 Cédric Augonnet
|
e49ca9715c
Factorize the code to detect whether prefetch should be used or not
|
15 년 전 |
|
Cédric Augonnet
|
710e89d51f
- with STARPU_CALIBRATE=2, StarPU calibrates performance models, but overwrites
|
15 년 전 |
|
Cédric Augonnet
|
aed837b56a
Remove a debug message that was commited by mistake in the previous commit.
|
15 년 전 |
|
Cédric Augonnet
|
979164bb41
Fix a confusion when defining a filter that manipulates a piece of data with a
|
15 년 전 |
|
Cédric Augonnet
|
4d75c11e35
Remove the totally obsolete Makefiles that are still wandering in the
|
15 년 전 |
 Samuel Thibault
|
aa8cce574e
reference-count starpu initialization
|
15 년 전 |
|
Cédric Augonnet
|
c037406638
Make it possible to select the number of tasks at runtime with the "-ntasks"
|
15 년 전 |
|
Cédric Augonnet
|
feff8b1018
Also display the distribution of the tasks between the different processing units
|
15 년 전 |
|
Cédric Augonnet
|
d99af315c6
Document STARPU_TASK_INITIALIZER
|
15 년 전 |
|
Cédric Augonnet
|
be0570c9d2
- Use a write-back cache policy for the "counter" data
|
15 년 전 |
|
Cédric Augonnet
|
6b2bcd48a5
- Rather than generating the random numbers on the host (which introduces a
|
15 년 전 |
|
Cédric Augonnet
|
bc49f72836
Do not spawn too many threads if the tasks are too big. While we now have a
|
15 년 전 |
|
Cédric Augonnet
|
66f82bac9e
Mention hwloc in the requirements section of the README
|
15 년 전 |
|
Cédric Augonnet
|
8fd98c847d
provide a performance model name for the pi example
|
15 년 전 |
|
Cédric Augonnet
|
d6d00f1f8b
Fix compilation if CUDA is not available
|
15 년 전 |
|
Cédric Augonnet
|
d2a2854b33
Use a reduction to sum the number of successful shots within a block
|
15 년 전 |
|
Cédric Augonnet
|
76f2fd6053
Implement a (really naive) CUDA kernel that computes pi with a Monte Carlo
|
15 년 전 |
|
Cédric Augonnet
|
7b30fad879
Add a very simple monte carlo example to compute pi (CPU only yet)
|
15 년 전 |
|
Cédric Augonnet
|
fb3d7bb6a9
The "id" field was previously renamed into "devid"
|
15 년 전 |
|
Cédric Augonnet
|
d996551e3a
Document how to install FxT so that StarPU can generate traces
|
15 년 전 |
|
Cédric Augonnet
|
42676df753
Only built the CUDA kernels for SpMV if CUDA is enabled
|
15 년 전 |
|
Cédric Augonnet
|
37679a5020
Prefix the last calls to get_sub_data
|
15 년 전 |
|
Cédric Augonnet
|
60a17cf50b
There is no point in maintaining total_number_of_jobs in the FIFO structure
|
15 년 전 |
|
Cédric Augonnet
|
e98a9d724d
Some calls to get_sub_data in examples that are only available with a BLAS lib
|
15 년 전 |
 Nathalie Furmento
|
815b70f3ac
followup to r1803: Rename the "blas" interface into "matrix"
|
15 년 전 |
|
Nathalie Furmento
|
432e1e6dd8
fix CPPFLAGS when building outside of source directory
|
15 년 전 |
|
Samuel Thibault
|
498359fe4d
minimal documentation of what a memory node is
|
15 년 전 |
|
Cédric Augonnet
|
b9e7a62fd0
The CSR filter was using a float pointer to compute the offset within the data
|
15 년 전 |
|
Samuel Thibault
|
ebae02ad27
the device mask uses the STARPU_ prefix
|
15 년 전 |
|
Cédric Augonnet
|
77c98167af
Fix typos in the documentation, thanks to Sylvain Gault for reporting them !
|
15 년 전 |
