lanczos type" -#endif - return fSampleWeight; -} -#endif - -FfxFloat32 ComputeMaxKernelWeight() { - const FfxFloat32 fKernelSizeBias = 1.0f; - - FfxFloat32 fKernelWeight = FfxFloat32(1) + (FfxFloat32(1.0f) / FfxFloat32x2(DownscaleFactor()) - FfxFloat32(1)).x * FfxFloat32(fKernelSizeBias); - - return ffxMin(FfxFloat32(1.99f), fKernelWeight); -} - -FfxFloat32x4 ComputeUpsampledColorAndWeight(const AccumulationPassCommonParams params, - FFX_PARAMETER_INOUT RectificationBox clippingBox, FfxFloat32 fReactiveFactor) -{ - #if FFX_FSR2_OPTION_UPSAMPLE_SAMPLERS_USE_DATA_HALF && FFX_HALF - #include "ffx_fsr2_force16_begin.h" - #endif - // We compute a sliced lanczos filter with 2 lobes (other slices are accumulated temporaly) - FfxFloat32x2 fDstOutputPos = FfxFloat32x2(params.iPxHrPos) + FFX_BROADCAST_FLOAT32X2(0.5f); // Destination resolution output pixel center position - FfxFloat32x2 fSrcOutputPos = fDstOutputPos * DownscaleFactor(); // Source resolution output pixel center position - FfxInt32x2 iSrcInputPos = FfxInt32x2(floor(fSrcOutputPos)); // TODO: what about weird upscale factors... - - #if FFX_FSR2_OPTION_UPSAMPLE_SAMPLERS_USE_DATA_HALF && FFX_HALF - #include "ffx_fsr2_force16_end.h" - #endif - - FfxFloat32x3 fSamples[iLanczos2SampleCount]; - - FfxFloat32x2 fSrcUnjitteredPos = (FfxFloat32x2(iSrcInputPos) + FfxFloat32x2(0.5f, 0.5f)) - Jitter(); // This is the un-jittered position of the sample at offset 0,0 - - FfxInt32x2 offsetTL; - offsetTL.x = (fSrcUnjitteredPos.x > fSrcOutputPos.x) ? FfxInt32(-2) : FfxInt32(-1); - offsetTL.y = (fSrcUnjitteredPos.y > fSrcOutputPos.y) ? FfxInt32(-2) : FfxInt32(-1); - - //Load samples - // If fSrcUnjitteredPos.y > fSrcOutputPos.y, indicates offsetTL.y = -2, sample offset Y will be [-2, 1], clipbox will be rows [1, 3]. - // Flip row# for sampling offset in this case, so first 0~2 rows in the sampled array can always be used for computing the clipbox. - // This reduces branch or cmove on sampled colors, but moving this overhead to sample position / weight calculation time which apply to less values. - const FfxBoolean bFlipRow = fSrcUnjitteredPos.y > fSrcOutputPos.y; - const FfxBoolean bFlipCol = fSrcUnjitteredPos.x > fSrcOutputPos.x; - - FfxFloat32x2 fOffsetTL = FfxFloat32x2(offsetTL); - - FFX_UNROLL - for (FfxInt32 row = 0; row < 3; row++) { - - FFX_UNROLL - for (FfxInt32 col = 0; col < 3; col++) { - FfxInt32 iSampleIndex = col + (row << 2); - - FfxInt32x2 sampleColRow = FfxInt32x2(bFlipCol ? (3 - col) : col, bFlipRow ? (3 - row) : row); - FfxInt32x2 iSrcSamplePos = FfxInt32x2(iSrcInputPos) + offsetTL + sampleColRow; - - const FfxInt32x2 sampleCoord = ClampLoad(iSrcSamplePos, FfxInt32x2(0, 0), FfxInt32x2(RenderSize())); - - fSamples[iSampleIndex] = LoadPreparedInputColor(FfxInt32x2(sampleCoord)); - } - } - - FfxFloat32x4 fColorAndWeight = FfxFloat32x4(0.0f, 0.0f, 0.0f, 0.0f); - - FfxFloat32x2 fBaseSampleOffset = FfxFloat32x2(fSrcUnjitteredPos - fSrcOutputPos); - - // Identify how much of each upsampled color to be used for this frame - const FfxFloat32 fKernelReactiveFactor = ffxMax(fReactiveFactor, FfxFloat32(params.bIsNewSample)); - const FfxFloat32 fKernelBiasMax = ComputeMaxKernelWeight() * (1.0f - fKernelReactiveFactor); - - const FfxFloat32 fKernelBiasMin = ffxMax(1.0f, ((1.0f + fKernelBiasMax) * 0.3f)); - const FfxFloat32 fKernelBiasFactor = ffxMax(0.0f, ffxMax(0.25f * params.fDepthClipFactor, fKernelReactiveFactor)); - const FfxFloat32 fKernelBias = ffxLerp(fKernelBiasMax, fKernelBiasMin, fKernelBiasFactor); - - const FfxFloat32 fRectificationCurveBias = ffxLerp(-2.0f, -3.0f, ffxSaturate(params.fHrVelocity / 50.0f)); - - FFX_UNROLL - for (FfxInt32 row = 0; row < 3; row++) { - FFX_UNROLL - for (FfxInt32 col = 0; col < 3; col++) { - FfxInt32 iSampleIndex = col + (row << 2); - - const FfxInt32x2 sampleColRow = FfxInt32x2(bFlipCol ? (3 - col) : col, bFlipRow ? (3 - row) : row); - const FfxFloat32x2 fOffset = fOffsetTL + FfxFloat32x2(sampleColRow); - FfxFloat32x2 fSrcSampleOffset = fBaseSampleOffset + fOffset; - - FfxInt32x2 iSrcSamplePos = FfxInt32x2(iSrcInputPos) + FfxInt32x2(offsetTL) + sampleColRow; - - const FfxFloat32 fOnScreenFactor = FfxFloat32(IsOnScreen(FfxInt32x2(iSrcSamplePos), FfxInt32x2(RenderSize()))); - FfxFloat32 fSampleWeight = fOnScreenFactor * FfxFloat32(GetUpsampleLanczosWeight(fSrcSampleOffset, fKernelBias)); - - fColorAndWeight += FfxFloat32x4(fSamples[iSampleIndex] * fSampleWeight, fSampleWeight); - - // Update rectification box - { - const FfxFloat32 fSrcSampleOffsetSq = dot(fSrcSampleOffset, fSrcSampleOffset); - const FfxFloat32 fBoxSampleWeight = exp(fRectificationCurveBias * fSrcSampleOffsetSq); - - const FfxBoolean bInitialSample = (row == 0) && (col == 0); - RectificationBoxAddSample(bInitialSample, clippingBox, fSamples[iSampleIndex], fBoxSampleWeight); - } - } - } - - RectificationBoxComputeVarianceBoxData(clippingBox); - - fColorAndWeight.w *= FfxFloat32(fColorAndWeight.w > FSR2_EPSILON); - - if (fColorAndWeight.w > FSR2_EPSILON) { - // Normalize for deringing (we need to compare colors) - fColorAndWeight.xyz = fColorAndWeight.xyz / fColorAndWeight.w; - fColorAndWeight.w *= fUpsampleLanczosWeightScale; - - Deringing(clippingBox, fColorAndWeight.xyz); - } - - #if FFX_FSR2_OPTION_UPSAMPLE_SAMPLERS_USE_DATA_HALF && FFX_HALF - #include "ffx_fsr2_force16_end.h" - #endif - - return fColorAndWeight; -} - -#endif //!defined( FFX_FSR2_UPSAMPLE_H ) diff --git a/Assets/Resources/FSR2/shaders/ffx_fsr2_upsample.h.meta b/Assets/Resources/FSR2/shaders/ffx_fsr2_upsample.h.meta deleted file mode 100644 index c9fa0a3..