-
-
Notifications
You must be signed in to change notification settings - Fork 5.6k
Bump LLVM to v20 #58142
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
Bump LLVM to v20 #58142
Conversation
1e6ec21
to
f20e031
Compare
Good news: Julia unit tests already pass! Only issues at the moment are
|
I guess the asan issue was not in fact fixed upstream Lines 1989 to 1999 in 37a7541
|
@nanosoldier |
Your benchmark job has completed - possible performance regressions were detected. A full report can be found here. |
Seems to be usual mixed bag of regressions and improvements, but with no major regressions for once. And SIMD benchmarks appear to have some significant boost. |
Can't think of an easy way to make the change to the gc push pop test (88a55e9) compatible with llvm <= 19.
#include <stddef.h>
int test(size_t nel, size_t elsz) {
size_t nbytes;
int overflow = __builtin_mul_overflow(nel, elsz, &nbytes);
int overflow2 = __builtin_add_overflow(nel, nbytes, &nbytes);
return overflow * overflow2;
}
|
Last static analyzer failure seems to be it complaining that Lines 289 to 306 in da99461
Called from Lines 1384 to 1394 in da99461
Lines 1298 to 1321 in da99461
doublesplit != 0
|
Good info. I marked those instead, so we can merge this once those finish getting through CI |
Fix #57021. Split out from #58142, not sure if there is a better way to fix this. --------- Co-authored-by: Jameson Nash <[email protected]>
Seems to be failing with
|
well reverting seems to have gotten the mac builds further, but I think the actual issue is something else' |
The linux asserts build is still failing so I doubt it was my change that caused this. |
That is looking pretty good. Can we drop f2a6b84 now? |
663b59f
to
a6b4cef
Compare
Alright, down to one false-positive clang-sa-subtype null-pointer detected. I'm not entirely sure why I don't get that one locally (I got a different one). Tried to add some annotations to make that happy, and some Makefile flags so that adding the annotations aren't a different error. |
Fix JuliaLang#57021. Split out from JuliaLang#58142, not sure if there is a better way to fix this. --------- Co-authored-by: Jameson Nash <[email protected]>
The NONNULL stuff has made clang very unhappy on a normal build |
@nanosoldier |
Fix analyzer and sanitizer issues with update The analyzer assumes that casts mean we used JL_GC_PUSHARGS. Simplify the typemap code to store the values in separate variables. Don't make the memcpy at all: it seems like it was UB in the first place (we need to wait for https://developers.redhat.com/articles/2024/12/11/making-memcpynull-null-0-well-defined to land) Workaround llvm/llvm-project#136292 Fix makefile dependencies for llvm in refresh_checksums
@nanosoldier |
Your benchmark job has completed, but no benchmarks were actually executed. Perhaps your tag predicate contains misspelled tags? cc @ |
@nanosoldier |
Your benchmark job has completed - possible performance regressions were detected. A full report can be found here. |
No description provided.