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/X86/copysign-constant-magnitude.ll

145 lines
4.5 KiB
LLVM
Raw Normal View History

[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
; RUN: llc -mtriple=x86_64-apple-macosx10.10.0 < %s | FileCheck %s
[X86] Teach FCOPYSIGN lowering to recognize constant magnitudes. For code like: float foo(float x) { return copysign(1.0, x); } We used to generate: andps <-0.000000e+00,0,0,0>, %xmm0 movss <1.000000e+00>, %xmm1 andps <nan>, %xmm1 orps %xmm0, %xmm1 Basically doing an abs(1.0f) in the two middle instructions. We now generate: andps <-0.000000e+00,0,0,0>, %xmm0 orps <1.000000e+00,0,0,0>, %xmm0 Builds on cleanups r223415, r223542. rdar://19049548 Differential Revision: http://reviews.llvm.org/D6555 llvm-svn: 225357
2015-01-08 01:33:03 +08:00
target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
; The assertions are *enhanced* from update_test_checks.ll to include
[x86] fix FCOPYSIGN lowering to create constants instead of ConstantPool loads This is similar to: https://reviews.llvm.org/rL279958 By not prematurely lowering to loads, we should be able to more easily eliminate the 'or' with zero instructions seen in copysign-constant-magnitude.ll. We should also be able to extend this code to handle vectors. llvm-svn: 282312
2016-09-24 07:17:29 +08:00
; the constant load values because those are important.
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
; CHECK: [[SIGNMASK1:L.+]]:
AsmPrinter: Use emitGlobalConstantFP to emit elements of constant data It's strange to duplicate the logic for emitting FP values into emitGlobalConstantDataSequential, and it's even stranger that we end up printing the verbose assembly comments differently between the two paths. Just call into emitGlobalConstantFP rather than crudely duplicating its logic. llvm-svn: 254988
2015-12-08 10:37:48 +08:00
; CHECK-NEXT: .quad -9223372036854775808 ## double -0
[x86] fix FCOPYSIGN lowering to create constants instead of ConstantPool loads This is similar to: https://reviews.llvm.org/rL279958 By not prematurely lowering to loads, we should be able to more easily eliminate the 'or' with zero instructions seen in copysign-constant-magnitude.ll. We should also be able to extend this code to handle vectors. llvm-svn: 282312
2016-09-24 07:17:29 +08:00
; CHECK-NEXT: .quad -9223372036854775808 ## double -0
[X86] Teach FCOPYSIGN lowering to recognize constant magnitudes. For code like: float foo(float x) { return copysign(1.0, x); } We used to generate: andps <-0.000000e+00,0,0,0>, %xmm0 movss <1.000000e+00>, %xmm1 andps <nan>, %xmm1 orps %xmm0, %xmm1 Basically doing an abs(1.0f) in the two middle instructions. We now generate: andps <-0.000000e+00,0,0,0>, %xmm0 orps <1.000000e+00,0,0,0>, %xmm0 Builds on cleanups r223415, r223542. rdar://19049548 Differential Revision: http://reviews.llvm.org/D6555 llvm-svn: 225357
2015-01-08 01:33:03 +08:00
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
define double @mag_pos0_double(double %x) nounwind {
; CHECK-LABEL: mag_pos0_double:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: ## %bb.0:
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
; CHECK-NEXT: andps [[SIGNMASK1]](%rip), %xmm0
; CHECK-NEXT: retq
;
%y = call double @copysign(double 0.0, double %x)
ret double %y
[X86] Teach FCOPYSIGN lowering to recognize constant magnitudes. For code like: float foo(float x) { return copysign(1.0, x); } We used to generate: andps <-0.000000e+00,0,0,0>, %xmm0 movss <1.000000e+00>, %xmm1 andps <nan>, %xmm1 orps %xmm0, %xmm1 Basically doing an abs(1.0f) in the two middle instructions. We now generate: andps <-0.000000e+00,0,0,0>, %xmm0 orps <1.000000e+00,0,0,0>, %xmm0 Builds on cleanups r223415, r223542. rdar://19049548 Differential Revision: http://reviews.llvm.org/D6555 llvm-svn: 225357
2015-01-08 01:33:03 +08:00
}
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
; CHECK: [[SIGNMASK2:L.+]]:
