fixup! AMDGPU: Report unaligned scratch access as fast if supported by tgt

macurtis-amd · macurtis-amd · commit 4f8e69e321fc · 2025-09-15T04:27:24.000-05:00
diff --git a/llvm/test/Transforms/AggressiveInstCombine/AMDGPU/fold-consecutive-loads.ll b/llvm/test/Transforms/AggressiveInstCombine/AMDGPU/fold-consecutive-loads.ll
@@ -229,10 +229,6 @@ bb:
   ret <16 x float> %call235
 }
 
-; Function Attrs: convergent nocallback nofree nosync nounwind willreturn memory(none)
 declare <16 x float> @llvm.amdgcn.mfma.f32.32x32x16.fp8.fp8(i64, i64, <16 x float>, i32 immarg, i32 immarg, i32 immarg) #0
 
-; uselistorder directives
-uselistorder ptr @llvm.amdgcn.mfma.f32.32x32x16.fp8.fp8, { 3, 2, 1, 0 }
-
 attributes #0 = { convergent nocallback nofree nosync nounwind willreturn memory(none) }
diff --git a/llvm/test/Transforms/AggressiveInstCombine/AMDGPU/lit.local.cfg b/llvm/test/Transforms/AggressiveInstCombine/AMDGPU/lit.local.cfg
@@ -0,0 +1,2 @@
+if not "AMDGPU" in config.root.targets:
+    config.unsupported = True

Original file line number	Diff line number	Diff line change
`@@ -229,10 +229,6 @@ bb:`
`229`	`229`	`ret <16 x float> %call235`
`230`	`230`	`}`
`231`	`231`
`232`		`-; Function Attrs: convergent nocallback nofree nosync nounwind willreturn memory(none)`
`233`	`232`	`declare <16 x float> @llvm.amdgcn.mfma.f32.32x32x16.fp8.fp8(i64, i64, <16 x float>, i32 immarg, i32 immarg, i32 immarg) #0`
`234`	`233`
`235`		`-; uselistorder directives`
`236`		`-uselistorder ptr @llvm.amdgcn.mfma.f32.32x32x16.fp8.fp8, { 3, 2, 1, 0 }`
`237`		`-`
`238`	`234`	`attributes #0 = { convergent nocallback nofree nosync nounwind willreturn memory(none) }`
Original file line number	Diff line number	Diff line change
`@@ -0,0 +1,2 @@`
	`1`	`+if not "AMDGPU" in config.root.targets:`
	`2`	`+ config.unsupported = True`