Master: new thread-specific-data (tsd) api #7936
Conversation
|
The IBM CI (GNU/Scale) build failed! Please review the log, linked below. Gist: https://gist.github.com/925c3fbc0ca22abcb96bfbdde1f93038 |
|
The IBM CI (XL) build failed! Please review the log, linked below. Gist: https://gist.github.com/60f22560db54b89ec2d79100f3f6faf9 |
|
The IBM CI (PGI) build failed! Please review the log, linked below. Gist: https://gist.github.com/f6f99e857a576b69886b24b9cc50e304 |
janjust
changed the title
janjust
force-pushed
the
master-new-tsd-thread-api
branch
from
0132c13 to
4e7f9c1
Compare
|
The IBM CI (GNU/Scale) build failed! Please review the log, linked below. Gist: https://gist.github.com/5d174a978a7743ea2b18192f0dc208c2 |
|
The IBM CI (XL) build failed! Please review the log, linked below. Gist: https://gist.github.com/1adbed4c69b0dfe754ffaddc956606bf |
|
The IBM CI (PGI) build failed! Please review the log, linked below. Gist: https://gist.github.com/4e1d191589fcb5d5c1e6dca1d58d459f |
janjust
force-pushed
the
master-new-tsd-thread-api
branch
from
4e7f9c1 to
dd7fabb
Compare
|
The IBM CI (GNU/Scale) build failed! Please review the log, linked below. Gist: https://gist.github.com/84c7d85a1f8b8ab3224789dce3b62f5a |
|
The IBM CI (XL) build failed! Please review the log, linked below. Gist: https://gist.github.com/70da53d41c63516b9c4c837284c4a4ad |
|
The IBM CI (PGI) build failed! Please review the log, linked below. Gist: https://gist.github.com/e3c073babee2bd482b15c35fddefa398 |
janjust
force-pushed
the
master-new-tsd-thread-api
branch
from
dd7fabb to
86f837f
Compare
|
bot:retest |
janjust
force-pushed
the
master-new-tsd-thread-api
branch
3 times, most recently
from
9a1c4b7 to
852adc3
Compare
|
bot:ompi:retest |
|
bot:retest |
|
Looking closely, it's IPv6. |
janjust
force-pushed
the
master-new-tsd-thread-api
branch
from
852adc3 to
6de1faf
Compare
|
@bosilca @hppritcha please have a look and approve if OK. |
|
P.S. Sorry for the type in the prev comment: "not cleared" = "now cleared" |
janjust
force-pushed
the
master-new-tsd-thread-api
branch
from
6de1faf to
59403de
Compare
|
Just fyi - forced push after finding a potential bug, and reverted key type from pointer back to static. |
janjust
force-pushed
the
master-new-tsd-thread-api
branch
from
59403de to
59aff0d
Compare
janjust
force-pushed
the
master-new-tsd-thread-api
branch
from
59aff0d to
708bdde
Compare
|
@janjust please rebase to resolve the conflict. |
Co-authored-by: Artem Polyakov <[email protected]> Signed-off-by: Tomislav Janjusic <[email protected]>
Co-authored by: Artem Polykaov <[email protected]> Signed-off-by: Tomislav Janjusic <[email protected]>
Co-authored by: Artem Polykaov <[email protected]> Signed-off-by: Tomislav Janjusic <[email protected]>
Co-authored by: Artem Polykaov <[email protected]> Signed-off-by: Tomislav Janjusic <[email protected]>
janjust
force-pushed
the
master-new-tsd-thread-api
branch
from
708bdde to
d809f6b
Compare
|
@artpol84 done |
This PR adds a new opal/thread/tsd API interface that avoids storing keys in a global structure.
Instead we introduce 2 types of API:
The user of the convenience API is responsible of calling the destruction explicitly instead of relying on MPI_Finalize (as it happens currently).
We also update/fix existing usages of the old API to retain functionality.
As is, this compiles, but hasn't been tested.
It's here to advance the discussion regarding the opal thread api changes that we're proposing.
I will be closing #7877 in favor of this PR.