From 130e77e4311c74b9ddb7931e1adc89a5d1132c7a Mon Sep 17 00:00:00 2001 From: Juergen Ributzka Date: Thu, 31 Jul 2014 06:25:33 +0000 Subject: [PATCH] [FastISel][AArch64] Add sqrt intrinsic support. Fixes . llvm-svn: 214388 --- llvm/lib/Target/AArch64/AArch64FastISel.cpp | 19 +++++++++++++++++++ llvm/test/CodeGen/AArch64/fast-isel-sqrt.ll | 20 ++++++++++++++++++++ 2 files changed, 39 insertions(+) create mode 100644 llvm/test/CodeGen/AArch64/fast-isel-sqrt.ll diff --git a/llvm/lib/Target/AArch64/AArch64FastISel.cpp b/llvm/lib/Target/AArch64/AArch64FastISel.cpp index 3d5fb66..647bb57 100644 --- a/llvm/lib/Target/AArch64/AArch64FastISel.cpp +++ b/llvm/lib/Target/AArch64/AArch64FastISel.cpp @@ -1708,6 +1708,25 @@ bool AArch64FastISel::FastLowerIntrinsicCall(const IntrinsicInst *II) { .addImm(1); return true; } + case Intrinsic::sqrt: { + Type *RetTy = II->getCalledFunction()->getReturnType(); + + MVT VT; + if (!isTypeLegal(RetTy, VT)) + return false; + + unsigned Op0Reg = getRegForValue(II->getOperand(0)); + if (!Op0Reg) + return false; + bool Op0IsKill = hasTrivialKill(II->getOperand(0)); + + unsigned ResultReg = FastEmit_r(VT, VT, ISD::FSQRT, Op0Reg, Op0IsKill); + if (!ResultReg) + return false; + + UpdateValueMap(II, ResultReg); + return true; + } case Intrinsic::sadd_with_overflow: case Intrinsic::uadd_with_overflow: case Intrinsic::ssub_with_overflow: diff --git a/llvm/test/CodeGen/AArch64/fast-isel-sqrt.ll b/llvm/test/CodeGen/AArch64/fast-isel-sqrt.ll new file mode 100644 index 0000000..44ad8ea --- /dev/null +++ b/llvm/test/CodeGen/AArch64/fast-isel-sqrt.ll @@ -0,0 +1,20 @@ +; RUN: llc -mtriple=arm64-apple-darwin < %s | FileCheck %s +; RUN: llc -mtriple=arm64-apple-darwin -fast-isel -fast-isel-abort < %s | FileCheck %s + +define float @test_sqrt_f32(float %a) { +; CHECK-LABEL: test_sqrt_f32 +; CHECK: fsqrt s0, s0 + %res = call float @llvm.sqrt.f32(float %a) + ret float %res +} +declare float @llvm.sqrt.f32(float) nounwind readnone + +define double @test_sqrt_f64(double %a) { +; CHECK-LABEL: test_sqrt_f64 +; CHECK: fsqrt d0, d0 + %res = call double @llvm.sqrt.f64(double %a) + ret double %res +} +declare double @llvm.sqrt.f64(double) nounwind readnone + + -- 2.7.4