[VE] enable unaligned load/store isel

[simoll]

Summary

Enable unaligned load/store isel for iN and fp32/64 and tests.

[merge_guards_bot]

Unit tests: pass. 62198 tests passed, 0 failed and 823 were skipped.

clang-tidy: fail. clang-tidy found 2 errors and 0 warnings. 0 of them are added as review comments below (why?).

clang-format: fail. Please format your changes with clang-format by running git-clang-format HEAD^ or applying this patch.

Build artifacts: diff.json, clang-tidy.txt, clang-format.patch, CMakeCache.txt, console-log.txt, test-results.xml

Pre-merge checks is in beta. Report issue. Please join beta or enable it for your project.

[simoll]

• clang-format

[merge_guards_bot]

Unit tests: pass. 62198 tests passed, 0 failed and 823 were skipped.

clang-tidy: fail. clang-tidy found 2 errors and 0 warnings. 0 of them are added as review comments below (why?).

clang-format: pass.

Build artifacts: diff.json, clang-tidy.txt, clang-format.patch, CMakeCache.txt, console-log.txt, test-results.xml

Pre-merge checks is in beta. Report issue. Please join beta or enable it for your project.

[arsenm]

llvm/lib/Target/VE/VEInstrInfo.td

777

There's a new dedicated Alignment predicate on patfrags, so you can avoid the code predicate

[simoll]

• unaligned pattern rewrite

[simoll]

• Stripped redundant unaligned load/store patterns..

[merge_guards_bot]

Unit tests: pass. 62222 tests passed, 0 failed and 824 were skipped.

clang-tidy: fail. clang-tidy found 2 errors and 0 warnings. 0 of them are added as review comments below (why?).

clang-format: pass.

Build artifacts: diff.json, clang-tidy.txt, clang-format.patch, CMakeCache.txt, console-log.txt, test-results.xml

Pre-merge checks is in beta. Report issue. Please join beta or enable it for your project.

llvm/lib/Target/VE/VEISelLowering.h

public:
 
   SDValue withTargetFlags(SDValue Op, unsigned TF, SelectionDAG &DAG) const;
   SDValue makeHiLoPair(SDValue Op, unsigned HiTF, unsigned LoTF,
                        SelectionDAG &DAG) const;
   SDValue makeAddress(SDValue Op, SelectionDAG &DAG) const;
 
   bool isFPImmLegal(const APFloat &Imm, EVT VT,
                     bool ForCodeSize) const override;
+  /// Returns true if the target allows unaligned memory accesses of the
+  /// specified type.
+  bool allowsMisalignedMemoryAccesses(EVT VT, unsigned AS, unsigned Align,
+                                      MachineMemOperand::Flags Flags,
+                                      bool *Fast) const override;
 };
 } // namespace llvm
 
 #endif // VE_ISELLOWERING_H

llvm/lib/Target/VE/VEISelLowering.cpp

 /// isFPImmLegal - Returns true if the target can instruction select the
 /// specified FP immediate natively. If false, the legalizer will
 /// materialize the FP immediate as a load from a constant pool.
 bool VETargetLowering::isFPImmLegal(const APFloat &Imm, EVT VT,
                                     bool ForCodeSize) const {
   return VT == MVT::f32 || VT == MVT::f64;
 }
 
+/// Determine if the target supports unaligned memory accesses.
+///
+/// This function returns true if the target allows unaligned memory accesses
+/// of the specified type in the given address space. If true, it also returns
+/// whether the unaligned memory access is "fast" in the last argument by
+/// reference. This is used, for example, in situations where an array
+/// copy/move/set is converted to a sequence of store operations. Its use
+/// helps to ensure that such replacements don't generate code that causes an
+/// alignment error (trap) on the target machine.
+bool VETargetLowering::allowsMisalignedMemoryAccesses(EVT VT,
+                                                      unsigned AddrSpace,
+                                                      unsigned Align,
+                                                      MachineMemOperand::Flags,
+                                                      bool *Fast) const {
+  if (Fast) {
+    // It's fast anytime on VE
+    *Fast = true;
+  }
+  return true;
+}
+
 VETargetLowering::VETargetLowering(const TargetMachine &TM,
                                    const VESubtarget &STI)
     : TargetLowering(TM), Subtarget(&STI) {
   // Instructions which use registers as conditionals examine all the
   // bits (as does the pseudo SELECT_CC expansion). I don't think it
   // matters much whether it's ZeroOrOneBooleanContent, or
   // ZeroOrNegativeOneBooleanContent, so, arbitrarily choose the
   // former.

llvm/lib/Target/VE/VEInstrInfo.td

def ST1Bri : RM<
     "st1b $sx, $addr",
     [(truncstorei8 i32:$sx, ADDRri:$addr)]>;
 }
 }
 
 def : Pat<(f64 (load ADDRri:$addr)), (LDSri ADDRri:$addr)>;
 def : Pat<(store f64:$sx, ADDRri:$addr), (STSri ADDRri:$addr, $sx)>;
 
