[PATCH] D87188: [InstCombine] Canonicalize SPF to abs intrinc

[Martin Storsjö via Phabricator via cfe-commits]

mstorsjo added a comment.

In D87188#2447958 <https://reviews.llvm.org/D87188#2447958>, @lebedev.ri wrote:

> Just to be sure, i just run the entire compilation of `vc1_block-aarch64.c` through alive2, and as far as i can tell, it did not report any IR-level miscompilations.
> So either the original code has bugs/UB's/whatever, or this is an AArch64 backend bug (cc @t.p.northover), or the IR problem is in a place alive2 can't find.

FWIW, the testcase does run without any remarks in ubsan at least...

I retried the issue now, and I've reduced it down to one standalone function that causes the errors. https://martin.st/temp/vc1_block-p-block2.c is this function separated into a standalone translation unit - that I can run in the test framework. Unfortunately, the rather small difference further up in the optimization chain cascades into quite a lot of differences at the end, and I don't know really where to go from there to pinpoint the issue.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D87188/new/

https://reviews.llvm.org/D87188