[LLVMbugs] [Bug 23464] New: [SSE] Scalar users should be able to fold MOVZQI2PQIrm/MOVDI2PDIrm

[bugzilla-daemon at llvm.org]

https://llvm.org/bugs/show_bug.cgi?id=23464

            Bug ID: 23464
           Summary: [SSE] Scalar users should be able to fold
                    MOVZQI2PQIrm/MOVDI2PDIrm
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: All
            Status: NEW
          Severity: normal
          Priority: P
         Component: Backend: X86
          Assignee: unassignedbugs at nondot.org
          Reporter: ahmed.bougacha at gmail.com
                CC: llvmbugs at cs.uiuc.edu
    Classification: Unclassified

It's trickier than MOVSS/MOVSD because they zero the upper lanes, but if we're
able to recognize that a load feeds into a scalar op (PR23349), this is
probably as easy as canFoldLoad=1 on QI2PQI/DI2PDI.

It's also much less useful, because most integer domain instructions are packed
anyway.  There are a few that aren't though;  consider:

    _mm256_broadcastq_epi64(_mm_loadl_epi64(src))

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20150509/ef5263ab/attachment.html>