+// Patterns for unaligned load
+
+def unaligned4load : PatFrag<(ops node:$ptr), (load node:$ptr), [{
+  return cast<LoadSDNode>(N)->getAlignment() < 4;
+}]>;
+def unaligned8load : PatFrag<(ops node:$ptr), (load node:$ptr), [{
+  return cast<LoadSDNode>(N)->getAlignment() < 8;
+}]>;
+def : Pat<(i32 (unaligned4load ADDRri:$addr)),
+          (LDLri MEMri:$addr)>;
+def : Pat<(f32 (unaligned4load ADDRri:$addr)),
+          (LDUri MEMri:$addr)>;
+def : Pat<(i64 (unaligned8load ADDRri:$addr)),
+          (LDSri ADDRri:$addr)>;
+def : Pat<(f64 (unaligned8load ADDRri:$addr)),
+          (LDSri ADDRri:$addr)>;
+
+// Patterns for unaligned store
+
+def unaligned4store : PatFrag<(ops node:$val, node:$ptr),
+                              (store node:$val, node:$ptr), [{
+  return cast<StoreSDNode>(N)->getAlignment() < 4;
+}]>;
+def unaligned8store : PatFrag<(ops node:$val, node:$ptr),
+                              (store node:$val, node:$ptr), [{
+  return cast<StoreSDNode>(N)->getAlignment() < 8;
+}]>;
+def : Pat<(unaligned4store i32:$sx, ADDRri:$addr),
+          (STLri ADDRri:$addr, $sx)>;
+def : Pat<(unaligned4store f32:$sx, ADDRri:$addr),
+          (STUri ADDRri:$addr, $sx)>;
+def : Pat<(unaligned8store i64:$sx, ADDRri:$addr),
+          (STSri ADDRri:$addr, $sx)>;
+def : Pat<(unaligned8store f64:$sx, ADDRri:$addr),
+          (STSri ADDRri:$addr, $sx)>;
+
+// Patterns for unaligned sextload/zextload/extload
+
+def unaligned2extloadi16 : PatFrag<(ops node:$ptr), (extloadi16 node:$ptr), [{
+  return cast<LoadSDNode>(N)->getAlignment() < 2;
+}]>;
+def unaligned2sextloadi16 : PatFrag<(ops node:$ptr), (sextloadi16 node:$ptr), [{
+  return cast<LoadSDNode>(N)->getAlignment() < 2;
+}]>;
+def unaligned2zextloadi16 : PatFrag<(ops node:$ptr), (zextloadi16 node:$ptr), [{
+  return cast<LoadSDNode>(N)->getAlignment() < 2;
+}]>;
+def unaligned4extloadi32 : PatFrag<(ops node:$ptr), (extloadi32 node:$ptr), [{
+  return cast<LoadSDNode>(N)->getAlignment() < 4;
+}]>;
+def unaligned4sextloadi32 : PatFrag<(ops node:$ptr), (sextloadi32 node:$ptr), [{
+  return cast<LoadSDNode>(N)->getAlignment() < 4;
+}]>;
+def unaligned4zextloadi32 : PatFrag<(ops node:$ptr), (zextloadi32 node:$ptr), [{
+  return cast<LoadSDNode>(N)->getAlignment() < 4;
+}]>;
+def : Pat<(i64 (unaligned2sextloadi16 ADDRri:$addr)),
+          (INSERT_SUBREG (i64 (IMPLICIT_DEF)), (LD2Bri MEMri:$addr), sub_i32)>;
+def : Pat<(i64 (unaligned2zextloadi16 ADDRri:$addr)),
+          (INSERT_SUBREG (i64 (IMPLICIT_DEF)), (LD2BUri MEMri:$addr), sub_i32)>;
+def : Pat<(i64 (unaligned2extloadi16 ADDRri:$addr)),
+          (INSERT_SUBREG (i64 (IMPLICIT_DEF)), (LD2BUri MEMri:$addr), sub_i32)>;
+def : Pat<(i64 (unaligned4sextloadi32 ADDRri:$addr)),
+          (INSERT_SUBREG (i64 (IMPLICIT_DEF)), (LDLri MEMri:$addr), sub_i32)>;
+def : Pat<(i64 (unaligned4zextloadi32 ADDRri:$addr)),
+          (INSERT_SUBREG (i64 (IMPLICIT_DEF)), (LDLUri MEMri:$addr), sub_i32)>;
+def : Pat<(i64 (unaligned4extloadi32 ADDRri:$addr)),
+          (INSERT_SUBREG (i64 (IMPLICIT_DEF)), (LDLri MEMri:$addr), sub_i32)>;
+
+// Patterns for unaligned truncstore
+
+def unaligned4truncstorei16 : PatFrag<(ops node:$val, node:$ptr),
+                                   (truncstorei16 node:$val, node:$ptr), [{
+  return cast<StoreSDNode>(N)->getAlignment() < 4;
+}]>;
+def unaligned8truncstorei32 : PatFrag<(ops node:$val, node:$ptr),
+                                   (truncstorei32 node:$val, node:$ptr), [{
+  return cast<StoreSDNode>(N)->getAlignment() < 8;
+}]>;
+def : Pat<(unaligned4truncstorei16 i64:$sx, ADDRri:$addr),
+          (ST2Bri ADDRri:$addr, (EXTRACT_SUBREG $sx, sub_i32))>;
+def : Pat<(unaligned8truncstorei32 i64:$sx, ADDRri:$addr),
+          (STLri ADDRri:$addr, (EXTRACT_SUBREG $sx, sub_i32))>;
+
 // Return instruction is also a special case of jump.
 let cx = 0, cx2 = 0, bpf = 0 /* NONE */, cf = 15 /* AT */, cy = 0, sy = 0,
     cz = 1, sz = 0x10 /* SX10 */, imm32 = 0, Uses = [SX10],
     isReturn = 1, isTerminator = 1, hasDelaySlot = 1, isBarrier = 1,
     isCodeGenOnly = 1, hasSideEffects = 0 in
 def RET : CF<
     0x19, (outs), (ins),
     "b.l (,%lr)",

