 Nathalie Furmento
|
4d30cef9dc
mpi/src: properly deal with unknown datatypes
|
12 years ago |
 Simon Archipoff
|
1b44397d62
a lot of comments in node_sched.h
|
12 years ago |
|
Simon Archipoff
|
f43398996f
bug fix, sched_policy.c:_get_next_sched_ctx_to_pop_into made a recursive call
|
12 years ago |
|
Simon Archipoff
|
0a293ff950
fix init bug, sched_policy.c:_get_next_sched_ctx_to_pop_into() made a recursive call to wait, with no optimisation this lead to stack overflow
|
12 years ago |
|
Simon Archipoff
|
d8872d542c
dead lock fixed : a worker of a combined worker first free his mutex before waking up his mates
|
12 years ago |
|
Nathalie Furmento
|
9d993d0369
mpi/src/starpu_mpi_datatype.c: deal with unknown datatype
|
12 years ago |
|
Nathalie Furmento
|
6962ff103c
merge branches/starpu-1.1@10328: mpi/tests: add new test to do a redux by hand + modify test not to use same tag for each node + needs to be fixed as test sometimes hangs
|
12 years ago |
 Samuel Thibault
|
27a70c8cc6
document measurement unit
|
12 years ago |
|
Nathalie Furmento
|
69af6ec637
tests: new test to test insert task with codelet NULL and a callback
|
12 years ago |
|
Nathalie Furmento
|
0d8d65bebb
data management: New functions
|
12 years ago |
|
Samuel Thibault
|
12aa69aa32
It does not make sense to test for sequential_consistency flag when unlocking the post_sync_tasks: we are at execution time, not submission time. post_sync_tasks_cnt tells us whether post_sync tasks have been submitted or not anyway
|
12 years ago |
|
Samuel Thibault
|
f2fffd0544
do not add the post_sync task if there is none. Fixes acquiring a data whose consistency flag is cleared
|
12 years ago |
|
Samuel Thibault
|
912cad9cce
make sure pre/post_sync_task are set in the non-sequential-consistency mode
|
12 years ago |
|
Simon Archipoff
|
ec7ccb4070
dirty hack to have combined worker before scheduler unitialization
|
12 years ago |
|
Samuel Thibault
|
582fe879e4
comments for the LRU algorithm
|
12 years ago |
|
Samuel Thibault
|
f7c329c564
port 10304 from 1.1: Advertise trace change
|
12 years ago |
|
Simon Archipoff
|
3d50031385
first lines of documentation
|
12 years ago |
|
Samuel Thibault
|
5c6d861719
port r10284 from 1.1: Use a different color in paje traces for the overhead parts of StarPU, distinguished from the sleep parts
|
12 years ago |
|
Simon Archipoff
|
aecb30d21d
compile fix
|
12 years ago |
|
Simon Archipoff
|
53d63b44a0
merge trunk up to 10281
|
12 years ago |
|
Simon Archipoff
|
a1a31ff6b9
minor bug fixes
|
12 years ago |
|
Nathalie Furmento
|
f947d013a6
doc/doxygen: add new faq starpu mkl (thanks to Benoit Lize)
|
12 years ago |
|
Simon Archipoff
|
befbeaa725
update prediction function to new data structure
|
12 years ago |
|
Simon Archipoff
|
21b18e5965
big changes in node_worker.c
|
12 years ago |
|
Nathalie Furmento
|
7b11a5754c
doc/doxygen: typo
|
12 years ago |
|
Nathalie Furmento
|
a38e803a5e
doc/doxygen: explicitely list all .h files from directory srcdir/include (giving the directory might lead to include starpu_config.h)
|
12 years ago |
|
Samuel Thibault
|
039b1f4ff1
fix doc: STARPU_MIN_WORKERSIZE and STARPU_MAX_WORKERSIZE have effect in all cases
|
12 years ago |
|
Nathalie Furmento
|
7e91a33326
configure.ac: documentation now needs doxygen
|
12 years ago |
|
Samuel Thibault
|
f5376f57de
Fix missing unlock when a data tree could be locked, but the data couldn't be freed
|
12 years ago |
 Andra Hugo
|
f2d6fafd49
fix examples of hyp compilation + add default value for the resize triggering criteria
|
12 years ago |
