Vertex Loader: SSE implementations of more position/texcoord/normal formats

~35-45% faster NFS:HP2, possibly other vertex-bound games.
2014-11-11 01:48:38 -08:00
parent da962a3d2b
commit 3ddf82a318
7 changed files with 299 additions and 106 deletions
@@ -53,7 +53,8 @@ u32 VertexManager::GetRemainingSize()

 void VertexManager::PrepareForAdditionalData(int primitive, u32 count, u32 stride)
 {
-	u32 const needed_vertex_bytes = count * stride;
+	// The SSE vertex loader can write up to 4 bytes past the end
+	u32 const needed_vertex_bytes = count * stride + 4;

 	// We can't merge different kinds of primitives, so we have to flush here
 	if (current_primitive_type != primitive_from_gx[primitive])