 Cédric Augonnet
|
61cdbb41bd
Add support for parallel tasks:
|
il y a 14 ans |
 Nathalie Furmento
|
db63f0fb70
doc: minor fix
|
il y a 14 ans |
|
Cédric Augonnet
|
5afb322c21
When creating per-worker replicates, we don't create more replicates than there are workers
|
il y a 14 ans |
|
Cédric Augonnet
|
3096ff03f4
The _starpu_worker_get_relative_speedup function now takes a starpu_perf_arch
|
il y a 14 ans |
|
Nathalie Furmento
|
2bb283fb4f
OpenCL: use 'opencl kernels' instead of 'opencl codelet' to avoid confusion with starpu codelet object
|
il y a 14 ans |
|
Nathalie Furmento
|
26d55a4e58
OpenCL: use 'opencl kernels' instead of 'opencl codelet' to avoid confusion with starpu codelet object
|
il y a 14 ans |
|
Nathalie Furmento
|
38808dce37
OpenCL: use 'opencl kernels' instead of 'opencl codelet' to avoid confusion with starpu codelet object
|
il y a 14 ans |
|
Nathalie Furmento
|
cafb22e85f
Makefile: make -C tests check runs coverage/coverage.sh only when coverage is enabled
|
il y a 14 ans |
|
Cédric Augonnet
|
babde97945
Store tasks instead of jobs in the per-worker lists.
|
il y a 14 ans |
 Samuel Thibault
|
b841a828c3
make it really clear about the value at entry of the function
|
il y a 14 ans |
|
Samuel Thibault
|
eb001abd5b
(non-)document relation between scratch and the rest
|
il y a 14 ans |
|
Samuel Thibault
|
cafe57bf6c
explicit a bit more about SCRATCH
|
il y a 14 ans |
|
Nathalie Furmento
|
c099c48e8f
autogen.sh: specify shell to execute the script (this is needed by buildbot)
|
il y a 14 ans |
|
Cédric Augonnet
|
e06529d93f
Implement tree-based reductions
|
il y a 14 ans |
|
Cédric Augonnet
|
c384b4f4e7
When starting a kernel while there remains pending data requests, the measured
|
il y a 14 ans |
|
Cédric Augonnet
|
e972e93da4
Fix make distcheck
|
il y a 14 ans |
|
Cédric Augonnet
|
56b6ce49ca
Remove the examples/coverage directory because its content was moved into tests/coverage/
|
il y a 14 ans |
|
Cédric Augonnet
|
77e1ff6d24
The examples are not in the current directory anymore
|
il y a 14 ans |
|
Cédric Augonnet
|
5a6e53c653
Remove superfluous cudaThreadSynchronize() calls.
|
il y a 14 ans |
|
Samuel Thibault
|
b3cae45f6c
make sure libtool is installed
|
il y a 14 ans |
|
Cédric Augonnet
|
975cfaa576
Move coverage.sh test into the tests/ directory as this is not something that
|
il y a 14 ans |
|
Cédric Augonnet
|
d016ed967c
Move the dot_product example into the examples/reductions/ directory
|
il y a 14 ans |
|
Cédric Augonnet
|
6740d0e913
Minor cleanups
|
il y a 14 ans |
|
Cédric Augonnet
|
bb17b419fe
The example may only execute on CPU workers
|
il y a 14 ans |
|
Cédric Augonnet
|
d24f3f9b7b
Use reductions to compute the min and the max entry of an array
|
il y a 14 ans |
|
Nathalie Furmento
|
f38aec272a
drivers/opencl: using fread does not work on all OSes (namely Windows ...)
|
il y a 14 ans |
|
Nathalie Furmento
|
fe5a0dc0fd
core/perfmodel/perfmodel_bus.c: cast so that the comparison is correct
|
il y a 14 ans |
|
Nathalie Furmento
|
8e358b9e9c
driver/opencl: typo
|
il y a 14 ans |
|
Nathalie Furmento
|
61c061f527
drivers/opencl: also detect accelerator devices (as defined by Movidius implementation)
|
il y a 14 ans |
|
Nathalie Furmento
|
e976dd317e
drivers/opencl: improve error checking
|
il y a 14 ans |
