| .. |
|
dependencies
|
f43bf86292
Fix complexity of implicit task/data dependency, from quadratic to linear
|
лет назад: 12 |
|
disk_ops
|
909d76588e
Make config.h check for unistd.h inclusion before leveldb includes it
|
лет назад: 12 |
|
perfmodel
|
53f712b7e9
perf_model: make sure combination is always created (before it was created only when the perfmodel was dumped on file)
|
лет назад: 12 |
|
combined_workers.c
|
f885c06876
merge trunk
|
лет назад: 12 |
|
combined_workers.h
|
0d9f08ae8d
Fix copyrights, thanks Ludo!
|
лет назад: 15 |
|
debug.c
|
cf2fcb6937
update to temanejo 1.0rc2
|
лет назад: 13 |
|
debug.h
|
1988972e97
port r11383 from 1.1: Add a watchdog which permits to easily trigger a crash when StarPU gets stuck.
|
лет назад: 12 |
|
detect_combined_workers.c
|
5593439277
changing perf_model structure: arch now contains several devices, so we can have for eg one device STARPU_CPU with 2 cores one device STARPU_CUDA_WORKER with 1 core
|
лет назад: 12 |
|
detect_combined_workers.h
|
f9bdb5ae49
src: move sched_policies/detect_combined_workers.[ch] to core/detect_combined_workers.[ch]
|
лет назад: 13 |
|
disk.c
|
fcae2dca11
fix assert bug + reduce warning
|
лет назад: 13 |
|
disk.h
|
d5d7e412c3
merge trunk
|
лет назад: 13 |
|
errorcheck.c
|
0bd2fe9020
src: rename internal types following coding conventions
|
лет назад: 14 |
|
errorcheck.h
|
30d5f814d3
Show 'scheduling' state in paje trace
|
лет назад: 12 |
|
jobs.c
|
ec97de8ce5
merge trunk
|
лет назад: 12 |
|
jobs.h
|
a92760f1c6
Make sure we don't unconditionnally include unistd.h, which is not available on windows
|
лет назад: 12 |
|
parallel_task.c
|
7c03c945f7
Add support for letting a schedule choose its own set of workers without having to care about combined workers. This only works for SPMD however, not FORKJOIN
|
лет назад: 12 |
|
progress_hook.c
|
21defdfaa0
port r11390 from 1.1: Do not bother taking the rwlock for testing active_hook_cnt, this is already atomic
|
лет назад: 12 |
|
progress_hook.h
|
f3bbbf58e8
src: move definition of function _starpu_execute_registered_progression_hooks() from workers.h to progress_hook.h
|
лет назад: 13 |
|
sched_ctx.c
|
af26be4c58
fix mem corruptions
|
лет назад: 12 |
|
sched_ctx.h
|
afade57efa
allow ctx to execute parallel codes by reusing starpu workers (keep the workers alive instead of getting them to sleep) -> TODO try to factorize with parallel tasks on combined workers (doing practically the same, duplicate tasks and barrier_wait before & after exec)
|
лет назад: 12 |
|
sched_ctx_list.c
|
e28f7852a3
hypervisor: steal work from other ctxs when the worker is idle
|
лет назад: 12 |
|
sched_ctx_list.h
|
e28f7852a3
hypervisor: steal work from other ctxs when the worker is idle
|
лет назад: 12 |
|
sched_policy.c
|
46367259cb
refatctoring to use j->task_size = nworkers of ctx
|
лет назад: 12 |
|
sched_policy.h
|
3264633590
Add locality work stealing from Alfredo Buttari
|
лет назад: 12 |
|
simgrid.c
|
a92760f1c6
Make sure we don't unconditionnally include unistd.h, which is not available on windows
|
лет назад: 12 |
|
simgrid.h
|
7a82eb9b6b
Add initial simgrid support for MPI, not working yet
|
лет назад: 12 |
|
task.c
|
f885c06876
merge trunk
|
лет назад: 12 |
|
task.h
|
9bb7bdb07f
Increment/decrement submitted/ready tasks per context (instead of doing it per context & globally)
|
лет назад: 12 |
|
task_bundle.c
|
30860cac02
Renaming of diverse functions and datatypes.
|
лет назад: 13 |
|
task_bundle.h
|
30860cac02
Renaming of diverse functions and datatypes.
|
лет назад: 13 |
|
topology.c
|
f885c06876
merge trunk
|
лет назад: 12 |
|
topology.h
|
8c75841bf9
consider the hyperthreading when constructing the list of workers
|
лет назад: 12 |
|
tree.c
|
dcc987b0b7
Browse neighbours from ourselves to preserve locality
|
лет назад: 12 |
|
workers.c
|
5481b186a6
merge trunk
|
лет назад: 12 |
|
workers.h
|
529e9bf200
Improve static scheduling by adding support for specifying the task execution order.
|
лет назад: 12 |
Nathalie Furmento
53f712b7e9
perf_model: make sure combination is always created (before it was created only when the perfmodel was dumped on file)