For the Matrix Multiply (GeMM) library element the list of configurable parameters and default values is presented below.
Name | Type | Default | Description |
---|---|---|---|
T_DATA_A | typename | cint16 | Input A Data Type. |
T_DATA_B | typename | cint16 | Input B Data Type. |
P_DIM_A | unsigned | 16 | Input A Dimension |
P_DIM_AB | unsigned | 16 | Input AB Common Dimension. |
P_DIM_B | unsigned | 16 | Input B Dimension. |
SHIFT | unsigned | 20 | Acc results shift down value. |
ROUND_MODE | unsigned | 0 | Rounding mode. |
P_CASC_LEN | unsigned | 1 | Cascade length. |
P_DIM_A_LEADING | unsigned | 0 | ROW_MAJOR = 0 COL_MAJOR = 1 |
P_DIM_B_LEADING | unsigned | 1 | ROW_MAJOR = 0 COL_MAJOR = 1 |
P_DIM_OUT_LEADING | unsigned | 0 | ROW_MAJOR = 0 COL_MAJOR = 1 |
P_ADD_TILING_A | unsigned | 1 | no additional tiling kernel = 0 add additional tiling kernel = 1 |
P_ADD_TILING_B | unsigned | 1 | no additional tiling kernel = 0 add additional tiling kernel = 1 |
P_ADD_DETILING_OUT | unsigned | 1 | no additional detiling kernel = 0 add additional detiling kernel = 1 |
NITER | unsigned | 16 | Number of iterations to execute. |
DIFF_TOLERANCE | unsigned | 0 | Tolerance value when comparing output sample with reference model, e.g. 0.0025 for floats and cfloats. |
STIM_TYPE_A | unsigned | 0 | Supported types: 0 - random 3 - impulse 4 - all ones 5 - incrementing pattern 6 - sym incrementing pattern 8 - sine wave |
STIM_TYPE_B | unsigned | 0 | Supported types: 0 - random 3 - impulse 4 - all ones 5 - incrementing pattern 6 - sym incrementing pattern 8 - sine wave |
Note
The above configurable parameters range may exceed a library element’s maximum supported range, in which case the compilation will end with a static_assert error informing about the exceeded range.