Fix buggy fcopysign lowering.
This
define float @foo(float %x, float %y) nounwind readnone {
entry:
%0 = tail call float @copysignf(float %x, float %y) nounwind readnone
ret float %0
}
Was compiled to:
vmov s0, r1
bic r0, r0, #-2147483648
vmov s1, r0
vcmpe.f32 s0, #0
vmrs apsr_nzcv, fpscr
it lt
vneglt.f32 s1, s1
vmov r0, s1
bx lr
This fails to copy the sign of -0.0f because it's lost during the float to int
conversion. Also, it's sub-optimal when the inputs are in GPR registers.
Now it uses integer and + or operations when it's profitable. And it's correct!
lsrs r1, r1, #31
bfi r0, r1, #31, #1
bx lr
rdar://8984306
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@125357 91177308-0d34-0410-b5e6-96231b3b80d8
2011-02-11 02:28:55 +00:00
|
|
|
; RUN: llc < %s -mtriple=armv7-apple-darwin -mcpu=cortex-a8 | FileCheck %s -check-prefix=SOFT
|
|
|
|
; RUN: llc < %s -mtriple=armv7-gnueabi -float-abi=hard -mcpu=cortex-a8 | FileCheck %s -check-prefix=HARD
|
2007-01-04 14:24:32 +00:00
|
|
|
|
Fix buggy fcopysign lowering.
This
define float @foo(float %x, float %y) nounwind readnone {
entry:
%0 = tail call float @copysignf(float %x, float %y) nounwind readnone
ret float %0
}
Was compiled to:
vmov s0, r1
bic r0, r0, #-2147483648
vmov s1, r0
vcmpe.f32 s0, #0
vmrs apsr_nzcv, fpscr
it lt
vneglt.f32 s1, s1
vmov r0, s1
bx lr
This fails to copy the sign of -0.0f because it's lost during the float to int
conversion. Also, it's sub-optimal when the inputs are in GPR registers.
Now it uses integer and + or operations when it's profitable. And it's correct!
lsrs r1, r1, #31
bfi r0, r1, #31, #1
bx lr
rdar://8984306
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@125357 91177308-0d34-0410-b5e6-96231b3b80d8
2011-02-11 02:28:55 +00:00
|
|
|
; rdar://8984306
|
|
|
|
define float @test1(float %x, float %y) nounwind {
|
|
|
|
entry:
|
|
|
|
; SOFT: test1:
|
|
|
|
; SOFT: lsr r1, r1, #31
|
|
|
|
; SOFT: bfi r0, r1, #31, #1
|
|
|
|
|
|
|
|
; HARD: test1:
|
2011-02-23 02:24:55 +00:00
|
|
|
; HARD: vmov.i32 [[REG1:(d[0-9]+)]], #0x80000000
|
|
|
|
; HARD: vbsl [[REG1]], d2, d0
|
Fix buggy fcopysign lowering.
This
define float @foo(float %x, float %y) nounwind readnone {
entry:
%0 = tail call float @copysignf(float %x, float %y) nounwind readnone
ret float %0
}
Was compiled to:
vmov s0, r1
bic r0, r0, #-2147483648
vmov s1, r0
vcmpe.f32 s0, #0
vmrs apsr_nzcv, fpscr
it lt
vneglt.f32 s1, s1
vmov r0, s1
bx lr
This fails to copy the sign of -0.0f because it's lost during the float to int
conversion. Also, it's sub-optimal when the inputs are in GPR registers.
