 Terry Cojean
|
818ff26066
Fix locking scheme for nworkers_able_to_execute_tasks. It should be a rdlock since the context isn't modified
|
il y a 9 ans |
 Olivier Aumage
|
32974f7b85
since there is only one mutex left for sched_ctx, give it a shorter name
|
il y a 9 ans |
|
Olivier Aumage
|
849f26ab28
- drop unused fields
|
il y a 9 ans |
|
Olivier Aumage
|
636e653c52
drop waiting_tasks_mutex in favor of existing changing_ctx_mutex
|
il y a 9 ans |
|
Olivier Aumage
|
7010fe7dce
drop empty_ctx_mutex in favor of existing changing_ctx_mutex
|
il y a 9 ans |
|
Olivier Aumage
|
301a948922
drop sched_ctx_list_mutex in favor of existing changing_ctx_mutex
|
il y a 9 ans |
 Samuel Thibault
|
e61dc2aa4d
Add callback alias for user thread state
|
il y a 9 ans |
|
Olivier Aumage
|
7463ed1261
streamline locking in fetch_tasks_from_empty_ctx_list
|
il y a 9 ans |
|
Olivier Aumage
|
145857c90e
rework locking scheme in sched_ctx_delete
|
il y a 9 ans |
|
Samuel Thibault
|
59b607325d
Add -no-events option to starpu_fxt_tool
|
il y a 9 ans |
|
Samuel Thibault
|
65b7113191
Add -no-flops option to starpu_fxt_tool to reduce trace size
|
il y a 9 ans |
|
Samuel Thibault
|
4123268114
Add missing part of commit
|
il y a 9 ans |
|
Olivier Aumage
|
ce7a2c646e
set_priority already done in add_workers
|
il y a 9 ans |
|
Olivier Aumage
|
77e0637657
protect add_workers/remove_workers
|
il y a 9 ans |
|
Olivier Aumage
|
2207051181
allow workers to recurse on their own sched_mutex
|
il y a 9 ans |
|
Samuel Thibault
|
332eaad4b0
Add function to get job_id of a task for schedulers
|
il y a 9 ans |
|
Samuel Thibault
|
615406b525
avoid starpu internals use in scheduler
|
il y a 9 ans |
|
Olivier Aumage
|
fe1082e42b
add wrapper
|
il y a 9 ans |
|
Olivier Aumage
|
e47d845df0
encompass a larger scope for transient pop sched op protection
|
il y a 9 ans |
|
Olivier Aumage
|
5efd860ef0
fix race condition between starpu_sched_ctx_delete and pop operations
|
il y a 9 ans |
|
Samuel Thibault
|
0645ff44a6
simgrid: Add GPU synchronization point on cudaFree, like CUDA does
|
il y a 9 ans |
 Corentin Salingue
|
4c98627776
fix compute bandwidth of MPI Master-Slave
|
il y a 9 ans |
|
Samuel Thibault
|
2374c1ce19
No need to initialize a whole modular heft prio scheduler when there is only one memory node
|
il y a 9 ans |
 Nathalie Furmento
|
52ba443789
mpi: update tests and examples to call MPI_Init_Thread
|
il y a 9 ans |
|
Nathalie Furmento
|
aa53063ac8
doc: explain how to initialize mpi
|
il y a 9 ans |
|
Samuel Thibault
|
ffe16a3332
Fix confusion in simgrid API evolution
|
il y a 9 ans |
|
Samuel Thibault
|
c2f600c387
complete tested versions of simgrid
|
il y a 9 ans |
|
Samuel Thibault
|
83465cc907
Fix build with simgrid 3.13
|
il y a 9 ans |
|
Corentin Salingue
|
8bba78df50
skip microbenchs because mpirun has strange behaviors with bash scripts
|
il y a 9 ans |
|
Corentin Salingue
|
b6f2af3387
remove useless timeout
|
il y a 9 ans |
