Samuel Thibault
|
a3f27f08ad
Make some warnings clearer. Avoid a very odd warning when the codelet has more parameters than allowed
|
12 years ago |
Nathalie Furmento
|
786b8fb4b7
move _STARPU_PTHREAD_XXX macros to public API starpu_thread_util and rename them to STARPU_PTHREAD_XXX
|
12 years ago |
Nathalie Furmento
|
097a832e3c
Allow to have a dynamically allocated number of buffers per task, and
|
12 years ago |
Nathalie Furmento
|
2cb9fcd148
use starpu_ssize_t instead of ssize_t to be windows-able
|
12 years ago |
Samuel Thibault
|
021aef6b24
fix simgrid build
|
12 years ago |
Samuel Thibault
|
d71b88efa7
Fix showing GPU size in starpu_machine_display
|
12 years ago |
Samuel Thibault
|
78e6c17e27
Fix setting memory limit above 2GiB
|
12 years ago |
Nathalie Furmento
|
44ef2f971d
fxt: fix STARPU_TRACES macros when fxt is not enabled
|
12 years ago |
Nathalie Furmento
|
eb50762352
drivers: fix type for limit variable
|
12 years ago |
Nathalie Furmento
|
294e0e4e26
src/drivers/cuda/driver_cuda.c: fix code which is or is not enabled when cuda is or is not enabled
|
12 years ago |
Samuel Thibault
|
8e4534513a
Add support for limiting gpu memory in simgrid
|
12 years ago |
Nathalie Furmento
|
6b2e237ad5
src/drivers: functions _starpu_cuda_get_global_mem_size and _starpu_opencl_get_global_mem_size always need to be defined, they return 0 when the driver is disabled
|
12 years ago |
Nathalie Furmento
|
d0591ff302
memory manager: simplify management of global memory size
|
12 years ago |
Samuel Thibault
|
e0167e89bc
fix build with simgrid without cuda
|
12 years ago |
Samuel Thibault
|
bc1a386d12
Make drivers push back a task through _starpu_push_task_to_workers instead of _starpu_push_task, so such tasks are not accounted twice for the nready counter
|
12 years ago |
Nathalie Furmento
|
52a7ff482a
src/drivers/cuda/driver_cuda.c: function _starpu_cuda_limit_gpu_mem_if_needed also needs to be defined when simgrid is enabled
|
12 years ago |
Nathalie Furmento
|
e327177d94
Introduce new variables STARPU_LIMIT_CUDA_devid_MEM and
|
12 years ago |
Nathalie Furmento
|
5a5e86e8cd
Discard environment variable STARPU_LIMIT_GPU_MEM and define instead
|
12 years ago |
Nathalie Furmento
|
f2b45203d6
src/drivers/cuda/driver_cuda.c: reimplement limit_gpu_if_needed, we no
|
12 years ago |
Nathalie Furmento
|
842b7b131a
src: check if there is enough memory available on a node before allocation
|
12 years ago |
Nathalie Furmento
|
b770541f52
src: memory_nodes function renaming
|
12 years ago |
Samuel Thibault
|
2b30d49db1
Make it possible to use simgrid without having CUDA/OpenCL installed
|
12 years ago |
Nathalie Furmento
|
eb32a9b3d4
src: fix coding style
|
12 years ago |
Samuel Thibault
|
9d2589a240
Add starpu_worker_get_by_type and starpu_worker_get_by_devid
|
12 years ago |
Samuel Thibault
|
427214d42e
Do not enable gpu-gpu direct transfer, it seems to just slow down cuda ATM...
|
12 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
|
12 years ago |
Samuel Thibault
|
8510692204
Simply use devid everywhere instead of args->devid
|
13 years ago |
Nathalie Furmento
|
f2acd4d240
src/drivers: use args->devid as devid may not be defined
|
13 years ago |
Samuel Thibault
|
4492f76e0b
Add --enable-opencl-simulator option to combine PePU with simgrid
|
13 years ago |
Samuel Thibault
|
f60eb1288d
Add initial simgrid support
|
13 years ago |