[llvm] 908b3b6 - [AArch64] Use fast-math-flags in isAssociativeAndCommutative
David Green via llvm-commits
llvm-commits at lists.llvm.org
Mon Sep 19 03:34:09 PDT 2022
Author: David Green
Date: 2022-09-19T11:34:00+01:00
New Revision: 908b3b6ccb486396eb22ce4801f094ea6f17e8c5
URL: https://github.com/llvm/llvm-project/commit/908b3b6ccb486396eb22ce4801f094ea6f17e8c5
DIFF: https://github.com/llvm/llvm-project/commit/908b3b6ccb486396eb22ce4801f094ea6f17e8c5.diff
LOG: [AArch64] Use fast-math-flags in isAssociativeAndCommutative
Previously only using the UnsafeFPMath option, this now looks for the
fast moth flags on the instructions, using the same flag flags as other
backends.
Added:
Modified:
llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
llvm/test/CodeGen/AArch64/machine-combiner.ll
Removed:
################################################################################
diff --git a/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp b/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
index 35a37a0519518..4607a5b2867e2 100644
--- a/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
+++ b/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
@@ -4912,7 +4912,9 @@ bool AArch64InstrInfo::isAssociativeAndCommutative(
case AArch64::FMULv2f32:
case AArch64::FMULv2f64:
case AArch64::FMULv4f32:
- return Inst.getParent()->getParent()->getTarget().Options.UnsafeFPMath;
+ return Inst.getParent()->getParent()->getTarget().Options.UnsafeFPMath ||
+ (Inst.getFlag(MachineInstr::MIFlag::FmReassoc) &&
+ Inst.getFlag(MachineInstr::MIFlag::FmNsz));
default:
return false;
}
diff --git a/llvm/test/CodeGen/AArch64/machine-combiner.ll b/llvm/test/CodeGen/AArch64/machine-combiner.ll
index aee9c68d6d917..62430287e30a0 100644
--- a/llvm/test/CodeGen/AArch64/machine-combiner.ll
+++ b/llvm/test/CodeGen/AArch64/machine-combiner.ll
@@ -1,31 +1,82 @@
-; RUN: llc -mtriple=aarch64-gnu-linux -mcpu=cortex-a57 -enable-unsafe-fp-math -disable-post-ra < %s | FileCheck %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=aarch64-gnu-linux -mcpu=cortex-a57 < %s | FileCheck %s --check-prefixes=CHECK,CHECK-STD
+; RUN: llc -mtriple=aarch64-gnu-linux -mcpu=cortex-a57 -enable-unsafe-fp-math < %s | FileCheck %s --check-prefixes=CHECK,CHECK-UNSAFE
; Incremental updates of the instruction depths should be enough for this test
; case.
; RUN: llc -mtriple=aarch64-gnu-linux -mcpu=cortex-a57 -enable-unsafe-fp-math \
-; RUN: -disable-post-ra -machine-combiner-inc-threshold=0 -machine-combiner-verify-pattern-order=true < %s | FileCheck %s
+; RUN: -machine-combiner-inc-threshold=0 -machine-combiner-verify-pattern-order=true < %s | FileCheck %s --check-prefixes=CHECK,CHECK-UNSAFE
; Verify that the first two adds are independent regardless of how the inputs are
; commuted. The destination registers are used as source registers for the third add.