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_fsr2_upsample.h.meta +++ /dev/null @@ -1,60 +0,0 @@ -fileFormatVersion: 2 -guid: 1ff3a385cfe07db4387e4d7b457238f8 -PluginImporter: - externalObjects: {} - serializedVersion: 2 - iconMap: {} - executionOrder: {} - defineConstraints: [] - isPreloaded: 0 - isOverridable: 0 - isExplicitlyReferenced: 0 - validateReferences: 1 - platformData: - - first: - : Any - second: - enabled: 0 - settings: - Exclude Editor: 1 - Exclude GameCoreScarlett: 1 - Exclude GameCoreXboxOne: 1 - Exclude Linux64: 1 - Exclude OSXUniversal: 1 - Exclude PS4: 1 - Exclude PS5: 1 - Exclude WebGL: 1 - Exclude Win: 1 - Exclude Win64: 1 - - first: - Any: - second: - enabled: 0 - settings: {} - - first: - Editor: Editor - second: - enabled: 0 - settings: - DefaultValueInitialized: true - - first: - Standalone: Linux64 - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win64 - second: - enabled: 0 - settings: - CPU: None - userData: - assetBundleName: - assetBundleVariant: diff --git a/Assets/Resources/FSR2/shaders/ffx_spd.h b/Assets/Resources/FSR2/shaders/ffx_spd.h deleted file mode 100644 index b94f46f..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_spd.h +++ /dev/null @@ -1,929 +0,0 @@ -// This file is part of the FidelityFX SDK. -// -// Copyright (c) 2022-2023 Advanced Micro Devices, Inc. All rights reserved. -// -// Permission is hereby granted, free of charge, to any person obtaining a copy -// of this software and associated documentation files (the "Software"), to deal -// in the Software without restriction, including without limitation the rights -// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell -// copies of the Software, and to permit persons to whom the Software is -// furnished to do so, subject to the following conditions: -// The above copyright notice and this permission notice shall be included in -// all copies or substantial portions of the Software. -// -// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN -// THE SOFTWARE. - -#ifdef FFX_CPU -FFX_STATIC void SpdSetup(FfxUInt32x2 dispatchThreadGroupCountXY, // CPU side: dispatch thread group count xy - FfxUInt32x2 workGroupOffset, // GPU side: pass in as constant - FfxUInt32x2 numWorkGroupsAndMips, // GPU side: pass in as constant - FfxUInt32x4 rectInfo, // left, top, width, height - FfxInt32 mips) // optional: if -1, calculate based on rect width and height -{ - workGroupOffset[0] = rectInfo[0] / 64; // rectInfo[0] = left - workGroupOffset[1] = rectInfo[1] / 64; // rectInfo[1] = top - - FfxUInt32 endIndexX = (rectInfo[0] + rectInfo[2] - 1) / 64; // rectInfo[0] = left, rectInfo[2] = width - FfxUInt32 endIndexY = (rectInfo[1] + rectInfo[3] - 1) / 64; // rectInfo[1] = top, rectInfo[3] = height - - dispatchThreadGroupCountXY[0] = endIndexX + 1 - workGroupOffset[0]; - dispatchThreadGroupCountXY[1] = endIndexY + 1 - workGroupOffset[1]; - - numWorkGroupsAndMips[0] = (dispatchThreadGroupCountXY[0]) * (dispatchThreadGroupCountXY[1]); - - if (mips >= 0) - { - numWorkGroupsAndMips[1] = FfxUInt32(mips); - } - else - { - // calculate based on rect width and height - FfxUInt32 resolution = ffxMax(rectInfo[2], rectInfo[3]); - numWorkGroupsAndMips[1] = FfxUInt32((ffxMin(floor(log2(FfxFloat32(resolution))), FfxFloat32(12)))); - } -} - -FFX_STATIC void SpdSetup(FfxUInt32x2 dispatchThreadGroupCountXY, // CPU side: dispatch thread group count xy - FfxUInt32x2 workGroupOffset, // GPU side: pass in as constant - FfxUInt32x2 numWorkGroupsAndMips, // GPU side: pass in as constant - FfxUInt32x4 rectInfo) // left, top, width, height -{ - SpdSetup(dispatchThreadGroupCountXY, workGroupOffset, numWorkGroupsAndMips, rectInfo, -1); -} -#endif // #ifdef FFX_CPU - - -//============================================================================================================================== -// NON-PACKED VERSION -//============================================================================================================================== -#ifdef FFX_GPU -#ifdef SPD_PACKED_ONLY -// Avoid compiler error -FfxFloat32x4 SpdLoadSourceImage(FfxInt32x2 p, FfxUInt32 slice) -{ - return FfxFloat32x4(0.0, 0.0, 0.0, 0.0); -} - -FfxFloat32x4 SpdLoad(FfxInt32x2 p, FfxUInt32 slice) -{ - return FfxFloat32x4(0.0, 0.0, 0.0, 0.0); -} -void SpdStore(FfxInt32x2 p, FfxFloat32x4 value, FfxUInt32 mip, FfxUInt32 slice) -{ -} -FfxFloat32x4 SpdLoadIntermediate(FfxUInt32 x, FfxUInt32 y) -{ - return FfxFloat32x4(0.0, 0.0, 0.0, 0.0); -} -void SpdStoreIntermediate(FfxUInt32 x, FfxUInt32 y, FfxFloat32x4 value) -{ -} -FfxFloat32x4 SpdReduce4(FfxFloat32x4 v0, FfxFloat32x4 v1, FfxFloat32x4 v2, FfxFloat32x4 v3) -{ - return FfxFloat32x4(0.0, 0.0, 0.0, 0.0); -} -#endif // #ifdef SPD_PACKED_ONLY - -//_____________________________________________________________/\_______________________________________________________________ - -void SpdWorkgroupShuffleBarrier() -{ -#ifdef FFX_GLSL - barrier(); -#endif -#ifdef FFX_HLSL - GroupMemoryBarrierWithGroupSync(); -#endif -} - -// Only last active workgroup should proceed -bool SpdExitWorkgroup(FfxUInt32 numWorkGroups, FfxUInt32 localInvocationIndex, FfxUInt32 slice) -{ - // global atomic counter - if (localInvocationIndex == 0) - { - SpdIncreaseAtomicCounter(slice); - } - - SpdWorkgroupShuffleBarrier(); - return (SpdGetAtomicCounter() != (numWorkGroups - 1)); -} - -// User defined: FfxFloat32x4 SpdReduce4(FfxFloat32x4 v0, FfxFloat32x4 v1, FfxFloat32x4 v2, FfxFloat32x4 v3); -FfxFloat32x4 SpdReduceQuad(FfxFloat32x4 v) -{ -#if defined(FFX_GLSL) && !defined(SPD_NO_WAVE_OPERATIONS) - - FfxFloat32x4 v0 = v; - FfxFloat32x4 v1 = subgroupQuadSwapHorizontal(v); - FfxFloat32x4 v2 = subgroupQuadSwapVertical(v); - FfxFloat32x4 v3 = subgroupQuadSwapDiagonal(v); - return SpdReduce4(v0, v1, v2, v3); - -#elif defined(FFX_HLSL) && !defined(SPD_NO_WAVE_OPERATIONS) - - // requires SM6.0 - FfxUInt32 quad = WaveGetLaneIndex() & (~0x3); - FfxFloat32x4 v0 = v; - FfxFloat32x4 v1 = WaveReadLaneAt(v, quad | 1); - FfxFloat32x4 v2 = WaveReadLaneAt(v, quad | 2); - FfxFloat32x4 v3 = WaveReadLaneAt(v, quad | 3); - return SpdReduce4(v0, v1, v2, v3); -/* - // if SM6.0 is not available, you can use the AMD shader intrinsics - // the AMD shader intrinsics are available in AMD GPU Services (AGS) library: - // https://gpuopen.com/amd-gpu-services-ags-library/ - // works for DX11 - FfxFloat32x4 v0 = v; - FfxFloat32x4 v1; - v1.x = AmdExtD3DShaderIntrinsics_SwizzleF(v.x, AmdExtD3DShaderIntrinsicsSwizzle_SwapX1); - v1.y = AmdExtD3DShaderIntrinsics_SwizzleF(v.y, AmdExtD3DShaderIntrinsicsSwizzle_SwapX1); - v1.z = AmdExtD3DShaderIntrinsics_SwizzleF(v.z, AmdExtD3DShaderIntrinsicsSwizzle_SwapX1); - v1.w = AmdExtD3DShaderIntrinsics_SwizzleF(v.w, AmdExtD3DShaderIntrinsicsSwizzle_SwapX1); - FfxFloat32x4 v2; - v2.x = AmdExtD3DShaderIntrinsics_SwizzleF(v.x, AmdExtD3DShaderIntrinsicsSwizzle_SwapX2); - v2.y = AmdExtD3DShaderIntrinsics_SwizzleF(v.y, AmdExtD3DShaderIntrinsicsSwizzle_SwapX2); - v2.z = AmdExtD3DShaderIntrinsics_SwizzleF(v.z, AmdExtD3DShaderIntrinsicsSwizzle_SwapX2); - v2.w = AmdExtD3DShaderIntrinsics_SwizzleF(v.w, AmdExtD3DShaderIntrinsicsSwizzle_SwapX2); - FfxFloat32x4 v3; - v3.x = AmdExtD3DShaderIntrinsics_SwizzleF(v.x, AmdExtD3DShaderIntrinsicsSwizzle_ReverseX4); - v3.y = AmdExtD3DShaderIntrinsics_SwizzleF(v.y, AmdExtD3DShaderIntrinsicsSwizzle_ReverseX4); - v3.z = AmdExtD3DShaderIntrinsics_SwizzleF(v.z, AmdExtD3DShaderIntrinsicsSwizzle_ReverseX4); - v3.w = AmdExtD3DShaderIntrinsics_SwizzleF(v.w, AmdExtD3DShaderIntrinsicsSwizzle_ReverseX4); - return SpdReduce4(v0, v1, v2, v3); - */ -#endif - return v; -} - -FfxFloat32x4 SpdReduceIntermediate(FfxUInt32x2 i0, FfxUInt32x2 i1, FfxUInt32x2 i2, FfxUInt32x2 i3) -{ - FfxFloat32x4 v0 = SpdLoadIntermediate(i0.x, i0.y); - FfxFloat32x4 v1 = SpdLoadIntermediate(i1.x, i1.y); - FfxFloat32x4 v2 = SpdLoadIntermediate(i2.x, i2.y); - FfxFloat32x4 v3 = SpdLoadIntermediate(i3.x, i3.y); - return SpdReduce4(v0, v1, v2, v3); -} - -FfxFloat32x4 SpdReduceLoad4(FfxUInt32x2 i0, FfxUInt32x2 i1, FfxUInt32x2 i2, FfxUInt32x2 i3, FfxUInt32 slice) -{ - FfxFloat32x4 v0 = SpdLoad(FfxInt32x2(i0), slice); - FfxFloat32x4 v1 = SpdLoad(FfxInt32x2(i1), slice); - FfxFloat32x4 v2 = SpdLoad(FfxInt32x2(i2), slice); - FfxFloat32x4 v3 = SpdLoad(FfxInt32x2(i3), slice); - return SpdReduce4(v0, v1, v2, v3); -} - -FfxFloat32x4 SpdReduceLoad4(FfxUInt32x2 base, FfxUInt32 slice) -{ - return SpdReduceLoad4(FfxUInt32x2(base + FfxUInt32x2(0, 0)), FfxUInt32x2(base + FfxUInt32x2(0, 1)), FfxUInt32x2(base + FfxUInt32x2(1, 0)), FfxUInt32x2(base + FfxUInt32x2(1, 1)), slice); -} - -FfxFloat32x4 SpdReduceLoadSourceImage4(FfxUInt32x2 i0, FfxUInt32x2 i1, FfxUInt32x2 i2, FfxUInt32x2 i3, FfxUInt32 slice) -{ - FfxFloat32x4 v0 = SpdLoadSourceImage(FfxInt32x2(i0), slice); - FfxFloat32x4 v1 = SpdLoadSourceImage(FfxInt32x2(i1), slice); - FfxFloat32x4 v2 = SpdLoadSourceImage(FfxInt32x2(i2), slice); - FfxFloat32x4 v3 = SpdLoadSourceImage(FfxInt32x2(i3), slice); - return SpdReduce4(v0, v1, v2, v3); -} - -FfxFloat32x4 SpdReduceLoadSourceImage(FfxUInt32x2 base, FfxUInt32 slice) -{ -#ifdef SPD_LINEAR_SAMPLER - return SpdLoadSourceImage(FfxInt32x2(base), slice); -#else - return SpdReduceLoadSourceImage4(FfxUInt32x2(base + FfxUInt32x2(0, 0)), FfxUInt32x2(base + FfxUInt32x2(0, 1)), FfxUInt32x2(base + FfxUInt32x2(1, 0)), FfxUInt32x2(base + FfxUInt32x2(1, 1)), slice); -#endif -} - -void SpdDownsampleMips_0_1_Intrinsics(FfxUInt32 x, FfxUInt32 y, FfxUInt32x2 workGroupID, FfxUInt32 localInvocationIndex, FfxUInt32 mip, FfxUInt32 slice) -{ - FfxFloat32x4 v[4]; - - FfxInt32x2 tex = FfxInt32x2(workGroupID.xy * 64) + FfxInt32x2(x * 2, y * 2); - FfxInt32x2 pix = FfxInt32x2(workGroupID.xy * 32) + FfxInt32x2(x, y); - v[0] = SpdReduceLoadSourceImage(tex, slice); - SpdStore(pix, v[0], 0, slice); - - tex = FfxInt32x2(workGroupID.xy * 64) + FfxInt32x2(x * 2 + 32, y * 2); - pix = FfxInt32x2(workGroupID.xy * 32) + FfxInt32x2(x + 16, y); - v[1] = SpdReduceLoadSourceImage(tex, slice); - SpdStore(pix, v[1], 0, slice); - - tex = FfxInt32x2(workGroupID.xy * 64) + FfxInt32x2(x * 2, y * 2 + 32); - pix = FfxInt32x2(workGroupID.xy * 32) + FfxInt32x2(x, y + 16); - v[2] = SpdReduceLoadSourceImage(tex, slice); - SpdStore(pix, v[2], 0, slice); - - tex = FfxInt32x2(workGroupID.xy * 64) + FfxInt32x2(x * 2 + 32, y * 2 + 32); - pix = FfxInt32x2(workGroupID.xy * 32) + FfxInt32x2(x + 16, y + 16); - v[3] = SpdReduceLoadSourceImage(tex, slice); - SpdStore(pix, v[3], 0, slice); - - if (mip <= 