[LLVMbugs] [Bug 17850] New: missed opportunities to use lower precision cmath functions

[bugzilla-daemon at llvm.org]

http://llvm.org/bugs/show_bug.cgi?id=17850

            Bug ID: 17850
           Summary: missed opportunities to use lower precision cmath
                    functions
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: All
            Status: NEW
          Severity: normal
          Priority: P
         Component: Transformation Utilities
          Assignee: unassignedbugs at nondot.org
          Reporter: kkhoo at perfwizard.com
                CC: llvmbugs at cs.uiuc.edu
    Classification: Unclassified

LLVM isn't optimizing math.h/cmath library calls based on precision of the
inputs and outputs:

$ cat ~/Desktop/cos.c
#include <math.h>
float foo(float x) { return cos(x); }

$ ./clang -O3 -S -o - ~/Desktop/cos.c
...
    cvtss2sd    %xmm0, %xmm0
    callq    _cos
    cvtsd2ss    %xmm0, %xmm0
    popq    %rbp
    ret
...

GCC gets rid of the precision conversions and makes this a call to _cosf.

Based on the code in LibCallSimplifierImpl::lookupOptimization(), it appears
that LLVM should be making this optimization assuming unsafe math, but there's
no difference if I use -ffast-math.

So there are 2 potential bugs here:
1. Why is LLVM failing to optimize this code with fast-math?
2. Why doesn't LLVM do this optimization regardless of fast-math (unsafe FP)?


This is with:
$ ./clang -v
clang version 3.4 (trunk 194153)
Target: x86_64-apple-darwin11.4.2
Thread model: posix


And here's the IR:
target datalayout =
"e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
target triple = "x86_64-apple-macosx10.7.0"

; Function Attrs: nounwind readnone ssp uwtable
define float @foo(float %x) #0 {
entry:
  %conv = fpext float %x to double
  %call = tail call double @cos(double %conv) #2
  %conv1 = fptrunc double %call to float
  ret float %conv1
}

; Function Attrs: nounwind readnone
declare double @cos(double) #1

attributes #0 = { nounwind readnone ssp uwtable "less-precise-fpmad"="false"
"no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf"
"no-infs-fp-math"="false" "no-nans-fp-math"="false"
"stack-protector-buffer-size"="8" "unsafe-fp-math"="false"
"use-soft-float"="false" }
attributes #1 = { nounwind readnone "less-precise-fpmad"="false"
"no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf"
"no-infs-fp-math"="false" "no-nans-fp-math"="false"
"stack-protector-buffer-size"="8" "unsafe-fp-math"="false"
"use-soft-float"="false" }
attributes #2 = { nounwind readnone }

!llvm.ident = !{!0}

!0 = metadata !{metadata !"clang version 3.4 (trunk 194153)"}

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20131108/a22c9d21/attachment.html>