define float @reassociate_adds1(float %x0, float %x1, float %x2, float %x3) {
-; CHECK-LABEL: reassociate_adds1:
-; CHECK: fadd s0, s0, s1
-; CHECK-NEXT: fadd s1, s2, s3
-; CHECK-NEXT: fadd s0, s0, s1
-; CHECK-NEXT: ret
+; CHECK-STD-LABEL: reassociate_adds1:
+; CHECK-STD: // %bb.0:
+; CHECK-STD-NEXT: fadd s0, s0, s1
+; CHECK-STD-NEXT: fadd s0, s0, s2
+; CHECK-STD-NEXT: fadd s0, s0, s3
+; CHECK-STD-NEXT: ret
+;
+; CHECK-UNSAFE-LABEL: reassociate_adds1:
+; CHECK-UNSAFE: // %bb.0:
+; CHECK-UNSAFE-NEXT: fadd s0, s0, s1
+; CHECK-UNSAFE-NEXT: fadd s1, s2, s3
+; CHECK-UNSAFE-NEXT: fadd s0, s0, s1
+; CHECK-UNSAFE-NEXT: ret
%t0 = fadd float %x0, %x1
%t1 = fadd float %t0, %x2
%t2 = fadd float %t1, %x3
ret float %t2
}
-define float @reassociate_adds2(float %x0, float %x1, float %x2, float %x3) {
-; CHECK-LABEL: reassociate_adds2:
-; CHECK: fadd s0, s0, s1
-; CHECK-NEXT: fadd s1, s2, s3
-; CHECK-NEXT: fadd s0, s0, s1
+define float @reassociate_adds1_fast(float %x0, float %x1, float %x2, float %x3) {
+; CHECK-LABEL: reassociate_adds1_fast:
+; CHECK: // %bb.0:
+; CHECK-NEXT: fadd s0, s0, s1
+; CHECK-NEXT: fadd s1, s2, s3
+; CHECK-NEXT: fadd s0, s0, s1
; CHECK-NEXT: ret
+ %t0 = fadd fast float %x0, %x1
+ %t1 = fadd fast float %t0, %x2
+ %t2 = fadd fast float %t1, %x3
+ ret float %t2
+}
+
+define float @reassociate_adds1_reassoc(float %x0, float %x1, float %x2, float %x3) {
+; CHECK-STD-LABEL: reassociate_adds1_reassoc:
+; CHECK-STD: // %bb.0:
+; CHECK-STD-NEXT: fadd s0, s0, s1
+; CHECK-STD-NEXT: fadd s0, s0, s2
+; CHECK-STD-NEXT: fadd s0, s0, s3
+; CHECK-STD-NEXT: ret
+;
+; CHECK-UNSAFE-LABEL: reassociate_adds1_reassoc:
+; CHECK-UNSAFE: // %bb.0:
+; CHECK-UNSAFE-NEXT: fadd s0, s0, s1
+; CHECK-UNSAFE-NEXT: fadd s1, s2, s3
+; CHECK-UNSAFE-NEXT: fadd s0, s0, s1
+; CHECK-UNSAFE-NEXT: ret
+ %t0 = fadd reassoc float %x0, %x1
+ %t1 = fadd reassoc float %t0, %x2
+ %t2 = fadd reassoc float %t1, %x3
+ ret float %t2
+}
+
+define float @reassociate_adds2(float %x0, float %x1, float %x2, float %x3) {
+; CHECK-STD-LABEL: reassociate_adds2:
+; CHECK-STD: // %bb.0:
+; CHECK-STD-NEXT: fadd s0, s0, s1
+; CHECK-STD-NEXT: fadd s0, s2, s0
+; CHECK-STD-NEXT: fadd s0, s0, s3
+; CHECK-STD-NEXT: ret
+;
+; CHECK-UNSAFE-LABEL: reassociate_adds2:
+; CHECK-UNSAFE: // %bb.0:
+; CHECK-UNSAFE-NEXT: fadd s0, s0, s1
+; CHECK-UNSAFE-NEXT: fadd s1, s2, s3
+; CHECK-UNSAFE-NEXT: fadd s0, s0, s1
+; CHECK-UNSAFE-NEXT: ret
%t0 = fadd float %x0, %x1
%t1 = fadd float %x2, %t0
%t2 = fadd float %t1, %x3
@@ -33,11 +84,19 @@ define float @reassociate_adds2(float %x0, float %x1, float %x2, float %x3) {
}
define float @reassociate_adds3(float %x0, float %x1, float %x2, float %x3) {
-; CHECK-LABEL: reassociate_adds3:
-; CHECK: s0, s0, s1
-; CHECK-NEXT: s1, s2, s3
-; CHECK-NEXT: s0, s0, s1
