-
-
Notifications
You must be signed in to change notification settings - Fork 5.5k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Upgrade to LLVM 14.0.2 #45195
Upgrade to LLVM 14.0.2 #45195
Conversation
I think the current build error can be fixed by moving |
3d35596
to
b3febcb
Compare
Win64 looks like a fun one:
|
@nanosoldier |
@nanosoldier |
Your benchmark job has completed - possible performance regressions were detected. A full report can be found here. |
foldl seems to be the biggest regression here (~8X), and some of the summation benchmarks seem to have improved for floats. |
That big of change sounds like a failure to SIMD |
Your package evaluation job has completed - possible new issues were detected. A full report can be found here. |
The aarch64 failure is known:
It's the reason we switched to JITLink for aarch64-darwin. cc: @lhames @weliveindetail I will happily offer support if the GSoC student wants use Julia as a stress-test for JITLink on Elf |
on aarch64-darwin and
from PkgEval seem to be the biggest issues. Both are GC related. |
@nanosoldier |
Your package evaluation job has completed - possible new issues were detected. A full report can be found here. |
|
@nanosoldier |
1a4b425
to
26d4516
Compare
@vtjnash I am stuck on the Windows test somehow a linux path creeps into the backtrace. |
https://reviews.llvm.org/D100944 introduces will split sections if the metadata does not match. The GlobalVariables are `RW_` and the `.text` section is `R_X`. Currently there is no facility to actually mark the GV as `R_X` and we only need to write them during code-emission before we flip the permissions.
@nanosoldier |
@nanosoldier |
Your benchmark job has completed - possible performance regressions were detected. A full report can be found here. |
Your package evaluation job has completed - possible new issues were detected. A full report can be found here. |
Maybe want to run PkgEval with assertions for PRs like this? |
That would be a great option to have! |
@DilumAluthge @tkf and I chatted about ASAN and we decided that for now we should make it "allowed to fail" best to keep the "can we build" bit though. |
I believe we already have this option. Control-F https://github.com/JuliaCI/Nanosoldier.jl/blob/master/README.md for "assert". |
@nanosoldier |
Your package evaluation job has completed - possible new issues were detected. A full report can be found here. |
Feels like this should give some comment about all the assertion failures in the PkgEval report. |
It probably also should have been a comparison to master with assertions, so that any assertion reported would have been a new one. |
... I am blind. We really need to improve the formatting (and order of the report). I looked at the report on my phone and mentally lumped the process aborted and segmentation fault reports under "Networking related" since the heading got completely lost. Sorry about that. |
We should probably run PkgEval with assertions periodically to catch some assertion specific errors earlier. |
TODO:
foldl