Abuse SVECTOR padding field to store the normal index, so we don't need to do an extra lookup inside the draw loop.

main.c

@@ -410,26 +410,24 @@ void drawModel(MATRIX *view_matrix, ps1mdl_t *model, ps1skin_t *skin, int frameC
 	gte_SetRotMatrix(&model_mtx);
 	gte_SetTransMatrix(&model_mtx);
 
-	ps1mdl_vertex_t *vin = baseVertex;
-	for (unsigned short vertIdx = 0; vertIdx < vertexCount; ++vertIdx, ++vin)
+	// Scale, translate and convert vertex data to SVECTOR ahead of time for all vertices.
+	// This eliminates duplicate calculations on vertices shared by multiple triangles.
+	SVECTOR *v = vertexBuffer;
+	for (unsigned short vertIdx = 0; vertIdx < vertexCount; ++vertIdx, ++baseVertex, ++v)
 	{
-		SVECTOR *v = &vertexBuffer[vertIdx];
-		setVector(v, SCTR(vin,0), SCTR(vin,1), SCTR(vin,2));
+		setVector(v, SCTR(baseVertex,0), SCTR(baseVertex,1), SCTR(baseVertex,2));
+		v->pad = baseVertex->normalIndex;
 	}
 	
 	for (unsigned short triIdx = 0; triIdx < triangleCount; ++triIdx)
 	{
 		ps1mdl_triangle_t *tri = &model->triangles[triIdx];
 		
-		ps1mdl_vertex_t *v0 = &baseVertex[tri->vertexIndex[0]];
-		ps1mdl_vertex_t *v1 = &baseVertex[tri->vertexIndex[1]];
-		ps1mdl_vertex_t *v2 = &baseVertex[tri->vertexIndex[2]];
+		SVECTOR *v0 = &vertexBuffer[tri->vertexIndex[0]];
+		SVECTOR *v1 = &vertexBuffer[tri->vertexIndex[1]];
+		SVECTOR *v2 = &vertexBuffer[tri->vertexIndex[2]];
 
-		SVECTOR *pos0 = &vertexBuffer[tri->vertexIndex[0]];
-		SVECTOR *pos1 = &vertexBuffer[tri->vertexIndex[1]];
-		SVECTOR *pos2 = &vertexBuffer[tri->vertexIndex[2]];
-
-		gte_ldv3(pos0, pos1, pos2);
+		gte_ldv3(v0, v1, v2);
 		gte_rtpt();	// Rotation, Translation and Perspective triplet (all three vertices at once)
 
 		// Normal clipping for backface culling
@@ -480,7 +478,7 @@ void drawModel(MATRIX *view_matrix, ps1mdl_t *model, ps1skin_t *skin, int frameC
 		// Calculate vertex color based on normal
 		// TODO: we could probably speed this up by precalculating the lighting for each normal, if we need to draw more than 162 vertices
 		gte_ldrgb(&poly->r0);
-		gte_ldv3(anorms[v0->normalIndex], anorms[v1->normalIndex], anorms[v2->normalIndex]);
+		gte_ldv3(anorms[v0->pad], anorms[v1->pad], anorms[v2->pad]);
 		gte_nct();
 		gte_strgb3(&poly->r0, &poly->r1, &poly->r2);