Some micro-optimizations in the face drawing code. The PS1 really does favor simple counters over pointer arithmetic.

draw.h

@@ -170,20 +170,25 @@ static INLINE void draw_quadstrip_textured(STVECTOR *verts, u_char numVerts, u_s
         return;
 
     // Draw the face as a quad strip
-    const STVECTOR *v0, *v1, *v2, *v3;
-    const STVECTOR *head = verts;
-    const STVECTOR *tail = verts + numVerts;
+    u_char i0, i1, i2, i3;
+    u_char head = 0;
+    u_char tail = numVerts;
 
     // Initialize the first two vertices
-    v2 = --tail;
-    v3 = head++;
+    i2 = --tail;
+    i3 = head++;
 
+    STVECTOR *v0, *v1, *v2, *v3;
     for (u_char quadIdx = 0; quadIdx < numQuads; ++quadIdx)
     {
-        v0 = v2;
-        v1 = v3;
-        v2 = --tail;
-        v3 = head++;
+        i0 = i2;
+        i1 = i3;
+        i2 = --tail;
+        i3 = head++;
+
+        v0 = &verts[i0];
+        v1 = &verts[i1];
+        v2 = &verts[i2];
 
         // Naively draw the quad with GTE, nothing special or optimized about this
         gte_ldv3(v0, v1, v2);
@@ -194,7 +199,7 @@ static INLINE void draw_quadstrip_textured(STVECTOR *verts, u_char numVerts, u_s
         gte_stotz(&p);
         short depth = p >> 2;
         if (depth <= 0 || depth >= OTLEN)
-            continue;
+            return;
 
         // Draw a flat-shaded untextured colored quad
         POLY_GT4 *poly = (POLY_GT4*)mem_prim(sizeof(POLY_GT4));
@@ -203,6 +208,8 @@ static INLINE void draw_quadstrip_textured(STVECTOR *verts, u_char numVerts, u_s
         gte_stsxy1(&poly->x1);
         gte_stsxy2(&poly->x2);
 
+        v3 = &verts[i3];
+
         // Transform the fourth vertex to complete the quad
         gte_ldv0(v3);
         gte_rtps();
@@ -234,20 +241,25 @@ static INLINE void draw_quadstrip_water(STVECTOR *verts, u_char numVerts, u_shor
         return;
 
     // Draw the face as a quad strip
-    const STVECTOR *v0, *v1, *v2, *v3;
-    const STVECTOR *head = verts;
-    const STVECTOR *tail = verts + numVerts;
+    u_char i0, i1, i2, i3;
+    u_char head = 0;
+    u_char tail = numVerts;
 
     // Initialize the first two vertices
-    v2 = --tail;
-    v3 = head++;
+    i2 = --tail;
+    i3 = head++;
 
+    const STVECTOR *v0, *v1, *v2, *v3;
     for (u_char quadIdx = 0; quadIdx < numQuads; ++quadIdx)
     {
-        v0 = v2;
-        v1 = v3;
-        v2 = --tail;
-        v3 = head++;
+        i0 = i2;
+        i1 = i3;
+        i2 = --tail;
+        i3 = head++;
+
+        v0 = &verts[i0];
+        v1 = &verts[i1];
+        v2 = &verts[i2];
 
         // Naively draw the quad with GTE, nothing special or optimized about this
         gte_ldv3(v0, v1, v2);
@@ -267,6 +279,8 @@ static INLINE void draw_quadstrip_water(STVECTOR *verts, u_char numVerts, u_shor
         gte_stsxy1(&poly->x1);
         gte_stsxy2(&poly->x2);
 
+        v3 = &verts[i3];
+
         // Transform the fourth vertex to complete the quad
         gte_ldv0(v3);
         gte_rtps();

world.c

@@ -79,17 +79,17 @@ static INLINE char world_cull_backface(const world_t *world, const ps1bsp_face_t
     cam_vec.vy = face->center.vy - cam_pos.vy;
     cam_vec.vz = face->center.vz - cam_pos.vz;
 
+    // Check if the face's plane points towards the camera
+    const ps1bsp_plane_t *plane = &world->planes[face->planeId];
+    *dot = world_planeDot(&cam_vec, plane);
+    if ((*dot >= 0) ^ face->side)
+        return 1;
+
     // Check if the face is behind the camera 
     // TODO: this may cull faces close to the camera, use VectorNormalS to normalize and check for angles < -60 degrees
     // TODO: or, just check against all corners of the face...
     short camDot = m_dot12(cam_vec, cam_dir);
-    if (camDot < 0)
-        return 1;
-    
-    // Check if the face's plane points towards the camera
-    const ps1bsp_plane_t *plane = &world->planes[face->planeId];
-    *dot = world_planeDot(&cam_vec, plane);
-    return ((*dot >= 0) ^ face->side);
+    return camDot < 0;
 }
 
 static void world_drawface_fast(const world_t *world, const ps1bsp_face_t *face)
@@ -107,7 +107,9 @@ static void world_drawface_fast(const world_t *world, const ps1bsp_face_t *face)
     for (u_char vertIdx = 0; vertIdx < face->numFaceVertices; ++vertIdx, ++faceVertex, ++curVert)
     {
         const ps1bsp_vertex_t *vert = &world->vertices[faceVertex->index];
-        *curVert = *((SVECTOR*)vert);
+        curVert->vx = vert->x;
+        curVert->vy = vert->y;
+        curVert->vz = vert->z;
         curVert->pad = (unsigned short)(size_t)face & 0xFF; // TODO: apply averaged light * color value
     }
 
@@ -135,7 +137,9 @@ static void world_drawface_lit(const world_t *world, const ps1bsp_face_t *face)
         for (u_char vertIdx = 0; vertIdx < poly->numPolyVertices; ++vertIdx, ++polyVertex, ++curVert)
         {
             const ps1bsp_vertex_t *vert = &world->vertices[polyVertex->index];
-            *curVert = *((SVECTOR*)vert);
+            curVert->vx = vert->x;
+            curVert->vy = vert->y;
+            curVert->vz = vert->z;
             curVert->pad = polyVertex->light;
         }
 
@@ -169,7 +173,9 @@ static void world_drawface_textured(const world_t *world, const ps1bsp_face_t *f
         for (u_char vertIdx = 0; vertIdx < poly->numPolyVertices; ++vertIdx, ++polyVertex, ++curVert)
         {
             const ps1bsp_vertex_t *vert = &world->vertices[polyVertex->index];
-            *((SVECTOR*)curVert) = *((SVECTOR*)vert);
+            curVert->vx = vert->x;
+            curVert->vy = vert->y;
+            curVert->vz = vert->z;
             curVert->u = (u_short)polyVertex->u;
             curVert->v = (u_short)polyVertex->v;
             curVert->pad = polyVertex->light;