|
Uni20
Asynchronous tensor and runtime library.
|
One residual M or N loop around a rank-two GEMM projection. More...
#include <contraction_strides.hpp>
Public Attributes | |
| DirectContractionGemmPlan | gemm |
| uni20::index_type | lhs_offset_stride = 0 |
| uni20::index_type | loop_extent = 0 |
| uni20::index_type | output_offset_stride = 0 |
| uni20::index_type | rhs_offset_stride = 0 |
One residual M or N loop around a rank-two GEMM projection.
Each loop iteration addresses one disjoint output slice. The offset strides are measured in elements from the original mdspec descriptors or immediate handles.
| DirectContractionGemmPlan uni20::linalg::LoopedContractionGemmPlan::gemm |
| uni20::index_type uni20::linalg::LoopedContractionGemmPlan::lhs_offset_stride = 0 |
| uni20::index_type uni20::linalg::LoopedContractionGemmPlan::loop_extent = 0 |
| uni20::index_type uni20::linalg::LoopedContractionGemmPlan::output_offset_stride = 0 |
| uni20::index_type uni20::linalg::LoopedContractionGemmPlan::rhs_offset_stride = 0 |