[X86] Lowering of load atomic float via cast

X86 backend does not lower load atomic float, so it can be casted to an
integer before lowering.

Author: jofrn

Diff for: llvm/lib/Target/X86/X86ISelLowering.cpp

@@ -31158,6 +31158,14 @@ bool X86TargetLowering::needsCmpXchgNb(Type *MemType) const {
   return false;
 }
 
+TargetLoweringBase::AtomicExpansionKind
+X86TargetLowering::shouldCastAtomicLoadInIR(LoadInst *LI) const {
+  if (const auto VT = dyn_cast<VectorType>(LI->getType()))
+    if (VT->getElementType()->isFloatingPointTy())
+      return AtomicExpansionKind::CastToInteger;
+  return TargetLowering::shouldCastAtomicLoadInIR(LI);
+}
+
 TargetLoweringBase::AtomicExpansionKind
 X86TargetLowering::shouldExpandAtomicStoreInIR(StoreInst *SI) const {
   Type *MemType = SI->getValueOperand()->getType();

Diff for: llvm/lib/Target/X86/X86ISelLowering.h

@@ -1808,6 +1808,8 @@ namespace llvm {
     const MCPhysReg *getScratchRegisters(CallingConv::ID CC) const override;
     ArrayRef<MCPhysReg> getRoundingControlRegisters() const override;
 
+    TargetLoweringBase::AtomicExpansionKind
+    shouldCastAtomicLoadInIR(LoadInst *LI) const override;
     TargetLoweringBase::AtomicExpansionKind
     shouldExpandAtomicLoadInIR(LoadInst *LI) const override;
     TargetLoweringBase::AtomicExpansionKind

Diff for: llvm/test/CodeGen/X86/atomicvec-float.ll

@@ -0,0 +1,38 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc < %s --mtriple=x86_64 | FileCheck %s
+
+define float @load_atomic_float(ptr %src) {
+; CHECK-LABEL: load_atomic_float:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-NEXT:    retq
+  %ret = load atomic float, ptr %src acquire, align 4
+  ret float %ret
+}
+
+define <1 x float> @load_atomic_vector_float1(ptr %src) {
+; CHECK-LABEL: load_atomic_vector_float1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-NEXT:    retq
+  %ret = load atomic <1 x float>, ptr %src acquire, align 4
+  ret <1 x float> %ret
+}
+
+define <2 x float> @load_atomic_vector_float2(ptr %src) {
+; CHECK-LABEL: load_atomic_vector_float2:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    pushq %rax
+; CHECK-NEXT:    .cfi_def_cfa_offset 16
+; CHECK-NEXT:    movq %rdi, %rsi
+; CHECK-NEXT:    movq %rsp, %rdx
+; CHECK-NEXT:    movl $8, %edi
+; CHECK-NEXT:    movl $2, %ecx
+; CHECK-NEXT:    callq __atomic_load@PLT
+; CHECK-NEXT:    movsd {{.*#+}} xmm0 = mem[0],zero
+; CHECK-NEXT:    popq %rax
+; CHECK-NEXT:    .cfi_def_cfa_offset 8
+; CHECK-NEXT:    retq
+  %ret = load atomic <2 x float>, ptr %src acquire, align 4
+  ret <2 x float> %ret
+}

Diff for: llvm/test/Transforms/AtomicExpand/atomicvec-float.ll

@@ -0,0 +1,38 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt < %s --mtriple=x86_64 --passes=atomic-expand -S -o - | FileCheck %s
+
+define float @load_atomic_float(ptr %src) {
+; CHECK-LABEL: define float @load_atomic_float(
+; CHECK-SAME: ptr [[SRC:%.*]]) {
+; CHECK-NEXT:    [[TMP1:%.*]] = load atomic i32, ptr [[SRC]] acquire, align 4
+; CHECK-NEXT:    [[TMP2:%.*]] = bitcast i32 [[TMP1]] to float
+; CHECK-NEXT:    ret float [[TMP2]]
+;
+  %ret = load atomic float, ptr %src acquire, align 4
+  ret float %ret
+}
+
+define <1 x float> @load_atomic_vector_float1(ptr %src) {
+; CHECK-LABEL: define <1 x float> @load_atomic_vector_float1(
+; CHECK-SAME: ptr [[SRC:%.*]]) {
+; CHECK-NEXT:    [[TMP1:%.*]] = load atomic i32, ptr [[SRC]] acquire, align 4
+; CHECK-NEXT:    [[TMP2:%.*]] = bitcast i32 [[TMP1]] to <1 x float>
+; CHECK-NEXT:    ret <1 x float> [[TMP2]]
+;
+  %ret = load atomic <1 x float>, ptr %src acquire, align 4
+  ret <1 x float> %ret
+}
+
+define <2 x float> @load_atomic_vector_float2(ptr %src) {
+; CHECK-LABEL: define <2 x float> @load_atomic_vector_float2(
+; CHECK-SAME: ptr [[SRC:%.*]]) {
+; CHECK-NEXT:    [[TMP1:%.*]] = alloca <2 x float>, align 8
+; CHECK-NEXT:    call void @llvm.lifetime.start.p0(i64 8, ptr [[TMP1]])
+; CHECK-NEXT:    call void @__atomic_load(i64 8, ptr [[SRC]], ptr [[TMP1]], i32 2)
+; CHECK-NEXT:    [[TMP2:%.*]] = load <2 x float>, ptr [[TMP1]], align 8
+; CHECK-NEXT:    call void @llvm.lifetime.end.p0(i64 8, ptr [[TMP1]])
+; CHECK-NEXT:    ret <2 x float> [[TMP2]]
+;
+  %ret = load atomic <2 x float>, ptr %src acquire, align 4
+  ret <2 x float> %ret
+}