1) - return; - - v[0] = SpdReduceQuad(v[0]); - v[1] = SpdReduceQuad(v[1]); - v[2] = SpdReduceQuad(v[2]); - v[3] = SpdReduceQuad(v[3]); - - if ((localInvocationIndex % 4) == 0) - { - SpdStore(FfxInt32x2(workGroupID.xy * 16) + FfxInt32x2(x / 2, y / 2), v[0], 1, slice); - SpdStoreIntermediate(x / 2, y / 2, v[0]); - - SpdStore(FfxInt32x2(workGroupID.xy * 16) + FfxInt32x2(x / 2 + 8, y / 2), v[1], 1, slice); - SpdStoreIntermediate(x / 2 + 8, y / 2, v[1]); - - SpdStore(FfxInt32x2(workGroupID.xy * 16) + FfxInt32x2(x / 2, y / 2 + 8), v[2], 1, slice); - SpdStoreIntermediate(x / 2, y / 2 + 8, v[2]); - - SpdStore(FfxInt32x2(workGroupID.xy * 16) + FfxInt32x2(x / 2 + 8, y / 2 + 8), v[3], 1, slice); - SpdStoreIntermediate(x / 2 + 8, y / 2 + 8, v[3]); - } -} - -void SpdDownsampleMips_0_1_LDS(FfxUInt32 x, FfxUInt32 y, FfxUInt32x2 workGroupID, FfxUInt32 localInvocationIndex, FfxUInt32 mip, FfxUInt32 slice) -{ - FfxFloat32x4 v[4]; - - FfxInt32x2 tex = FfxInt32x2(workGroupID.xy * 64) + FfxInt32x2(x * 2, y * 2); - FfxInt32x2 pix = FfxInt32x2(workGroupID.xy * 32) + FfxInt32x2(x, y); - v[0] = SpdReduceLoadSourceImage(tex, slice); - SpdStore(pix, v[0], 0, slice); - - tex = FfxInt32x2(workGroupID.xy * 64) + FfxInt32x2(x * 2 + 32, y * 2); - pix = FfxInt32x2(workGroupID.xy * 32) + FfxInt32x2(x + 16, y); - v[1] = SpdReduceLoadSourceImage(tex, slice); - SpdStore(pix, v[1], 0, slice); - - tex = FfxInt32x2(workGroupID.xy * 64) + FfxInt32x2(x * 2, y * 2 + 32); - pix = FfxInt32x2(workGroupID.xy * 32) + FfxInt32x2(x, y + 16); - v[2] = SpdReduceLoadSourceImage(tex, slice); - SpdStore(pix, v[2], 0, slice); - - tex = FfxInt32x2(workGroupID.xy * 64) + FfxInt32x2(x * 2 + 32, y * 2 + 32); - pix = FfxInt32x2(workGroupID.xy * 32) + FfxInt32x2(x + 16, y + 16); - v[3] = SpdReduceLoadSourceImage(tex, slice); - SpdStore(pix, v[3], 0, slice); - - if (mip <= 1) - return; - - for (FfxUInt32 i = 0; i < 4; i++) - { - SpdStoreIntermediate(x, y, v[i]); - SpdWorkgroupShuffleBarrier(); - if (localInvocationIndex < 64) - { - v[i] = SpdReduceIntermediate(FfxUInt32x2(x * 2 + 0, y * 2 + 0), FfxUInt32x2(x * 2 + 1, y * 2 + 0), FfxUInt32x2(x * 2 + 0, y * 2 + 1), FfxUInt32x2(x * 2 + 1, y * 2 + 1)); - SpdStore(FfxInt32x2(workGroupID.xy * 16) + FfxInt32x2(x + (i % 2) * 8, y + (i / 2) * 8), v[i], 1, slice); - } - SpdWorkgroupShuffleBarrier(); - } - - if (localInvocationIndex < 64) - { - SpdStoreIntermediate(x + 0, y + 0, v[0]); - SpdStoreIntermediate(x + 8, y + 0, v[1]); - SpdStoreIntermediate(x + 0, y + 8, v[2]); - SpdStoreIntermediate(x + 8, y + 8, v[3]); - } -} - -void SpdDownsampleMips_0_1(FfxUInt32 x, FfxUInt32 y, FfxUInt32x2 workGroupID, FfxUInt32 localInvocationIndex, FfxUInt32 mip, FfxUInt32 slice) -{ -#ifdef SPD_NO_WAVE_OPERATIONS - SpdDownsampleMips_0_1_LDS(x, y, workGroupID, localInvocationIndex, mip, slice); -#else - SpdDownsampleMips_0_1_Intrinsics(x, y, workGroupID, localInvocationIndex, mip, slice); -#endif -} - - -void SpdDownsampleMip_2(FfxUInt32 x, FfxUInt32 y, FfxUInt32x2 workGroupID, FfxUInt32 localInvocationIndex, FfxUInt32 mip, FfxUInt32 slice) -{ -#ifdef SPD_NO_WAVE_OPERATIONS - if (localInvocationIndex < 64) - { - FfxFloat32x4 v = SpdReduceIntermediate(FfxUInt32x2(x * 2 + 0, y * 2 + 0), FfxUInt32x2(x * 2 + 1, y * 2 + 0), FfxUInt32x2(x * 2 + 0, y * 2 + 1), FfxUInt32x2(x * 2 + 1, y * 2 + 1)); - SpdStore(FfxInt32x2(workGroupID.xy * 8) + FfxInt32x2(x, y), v, mip, slice); - // store to LDS, try to reduce bank conflicts - // x 0 x 0 x 0 x 0 x 0 x 0 x 0 x 0 - // 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 - // 0 x 0 x 0 x 0 x 0 x 0 x 0 x 0 x - // 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 - // x 0 x 0 x 0 x 0 x 0 x 0 x 0 x 0 - // ... - // x 0 x 0 x 0 x 0 x 0 x 0 x 0 x 0 - SpdStoreIntermediate(x * 2 + y % 2, y * 2, v); - } -#else - FfxFloat32x4 v = SpdLoadIntermediate(x, y); - v = SpdReduceQuad(v); - // quad index 0 stores result - if (localInvocationIndex % 4 == 0) - { - SpdStore(FfxInt32x2(workGroupID.xy * 8) + FfxInt32x2(x / 2, y / 2), v, mip, slice); - SpdStoreIntermediate(x + (y / 2) % 2, y, v); - } -#endif -} - -void SpdDownsampleMip_3(FfxUInt32 x, FfxUInt32 y, FfxUInt32x2 workGroupID, FfxUInt32 localInvocationIndex, FfxUInt32 mip, FfxUInt32 slice) -{ -#ifdef SPD_NO_WAVE_OPERATIONS - if (localInvocationIndex < 16) - { - // x 0 x 0 - // 0 0 0 0 - // 0 x 0 x - // 0 0 0 0 - FfxFloat32x4 v = - SpdReduceIntermediate(FfxUInt32x2(x * 4 + 0 + 0, y * 4 + 0), FfxUInt32x2(x * 4 + 2 + 0, y * 4 + 0), FfxUInt32x2(x * 4 + 0 + 1, y * 4 + 2), FfxUInt32x2(x * 4 + 2 + 1, y * 4 + 2)); - SpdStore(FfxInt32x2(workGroupID.xy * 4) + FfxInt32x2(x, y), v, mip, slice); - // store to LDS - // x 0 0 0 x 0 0 0 x 0 0 0 x 0 0 0 - // 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 - // 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 - // 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 - // 0 x 0 0 0 x 0 0 0 x 0 0 0 x 0 0 - // ... - // 0 0 x 0 0 0 x 0 0 0 x 0 0 0 x 0 - // ... - // 0 0 0 x 0 0 0 x 0 0 0 x 0 0 0 x - // ... - SpdStoreIntermediate(x * 4 + y, y * 4, v); - } -#else - if (localInvocationIndex < 64) - { - FfxFloat32x4 v = SpdLoadIntermediate(x * 2 + y % 2, y * 2); - v = SpdReduceQuad(v); - // quad index 0 stores result - if (localInvocationIndex % 4 == 0) - { - SpdStore(FfxInt32x2(workGroupID.xy * 4) + FfxInt32x2(x / 2, y / 2), v, mip, slice); - SpdStoreIntermediate(x * 2 + y / 2, y * 2, v); - } - } -#endif -} - -void