[llvm] 8e51b22 - [AArch64][GlobalISel] Legalize G_LOAD for v4s8 Vector (#82989)

[via llvm-commits]

Author: chuongg3
Date: 2024-02-28T13:55:27Z
New Revision: 8e51b22ce21b01ae0be8267c5da3703ffd3b2c5b

URL: https://github.com/llvm/llvm-project/commit/8e51b22ce21b01ae0be8267c5da3703ffd3b2c5b
DIFF: https://github.com/llvm/llvm-project/commit/8e51b22ce21b01ae0be8267c5da3703ffd3b2c5b.diff

LOG: [AArch64][GlobalISel] Legalize G_LOAD for v4s8 Vector (#82989)

Lowers `v4s8 = G_LOAD %ptr ptr` into

`s32 = G_LOAD %ptr ptr`
`v4s8 = G_BITCAST s32`

Added: 
    

Modified: 
    llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp
    llvm/test/CodeGen/AArch64/load.ll

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp b/llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp
index 3205484bb0631d..91323e456a5ef8 100644
--- a/llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp
+++ b/llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp
@@ -388,8 +388,14 @@ AArch64LegalizerInfo::AArch64LegalizerInfo(const AArch64Subtarget &ST)
       .clampMaxNumElements(0, s32, 4)
       .clampMaxNumElements(0, s64, 2)
       .clampMaxNumElements(0, p0, 2)
+      // TODO: Use BITCAST for v2i8, v2i16 after G_TRUNC gets sorted out
+      .bitcastIf(typeInSet(0, {v4s8}),
+                 [=](const LegalityQuery &Query) {
+                   const LLT VecTy = Query.Types[0];
+                   return std::pair(0, LLT::scalar(VecTy.getSizeInBits()));
+                 })
       .customIf(IsPtrVecPred)
-      .scalarizeIf(typeIs(0, v2s16), 0);
+      .scalarizeIf(typeInSet(0, {v2s16, v2s8}), 0);
 
   getActionDefinitionsBuilder(G_STORE)
       .customIf([=](const LegalityQuery &Query) {

diff  --git a/llvm/test/CodeGen/AArch64/load.ll b/llvm/test/CodeGen/AArch64/load.ll
index d82c1fe55a2dd2..7f4540d915ab37 100644
--- a/llvm/test/CodeGen/AArch64/load.ll
+++ b/llvm/test/CodeGen/AArch64/load.ll
@@ -1,9 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
 ; RUN: llc -mtriple=aarch64-none-linux-gnu %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-SD
-; RUN: llc -mtriple=aarch64-none-linux-gnu -global-isel -global-isel-abort=2 %s -o - 2>&1 | FileCheck %s --check-prefixes=CHECK,CHECK-GI
-
-; CHECK-GI:         warning: Instruction selection used fallback path for load_v2i8
-; CHECK-GI-NEXT:    warning: Instruction selection used fallback path for load_v4i8
+; RUN: llc -mtriple=aarch64-none-linux-gnu -global-isel %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-GI
 
 ; ===== Legal Scalars =====
 
@@ -111,13 +108,21 @@ define <2 x i64> @load_v2i64(ptr %ptr){
 ; ===== Smaller/Larger Width Vectors with Legal Element Sizes =====
 
 define <2 x i8> @load_v2i8(ptr %ptr, <2 x i8> %b){
-; CHECK-LABEL: load_v2i8:
-; CHECK:       // %bb.0:
-; CHECK-NEXT:    ld1 { v0.b }[0], [x0]
-; CHECK-NEXT:    add x8, x0, #1
-; CHECK-NEXT:    ld1 { v0.b }[4], [x8]
-; CHECK-NEXT:    // kill: def $d0 killed $d0 killed $q0
-; CHECK-NEXT:    ret
+; CHECK-SD-LABEL: load_v2i8:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    ld1 { v0.b }[0], [x0]
+; CHECK-SD-NEXT:    add x8, x0, #1
+; CHECK-SD-NEXT:    ld1 { v0.b }[4], [x8]
+; CHECK-SD-NEXT:    // kill: def $d0 killed $d0 killed $q0
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: load_v2i8:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    ldr b0, [x0]
+; CHECK-GI-NEXT:    ldr b1, [x0, #1]
+; CHECK-GI-NEXT:    mov v0.s[1], v1.s[0]
+; CHECK-GI-NEXT:    // kill: def $d0 killed $d0 killed $q0
+; CHECK-GI-NEXT:    ret
     %a = load <2 x i8>, ptr %ptr
     ret <2 x i8> %a
 }