50 typename OperatorClass,
55 typename ElementAccumulator
66 typename ElementAccumulator>
75 static int const kAlignmentA = 1;
76 static int const kAlignmentB = 1;
80 static int const kStages = 2;
99 static int const kAlignmentA = 4;
100 static int const kAlignmentB = 4;
104 static int const kStages = 2;
123 typename ElementAccumulator>
125 arch::OpClassWmmaTensorOp,
130 ElementAccumulator> {
135 static int const kStages = 2;
153 typename ElementAccumulator>
155 arch::OpClassTensorOp,
160 ElementAccumulator> {
168 static int const kStages = 2;
186 typename ElementAccumulator>
188 arch::OpClassTensorOp,
193 ElementAccumulator> {
200 static int const kStages = 2;
214 arch::OpMultiplyAddSaturate, arch::OpMultiplyAdd>::type;
222 arch::OpClassTensorOp,
235 static int const kStages = 2;
248 arch::OpClassTensorOp,
261 static int const kStages = 2;
274 arch::OpClassTensorOp,
287 static int const kStages = 2;
300 arch::OpClassTensorOp,
313 static int const kStages = 2;
326 arch::OpClassTensorOp,
339 static int const kStages = 2;
352 arch::OpClassTensorOp,
365 static int const kStages = 2;
378 arch::OpClassTensorOp,
391 static int const kStages = 2;
404 arch::OpClassTensorOp,
417 static int const kStages = 2;
Definition: aligned_buffer.h:35
arch::OpMultiplyAddSaturate Operator
Definition: default_gemm_configuration.h:240
Definition: linear_combination.h:56
arch::OpMultiplyAddSaturate Operator
Definition: default_gemm_configuration.h:422
arch::OpMultiplyAddSaturate Operator
Definition: default_gemm_configuration.h:292
Definition: linear_combination_clamp.h:58
4-bit signed integer type
Definition: integer_subbyte.h:42
Functor performing linear scaling operations used by epilogues. Values are clamped before converting ...
Defines common types used for all GEMM-like operators.
typename platform::conditional< (platform::is_same< ElementA, int8_t >::value||platform::is_same< ElementA, int4b_t >::value||platform::is_same< ElementA, uint8_t >::value||platform::is_same< ElementA, uint4b_t >::value), arch::OpMultiplyAddSaturate, arch::OpMultiplyAdd >::type Operator
Definition: default_gemm_configuration.h:214
arch::OpMultiplyAddSaturate Operator
Definition: default_gemm_configuration.h:396
Templates exposing architecture support for multiply-add operations.
arch::OpMultiplyAdd Operator
Definition: default_gemm_configuration.h:144
Functor performing linear combination operations used by epilogues.
Defines the size of an element in bits.
Definition: numeric_types.h:42
arch::OpMultiplyAdd Operator
Definition: default_gemm_configuration.h:113
arch::OpMultiplyAddSaturate Operator
Definition: default_gemm_configuration.h:266
Top-level include for all CUTLASS numeric types.
Shape of a matrix multiply-add operation.
Definition: include/cutlass/gemm/gemm.h:57
arch::OpMultiplyAdd Operator
Definition: default_gemm_configuration.h:177
Definition: default_gemm_configuration.h:57
Defines tags for architecture-specific configurations.
Templates exposing architecture support for warp matrix multiply-add (WMMA) operations.
arch::OpMultiplyAddSaturate Operator
Definition: default_gemm_configuration.h:318
Basic include for CUTLASS.
arch::OpMultiplyAddSaturate Operator
Definition: default_gemm_configuration.h:344
arch::OpMultiplyAddSaturate Operator
Definition: default_gemm_configuration.h:370
arch::OpMultiplyAdd Operator
Definition: default_gemm_configuration.h:89