This is an archive of the discontinued LLVM Phabricator instance.

Fix for #19973
Needs ReviewPublic

Authored by kvanberendonck on Jun 16 2014, 3:21 AM.
This revision needs review, but there are no reviewers specified.

Details

Reviewers
None
Summary

Fix for #19973

http://llvm.org/bugs/show_bug.cgi?id=19973

Fails 1 test, which seems to be a test checking for the behaviour I don't think we want.

Diff Detail

Event Timeline

kvanberendonck retitled this revision from to Fix for #19973.
kvanberendonck updated this object.
kvanberendonck edited the test plan for this revision. (Show Details)

Sorry for the whitespace. Originally I had the constant offset check there, but I had to move it because shouldMergeGEP wasn't doing the right thing.

Got to write a regression test and remove the broken test case.

kvanberendonck added a subscriber: Unknown Object (MLST).Jun 16 2014, 7:56 PM
kvanberendonck updated this revision to Diff 10665.EditedJun 19 2014, 5:02 PM

Now removed the test which was failing (testing for incorrect behaviour) and added 2 new regression test cases -- one of which currently fails (test_gep_canon_2_add). I'm working on getting it right again.

There may be a constant floating problem in the direction we float constants outlined by the bottom tests. It's hard to pull constants out of stacks of arithmetic because they are moved to the bottom(?). It would be more efficient if they were floated to the top (i.e. closer to the things that use them), maybe.

chandlerc resigned from this revision.Mar 29 2015, 12:43 PM
chandlerc removed a reviewer: chandlerc.

This patch has gone quite stale. I don't really understand why this particular change is the right one, I think it'll need some comments or other clarification in that respect before we accept it... If you end up wanting to continue working on this, try to do that, update the patch, and re-add me so I can take a look.