mirror of
https://gitlab.com/libeigen/eigen.git
synced 2025-05-04 01:34:07 +08:00
Reverted a previous change that tripped nvcc when compiling in debug mode.
This commit is contained in:
parent
bbdabbb379
commit
bd7d901da9
@ -1274,7 +1274,7 @@ struct TensorEvaluator<const TensorContractionOp<Indices, LeftArgType, RightArgT
|
|||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
EIGEN_DEVICE_FUNC void evalTo(Scalar* buffer) const {
|
void evalTo(Scalar* buffer) const {
|
||||||
if (this->m_lhs_inner_dim_contiguous) {
|
if (this->m_lhs_inner_dim_contiguous) {
|
||||||
if (this->m_rhs_inner_dim_contiguous) {
|
if (this->m_rhs_inner_dim_contiguous) {
|
||||||
if (this->m_rhs_inner_dim_reordered) {
|
if (this->m_rhs_inner_dim_reordered) {
|
||||||
@ -1313,7 +1313,7 @@ struct TensorEvaluator<const TensorContractionOp<Indices, LeftArgType, RightArgT
|
|||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
template <bool lhs_inner_dim_contiguous, bool rhs_inner_dim_contiguous, bool rhs_inner_dim_reordered, int Alignment> EIGEN_DEVICE_FUNC
|
template <bool lhs_inner_dim_contiguous, bool rhs_inner_dim_contiguous, bool rhs_inner_dim_reordered, int Alignment>
|
||||||
void evalTyped(Scalar* buffer) const {
|
void evalTyped(Scalar* buffer) const {
|
||||||
// columns in left side, rows in right side
|
// columns in left side, rows in right side
|
||||||
const Index k = this->m_k_size;
|
const Index k = this->m_k_size;
|
||||||
@ -1362,7 +1362,7 @@ struct TensorEvaluator<const TensorContractionOp<Indices, LeftArgType, RightArgT
|
|||||||
const dim3 block_size(16, 16, 1);
|
const dim3 block_size(16, 16, 1);
|
||||||
LAUNCH_CUDA_KERNEL((EigenFloatContractionKernel16x16<Index, LhsMapper, RhsMapper, OutputMapper>), num_blocks, block_size, 0, this->m_device, lhs, rhs, output, m, n, k);
|
LAUNCH_CUDA_KERNEL((EigenFloatContractionKernel16x16<Index, LhsMapper, RhsMapper, OutputMapper>), num_blocks, block_size, 0, this->m_device, lhs, rhs, output, m, n, k);
|
||||||
} else {
|
} else {
|
||||||
const Index m_blocks = (m + 127) / 128;
|
const Index m_blocks = (m + 127) / 128;
|
||||||
const Index n_blocks = (n + 63) / 64;
|
const Index n_blocks = (n + 63) / 64;
|
||||||
const dim3 num_blocks(m_blocks, n_blocks, 1);
|
const dim3 num_blocks(m_blocks, n_blocks, 1);
|
||||||
const dim3 block_size(8, 32, 1);
|
const dim3 block_size(8, 32, 1);
|
||||||
|
Loading…
x
Reference in New Issue
Block a user