; CHECK-NEXT: .quad -9223372036854775808 ## double -0
[X86] Teach FCOPYSIGN lowering to recognize constant magnitudes. For code like: float foo(float x) { return copysign(1.0, x); } We used to generate: andps <-0.000000e+00,0,0,0>, %xmm0 movss <1.000000e+00>, %xmm1 andps <nan>, %xmm1 orps %xmm0, %xmm1 Basically doing an abs(1.0f) in the two middle instructions. We now generate: andps <-0.000000e+00,0,0,0>, %xmm0 orps <1.000000e+00,0,0,0>, %xmm0 Builds on cleanups r223415, r223542. rdar://19049548 Differential Revision: http://reviews.llvm.org/D6555 llvm-svn: 225357
2015-01-08 01:33:03 +08:00
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
define double @mag_neg0_double(double %x) nounwind {
; CHECK-LABEL: mag_neg0_double:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: ## %bb.0:
[x86] fix FCOPYSIGN lowering to create constants instead of ConstantPool loads This is similar to: https://reviews.llvm.org/rL279958 By not prematurely lowering to loads, we should be able to more easily eliminate the 'or' with zero instructions seen in copysign-constant-magnitude.ll. We should also be able to extend this code to handle vectors. llvm-svn: 282312
2016-09-24 07:17:29 +08:00
; CHECK-NEXT: movsd [[SIGNMASK2]](%rip), %xmm1
; CHECK-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0,0]
[x86] add folds for FP logic with vector zeros The 'or' case shows up in copysign. The copysign code also had redundant checking for a scalar zero operand with 'and', so I removed that. I'm not sure how to test vector 'and', 'andn', and 'xor' yet, but it seems better to just include all of the logic ops since we're fixing 'or' anyway. llvm-svn: 282546
2016-09-28 06:28:13 +08:00
; CHECK-NEXT: andps %xmm1, %xmm0
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
; CHECK-NEXT: retq
;
%y = call double @copysign(double -0.0, double %x)
ret double %y
}
[X86] Teach FCOPYSIGN lowering to recognize constant magnitudes. For code like: float foo(float x) { return copysign(1.0, x); } We used to generate: andps <-0.000000e+00,0,0,0>, %xmm0 movss <1.000000e+00>, %xmm1 andps <nan>, %xmm1 orps %xmm0, %xmm1 Basically doing an abs(1.0f) in the two middle instructions. We now generate: andps <-0.000000e+00,0,0,0>, %xmm0 orps <1.000000e+00,0,0,0>, %xmm0 Builds on cleanups r223415, r223542. rdar://19049548 Differential Revision: http://reviews.llvm.org/D6555 llvm-svn: 225357
2015-01-08 01:33:03 +08:00
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
; CHECK: [[SIGNMASK3:L.+]]:
; CHECK-NEXT: .quad -9223372036854775808 ## double -0
[x86] fix FCOPYSIGN lowering to create constants instead of ConstantPool loads This is similar to: https://reviews.llvm.org/rL279958 By not prematurely lowering to loads, we should be able to more easily eliminate the 'or' with zero instructions seen in copysign-constant-magnitude.ll. We should also be able to extend this code to handle vectors. llvm-svn: 282312
2016-09-24 07:17:29 +08:00
; CHECK-NEXT: .quad -9223372036854775808 ## double -0
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
; CHECK: [[ONE3:L.+]]:
; CHECK-NEXT: .quad 4607182418800017408 ## double 1
[X86] Teach FCOPYSIGN lowering to recognize constant magnitudes. For code like: float foo(float x) { return copysign(1.0, x); } We used to generate: andps <-0.000000e+00,0,0,0>, %xmm0 movss <1.000000e+00>, %xmm1 andps <nan>, %xmm1 orps %xmm0, %xmm1 Basically doing an abs(1.0f) in the two middle instructions. We now generate: andps <-0.000000e+00,0,0,0>, %xmm0 orps <1.000000e+00,0,0,0>, %xmm0 Builds on cleanups r223415, r223542. rdar://19049548 Differential Revision: http://reviews.llvm.org/D6555 llvm-svn: 225357
2015-01-08 01:33:03 +08:00
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
define double @mag_pos1_double(double %x) nounwind {
; CHECK-LABEL: mag_pos1_double:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: ## %bb.0:
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
