maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/test/CodeGen/ARM/fnmscs.ll

167 lines
3.2 KiB
LLVM
Raw Normal View History

[ARM] Add missing selection patterns for vnmla For the following function: double fn1(double d0, double d1, double d2) { double a = -d0 - d1 * d2; return a; } on ARM, LLVM generates code along the lines of vneg.f64 d0, d0 vmls.f64 d0, d1, d2 i.e., a negate and a multiply-subtract. The attached patch adds instruction selection patterns to allow it to generate the single instruction vnmla.f64 d0, d1, d2 (multiply-add with negation) instead, like GCC does. Committed on behalf of @gergo- (Gergö Barany) Differential Revision: https://reviews.llvm.org/D35911 llvm-svn: 313972
2017-09-22 17:50:52 +08:00
; RUN: llc -mtriple=arm-eabihf -mattr=+vfp2 %s -o - \
ARM: fixup more tests to specify the target more explicitly This changes the tests that were targeting ARM EABI to explicitly specify the environment rather than relying on the default. This breaks with the new Windows on ARM support when running the tests on Windows where the default environment is no longer EABI. Take the opportunity to avoid a pointless redirect (helps when trying to debug with providing a command line invocation which can be copy and pasted) and removing a few greps in favour of FileCheck. llvm-svn: 205541
2014-04-04 00:01:44 +08:00
; RUN: | FileCheck %s -check-prefix=VFP2
[ARM] Add missing selection patterns for vnmla For the following function: double fn1(double d0, double d1, double d2) { double a = -d0 - d1 * d2; return a; } on ARM, LLVM generates code along the lines of vneg.f64 d0, d0 vmls.f64 d0, d1, d2 i.e., a negate and a multiply-subtract. The attached patch adds instruction selection patterns to allow it to generate the single instruction vnmla.f64 d0, d1, d2 (multiply-add with negation) instead, like GCC does. Committed on behalf of @gergo- (Gergö Barany) Differential Revision: https://reviews.llvm.org/D35911 llvm-svn: 313972
2017-09-22 17:50:52 +08:00
; RUN: llc -mtriple=arm-eabihf -mattr=+vfp3 %s -o - \
; RUN: | FileCheck %s -check-prefix=VFP3
; RUN: llc -mtriple=arm-eabihf -mattr=+neon %s -o - \
ARM: fixup more tests to specify the target more explicitly This changes the tests that were targeting ARM EABI to explicitly specify the environment rather than relying on the default. This breaks with the new Windows on ARM support when running the tests on Windows where the default environment is no longer EABI. Take the opportunity to avoid a pointless redirect (helps when trying to debug with providing a command line invocation which can be copy and pasted) and removing a few greps in favour of FileCheck. llvm-svn: 205541
2014-04-04 00:01:44 +08:00
; RUN: | FileCheck %s -check-prefix=NEON
; RUN: llc -mtriple=arm-eabi -mcpu=cortex-a8 %s -o - \
; RUN: | FileCheck %s -check-prefix=A8
; RUN: llc -mtriple=arm-eabi -mcpu=cortex-a8 -regalloc=basic %s -o - \
; RUN: | FileCheck %s -check-prefix=A8
; RUN: llc -mtriple=arm-eabi -mcpu=cortex-a8 --enable-unsafe-fp-math %s -o - \
; RUN: | FileCheck %s -check-prefix=A8U
; RUN: llc -mtriple=arm-darwin -mcpu=cortex-a8 %s -o - \
; RUN: | FileCheck %s -check-prefix=A8U
Initial support for single-precision FP using NEON. Added "neonfp" attribute to enable. Added patterns for some binary FP operations. llvm-svn: 78081
2009-08-05 01:53:06 +08:00
Add some missing isel predicates on def : pat patterns to avoid generating VFP vmla / vmls (they cause stalls). Disabling them in isel is properly not a right solution, I'll look into a proper solution next. llvm-svn: 118922
2010-11-13 04:32:20 +08:00
define float @t1(float %acc, float %a, float %b) nounwind {
