On Fri, 13 Aug 2021 at 16:56, Conor McCarthy <cmccarthy(a)codeweavers.com> wrote:
--- tests/d3d12.c | 325 ++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 325 insertions(+)
I think it makes sense to split this into two patches, one for each test, as a matter of principle.
+ static const DWORD cs_code[] = + { + /* Compiled with /res_may_alias (but it has no effect on the output from fxc 10.1). */ +#if 0 + struct cb + { + uint value; + }; + ConstantBuffer<cb> c1[] : register(b2, space1); + + Buffer<uint> t1[] : register(t2, space1); + + RWBuffer<uint> u1[] : register(u1, space1); + RWBuffer<uint> u2[] : register(u2, space2); + RWBuffer<uint> u3[] : register(u1, space3); + + [numthreads(64, 1, 1)] + void main(uint id : SV_DispatchThreadID) + { + uint i = c1[NonUniformResourceIndex(id)].value; + if (id < 64) // Workaround for an fxc bug. + { So what is the bug? :) Incidentally, I don't think C99 comments are any better in HLSL either.