 Samuel Thibault
|
b08d01d0e2
cublasSetKernelStream was added in CUDA 3.1 only
|
14 年 前 |
|
Samuel Thibault
|
04bd4040d2
Better expose the needed host_defines.h change
|
14 年 前 |
 Nathalie Furmento
|
cc375f6e35
Driver CUDA: revert latest modifications. This feature needs to be re-implemented properly.
|
14 年 前 |
|
Nathalie Furmento
|
5b909055db
CUDA driver: do not abort when number of CUDA devices required is more than what is available. Just print warning (this needs to be improved)
|
14 年 前 |
|
Nathalie Furmento
|
7560e696db
src/common/utils.h: add a flush to _STARPU_DEBUG
|
14 年 前 |
|
Samuel Thibault
|
7ff23acdf9
drop unrelated changes
|
14 年 前 |
|
Samuel Thibault
|
2668a39263
Fix gotoblas detection
|
14 年 前 |
|
Samuel Thibault
|
c9d232f0b7
add path for basic examples before installation
|
14 年 前 |
|
Nathalie Furmento
|
47fa9111ba
core/workers.h: revert wrong commit
|
14 年 前 |
|
Nathalie Furmento
|
9347d5f45d
Driver CUDA: fails when number of CUDA devices required is higher than what is available
|
14 年 前 |
|
Nathalie Furmento
|
be21dbf0a3
CUDA driver: allow configure option --enable-maxcudadev to specify a number lower than the number of available CUDA devices. Warn users that not all CUDA devices are enabled
|
14 年 前 |
|
Nathalie Furmento
|
133839b017
tests/coverage/coverage.sh: remove un-needed test
|
14 年 前 |
|
Samuel Thibault
|
fae33ed81b
hint to change the model symbol name
|
14 年 前 |
|
Samuel Thibault
|
9aa661f1ba
Actually achieve parallelism instead of performing the computation 1024 times.
|
14 年 前 |
|
Samuel Thibault
|
00c5b0737b
cope with more accelerators than cpus
|
14 年 前 |
|
Samuel Thibault
|
7ecda6ea8f
Do not flush back output into main memory when not showing it.
|
14 年 前 |
|
Samuel Thibault
|
51cd5f410b
Add -demozoom parameter to mandelbrot example, to easily select the demo zoom speed
|
14 年 前 |
|
Samuel Thibault
|
97e93682e2
fix typo
|
14 年 前 |
|
Samuel Thibault
|
c74b17beff
Show the idle consumption in the energy consumption summary
|
14 年 前 |
|
Samuel Thibault
|
595249abc7
starpufft needs fftwf, not fftwl
|
14 年 前 |
|
Samuel Thibault
|
f5ea44c6a0
Add missing profiling collect.
|
14 年 前 |
|
Nathalie Furmento
|
87ed874e79
configure.ac: typo
|
14 年 前 |
|
Nathalie Furmento
|
ed65739272
tests/loader: increase default timeout after when unfinished tests are killed
|
14 年 前 |
 Cédric Augonnet
|
689896a8ef
The performance of the MPI Cholesky kernel is obtained by taking the
|
14 年 前 |
|
Samuel Thibault
|
009817b083
mention that StarPU already sets the local stream for blas operations
|
14 年 前 |
|
Cédric Augonnet
|
08e2b87a27
In case MAGMA is available (a recent version with pkg-config is now required),
|
14 年 前 |
|
Cédric Augonnet
|
aed2c35fd7
Improve the CPU kernel of the pi_redux benchmark by using the erand48_r which
|
14 年 前 |
 François Tessier
|
894a14c4dc
Define task->cl_arg_size in block example
|
14 年 前 |
|
Nathalie Furmento
|
7560ff696a
mpi/examples/cholesky: fix output
|
14 年 前 |
|
Nathalie Furmento
|
96f41d8c73
configure.ac: add -lfxt to LDFLAGS when using user-installed FxT
|
14 年 前 |
