[LLVMbugs] [Bug 2311] New: LLVM codegen unnecessarily realigns stack for x86 SSE code

[bugzilla-daemon at cs.uiuc.edu]

http://llvm.org/bugs/show_bug.cgi?id=2311

           Summary: LLVM codegen unnecessarily realigns stack for x86 SSE
                    code
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: normal
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: sharparrow1 at yahoo.com
                CC: llvmbugs at cs.uiuc.edu


#include <emmintrin.h>

typedef short vSInt16 __attribute__ ((__vector_size__ (16)));

static const vSInt16 a = {- 22725, - 12873, - 22725, - 12873, - 22725, - 12873,
- 22725, - 12873};;

vSInt16 madd(vSInt16 b)
{
    return _mm_madd_epi16(a, b);
}

Generated code:
madd:
        pushl   %ebp
        movl    %esp, %ebp
        andl    $-16, %esp
        movaps  .LCPI1_0, %xmm1
        pmaddwd %xmm1, %xmm0
        movl    %ebp, %esp
        popl    %ebp
        ret

The stack alignment is unnecessary because nothing is spilled onto the stack.

Probably not a big issue, but it looks silly.


-- 
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.