Initial support for single-precision FP using NEON. Added "neonfp" attribute to enable. Added patterns for some binary FP operations. llvm-svn: 78081
2009-08-05 01:53:06 +08:00
entry:
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC: *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done llvm-svn: 186280
2013-07-14 14:24:09 +08:00
; VFP2-LABEL: t1:
Add some missing isel predicates on def : pat patterns to avoid generating VFP vmla / vmls (they cause stalls). Disabling them in isel is properly not a right solution, I'll look into a proper solution next. llvm-svn: 118922
2010-11-13 04:32:20 +08:00
; VFP2: vnmla.f32
[ARM] Add missing selection patterns for vnmla For the following function: double fn1(double d0, double d1, double d2) { double a = -d0 - d1 * d2; return a; } on ARM, LLVM generates code along the lines of vneg.f64 d0, d0 vmls.f64 d0, d1, d2 i.e., a negate and a multiply-subtract. The attached patch adds instruction selection patterns to allow it to generate the single instruction vnmla.f64 d0, d1, d2 (multiply-add with negation) instead, like GCC does. Committed on behalf of @gergo- (Gergö Barany) Differential Revision: https://reviews.llvm.org/D35911 llvm-svn: 313972
2017-09-22 17:50:52 +08:00
; VFP3-LABEL: t1:
; VFP3: vnmla.f32
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC: *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done llvm-svn: 186280
2013-07-14 14:24:09 +08:00
; NEON-LABEL: t1:
Add some missing isel predicates on def : pat patterns to avoid generating VFP vmla / vmls (they cause stalls). Disabling them in isel is properly not a right solution, I'll look into a proper solution next. llvm-svn: 118922
2010-11-13 04:32:20 +08:00
; NEON: vnmla.f32
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC: *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done llvm-svn: 186280
2013-07-14 14:24:09 +08:00
; A8U-LABEL: t1:
Avoid NEON SP-FP unless unsafe-math or Darwin NEON is not IEEE 754 compliant, so we should avoid lowering single-precision floating point operations with NEON unless unsafe-math is turned on. The equivalent VFP instructions are IEEE 754 compliant, but in some cores they're much slower, so some archs/OSs might still request it to be on by default, such as Swift and Darwin. llvm-svn: 177651
2013-03-22 02:47:47 +08:00
; A8U: vnmul.f32 s{{[0-9]}}, s{{[0-9]}}, s{{[0-9]}}
; A8U: vsub.f32 d{{[0-9]}}, d{{[0-9]}}, d{{[0-9]}}
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC: *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done llvm-svn: 186280
2013-07-14 14:24:09 +08:00
; A8-LABEL: t1:
Fix ARM tests to be register allocator independent. llvm-svn: 128680
2011-04-01 06:14:03 +08:00
; A8: vnmul.f32 s{{[0-9]}}, s{{[0-9]}}, s{{[0-9]}}
Avoid NEON SP-FP unless unsafe-math or Darwin NEON is not IEEE 754 compliant, so we should avoid lowering single-precision floating point operations with NEON unless unsafe-math is turned on. The equivalent VFP instructions are IEEE 754 compliant, but in some cores they're much slower, so some archs/OSs might still request it to be on by default, such as Swift and Darwin. llvm-svn: 177651
2013-03-22 02:47:47 +08:00
; A8: vsub.f32 s{{[0-9]}}, s{{[0-9]}}, s{{[0-9]}}
Initial support for single-precision FP using NEON. Added "neonfp" attribute to enable. Added patterns for some binary FP operations. llvm-svn: 78081
2009-08-05 01:53:06 +08:00
%0 = fmul float %a, %b
Fix test. llvm-svn: 78606
2009-08-11 06:31:04 +08:00
%1 = fsub float -0.0, %0
Initial support for single-precision FP using NEON. Added "neonfp" attribute to enable. Added patterns for some binary FP operations. llvm-svn: 78081
2009-08-05 01:53:06 +08:00
