| commit 0ec57972967dfb43fc022c2e3788be041d1db730 |
| Author: John Brawn <john.brawn@arm.com> |
| Date: Thu Feb 13 14:52:33 2020 +0000 |
| |
| [ARM] Fix infinite loop when lowering STRICT_FP_EXTEND |
| |
| If the target has FP64 but not FP16 then we have custom lowering for FP_EXTEND |
| and STRICT_FP_EXTEND with type f64. However if the extend is from f32 to f64 the |
| current implementation will cause in infinite loop for STRICT_FP_EXTEND due to |
| emitting a merge_values of the original node which after replacement becomes a |
| merge_values of itself. |
| |
| Fix this by not doing anything for f32 to f64 extend when we have FP64, though |
| for STRICT_FP_EXTEND we have to do the strict-to-nonstrict mutation as that |
| doesn't happen automatically for opcodes with custom lowering. |
| |
| Differential Revision: https://reviews.llvm.org/D74559 |
| |
| diff --git a/llvm/lib/Target/ARM/ARMISelLowering.cpp b/llvm/lib/Target/ARM/ARMISelLowering.cpp |
| index e42f4deae0c..5c01ac1abdc 100644 |
| --- a/llvm/lib/Target/ARM/ARMISelLowering.cpp |
| +++ b/llvm/lib/Target/ARM/ARMISelLowering.cpp |
| @@ -16528,6 +16528,15 @@ SDValue ARMTargetLowering::LowerFP_EXTEND(SDValue Op, SelectionDAG &DAG) const { |
| assert(!(DstSz == 32 && Subtarget->hasFP16()) && |
| "With FP16, 16 to 32 conversion is legal!"); |
| |
| + // Converting from 32 -> 64 is valid if we have FP64. |
| + if (SrcSz == 32 && DstSz == 64 && Subtarget->hasFP64()) { |
| + // FIXME: Remove this when we have strict fp instruction selection patterns |
| + if (IsStrict) { |
| + DAG.mutateStrictFPToFP(Op.getNode()); |
| + } |
| + return Op; |
| + } |
| + |
| // Either we are converting from 16 -> 64, without FP16 and/or |
| // FP.double-precision or without Armv8-fp. So we must do it in two |
| // steps. |
| diff --git a/llvm/test/CodeGen/ARM/fp-intrinsics.ll b/llvm/test/CodeGen/ARM/fp-intrinsics.ll |
| index 4352548ef3c..c5746ef7b30 100644 |
| --- a/llvm/test/CodeGen/ARM/fp-intrinsics.ll |
| +++ b/llvm/test/CodeGen/ARM/fp-intrinsics.ll |
| @@ -1,7 +1,11 @@ |
| -; RUN: llc -mtriple=armv8a-none-eabi %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-SP,CHECK-DP |
| +; RUN: llc -mtriple=armv8a-none-eabi %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-SP,CHECK-DP,CHECK-SP-V8,CHECK-DP-V8 |
| ; RUN: llc -mtriple=thumbv8m.main-none-eabi %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-NOSP,CHECK-NODP |
| -; RUN: llc -mtriple=thumbv8m.main-none-eabi %s -o - -mattr=fp-armv8 | FileCheck %s --check-prefixes=CHECK,CHECK-SP,CHECK-DP |
| -; RUN: llc -mtriple=thumbv8m.main-none-eabi %s -o - -mattr=fp-armv8sp | FileCheck %s --check-prefixes=CHECK,CHECK-SP,CHECK-NODP |
| +; RUN: llc -mtriple=thumbv8m.main-none-eabi %s -o - -mattr=fp-armv8 | FileCheck %s --check-prefixes=CHECK,CHECK-SP,CHECK-DP,CHECK-SP-V8,CHECK-DP-V8 |
| +; RUN: llc -mtriple=thumbv8m.main-none-eabi %s -o - -mattr=fp-armv8sp | FileCheck %s --check-prefixes=CHECK,CHECK-SP,CHECK-NODP,CHECK-SP-V8 |
| +; RUN: llc -mtriple=armv7a-none-eabi %s -o - -mattr=vfp4 | FileCheck %s --check-prefixes=CHECK,CHECK-SP,CHECK-DP,CHECK-SP-NOV8,CHECK-DP-NOV8 |
| +; RUN: llc -mtriple=thumbv7m-none-eabi %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-NOSP,CHECK-NODP |
| +; RUN: llc -mtriple=thumbv7m-none-eabi %s -o - -mattr=vfp4 | FileCheck %s --check-prefixes=CHECK,CHECK-SP,CHECK-DP,CHECK-SP-NOV8,CHECK-DP-NOV8 |
