| Commit message (Collapse) | Author | Age |
| |
|
| |
|
| |
|
| |
|
|
|
|
| |
floats on CUDA devices
|
|
|
|
| |
using an older version of CUDA
|
| |
|
| |
|
|
|
|
| |
convert floats into half floats and vice versa
|
| |
|
|
|
|
| |
tensor expression.
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
| |
systems don't support the __uin128_t type
|
| |
|
| |
|
|
|
|
| |
compatible with pairs of element.
|
|
|
|
| |
some compiler warnings
|
| |
|
|
|
|
| |
function is not allowed" messages
|
|\
| |
| |
| | |
Add constructor for long types.
|
| | |
|
| |
| |
| |
| | |
required stride isn't available.
|
| | |
|
| |\
| |/
|/| |
|
| | |
|
| | |
|
| | |
|
| |
| |
| |
| | |
a tensor expression on a CUDA device. This makesit possible to set aside streaming multiprocessors for other computations.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
| |
| |
| | |
more resistant to numerical noise to prevent spurious test failures.
|
| |
| |
| |
| | |
in the tensor thread pool test
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
|\ \ |
|
| | | |
|