 Cédric Augonnet
|
c054754a75
Implement bus performance sampling: this is no longer hardcoded.
|
il y a 16 ans |
|
Cédric Augonnet
|
a3be10b313
The performande models associated to the codelets are now put in
|
il y a 16 ans |
|
Cédric Augonnet
|
3cad2c6bbb
The src and dst nodes were inverted in the matrices
|
il y a 16 ans |
|
Cédric Augonnet
|
e9b8da269e
add some dirty program to measure the bandwith/latency used by the performance models
|
il y a 16 ans |
|
Cédric Augonnet
|
0b80205b82
It is possible to generate a file with the distribution of the duration of the
|
il y a 16 ans |
|
Cédric Augonnet
|
bcd9c67fc4
Fix debug assertions, it is possible that no request is associated to a data
|
il y a 16 ans |
|
Cédric Augonnet
|
e561dd511a
We now assume that the header lock is already taken when calling
|
il y a 16 ans |
|
Cédric Augonnet
|
eb5c03ad79
BUG FIX: when a data dependency is unlocked, we have to update the current mode
|
il y a 16 ans |
|
Cédric Augonnet
|
485ce7b82f
cudaMemset only fills one byte at a time, so we use a char array instead of an
|
il y a 16 ans |
|
Cédric Augonnet
|
20420637e6
starpu_wait_all_tasks and starpu_wait_task are blocking too.
|
il y a 16 ans |
|
Cédric Augonnet
|
6f21718f48
The helper functions to allocate/liberate pinned memory use tasks internally,
|
il y a 16 ans |
|
Cédric Augonnet
|
5284a61274
When a data request needs to actually perform a data transfer from a source, we
|
il y a 16 ans |
|
Cédric Augonnet
|
499cbfed6a
Also check that we don't synchronize with a tag while blocking is prohibited.
|
il y a 16 ans |
|
Cédric Augonnet
|
02119279fc
We now check whether the following blocking calls are called from a callback or
|
il y a 16 ans |
|
Cédric Augonnet
|
a9394e1af3
Keep the status of each worker so that we can detect whether a blocking call is
|
il y a 16 ans |
|
Cédric Augonnet
|
4ac2e92f42
Add another classical synchronization problem to stress data management.
|
il y a 16 ans |
|
Cédric Augonnet
|
152b6a7329
Add a the classical dining philosopher example: we have to explicitely order
|
il y a 16 ans |
|
Cédric Augonnet
|
9e1bd3857d
Make it possible to select the maximum number of buffers per task at configure
|
il y a 16 ans |
|
Cédric Augonnet
|
da38bcd033
Implement async data transfers for the "block" data interface.
|
il y a 16 ans |
|
Cédric Augonnet
|
d3cfd654a8
Cleanup FxT tool, there is no svg output anymore (Vite is much better in every
|
il y a 16 ans |
|
Cédric Augonnet
|
d8132d45cf
Make it possible to select the colour of the tasks in the DOT output depending
|
il y a 16 ans |
|
Cédric Augonnet
|
82828a010d
On some systems (eg. MacOS), it's just not possible to bind threads at all, so
|
il y a 16 ans |
|
Cédric Augonnet
|
98540ae1f7
don't compile CUDA functions if CUDA is not available
|
il y a 16 ans |
|
Cédric Augonnet
|
5c984e9279
make sure that FFTW is available for the starpufft lib
|
il y a 16 ans |
|
Cédric Augonnet
|
1773849014
Properly handle the case of write-only buffers when prefetching (in this case,
|
il y a 17 ans |
|
Cédric Augonnet
|
277365f5c4
if a piece of data is accessed in a write only mode, there is no transfer
|
il y a 17 ans |
|
Cédric Augonnet
|
272db710fc
We now store what is the mask associated to each worker, so that we can test if
|
il y a 17 ans |
|
Cédric Augonnet
|
dc04c50247
Make the test more ""challenging"" (test the case where only a subset of the
|
il y a 17 ans |
|
Cédric Augonnet
|
2e5a4054ec
cleanup tests
|
il y a 17 ans |
|
Cédric Augonnet
|
e3defda082
Properly handle the case of 'write-only' buffers: there is no need for having a
|
il y a 17 ans |