%2 = fsub float %1, %acc
ret float %2
}
Add some missing isel predicates on def : pat patterns to avoid generating VFP vmla / vmls (they cause stalls). Disabling them in isel is properly not a right solution, I'll look into a proper solution next. llvm-svn: 118922
2010-11-13 04:32:20 +08:00
define float @t2(float %acc, float %a, float %b) nounwind {
Improve tests. llvm-svn: 78083
2009-08-05 02:11:59 +08:00
entry:
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC: *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done llvm-svn: 186280
2013-07-14 14:24:09 +08:00
; VFP2-LABEL: t2:
Add some missing isel predicates on def : pat patterns to avoid generating VFP vmla / vmls (they cause stalls). Disabling them in isel is properly not a right solution, I'll look into a proper solution next. llvm-svn: 118922
2010-11-13 04:32:20 +08:00
; VFP2: vnmla.f32
[ARM] Add missing selection patterns for vnmla For the following function: double fn1(double d0, double d1, double d2) { double a = -d0 - d1 * d2; return a; } on ARM, LLVM generates code along the lines of vneg.f64 d0, d0 vmls.f64 d0, d1, d2 i.e., a negate and a multiply-subtract. The attached patch adds instruction selection patterns to allow it to generate the single instruction vnmla.f64 d0, d1, d2 (multiply-add with negation) instead, like GCC does. Committed on behalf of @gergo- (Gergö Barany) Differential Revision: https://reviews.llvm.org/D35911 llvm-svn: 313972
2017-09-22 17:50:52 +08:00
; VFP3-LABEL: t2:
; VFP3: vnmla.f32
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC: *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done llvm-svn: 186280
2013-07-14 14:24:09 +08:00
; NEON-LABEL: t2:
Add some missing isel predicates on def : pat patterns to avoid generating VFP vmla / vmls (they cause stalls). Disabling them in isel is properly not a right solution, I'll look into a proper solution next. llvm-svn: 118922
2010-11-13 04:32:20 +08:00
; NEON: vnmla.f32
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC: *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done llvm-svn: 186280
2013-07-14 14:24:09 +08:00
; A8U-LABEL: t2:
Avoid NEON SP-FP unless unsafe-math or Darwin NEON is not IEEE 754 compliant, so we should avoid lowering single-precision floating point operations with NEON unless unsafe-math is turned on. The equivalent VFP instructions are IEEE 754 compliant, but in some cores they're much slower, so some archs/OSs might still request it to be on by default, such as Swift and Darwin. llvm-svn: 177651
2013-03-22 02:47:47 +08:00
; A8U: vnmul.f32 s{{[01234]}}, s{{[01234]}}, s{{[01234]}}
; A8U: vsub.f32 d{{[0-9]}}, d{{[0-9]}}, d{{[0-9]}}
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC: *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done llvm-svn: 186280
2013-07-14 14:24:09 +08:00
; A8-LABEL: t2:
Make the test less likely to fail with minor changes. llvm-svn: 130778
2011-05-04 03:09:32 +08:00
; A8: vnmul.f32 s{{[01234]}}, s{{[01234]}}, s{{[01234]}}
Avoid NEON SP-FP unless unsafe-math or Darwin NEON is not IEEE 754 compliant, so we should avoid lowering single-precision floating point operations with NEON unless unsafe-math is turned on. The equivalent VFP instructions are IEEE 754 compliant, but in some cores they're much slower, so some archs/OSs might still request it to be on by default, such as Swift and Darwin. llvm-svn: 177651
2013-03-22 02:47:47 +08:00
; A8: vsub.f32 s{{[0-9]}}, s{{[0-9]}}, s{{[0-9]}}
Improve tests. llvm-svn: 78083
2009-08-05 02:11:59 +08:00
%0 = fmul float %a, %b
%1 = fmul float -1.0, %0
%2 = fsub float %1, %acc
ret float %2
}
