 Nathalie Furmento
|
7c1cf3c779
src/core/workers.c: simplify the waiting for worker initialisation by using the field run_by_starpu instead of re-calling starpu_may_launch_driver()
|
8 anni fa |
|
Nathalie Furmento
|
58fda940f8
examples/reductions/dot_product.c: information about error
|
8 anni fa |
 Samuel Thibault
|
0452a05f97
Warn about missing concurrent kernel execution support even when running multistream workers in separate threads
|
8 anni fa |
|
Samuel Thibault
|
fc69bb54c7
Use a mutex to protect cublas initialization and shutdown, it seems that it is actually not threadsafe for the same device
|
8 anni fa |
|
Samuel Thibault
|
f44b1a5aed
Avoid variable name conflict with older versions of fxt
|
8 anni fa |
|
Samuel Thibault
|
3a366c3041
typo
|
8 anni fa |
|
Samuel Thibault
|
0bd7fce015
Add missing unlock
|
8 anni fa |
|
Samuel Thibault
|
d43771c32f
Fix build with simgrid
|
8 anni fa |
|
Nathalie Furmento
|
3056cff4a8
socl: unset environment variables STARPU_NCPU and STARPU_NCUDA, it might be better to have a variable in starpu_conf to indicate to overwrite environment variables
|
8 anni fa |
|
Nathalie Furmento
|
009872de63
tests/main/driver_api/init_run_deinit.c: remove spurious line
|
8 anni fa |
|
Nathalie Furmento
|
9efb3060ae
tests/disk/disk_pack.c: unset environment variables
|
8 anni fa |
|
Nathalie Furmento
|
41aa87e4b2
src/datawizard/coherency.c: fix code when fxt is disabled
|
8 anni fa |
|
Nathalie Furmento
|
e30878eb85
src/common/fxt.h: avoid unused variable warnings when fxt is disabled
|
8 anni fa |
|
Nathalie Furmento
|
367ae7dfa2
src/core/workers.h: turn cuda_th_per_stream and cuda_th_per_dev int
|
8 anni fa |
|
Nathalie Furmento
|
9d8907e6d3
mpi: fix return type, and remove now un-needed function parameter
|
8 anni fa |
|
Nathalie Furmento
|
857811a387
tests/main/driver_api: unset environment variables as we want to force the number of workers
|
8 anni fa |
|
Nathalie Furmento
|
e614701bff
doc/doxygen/chapters/api/workers.doxy: typo
|
8 anni fa |
|
Nathalie Furmento
|
3629e1e317
tests/main/driver_api: fix number of workers
|
8 anni fa |
|
Nathalie Furmento
|
bf074a7831
src: include core/task.h when needed
|
8 anni fa |
|
Samuel Thibault
|
bfcac01ad1
Avoid emitting progress probes repeatedly, allowing to re-enable them in the trace
|
8 anni fa |
|
Samuel Thibault
|
9d562f013b
Fix handling memnode state: when several transfers are queued, states should be pushed and poped, otherwise on termination of the first transfer the trace shows the memnode as idle
|
8 anni fa |
|
Samuel Thibault
|
82cc951ab9
Smooth gflop computations a bit
|
8 anni fa |
|
Samuel Thibault
|
a2e38ce59c
Set the device ID when running an initialization codelet
|
8 anni fa |
|
Samuel Thibault
|
ecac7ec55c
Also always set the current CUDA devid when freeing, it is not very costly and makes things more flexible
|
8 anni fa |
|
Samuel Thibault
|
922ff28a1e
Make datawizard_progress use worker 0 of worker sets as reference for determining which memory nodes to make progress
|
8 anni fa |
|
Samuel Thibault
|
384397546c
Label the CUDA driver just 'CUDA' one using just one thread
|
8 anni fa |
|
Samuel Thibault
|
283d125846
fix build without CUDA
|
8 anni fa |
|
Samuel Thibault
|
7cdeadd447
Add STARPU_CUDA_THREAD_PER_DEV environment variable to support driving all
|
8 anni fa |
|
Samuel Thibault
|
2ea92eb0fb
Always set the current CUDA devid when allocating, it is not very costly and makes things more flexible
|
8 anni fa |
|
Samuel Thibault
|
43a3c904ed
directly lookup destination node instead of relying on the node of the current worker
|
8 anni fa |
