Tex Riddell

Technically, HLSL argument passing is all by-value, meaning copy-in (and copy-out for inout/out args). This means that if you pass a structure from the StructuredBuffer to the function, it's loading...

Test failures are due to our preservation of value names in debug builds.

I have a PR #4452 which fixes related issue #4421. That may fix this issue as well, but I haven't had a chance to construct/try this repro. Artifacts for the...

There is already a supported attribute for DXIL library targets, like: `[shader("pixel")]`, for identifying the entry functions in a library. The missing piece is producing an array of normal (non-library)...

Generally, in DXC, uniform parameters to entry points are not supported this way. This was a change from FXC. However, DXC doesn't provide good error reporting for this case in...

Note: you can do this from a filecheck test more easily. See example for compilation + linking in `tools/clang/test/HLSLFileCheck/shader_targets/library/link_mismatch_target.hlsl`. It basically involves compiling with `-Fo ` to capture to virtual...

The intention is to change to match StructuredBuffer layout for the new cbufferLoad. However, there are various known bugs when enabling this option, and there's a PR that needs a...

I think this would make the most sense as a rewriter mode. One that could pre-assign registers to resources and emit a resource binding table define for use with the...

I think the root cause of this issue is that we generate DXIL op expansion log2+mul+exp (since we had no constant evaluation of the HL operation during CodeGen), and when...

If you think is a DXC compiler issue, can we get an HLSL repro that shows the bad DXIL output?