[llvm] [AMDGPU] Implement isSDNodeAlwaysUniform for INTRINSIC_W_CHAIN (PR #110114)

via llvm-commits llvm-commits at lists.llvm.org
Thu Sep 26 05:02:42 PDT 2024


llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-backend-amdgpu

Author: Jay Foad (jayfoad)

<details>
<summary>Changes</summary>

There are no always uniform side-effecting intrinsics upstream to test
this with, but we have examples downstream.


---
Full diff: https://github.com/llvm/llvm-project/pull/110114.diff


1 Files Affected:

- (modified) llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp (+4) 


``````````diff
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
index fad51ce8285e01..94fdf4effa10a1 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
@@ -900,6 +900,10 @@ bool AMDGPUTargetLowering::isSDNodeAlwaysUniform(const SDNode *N) const {
     unsigned IntrID = N->getConstantOperandVal(0);
     return AMDGPU::isIntrinsicAlwaysUniform(IntrID);
   }
+  case ISD::INTRINSIC_W_CHAIN: {
+    unsigned IntrID = N->getConstantOperandVal(1);
+    return AMDGPU::isIntrinsicAlwaysUniform(IntrID);
+  }
   case ISD::LOAD:
     if (cast<LoadSDNode>(N)->getMemOperand()->getAddrSpace() ==
         AMDGPUAS::CONSTANT_ADDRESS_32BIT)

``````````

</details>


https://github.com/llvm/llvm-project/pull/110114


More information about the llvm-commits mailing list