- 10 Nov, 2018 1 commit
-
-
Morris Jette authored
Without this patch test35.4 generates errors of this sort: error: select/cons_tres: _compute_c_b_task_dist: Invalid allocation for JobId=336012: job_res->cpus is NULL
-
- 09 Nov, 2018 5 commits
-
-
Morris Jette authored
-
Morris Jette authored
-
Morris Jette authored
-
Morris Jette authored
Based upon commit c2ff383a bug 5976
-
Michael Hinton authored
bug 5402
-
- 08 Nov, 2018 4 commits
-
-
Broderick Gardner authored
Affects account names, partition names, user names, and others. Bug 5992
-
Morris Jette authored
Bug 5402
-
Michael Hinton authored
Record which GPUs are connected to which via nvlinks. Determine if each nvlink is single, double, no connect, or self. Create device index to PCI bus id lookup table. Fix test 39.18 by reverting parsable prints to always use info(). Add better valgrind support for test 39.18. Update comments and remove whitespace. We want the print statements meant for the regression tests to print out everytime, even if user does not specify debug level. Bug 5402
-
Morris Jette authored
-
- 07 Nov, 2018 8 commits
-
-
Morris Jette authored
Coverity CID 189596
-
Morris Jette authored
-
Morris Jette authored
-
Morris Jette authored
bug 5402
-
Danny Auble authored
-
Morris Jette authored
Bug 5826
-
Danny Auble authored
Bug 5793
-
Morris Jette authored
bug 5402
-
- 06 Nov, 2018 8 commits
-
-
Michael Hinton authored
Required for changes in commit 78830a1d Bug 5402
-
Michael Hinton authored
Auto-detect system GPUs, so no need to specify in gres.conf. Add ignore syntax to gres.conf. Test for NVML library existence. Create tests for GRES GPU code. Update docs. Bug 5402
-
Danny Auble authored
-
Michael Hinton authored
when trying to register. If slurmd never connects to slurmctld and then slurmd is shutdown, there is a small memory leak. This is not an important memory leak; however, this will fix things for valgrind when testing a slurmd by itself. Bug 5989
-
Morris Jette authored
Finish NVLink optimization for all of the new GPU options: --gpus (per job),--gpus-per-node (or --gres), --gres-per-socket, and --gres-per-task
-
Morris Jette authored
without this change, the GRES allocate count on a node will be wrong if multiple attempts are made to schedule the job
-
Morris Jette authored
-
Morris Jette authored
flesh out gpu selection optimization based upon nvlink connectivity whe the --gpu (per node) specification is used
-
- 05 Nov, 2018 6 commits
-
-
Michael Hinton authored
Set errno to return value of pthread_* functions during error. All pthread macros now handle errors in the same way. Do not set errno immediately, because in threaded programs errno is a macro that expands into a function call. We don't want to incur that performance cost unless it's actually an error. See The Linux Programming Interface, 3.5.2 (pg.53) and 29.2 (pg 621). Bug 5901
-
Morris Jette authored
no change in logic
-
Danny Auble authored
mixed case compares. Bug 5976
-
Danny Auble authored
-
Morris Jette authored
Bug 5976
-
Artem Polyakov authored
Bug 5917
-
- 03 Nov, 2018 8 commits
-
-
Tim Wickberg authored
Keep the same enum value since this will be the same functional error.
-
Tim Wickberg authored
Rename slurm_crypto_type() to slurm_cred_type(). And rename related variables as well.
-
Tim Wickberg authored
-
Tim Wickberg authored
-
Tim Wickberg authored
Rename _slurm_crypto_{init,fini} to _slurm_cred_{init,fini} as well.
-
Tim Wickberg authored
All these lengths will need to be packed/unpacked through packmem/unpackmem, which will needs to handle them as uint32_t anyways.
-
Tim Wickberg authored
-
Tim Wickberg authored
Start of process to rename CryptoType to CredType. Bug 5916.
-