[PATCH] D141649: [DX] Begin adding support for pipeline state

Xiang Li via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Tue Jan 17 08:56:23 PST 2023


python3kgae added inline comments.


================
Comment at: llvm/include/llvm/BinaryFormat/DXContainer.h:203
+  uint8_t SampleFrequency;
+  uint8_t Unused[2];
+
----------------
Why we need Unused here?
If we need 32bit align, VertexPSVInfo need Unused too.


================
Comment at: llvm/include/llvm/BinaryFormat/DXContainer.h:287
+
+  void swapBytes(Triple::EnvironmentType Stage) { StageInfo.swapBytes(Stage); }
+};
----------------
Both swapBytes() and swapBytes(Stage) changed data.
Do we call both of them?


================
Comment at: llvm/test/ObjectYAML/DXContainer/PSVv0-amplification.yaml:28
+      DXILMajorVersion: 0
+      DXILMinorVersion: 1
+      DXILSize:        0
----------------
amplification shader needs shader model 6.5


================
Comment at: llvm/test/ObjectYAML/DXContainer/PSVv0-mesh.yaml:32
+      DXILMajorVersion: 0
+      DXILMinorVersion: 1
+      DXILSize:        0
----------------
mesh shader needs shader model 6.5


================
Comment at: llvm/test/ObjectYAML/DXContainer/PSVv2-domain.yaml:29
+      SigOutputVectors: [ 8, 16, 32, 64 ]
+      NumThreadsX:     512
+      NumThreadsY:     1024
----------------
Only mesh/compute/amplification shader set NumThreadX/Y/Z.


================
Comment at: llvm/test/ObjectYAML/DXContainer/PSVv2-hull.yaml:29
+      SigInputVectors: 64
+      SigOutputVectors: [ 8, 16, 32, 64 ]
+      NumThreadsX:     512
----------------
We cannot have NumThreadsX and Signature at the same time.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D141649/new/

https://reviews.llvm.org/D141649



More information about the llvm-commits mailing list