Samuel Thibault
|
e180da0966
Allow the application not to define an MPI tag when the data is not to be transferred
|
8 years ago |
Samuel Thibault
|
bd9d103c1a
Fix case when handle gets unregistered just before we check whether there are arbitered waiters
|
8 years ago |
Samuel Thibault
|
e50a934f58
Allow for more concurrent readers with arbiters: we may also want to notify even when the handle is being used, in read mode notably
|
8 years ago |
Samuel Thibault
|
00e62ce185
sched_ctx/sched_ctx_hierarchy can't work on simgrid yet
|
8 years ago |
Samuel Thibault
|
51e73cb81a
Add missing signaling when buffers are transferred
|
8 years ago |
Samuel Thibault
|
b1451ecff2
revert 20410: opencl does not driver several devices
|
8 years ago |
Samuel Thibault
|
c14ba419e6
don't omit setting the local worker key, to make sure _starpu_fetch_task_input_tail works with the proper device
|
8 years ago |
Samuel Thibault
|
97292596ab
write down why the ugly hack
|
8 years ago |
Luka Stanisic
|
76c48b0915
traces: prepending JobId with the prefix (MPI node number)
|
8 years ago |
Samuel Thibault
|
69fe1a85a9
Fix late access to codelet freed by data_unregister
|
8 years ago |
Corentin Salingue
|
2809f55640
factorize some MPI MASTER-SLAVE variables
|
8 years ago |
Luka Stanisic
|
3fab90c341
doc: correcting typo for PERF_MODEL_HOMOGENEOUS
|
8 years ago |
Samuel Thibault
|
7b7f03c698
fix memleak
|
8 years ago |
Nathalie Furmento
|
90b13d61d1
mpi/src/starpu_mpi_cache.c: avoid thread deadlock
|
8 years ago |
Samuel Thibault
|
c7f166c27c
Add mpi communications in dag.dot
|
8 years ago |
Samuel Thibault
|
a3b0152013
Fix gantt chart for starpu_mpi_irecv: it should use the termination time of the request, not the submission time
|
8 years ago |
Corentin Salingue
|
fe58870642
use correct number of CPUs when checking in master-slave mode
|
8 years ago |
Samuel Thibault
|
d1bbfcfdc8
improve data fetch efficiency, by not releasing data after the asynchronous transfers have completed
|
8 years ago |
Corentin Salingue
|
bebe96a109
little optimization to reduce number of barriers when benchmarking Master-Slave
|
8 years ago |
Luka Stanisic
|
5c6a4fda73
mlr: more descriptive name for the internal function doing validation of mlr models (so potential warning is easier to understand)
|
8 years ago |
Samuel Thibault
|
194ed948af
Finish fix
|
8 years ago |
Luka Stanisic
|
5c050ae61f
this assert is no longer valid with the recent code improvements from Samuel
|
8 years ago |
Corentin Salingue
|
d0d59fb514
use mpiexec when lauching tools in master-slave mode
|
8 years ago |
Samuel Thibault
|
80daa8c977
Apparently Nathalie fixed them all :)
|
8 years ago |
Samuel Thibault
|
b4c084694d
Update line number
|
8 years ago |
Samuel Thibault
|
74ecd02370
Fix boolean vs binary test
|
8 years ago |
Samuel Thibault
|
94761f747d
Remove useless assignment
|
8 years ago |
Samuel Thibault
|
ed7967a2bf
fix memleak
|
8 years ago |
Samuel Thibault
|
a8f1d7ac04
Make sure we have assigned a master
|
8 years ago |
Samuel Thibault
|
debcbc561e
Fix terminating string
|
8 years ago |