[LLVMbugs] [Bug 11251] New: lshr of character vector inefficient on x86

[bugzilla-daemon at llvm.org]

http://llvm.org/bugs/show_bug.cgi?id=11251

           Summary: lshr of character vector inefficient on x86
           Product: new-bugs
           Version: unspecified
          Platform: PC
        OS/Version: All
            Status: NEW
          Severity: normal
          Priority: P
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: pete.cooper at gmail.com
                CC: llvmbugs at cs.uiuc.edu


The following code is expanded out to 16 extracts, 16 byte shifts, and 16
inserts.

As neighbouring bytes (0 1, 2 3, etc) have equal shift amounts, this could be
done with an 8xi16 shift and then a mask on the high bits of the odd bytes
which were shifted down from the even byte in the pair but which should instead
be 0.

define <16 x i8> @shift_vec16x8(<16 x i8> %var) {
entry:
%0 = lshr <16 x i8> %var, <i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2,
                           i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2>
ret <16 x i8> %0
}

-- 
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.