-; CHECK-NEXT: ret
+; CHECK-STD-LABEL: reassociate_adds3:
+; CHECK-STD: // %bb.0:
+; CHECK-STD-NEXT: fadd s0, s0, s1
+; CHECK-STD-NEXT: fadd s0, s0, s2
+; CHECK-STD-NEXT: fadd s0, s3, s0
+; CHECK-STD-NEXT: ret
+;
+; CHECK-UNSAFE-LABEL: reassociate_adds3:
+; CHECK-UNSAFE: // %bb.0:
+; CHECK-UNSAFE-NEXT: fadd s0, s0, s1
+; CHECK-UNSAFE-NEXT: fadd s1, s2, s3
+; CHECK-UNSAFE-NEXT: fadd s0, s0, s1
+; CHECK-UNSAFE-NEXT: ret
%t0 = fadd float %x0, %x1
%t1 = fadd float %t0, %x2
%t2 = fadd float %x3, %t1
@@ -45,11 +104,19 @@ define float @reassociate_adds3(float %x0, float %x1, float %x2, float %x3) {
}
define float @reassociate_adds4(float %x0, float %x1, float %x2, float %x3) {
-; CHECK-LABEL: reassociate_adds4:
-; CHECK: s0, s0, s1
-; CHECK-NEXT: s1, s2, s3
-; CHECK-NEXT: s0, s0, s1
-; CHECK-NEXT: ret
+; CHECK-STD-LABEL: reassociate_adds4:
+; CHECK-STD: // %bb.0:
+; CHECK-STD-NEXT: fadd s0, s0, s1
+; CHECK-STD-NEXT: fadd s0, s2, s0
+; CHECK-STD-NEXT: fadd s0, s3, s0
+; CHECK-STD-NEXT: ret
+;
+; CHECK-UNSAFE-LABEL: reassociate_adds4:
+; CHECK-UNSAFE: // %bb.0:
+; CHECK-UNSAFE-NEXT: fadd s0, s0, s1
+; CHECK-UNSAFE-NEXT: fadd s1, s2, s3
+; CHECK-UNSAFE-NEXT: fadd s0, s0, s1
+; CHECK-UNSAFE-NEXT: ret
%t0 = fadd float %x0, %x1
%t1 = fadd float %x2, %t0
%t2 = fadd float %x3, %t1
@@ -60,15 +127,27 @@ define float @reassociate_adds4(float %x0, float %x1, float %x2, float %x3) {
; produced because that would cost more compile time.
define float @reassociate_adds5(float %x0, float %x1, float %x2, float %x3, float %x4, float %x5, float %x6, float %x7) {
-; CHECK-LABEL: reassociate_adds5:
-; CHECK: fadd s0, s0, s1
-; CHECK-NEXT: fadd s1, s2, s3
-; CHECK-NEXT: fadd s0, s0, s1
-; CHECK-NEXT: fadd s1, s4, s5
-; CHECK-NEXT: fadd s1, s1, s6
-; CHECK-NEXT: fadd s0, s0, s1
-; CHECK-NEXT: fadd s0, s0, s7
-; CHECK-NEXT: ret
+; CHECK-STD-LABEL: reassociate_adds5:
+; CHECK-STD: // %bb.0:
+; CHECK-STD-NEXT: fadd s0, s0, s1
+; CHECK-STD-NEXT: fadd s0, s0, s2
+; CHECK-STD-NEXT: fadd s0, s0, s3
+; CHECK-STD-NEXT: fadd s0, s0, s4
+; CHECK-STD-NEXT: fadd s0, s0, s5
+; CHECK-STD-NEXT: fadd s0, s0, s6
+; CHECK-STD-NEXT: fadd s0, s0, s7
+; CHECK-STD-NEXT: ret
+;
+; CHECK-UNSAFE-LABEL: reassociate_adds5:
+; CHECK-UNSAFE: // %bb.0:
+; CHECK-UNSAFE-NEXT: fadd s0, s0, s1
+; CHECK-UNSAFE-NEXT: fadd s1, s2, s3
+; CHECK-UNSAFE-NEXT: fadd s0, s0, s1
+; CHECK-UNSAFE-NEXT: fadd s1, s4, s5
+; CHECK-UNSAFE-NEXT: fadd s1, s1, s6
+; CHECK-UNSAFE-NEXT: fadd s0, s0, s1
+; CHECK-UNSAFE-NEXT: fadd s0, s0, s7
+; CHECK-UNSAFE-NEXT: ret
%t0 = fadd float %x0, %x1
%t1 = fadd float %t0, %x2
%t2 = fadd float %t1, %x3
@@ -85,11 +164,19 @@ define float @reassociate_adds5(float %x0, float %x1, float %x2, float %x3, floa
; division. The latter reassociation would not improve anything.
