Don't set RaytracingTier1_1 based on subobjects; enable flag validation #6320
Conversation
In validator version 1.7 and below, the RaytracingTier1_1 module flag was set on every function if any StateObjectConfig subobject had the AllowStateObjectAdditions flag set. This was incorrect, as the requirement is validated in the runtime based on the use of the subobject instead. Subobjects are removed from the module and placed in RDAT during container serialization, so the requirement would be lost when recomputing the flags in validation. This didn't break validation because flag validation was completely disabled for libraries! This change fixes this problem, and allows DxilValidation to validate ShaderFlags because they will no longer mismatch due to this issue. Running CollectShaderFlagsForModule is also necessary for collecting per-function shader flags which will be used by call graph validation in a subsequent change, so enabling flag validation unblocks that as well.
tex3d
force-pushed
the
no-rt11-flag-from-subobject
branch
from
8a7b441 to
06ab69a
Compare
...clang/test/HLSLFileCheck/shader_targets/raytracing/subobjects_raytracingPipelineConfig1.hlsl
Outdated
Show resolved
Hide resolved
There was a problem hiding this comment.
I think it would be helpful to acknowledge #6207 and the changes it made to this code. The description refers to 1.7 behavior, but is actually removing code that was added by that change which confused me at first.
In particular, I'd appreciate an explanation of why any consideration of the subobjects isn't needed to determine whether raytracing 1.1 is required. After this change, it seems that the only way to set the raytracing 1.1 flag is if these opcodes are used:
DirectXShaderCompiler/lib/DXIL/DxilShaderFlags.cpp
Lines 634 to 636 in df588be
That doesn't sound like checking for the use of subobjects to me and I think there are other ops that make use of them?
The code was basically moved, not exactly added. I added a note to issue #6321. Let me know if that helps regarding the churn of that code. That issue should also explain why this flag should not be set from the subobject, since for subobjects the feature requirement is checked independently in the runtime, and the subobjects are part of RDAT, and not part of the module. There are multiple things that could require raytracing 1.1 though. As you see based on the link, a function could use a DXR 1.1 feature directly, in which case, the function that uses it will reflect the raytracing 1.1 requirement flag, and this requirement will bubble up to the module flags as well. |
…on (microsoft#6320) In validator version 1.7 and below, the RaytracingTier1_1 module flag was set on every function if any StateObjectConfig subobject had the AllowStateObjectAdditions flag set. This was incorrect, as the requirement is validated in the runtime based on the use of the subobject instead. Subobjects are removed from the module and placed in RDAT during container serialization, so the requirement would be lost when recomputing the flags in validation. This didn't break validation because flag validation was completely disabled for libraries! This change fixes this problem, and allows DxilValidation to validate ShaderFlags because they will no longer mismatch due to this issue. Running CollectShaderFlagsForModule is also necessary for collecting per-function shader flags which will be used by call graph validation in a subsequent change, so enabling flag validation unblocks that as well. Fixes microsoft#6321 (cherry picked from commit fafbc42)
…on (#6320) (#6322) In validator version 1.7 and below, the RaytracingTier1_1 module flag was set on every function if any StateObjectConfig subobject had the AllowStateObjectAdditions flag set. This was incorrect, as the requirement is validated in the runtime based on the use of the subobject instead. Subobjects are removed from the module and placed in RDAT during container serialization, so the requirement would be lost when recomputing the flags in validation. This didn't break validation because flag validation was completely disabled for libraries! This change fixes this problem, and allows DxilValidation to validate ShaderFlags because they will no longer mismatch due to this issue. Running CollectShaderFlagsForModule is also necessary for collecting per-function shader flags which will be used by call graph validation in a subsequent change, so enabling flag validation unblocks that as well. Fixes #6321 (cherry picked from commit fafbc42)
In validator version 1.7 and below, the RaytracingTier1_1 module flag was set on every function if any StateObjectConfig subobject had the AllowStateObjectAdditions flag set. This was incorrect, as the requirement is validated in the runtime based on the use of the subobject instead. Subobjects are removed from the module and placed in RDAT during container serialization, so the requirement would be lost when recomputing the flags in validation. This didn't break validation because flag validation was completely disabled for libraries!
This change fixes this problem, and allows DxilValidation to validate ShaderFlags because they will no longer mismatch due to this issue. Running CollectShaderFlagsForModule is also necessary for collecting per-function shader flags which will be used by call graph validation in a subsequent change, so enabling flag validation unblocks that as well.
Fixes #6321