The data types of input ports A and B (inA and inB) are controlled by TT_DATA_A
and TT_DATA_B
respectively.
Both inputs may take one of the 6 choices: int16, int32, cint16, cint32, float and cfloat. It must be kept in mind that depending on the input type combinations, output type will be determined by the IP.
Please see table Hadamard Supported Combinations of Input/Output data types: for allowed input data type combinations and resultant output type. In addition, the vector size granularity is given for the data type combinations. TP_DIM
should be a multiple of the listed vector size.
Input Data Type A | Input Data Type B | Output Data Type | Vector Size (window/stream) |
---|---|---|---|
int16 | int16 | int16 | 16 / 8 |
int16 | int32 | int32 | 8 / 8 |
int16 | cint16 | cint16 | 8 / 8 |
int16 | cint32 | cint32 | 8 / 8 |
int32 | int16 | int32 | 8 / 8 |
int32 | int32 | int32 | 8 / 4 |
int32 | cint16 | cint32 * | 4 / 4 |
int32 | cint32 | cint32 | 4 / 4 |
cint16 | int16 | cint16 | 8 / 8 |
cint16 | int32 | cint32 * | 4 / 4 |
cint16 | cint16 | cint16 | 8 / 4 |
cint16 | cint32 | cint32 | 4 / 4 |
cint32 | int16 | cint32 | 8 / 8 |
cint32 | int32 | cint32 | 4 / 4 |
cint32 | cint16 | cint32 | 4 / 4 |
cint32 | cint32 | cint32 | 4 / 2 |
float | float | float | 8 / 4 |
float | cfloat | cfloat | 4 / 4 |
cfloat | float | cfloat | 4 / 4 |
cfloat | cfloat | cfloat | 4 / 2 |
Note
- Type combination is not supported by AIE-ML device.