mirror/eigen
2
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2025-01-30 17:40:05 +08:00
Code Issues Packages Projects Releases Wiki Activity

Fixed the evaluation of expressions involving tensors of 2 or 3 elements on CUDA devices.

Browse Source
This commit is contained in:
Benoit Steiner 2014-11-18 14:32:41 -08:00
parent 1d3c8306f8
commit b33cf92878
1 changed files with 2 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
unsupported/Eigen/CXX11/src/Tensor/TensorExecutor.h
View File

@ -168,11 +168,10 @@ __launch_bounds__(1024)
const Index PacketSize = unpacket_traits<typename Evaluator::PacketReturnType>::size;
const Index vectorized_step_size = step_size * PacketSize;
const Index vectorized_size = (size / PacketSize) * PacketSize;
Index i = first_index * PacketSize;
for ( ; i < vectorized_size; i += vectorized_step_size) {
for (Index i = first_index * PacketSize; i < vectorized_size; i += vectorized_step_size) {
eval.evalPacket(i);
}
for ( ; i < size; i += step_size) {
for (Index i = vectorized_size + first_index; i < size; i += step_size) {
eval.evalScalar(i);
}
}