; CHECK-NEXT: andps [[SIGNMASK3]](%rip), %xmm0
[x86] fix FCOPYSIGN lowering to create constants instead of ConstantPool loads This is similar to: https://reviews.llvm.org/rL279958 By not prematurely lowering to loads, we should be able to more easily eliminate the 'or' with zero instructions seen in copysign-constant-magnitude.ll. We should also be able to extend this code to handle vectors. llvm-svn: 282312
2016-09-24 07:17:29 +08:00
; CHECK-NEXT: movsd [[ONE3]](%rip), %xmm1
; CHECK-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0,0]
; CHECK-NEXT: orps %xmm1, %xmm0
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
; CHECK-NEXT: retq
;
%y = call double @copysign(double 1.0, double %x)
ret double %y
}
[X86] Teach FCOPYSIGN lowering to recognize constant magnitudes. For code like: float foo(float x) { return copysign(1.0, x); } We used to generate: andps <-0.000000e+00,0,0,0>, %xmm0 movss <1.000000e+00>, %xmm1 andps <nan>, %xmm1 orps %xmm0, %xmm1 Basically doing an abs(1.0f) in the two middle instructions. We now generate: andps <-0.000000e+00,0,0,0>, %xmm0 orps <1.000000e+00,0,0,0>, %xmm0 Builds on cleanups r223415, r223542. rdar://19049548 Differential Revision: http://reviews.llvm.org/D6555 llvm-svn: 225357
2015-01-08 01:33:03 +08:00
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
; CHECK: [[SIGNMASK4:L.+]]:
; CHECK-NEXT: .quad -9223372036854775808 ## double -0
[x86] fix FCOPYSIGN lowering to create constants instead of ConstantPool loads This is similar to: https://reviews.llvm.org/rL279958 By not prematurely lowering to loads, we should be able to more easily eliminate the 'or' with zero instructions seen in copysign-constant-magnitude.ll. We should also be able to extend this code to handle vectors. llvm-svn: 282312
2016-09-24 07:17:29 +08:00
; CHECK-NEXT: .quad -9223372036854775808 ## double -0
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
; CHECK: [[ONE4:L.+]]:
; CHECK-NEXT: .quad 4607182418800017408 ## double 1
[x86] fix FCOPYSIGN lowering to create constants instead of ConstantPool loads This is similar to: https://reviews.llvm.org/rL279958 By not prematurely lowering to loads, we should be able to more easily eliminate the 'or' with zero instructions seen in copysign-constant-magnitude.ll. We should also be able to extend this code to handle vectors. llvm-svn: 282312
2016-09-24 07:17:29 +08:00
; CHECK-NEXT: .quad 4607182418800017408 ## double 1
[X86] Teach FCOPYSIGN lowering to recognize constant magnitudes. For code like: float foo(float x) { return copysign(1.0, x); } We used to generate: andps <-0.000000e+00,0,0,0>, %xmm0 movss <1.000000e+00>, %xmm1 andps <nan>, %xmm1 orps %xmm0, %xmm1 Basically doing an abs(1.0f) in the two middle instructions. We now generate: andps <-0.000000e+00,0,0,0>, %xmm0 orps <1.000000e+00,0,0,0>, %xmm0 Builds on cleanups r223415, r223542. rdar://19049548 Differential Revision: http://reviews.llvm.org/D6555 llvm-svn: 225357
2015-01-08 01:33:03 +08:00
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
define double @mag_neg1_double(double %x) nounwind {
; CHECK-LABEL: mag_neg1_double:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: ## %bb.0:
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
; CHECK-NEXT: andps [[SIGNMASK4]](%rip), %xmm0
; CHECK-NEXT: orps [[ONE4]](%rip), %xmm0
; CHECK-NEXT: retq
;
%y = call double @copysign(double -1.0, double %x)
ret double %y
}
[X86] Teach FCOPYSIGN lowering to recognize constant magnitudes. For code like: float foo(float x) { return copysign(1.0, x); } We used to generate: andps <-0.000000e+00,0,0,0>, %xmm0 movss <1.000000e+00>, %xmm1 andps <nan>, %xmm1 orps %xmm0, %xmm1 Basically doing an abs(1.0f) in the two middle instructions. We now generate: andps <-0.000000e+00,0,0,0>, %xmm0 orps <1.000000e+00,0,0,0>, %xmm0 Builds on cleanups r223415, r223542. rdar://19049548 Differential Revision: http://reviews.llvm.org/D6555 llvm-svn: 225357
2015-01-08 01:33:03 +08:00
