- User Since
- Jul 7 2012, 2:54 PM (318 w, 4 d)
Include a minor update to the documentation to make it clear that this
component is now implemented, and only jumptables remain unhandled (but can be
handled with retpolines).
Update this to be the complete implementation of pre- and post-instruction
Try to improve comments.
Now using API suggested by Reid.
Switch to API suggested in review.
Tue, Aug 14
Try to shrink the code of the ExtraInfo class a bit and add some helpful comments about how all of these moving pieces interact.
Update to remove UB from PointerSumType after consulting with Richard Smith.
This should also make the patterns used more likely to work with our supported
Thanks all. Think I have everything so landing. Give a shout if anything else jumps out.
Mon, Aug 13
I think all done.
Address review feedback.
Update to include the optimization where we store a single MMO inline to avoid
allocation. Almost nothing changed, all tests continue to pass.
Doh, this doesn't have the change to do the clever optimization mentioned in the change description. I'll update it with that momentarily. But that will be just a tiny implementation detail, so probably fin eto start reviewing.
This makes sense to me, but definitely wait for someone more familiar w/ Clang's IR gen to review...
Mon, Aug 6
Sun, Aug 5
Sat, Aug 4
This looks really good to me and seems like a nice clarification of historical practice. =D Thanks so much for driving an actual documentation update for folks that simply would never know about these practices otherwise, I think it will help folks a lot.
Wed, Aug 1
Tue, Jul 31
Mon, Jul 30
We have many, many for loops still using iterators for this....
Sat, Jul 28
Now with correct usage of the ABI breaking checks macro and fixing a silly
Update with actual documentation (instead of it being missing or talking about
the wrong thing) and a more direct implementation strategy.
Fri, Jul 27
LGTM! Thanks for the improved comment. Hopefully at least helps future me when I get confused reading it again.
LGTM, but *please* update the comments in the update API to make this clear. I specifically thought there was no difference here after reading the comments, so I think they need to make this way more clear.
After thinking more about it, this seems like the only sensible course.
Thu, Jul 26
Wed, Jul 25
Before anyone gets too excited about the performance improvements here, I should mention that I am working on implementing another component of the high level SLH design that I'm worried will completely defeat all of these improvements. Maybe even hold off reviewing, as this whole thing may not be relevant. =/ Sorry for the noise.
Tue, Jul 24
All done and submitting! Thanks!
Confirmed that this does-the-right-thing with retpolines (as they get
a different instruction).
Rebase on top of the cleaned up folding patch, and now without any copy/paste
code smells. We just call out to the generic low-level register hardening
Use NewMI.getDesc() rather than looking it up in TII again. Thanks to Craig for the suggestion.
Mon, Jul 23
Split the TCRETURN folding into https://reviews.llvm.org/D49717 to make the
messy situation there more clear.
Sun, Jul 22
Thu, Jul 19
Wed, Jul 18
(Feel free to commandeer back if someone else wants to land this)