 Samuel Thibault
|
505aaf5a25
fix build
|
12 yıl önce |
|
Samuel Thibault
|
dc60b6bd7a
Avoid redefining cudaStream_t, in case the caller defines STARPU_DONT_INCLUDE_CUDA_HEADERS and still includes some cuda header itself
|
12 yıl önce |
|
Samuel Thibault
|
1f81323901
fix build
|
12 yıl önce |
|
Samuel Thibault
|
f3d488b260
Avoid including cublas.h to avoid conflicting with cublas_v2.h
|
12 yıl önce |
|
Samuel Thibault
|
34248f4d14
explicit versions of the blas
|
12 yıl önce |
 Nathalie Furmento
|
36fe1cbc6e
src/core/perfmodel: store at the beginning of the application the names for the different perfmodel directories, there is no need to recalcute them at each function call
|
12 yıl önce |
 Olivier Aumage
|
a0c36cc735
- execute Cuda/OpenCL omp target tasks without StarPU/OpenMP wrapper
|
12 yıl önce |
|
Samuel Thibault
|
353019fea0
Enable integer resolution for gemm, it's not costly here
|
12 yıl önce |
|
Samuel Thibault
|
c53f16a65d
Add gemm perfmodels
|
12 yıl önce |
|
Samuel Thibault
|
c66eca1555
increase default matrix size
|
12 yıl önce |
|
Samuel Thibault
|
be9be0e496
Fix flops computation in gemm
|
12 yıl önce |
|
Samuel Thibault
|
f7fe21380e
TODO
|
12 yıl önce |
|
Samuel Thibault
|
8ede268af6
Add bound support to gemm
|
12 yıl önce |
|
Samuel Thibault
|
ed082a5e41
Integrate bound in lu output
|
12 yıl önce |
|
Samuel Thibault
|
3192fc5b08
Make some example outputs nicer to plot
|
12 yıl önce |
|
Samuel Thibault
|
24a6bf49c0
Fix footprint computation without bound
|
12 yıl önce |
|
Samuel Thibault
|
d6b04d16b6
Fix crash when enabling bounds
|
12 yıl önce |
|
Samuel Thibault
|
d045a58e37
Add stencil perfmodel
|
12 yıl önce |
|
Samuel Thibault
|
0630c33945
Fix running stencil in simgrid
|
12 yıl önce |
|
Samuel Thibault
|
699f63f475
Use bigger tasks by default
|
12 yıl önce |
|
Samuel Thibault
|
0fee7354fd
Increase the default matrix size to get about 1GiB data
|
12 yıl önce |
|
Samuel Thibault
|
e4d72271e7
port r14072 from 1.1: Document a bit the mirage architecture
|
12 yıl önce |
|
Samuel Thibault
|
4b28886ba9
Add more measurements
|
12 yıl önce |
|
Samuel Thibault
|
c7c9ae1aa7
Add missing flops
|
12 yıl önce |
|
Samuel Thibault
|
95993904fc
Use approximation when comparing declared flops, since when printing the performance file we get some rounding
|
12 yıl önce |
|
Samuel Thibault
|
60b85b966b
add lu double perfmodel
|
12 yıl önce |
|
Samuel Thibault
|
6f02123c98
assume double support is available in simgrid mode
|
12 yıl önce |
|
Samuel Thibault
|
41d4c8a7fe
Also enable ASYNC in simgrid mode
|
12 yıl önce |
|
Samuel Thibault
|
e09b976171
Ajout perfmodels lu float
|
12 yıl önce |
|
Samuel Thibault
|
eb7b6a9fbc
Add performance models for mirage
|
12 yıl önce |