SpdDownsampleMip_4(FfxUInt32 x, FfxUInt32 y, FfxUInt32x2 workGroupID, FfxUInt32 localInvocationIndex, FfxUInt32 mip, FfxUInt32 slice) -{ -#ifdef SPD_NO_WAVE_OPERATIONS - if (localInvocationIndex < 4) - { - // x 0 0 0 x 0 0 0 - // ... - // 0 x 0 0 0 x 0 0 - FfxFloat32x4 v = SpdReduceIntermediate(FfxUInt32x2(x * 8 + 0 + 0 + y * 2, y * 8 + 0), - FfxUInt32x2(x * 8 + 4 + 0 + y * 2, y * 8 + 0), - FfxUInt32x2(x * 8 + 0 + 1 + y * 2, y * 8 + 4), - FfxUInt32x2(x * 8 + 4 + 1 + y * 2, y * 8 + 4)); - SpdStore(FfxInt32x2(workGroupID.xy * 2) + FfxInt32x2(x, y), v, mip, slice); - // store to LDS - // x x x x 0 ... - // 0 ... - SpdStoreIntermediate(x + y * 2, 0, v); - } -#else - if (localInvocationIndex < 16) - { - FfxFloat32x4 v = SpdLoadIntermediate(x * 4 + y, y * 4); - v = SpdReduceQuad(v); - // quad index 0 stores result - if (localInvocationIndex % 4 == 0) - { - SpdStore(FfxInt32x2(workGroupID.xy * 2) + FfxInt32x2(x / 2, y / 2), v, mip, slice); - SpdStoreIntermediate(x / 2 + y, 0, v); - } - } -#endif -} - -void SpdDownsampleMip_5(FfxUInt32x2 workGroupID, FfxUInt32 localInvocationIndex, FfxUInt32 mip, FfxUInt32 slice) -{ -#ifdef SPD_NO_WAVE_OPERATIONS - if (localInvocationIndex < 1) - { - // x x x x 0 ... - // 0 ... - FfxFloat32x4 v = SpdReduceIntermediate(FfxUInt32x2(0, 0), FfxUInt32x2(1, 0), FfxUInt32x2(2, 0), FfxUInt32x2(3, 0)); - SpdStore(FfxInt32x2(workGroupID.xy), v, mip, slice); - } -#else - if (localInvocationIndex < 4) - { - FfxFloat32x4 v = SpdLoadIntermediate(localInvocationIndex, 0); - v = SpdReduceQuad(v); - // quad index 0 stores result - if (localInvocationIndex % 4 == 0) - { - SpdStore(FfxInt32x2(workGroupID.xy), v, mip, slice); - } - } -#endif -} - -void SpdDownsampleMips_6_7(FfxUInt32 x, FfxUInt32 y, FfxUInt32 mips, FfxUInt32 slice) -{ - FfxInt32x2 tex = FfxInt32x2(x * 4 + 0, y * 4 + 0); - FfxInt32x2 pix = FfxInt32x2(x * 2 + 0, y * 2 + 0); - FfxFloat32x4 v0 = SpdReduceLoad4(tex, slice); - SpdStore(pix, v0, 6, slice); - - tex = FfxInt32x2(x * 4 + 2, y * 4 + 0); - pix = FfxInt32x2(x * 2 + 1, y * 2 + 0); - FfxFloat32x4 v1 = SpdReduceLoad4(tex, slice); - SpdStore(pix, v1, 6, slice); - - tex = FfxInt32x2(x * 4 + 0, y * 4 + 2); - pix = FfxInt32x2(x * 2 + 0, y * 2 + 1); - FfxFloat32x4 v2 = SpdReduceLoad4(tex, slice); - SpdStore(pix, v2, 6, slice); - - tex = FfxInt32x2(x * 4 + 2, y * 4 + 2); - pix = FfxInt32x2(x * 2 + 1, y * 2 + 1); - FfxFloat32x4 v3 = SpdReduceLoad4(tex, slice); - SpdStore(pix, v3, 6, slice); - - if (mips <= 7) - return; - // no barrier needed, working on values only from the same thread - - FfxFloat32x4 v = SpdReduce4(v0, v1, v2, v3); - SpdStore(FfxInt32x2(x, y), v, 7, slice); - SpdStoreIntermediate(x, y, v); -} - -void SpdDownsampleNextFour(FfxUInt32 x, FfxUInt32 y, FfxUInt32x2 workGroupID, FfxUInt32 localInvocationIndex, FfxUInt32 baseMip, FfxUInt32 mips, FfxUInt32 slice) -{ - if (mips <= baseMip) - return; - SpdWorkgroupShuffleBarrier(); - SpdDownsampleMip_2(x, y, workGroupID, localInvocationIndex, baseMip, slice); - - if (mips <= baseMip + 1) - return; - SpdWorkgroupShuffleBarrier(); - SpdDownsampleMip_3(x, y, workGroupID, localInvocationIndex, baseMip + 1, slice); - - if (mips <= baseMip + 2) - return; - SpdWorkgroupShuffleBarrier(); - SpdDownsampleMip_4(x, y, workGroupID, localInvocationIndex, baseMip + 2, slice); - - if (mips <= baseMip + 3) - return; - SpdWorkgroupShuffleBarrier(); - SpdDownsampleMip_5(workGroupID, localInvocationIndex, baseMip + 3, slice); -} - -void SpdDownsample(FfxUInt32x2 workGroupID, FfxUInt32 localInvocationIndex, FfxUInt32 mips, FfxUInt32 numWorkGroups, FfxUInt32 slice) -{ - FfxUInt32x2 sub_xy = ffxRemapForWaveReduction(localInvocationIndex % 64); - FfxUInt32 x = sub_xy.x + 8 * ((localInvocationIndex >> 6) % 2); - FfxUInt32 y = sub_xy.y + 8 * ((localInvocationIndex >> 7)); - SpdDownsampleMips_0_1(x, y, workGroupID, localInvocationIndex, mips, slice); - - SpdDownsampleNextFour(x, y, workGroupID, localInvocationIndex, 2, mips, slice); - - if (mips <= 6) - return; - - if (SpdExitWorkgroup(numWorkGroups, localInvocationIndex, slice)) - return; - - SpdResetAtomicCounter(slice); - - // After mip 6 there is only a single workgroup left that downsamples the remaining up to 64x64 texels. - SpdDownsampleMips_6_7(x, y, mips, slice); - - SpdDownsampleNextFour(x, y, FfxUInt32x2(0, 0), localInvocationIndex, 8, mips, slice); -} - -void SpdDownsample(FfxUInt32x2 workGroupID, FfxUInt32 localInvocationIndex, FfxUInt32 mips, FfxUInt32 numWorkGroups, FfxUInt32 slice, FfxUInt32x2 workGroupOffset) -{ - SpdDownsample(workGroupID + workGroupOffset, localInvocationIndex, mips, numWorkGroups, slice); -} - -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// -//////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// - -//============================================================================================================================== -// PACKED VERSION -//============================================================================================================================== - -#if FFX_HALF - -FfxFloat16x4 SpdReduceQuadH(FfxFloat16x4 v) -{ -#if defined(FFX_GLSL) && !defined(SPD_NO_WAVE_OPERATIONS) - FfxFloat16x4 v0 = v; - FfxFloat16x4 v1 = subgroupQuadSwapHorizontal(v); - FfxFloat16x4 v2 = subgroupQuadSwapVertical(v); - FfxFloat16x4 v3 = subgroupQuadSwapDiagonal(v); - return SpdReduce4H(v0, v1, v2, v3); -#elif