mirror/eigen
2
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2024-12-21 07:19:46 +08:00
Code Issues Packages Projects Releases Wiki Activity

Avoid relying on a default value for the Vectorizable template parameter of the EvalRange functor

Browse Source
This commit is contained in:
Benoit Steiner 2015-07-15 17:17:04 -07:00
parent 4b3d697e12
commit b900fe47d5
1 changed files with 3 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
unsupported/Eigen/CXX11/src/Tensor/TensorExecutor.h
View File

@ -75,7 +75,7 @@ class TensorExecutor<Expression, DefaultDevice, true>
// Multicore strategy: the index space is partitioned and each partition is executed on a single core
#ifdef EIGEN_USE_THREADS
template <typename Evaluator, typename Index, bool Vectorizable = Evaluator::PacketAccess>
template <typename Evaluator, typename Index, bool Vectorizable>
struct EvalRange {
static void run(Evaluator evaluator, const Index first, const Index last) {
eigen_assert(last > first);
@ -129,11 +129,11 @@ class TensorExecutor<Expression, ThreadPoolDevice, Vectorizable>
std::vector<Notification*> results;
results.reserve(numblocks);
for (int i = 0; i < numblocks; ++i) {
results.push_back(device.enqueue(&EvalRange<Evaluator, Index>::run, evaluator, i*blocksize, (i+1)*blocksize));
results.push_back(device.enqueue(&EvalRange<Evaluator, Index, Vectorizable>::run, evaluator, i*blocksize, (i+1)*blocksize));
}
if (numblocks * blocksize < size) {
EvalRange<Evaluator, Index>::run(evaluator, numblocks * blocksize, size);
EvalRange<Evaluator, Index, Vectorizable>::run(evaluator, numblocks * blocksize, size);
}
for (int i = 0; i < numblocks; ++i) {