Re: [PATCH v4 0/5] MR278: vkd3d-shader/dxil: Emit the shader instructions.

[Conor McCarthy (＠cmccarthy)]

On Fri Jul 21 01:25:37 2023 +0000, Henri Verbeet wrote:
> > For the d3d12 and shader runner tests I've been using an environment
> > variable to specify the path to DXC, HLSL source declarations in
> > d3d12.c, and an `[enable]` directive in shader runner. Are those
> > acceptable for upstream?
> > It's of course not ideal to depend on DXC/dxcompiler for the DXIL shader
> > runner tests like that, but I think we can work with that, yes. There's
> > some precedent as well, because we can also use d3dcompiler for running
> > shader runner tests.
> > In terms of the checksum, the most practical way to make that work in
> > the shader runner is probably to introduce a flag like
> > "VKD3D_SHADER_PARSE_DXBC_IGNORE_CHECKSUM" for vkd3d_shader_parse_dxbc(),
> > then use that to parse the DXBCs produced by Linux DXC, and use
> > vkd3d_shader_serialize_dxbc() to create a DXBC blob to pass to
> > d3d12/vkd3d. Also, would we need to use DXC, or could we use the
> > dxcompiler library?
> > Using dxcompiler would be preferable, but instead of COM interfaces it uses C++ linking, and when I tried using it a while ago I couldn't get the calling conventions to match up, even if the C declaration was explicitly declared the same convention as the dxcompiler function. It worked if I added a wrapper in a C++ source file, but we probably don't want that. I'll take another look and see if they can be made to match up.

-- 
https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/278#note_39702