 Terry Cojean
|
b9cda4509e
Improve my solution to fix work stealing with contexts
|
9 years ago |
|
Terry Cojean
|
7f9c179b81
Fix work stealing policy with contexts
|
9 years ago |
 Samuel Thibault
|
af3ffe49df
lws actually *requires* trees when built with hwloc
|
9 years ago |
|
Samuel Thibault
|
2d064f8300
do not use tree worker iterators by default, they are too expensive
|
9 years ago |
|
Samuel Thibault
|
3c81d54382
Reduce a big overhead in starpu_pthread_mutex_lock/unlock by only checking for the sched mutex when locking a sched mutex
|
9 years ago |
 Samuel Pitoiset
|
f03fbf6557
sched_policies: refactor ws and lws scheduling policies
|
9 years ago |
|
Samuel Pitoiset
|
1e5664626d
sched_policies: add a select_victim() helper for ws sched
|
9 years ago |
|
Samuel Pitoiset
|
1d6d838d5a
sched_policies: add a comment related to USE_OVERLOAD experimental code
|
9 years ago |
|
Samuel Pitoiset
|
fa027e9388
src/sched_policies: do not lock all workers when selecting a victim with ws
|
9 years ago |
|
Samuel Pitoiset
|
d4541060ed
src/sched_policies: remove useless initialization for work stealing policies
|
9 years ago |
|
Samuel Pitoiset
|
64ed3c52ca
src/sched_policies: remove useless call to starpu_worker_get_sched_condition()
|
9 years ago |
|
Samuel Pitoiset
|
899702825f
sched_policies: use _starpu_fifo_taskq instead of _starpu_deque_jobq for ws
|
9 years ago |
|
Samuel Pitoiset
|
5b5aec21b3
sched_policies: do not allocate too much memory for ws sched
|
9 years ago |
|
Samuel Pitoiset
|
0edd73e32f
sched_policies: remove unused fields in _starpu_work_stealing_data
|
9 years ago |
|
Samuel Pitoiset
|
f29efa4f4d
sched_policies: create/remove worker collections outside scheduler policies
|
9 years ago |
|
Samuel Thibault
|
c277c9b080
Add starpu_wake_worker_locked and starpu_wakeup_worker_locked. Use starpu_wake{,up}_worker{,_locked} in all code instead of explicitly waking the scheduling condition
|
9 years ago |
|
Samuel Thibault
|
f9948fbc3f
Fix restarting iterator for the unlock
|
9 years ago |
|
Samuel Thibault
|
c8efbb83e8
Inline list heads instead of keeping allocating them
|
10 years ago |
 Nathalie Furmento
|
bb61f3808d
update cnrs copyright
|
10 years ago |
|
Samuel Thibault
|
4b1a16edde
Fix University name after the fusion..
|
10 years ago |
|
Samuel Thibault
|
717fbf7ab1
Always call init_iterator, it does not make sense otherwise
|
10 years ago |
|
Samuel Thibault
|
b25760aa94
fix warnings
|
11 years ago |
|
Nathalie Furmento
|
9bb5a09263
src: minor fixes
|
11 years ago |
|
Samuel Thibault
|
19ed150528
port r11354 and 11355 from 1.1: There is no need to signal workers in non-blocking mode
|
11 years ago |
|
Samuel Thibault
|
e1501b5592
port r11308 from 1.1: record allocation size in trace
|
11 years ago |
|
Samuel Thibault
|
aea35461df
Document each racy spot normally reported by helgrind
|
12 years ago |
|
Samuel Thibault
|
0d1b0375b3
backport r11155 from 1.1: Disabling access to a variable locally does not actually seem to prevent helgrind from raising the race issues, on the contrary it prevents it from recording where the race comes from... Simply just disable checking these variables completely.
|
12 years ago |
|
Samuel Thibault
|
54460ab919
port r11138 from 1.1: Do not lie to helgrind about having taken a mutex, it produces erroneous races inside mutexes. Rather make helgrind ignore the precise memory accesses we know we do not need to care about
|
12 years ago |
|
Samuel Thibault
|
cf2fcb6937
update to temanejo 1.0rc2
|
12 years ago |
 Andra Hugo
|
dc4e81b588
indicate workers sharing tasks lists for all the concerned schedulers
|
12 years ago |
