| Commit message (Collapse) | Author | Age |
| |
|
|
|
|
| |
support it.
|
| |
|
| |
|
| |
|
|
|
|
| |
some nvcc limitations.
|
| |
|
| |
|
| |
|
| |
|
|
|
|
| |
floats on CUDA devices
|
|
|
|
| |
using an older version of CUDA
|
|
|
|
| |
convert floats into half floats and vice versa
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
|\ |
|
| | |
|
| |
| |
| |
| |
| |
| |
| |
| | |
This changeset add two specializations for float/double on SSE. Those
are mostly usefull with GCC for which std::sqrt add an extra and costly
check on the result of _mm_sqrt_*. Clang does not add this burden.
In this changeset, only DenseBase::norm() makes use of it.
|
| | |
|
| |\
| |/
|/| |
|
| | |
|
|/ |
|
| |
|
| |
|
| |
|
| |
|
|\
| |
| |
| | |
Add special functions to eigen: lgamma, erf, erfc.
|
| | |
|
|/
|
|
| |
Includes CUDA support and unit tests.
|
| |
|
| |
|
|\
| |
| |
| | |
Add round, ceil and floor for SSE4.1/AVX (Bug #70)
|
| |
| |
| |
| | |
implementations older than 3.5
|
|/ |
|
| |
|
| |
|
|
|
|
| |
SSE3 integer vectorization, plus minor tweaks)
|
| |
|
| |
|
| |
|
| |
|
|
|
|
| |
only in fast-math mode (as SSE)
|
|
|
|
| |
everywhere
|
| |
|
|
|
|
| |
world as in microbenchmark.
|