mirror/eigen
2
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2024-12-27 07:29:52 +08:00
Code Issues Packages Projects Releases Wiki Activity

Revert code lost in merge

Browse Source
This commit is contained in:
Eugene Zhulenev 2018-09-27 12:08:17 -07:00
parent a7a3e9f2b6
commit 9f33e71e9d
1 changed files with 5 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
unsupported/Eigen/CXX11/src/Tensor/TensorContraction.h
View File

@ -672,11 +672,12 @@ struct TensorContractionEvaluatorBase
}
template <bool lhs_inner_dim_contiguous, bool rhs_inner_dim_contiguous, bool rhs_inner_dim_reordered, int Alignment>
EIGEN_DEVICE_FUNC void evalGemmPartial(Scalar* buffer, Index k_start, Index k_end, int /*num_threads*/) const {
EIGEN_DEVICE_FUNC void evalGemmPartial(Scalar* buffer, Index k_start, Index k_end, int num_threads) const {
// columns in left side, rows in right side
const Index k = this->m_k_size;
eigen_assert(k_end >= k_start && k_start >= 0 && k_end <= k);
const Index k_slice = k_end - k_start;
// rows in left side
const Index m = this->m_i_size;
@ -722,7 +723,9 @@ struct TensorContractionEvaluatorBase
OutputMapper output(buffer, m);
// Sizes of the blocks to load in cache. See the Goto paper for details.
internal::TensorContractionBlocking<Scalar, LhsScalar, RhsScalar, Index, internal::ShardByCol> blocking(k, m, n, 1);
internal::TensorContractionBlocking<Scalar, LhsScalar, RhsScalar,
Index, internal::ShardByCol>
blocking(k_slice, m, n, num_threads);
const Index kc = blocking.kc();
const Index mc = numext::mini(m, blocking.mc());
const Index nc = numext::mini(n, blocking.nc());