[LLVMbugs] [Bug 7703] New: All shuffle intrinsics use PSHUFD (even when SHUFPS is requested and smaller)

[bugzilla-daemon at llvm.org]

http://llvm.org/bugs/show_bug.cgi?id=7703

           Summary: All shuffle intrinsics use PSHUFD (even when SHUFPS is
                    requested and smaller)
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: normal
          Priority: P
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: jed at 59a2.org
                CC: llvmbugs at cs.uiuc.edu


void f(const float *x,float *y) {
  __m128 p = _mm_load_ps(x);
  p = _mm_shuffle_ps(p,p,_MM_SHUFFLE(0,2,1,3));
  _mm_store_ps(y,p);
}

Compile with -Os -fomit-frame-pointer (clang version 1.1
(branches/release_27)):

00000000004004c4 <f> movaps xmm0,XMMWORD PTR [rdi]
00000000004004c7 <f+0x3> pshufd xmm0,xmm0,0x27
00000000004004cc <f+0x8> movaps XMMWORD PTR [rsi],xmm0
00000000004004cf <f+0xb> ret

Compare to gcc (4.5.0) which is one byte shorter and actually uses the
instruction specified by the intrinsic:

00000000004004c4 <f> movaps xmm0,XMMWORD PTR [rdi]
00000000004004c7 <f+0x3> shufps xmm0,xmm0,0x27
00000000004004cb <f+0x7> movaps XMMWORD PTR [rsi],xmm0
00000000004004ce <f+0xa> ret

-- 
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.