llvm/test/CodeGen/VE/load-align1.ll

+; RUN: llc < %s -mtriple=ve-unknown-unknown | FileCheck %s
+
+@vi8 = common dso_local local_unnamed_addr global i8 0, align 1
+@vi16 = common dso_local local_unnamed_addr global i16 0, align 1
+@vi32 = common dso_local local_unnamed_addr global i32 0, align 1
+@vi64 = common dso_local local_unnamed_addr global i64 0, align 1
+@vf32 = common dso_local local_unnamed_addr global float 0.000000e+00, align 1
+@vf64 = common dso_local local_unnamed_addr global double 0.000000e+00, align 1
+
+; Function Attrs: norecurse nounwind readonly
+define double @loadf64stk() {
+; CHECK-LABEL: loadf64stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    ld %s0, 184(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca double, align 1
+  %1 = load double, double* %addr, align 1
+  ret double %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define float @loadf32stk() {
+; CHECK-LABEL: loadf32stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    ldu %s0, 188(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca float, align 1
+  %1 = load float, float* %addr, align 1
+  ret float %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i64 @loadi64stk() {
+; CHECK-LABEL: loadi64stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    ld %s0, 184(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i64, align 1
+  %1 = load i64, i64* %addr, align 1
+  ret i64 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i32 @loadi32stk() {
+; CHECK-LABEL: loadi32stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    ldl.sx %s0, 188(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i32, align 1
+  %1 = load i32, i32* %addr, align 1
+  ret i32 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i16 @loadi16stk() {
+; CHECK-LABEL: loadi16stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    ld2b.zx %s0, 190(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i16, align 1
+  %1 = load i16, i16* %addr, align 1
+  ret i16 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i8 @loadi8stk() {
+; CHECK-LABEL: loadi8stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    ld1b.zx %s0, 191(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i8, align 1
+  %1 = load i8, i8* %addr, align 1
+  ret i8 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define double @loadf64com() {
+; CHECK-LABEL: loadf64com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s0, vf64@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s0, vf64@hi(%s0)
+; CHECK-NEXT:    ld %s0, (,%s0)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %1 = load double, double* @vf64, align 1
+  ret double %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define float @loadf32com() {
+; CHECK-LABEL: loadf32com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s0, vf32@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s0, vf32@hi(%s0)
+; CHECK-NEXT:    ldu %s0, (,%s0)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %1 = load float, float* @vf32, align 1
+  ret float %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i64 @loadi64com() {
+; CHECK-LABEL: loadi64com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s0, vi64@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s0, vi64@hi(%s0)
+; CHECK-NEXT:    ld %s0, (,%s0)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %1 = load i64, i64* @vi64, align 1
+  ret i64 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i32 @loadi32com() {
+; CHECK-LABEL: loadi32com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s0, vi32@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s0, vi32@hi(%s0)
+; CHECK-NEXT:    ldl.sx %s0, (,%s0)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %1 = load i32, i32* @vi32, align 1
+  ret i32 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i16 @loadi16com() {
+; CHECK-LABEL: loadi16com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s0, vi16@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s0, vi16@hi(%s0)
+; CHECK-NEXT:    ld2b.zx %s0, (,%s0)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %1 = load i16, i16* @vi16, align 1
+  ret i16 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i8 @loadi8com() {
+; CHECK-LABEL: loadi8com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s0, vi8@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s0, vi8@hi(%s0)
+; CHECK-NEXT:    ld1b.zx %s0, (,%s0)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %1 = load i8, i8* @vi8, align 1
+  ret i8 %1
+}
+

