Cutlass
CUDA Templates for Linear Algebra Subroutines and Solvers
|
#include <gemm_global_stream.h>
Public Types | |
typedef GlobalLoadStreamBase< LoadIterator_, StoreIterator_, Transformer_ > | Base |
The base class. More... | |
![]() | |
typedef LoadIterator_ | LoadIterator |
The load iterator. More... | |
typedef Transformer_ | Transformer |
The transformer. More... | |
typedef StoreIterator_ | StoreIterator |
The store iterator to write to shared memory. More... | |
typedef LoadIterator::Fragment | FetchedFragment |
The fragment that is copied from shared memory. More... | |
typedef Transformer::OutputFragment | TransformedFragment |
The fragment that is obtained after the transformation by the transformer. More... | |
typedef TransformedFragment | Fragment |
Make sure the fragments match. More... | |
typedef LoadIterator::Scalar | Scalar |
The scalar type of the iterator. More... | |
typedef LoadIterator::Pointer | Pointer |
The pointer. More... | |
typedef LoadIterator::Index | Index |
The index. More... | |
typedef StoreIterator::SharedStorage | SharedStoreStorage |
The amount of storage in shared memory needed to store the tile. More... | |
Public Member Functions | |
CUTLASS_DEVICE | GlobalLoadStream (typename Base::Params const ¶ms, typename Base::SharedStorage &shared_storage, Coord< 3 > const &bounds, Coord< 3 > const &block) |
Ctor. More... | |
![]() | |
CUTLASS_DEVICE | GlobalLoadStreamBase (Params const ¶ms, SharedStorage &shared_storage, Coord< 3 > const bounds, Coord< 3 > const &block) |
Ctor. More... | |
CUTLASS_DEVICE void | copy () |
Load the data from shared memory to the fetch fragment. More... | |
CUTLASS_DEVICE void | commit () |
Commit the data. More... | |
CUTLASS_DEVICE void | residue (Index k, bool skip_clear=false) |
Execute the residue code. More... | |
Additional Inherited Members | |
![]() | |
LoadIterator | load_iterator |
The iterator. More... | |
FetchedFragment | fetched_fragment |
The fragment to fetch from shared memory. More... | |
Transformer | transformer |
The transformer. More... | |
TransformedFragment | transformed_fragment |
The fragment to convert the data after it has been fetched from shared memory. More... | |
StoreIterator | store_iterator |
The store iterator. More... | |
![]() | |
static MatrixLayout::Kind const | kLayout = LoadIterator::kLayout |
Make sure the transformed fragment is the same as the store fragment. More... | |
typedef GlobalLoadStreamBase<LoadIterator_, StoreIterator_, Transformer_> cutlass::gemm::GlobalLoadStream< LoadIterator_, StoreIterator_, Transformer_ >::Base |
|
inline |