// Each #kernel tells which function to compile; you can have many kernels
#pragma kernel CS

// Create a RenderTexture with enableRandomWrite flag and set it with cs.SetTexture
// NOTE: the .hlsl files conveniently contain a list of inputs and outputs in their header comments
//Texture2D<float2> r_exposure; // Expose is not used at the moment, so no need to bind it
Texture2D<float4> r_rcas_input;
RWTexture2D<float4> rw_upscaled_output;

// Some global defines are needed
#define FFX_GPU         // Compiling for GPU
#define FFX_HLSL        // Compile for plain HLSL

//#define FFX_FSR2_OPTION_HDR_COLOR_INPUT     // This ought to be supported in the long run

#define FSR_RCAS_F      // Using 32-bit float mode (not half)
#define FSR2_BIND_UAV_UPSCALED_OUTPUT    // Create an SRV binding for the output we need

#include "ffx_core.h"   // With the above defines this all compiles cleanly, yay!

// This is from ffx_fsr2_callbacks_hlsl.h (bunch of macro use to declare SRVs, may need conversion)
FfxFloat32 Exposure()
{
    // return 1.0f;
    #if defined(FSR2_BIND_SRV_EXPOSURE) || defined(FFX_INTERNAL)
    FfxFloat32 exposure = r_exposure[FfxUInt32x2(0, 0)].x;
    #else
    FfxFloat32 exposure = 1.f;
    #endif

    if (exposure == 0.0f) {
        exposure = 1.0f;
    }

    return exposure;
}

FfxFloat32 PreExposure()
{
    //return fPreExposure;  // TODO: requires constant buffer bindings, uhhh... (see FSR2_BIND_CB_FSR2)
    return 1.0f;
}

void StoreUpscaledOutput(FfxUInt32x2 iPxPos, FfxFloat32x3 fColor)
{
    #if defined(FSR2_BIND_UAV_UPSCALED_OUTPUT) || defined(FFX_INTERNAL)
    rw_upscaled_output[iPxPos] = FfxFloat32x4(fColor * PreExposure(), 1.f);
    #endif
}

//#include "ffx_fsr2_common.h"  // TODO: still depends on a bunch of CB's from callbacks_hlsl

// Below is all converted from ffx_fsr2_rcas_pass.hlsl
uint4 RCASConfig()
{
    // These are not your typical config parameters... these are packed floats-as-ints, i.e. complete nonsense values
    return uint4(1061290752, 974141968, 0, 0);   // TODO: this needs to be a constant buffer value
}

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
void CS(uint3 LocalThreadId : SV_GroupThreadID, uint3 WorkGroupId : SV_GroupID, uint3 Dtid : SV_DispatchThreadID)
{
    RCAS(LocalThreadId, WorkGroupId, Dtid);
}