llvm/test/CodeGen/VE/load-align2.ll

+; RUN: llc < %s -mtriple=ve-unknown-unknown | FileCheck %s
+
+@vi8 = common dso_local local_unnamed_addr global i8 0, align 2
+@vi16 = common dso_local local_unnamed_addr global i16 0, align 2
+@vi32 = common dso_local local_unnamed_addr global i32 0, align 2
+@vi64 = common dso_local local_unnamed_addr global i64 0, align 2
+@vf32 = common dso_local local_unnamed_addr global float 0.000000e+00, align 2
+@vf64 = common dso_local local_unnamed_addr global double 0.000000e+00, align 2
+
+; Function Attrs: norecurse nounwind readonly
+define double @loadf64stk() {
+; CHECK-LABEL: loadf64stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    ld %s0, 184(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca double, align 2
+  %1 = load double, double* %addr, align 2
+  ret double %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define float @loadf32stk() {
+; CHECK-LABEL: loadf32stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    ldu %s0, 188(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca float, align 2
+  %1 = load float, float* %addr, align 2
+  ret float %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i64 @loadi64stk() {
+; CHECK-LABEL: loadi64stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    ld %s0, 184(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i64, align 2
+  %1 = load i64, i64* %addr, align 2
+  ret i64 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i32 @loadi32stk() {
+; CHECK-LABEL: loadi32stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    ldl.sx %s0, 188(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i32, align 2
+  %1 = load i32, i32* %addr, align 2
+  ret i32 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i16 @loadi16stk() {
+; CHECK-LABEL: loadi16stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    ld2b.zx %s0, 190(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i16, align 2
+  %1 = load i16, i16* %addr, align 2
+  ret i16 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i8 @loadi8stk() {
+; CHECK-LABEL: loadi8stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    ld1b.zx %s0, 190(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i8, align 2
+  %1 = load i8, i8* %addr, align 2
+  ret i8 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define double @loadf64com() {
+; CHECK-LABEL: loadf64com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s0, vf64@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s0, vf64@hi(%s0)
+; CHECK-NEXT:    ld %s0, (,%s0)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %1 = load double, double* @vf64, align 2
+  ret double %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define float @loadf32com() {
+; CHECK-LABEL: loadf32com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s0, vf32@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s0, vf32@hi(%s0)
+; CHECK-NEXT:    ldu %s0, (,%s0)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %1 = load float, float* @vf32, align 2
+  ret float %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i64 @loadi64com() {
+; CHECK-LABEL: loadi64com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s0, vi64@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s0, vi64@hi(%s0)
+; CHECK-NEXT:    ld %s0, (,%s0)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %1 = load i64, i64* @vi64, align 2
+  ret i64 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i32 @loadi32com() {
+; CHECK-LABEL: loadi32com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s0, vi32@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s0, vi32@hi(%s0)
+; CHECK-NEXT:    ldl.sx %s0, (,%s0)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %1 = load i32, i32* @vi32, align 2
+  ret i32 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i16 @loadi16com() {
+; CHECK-LABEL: loadi16com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s0, vi16@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s0, vi16@hi(%s0)
+; CHECK-NEXT:    ld2b.zx %s0, (,%s0)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %1 = load i16, i16* @vi16, align 2
+  ret i16 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i8 @loadi8com() {
+; CHECK-LABEL: loadi8com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s0, vi8@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s0, vi8@hi(%s0)
+; CHECK-NEXT:    ld1b.zx %s0, (,%s0)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %1 = load i8, i8* @vi8, align 2
+  ret i8 %1
+}
+

llvm/test/CodeGen/VE/load-align4.ll

+; RUN: llc < %s -mtriple=ve-unknown-unknown | FileCheck %s
+
+@vi8 = common dso_local local_unnamed_addr global i8 0, align 4
+@vi16 = common dso_local local_unnamed_addr global i16 0, align 4
+@vi32 = common dso_local local_unnamed_addr global i32 0, align 4
+@vi64 = common dso_local local_unnamed_addr global i64 0, align 4
+@vf32 = common dso_local local_unnamed_addr global float 0.000000e+00, align 4
+@vf64 = common dso_local local_unnamed_addr global double 0.000000e+00, align 4
+
+; Function Attrs: norecurse nounwind readonly
+define double @loadf64stk() {
+; CHECK-LABEL: loadf64stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    ld %s0, 184(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca double, align 4
+  %1 = load double, double* %addr, align 4
+  ret double %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define float @loadf32stk() {
+; CHECK-LABEL: loadf32stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    ldu %s0, 188(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca float, align 4
+  %1 = load float, float* %addr, align 4
+  ret float %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i64 @loadi64stk() {
+; CHECK-LABEL: loadi64stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    ld %s0, 184(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i64, align 4
+  %1 = load i64, i64* %addr, align 4
+  ret i64 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i32 @loadi32stk() {
+; CHECK-LABEL: loadi32stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    ldl.sx %s0, 188(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i32, align 4
+  %1 = load i32, i32* %addr, align 4
+  ret i32 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i16 @loadi16stk() {
+; CHECK-LABEL: loadi16stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    ld2b.zx %s0, 188(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i16, align 4
+  %1 = load i16, i16* %addr, align 4
+  ret i16 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i8 @loadi8stk() {
+; CHECK-LABEL: loadi8stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    ld1b.zx %s0, 188(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i8, align 4
+  %1 = load i8, i8* %addr, align 4
+  ret i8 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define double @loadf64com() {
+; CHECK-LABEL: loadf64com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s0, vf64@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s0, vf64@hi(%s0)
+; CHECK-NEXT:    ld %s0, (,%s0)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %1 = load double, double* @vf64, align 4
+  ret double %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define float @loadf32com() {
+; CHECK-LABEL: loadf32com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s0, vf32@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s0, vf32@hi(%s0)
+; CHECK-NEXT:    ldu %s0, (,%s0)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %1 = load float, float* @vf32, align 4
+  ret float %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i64 @loadi64com() {
+; CHECK-LABEL: loadi64com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s0, vi64@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s0, vi64@hi(%s0)
+; CHECK-NEXT:    ld %s0, (,%s0)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %1 = load i64, i64* @vi64, align 4
+  ret i64 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i32 @loadi32com() {
+; CHECK-LABEL: loadi32com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s0, vi32@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s0, vi32@hi(%s0)
+; CHECK-NEXT:    ldl.sx %s0, (,%s0)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %1 = load i32, i32* @vi32, align 4
+  ret i32 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i16 @loadi16com() {
+; CHECK-LABEL: loadi16com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s0, vi16@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s0, vi16@hi(%s0)
+; CHECK-NEXT:    ld2b.zx %s0, (,%s0)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %1 = load i16, i16* @vi16, align 4
+  ret i16 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i8 @loadi8com() {
+; CHECK-LABEL: loadi8com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s0, vi8@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s0, vi8@hi(%s0)
+; CHECK-NEXT:    ld1b.zx %s0, (,%s0)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %1 = load i8, i8* @vi8, align 4
+  ret i8 %1
+}
+