define float @reassociate_adds6(float %x0, float %x1, float %x2, float %x3) {
-; CHECK-LABEL: reassociate_adds6:
-; CHECK: fdiv s0, s0, s1
-; CHECK-NEXT: fadd s1, s2, s3
-; CHECK-NEXT: fadd s0, s0, s1
-; CHECK-NEXT: ret
+; CHECK-STD-LABEL: reassociate_adds6:
+; CHECK-STD: // %bb.0:
+; CHECK-STD-NEXT: fdiv s0, s0, s1
+; CHECK-STD-NEXT: fadd s0, s2, s0
+; CHECK-STD-NEXT: fadd s0, s3, s0
+; CHECK-STD-NEXT: ret
+;
+; CHECK-UNSAFE-LABEL: reassociate_adds6:
+; CHECK-UNSAFE: // %bb.0:
+; CHECK-UNSAFE-NEXT: fdiv s0, s0, s1
+; CHECK-UNSAFE-NEXT: fadd s1, s2, s3
+; CHECK-UNSAFE-NEXT: fadd s0, s0, s1
+; CHECK-UNSAFE-NEXT: ret
%t0 = fdiv float %x0, %x1
%t1 = fadd float %x2, %t0
%t2 = fadd float %x3, %t1
@@ -99,11 +186,19 @@ define float @reassociate_adds6(float %x0, float %x1, float %x2, float %x3) {
; Verify that scalar single-precision multiplies are reassociated.
define float @reassociate_muls1(float %x0, float %x1, float %x2, float %x3) {
-; CHECK-LABEL: reassociate_muls1:
-; CHECK: fdiv s0, s0, s1
-; CHECK-NEXT: fmul s1, s2, s3
-; CHECK-NEXT: fmul s0, s0, s1
-; CHECK-NEXT: ret
+; CHECK-STD-LABEL: reassociate_muls1:
+; CHECK-STD: // %bb.0:
+; CHECK-STD-NEXT: fdiv s0, s0, s1
+; CHECK-STD-NEXT: fmul s1, s2, s0
+; CHECK-STD-NEXT: fmul s0, s3, s1
+; CHECK-STD-NEXT: ret
+;
+; CHECK-UNSAFE-LABEL: reassociate_muls1:
+; CHECK-UNSAFE: // %bb.0:
+; CHECK-UNSAFE-NEXT: fdiv s0, s0, s1
+; CHECK-UNSAFE-NEXT: fmul s1, s2, s3
+; CHECK-UNSAFE-NEXT: fmul s0, s0, s1
+; CHECK-UNSAFE-NEXT: ret
%t0 = fdiv float %x0, %x1
%t1 = fmul float %x2, %t0
%t2 = fmul float %x3, %t1
@@ -113,11 +208,19 @@ define float @reassociate_muls1(float %x0, float %x1, float %x2, float %x3) {
; Verify that scalar double-precision adds are reassociated.
define double @reassociate_adds_double(double %x0, double %x1, double %x2, double %x3) {
-; CHECK-LABEL: reassociate_adds_double:
-; CHECK: fdiv d0, d0, d1
-; CHECK-NEXT: fadd d1, d2, d3
-; CHECK-NEXT: fadd d0, d0, d1
-; CHECK-NEXT: ret
+; CHECK-STD-LABEL: reassociate_adds_double:
+; CHECK-STD: // %bb.0:
+; CHECK-STD-NEXT: fdiv d0, d0, d1
+; CHECK-STD-NEXT: fadd d0, d2, d0
+; CHECK-STD-NEXT: fadd d0, d3, d0
+; CHECK-STD-NEXT: ret
+;
+; CHECK-UNSAFE-LABEL: reassociate_adds_double:
+; CHECK-UNSAFE: // %bb.0:
+; CHECK-UNSAFE-NEXT: fdiv d0, d0, d1
+; CHECK-UNSAFE-NEXT: fadd d1, d2, d3
+; CHECK-UNSAFE-NEXT: fadd d0, d0, d1
+; CHECK-UNSAFE-NEXT: ret
%t0 = fdiv double %x0, %x1
%t1 = fadd double %x2, %t0
%t2 = fadd double %x3, %t1
@@ -127,11 +230,19 @@ define double @reassociate_adds_double(double %x0, double %x1, double %x2, doubl
; Verify that scalar double-precision multiplies are reassociated.
