[Clang][CodeGen] Optimised LLVM IR for atomic increments/decrements on floats

clang/lib/CodeGen/CGExprScalar.cpp

@@ -2792,6 +2792,19 @@ ScalarExprEmitter::EmitScalarPrePostIncDec(const UnaryOperator *E, LValue LV,
                                   llvm::AtomicOrdering::SequentiallyConsistent);
       return isPre ? Builder.CreateBinOp(op, old, amt) : old;
     }
+    // Special case for atomic increment/decrement on floats
+    if (type->isFloatingType()) {
+      llvm::AtomicRMWInst::BinOp aop =
+          isInc ? llvm::AtomicRMWInst::FAdd : llvm::AtomicRMWInst::FSub;
+      llvm::Instruction::BinaryOps op =
+          isInc ? llvm::Instruction::FAdd : llvm::Instruction::FSub;
+      llvm::Value *amt = llvm::ConstantFP::get(
+          VMContext, llvm::APFloat(static_cast<float>(1.0)));
+      llvm::Value *old =
+          Builder.CreateAtomicRMW(aop, LV.getAddress(CGF), amt,
+                                  llvm::AtomicOrdering::SequentiallyConsistent);
+      return isPre ? Builder.CreateBinOp(op, old, amt) : old;
+    }
     value = EmitLoadOfLValue(LV, E->getExprLoc());
     input = value;
     // For every other atomic operation, we need to emit a load-op-cmpxchg loop

clang/test/CodeGen/X86/x86-atomic-float.c

@@ -0,0 +1,69 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 4
+// RUN: %clang_cc1 -triple x86_64-linux-gnu -target-cpu core2 %s -S -emit-llvm -o - | FileCheck -check-prefixes=CHECK,CHECK64 %s
+// RUN: %clang_cc1 -triple i686-linux-gnu -target-cpu core2 %s -S -emit-llvm -o - | FileCheck -check-prefixes=CHECK,CHECK32 %s
+
+
+// CHECK-LABEL: define dso_local i32 @test_int_inc(
+// CHECK-SAME: ) #[[ATTR0:[0-9]+]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = atomicrmw add ptr @test_int_inc.n, i32 1 seq_cst, align 4
+// CHECK-NEXT:    ret i32 [[TMP0]]
+//
+int test_int_inc()
+{
+    static _Atomic int n;
+    return n++;
+}
+
+// CHECK-LABEL: define dso_local float @test_float_post_inc(
+// CHECK-SAME: ) #[[ATTR0]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = atomicrmw fadd ptr @test_float_post_inc.n, float 1.000000e+00 seq_cst, align 4
+// CHECK-NEXT:    ret float [[TMP0]]
+//
+float test_float_post_inc()
+{
+    static _Atomic float n;
+    return n++;
+}
+
+// CHECK-LABEL: define dso_local float @test_float_post_dc(
+// CHECK-SAME: ) #[[ATTR0]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = atomicrmw fsub ptr @test_float_post_dc.n, float 1.000000e+00 seq_cst, align 4
+// CHECK-NEXT:    ret float [[TMP0]]
+//
+float test_float_post_dc()
+{
+    static _Atomic float n;
+    return n--;
+}
+
+// CHECK-LABEL: define dso_local float @test_float_pre_dc(
+// CHECK-SAME: ) #[[ATTR0]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = atomicrmw fsub ptr @test_float_pre_dc.n, float 1.000000e+00 seq_cst, align 4
+// CHECK-NEXT:    [[TMP1:%.*]] = fsub float [[TMP0]], 1.000000e+00
+// CHECK-NEXT:    ret float [[TMP1]]
+//
+float test_float_pre_dc()
+{
+    static _Atomic float n;
+    return --n;
+}
+
+// CHECK-LABEL: define dso_local float @test_float_pre_inc(
+// CHECK-SAME: ) #[[ATTR0]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = atomicrmw fadd ptr @test_float_pre_inc.n, float 1.000000e+00 seq_cst, align 4
+// CHECK-NEXT:    [[TMP1:%.*]] = fadd float [[TMP0]], 1.000000e+00
+// CHECK-NEXT:    ret float [[TMP1]]
+//
+float test_float_pre_inc()
+{
+    static _Atomic float n;
+    return ++n;
+}
+//// NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+// CHECK32: {{.*}}
+// CHECK64: {{.*}}