llvm/test/CodeGen/VE/load-align8.ll

+; RUN: llc < %s -mtriple=ve-unknown-unknown | FileCheck %s
+
+@vi8 = common dso_local local_unnamed_addr global i8 0, align 8
+@vi16 = common dso_local local_unnamed_addr global i16 0, align 8
+@vi32 = common dso_local local_unnamed_addr global i32 0, align 8
+@vi64 = common dso_local local_unnamed_addr global i64 0, align 8
+@vf32 = common dso_local local_unnamed_addr global float 0.000000e+00, align 8
+@vf64 = common dso_local local_unnamed_addr global double 0.000000e+00, align 8
+
+; Function Attrs: norecurse nounwind readonly
+define double @loadf64stk() {
+; CHECK-LABEL: loadf64stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    ld %s0, 184(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca double, align 8
+  %1 = load double, double* %addr, align 8
+  ret double %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define float @loadf32stk() {
+; CHECK-LABEL: loadf32stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    ldu %s0, 184(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca float, align 8
+  %1 = load float, float* %addr, align 8
+  ret float %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i64 @loadi64stk() {
+; CHECK-LABEL: loadi64stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    ld %s0, 184(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i64, align 8
+  %1 = load i64, i64* %addr, align 8
+  ret i64 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i32 @loadi32stk() {
+; CHECK-LABEL: loadi32stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    ldl.sx %s0, 184(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i32, align 8
+  %1 = load i32, i32* %addr, align 8
+  ret i32 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i16 @loadi16stk() {
+; CHECK-LABEL: loadi16stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    ld2b.zx %s0, 184(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i16, align 8
+  %1 = load i16, i16* %addr, align 8
+  ret i16 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i8 @loadi8stk() {
+; CHECK-LABEL: loadi8stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    ld1b.zx %s0, 184(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i8, align 8
+  %1 = load i8, i8* %addr, align 8
+  ret i8 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define double @loadf64com() {
+; CHECK-LABEL: loadf64com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s0, vf64@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s0, vf64@hi(%s0)
+; CHECK-NEXT:    ld %s0, (,%s0)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %1 = load double, double* @vf64, align 8
+  ret double %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define float @loadf32com() {
+; CHECK-LABEL: loadf32com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s0, vf32@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s0, vf32@hi(%s0)
+; CHECK-NEXT:    ldu %s0, (,%s0)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %1 = load float, float* @vf32, align 8
+  ret float %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i64 @loadi64com() {
+; CHECK-LABEL: loadi64com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s0, vi64@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s0, vi64@hi(%s0)
+; CHECK-NEXT:    ld %s0, (,%s0)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %1 = load i64, i64* @vi64, align 8
+  ret i64 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i32 @loadi32com() {
+; CHECK-LABEL: loadi32com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s0, vi32@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s0, vi32@hi(%s0)
+; CHECK-NEXT:    ldl.sx %s0, (,%s0)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %1 = load i32, i32* @vi32, align 8
+  ret i32 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i16 @loadi16com() {
+; CHECK-LABEL: loadi16com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s0, vi16@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s0, vi16@hi(%s0)
+; CHECK-NEXT:    ld2b.zx %s0, (,%s0)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %1 = load i16, i16* @vi16, align 8
+  ret i16 %1
+}
+
+; Function Attrs: norecurse nounwind readonly
+define i8 @loadi8com() {
+; CHECK-LABEL: loadi8com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s0, vi8@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s0, vi8@hi(%s0)
+; CHECK-NEXT:    ld1b.zx %s0, (,%s0)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %1 = load i8, i8* @vi8, align 8
+  ret i8 %1
+}
+