define double @reassociate_muls_double(double %x0, double %x1, double %x2, double %x3) {
-; CHECK-LABEL: reassociate_muls_double:
-; CHECK: fdiv d0, d0, d1
-; CHECK-NEXT: fmul d1, d2, d3
-; CHECK-NEXT: fmul d0, d0, d1
-; CHECK-NEXT: ret
+; CHECK-STD-LABEL: reassociate_muls_double:
+; CHECK-STD: // %bb.0:
+; CHECK-STD-NEXT: fdiv d0, d0, d1
+; CHECK-STD-NEXT: fmul d1, d2, d0
+; CHECK-STD-NEXT: fmul d0, d3, d1
+; CHECK-STD-NEXT: ret
+;
+; CHECK-UNSAFE-LABEL: reassociate_muls_double:
+; CHECK-UNSAFE: // %bb.0:
+; CHECK-UNSAFE-NEXT: fdiv d0, d0, d1
+; CHECK-UNSAFE-NEXT: fmul d1, d2, d3
+; CHECK-UNSAFE-NEXT: fmul d0, d0, d1
+; CHECK-UNSAFE-NEXT: ret
%t0 = fdiv double %x0, %x1
%t1 = fmul double %x2, %t0
%t2 = fmul double %x3, %t1
@@ -141,11 +252,19 @@ define double @reassociate_muls_double(double %x0, double %x1, double %x2, doubl
; Verify that we reassociate vector instructions too.
define <4 x float> @vector_reassociate_adds1(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, <4 x float> %x3) {
-; CHECK-LABEL: vector_reassociate_adds1:
-; CHECK: fadd v0.4s, v0.4s, v1.4s
-; CHECK-NEXT: fadd v1.4s, v2.4s, v3.4s
-; CHECK-NEXT: fadd v0.4s, v0.4s, v1.4s
-; CHECK-NEXT: ret
+; CHECK-STD-LABEL: vector_reassociate_adds1:
+; CHECK-STD: // %bb.0:
+; CHECK-STD-NEXT: fadd v0.4s, v0.4s, v1.4s
+; CHECK-STD-NEXT: fadd v0.4s, v0.4s, v2.4s
+; CHECK-STD-NEXT: fadd v0.4s, v0.4s, v3.4s
+; CHECK-STD-NEXT: ret
+;
+; CHECK-UNSAFE-LABEL: vector_reassociate_adds1:
+; CHECK-UNSAFE: // %bb.0:
+; CHECK-UNSAFE-NEXT: fadd v0.4s, v0.4s, v1.4s
+; CHECK-UNSAFE-NEXT: fadd v1.4s, v2.4s, v3.4s
+; CHECK-UNSAFE-NEXT: fadd v0.4s, v0.4s, v1.4s
+; CHECK-UNSAFE-NEXT: ret
%t0 = fadd <4 x float> %x0, %x1
%t1 = fadd <4 x float> %t0, %x2
%t2 = fadd <4 x float> %t1, %x3
@@ -153,10 +272,19 @@ define <4 x float> @vector_reassociate_adds1(<4 x float> %x0, <4 x float> %x1, <
}
define <4 x float> @vector_reassociate_adds2(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, <4 x float> %x3) {
-; CHECK-LABEL: vector_reassociate_adds2:
-; CHECK: fadd v0.4s, v0.4s, v1.4s
-; CHECK-NEXT: fadd v1.4s, v2.4s, v3.4s
-; CHECK-NEXT: fadd v0.4s, v0.4s, v1.4s
+; CHECK-STD-LABEL: vector_reassociate_adds2:
+; CHECK-STD: // %bb.0:
+; CHECK-STD-NEXT: fadd v0.4s, v0.4s, v1.4s
+; CHECK-STD-NEXT: fadd v0.4s, v2.4s, v0.4s
+; CHECK-STD-NEXT: fadd v0.4s, v0.4s, v3.4s
+; CHECK-STD-NEXT: ret
+;
+; CHECK-UNSAFE-LABEL: vector_reassociate_adds2:
+; CHECK-UNSAFE: // %bb.0:
+; CHECK-UNSAFE-NEXT: fadd v0.4s, v0.4s, v1.4s
+; CHECK-UNSAFE-NEXT: fadd v1.4s, v2.4s, v3.4s
+; CHECK-UNSAFE-NEXT: fadd v0.4s, v0.4s, v1.4s
+; CHECK-UNSAFE-NEXT: ret
%t0 = fadd <4 x float> %x0, %x1
%t1 = fadd <4 x float> %x2, %t0
%t2 = fadd <4 x float> %t1, %x3
@@ -164,10 +292,19 @@ define <4 x float> @vector_reassociate_adds2(<4 x float> %x0, <4 x float> %x1, <
}
define <4 x float> @vector_reassociate_adds3(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, <4 x float> %x3) {
