 Cédric Augonnet
|
722ba734ee
Allocate pinned memory in an existing context rather than from the application
|
%!s(int64=16) %!d(string=hai) anos |
|
Cédric Augonnet
|
13dd236182
Keep this function public
|
%!s(int64=16) %!d(string=hai) anos |
 Nathalie Furmento
|
3c7a6bade8
Rename cuda functions
|
%!s(int64=16) %!d(string=hai) anos |
 Samuel Thibault
|
fd11fd4590
TODO: use hwloc for binding for cuda perfs
|
%!s(int64=16) %!d(string=hai) anos |
|
Cédric Augonnet
|
f838d1aae0
Use an array of temporary buffers for "task 11" instead of a single buffer.
|
%!s(int64=16) %!d(string=hai) anos |
|
Cédric Augonnet
|
1d37506834
by default, disable debugging messages
|
%!s(int64=16) %!d(string=hai) anos |
|
Samuel Thibault
|
a378191763
debugging message according to C convention
|
%!s(int64=16) %!d(string=hai) anos |
|
Samuel Thibault
|
d22f7b3aef
fix tag dependency, seems to work better now
|
%!s(int64=16) %!d(string=hai) anos |
|
Samuel Thibault
|
34e12b640f
more readable debugging
|
%!s(int64=16) %!d(string=hai) anos |
|
Samuel Thibault
|
2c2170c9c8
initialize thread support, warn if no thread support is available
|
%!s(int64=16) %!d(string=hai) anos |
|
Samuel Thibault
|
fd84a2abd0
cosmetic changes
|
%!s(int64=16) %!d(string=hai) anos |
|
Samuel Thibault
|
31ec191f32
fix parameter name coherencey
|
%!s(int64=16) %!d(string=hai) anos |
|
Samuel Thibault
|
c5cc01e6d5
Fix MPI scatter
|
%!s(int64=16) %!d(string=hai) anos |
|
Cédric Augonnet
|
417c14b7f0
use the good parameters for CUDA calls
|
%!s(int64=16) %!d(string=hai) anos |
|
Samuel Thibault
|
c05e4565d4
only the owner of the block should send it to node 0
|
%!s(int64=16) %!d(string=hai) anos |
|
Cédric Augonnet
|
578c319b4a
Fix bug and add error checking
|
%!s(int64=16) %!d(string=hai) anos |
|
Cédric Augonnet
|
4408a54de9
add an experiment to measure the latency between 2 GPUs
|
%!s(int64=16) %!d(string=hai) anos |
|
Cédric Augonnet
|
c1273c8f02
- add a (pretty naive) test to measure the latency between 2 GPUs
|
%!s(int64=16) %!d(string=hai) anos |
|
Cédric Augonnet
|
516411357c
forgot to commit the public headers in the previous commit
|
%!s(int64=16) %!d(string=hai) anos |
|
Cédric Augonnet
|
3ea8462620
- Use events instead of streams to check whether a data transfer is terminated
|
%!s(int64=16) %!d(string=hai) anos |
|
Cédric Augonnet
|
b308e85842
The user events are typically not executed by a worker, so in case the event is
|
%!s(int64=16) %!d(string=hai) anos |
|
Cédric Augonnet
|
3946a1e554
Forgot to move the cuda kernel associated to the sync_and_notify_data test.
|
%!s(int64=16) %!d(string=hai) anos |
|
Cédric Augonnet
|
c7d8e962c0
The progression hook is a mechanism which should be use in very specific cases.
|
%!s(int64=16) %!d(string=hai) anos |
|
Cédric Augonnet
|
5a38bde976
tag-wait-api is not a microbench either
|
%!s(int64=16) %!d(string=hai) anos |
|
Cédric Augonnet
|
59a0283279
move the sync_and_notify_data example in a more appropriate directory too
|
%!s(int64=16) %!d(string=hai) anos |
|
Cédric Augonnet
|
7363413be1
Move the "dsm stress" example in a more appropriate directory, this is not a
|
%!s(int64=16) %!d(string=hai) anos |
|
Cédric Augonnet
|
89c1870f00
Define a new header which should contain the public functions that should only
|
%!s(int64=16) %!d(string=hai) anos |
|
Cédric Augonnet
|
4a8fe3239e
improve error checking
|
%!s(int64=16) %!d(string=hai) anos |
|
Cédric Augonnet
|
dc7a276f27
add some debug messages in the MPI lib
|
%!s(int64=16) %!d(string=hai) anos |
|
Cédric Augonnet
|
d247fadc97
add a -display flag
|
%!s(int64=16) %!d(string=hai) anos |