llvm/test/CodeGen/VE/store-align1.ll

+; RUN: llc < %s -mtriple=ve-unknown-unknown | FileCheck %s
+
+@vi8 = common dso_local local_unnamed_addr global i8 0, align 1
+@vi16 = common dso_local local_unnamed_addr global i16 0, align 1
+@vi32 = common dso_local local_unnamed_addr global i32 0, align 1
+@vi64 = common dso_local local_unnamed_addr global i64 0, align 1
+@vf32 = common dso_local local_unnamed_addr global float 0.000000e+00, align 1
+@vf64 = common dso_local local_unnamed_addr global double 0.000000e+00, align 1
+
+; Function Attrs: norecurse nounwind readonly
+define void @storef64stk(double %0) {
+; CHECK-LABEL: storef64stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    st %s0, 184(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca double, align 1
+  store double %0, double* %addr, align 1
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storef32stk(float %0) {
+; CHECK-LABEL: storef32stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    stu %s0, 188(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca float, align 1
+  store float %0, float* %addr, align 1
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei64stk(i64 %0) {
+; CHECK-LABEL: storei64stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    st %s0, 184(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i64, align 1
+  store i64 %0, i64* %addr, align 1
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei32stk(i32 %0) {
+; CHECK-LABEL: storei32stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    stl %s0, 188(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i32, align 1
+  store i32 %0, i32* %addr, align 1
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei16stk(i16 %0) {
+; CHECK-LABEL: storei16stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    st2b %s0, 190(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i16, align 1
+  store i16 %0, i16* %addr, align 1
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei8stk(i8 %0) {
+; CHECK-LABEL: storei8stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    st1b %s0, 191(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i8, align 1
+  store i8 %0, i8* %addr, align 1
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storef64com(double %0) {
+; CHECK-LABEL: storef64com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s1, vf64@lo
+; CHECK-NEXT:    and %s1, %s1, (32)0
+; CHECK-NEXT:    lea.sl %s1, vf64@hi(%s1)
+; CHECK-NEXT:    st %s0, (,%s1)
+; CHECK-NEXT:    or %s11, 0, %s9
+  store double %0, double* @vf64, align 1
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storef32com(float %0) {
+; CHECK-LABEL: storef32com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s1, vf32@lo
+; CHECK-NEXT:    and %s1, %s1, (32)0
+; CHECK-NEXT:    lea.sl %s1, vf32@hi(%s1)
+; CHECK-NEXT:    stu %s0, (,%s1)
+; CHECK-NEXT:    or %s11, 0, %s9
+  store float %0, float* @vf32, align 1
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei64com(i64 %0) {
+; CHECK-LABEL: storei64com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s1, vi64@lo
+; CHECK-NEXT:    and %s1, %s1, (32)0
+; CHECK-NEXT:    lea.sl %s1, vi64@hi(%s1)
+; CHECK-NEXT:    st %s0, (,%s1)
+; CHECK-NEXT:    or %s11, 0, %s9
+  store i64 %0, i64* @vi64, align 1
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei32com(i32 %0) {
+; CHECK-LABEL: storei32com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s1, vi32@lo
+; CHECK-NEXT:    and %s1, %s1, (32)0
+; CHECK-NEXT:    lea.sl %s1, vi32@hi(%s1)
+; CHECK-NEXT:    stl %s0, (,%s1)
+; CHECK-NEXT:    or %s11, 0, %s9
+  store i32 %0, i32* @vi32, align 1
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei16com(i16 %0) {
+; CHECK-LABEL: storei16com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s1, vi16@lo
+; CHECK-NEXT:    and %s1, %s1, (32)0
+; CHECK-NEXT:    lea.sl %s1, vi16@hi(%s1)
+; CHECK-NEXT:    st2b %s0, (,%s1)
+; CHECK-NEXT:    or %s11, 0, %s9
+  store i16 %0, i16* @vi16, align 1
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei8com(i8 %0) {
+; CHECK-LABEL: storei8com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s1, vi8@lo
+; CHECK-NEXT:    and %s1, %s1, (32)0
+; CHECK-NEXT:    lea.sl %s1, vi8@hi(%s1)
+; CHECK-NEXT:    st1b %s0, (,%s1)
+; CHECK-NEXT:    or %s11, 0, %s9
+  store i8 %0, i8* @vi8, align 1
+  ret void
+}
+

llvm/test/CodeGen/VE/store-align2.ll

+; RUN: llc < %s -mtriple=ve-unknown-unknown | FileCheck %s
+
+@vi8 = common dso_local local_unnamed_addr global i8 0, align 2
+@vi16 = common dso_local local_unnamed_addr global i16 0, align 2
+@vi32 = common dso_local local_unnamed_addr global i32 0, align 2
+@vi64 = common dso_local local_unnamed_addr global i64 0, align 2
+@vf32 = common dso_local local_unnamed_addr global float 0.000000e+00, align 2
+@vf64 = common dso_local local_unnamed_addr global double 0.000000e+00, align 2
+
+; Function Attrs: norecurse nounwind readonly
+define void @storef64stk(double %0) {
+; CHECK-LABEL: storef64stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    st %s0, 184(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca double, align 2
+  store double %0, double* %addr, align 2
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storef32stk(float %0) {
+; CHECK-LABEL: storef32stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    stu %s0, 188(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca float, align 2
+  store float %0, float* %addr, align 2
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei64stk(i64 %0) {
+; CHECK-LABEL: storei64stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    st %s0, 184(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i64, align 2
+  store i64 %0, i64* %addr, align 2
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei32stk(i32 %0) {
+; CHECK-LABEL: storei32stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    stl %s0, 188(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i32, align 2
+  store i32 %0, i32* %addr, align 2
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei16stk(i16 %0) {
+; CHECK-LABEL: storei16stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    st2b %s0, 190(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i16, align 2
+  store i16 %0, i16* %addr, align 2
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei8stk(i8 %0) {
+; CHECK-LABEL: storei8stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    st1b %s0, 190(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i8, align 2
+  store i8 %0, i8* %addr, align 2
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storef64com(double %0) {
+; CHECK-LABEL: storef64com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s1, vf64@lo
+; CHECK-NEXT:    and %s1, %s1, (32)0
+; CHECK-NEXT:    lea.sl %s1, vf64@hi(%s1)
+; CHECK-NEXT:    st %s0, (,%s1)
+; CHECK-NEXT:    or %s11, 0, %s9
+  store double %0, double* @vf64, align 2
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storef32com(float %0) {
+; CHECK-LABEL: storef32com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s1, vf32@lo
+; CHECK-NEXT:    and %s1, %s1, (32)0
+; CHECK-NEXT:    lea.sl %s1, vf32@hi(%s1)
+; CHECK-NEXT:    stu %s0, (,%s1)
+; CHECK-NEXT:    or %s11, 0, %s9
+  store float %0, float* @vf32, align 2
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei64com(i64 %0) {
+; CHECK-LABEL: storei64com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s1, vi64@lo
+; CHECK-NEXT:    and %s1, %s1, (32)0
+; CHECK-NEXT:    lea.sl %s1, vi64@hi(%s1)
+; CHECK-NEXT:    st %s0, (,%s1)
+; CHECK-NEXT:    or %s11, 0, %s9
+  store i64 %0, i64* @vi64, align 2
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei32com(i32 %0) {
+; CHECK-LABEL: storei32com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s1, vi32@lo
+; CHECK-NEXT:    and %s1, %s1, (32)0
+; CHECK-NEXT:    lea.sl %s1, vi32@hi(%s1)
+; CHECK-NEXT:    stl %s0, (,%s1)
+; CHECK-NEXT:    or %s11, 0, %s9
+  store i32 %0, i32* @vi32, align 2
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei16com(i16 %0) {
+; CHECK-LABEL: storei16com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s1, vi16@lo
+; CHECK-NEXT:    and %s1, %s1, (32)0
+; CHECK-NEXT:    lea.sl %s1, vi16@hi(%s1)
+; CHECK-NEXT:    st2b %s0, (,%s1)
+; CHECK-NEXT:    or %s11, 0, %s9
+  store i16 %0, i16* @vi16, align 2
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei8com(i8 %0) {
+; CHECK-LABEL: storei8com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s1, vi8@lo
+; CHECK-NEXT:    and %s1, %s1, (32)0
+; CHECK-NEXT:    lea.sl %s1, vi8@hi(%s1)
+; CHECK-NEXT:    st1b %s0, (,%s1)
+; CHECK-NEXT:    or %s11, 0, %s9
+  store i8 %0, i8* @vi8, align 2
+  ret void
+}
+

