[llvm-branch-commits] [llvm] AMDGPU/GlobalISel: Fix extending load narrow scalar (PR #210097)

[via llvm-branch-commits]

llvmorg-github-actions[bot] wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-backend-amdgpu

Author: Petar Avramovic (petar-avramovic)

<details>
<summary>Changes</summary>

isAnyScalar is explicit LLT::scalar check but we want to narrow scalar
integer types as well.

---
Full diff: https://github.com/llvm/llvm-project/pull/210097.diff


2 Files Affected:

- (modified) llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp (+1-1) 
- (modified) llvm/test/CodeGen/AMDGPU/GlobalISel/zextload.ll (+6-7) 


``````````diff
diff --git a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
index c2efcd626a9c1..8d0fed4c635d3 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
@@ -1818,7 +1818,7 @@ AMDGPULegalizerInfo::AMDGPULegalizerInfo(const GCNSubtarget &ST_,
   ExtLoads.narrowScalarIf(
       [](const LegalityQuery &Query) {
         LLT MemTy = Query.MMODescrs[0].MemoryTy;
-        return MemTy.isAnyScalar() && MemTy.getSizeInBits() > 32 &&
+        return MemTy.isScalar() && MemTy.getSizeInBits() > 32 &&
                Query.Types[0].getSizeInBits() > MemTy.getSizeInBits();
       }, // For large MemSize, narrowscalar to MemSize (load MemSize + ext)
       getScalarTypeFromMemDesc(0));
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/zextload.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/zextload.ll
index ae7bf4fd17d4d..4e3203bf972ce 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/zextload.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/zextload.ll
@@ -1,9 +1,9 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc -global-isel -global-isel-abort=2 -mtriple=amdgpu9.00-amd-amdpal < %s | FileCheck --check-prefix=GFX9 %s
-; RUN: llc -global-isel -global-isel-abort=2 -mtriple=amdgpu8.03-amd-amdpal < %s | FileCheck --check-prefix=GFX8 %s
-; RUN: llc -global-isel -global-isel-abort=2 -mtriple=amdgpu6.00-amd-amdpal < %s | FileCheck --check-prefix=GFX6 %s
-; RUN: llc -global-isel -global-isel-abort=2 -mtriple=amdgpu10.10-amd-amdpal < %s | FileCheck --check-prefix=GFX10 %s
-; RUN: llc -global-isel -global-isel-abort=2 -mtriple=amdgpu11.00-amd-amdpal < %s | FileCheck --check-prefix=GFX11 %s
+; RUN: llc -global-isel -mtriple=amdgpu9.00-amd-amdpal < %s | FileCheck --check-prefix=GFX9 %s
+; RUN: llc -global-isel -mtriple=amdgpu8.03-amd-amdpal < %s | FileCheck --check-prefix=GFX8 %s
+; RUN: llc -global-isel -mtriple=amdgpu6.00-amd-amdpal < %s | FileCheck --check-prefix=GFX6 %s
+; RUN: llc -global-isel -mtriple=amdgpu10.10-amd-amdpal < %s | FileCheck --check-prefix=GFX10 %s
+; RUN: llc -global-isel -mtriple=amdgpu11.00-amd-amdpal < %s | FileCheck --check-prefix=GFX11 %s
 
 define i32 @zextload_global_i1_to_i32(ptr addrspace(1) %ptr) {
 ; GFX9-LABEL: zextload_global_i1_to_i32:
@@ -468,8 +468,7 @@ define i128 @zextload_global_i64_to_i128(ptr addrspace(1) %ptr) {
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    s_mov_b32 s6, 0
 ; GFX6-NEXT:    s_mov_b32 s7, 0xf000
-; GFX6-NEXT:    s_mov_b32 s4, s6
-; GFX6-NEXT:    s_mov_b32 s5, s6
+; GFX6-NEXT:    s_mov_b64 s[4:5], 0
 ; GFX6-NEXT:    buffer_load_dwordx2 v[0:1], v[0:1], s[4:7], 0 addr64
 ; GFX6-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX6-NEXT:    v_mov_b32_e32 v3, 0

``````````

</details>


https://github.com/llvm/llvm-project/pull/210097