Add some missing isel predicates on def : pat patterns to avoid generating VFP vmla / vmls (they cause stalls). Disabling them in isel is properly not a right solution, I'll look into a proper solution next. llvm-svn: 118922
2010-11-13 04:32:20 +08:00
define double @t3(double %acc, double %a, double %b) nounwind {
entry:
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC: *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done llvm-svn: 186280
2013-07-14 14:24:09 +08:00
; VFP2-LABEL: t3:
Add some missing isel predicates on def : pat patterns to avoid generating VFP vmla / vmls (they cause stalls). Disabling them in isel is properly not a right solution, I'll look into a proper solution next. llvm-svn: 118922
2010-11-13 04:32:20 +08:00
; VFP2: vnmla.f64
[ARM] Add missing selection patterns for vnmla For the following function: double fn1(double d0, double d1, double d2) { double a = -d0 - d1 * d2; return a; } on ARM, LLVM generates code along the lines of vneg.f64 d0, d0 vmls.f64 d0, d1, d2 i.e., a negate and a multiply-subtract. The attached patch adds instruction selection patterns to allow it to generate the single instruction vnmla.f64 d0, d1, d2 (multiply-add with negation) instead, like GCC does. Committed on behalf of @gergo- (Gergö Barany) Differential Revision: https://reviews.llvm.org/D35911 llvm-svn: 313972
2017-09-22 17:50:52 +08:00
; VFP3-LABEL: t3:
; VFP3: vnmla.f64
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC: *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done llvm-svn: 186280
2013-07-14 14:24:09 +08:00
; NEON-LABEL: t3:
Add some missing isel predicates on def : pat patterns to avoid generating VFP vmla / vmls (they cause stalls). Disabling them in isel is properly not a right solution, I'll look into a proper solution next. llvm-svn: 118922
2010-11-13 04:32:20 +08:00
; NEON: vnmla.f64
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC: *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done llvm-svn: 186280
2013-07-14 14:24:09 +08:00
; A8U-LABEL: t3:
Avoid NEON SP-FP unless unsafe-math or Darwin NEON is not IEEE 754 compliant, so we should avoid lowering single-precision floating point operations with NEON unless unsafe-math is turned on. The equivalent VFP instructions are IEEE 754 compliant, but in some cores they're much slower, so some archs/OSs might still request it to be on by default, such as Swift and Darwin. llvm-svn: 177651
2013-03-22 02:47:47 +08:00
; A8U: vnmul.f64 d
; A8U: vsub.f64 d
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC: *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done llvm-svn: 186280
2013-07-14 14:24:09 +08:00
; A8-LABEL: t3:
Remove some register allocation order dependencies. llvm-svn: 172874
2013-01-19 08:03:32 +08:00
; A8: vnmul.f64 d
; A8: vsub.f64 d
Add some missing isel predicates on def : pat patterns to avoid generating VFP vmla / vmls (they cause stalls). Disabling them in isel is properly not a right solution, I'll look into a proper solution next. llvm-svn: 118922
2010-11-13 04:32:20 +08:00
%0 = fmul double %a, %b
%1 = fsub double -0.0, %0
%2 = fsub double %1, %acc
ret double %2
}
define double @t4(double %acc, double %a, double %b) nounwind {
entry:
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC: *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done llvm-svn: 186280
2013-07-14 14:24:09 +08:00
; VFP2-LABEL: t4:
Add some missing isel predicates on def : pat patterns to avoid generating VFP vmla / vmls (they cause stalls). Disabling them in isel is properly not a right solution, I'll look into a proper solution next. llvm-svn: 118922
2010-11-13 04:32:20 +08:00
; VFP2: vnmla.f64
