- User Since
- Jan 3 2020, 4:17 AM (37 w, 1 d)
I'm abandoning this revision for now
Add note to explain purpose of method.
Ready for review now.
Update test case to pass by using previously unused value.
Note that this is not complete and was uploaded at the current state to share a patch to reproduce a bug.
Tue, Sep 15
Jul 29 2020
Jul 27 2020
Correct loop space of the second for loop to be a suffix rather than a prefix.
Jul 21 2020
Jul 17 2020
Jul 15 2020
Jul 7 2020
Refactor code. No new behavior was added.
Jul 6 2020
Jul 1 2020
Thank you for handling this case and the patience in waiting for the reproducer when I originally rolled this back. I'm resigning as a reviewer just to make it clear that I'm not qualified to review this change.
Jun 30 2020
Jun 23 2020
I'm trying to get some help from more experienced coworkers in California side to see how I could help prove this. I also was off in the 2x slowdown. Compilation of one object file is taking slightly under 3x longer and then being stopped by a timeout in infrastructure that I haven't yet found a workaround for. I'm looking into some parts to see if maybe there's an infinite loop while waiting for insights from others.
Jun 22 2020
I reverted the follow up change that enabled this pass as it is causing 2x slowdowns on compilation of some large binaries at Google and assume that it will be affecting others also. It looked maybe related to ThinLTO, but I don't know if that's relevant or not.
Jun 18 2020
Add some tests for individual ops.
I'm going to land this, as I handled Sean's comment, so there doesn't seem to be anything contentious left.
Jun 17 2020
Change constraint remove to replace cstr_ ops with passing witnesses.
Jun 15 2020
I haven't finished handling Sean's comment, but wanted to share my thoughts and pieces so far.
Match MLIR style.
Jun 10 2020
Jun 9 2020
I think you might have deleted your changes to ShapeOps.td, but that might also be me getting confused by this website.
Jun 6 2020
As more people are complaining, I'm going ahead and reverting the CLs that added and rely on this new file.
First, sorry for the breakage.
Jun 5 2020
Jun 4 2020
Replace with folding and handle non-const but equal inputs.
Explain commutativity of any.
Switch to folding and also handle repeated inputs.
Instead of canonicalization, use folding.
Instead of true_witness, have const_witness that can be true or false.
May 29 2020
May 25 2020
Just a note of a typo in your commit message "sahpe". I see you updated the diff with an updated commit message, so I'm guessing it's fixed, but arc doesn't propagate updated commit messages unfortunately.
May 20 2020
Remove named TODO.
I split this commit up into smaller pieces.
@silvas I was going to wait because you hadn't accepted the commit, but because (1) I am not working the next 4 days and (2) have other commits using these operations, I'm going to submit this, but can follow up on any other feedback.
Match example code in documentation to the actual custom form.
May 19 2020
May 18 2020
May 15 2020
Update op descriptions and test based on feedback.
May 12 2020
Update descriptions of operations.
May 11 2020
Please don't review yet. This is a work in progress including the op definitions.
Apr 30 2020
Apr 29 2020
Apr 28 2020
Reorder definition of ops and handle a rebase.
Apr 27 2020
The failing tests seem to have just been failing at head. Failures now are just from the clang-tidy warning of a Twine being used.
std::move last use of string.
Apr 24 2020
Update variable names to llvm style.
Update comment on the kernel attribute of gpu.launch_func.
Creating a new revision to confirm that build failures were unrelated to my change.
Apr 23 2020
Thank you for the quick solution!
Apr 14 2020
Replace some types with auto.
Apr 8 2020
Apr 1 2020
Uday, I landed this thinking you were doing more of a drive-by review, but I'll do a follow up commit if you have any more concerns.
Mar 31 2020
Update map# on test case to allow splitting of test cases.
Follow naming convention for atBlockBegin and atBlockEnd.