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
; CHECK: [[SIGNMASK5:L.+]]:
; CHECK-NEXT: .long 2147483648 ## float -0
[x86] fix FCOPYSIGN lowering to create constants instead of ConstantPool loads This is similar to: https://reviews.llvm.org/rL279958 By not prematurely lowering to loads, we should be able to more easily eliminate the 'or' with zero instructions seen in copysign-constant-magnitude.ll. We should also be able to extend this code to handle vectors. llvm-svn: 282312
2016-09-24 07:17:29 +08:00
; CHECK-NEXT: .long 2147483648 ## float -0
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
define float @mag_pos0_float(float %x) nounwind {
; CHECK-LABEL: mag_pos0_float:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: ## %bb.0:
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
; CHECK-NEXT: andps [[SIGNMASK5]](%rip), %xmm0
; CHECK-NEXT: retq
;
%y = call float @copysignf(float 0.0, float %x)
ret float %y
}
[X86] Teach FCOPYSIGN lowering to recognize constant magnitudes. For code like: float foo(float x) { return copysign(1.0, x); } We used to generate: andps <-0.000000e+00,0,0,0>, %xmm0 movss <1.000000e+00>, %xmm1 andps <nan>, %xmm1 orps %xmm0, %xmm1 Basically doing an abs(1.0f) in the two middle instructions. We now generate: andps <-0.000000e+00,0,0,0>, %xmm0 orps <1.000000e+00,0,0,0>, %xmm0 Builds on cleanups r223415, r223542. rdar://19049548 Differential Revision: http://reviews.llvm.org/D6555 llvm-svn: 225357
2015-01-08 01:33:03 +08:00
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
; CHECK: [[SIGNMASK6:L.+]]:
; CHECK-NEXT: .long 2147483648 ## float -0
define float @mag_neg0_float(float %x) nounwind {
; CHECK-LABEL: mag_neg0_float:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: ## %bb.0:
[x86] fix FCOPYSIGN lowering to create constants instead of ConstantPool loads This is similar to: https://reviews.llvm.org/rL279958 By not prematurely lowering to loads, we should be able to more easily eliminate the 'or' with zero instructions seen in copysign-constant-magnitude.ll. We should also be able to extend this code to handle vectors. llvm-svn: 282312
2016-09-24 07:17:29 +08:00
; CHECK-NEXT: movss [[SIGNMASK6]](%rip), %xmm1
; CHECK-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,0,0,0]
[x86] add folds for FP logic with vector zeros The 'or' case shows up in copysign. The copysign code also had redundant checking for a scalar zero operand with 'and', so I removed that. I'm not sure how to test vector 'and', 'andn', and 'xor' yet, but it seems better to just include all of the logic ops since we're fixing 'or' anyway. llvm-svn: 282546
2016-09-28 06:28:13 +08:00
; CHECK-NEXT: andps %xmm1, %xmm0
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
; CHECK-NEXT: retq
;
%y = call float @copysignf(float -0.0, float %x)
ret float %y
}
[X86] Teach FCOPYSIGN lowering to recognize constant magnitudes. For code like: float foo(float x) { return copysign(1.0, x); } We used to generate: andps <-0.000000e+00,0,0,0>, %xmm0 movss <1.000000e+00>, %xmm1 andps <nan>, %xmm1 orps %xmm0, %xmm1 Basically doing an abs(1.0f) in the two middle instructions. We now generate: andps <-0.000000e+00,0,0,0>, %xmm0 orps <1.000000e+00,0,0,0>, %xmm0 Builds on cleanups r223415, r223542. rdar://19049548 Differential Revision: http://reviews.llvm.org/D6555 llvm-svn: 225357
2015-01-08 01:33:03 +08:00
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
; CHECK: [[SIGNMASK7:L.+]]:
; CHECK-NEXT: .long 2147483648 ## float -0
[x86] fix FCOPYSIGN lowering to create constants instead of ConstantPool loads This is similar to: https://reviews.llvm.org/rL279958 By not prematurely lowering to loads, we should be able to more easily eliminate the 'or' with zero instructions seen in copysign-constant-magnitude.ll. We should also be able to extend this code to handle vectors. llvm-svn: 282312
2016-09-24 07:17:29 +08:00
; CHECK-NEXT: .long 2147483648 ## float -0