Now it uses integer and + or operations when it's profitable. And it's correct!
lsrs r1, r1, #31
bfi r0, r1, #31, #1
bx lr
rdar://8984306
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@125357 91177308-0d34-0410-b5e6-96231b3b80d8
2011-02-11 02:28:55 +00:00
|
|
|
%0 = tail call float @copysignf(float %x, float %y) nounwind
|
|
|
|
ret float %0
|
|
|
|
}
|
|
|
|
|
|
|
|
define double @test2(double %x, double %y) nounwind {
|
|
|
|
entry:
|
|
|
|
; SOFT: test2:
|
|
|
|
; SOFT: lsr r2, r3, #31
|
|
|
|
; SOFT: bfi r1, r2, #31, #1
|
|
|
|
|
|
|
|
; HARD: test2:
|
2011-02-23 02:24:55 +00:00
|
|
|
; HARD: vmov.i32 [[REG2:(d[0-9]+)]], #0x80000000
|
|
|
|
; HARD: vshl.i64 [[REG2]], [[REG2]], #32
|
|
|
|
; HARD: vbsl [[REG2]], d1, d0
|
Fix buggy fcopysign lowering.
This
define float @foo(float %x, float %y) nounwind readnone {
entry:
%0 = tail call float @copysignf(float %x, float %y) nounwind readnone
ret float %0
}
Was compiled to:
vmov s0, r1
bic r0, r0, #-2147483648
vmov s1, r0
vcmpe.f32 s0, #0
vmrs apsr_nzcv, fpscr
it lt
vneglt.f32 s1, s1
vmov r0, s1
bx lr
This fails to copy the sign of -0.0f because it's lost during the float to int
conversion. Also, it's sub-optimal when the inputs are in GPR registers.
Now it uses integer and + or operations when it's profitable. And it's correct!
lsrs r1, r1, #31
bfi r0, r1, #31, #1
bx lr
rdar://8984306
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@125357 91177308-0d34-0410-b5e6-96231b3b80d8
2011-02-11 02:28:55 +00:00
|
|
|
%0 = tail call double @copysign(double %x, double %y) nounwind
|
|
|
|
ret double %0
|
2007-01-04 14:24:32 +00:00
|
|
|
}
|
|
|
|
|
Fix buggy fcopysign lowering.
This
define float @foo(float %x, float %y) nounwind readnone {
entry:
%0 = tail call float @copysignf(float %x, float %y) nounwind readnone
ret float %0
}
Was compiled to:
vmov s0, r1
bic r0, r0, #-2147483648
vmov s1, r0
vcmpe.f32 s0, #0
vmrs apsr_nzcv, fpscr
it lt
vneglt.f32 s1, s1
vmov r0, s1
bx lr
This fails to copy the sign of -0.0f because it's lost during the float to int
conversion. Also, it's sub-optimal when the inputs are in GPR registers.
Now it uses integer and + or operations when it's profitable. And it's correct!
lsrs r1, r1, #31
bfi r0, r1, #31, #1
bx lr
rdar://8984306
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@125357 91177308-0d34-0410-b5e6-96231b3b80d8
2011-02-11 02:28:55 +00:00
|
|
|
define double @test3(double %x, double %y, double %z) nounwind {
|
|
|
|
entry:
|
|
|
|
; SOFT: test3:
|
2011-02-23 02:24:55 +00:00
|
|
|
; SOFT: vmov.i32 [[REG3:(d[0-9]+)]], #0x80000000
|
|
|
|
; SOFT: vshl.i64 [[REG3]], [[REG3]], #32
|
|
|
|
; SOFT: vbsl [[REG3]],
|
Fix buggy fcopysign lowering.
This
define float @foo(float %x, float %y) nounwind readnone {
entry:
%0 = tail call float @copysignf(float %x, float %y) nounwind readnone
ret float %0
}
Was compiled to:
vmov s0, r1
bic r0, r0, #-2147483648
vmov s1, r0
vcmpe.f32 s0, #0
vmrs apsr_nzcv, fpscr
it lt
vneglt.f32 s1, s1
vmov r0, s1
bx lr
This fails to copy the sign of -0.0f because it's lost during the float to int
conversion. Also, it's sub-optimal when the inputs are in GPR registers.
