 Samuel Thibault
|
51c2f1bb49
fix simgrid check with quick_check
|
vor 9 Jahren |
|
Samuel Thibault
|
cd11bb3f79
revert r17989, blas availability in simgrid mode was actually already fixed globally
|
vor 9 Jahren |
|
Samuel Thibault
|
02b82caf1b
allow building sgemm & dgemm in simgrid mode without a blas lib
|
vor 9 Jahren |
|
Samuel Thibault
|
1bb4b3b61e
drop spurious line
|
vor 9 Jahren |
|
Samuel Thibault
|
18670a9934
Add starpu_fxt_autostart_profiling to be able to avoid autostart.
|
vor 9 Jahren |
|
Samuel Thibault
|
ab4f759b2a
fix assertion
|
vor 9 Jahren |
|
Samuel Thibault
|
8f654dead2
Add "folded" memory allocation in simulation mode.
|
vor 9 Jahren |
|
Samuel Thibault
|
b74dfb5713
port r16718 from trunk: Make xgemm advise returning C pieces to main memory once finished with them
|
vor 9 Jahren |
|
Samuel Thibault
|
477690d953
Add -size parameter to xgemm
|
vor 9 Jahren |
|
Samuel Thibault
|
a95613dc3b
Enable codelets using BLAS on MIC
|
vor 9 Jahren |
|
Samuel Thibault
|
00b1ef6f74
Add more verbose comments in examples source code
|
vor 9 Jahren |
 Nathalie Furmento
|
bb61f3808d
update cnrs copyright
|
vor 10 Jahren |
|
Samuel Thibault
|
d9ebdb78f8
revert spurious change
|
vor 10 Jahren |
|
Samuel Thibault
|
27f18f4c66
upgrade on the proper list
|
vor 10 Jahren |
|
Samuel Thibault
|
c6ede8b07d
always enable CUDA gemm kernel in simgrid mode
|
vor 10 Jahren |
|
Samuel Thibault
|
353019fea0
Enable integer resolution for gemm, it's not costly here
|
vor 10 Jahren |
|
Samuel Thibault
|
c66eca1555
increase default matrix size
|
vor 10 Jahren |
|
Samuel Thibault
|
be9be0e496
Fix flops computation in gemm
|
vor 10 Jahren |
|
Samuel Thibault
|
8ede268af6
Add bound support to gemm
|
vor 10 Jahren |
|
Samuel Thibault
|
6757bca1b5
Remove implementation array terminator in tests, examples, and documentation, since it is not actually used
|
vor 10 Jahren |
|
Samuel Thibault
|
508f2cc47d
Add flops
|
vor 10 Jahren |
|
Samuel Thibault
|
f7bc4f304f
Fix warnings
|
vor 10 Jahren |
|
Samuel Thibault
|
d84a627a42
Do not factorize code in xgemm codelet, to avoid confusing the reader
|
vor 10 Jahren |
|
Samuel Thibault
|
605deae422
Make sure people don't try to compile xgemm.c directly
|
vor 10 Jahren |
|
Samuel Thibault
|
4b1a16edde
Fix University name after the fusion..
|
vor 10 Jahren |
|
Samuel Thibault
|
1ee69f6aaa
drop including sys/time.h, now useless
|
vor 10 Jahren |
|
Samuel Thibault
|
7d2bf0c630
Enable asynchronous CUDA kernels wherever possible
|
vor 11 Jahren |
 Corentin Salingue
|
6e5d499f17
replace 0 with macro part 6
|
vor 12 Jahren |
|
Samuel Thibault
|
b227dbcfdc
Use STARPU_ATTRIBUTE_* macros
|
vor 12 Jahren |
|
Samuel Thibault
|
df3534242c
make xgemm abort on erroneous parameter
|
vor 12 Jahren |
