[llvm-bugs] [Bug 43542] Poor codegen for targets with no native shifts

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=43542

Sanjay Patel <spatel+llvm at rotateright.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
             Status|NEW                         |RESOLVED
 Fixed By Commit(s)|                            |r373666

--- Comment #11 from Sanjay Patel <spatel+llvm at rotateright.com> ---
(In reply to Joan Lluch from comment #10)
> On the other hand, some issues are difficult to explain because they never
> show on the trunk targets due to a different combination of features. It's
> also understandable that it's hard to get approval of changes that will not
> bring any real improvement (or fixes to bugs that don't show on the existing
> backends, in spite that they may be actual bugs in the LLVM code).

Yes, I agree - it's going to take more effort to make changes without a known
test/improvement. Feel free to raise that on llvm-dev.

Hopefully, this report has provided a recipe for fixing the simplest cases.

As I mentioned before, changing instcombine canonicalization is generally a
higher degree of difficulty.

For CodeGenPrepare patches, if you can show the problem in IR alone (use "opt
-codegenprepare"), that is preferred to creating an explicit test for a
specific target's asm, and it might make your job easier.


https://reviews.llvm.org/rL373666

-- 
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/20191003/f2ba839a/attachment-0001.html>