-; CHECK-LABEL: vector_reassociate_adds3:
-; CHECK: fadd v0.4s, v0.4s, v1.4s
-; CHECK-NEXT: fadd v1.4s, v2.4s, v3.4s
-; CHECK-NEXT: fadd v0.4s, v0.4s, v1.4s
+; CHECK-STD-LABEL: vector_reassociate_adds3:
+; CHECK-STD: // %bb.0:
+; CHECK-STD-NEXT: fadd v0.4s, v0.4s, v1.4s
+; CHECK-STD-NEXT: fadd v0.4s, v0.4s, v2.4s
+; CHECK-STD-NEXT: fadd v0.4s, v3.4s, v0.4s
+; CHECK-STD-NEXT: ret
+;
+; CHECK-UNSAFE-LABEL: vector_reassociate_adds3:
+; CHECK-UNSAFE: // %bb.0:
+; CHECK-UNSAFE-NEXT: fadd v0.4s, v0.4s, v1.4s
+; CHECK-UNSAFE-NEXT: fadd v1.4s, v2.4s, v3.4s
+; CHECK-UNSAFE-NEXT: fadd v0.4s, v0.4s, v1.4s
+; CHECK-UNSAFE-NEXT: ret
%t0 = fadd <4 x float> %x0, %x1
%t1 = fadd <4 x float> %t0, %x2
%t2 = fadd <4 x float> %x3, %t1
@@ -175,10 +312,19 @@ define <4 x float> @vector_reassociate_adds3(<4 x float> %x0, <4 x float> %x1, <
}
define <4 x float> @vector_reassociate_adds4(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, <4 x float> %x3) {
-; CHECK-LABEL: vector_reassociate_adds4:
-; CHECK: fadd v0.4s, v0.4s, v1.4s
-; CHECK-NEXT: fadd v1.4s, v2.4s, v3.4s
-; CHECK-NEXT: fadd v0.4s, v0.4s, v1.4s
+; CHECK-STD-LABEL: vector_reassociate_adds4:
+; CHECK-STD: // %bb.0:
+; CHECK-STD-NEXT: fadd v0.4s, v0.4s, v1.4s
+; CHECK-STD-NEXT: fadd v0.4s, v2.4s, v0.4s
+; CHECK-STD-NEXT: fadd v0.4s, v3.4s, v0.4s
+; CHECK-STD-NEXT: ret
+;
+; CHECK-UNSAFE-LABEL: vector_reassociate_adds4:
+; CHECK-UNSAFE: // %bb.0:
+; CHECK-UNSAFE-NEXT: fadd v0.4s, v0.4s, v1.4s
+; CHECK-UNSAFE-NEXT: fadd v1.4s, v2.4s, v3.4s
+; CHECK-UNSAFE-NEXT: fadd v0.4s, v0.4s, v1.4s
+; CHECK-UNSAFE-NEXT: ret
%t0 = fadd <4 x float> %x0, %x1
%t1 = fadd <4 x float> %x2, %t0
%t2 = fadd <4 x float> %x3, %t1
@@ -187,11 +333,19 @@ define <4 x float> @vector_reassociate_adds4(<4 x float> %x0, <4 x float> %x1, <
; Verify that 128-bit vector single-precision multiplies are reassociated.
define <4 x float> @reassociate_muls_v4f32(<4 x float> %x0, <4 x float> %x1, <4 x float> %x2, <4 x float> %x3) {
-; CHECK-LABEL: reassociate_muls_v4f32:
-; CHECK: fadd v0.4s, v0.4s, v1.4s
-; CHECK-NEXT: fmul v1.4s, v2.4s, v3.4s
-; CHECK-NEXT: fmul v0.4s, v0.4s, v1.4s
-; CHECK-NEXT: ret
+; CHECK-STD-LABEL: reassociate_muls_v4f32:
+; CHECK-STD: // %bb.0:
+; CHECK-STD-NEXT: fadd v0.4s, v0.4s, v1.4s
+; CHECK-STD-NEXT: fmul v0.4s, v2.4s, v0.4s
+; CHECK-STD-NEXT: fmul v0.4s, v3.4s, v0.4s
+; CHECK-STD-NEXT: ret
+;
+; CHECK-UNSAFE-LABEL: reassociate_muls_v4f32:
+; CHECK-UNSAFE: // %bb.0:
+; CHECK-UNSAFE-NEXT: fadd v0.4s, v0.4s, v1.4s
+; CHECK-UNSAFE-NEXT: fmul v1.4s, v2.4s, v3.4s
+; CHECK-UNSAFE-NEXT: fmul v0.4s, v0.4s, v1.4s
+; CHECK-UNSAFE-NEXT: ret
%t0 = fadd <4 x float> %x0, %x1
%t1 = fmul <4 x float> %x2, %t0
%t2 = fmul <4 x float> %x3, %t1
@@ -201,11 +355,19 @@ define <4 x float> @reassociate_muls_v4f32(<4 x float> %x0, <4 x float> %x1, <4
; Verify that 128-bit vector double-precision multiplies are reassociated.
