[llvm-bugs] [Bug 46461] [X86] reduceBuildVecExtToExtBuildVec breaks uses of opportunity to use broadcast

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=46461

Simon Pilgrim <llvm-dev at redking.me.uk> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
 Fixed By Commit(s)|                            |6bdb3ce4529ffbaad0354b052f2
                   |                            |e9f06b4431085
             Status|NEW                         |RESOLVED

--- Comment #3 from Simon Pilgrim <llvm-dev at redking.me.uk> ---
(In reply to Craig Topper from comment #0)
> Should we disable reduceBuildVecExtToExtBuildVec for splat vectors or should
> we improve our broadcast matching? I kind of think that even without native
> broadcast we could still use pshufd to broadcast instead of multiple inserts.

Fixed by
https://github.com/llvm/llvm-project/commit/6bdb3ce4529ffbaad0354b052f2e9f06b4431085

Restricting reduceBuildVecExtToExtBuildVec zext cases to non-splat buildvectors
helped improve some broadcastm codegen as well.

-- 
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/20200627/316c8e69/attachment-0001.html>