 Sylvain Henry
|
3607f8b341
Use a specific queue for allocations to avoid barriers in other queues
|
12 years ago |
|
Sylvain Henry
|
07b7432fc7
SOCL: fix StarPU's configuration (must not be a stack var)
|
12 years ago |
 Samuel Thibault
|
029890da24
really fix build
|
12 years ago |
|
Samuel Thibault
|
ccfcf93d72
fix build
|
12 years ago |
|
Samuel Thibault
|
46f76d7629
Fix crash on no user conf
|
12 years ago |
 Andra Hugo
|
676eb70644
fix last commit
|
12 years ago |
|
Andra Hugo
|
d6a15c7e7c
* create a public function starpu_start_fxt_profiling
|
12 years ago |
|
Samuel Thibault
|
1997bc7549
Take GPU transfers sequentialization into account in simgrid
|
12 years ago |
|
Samuel Thibault
|
e0313a071e
Rework simgrid transfers to always use a transfer structure
|
12 years ago |
|
Andra Hugo
|
7eddde2a77
factorize init of worker_collection
|
12 years ago |
|
Samuel Thibault
|
e3830cb1b7
refactor simgrid transfer submission, pass node numbers
|
12 years ago |
|
Samuel Thibault
|
b70256d333
drop spurious visibility
|
12 years ago |
|
Samuel Thibault
|
e373fad5cb
Move simgrid code for transfers into simgrid.c. No code change
|
12 years ago |
|
Samuel Thibault
|
bcc603c16e
document the simgrid environment variables
|
12 years ago |
|
Samuel Thibault
|
9a152a018e
Warn if hwloc didn't report core objects
|
12 years ago |
|
Sylvain Henry
|
db97716c0f
SOCL: fix dependency
|
13 years ago |
 Nathalie Furmento
|
13f3b8fb8f
ChangeLog: update w.r.t r8374
|
13 years ago |
|
Sylvain Henry
|
d1301e0444
SOCL: display command durations in basic example
|
13 years ago |
|
Sylvain Henry
|
003007bb8f
SOCL: fix warnings
|
13 years ago |
|
Sylvain Henry
|
1f289d074f
SOCL: "workaround" for performance models
|
13 years ago |
|
Sylvain Henry
|
75ae9dba3b
SOCL: enhanced entity management
|
13 years ago |
|
Sylvain Henry
|
28074351ce
SOCL: add "mapping" example
|
13 years ago |
|
Sylvain Henry
|
f005e6f1ff
SOCL: fix buffer release
|
13 years ago |
|
Sylvain Henry
|
42bfecf7f9
SOCL: minor change in basic example
|
13 years ago |
|
Samuel Thibault
|
7da73d5bf4
comment about scheduler overhead
|
13 years ago |
|
Samuel Thibault
|
427214d42e
Do not enable gpu-gpu direct transfer, it seems to just slow down cuda ATM...
|
13 years ago |
|
Samuel Thibault
|
86cd122211
Comment that in GPUDirect mode, peer transfer is probably better when issued from the destination too
|
13 years ago |
|
Samuel Thibault
|
27960c28bc
Use a separate stream for gpu in/out, and peer, as some cards can actually do some transfers in parallel
|
13 years ago |
|
Samuel Thibault
|
159317c1dc
Make starpu_interface_matrix_ops public
|
13 years ago |
|
Samuel Thibault
|
365ccd855e
fix compilation warning
|
13 years ago |