llvm/test/CodeGen/VE/store-align4.ll

+; RUN: llc < %s -mtriple=ve-unknown-unknown | FileCheck %s
+
+@vi8 = common dso_local local_unnamed_addr global i8 0, align 4
+@vi16 = common dso_local local_unnamed_addr global i16 0, align 4
+@vi32 = common dso_local local_unnamed_addr global i32 0, align 4
+@vi64 = common dso_local local_unnamed_addr global i64 0, align 4
+@vf32 = common dso_local local_unnamed_addr global float 0.000000e+00, align 4
+@vf64 = common dso_local local_unnamed_addr global double 0.000000e+00, align 4
+
+; Function Attrs: norecurse nounwind readonly
+define void @storef64stk(double %0) {
+; CHECK-LABEL: storef64stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    st %s0, 184(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca double, align 4
+  store double %0, double* %addr, align 4
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storef32stk(float %0) {
+; CHECK-LABEL: storef32stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    stu %s0, 188(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca float, align 4
+  store float %0, float* %addr, align 4
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei64stk(i64 %0) {
+; CHECK-LABEL: storei64stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    st %s0, 184(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i64, align 4
+  store i64 %0, i64* %addr, align 4
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei32stk(i32 %0) {
+; CHECK-LABEL: storei32stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    stl %s0, 188(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i32, align 4
+  store i32 %0, i32* %addr, align 4
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei16stk(i16 %0) {
+; CHECK-LABEL: storei16stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    st2b %s0, 188(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i16, align 4
+  store i16 %0, i16* %addr, align 4
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei8stk(i8 %0) {
+; CHECK-LABEL: storei8stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    st1b %s0, 188(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i8, align 4
+  store i8 %0, i8* %addr, align 4
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storef64com(double %0) {
+; CHECK-LABEL: storef64com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s1, vf64@lo
+; CHECK-NEXT:    and %s1, %s1, (32)0
+; CHECK-NEXT:    lea.sl %s1, vf64@hi(%s1)
+; CHECK-NEXT:    st %s0, (,%s1)
+; CHECK-NEXT:    or %s11, 0, %s9
+  store double %0, double* @vf64, align 4
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storef32com(float %0) {
+; CHECK-LABEL: storef32com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s1, vf32@lo
+; CHECK-NEXT:    and %s1, %s1, (32)0
+; CHECK-NEXT:    lea.sl %s1, vf32@hi(%s1)
+; CHECK-NEXT:    stu %s0, (,%s1)
+; CHECK-NEXT:    or %s11, 0, %s9
+  store float %0, float* @vf32, align 4
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei64com(i64 %0) {
+; CHECK-LABEL: storei64com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s1, vi64@lo
+; CHECK-NEXT:    and %s1, %s1, (32)0
+; CHECK-NEXT:    lea.sl %s1, vi64@hi(%s1)
+; CHECK-NEXT:    st %s0, (,%s1)
+; CHECK-NEXT:    or %s11, 0, %s9
+  store i64 %0, i64* @vi64, align 4
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei32com(i32 %0) {
+; CHECK-LABEL: storei32com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s1, vi32@lo
+; CHECK-NEXT:    and %s1, %s1, (32)0
+; CHECK-NEXT:    lea.sl %s1, vi32@hi(%s1)
+; CHECK-NEXT:    stl %s0, (,%s1)
+; CHECK-NEXT:    or %s11, 0, %s9
+  store i32 %0, i32* @vi32, align 4
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei16com(i16 %0) {
+; CHECK-LABEL: storei16com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s1, vi16@lo
+; CHECK-NEXT:    and %s1, %s1, (32)0
+; CHECK-NEXT:    lea.sl %s1, vi16@hi(%s1)
+; CHECK-NEXT:    st2b %s0, (,%s1)
+; CHECK-NEXT:    or %s11, 0, %s9
+  store i16 %0, i16* @vi16, align 4
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei8com(i8 %0) {
+; CHECK-LABEL: storei8com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s1, vi8@lo
+; CHECK-NEXT:    and %s1, %s1, (32)0
+; CHECK-NEXT:    lea.sl %s1, vi8@hi(%s1)
+; CHECK-NEXT:    st1b %s0, (,%s1)
+; CHECK-NEXT:    or %s11, 0, %s9
+  store i8 %0, i8* @vi8, align 4
+  ret void
+}
+