[ARM] Add missing selection patterns for vnmla For the following function: double fn1(double d0, double d1, double d2) { double a = -d0 - d1 * d2; return a; } on ARM, LLVM generates code along the lines of vneg.f64 d0, d0 vmls.f64 d0, d1, d2 i.e., a negate and a multiply-subtract. The attached patch adds instruction selection patterns to allow it to generate the single instruction vnmla.f64 d0, d1, d2 (multiply-add with negation) instead, like GCC does. Committed on behalf of @gergo- (Gergö Barany) Differential Revision: https://reviews.llvm.org/D35911 llvm-svn: 313972
2017-09-22 17:50:52 +08:00
; VFP3-LABEL: t4:
; VFP3: vnmla.f64
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC: *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done llvm-svn: 186280
2013-07-14 14:24:09 +08:00
; NEON-LABEL: t4:
Add some missing isel predicates on def : pat patterns to avoid generating VFP vmla / vmls (they cause stalls). Disabling them in isel is properly not a right solution, I'll look into a proper solution next. llvm-svn: 118922
2010-11-13 04:32:20 +08:00
; NEON: vnmla.f64
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC: *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done llvm-svn: 186280
2013-07-14 14:24:09 +08:00
; A8U-LABEL: t4:
Avoid NEON SP-FP unless unsafe-math or Darwin NEON is not IEEE 754 compliant, so we should avoid lowering single-precision floating point operations with NEON unless unsafe-math is turned on. The equivalent VFP instructions are IEEE 754 compliant, but in some cores they're much slower, so some archs/OSs might still request it to be on by default, such as Swift and Darwin. llvm-svn: 177651
2013-03-22 02:47:47 +08:00
; A8U: vnmul.f64 d
; A8U: vsub.f64 d
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC: *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done llvm-svn: 186280
2013-07-14 14:24:09 +08:00
; A8-LABEL: t4:
Remove some register allocation order dependencies. llvm-svn: 172874
2013-01-19 08:03:32 +08:00
; A8: vnmul.f64 d
; A8: vsub.f64 d
Add some missing isel predicates on def : pat patterns to avoid generating VFP vmla / vmls (they cause stalls). Disabling them in isel is properly not a right solution, I'll look into a proper solution next. llvm-svn: 118922
2010-11-13 04:32:20 +08:00
%0 = fmul double %a, %b
%1 = fmul double -1.0, %0
%2 = fsub double %1, %acc
ret double %2
}
[ARM] Add missing selection patterns for vnmla For the following function: double fn1(double d0, double d1, double d2) { double a = -d0 - d1 * d2; return a; } on ARM, LLVM generates code along the lines of vneg.f64 d0, d0 vmls.f64 d0, d1, d2 i.e., a negate and a multiply-subtract. The attached patch adds instruction selection patterns to allow it to generate the single instruction vnmla.f64 d0, d1, d2 (multiply-add with negation) instead, like GCC does. Committed on behalf of @gergo- (Gergö Barany) Differential Revision: https://reviews.llvm.org/D35911 llvm-svn: 313972
2017-09-22 17:50:52 +08:00
define double @t5(double %acc, double %a, double %b) nounwind {
entry:
; VFP2-LABEL: t5:
; VFP2: vnmla.f64
; VFP3-LABEL: t5:
; VFP3: vnmla.f64
; NEON-LABEL: t5:
; NEON: vnmla.f64
; A8U-LABEL: t5:
; A8U: vmul.f64 d
; A8U: vsub.f64 d
; A8-LABEL: t5:
; A8: vmul.f64 d
; A8: vsub.f64 d
%0 = fsub double -0.0, %acc
%1 = fmul double %a, %b
%2 = fsub double %0, %1
ret double %2
}
define float @t6(float %acc, float %a, float %b) nounwind {
entry:
; VFP2-LABEL: t6:
; VFP2: vnmla.f32
; VFP3-LABEL: t6:
; VFP3: vnmla.f32
; NEON-LABEL: t6:
; NEON: vnmla.f32
; A8U-LABEL: t6:
; A8U: vmul.f32 d
; A8U: vsub.f32 d
; A8-LABEL: t6:
; A8: vmul.f32 s
; A8: vsub.f32 s
%0 = fsub float -0.0, %acc
%1 = fmul float %a, %b
%2 = fsub float %0, %1
ret float %2
}