Samuel Thibault
|
b47d146631
Avoid exposing to valgrind/tsan optimizations which are thread-unsafe but scheduling-safe
|
10 anni fa |
Nathalie Furmento
|
bb61f3808d
update cnrs copyright
|
10 anni fa |
Andra Hugo
|
777eba975c
move_task_to_ctx should push directly the job & the prolog should be called only for contexts needing it + example jean-marie
|
10 anni fa |
Samuel Thibault
|
0f1149cf22
port r14531 from 1.1: Also advise to try dmdas
|
10 anni fa |
Samuel Thibault
|
113cdbb1a3
Support pipeline mode in simgrid too. The results match the reality!
|
11 anni fa |
Samuel Thibault
|
3a33757343
Add starpu_worker_can_execute_task_impl and starpu_worker_can_execute_task_first_impl to optimize getting the working implementations
|
11 anni fa |
Samuel Thibault
|
4b1a16edde
Fix University name after the fusion..
|
11 anni fa |
Samuel Thibault
|
717fbf7ab1
Always call init_iterator, it does not make sense otherwise
|
11 anni fa |
Samuel Thibault
|
4497fe7d38
Fix eager policies in blocking mode: the policy mutex mustn't be held while calling starpu_wakeup_worker, since in other parts of the code we take the mutexes the other way
|
11 anni fa |
Andra Hugo
|
ab7bd6810c
nested scheduling contexts + worker collection keeping track of blocked workers
|
11 anni fa |
Andra Hugo
|
2e4399f771
bindid fixes
|
11 anni fa |
Andra Hugo
|
ef81fe3c52
sorry: rollback again eager (don't use STARPU_WORKER_TREE yet)
|
11 anni fa |
Andra Hugo
|
491a6c6124
fix hierarchical resizing of contexts
|
11 anni fa |
Andra Hugo
|
994862e3ac
rollback eager (don't use STARPU_WORKER_TREE yet)
|
11 anni fa |
Andra Hugo
|
60d44eb737
resize hierarchic ctxs hierarchically :)
|
11 anni fa |
Andra Hugo
|
bfb858fe1d
Add hierarchical contexts
|
11 anni fa |
Andra Hugo
|
6266721e89
do not take the mutex if the current worker is calling the code
|
12 anni fa |
Samuel Thibault
|
c3d255416f
port r11361 from 1.1: Fix using fields of the just-pushed task without keeping the mutex, otherwise it might get executed and freed in between. Also make the blocking mode wake only workers which can execute the task
|
12 anni fa |
Samuel Thibault
|
8b8f4bffe4
port r11359 from 1.1: Make eager release pressure on the mutexes by using a bitmap for signaling which workers are waiting for a task.
|
12 anni fa |
Andra Hugo
|
52459c8508
*eager: wake up workers when we add them to a context
|
12 anni fa |
Samuel Thibault
|
aea35461df
Document each racy spot normally reported by helgrind
|
12 anni fa |
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 anni fa |
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 anni fa |
Andra Hugo
|
646c61a828
improve idle computations in order to better compute max_workers for the resizing lp
|
12 anni fa |
Pierre-André Wacrenier
|
5a626f84cc
do not wake workers when macro STARPU_NON_BLOCKING_DRIVERS is not set
|
12 anni fa |
Pierre-André Wacrenier
|
c97cc4c390
eager central policy : wake up a single worker when submitting a task
|
12 anni fa |
Andra Hugo
|
6e52308e70
Indicate whether a worker shares tasks lists with other workers (if yes if we remove it from the context we should do it right away, the other workers are in charge of the remaining tasks)
|
12 anni fa |
Simon Archipoff
|
8160206f31
make eager_central_policy dont respect priorities
|
12 anni fa |
Nathalie Furmento
|
786b8fb4b7
move _STARPU_PTHREAD_XXX macros to public API starpu_thread_util and rename them to STARPU_PTHREAD_XXX
|
12 anni fa |
Nathalie Furmento
|
9e75014ab9
src/sched_policies: remove useless include files
|
12 anni fa |