define <2 x double> @reassociate_muls_v2f64(<2 x double> %x0, <2 x double> %x1, <2 x double> %x2, <2 x double> %x3) {
-; CHECK-LABEL: reassociate_muls_v2f64:
-; CHECK: fadd v0.2d, v0.2d, v1.2d
-; CHECK-NEXT: fmul v1.2d, v2.2d, v3.2d
-; CHECK-NEXT: fmul v0.2d, v0.2d, v1.2d
-; CHECK-NEXT: ret
+; CHECK-STD-LABEL: reassociate_muls_v2f64:
+; CHECK-STD: // %bb.0:
+; CHECK-STD-NEXT: fadd v0.2d, v0.2d, v1.2d
+; CHECK-STD-NEXT: fmul v0.2d, v2.2d, v0.2d
+; CHECK-STD-NEXT: fmul v0.2d, v3.2d, v0.2d
+; CHECK-STD-NEXT: ret
+;
+; CHECK-UNSAFE-LABEL: reassociate_muls_v2f64:
+; CHECK-UNSAFE: // %bb.0:
+; CHECK-UNSAFE-NEXT: fadd v0.2d, v0.2d, v1.2d
+; CHECK-UNSAFE-NEXT: fmul v1.2d, v2.2d, v3.2d
+; CHECK-UNSAFE-NEXT: fmul v0.2d, v0.2d, v1.2d
+; CHECK-UNSAFE-NEXT: ret
%t0 = fadd <2 x double> %x0, %x1
%t1 = fmul <2 x double> %x2, %t0
%t2 = fmul <2 x double> %x3, %t1
@@ -218,17 +380,55 @@ define <2 x double> @reassociate_muls_v2f64(<2 x double> %x0, <2 x double> %x1,
declare double @bar()
define double @reassociate_adds_from_calls() {
-; CHECK-LABEL: reassociate_adds_from_calls:
-; CHECK: bl bar
-; CHECK-NEXT: fmov d8, d0
-; CHECK-NEXT: bl bar
-; CHECK-NEXT: fmov d9, d0
-; CHECK-NEXT: bl bar
-; CHECK-NEXT: fmov d10, d0
-; CHECK-NEXT: bl bar
-; CHECK: fadd d1, d8, d9
-; CHECK-NEXT: fadd d0, d10, d0
-; CHECK-NEXT: fadd d0, d1, d0
+; CHECK-STD-LABEL: reassociate_adds_from_calls:
+; CHECK-STD: // %bb.0:
+; CHECK-STD-NEXT: str d10, [sp, #-32]! // 8-byte Folded Spill
+; CHECK-STD-NEXT: .cfi_def_cfa_offset 32
+; CHECK-STD-NEXT: stp d9, d8, [sp, #8] // 16-byte Folded Spill
+; CHECK-STD-NEXT: str x30, [sp, #24] // 8-byte Folded Spill
+; CHECK-STD-NEXT: .cfi_offset w30, -8
+; CHECK-STD-NEXT: .cfi_offset b8, -16
+; CHECK-STD-NEXT: .cfi_offset b9, -24
+; CHECK-STD-NEXT: .cfi_offset b10, -32
+; CHECK-STD-NEXT: bl bar
+; CHECK-STD-NEXT: fmov d8, d0
+; CHECK-STD-NEXT: bl bar
+; CHECK-STD-NEXT: fmov d9, d0
+; CHECK-STD-NEXT: bl bar
+; CHECK-STD-NEXT: fmov d10, d0
+; CHECK-STD-NEXT: bl bar
+; CHECK-STD-NEXT: fadd d1, d8, d9
+; CHECK-STD-NEXT: ldp d9, d8, [sp, #8] // 16-byte Folded Reload
+; CHECK-STD-NEXT: ldr x30, [sp, #24] // 8-byte Folded Reload
+; CHECK-STD-NEXT: fadd d1, d1, d10
+; CHECK-STD-NEXT: fadd d0, d1, d0
+; CHECK-STD-NEXT: ldr d10, [sp], #32 // 8-byte Folded Reload
+; CHECK-STD-NEXT: ret
+;
+; CHECK-UNSAFE-LABEL: reassociate_adds_from_calls:
+; CHECK-UNSAFE: // %bb.0:
