[llvm-bugs] [Bug 45941] New: Suboptimal optimization for vector-move operation

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=45941

            Bug ID: 45941
           Summary: Suboptimal optimization for vector-move operation
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Backend: X86
          Assignee: unassignedbugs at nondot.org
          Reporter: daan at dsprenkels.com
                CC: craig.topper at gmail.com, llvm-bugs at lists.llvm.org,
                    llvm-dev at redking.me.uk, spatel+llvm at rotateright.com

https://godbolt.org/z/a3mRRf

---

On X86, the shuffle operation

  ymm: [x0: i32, x1: i32, x2: i32, x3: i32, ...]
=>
  ymm: [x0: i64, x1: i64, x2: i64, x3: i64]

can be implemented efficiently using a `vpmovzxdq` instruction.

LLVM uses a normal `vpshufb` instruction, which needs to load the list of
indices. Using a `vpmovzxdq` instruction would erase this load.

-- 
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/20200515/342879ec/attachment-0001.html>