b14e83d1a4
We currently have log, log2, log10, exp and exp2 intrinsics. Add exp10 to fix this asymmetry. AMDGPU already has most of the code for f32 exp10 expansion implemented alongside exp, so the current implementation is duplicating nearly identical effort between the compiler and library which is inconvenient. https://reviews.llvm.org/D157871
82 lines
2.4 KiB
LLVM
82 lines
2.4 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
|
|
; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s
|
|
; RUN: llc < %s -mtriple=i386-pc-win32 | FileCheck %s -check-prefix=CHECK-WIN
|
|
|
|
define float @call_exp10f(float %a) {
|
|
; CHECK-LABEL: call_exp10f:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: pushq %rax
|
|
; CHECK-NEXT: .cfi_def_cfa_offset 16
|
|
; CHECK-NEXT: callq exp10f@PLT
|
|
; CHECK-NEXT: popq %rax
|
|
; CHECK-NEXT: .cfi_def_cfa_offset 8
|
|
; CHECK-NEXT: retq
|
|
;
|
|
; CHECK-WIN-LABEL: call_exp10f:
|
|
; CHECK-WIN: # %bb.0:
|
|
; CHECK-WIN-NEXT: pushl %eax
|
|
; CHECK-WIN-NEXT: flds {{[0-9]+}}(%esp)
|
|
; CHECK-WIN-NEXT: fstps (%esp)
|
|
; CHECK-WIN-NEXT: calll _exp10f
|
|
; CHECK-WIN-NEXT: popl %eax
|
|
; CHECK-WIN-NEXT: retl
|
|
%result = call float @exp10f(float %a)
|
|
ret float %result
|
|
}
|
|
|
|
define double @call_exp10(double %a) {
|
|
; CHECK-LABEL: call_exp10:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: pushq %rax
|
|
; CHECK-NEXT: .cfi_def_cfa_offset 16
|
|
; CHECK-NEXT: callq exp10@PLT
|
|
; CHECK-NEXT: popq %rax
|
|
; CHECK-NEXT: .cfi_def_cfa_offset 8
|
|
; CHECK-NEXT: retq
|
|
;
|
|
; CHECK-WIN-LABEL: call_exp10:
|
|
; CHECK-WIN: # %bb.0:
|
|
; CHECK-WIN-NEXT: subl $8, %esp
|
|
; CHECK-WIN-NEXT: fldl {{[0-9]+}}(%esp)
|
|
; CHECK-WIN-NEXT: fstpl (%esp)
|
|
; CHECK-WIN-NEXT: calll _exp10
|
|
; CHECK-WIN-NEXT: addl $8, %esp
|
|
; CHECK-WIN-NEXT: retl
|
|
%result = call double @exp10(double %a)
|
|
ret double %result
|
|
}
|
|
|
|
define x86_fp80 @call_exp10l(x86_fp80 %a) {
|
|
; CHECK-LABEL: call_exp10l:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: subq $24, %rsp
|
|
; CHECK-NEXT: .cfi_def_cfa_offset 32
|
|
; CHECK-NEXT: fldt {{[0-9]+}}(%rsp)
|
|
; CHECK-NEXT: fstpt (%rsp)
|
|
; CHECK-NEXT: callq exp10l@PLT
|
|
; CHECK-NEXT: addq $24, %rsp
|
|
; CHECK-NEXT: .cfi_def_cfa_offset 8
|
|
; CHECK-NEXT: retq
|
|
;
|
|
; CHECK-WIN-LABEL: call_exp10l:
|
|
; CHECK-WIN: # %bb.0:
|
|
; CHECK-WIN-NEXT: pushl %ebp
|
|
; CHECK-WIN-NEXT: movl %esp, %ebp
|
|
; CHECK-WIN-NEXT: andl $-16, %esp
|
|
; CHECK-WIN-NEXT: subl $32, %esp
|
|
; CHECK-WIN-NEXT: fldt 8(%ebp)
|
|
; CHECK-WIN-NEXT: fstpt (%esp)
|
|
; CHECK-WIN-NEXT: calll _exp10l
|
|
; CHECK-WIN-NEXT: movl %ebp, %esp
|
|
; CHECK-WIN-NEXT: popl %ebp
|
|
; CHECK-WIN-NEXT: retl
|
|
%result = call x86_fp80 @exp10l(x86_fp80 %a)
|
|
ret x86_fp80 %result
|
|
}
|
|
|
|
declare float @exp10f(float %a) #0
|
|
declare double @exp10(double %a) #0
|
|
declare x86_fp80 @exp10l(x86_fp80 %a) #0
|
|
|
|
attributes #0 = { nounwind readonly }
|