+; CHECK-UNSAFE-NEXT: str d10, [sp, #-32]! // 8-byte Folded Spill
+; CHECK-UNSAFE-NEXT: .cfi_def_cfa_offset 32
+; CHECK-UNSAFE-NEXT: stp d9, d8, [sp, #8] // 16-byte Folded Spill
+; CHECK-UNSAFE-NEXT: str x30, [sp, #24] // 8-byte Folded Spill
+; CHECK-UNSAFE-NEXT: .cfi_offset w30, -8
+; CHECK-UNSAFE-NEXT: .cfi_offset b8, -16
+; CHECK-UNSAFE-NEXT: .cfi_offset b9, -24
+; CHECK-UNSAFE-NEXT: .cfi_offset b10, -32
+; CHECK-UNSAFE-NEXT: bl bar
+; CHECK-UNSAFE-NEXT: fmov d8, d0
+; CHECK-UNSAFE-NEXT: bl bar
+; CHECK-UNSAFE-NEXT: fmov d9, d0
+; CHECK-UNSAFE-NEXT: bl bar
+; CHECK-UNSAFE-NEXT: fmov d10, d0
+; CHECK-UNSAFE-NEXT: bl bar
+; CHECK-UNSAFE-NEXT: fadd d1, d8, d9
+; CHECK-UNSAFE-NEXT: fadd d0, d10, d0
+; CHECK-UNSAFE-NEXT: ldr x30, [sp, #24] // 8-byte Folded Reload
+; CHECK-UNSAFE-NEXT: ldp d9, d8, [sp, #8] // 16-byte Folded Reload
+; CHECK-UNSAFE-NEXT: fadd d0, d1, d0
+; CHECK-UNSAFE-NEXT: ldr d10, [sp], #32 // 8-byte Folded Reload
+; CHECK-UNSAFE-NEXT: ret
%x0 = call double @bar()
%x1 = call double @bar()
%x2 = call double @bar()
@@ -241,16 +441,29 @@ define double @reassociate_adds_from_calls() {
define double @already_reassociated() {
; CHECK-LABEL: already_reassociated:
-; CHECK: bl bar
-; CHECK-NEXT: fmov d8, d0
-; CHECK-NEXT: bl bar
-; CHECK-NEXT: fmov d9, d0
-; CHECK-NEXT: bl bar
-; CHECK-NEXT: fmov d10, d0
-; CHECK-NEXT: bl bar
-; CHECK: fadd d1, d8, d9
-; CHECK-NEXT: fadd d0, d10, d0
-; CHECK-NEXT: fadd d0, d1, d0
+; CHECK: // %bb.0:
+; CHECK-NEXT: str d10, [sp, #-32]! // 8-byte Folded Spill
+; CHECK-NEXT: .cfi_def_cfa_offset 32
+; CHECK-NEXT: stp d9, d8, [sp, #8] // 16-byte Folded Spill
+; CHECK-NEXT: str x30, [sp, #24] // 8-byte Folded Spill
+; CHECK-NEXT: .cfi_offset w30, -8
+; CHECK-NEXT: .cfi_offset b8, -16
+; CHECK-NEXT: .cfi_offset b9, -24
+; CHECK-NEXT: .cfi_offset b10, -32
+; CHECK-NEXT: bl bar
+; CHECK-NEXT: fmov d8, d0
+; CHECK-NEXT: bl bar
+; CHECK-NEXT: fmov d9, d0
+; CHECK-NEXT: bl bar
+; CHECK-NEXT: fmov d10, d0
+; CHECK-NEXT: bl bar
+; CHECK-NEXT: fadd d1, d8, d9
+; CHECK-NEXT: fadd d0, d10, d0
+; CHECK-NEXT: ldr x30, [sp, #24] // 8-byte Folded Reload
+; CHECK-NEXT: ldp d9, d8, [sp, #8] // 16-byte Folded Reload
+; CHECK-NEXT: fadd d0, d1, d0
+; CHECK-NEXT: ldr d10, [sp], #32 // 8-byte Folded Reload
+; CHECK-NEXT: ret
%x0 = call double @bar()
%x1 = call double @bar()
%x2 = call double @bar()
More information about the llvm-commits
mailing list