llvm/test/CodeGen/VE/store-align8.ll

+; RUN: llc < %s -mtriple=ve-unknown-unknown | FileCheck %s
+
+@vi8 = common dso_local local_unnamed_addr global i8 0, align 8
+@vi16 = common dso_local local_unnamed_addr global i16 0, align 8
+@vi32 = common dso_local local_unnamed_addr global i32 0, align 8
+@vi64 = common dso_local local_unnamed_addr global i64 0, align 8
+@vf32 = common dso_local local_unnamed_addr global float 0.000000e+00, align 8
+@vf64 = common dso_local local_unnamed_addr global double 0.000000e+00, align 8
+
+; Function Attrs: norecurse nounwind readonly
+define void @storef64stk(double %0) {
+; CHECK-LABEL: storef64stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    st %s0, 184(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca double, align 8
+  store double %0, double* %addr, align 8
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storef32stk(float %0) {
+; CHECK-LABEL: storef32stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    stu %s0, 184(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca float, align 8
+  store float %0, float* %addr, align 8
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei64stk(i64 %0) {
+; CHECK-LABEL: storei64stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    st %s0, 184(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i64, align 8
+  store i64 %0, i64* %addr, align 8
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei32stk(i32 %0) {
+; CHECK-LABEL: storei32stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    stl %s0, 184(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i32, align 8
+  store i32 %0, i32* %addr, align 8
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei16stk(i16 %0) {
+; CHECK-LABEL: storei16stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    st2b %s0, 184(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i16, align 8
+  store i16 %0, i16* %addr, align 8
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei8stk(i8 %0) {
+; CHECK-LABEL: storei8stk:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    st1b %s0, 184(,%s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %addr = alloca i8, align 8
+  store i8 %0, i8* %addr, align 8
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storef64com(double %0) {
+; CHECK-LABEL: storef64com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s1, vf64@lo
+; CHECK-NEXT:    and %s1, %s1, (32)0
+; CHECK-NEXT:    lea.sl %s1, vf64@hi(%s1)
+; CHECK-NEXT:    st %s0, (,%s1)
+; CHECK-NEXT:    or %s11, 0, %s9
+  store double %0, double* @vf64, align 8
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storef32com(float %0) {
+; CHECK-LABEL: storef32com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s1, vf32@lo
+; CHECK-NEXT:    and %s1, %s1, (32)0
+; CHECK-NEXT:    lea.sl %s1, vf32@hi(%s1)
+; CHECK-NEXT:    stu %s0, (,%s1)
+; CHECK-NEXT:    or %s11, 0, %s9
+  store float %0, float* @vf32, align 8
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei64com(i64 %0) {
+; CHECK-LABEL: storei64com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s1, vi64@lo
+; CHECK-NEXT:    and %s1, %s1, (32)0
+; CHECK-NEXT:    lea.sl %s1, vi64@hi(%s1)
+; CHECK-NEXT:    st %s0, (,%s1)
+; CHECK-NEXT:    or %s11, 0, %s9
+  store i64 %0, i64* @vi64, align 8
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei32com(i32 %0) {
+; CHECK-LABEL: storei32com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s1, vi32@lo
+; CHECK-NEXT:    and %s1, %s1, (32)0
+; CHECK-NEXT:    lea.sl %s1, vi32@hi(%s1)
+; CHECK-NEXT:    stl %s0, (,%s1)
+; CHECK-NEXT:    or %s11, 0, %s9
+  store i32 %0, i32* @vi32, align 8
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei16com(i16 %0) {
+; CHECK-LABEL: storei16com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s1, vi16@lo
+; CHECK-NEXT:    and %s1, %s1, (32)0
+; CHECK-NEXT:    lea.sl %s1, vi16@hi(%s1)
+; CHECK-NEXT:    st2b %s0, (,%s1)
+; CHECK-NEXT:    or %s11, 0, %s9
+  store i16 %0, i16* @vi16, align 8
+  ret void
+}
+
+; Function Attrs: norecurse nounwind readonly
+define void @storei8com(i8 %0) {
+; CHECK-LABEL: storei8com:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s1, vi8@lo
+; CHECK-NEXT:    and %s1, %s1, (32)0
+; CHECK-NEXT:    lea.sl %s1, vi8@hi(%s1)
+; CHECK-NEXT:    st1b %s0, (,%s1)
+; CHECK-NEXT:    or %s11, 0, %s9
+  store i8 %0, i8* @vi8, align 8
+  ret void
+}
+