defined(FFX_HLSL) && !defined(SPD_NO_WAVE_OPERATIONS) - // requires SM6.0 - FfxUInt32 quad = WaveGetLaneIndex() & (~0x3); - FfxFloat16x4 v0 = v; - FfxFloat16x4 v1 = WaveReadLaneAt(v, quad | 1); - FfxFloat16x4 v2 = WaveReadLaneAt(v, quad | 2); - FfxFloat16x4 v3 = WaveReadLaneAt(v, quad | 3); - return SpdReduce4H(v0, v1, v2, v3); -/* - // if SM6.0 is not available, you can use the AMD shader intrinsics - // the AMD shader intrinsics are available in AMD GPU Services (AGS) library: - // https://gpuopen.com/amd-gpu-services-ags-library/ - // works for DX11 - FfxFloat16x4 v0 = v; - FfxFloat16x4 v1; - v1.x = AmdExtD3DShaderIntrinsics_SwizzleF(v.x, AmdExtD3DShaderIntrinsicsSwizzle_SwapX1); - v1.y = AmdExtD3DShaderIntrinsics_SwizzleF(v.y, AmdExtD3DShaderIntrinsicsSwizzle_SwapX1); - v1.z = AmdExtD3DShaderIntrinsics_SwizzleF(v.z, AmdExtD3DShaderIntrinsicsSwizzle_SwapX1); - v1.w = AmdExtD3DShaderIntrinsics_SwizzleF(v.w, AmdExtD3DShaderIntrinsicsSwizzle_SwapX1); - FfxFloat16x4 v2; - v2.x = AmdExtD3DShaderIntrinsics_SwizzleF(v.x, AmdExtD3DShaderIntrinsicsSwizzle_SwapX2); - v2.y = AmdExtD3DShaderIntrinsics_SwizzleF(v.y, AmdExtD3DShaderIntrinsicsSwizzle_SwapX2); - v2.z = AmdExtD3DShaderIntrinsics_SwizzleF(v.z, AmdExtD3DShaderIntrinsicsSwizzle_SwapX2); - v2.w = AmdExtD3DShaderIntrinsics_SwizzleF(v.w, AmdExtD3DShaderIntrinsicsSwizzle_SwapX2); - FfxFloat16x4 v3; - v3.x = AmdExtD3DShaderIntrinsics_SwizzleF(v.x, AmdExtD3DShaderIntrinsicsSwizzle_ReverseX4); - v3.y = AmdExtD3DShaderIntrinsics_SwizzleF(v.y, AmdExtD3DShaderIntrinsicsSwizzle_ReverseX4); - v3.z = AmdExtD3DShaderIntrinsics_SwizzleF(v.z, AmdExtD3DShaderIntrinsicsSwizzle_ReverseX4); - v3.w = AmdExtD3DShaderIntrinsics_SwizzleF(v.w, AmdExtD3DShaderIntrinsicsSwizzle_ReverseX4); - return SpdReduce4H(v0, v1, v2, v3); - */ -#endif - return FfxFloat16x4(0.0, 0.0, 0.0, 0.0); -} - -FfxFloat16x4 SpdReduceIntermediateH(FfxUInt32x2 i0, FfxUInt32x2 i1, FfxUInt32x2 i2, FfxUInt32x2 i3) -{ - FfxFloat16x4 v0 = SpdLoadIntermediateH(i0.x, i0.y); - FfxFloat16x4 v1 = SpdLoadIntermediateH(i1.x, i1.y); - FfxFloat16x4 v2 = SpdLoadIntermediateH(i2.x, i2.y); - FfxFloat16x4 v3 = SpdLoadIntermediateH(i3.x, i3.y); - return SpdReduce4H(v0, v1, v2, v3); -} - -FfxFloat16x4 SpdReduceLoad4H(FfxUInt32x2 i0, FfxUInt32x2 i1, FfxUInt32x2 i2, FfxUInt32x2 i3, FfxUInt32 slice) -{ - FfxFloat16x4 v0 = SpdLoadH(FfxInt32x2(i0), slice); - FfxFloat16x4 v1 = SpdLoadH(FfxInt32x2(i1), slice); - FfxFloat16x4 v2 = SpdLoadH(FfxInt32x2(i2), slice); - FfxFloat16x4 v3 = SpdLoadH(FfxInt32x2(i3), slice); - return SpdReduce4H(v0, v1, v2, v3); -} - -FfxFloat16x4 SpdReduceLoad4H(FfxUInt32x2 base, FfxUInt32 slice) -{ - return SpdReduceLoad4H(FfxUInt32x2(base + FfxUInt32x2(0, 0)), FfxUInt32x2(base + FfxUInt32x2(0, 1)), FfxUInt32x2(base + FfxUInt32x2(1, 0)), FfxUInt32x2(base + FfxUInt32x2(1, 1)), slice); -} - -FfxFloat16x4 SpdReduceLoadSourceImage4H(FfxUInt32x2 i0, FfxUInt32x2 i1, FfxUInt32x2 i2, FfxUInt32x2 i3, FfxUInt32 slice) -{ - FfxFloat16x4 v0 = SpdLoadSourceImageH(FfxInt32x2(i0), slice); - FfxFloat16x4 v1 = SpdLoadSourceImageH(FfxInt32x2(i1), slice); - FfxFloat16x4 v2 = SpdLoadSourceImageH(FfxInt32x2(i2), slice); - FfxFloat16x4 v3 = SpdLoadSourceImageH(FfxInt32x2(i3), slice); - return SpdReduce4H(v0, v1, v2, v3); -} - -FfxFloat16x4 SpdReduceLoadSourceImageH(FfxUInt32x2 base, FfxUInt32 slice) -{ -#ifdef SPD_LINEAR_SAMPLER - return SpdLoadSourceImageH(FfxInt32x2(base), slice); -#else - return SpdReduceLoadSourceImage4H(FfxUInt32x2(base + FfxUInt32x2(0, 0)), FfxUInt32x2(base + FfxUInt32x2(0, 1)), FfxUInt32x2(base + FfxUInt32x2(1, 0)), FfxUInt32x2(base + FfxUInt32x2(1, 1)), slice); -#endif -} - -void SpdDownsampleMips_0_1_IntrinsicsH(FfxUInt32 x, FfxUInt32 y, FfxUInt32x2 workGroupID, FfxUInt32 localInvocationIndex, FfxUInt32 mips, FfxUInt32 slice) -{ - FfxFloat16x4 v[4]; - - FfxInt32x2 tex = FfxInt32x2(workGroupID.xy * 64) + FfxInt32x2(x * 2, y * 2); - FfxInt32x2 pix = FfxInt32x2(workGroupID.xy * 32) + FfxInt32x2(x, y); - v[0] = SpdReduceLoadSourceImageH(tex, slice); - SpdStoreH(pix, v[0], 0, slice); - - tex = FfxInt32x2(workGroupID.xy * 64) + FfxInt32x2(x * 2 + 32, y * 2); - pix = FfxInt32x2(workGroupID.xy * 32) + FfxInt32x2(x + 16, y); - v[1] = SpdReduceLoadSourceImageH(tex, slice); - SpdStoreH(pix, v[1], 0, slice); - - tex = FfxInt32x2(workGroupID.xy * 64) + FfxInt32x2(x * 2, y * 2 + 32); - pix = FfxInt32x2(workGroupID.xy * 32) + FfxInt32x2(x, y + 16); - v[2] = SpdReduceLoadSourceImageH(tex, slice); - SpdStoreH(pix, v[2], 0, slice); - - tex = FfxInt32x2(workGroupID.xy * 64) + FfxInt32x2(x * 2 + 32, y * 2 + 32); - pix = FfxInt32x2(workGroupID.xy * 32) + FfxInt32x2(x + 16, y + 16); - v[3] = SpdReduceLoadSourceImageH(tex, slice); - SpdStoreH(pix, v[3], 0, slice); - - if (mips <= 1) - return; - - v[0] = SpdReduceQuadH(v[0]); - v[1] = SpdReduceQuadH(v[1]); - v[2] = SpdReduceQuadH(v[2]); - v[3] = SpdReduceQuadH(v[3]); - - if ((localInvocationIndex % 4) == 0) - { - SpdStoreH(FfxInt32x2(workGroupID.xy * 16) + FfxInt32x2(x / 2, y / 2), v[0], 1, slice); - SpdStoreIntermediateH(x / 2, y / 2, v[0]); - - SpdStoreH(FfxInt32x2(workGroupID.xy * 16) + FfxInt32x2(x / 2 + 8, y / 2), v[1], 1, slice); - SpdStoreIntermediateH(x / 2 + 8, y / 2, v[1]); - - SpdStoreH(FfxInt32x2(workGroupID.xy * 16) + FfxInt32x2(x / 2, y / 2 + 8), v[2], 1, slice); - SpdStoreIntermediateH(x / 2, y / 2 + 8, v[2]); - - SpdStoreH(FfxInt32x2(workGroupID.xy * 16) + FfxInt32x2(x / 2 + 8, y / 2 + 8), v[3], 1, slice); - SpdStoreIntermediateH(x / 2 + 8, y / 2 + 8, v[3]); - } -} - -void SpdDownsampleMips_0_1_LDSH(FfxUInt32 x, FfxUInt32 y, FfxUInt32x2 workGroupID, FfxUInt32 localInvocationIndex, FfxUInt32 mips, FfxUInt32 slice) -{ - FfxFloat16x4 v[4]; - - FfxInt32x2 tex = FfxInt32x2(workGroupID.xy * 64) + FfxInt32x2(x * 2, y * 2); - FfxInt32x2 pix = FfxInt32x2(workGroupID.xy * 32) + FfxInt32x2(x, y); - v[0] = SpdReduceLoadSourceImageH(tex, slice); - SpdStoreH(pix, v[0], 0, slice); - - tex = FfxInt32x2(workGroupID.xy * 64) + FfxInt32x2(x * 2 + 32, y * 2); - pix = FfxInt32x2(workGroupID.xy * 32) + FfxInt32x2(x + 16, y); - v[1] = SpdReduceLoadSourceImageH(tex, slice); - SpdStoreH(pix, v[1], 0, slice); - - tex = FfxInt32x2(workGroupID.xy * 64) + FfxInt32x2(x * 2, y * 2 + 32); - pix = FfxInt32x2(workGroupID.xy * 32) + FfxInt32x2(x, y + 16); - v[2] = SpdReduceLoadSourceImageH(tex, slice); - SpdStoreH(pix, v[2], 0, slice); - - tex = FfxInt32x2(workGroupID.xy * 64) + FfxInt32x2(x * 2 + 32, y * 2 + 32); - pix = FfxInt32x2(workGroupID.xy * 32) + FfxInt32x2(x + 16, y + 16); - v[3] = SpdReduceLoadSourceImageH(tex, slice); - SpdStoreH(pix, v[3], 0, slice); - - if (mips <= 1) - return; - - for (FfxInt32 i = 0; i < 4; i++) - { - SpdStoreIntermediateH(x, y, v[i]); - SpdWorkgroupShuffleBarrier(); - if (localInvocationIndex < 64) - { - v[i] = SpdReduceIntermediateH(FfxUInt32x2(x * 2 + 0, y * 2 + 0), FfxUInt32x2(x * 2 + 1, y * 2 + 0), FfxUInt32x2(x * 2 + 0, y * 2 + 1), FfxUInt32x2(x * 2 + 1, y * 2 + 1)); - SpdStoreH(FfxInt32x2(workGroupID.xy * 16) + FfxInt32x2(x + (i % 2) * 8, y + (i / 2) * 8), v[i], 1, slice); - } - SpdWorkgroupShuffleBarrier(); - } - - if (localInvocationIndex < 64) - { - SpdStoreIntermediateH(x + 0, y + 0, v[0]); - SpdStoreIntermediateH(x + 8, y + 0, v[1]); - SpdStoreIntermediateH(x + 0, y + 8, v[2]); - SpdStoreIntermediateH(x + 8, y + 8, v[3]); - } -} - -void SpdDownsampleMips_0_1H(FfxUInt32 x, FfxUInt32 y, FfxUInt32x2 workGroupID, FfxUInt32 localInvocationIndex, FfxUInt32 mips, FfxUInt32 slice) -{ -#ifdef SPD_NO_WAVE_OPERATIONS - SpdDownsampleMips_0_1_LDSH(x, y, workGroupID, localInvocationIndex, mips, slice); -#else - SpdDownsampleMips_0_1_IntrinsicsH(x, y, workGroupID, localInvocationIndex, mips, slice); -#endif -} - - -void SpdDownsampleMip_2H(FfxUInt32 x, FfxUInt32 y, FfxUInt32x2 workGroupID, FfxUInt32 localInvocationIndex, FfxUInt32 mip, FfxUInt32 slice) -{ -#ifdef SPD_NO_WAVE_OPERATIONS - if (localInvocationIndex < 64) - { - FfxFloat16x4 v = SpdReduceIntermediateH(FfxUInt32x2(x * 2 + 0, y * 2 + 0), FfxUInt32x2(x * 2 + 1, y * 2 + 0), FfxUInt32x2(x * 2 + 0, y * 2 + 1), FfxUInt32x2(x * 2 + 1, y * 2 + 1)); - SpdStoreH(FfxInt32x2(workGroupID.xy * 8) + FfxInt32x2(x, y), v, mip, slice); - // store to LDS, try to reduce bank conflicts - // x 0 x 0 x 0 x 0 x 0 x 0 x 0 x 0 - // 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 - // 0 x 0 x 0 x 0 x 0 x 0 x 0 x 0 x - // 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 - // x 0 x 0 x 0 x 0 x 0 x 0 x 0 x 0 - // ... - // x 0 x 0 x 0 x 0 x 0 x 0 x 0 x 0 - SpdStoreIntermediateH(x * 2 + y % 2, y * 2, v); - } -#else - FfxFloat16x4 v = SpdLoadIntermediateH(x, y); - v = SpdReduceQuadH(v); - // quad index 0 stores result - if (localInvocationIndex % 4 == 0) - { - SpdStoreH(FfxInt32x2(workGroupID.xy * 8) + FfxInt32x2(x / 2, y / 2), v, mip, slice); - SpdStoreIntermediateH(x + (y / 2) % 2, y, v); - } -#endif -} - -void SpdDownsampleMip_3H(FfxUInt32 x, FfxUInt32 y, FfxUInt32x2 workGroupID, FfxUInt32 localInvocationIndex, FfxUInt32 mip, FfxUInt32 slice) -{ -#ifdef SPD_NO_WAVE_OPERATIONS - if (localInvocationIndex < 16) - { - // x 0 x 0 - // 0 0 0 0 - // 0 x 0 x - // 0 0 0 0 - FfxFloat16x4 v = - SpdReduceIntermediateH(FfxUInt32x2(x * 4 + 0 + 0, y * 4 + 0), FfxUInt32x2(x * 4 + 2 + 0, y * 4 + 0), FfxUInt32x2(x * 4 + 0 + 1, y * 4 + 2), FfxUInt32x2(x * 4 + 2 + 1, y * 4 + 2)); - SpdStoreH(FfxInt32x2(workGroupID.xy * 4) + FfxInt32x2(x, y), v, mip, slice); - // store to LDS - // x 0 0 0 x 0 0 0 x 0 0 0 x 0 0 0 - // 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 - // 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 - // 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 - // 0 x 0 0 0 x 0 0 0 x 0 0 0 x 0 0 - // ... - // 0 0 x 0 0 0 x 0 0 0 x 0 0 0 x 0 - // ... - // 0 0 0 x 0 0 0 x 0 0 0 x 0 0 0 x - // ... - SpdStoreIntermediateH(x * 4 + y, y * 4, v); - } -#else - if (localInvocationIndex < 64) - { - FfxFloat16x4 v = SpdLoadIntermediateH(x * 2 + y % 2, y * 2); - v = SpdReduceQuadH(v); - // quad index 0 stores result - if (localInvocationIndex % 4 == 0) - { - SpdStoreH(FfxInt32x2(workGroupID.xy * 4) + FfxInt32x2(x / 2, y / 2), v, mip, slice); - SpdStoreIntermediateH(x * 2 + y / 2, y * 2, v); - } - } -#endif -} - -void SpdDownsampleMip_4H(FfxUInt32 x, FfxUInt32 y, FfxUInt32x2 workGroupID, FfxUInt32 localInvocationIndex, FfxUInt32 mip, FfxUInt32 slice) -{ -#ifdef SPD_NO_WAVE_OPERATIONS - if (localInvocationIndex < 4) - { - // x 0 0 0 x 0 0 0 - // ... - // 0 x 0 0 0 x 0 0 - FfxFloat16x4 v = SpdReduceIntermediateH(FfxUInt32x2(x * 8 + 0 + 0 + y * 2, y * 8 + 0), - FfxUInt32x2(x * 8 + 4 + 0 + y * 2, y * 8 + 0), - FfxUInt32x2(x * 8 + 0 + 1 + y * 2, y * 8 + 4), - FfxUInt32x2(x * 8 + 4 + 1 + y * 2, y * 8 + 4)); - SpdStoreH(FfxInt32x2(workGroupID.xy * 