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
; CHECK: [[ONE7:L.+]]:
; CHECK-NEXT: .long 1065353216 ## float 1
define float @mag_pos1_float(float %x) nounwind {
; CHECK-LABEL: mag_pos1_float:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: ## %bb.0:
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
; CHECK-NEXT: andps [[SIGNMASK7]](%rip), %xmm0
[x86] fix FCOPYSIGN lowering to create constants instead of ConstantPool loads This is similar to: https://reviews.llvm.org/rL279958 By not prematurely lowering to loads, we should be able to more easily eliminate the 'or' with zero instructions seen in copysign-constant-magnitude.ll. We should also be able to extend this code to handle vectors. llvm-svn: 282312
2016-09-24 07:17:29 +08:00
; CHECK-NEXT: movss [[ONE7]](%rip), %xmm1
; CHECK-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,0,0,0]
; CHECK-NEXT: orps %xmm1, %xmm0
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
; CHECK-NEXT: retq
;
%y = call float @copysignf(float 1.0, float %x)
ret float %y
}
[X86] Teach FCOPYSIGN lowering to recognize constant magnitudes. For code like: float foo(float x) { return copysign(1.0, x); } We used to generate: andps <-0.000000e+00,0,0,0>, %xmm0 movss <1.000000e+00>, %xmm1 andps <nan>, %xmm1 orps %xmm0, %xmm1 Basically doing an abs(1.0f) in the two middle instructions. We now generate: andps <-0.000000e+00,0,0,0>, %xmm0 orps <1.000000e+00,0,0,0>, %xmm0 Builds on cleanups r223415, r223542. rdar://19049548 Differential Revision: http://reviews.llvm.org/D6555 llvm-svn: 225357
2015-01-08 01:33:03 +08:00
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
; CHECK: [[SIGNMASK8:L.+]]:
; CHECK-NEXT: .long 2147483648 ## float -0
[x86] fix FCOPYSIGN lowering to create constants instead of ConstantPool loads This is similar to: https://reviews.llvm.org/rL279958 By not prematurely lowering to loads, we should be able to more easily eliminate the 'or' with zero instructions seen in copysign-constant-magnitude.ll. We should also be able to extend this code to handle vectors. llvm-svn: 282312
2016-09-24 07:17:29 +08:00
; CHECK-NEXT: .long 2147483648 ## float -0
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
; CHECK: [[ONE8:L.+]]:
; CHECK-NEXT: .long 1065353216 ## float 1
[x86] fix FCOPYSIGN lowering to create constants instead of ConstantPool loads This is similar to: https://reviews.llvm.org/rL279958 By not prematurely lowering to loads, we should be able to more easily eliminate the 'or' with zero instructions seen in copysign-constant-magnitude.ll. We should also be able to extend this code to handle vectors. llvm-svn: 282312
2016-09-24 07:17:29 +08:00
; CHECK-NEXT: .long 1065353216 ## float 1
; CHECK-NEXT: .long 1065353216 ## float 1
; CHECK-NEXT: .long 1065353216 ## float 1
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
define float @mag_neg1_float(float %x) nounwind {
; CHECK-LABEL: mag_neg1_float:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: ## %bb.0:
[x86] split up tests, regenerate checks Note that we fail to eliminate 'or' with 0! llvm-svn: 282005
2016-09-21 03:31:30 +08:00
; CHECK-NEXT: andps [[SIGNMASK8]](%rip), %xmm0
; CHECK-NEXT: orps [[ONE8]](%rip), %xmm0
; CHECK-NEXT: retq
;
%y = call float @copysignf(float -1.0, float %x)
ret float %y
[X86] Teach FCOPYSIGN lowering to recognize constant magnitudes. For code like: float foo(float x) { return copysign(1.0, x); } We used to generate: andps <-0.000000e+00,0,0,0>, %xmm0 movss <1.000000e+00>, %xmm1 andps <nan>, %xmm1 orps %xmm0, %xmm1 Basically doing an abs(1.0f) in the two middle instructions. We now generate: andps <-0.000000e+00,0,0,0>, %xmm0 orps <1.000000e+00,0,0,0>, %xmm0 Builds on cleanups r223415, r223542. rdar://19049548 Differential Revision: http://reviews.llvm.org/D6555 llvm-svn: 225357
2015-01-08 01:33:03 +08:00
}
declare double @copysign(double, double) nounwind readnone
declare float @copysignf(float, float) nounwind readnone