mirror of
https://github.com/c64scene-ar/llvm-6502.git
synced 2024-12-26 05:32:25 +00:00
1d02acb7a0
This trades a (register-renamer-friendly) movaps for a floating point / integer domain cross. That is a very bad trade, even on architectures where domain crossing is relatively fast. On any chip where there is even a cycle stall, this is a Very Bad Idea. It doesn't even seem likely to cause a spill to be introduced because the reason for the copy is to destructively shuffle in place. Thanks to Ben Kramer for fixing a bug in this code that my new shuffle lowering exposed and highlighting that perhaps it should just go away. =] git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@219090 91177308-0d34-0410-b5e6-96231b3b80d8
70 lines
2.0 KiB
LLVM
70 lines
2.0 KiB
LLVM
; RUN: llc < %s -mtriple=x86_64-apple-macosx10.9.0 -mcpu=core2 | FileCheck %s --check-prefix=OSX_SINCOS
|
|
; RUN: llc < %s -mtriple=x86_64-apple-macosx10.8.0 -mcpu=core2 | FileCheck %s --check-prefix=OSX_NOOPT
|
|
; RUN: llc < %s -mtriple=x86_64-pc-linux-gnu -mcpu=core2 -enable-unsafe-fp-math | FileCheck %s --check-prefix=GNU_SINCOS
|
|
|
|
; Combine sin / cos into a single call.
|
|
; rdar://13087969
|
|
; rdar://13599493
|
|
|
|
define float @test1(float %x) nounwind {
|
|
entry:
|
|
; GNU_SINCOS-LABEL: test1:
|
|
; GNU_SINCOS: callq sincosf
|
|
; GNU_SINCOS: movss 4(%rsp), %xmm0
|
|
; GNU_SINCOS: addss (%rsp), %xmm0
|
|
|
|
; OSX_SINCOS-LABEL: test1:
|
|
; OSX_SINCOS: callq ___sincosf_stret
|
|
; OSX_SINCOS: movaps %xmm0, %xmm1
|
|
; OSX_SINCOS: shufps {{.*}} ## xmm1 = xmm1[1,1,2,3]
|
|
; OSX_SINCOS: addss %xmm0, %xmm1
|
|
|
|
; OSX_NOOPT: test1
|
|
; OSX_NOOPT: callq _sinf
|
|
; OSX_NOOPT: callq _cosf
|
|
%call = tail call float @sinf(float %x) nounwind readnone
|
|
%call1 = tail call float @cosf(float %x) nounwind readnone
|
|
%add = fadd float %call, %call1
|
|
ret float %add
|
|
}
|
|
|
|
define double @test2(double %x) nounwind {
|
|
entry:
|
|
; GNU_SINCOS-LABEL: test2:
|
|
; GNU_SINCOS: callq sincos
|
|
; GNU_SINCOS: movsd 16(%rsp), %xmm0
|
|
; GNU_SINCOS: addsd 8(%rsp), %xmm0
|
|
|
|
; OSX_SINCOS-LABEL: test2:
|
|
; OSX_SINCOS: callq ___sincos_stret
|
|
; OSX_SINCOS: addsd %xmm1, %xmm0
|
|
|
|
; OSX_NOOPT: test2
|
|
; OSX_NOOPT: callq _sin
|
|
; OSX_NOOPT: callq _cos
|
|
%call = tail call double @sin(double %x) nounwind readnone
|
|
%call1 = tail call double @cos(double %x) nounwind readnone
|
|
%add = fadd double %call, %call1
|
|
ret double %add
|
|
}
|
|
|
|
define x86_fp80 @test3(x86_fp80 %x) nounwind {
|
|
entry:
|
|
; GNU_SINCOS-LABEL: test3:
|
|
; GNU_SINCOS: callq sinl
|
|
; GNU_SINCOS: callq cosl
|
|
; GNU_SINCOS: ret
|
|
%call = tail call x86_fp80 @sinl(x86_fp80 %x) nounwind
|
|
%call1 = tail call x86_fp80 @cosl(x86_fp80 %x) nounwind
|
|
%add = fadd x86_fp80 %call, %call1
|
|
ret x86_fp80 %add
|
|
}
|
|
|
|
declare float @sinf(float) readonly
|
|
declare double @sin(double) readonly
|
|
declare float @cosf(float) readonly
|
|
declare double @cos(double) readonly
|
|
|
|
declare x86_fp80 @sinl(x86_fp80)
|
|
declare x86_fp80 @cosl(x86_fp80)
|