2) + FfxInt32x2(x, y), v, mip, slice); - // store to LDS - // x x x x 0 ... - // 0 ... - SpdStoreIntermediateH(x + y * 2, 0, v); - } -#else - if (localInvocationIndex < 16) - { - FfxFloat16x4 v = SpdLoadIntermediateH(x * 4 + y, y * 4); - v = SpdReduceQuadH(v); - // quad index 0 stores result - if (localInvocationIndex % 4 == 0) - { - SpdStoreH(FfxInt32x2(workGroupID.xy * 2) + FfxInt32x2(x / 2, y / 2), v, mip, slice); - SpdStoreIntermediateH(x / 2 + y, 0, v); - } - } -#endif -} - -void SpdDownsampleMip_5H(FfxUInt32x2 workGroupID, FfxUInt32 localInvocationIndex, FfxUInt32 mip, FfxUInt32 slice) -{ -#ifdef SPD_NO_WAVE_OPERATIONS - if (localInvocationIndex < 1) - { - // x x x x 0 ... - // 0 ... - FfxFloat16x4 v = SpdReduceIntermediateH(FfxUInt32x2(0, 0), FfxUInt32x2(1, 0), FfxUInt32x2(2, 0), FfxUInt32x2(3, 0)); - SpdStoreH(FfxInt32x2(workGroupID.xy), v, mip, slice); - } -#else - if (localInvocationIndex < 4) - { - FfxFloat16x4 v = SpdLoadIntermediateH(localInvocationIndex, 0); - v = SpdReduceQuadH(v); - // quad index 0 stores result - if (localInvocationIndex % 4 == 0) - { - SpdStoreH(FfxInt32x2(workGroupID.xy), v, mip, slice); - } - } -#endif -} - -void SpdDownsampleMips_6_7H(FfxUInt32 x, FfxUInt32 y, FfxUInt32 mips, FfxUInt32 slice) -{ - FfxInt32x2 tex = FfxInt32x2(x * 4 + 0, y * 4 + 0); - FfxInt32x2 pix = FfxInt32x2(x * 2 + 0, y * 2 + 0); - FfxFloat16x4 v0 = SpdReduceLoad4H(tex, slice); - SpdStoreH(pix, v0, 6, slice); - - tex = FfxInt32x2(x * 4 + 2, y * 4 + 0); - pix = FfxInt32x2(x * 2 + 1, y * 2 + 0); - FfxFloat16x4 v1 = SpdReduceLoad4H(tex, slice); - SpdStoreH(pix, v1, 6, slice); - - tex = FfxInt32x2(x * 4 + 0, y * 4 + 2); - pix = FfxInt32x2(x * 2 + 0, y * 2 + 1); - FfxFloat16x4 v2 = SpdReduceLoad4H(tex, slice); - SpdStoreH(pix, v2, 6, slice); - - tex = FfxInt32x2(x * 4 + 2, y * 4 + 2); - pix = FfxInt32x2(x * 2 + 1, y * 2 + 1); - FfxFloat16x4 v3 = SpdReduceLoad4H(tex, slice); - SpdStoreH(pix, v3, 6, slice); - - if (mips < 8) - return; - // no barrier needed, working on values only from the same thread - - FfxFloat16x4 v = SpdReduce4H(v0, v1, v2, v3); - SpdStoreH(FfxInt32x2(x, y), v, 7, slice); - SpdStoreIntermediateH(x, y, v); -} - -void SpdDownsampleNextFourH(FfxUInt32 x, FfxUInt32 y, FfxUInt32x2 workGroupID, FfxUInt32 localInvocationIndex, FfxUInt32 baseMip, FfxUInt32 mips, FfxUInt32 slice) -{ - if (mips <= baseMip) - return; - SpdWorkgroupShuffleBarrier(); - SpdDownsampleMip_2H(x, y, workGroupID, localInvocationIndex, baseMip, slice); - - if (mips <= baseMip + 1) - return; - SpdWorkgroupShuffleBarrier(); - SpdDownsampleMip_3H(x, y, workGroupID, localInvocationIndex, baseMip + 1, slice); - - if (mips <= baseMip + 2) - return; - SpdWorkgroupShuffleBarrier(); - SpdDownsampleMip_4H(x, y, workGroupID, localInvocationIndex, baseMip + 2, slice); - - if (mips <= baseMip + 3) - return; - SpdWorkgroupShuffleBarrier(); - SpdDownsampleMip_5H(workGroupID, localInvocationIndex, baseMip + 3, slice); -} - -void SpdDownsampleH(FfxUInt32x2 workGroupID, FfxUInt32 localInvocationIndex, FfxUInt32 mips, FfxUInt32 numWorkGroups, FfxUInt32 slice) -{ - FfxUInt32x2 sub_xy = ffxRemapForWaveReduction(localInvocationIndex % 64); - FfxUInt32 x = sub_xy.x + 8 * ((localInvocationIndex >> 6) % 2); - FfxUInt32 y = sub_xy.y + 8 * ((localInvocationIndex >> 7)); - - SpdDownsampleMips_0_1H(x, y, workGroupID, localInvocationIndex, mips, slice); - - SpdDownsampleNextFourH(x, y, workGroupID, localInvocationIndex, 2, mips, slice); - - if (mips < 7) - return; - - if (SpdExitWorkgroup(numWorkGroups, localInvocationIndex, slice)) - return; - - SpdResetAtomicCounter(slice); - - // After mip 6 there is only a single workgroup left that downsamples the remaining up to 64x64 texels. - SpdDownsampleMips_6_7H(x, y, mips, slice); - - SpdDownsampleNextFourH(x, y, FfxUInt32x2(0, 0), localInvocationIndex, 8, mips, slice); -} - -void SpdDownsampleH(FfxUInt32x2 workGroupID, FfxUInt32 localInvocationIndex, FfxUInt32 mips, FfxUInt32 numWorkGroups, FfxUInt32 slice, FfxUInt32x2 workGroupOffset) -{ - SpdDownsampleH(workGroupID + workGroupOffset, localInvocationIndex, mips, numWorkGroups, slice); -} - -#endif // #if FFX_HALF -#endif // #ifdef FFX_GPU diff --git a/Assets/Resources/FSR2/shaders/ffx_spd.h.meta b/Assets/Resources/FSR2/shaders/ffx_spd.h.meta deleted file mode 100644 index d896fc1..0000000 --- a/Assets/Resources/FSR2/shaders/ffx_spd.h.meta +++ /dev/null @@ -1,60 +0,0 @@ -fileFormatVersion: 2 -guid: 3ef69a900a925bb498651c10581e0979 -PluginImporter: - externalObjects: {} - serializedVersion: 2 - iconMap: {} - executionOrder: {} - defineConstraints: [] - isPreloaded: 0 - isOverridable: 0 - isExplicitlyReferenced: 0 - validateReferences: 1 - platformData: - - first: - : Any - second: - enabled: 0 - settings: - Exclude Editor: 1 - Exclude GameCoreScarlett: 1 - Exclude GameCoreXboxOne: 1 - Exclude Linux64: 1 - Exclude OSXUniversal: 1 - Exclude PS4: 1 - Exclude PS5: 1 - Exclude WebGL: 1 - Exclude Win: 1 - Exclude Win64: 1 - - first: - Any: - second: - enabled: 0 - settings: {} - - first: - Editor: Editor - second: - enabled: 0 - settings: - DefaultValueInitialized: true - - first: - Standalone: Linux64 - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win - second: - enabled: 0 - settings: - CPU: None - - first: - Standalone: Win64 - second: - enabled: 0 - settings: - CPU: None - userData: - assetBundleName: - assetBundleVariant: