 Samuel Thibault
|
f3d488b260
Avoid including cublas.h to avoid conflicting with cublas_v2.h
|
11 years ago |
|
Samuel Thibault
|
48eed7f291
also print the codelet name
|
11 years ago |
|
Samuel Thibault
|
7e8b37842d
print the error string, not only the value
|
11 years ago |
|
Samuel Thibault
|
907cc16e0a
be more specific about the task and codelet which failed to get submitted
|
11 years ago |
|
Samuel Thibault
|
2e3c01cb38
Always make sure that kernels without the async flag are waiting for kernel termination
|
11 years ago |
 Olivier Aumage
|
793eba8411
- modify the way to unregister OpenMP tasks' data
|
11 years ago |
|
Samuel Thibault
|
791ad827f7
merge trunk
|
11 years ago |
|
Samuel Thibault
|
113cdbb1a3
Support pipeline mode in simgrid too. The results match the reality!
|
11 years ago |
 Nathalie Furmento
|
cc9aecb792
merge trunk
|
11 years ago |
|
Samuel Thibault
|
4b1a16edde
Fix University name after the fusion..
|
11 years ago |
|
Nathalie Furmento
|
836c1e148c
merge trunk
|
11 years ago |
|
Nathalie Furmento
|
bbed82f445
merge trunk
|
11 years ago |
|
Nathalie Furmento
|
67951c8857
merge trunk
|
11 years ago |
|
Samuel Thibault
|
c3e6095f8d
Move the pause implementation into its own function, instead of _starpu_machine_is_running which is called in various places where the pause can actually lead to deadlocks
|
11 years ago |
|
Olivier Aumage
|
f4e9763817
- workaround allowing starpu_omp_shutdown to complete when called from a destructor, after CUDA as been used
|
11 years ago |
|
Nathalie Furmento
|
5481b186a6
merge trunk
|
11 years ago |
|
Samuel Thibault
|
b7b1a52587
truncat STARPU_*_PIPELINE to STARPU_MAX_PIPELINE
|
11 years ago |
|
Nathalie Furmento
|
406e57e284
merge trunk
|
11 years ago |
|
Nathalie Furmento
|
cac7d80ba8
src/drivers/cuda/driver_cuda.c: fix typo
|
11 years ago |
|
Samuel Thibault
|
3ed3e95a2a
Do not check for stream synchronization when we do it ourselves
|
11 years ago |
|
Nathalie Furmento
|
1ae61a95f9
CUDA and OpenCL drivers: when STARPU_CUDA_PIPELINE and STARPU_OPENCL_PIPELINE are set to 0, force synchronous execution of all kernels
|
11 years ago |
|
Nathalie Furmento
|
75c5c178ae
merge trunk
|
11 years ago |
 Andra Hugo
|
af26be4c58
fix mem corruptions
|
11 years ago |
|
Samuel Thibault
|
7ac2c135ad
protect worker status with the scheduler mutex
|
11 years ago |
|
Nathalie Furmento
|
f885c06876
merge trunk
|
11 years ago |
|
Samuel Thibault
|
b7efebe04d
Fix task accounting when the scheduler erroneously puts a non-cuda task to a cuda device
|
11 years ago |
|
Samuel Thibault
|
23636fb4f4
Fix simgrid build
|
11 years ago |
|
Samuel Thibault
|
77e107e4af
Warn if the user uses concurrent execution on a device which doesn't support it
|
11 years ago |
|
Samuel Thibault
|
62194c5c86
Fix filling the pipeline
|
11 years ago |
|
Samuel Thibault
|
f99ef13c88
Add CUDA kernel submission pipelining, to overlap costs and allow concurrent
|
11 years ago |