| +; RUN: llc -mtriple=thumbv7m-none-eabi %s -o - -mattr=vfp4sp | FileCheck %s --check-prefixes=CHECK,CHECK-SP,CHECK-NODP,CHECK-SP-NOV8 |
| |
| ; Check that constrained fp intrinsics are correctly lowered. In particular |
| ; check that the valid combinations of single-precision and double-precision |
| @@ -149,7 +153,8 @@ define float @exp2_f32(float %x) #0 { |
| |
| ; CHECK-LABEL: rint_f32: |
| ; CHECK-NOSP: bl rintf |
| -; CHECK-SP: vrintx.f32 |
| +; CHECK-SP-NOV8: bl rintf |
| +; CHECK-SP-V8: vrintx.f32 |
| define float @rint_f32(float %x) #0 { |
| %val = call float @llvm.experimental.constrained.rint.f32(float %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 |
| ret float %val |
| @@ -157,7 +162,8 @@ define float @rint_f32(float %x) #0 { |
| |
| ; CHECK-LABEL: nearbyint_f32: |
| ; CHECK-NOSP: bl nearbyintf |
| -; CHECK-SP: vrintr.f32 |
| +; CHECK-SP-NOV8: bl nearbyintf |
| +; CHECK-SP-V8: vrintr.f32 |
| define float @nearbyint_f32(float %x) #0 { |
| %val = call float @llvm.experimental.constrained.nearbyint.f32(float %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 |
| ret float %val |
| @@ -179,7 +185,8 @@ define i32 @llrint_f32(float %x) #0 { |
| |
| ; CHECK-LABEL: maxnum_f32: |
| ; CHECK-NOSP: bl fmaxf |
| -; CHECK-SP: vmaxnm.f32 |
| +; CHECK-SP-NOV8: bl fmaxf |
| +; CHECK-SP-V8: vmaxnm.f32 |
| define float @maxnum_f32(float %x, float %y) #0 { |
| %val = call float @llvm.experimental.constrained.maxnum.f32(float %x, float %y, metadata !"fpexcept.strict") #0 |
| ret float %val |
| @@ -187,7 +194,8 @@ define float @maxnum_f32(float %x, float %y) #0 { |
| |
| ; CHECK-LABEL: minnum_f32: |
| ; CHECK-NOSP: bl fminf |
| -; CHECK-SP: vminnm.f32 |
| +; CHECK-SP-NOV8: bl fminf |
| +; CHECK-SP-V8: vminnm.f32 |
| define float @minnum_f32(float %x, float %y) #0 { |
| %val = call float @llvm.experimental.constrained.minnum.f32(float %x, float %y, metadata !"fpexcept.strict") #0 |
| ret float %val |
| @@ -195,7 +203,8 @@ define float @minnum_f32(float %x, float %y) #0 { |
| |
| ; CHECK-LABEL: ceil_f32: |
| ; CHECK-NOSP: bl ceilf |
| -; CHECK-SP: vrintp.f32 |
| +; CHECK-SP-NOV8: bl ceilf |
| +; CHECK-SP-V8: vrintp.f32 |
| define float @ceil_f32(float %x) #0 { |
| %val = call float @llvm.experimental.constrained.ceil.f32(float %x, metadata !"fpexcept.strict") #0 |
| ret float %val |
| @@ -203,7 +212,8 @@ define float @ceil_f32(float %x) #0 { |
| |
| ; CHECK-LABEL: floor_f32: |
| ; CHECK-NOSP: bl floorf |
| -; CHECK-SP: vrintm.f32 |
| +; CHECK-SP-NOV8: bl floorf |
| +; CHECK-SP-V8: vrintm.f32 |
| define float @floor_f32(float %x) #0 { |
| %val = call float @llvm.experimental.constrained.floor.f32(float %x, metadata !"fpexcept.strict") #0 |
| ret float %val |
| @@ -225,7 +235,8 @@ define i32 @llround_f32(float %x) #0 { |
| |
| ; CHECK-LABEL: round_f32: |
| ; CHECK-NOSP: bl roundf |
| -; CHECK-SP: vrinta.f32 |
| +; CHECK-SP-NOV8: bl roundf |
| +; CHECK-SP-V8: vrinta.f32 |
| define float @round_f32(float %x) #0 { |
| %val = call float @llvm.experimental.constrained.round.f32(float %x, metadata !"fpexcept.strict") #0 |
| ret float %val |
| @@ -233,7 +244,8 @@ define float @round_f32(float %x) #0 { |
| |
| ; CHECK-LABEL: trunc_f32: |
| ; CHECK-NOSP: bl truncf |
| -; CHECK-SP: vrintz.f32 |
| +; CHECK-SP-NOV8: bl truncf |
| +; CHECK-SP-V8: vrintz.f32 |
| define float @trunc_f32(float %x) #0 { |
