 Samuel Thibault
|
5a5eb593d1
Always reset the local worker key to worker0 for datawizard to get the proper wait queue when using multiple workers in the same driver thread
|
8 anni fa |
|
Samuel Thibault
|
a41080d2a6
Fix setting the proper CUDA stream when using several streams from the same driver thread
|
8 anni fa |
|
Samuel Thibault
|
52d97c6ae9
Fix cublas initialization/shutdown when using one thread per stream with multistream
|
8 anni fa |
|
Samuel Thibault
|
27cfafb119
Move setting worker_set->nworkers to topology, where it makes more sense
|
8 anni fa |
|
Samuel Thibault
|
6dbd5add56
Fix indent
|
8 anni fa |
|
Samuel Thibault
|
d3d32d6451
Fix priorities for schedulers which don't have unbound priority values
|
8 anni fa |
|
Samuel Thibault
|
ace9f6b23e
Use better priorities in cholesky algorithm
|
8 anni fa |
|
Samuel Thibault
|
5278ab7d1b
typo
|
8 anni fa |
|
Samuel Thibault
|
14492fb09a
typo
|
8 anni fa |
 Nathalie Furmento
|
461ca6a408
mpi/src/starpu_mpi_cache_stats.c: remove internal array which is not needed
|
8 anni fa |
|
Nathalie Furmento
|
d8961fb495
mpi: move cache data in the starpu_data_handle_t
|
8 anni fa |
|
Samuel Thibault
|
6a9541027d
comments
|
8 anni fa |
|
Samuel Thibault
|
5e61b183a5
Fix reporting CUDA workers when using several workers per CUDA
|
8 anni fa |
|
Samuel Thibault
|
8a713e9521
Fix missing include, thanks Léo
|
8 anni fa |
|
Nathalie Furmento
|
4ee7c64768
mpi/tests/insert_task_node_choice.c: use long long to make sure test is valid on a 32-bit architecture
|
8 anni fa |
|
Samuel Thibault
|
cd847ea3d3
Fix build with upcoming simgrid release
|
8 anni fa |
|
Samuel Thibault
|
e71124580f
shut up warning
|
8 anni fa |
|
Samuel Thibault
|
c16ae6ed15
Fix warning with simgrid
|
8 anni fa |
|
Samuel Thibault
|
e219b37a8d
Fix warnings with older simgrid
|
8 anni fa |
|
Samuel Thibault
|
cc7c39c4f7
Fix build with upcoming simgrid 3.15
|
8 anni fa |
|
Samuel Thibault
|
b24f685b47
When trying to reuse a memchunk, avoid locking the handle when we easily know that a transfer is needed and thus we will not afford doing it
|
8 anni fa |
|
Samuel Thibault
|
6734d1fd95
Fix initial value for number of cuda devices
|
8 anni fa |
|
Samuel Thibault
|
3bb7af49d1
Fix build without allocation cache
|
8 anni fa |
|
Samuel Thibault
|
afcbc106ef
Fix build without allocation cache
|
8 anni fa |
|
Samuel Thibault
|
03c9afe025
Do not set ncudagpus each time _starpu_init_cuda is called, it is useless
|
8 anni fa |
|
Nathalie Furmento
|
e2a6f6ca5a
tools/starpu_machine_display.c: also display SCC and MPI master slaves nodes
|
8 anni fa |
|
Samuel Thibault
|
1879558d5f
Drop now-useless case
|
8 anni fa |
|
Samuel Thibault
|
2f6fe20825
Only reuse memchunks which are easy to throw away
|
8 anni fa |
|
Samuel Thibault
|
72453f9f98
Disable forcing available memory by default, since this makes StarPU spuriously free data when prefetching fills the memory. Clean buffers are enough to be able to allocate data easily anyway.
|
8 anni fa |
|
Samuel Thibault
|
33399e8f5a
Improve data reuse: before calling generic reclaiming, try to reuse potentially-in-use mc of the same interface. Avoids having generic reclaiming free data just to be able to reallocate exactly the same kind of data
|
8 anni fa |
