[LLVMbugs] [Bug 12590] New: Unnecessary FP multiply by zero generated

[bugzilla-daemon at llvm.org]

http://llvm.org/bugs/show_bug.cgi?id=12590

             Bug #: 12590
           Summary: Unnecessary FP multiply by zero generated
           Product: clang
           Version: 3.0
          Platform: PC
        OS/Version: All
            Status: NEW
          Severity: normal
          Priority: P
         Component: -New Bugs
        AssignedTo: unassignedclangbugs at nondot.org
        ReportedBy: ckmiller at gmail.com
                CC: llvmbugs at cs.uiuc.edu
    Classification: Unclassified


Created attachment 8408
  --> http://llvm.org/bugs/attachment.cgi?id=8408
Example program

Clang will emit multiply instructions when one of the operands is known to be
zero at compile time.  When run on the attached example, clang 3.0 (with -O3)
emits the following:

_dot:
pushq    %rbp
movq    %rsp,%rbp
pxor    %xmm3,%xmm3
mulss    %xmm3,%xmm2
mulss    %xmm1,%xmm3
addss    %xmm3,%xmm0
addss    %xmm2,%xmm0
popq    %rbp
ret

For comparison, gcc (with -O3) just produces an empty function, leaving the
input value in %xmm0 to be returned:

_dot:
pushq    %rbp
movq    %rsp,%rbp
leave
ret

Interestingly, clang does catch and eliminate the multiply by 1...

-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.