| %val = call float @llvm.experimental.constrained.trunc.f32(float %x, metadata !"fpexcept.strict") #0 |
| ret float %val |
| @@ -598,7 +610,8 @@ define double @exp2_f64(double %x) #0 { |
| |
| ; CHECK-LABEL: rint_f64: |
| ; CHECK-NODP: bl rint |
| -; CHECK-DP: vrintx.f64 |
| +; CHECK-DP-NOV8: bl rint |
| +; CHECK-DP-V8: vrintx.f64 |
| define double @rint_f64(double %x) #0 { |
| %val = call double @llvm.experimental.constrained.rint.f64(double %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 |
| ret double %val |
| @@ -606,7 +619,8 @@ define double @rint_f64(double %x) #0 { |
| |
| ; CHECK-LABEL: nearbyint_f64: |
| ; CHECK-NODP: bl nearbyint |
| -; CHECK-DP: vrintr.f64 |
| +; CHECK-DP-NOV8: bl nearbyint |
| +; CHECK-DP-V8: vrintr.f64 |
| define double @nearbyint_f64(double %x) #0 { |
| %val = call double @llvm.experimental.constrained.nearbyint.f64(double %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 |
| ret double %val |
| @@ -628,7 +642,8 @@ define i32 @llrint_f64(double %x) #0 { |
| |
| ; CHECK-LABEL: maxnum_f64: |
| ; CHECK-NODP: bl fmax |
| -; CHECK-DP: vmaxnm.f64 |
| +; CHECK-DP-NOV8: bl fmax |
| +; CHECK-DP-V8: vmaxnm.f64 |
| define double @maxnum_f64(double %x, double %y) #0 { |
| %val = call double @llvm.experimental.constrained.maxnum.f64(double %x, double %y, metadata !"fpexcept.strict") #0 |
| ret double %val |
| @@ -636,7 +651,8 @@ define double @maxnum_f64(double %x, double %y) #0 { |
| |
| ; CHECK-LABEL: minnum_f64: |
| ; CHECK-NODP: bl fmin |
| -; CHECK-DP: vminnm.f64 |
| +; CHECK-DP-NOV8: bl fmin |
| +; CHECK-DP-V8: vminnm.f64 |
| define double @minnum_f64(double %x, double %y) #0 { |
| %val = call double @llvm.experimental.constrained.minnum.f64(double %x, double %y, metadata !"fpexcept.strict") #0 |
| ret double %val |
| @@ -644,7 +660,8 @@ define double @minnum_f64(double %x, double %y) #0 { |
| |
| ; CHECK-LABEL: ceil_f64: |
| ; CHECK-NODP: bl ceil |
| -; CHECK-DP: vrintp.f64 |
| +; CHECK-DP-NOV8: bl ceil |
| +; CHECK-DP-V8: vrintp.f64 |
| define double @ceil_f64(double %x) #0 { |
| %val = call double @llvm.experimental.constrained.ceil.f64(double %x, metadata !"fpexcept.strict") #0 |
| ret double %val |
| @@ -652,7 +669,8 @@ define double @ceil_f64(double %x) #0 { |
| |
| ; CHECK-LABEL: floor_f64: |
| ; CHECK-NODP: bl floor |
| -; CHECK-DP: vrintm.f64 |
| +; CHECK-DP-NOV8: bl floor |
| +; CHECK-DP-V8: vrintm.f64 |
| define double @floor_f64(double %x) #0 { |
| %val = call double @llvm.experimental.constrained.floor.f64(double %x, metadata !"fpexcept.strict") #0 |
| ret double %val |
| @@ -674,7 +692,8 @@ define i32 @llround_f64(double %x) #0 { |
| |
| ; CHECK-LABEL: round_f64: |
| ; CHECK-NODP: bl round |
| -; CHECK-DP: vrinta.f64 |
| +; CHECK-DP-NOV8: bl round |
| +; CHECK-DP-V8: vrinta.f64 |
| define double @round_f64(double %x) #0 { |
| %val = call double @llvm.experimental.constrained.round.f64(double %x, metadata !"fpexcept.strict") #0 |
| ret double %val |
| @@ -682,7 +701,8 @@ define double @round_f64(double %x) #0 { |
| |
| ; CHECK-LABEL: trunc_f64: |
| ; CHECK-NODP: bl trunc |
| -; CHECK-DP: vrintz.f64 |
| +; CHECK-DP-NOV8: bl trunc |
| +; CHECK-DP-V8: vrintz.f64 |
| define double @trunc_f64(double %x) #0 { |
| %val = call double @llvm.experimental.constrained.trunc.f64(double %x, metadata !"fpexcept.strict") #0 |
| ret double %val |