[Mlir-commits] [mlir] [mlir][ArithToAMDGPU] limit scaling truncf/extf support to gfx950 (PR #155431)

[llvmlistbot at llvm.org]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-backend-amdgpu

Author: Muzammil (Muzammiluddin-Syed-ECE)

<details>
<summary>Changes</summary>

The current chip guard fails to prevent scaling_extf/truncf patterns from being applied on gfx1100 which does not have scaling support.

---
Full diff: https://github.com/llvm/llvm-project/pull/155431.diff


1 Files Affected:

- (modified) mlir/lib/Conversion/ArithToAMDGPU/ArithToAMDGPU.cpp (+1-1) 


``````````diff
diff --git a/mlir/lib/Conversion/ArithToAMDGPU/ArithToAMDGPU.cpp b/mlir/lib/Conversion/ArithToAMDGPU/ArithToAMDGPU.cpp
index 8230591123661..c0c063b4d4923 100644
--- a/mlir/lib/Conversion/ArithToAMDGPU/ArithToAMDGPU.cpp
+++ b/mlir/lib/Conversion/ArithToAMDGPU/ArithToAMDGPU.cpp
@@ -702,7 +702,7 @@ void mlir::arith::populateArithToAMDGPUConversionPatterns(
   if (allowPackedF16Rtz)
     patterns.add<TruncfToFloat16RewritePattern>(patterns.getContext(), benefit);
 
-  if (chipset >= kGfx950) {
+  if (chipset == kGfx950) {
     patterns.add<ScalingExtFRewritePattern>(patterns.getContext(), benefit);
     patterns.add<ScalingTruncFRewritePattern>(patterns.getContext(), benefit);
   }

``````````

</details>


https://github.com/llvm/llvm-project/pull/155431