[llvm-bugs] [Bug 31712] New: Do we need X86ISD::VSEXT and X86ISD::VZEXT?

via llvm-bugs llvm-bugs at lists.llvm.org
Fri Jan 20 12:25:20 PST 2017


https://llvm.org/bugs/show_bug.cgi?id=31712

            Bug ID: 31712
           Summary: Do we need X86ISD::VSEXT and X86ISD::VZEXT?
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Windows NT
            Status: NEW
          Severity: normal
          Priority: P
         Component: Backend: X86
          Assignee: unassignedbugs at nondot.org
          Reporter: llvm-dev at redking.me.uk
                CC: craig.topper at gmail.com, elena.demikhovsky at intel.com,
                    llvm-bugs at lists.llvm.org, mkuper at google.com,
                    spatel+llvm at rotateright.com
            Blocks: 30624
    Classification: Unclassified

Everything that X86ISD::VSEXT/X86ISD::VZEXT does can be performed with the ISD
SIGN/ZERO_EXTEND and SIGN/ZERO_EXTEND_VECTOR_IN_REG ops.

Using these directly would help us reuse existing generic combines and would
probably assist with (or at least avoid) various legalization/canonicalization
issues (such as what is going on with D28537).

AVX512 predicate mask extension may be tricky but it might make sense to keep
the X86ISD opcodes just for those cases, at least initially.

-- 
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/20170120/812e24f8/attachment.html>


More information about the llvm-bugs mailing list