tra abandoned this revision.
tra added a comment.
It appears that the issue that originally prompted this change is due to
suspected bug in glibc triggered by specific details of our internal build.
https://reviews.llvm.org/D49763
___
cfe-commits m
tra planned changes to this revision.
tra added a comment.
Ugh. Apparently moving this code up just disabled module destructor. :-( That
explains why we no longer crash.
https://reviews.llvm.org/D49763
___
cfe-commits mailing list
cfe-commits@lists
joerg added a comment.
Depends a bit on the platform, __cxa_atexit on most modern ELF systems,
fallback to atexit. If the global dtor is run too late, it smells like a
missing library dependency. They are executed in topological order after all.
https://reviews.llvm.org/D49763
_
tra added a comment.
In https://reviews.llvm.org/D49763#1174283, @joerg wrote:
> Can this ever end up in a shared library? If yes, please use the normal logic
> for creating a global destructor. atexit is not very friendly to dlopen...
Yes, it can end up in a shared library. What would be the
joerg added a comment.
Can this ever end up in a shared library? If yes, please use the normal logic
for creating a global destructor. atexit is not very friendly to dlopen...
https://reviews.llvm.org/D49763
___
cfe-commits mailing list
cfe-commits
jlebar accepted this revision.
jlebar added inline comments.
This revision is now accepted and ready to land.
Comment at: clang/lib/CodeGen/CGCUDANV.cpp:379
+ // Create destructor and register it with atexit() the way NVCC does it.
Doing
+ // it during regular destructor phase
tra created this revision.
tra added reviewers: jlebar, timshen.
Herald added subscribers: bixia, sanjoy.
There's apparently a race between fatbin destructors registered by us
and some internal calls registered by CUDA runtime from cudaRegisterFatbin.
Moving fatbin de-registration to atexit() was