Now it uses integer and + or operations when it's profitable. And it's correct!
lsrs r1, r1, #31
bfi r0, r1, #31, #1
bx lr
rdar://8984306
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@125357 91177308-0d34-0410-b5e6-96231b3b80d8
2011-02-11 02:28:55 +00:00
|
|
|
%0 = fmul double %x, %y
|
|
|
|
%1 = tail call double @copysign(double %0, double %z) nounwind
|
|
|
|
ret double %1
|
2007-01-19 09:20:23 +00:00
|
|
|
}
|
2007-01-04 14:24:32 +00:00
|
|
|
|
2011-02-28 18:45:27 +00:00
|
|
|
; rdar://9059537
|
|
|
|
define i32 @test4() ssp {
|
|
|
|
entry:
|
|
|
|
; SOFT: test4:
|
2011-04-15 01:31:00 +00:00
|
|
|
; SOFT: vmov.f64 [[REG4:(d[0-9]+)]], #1.000000e+00
|
2011-04-19 06:14:45 +00:00
|
|
|
; This S-reg must be the first sub-reg of the last D-reg on vbsl.
|
|
|
|
; SOFT: vcvt.f32.f64 {{s1?[02468]}}, [[REG4]]
|
2011-04-15 01:31:00 +00:00
|
|
|
; SOFT: vshr.u64 [[REG4]], [[REG4]], #32
|
|
|
|
; SOFT: vmov.i32 [[REG5:(d[0-9]+)]], #0x80000000
|
2011-04-19 00:14:43 +00:00
|
|
|
; SOFT: vbsl [[REG5]], [[REG4]], {{d[0-9]+}}
|
2011-02-28 18:45:27 +00:00
|
|
|
%call80 = tail call double @copysign(double 1.000000e+00, double undef)
|
|
|
|
%conv81 = fptrunc double %call80 to float
|
|
|
|
%tmp88 = bitcast float %conv81 to i32
|
|
|
|
ret i32 %tmp88
|
|
|
|
}
|
|
|
|
|
2011-04-15 01:31:00 +00:00
|
|
|
; rdar://9287902
|
|
|
|
define float @test5() nounwind {
|
|
|
|
entry:
|
|
|
|
; SOFT: test5:
|
|
|
|
; SOFT: vmov.i32 [[REG6:(d[0-9]+)]], #0x80000000
|
|
|
|
; SOFT: vmov [[REG7:(d[0-9]+)]], r0, r1
|
|
|
|
; SOFT: vshr.u64 [[REG7]], [[REG7]], #32
|
|
|
|
; SOFT: vbsl [[REG6]], [[REG7]],
|
|
|
|
%0 = tail call double (...)* @bar() nounwind
|
|
|
|
%1 = fptrunc double %0 to float
|
|
|
|
%2 = tail call float @copysignf(float 5.000000e-01, float %1) nounwind readnone
|
|
|
|
%3 = fadd float %1, %2
|
|
|
|
ret float %3
|
|
|
|
}
|
|
|
|
|
|
|
|
declare double @bar(...)
|
Fix buggy fcopysign lowering.
This
define float @foo(float %x, float %y) nounwind readnone {
entry:
%0 = tail call float @copysignf(float %x, float %y) nounwind readnone
ret float %0
}
Was compiled to:
vmov s0, r1
bic r0, r0, #-2147483648
vmov s1, r0
vcmpe.f32 s0, #0
vmrs apsr_nzcv, fpscr
it lt
vneglt.f32 s1, s1
vmov r0, s1
bx lr
This fails to copy the sign of -0.0f because it's lost during the float to int
conversion. Also, it's sub-optimal when the inputs are in GPR registers.
Now it uses integer and + or operations when it's profitable. And it's correct!
lsrs r1, r1, #31
bfi r0, r1, #31, #1
bx lr
rdar://8984306
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@125357 91177308-0d34-0410-b5e6-96231b3b80d8
2011-02-11 02:28:55 +00:00
|
|
|
declare double @copysign(double, double) nounwind
|
|
|
|
declare float @copysignf(float, float) nounwind
|