[LLVMbugs] [Bug 15155] New: PPC: Optimize vector splats with more than 3 bits better.

[bugzilla-daemon at llvm.org]

http://llvm.org/bugs/show_bug.cgi?id=15155

             Bug #: 15155
           Summary: PPC: Optimize vector splats with more than 3 bits
                    better.
           Product: libraries
           Version: trunk
          Platform: PC
        OS/Version: All
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Backend: PowerPC
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: benny.kra at gmail.com
                CC: llvmbugs at cs.uiuc.edu
    Classification: Unclassified


After r174325 we compile

define <4 x i32> @foo() {
  ret <4 x i32> <i32 16, i32 16, i32 16, i32 16>
}

into a load:
    ld 3, .LC1 at toc(2)
    lvx 2, 0, 3
    blr

instead of a splat:
    vspltisw 2, 8
    vadduwm 2, 2, 2
    blr

The code in PPCISelLowering.cpp that generates the splat was disabled in
r174325 because it fails to cope with constant folding of the ADD node.

-- 
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.