 Samuel Thibault
|
3621a2818e
benefit from magma 1.4.0 and later being able to use an application-provided stream
|
11 år sedan |
|
Samuel Thibault
|
90bbc50bb2
harmonize fixes with the distributed version
|
11 år sedan |
|
Samuel Thibault
|
2d00fb74e7
Only set stream when using magma. Without using magma, starpu_cublas_init is enough
|
11 år sedan |
|
Samuel Thibault
|
5f2955555e
always set the kernel stream in the codelet itself, since magma apparently changes it dynamically
|
11 år sedan |
|
Samuel Thibault
|
7fa7ccb485
Fix potrf kernel: benefit from the MKL implementation, and Magma is actually not asynchronous
|
11 år sedan |
|
Samuel Thibault
|
2fe69151e1
revert 14030, keep the same model name, to avoid confusions
|
11 år sedan |
|
Samuel Thibault
|
751c44cd97
Use a different model name when magma support is enabled
|
11 år sedan |
|
Samuel Thibault
|
17b687c854
Fix magma API use
|
11 år sedan |
|
Samuel Thibault
|
6757bca1b5
Remove implementation array terminator in tests, examples, and documentation, since it is not actually used
|
11 år sedan |
|
Samuel Thibault
|
f7bc4f304f
Fix warnings
|
11 år sedan |
|
Samuel Thibault
|
4b1a16edde
Fix University name after the fusion..
|
11 år sedan |
|
Samuel Thibault
|
b6d83183ea
port r12770 from 1.1: MKL now provides GEMM etc. symbols, prefix ours to avoid conflicting
|
11 år sedan |
|
Samuel Thibault
|
3725f99550
Fix build with magma
|
11 år sedan |
|
Samuel Thibault
|
87bf731e40
Make Cholesky use asynchronous CUDA tasks. Also factorize the codelet structures along the way.
|
11 år sedan |
|
Samuel Thibault
|
b227dbcfdc
Use STARPU_ATTRIBUTE_* macros
|
12 år sedan |
 Nathalie Furmento
|
ca7d7aa451
examples: only include starpu.h as it includes all the other starpu_ .h files
|
13 år sedan |
|
Samuel Thibault
|
29b7f32f77
Provide good examples by always using streams for cublas
|
13 år sedan |
|
Nathalie Furmento
|
b9476e6d4a
update code w.r.t coding style
|
14 år sedan |
 Cédric Augonnet
|
08e2b87a27
In case MAGMA is available (a recent version with pkg-config is now required),
|
14 år sedan |
|
Samuel Thibault
|
68c2135caf
fix yet more c++-like comments
|
14 år sedan |
|
Cédric Augonnet
|
c03a4fad7d
Fix POTRF kernel : cudaMemcpyAsync needs pinned memory
|
14 år sedan |
|
Cédric Augonnet
|
355f25c062
Implement a parallel version of the GEMM kernel for the cholesky example.
|
14 år sedan |
|
Cédric Augonnet
|
b8ffa4f5e8
Use streams instead of cudaThreadSynchronize
|
14 år sedan |
|
Cédric Augonnet
|
0de9ecb7a8
Massive cleanup in the examples/cholesky/ directory. Remove traces of the "dw_"
|
14 år sedan |
