[PowerPC] Fix incorrect store alignment for __builtin_vsx_build_pair() (#108606)

Fixes #107229

Author: lei137

clang/lib/CodeGen/CGBuiltin.cpp

@@ -18175,7 +18175,7 @@ Value *CodeGenFunction::EmitPPCBuiltinExpr(unsigned BuiltinID,
       CallOps.push_back(Ops[i]);
     llvm::Function *F = CGM.getIntrinsic(ID);
     Value *Call = Builder.CreateCall(F, CallOps);
-    return Builder.CreateAlignedStore(Call, Ops[0], MaybeAlign(64));
+    return Builder.CreateAlignedStore(Call, Ops[0], MaybeAlign());
   }
 
   case PPC::BI__builtin_ppc_compare_and_swap:

clang/test/CodeGen/PowerPC/builtins-ppc-build-pair-mma.c

@@ -99,7 +99,7 @@ void test1(unsigned char *vqp, unsigned char *vpp, vector unsigned char vc1, vec
 // CHECK-LE-NOOPT-NEXT:    [[TMP4:%.*]] = load <16 x i8>, ptr [[VC1_ADDR]], align 16
 // CHECK-LE-NOOPT-NEXT:    [[TMP5:%.*]] = load <16 x i8>, ptr [[VC2_ADDR]], align 16
 // CHECK-LE-NOOPT-NEXT:    [[TMP6:%.*]] = call <256 x i1> @llvm.ppc.vsx.assemble.pair(<16 x i8> [[TMP5]], <16 x i8> [[TMP4]])
-// CHECK-LE-NOOPT-NEXT:    store <256 x i1> [[TMP6]], ptr [[RES]], align 64
+// CHECK-LE-NOOPT-NEXT:    store <256 x i1> [[TMP6]], ptr [[RES]], align 32
 // CHECK-LE-NOOPT-NEXT:    [[TMP7:%.*]] = load <256 x i1>, ptr [[RES]], align 32
 // CHECK-LE-NOOPT-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[RESP_ADDR]], align 8
 // CHECK-LE-NOOPT-NEXT:    store <256 x i1> [[TMP7]], ptr [[TMP8]], align 32

clang/test/CodeGen/PowerPC/builtins-ppc-pair-mma-types.c

@@ -85,11 +85,11 @@ void testVQLocal(int *ptr, vector unsigned char vc) {
 // CHECK-NEXT:    [[TMP3:%.*]] = load <16 x i8>, ptr [[VC_ADDR]], align 16
 // CHECK-NEXT:    [[TMP4:%.*]] = load <16 x i8>, ptr [[VC_ADDR]], align 16
 // CHECK-NEXT:    [[TMP5:%.*]] = call <256 x i1> @llvm.ppc.vsx.assemble.pair(<16 x i8> [[TMP3]], <16 x i8> [[TMP4]])
-// CHECK-NEXT:    store <256 x i1> [[TMP5]], ptr [[VP2]], align 64
+// CHECK-NEXT:    store <256 x i1> [[TMP5]], ptr [[VP2]], align 32
 // CHECK-NEXT:    [[TMP6:%.*]] = load <16 x i8>, ptr [[VC_ADDR]], align 16
 // CHECK-NEXT:    [[TMP7:%.*]] = load <16 x i8>, ptr [[VC_ADDR]], align 16
 // CHECK-NEXT:    [[TMP8:%.*]] = call <256 x i1> @llvm.ppc.vsx.assemble.pair(<16 x i8> [[TMP7]], <16 x i8> [[TMP6]])
-// CHECK-NEXT:    store <256 x i1> [[TMP8]], ptr [[VP2]], align 64
+// CHECK-NEXT:    store <256 x i1> [[TMP8]], ptr [[VP2]], align 32
 // CHECK-NEXT:    [[TMP9:%.*]] = load <256 x i1>, ptr [[VP3]], align 32
 // CHECK-NEXT:    [[TMP10:%.*]] = load <16 x i8>, ptr [[VC_ADDR]], align 16
 // CHECK-NEXT:    [[TMP11:%.*]] = call <512 x i1> @llvm.ppc.mma.xvf64ger(<256 x i1> [[TMP9]], <16 x i8> [[TMP10]])
@@ -118,11 +118,11 @@ void testVQLocal(int *ptr, vector unsigned char vc) {
 // CHECK-BE-NEXT:    [[TMP3:%.*]] = load <16 x i8>, ptr [[VC_ADDR]], align 16
 // CHECK-BE-NEXT:    [[TMP4:%.*]] = load <16 x i8>, ptr [[VC_ADDR]], align 16
 // CHECK-BE-NEXT:    [[TMP5:%.*]] = call <256 x i1> @llvm.ppc.vsx.assemble.pair(<16 x i8> [[TMP3]], <16 x i8> [[TMP4]])
-// CHECK-BE-NEXT:    store <256 x i1> [[TMP5]], ptr [[VP2]], align 64
+// CHECK-BE-NEXT:    store <256 x i1> [[TMP5]], ptr [[VP2]], align 32
 // CHECK-BE-NEXT:    [[TMP6:%.*]] = load <16 x i8>, ptr [[VC_ADDR]], align 16
 // CHECK-BE-NEXT:    [[TMP7:%.*]] = load <16 x i8>, ptr [[VC_ADDR]], align 16
 // CHECK-BE-NEXT:    [[TMP8:%.*]] = call <256 x i1> @llvm.ppc.vsx.assemble.pair(<16 x i8> [[TMP6]], <16 x i8> [[TMP7]])
-// CHECK-BE-NEXT:    store <256 x i1> [[TMP8]], ptr [[VP2]], align 64
+// CHECK-BE-NEXT:    store <256 x i1> [[TMP8]], ptr [[VP2]], align 32
 // CHECK-BE-NEXT:    [[TMP9:%.*]] = load <256 x i1>, ptr [[VP3]], align 32
 // CHECK-BE-NEXT:    [[TMP10:%.*]] = load <16 x i8>, ptr [[VC_ADDR]], align 16
 // CHECK-BE-NEXT:    [[TMP11:%.*]] = call <512 x i1> @llvm.ppc.mma.xvf64ger(<256 x i1> [[TMP9]], <16 x i8> [[TMP10]])