All FIR filters can be configured for various types of data and coefficients. These types can be int16, int32, or float, and also real or complex. Certain combinations of data and coefficient type are not supported.
The following table lists the supported combinations of data type and coefficient type with notes for those combinations not supported.
Data Type | |||||||
Int16 | Cint16 | Int32 | Cint32 | Float | Cfloat (note 4) | ||
Coefficient type | Int16 | Supported (note 3) | Supported | Supported | Supported | note 2 | note 2 |
Cint16 | note 1 | Supported | note 1 | Supported | note 2 | note 2 | |
Int32 | Supported | Supported | Supported | Supported | note 2 | note 2 | |
Cint32 | note 1 | Supported | note 1 | Supported | note 2 | note 2 | |
Float | note 2 | note 2 | note 2 | note 2 | Supported | Supported | |
Cfloat (note 4) | note 2 note 2 | note 2 note 2 | note 2 note 2 | note 2 note 2 | note 1 note 1 | Supported Supported | |
3. The rate-changing FIR variants, i.e. fir_decimate_asym, fir_decimate_sym, fir_interpolate_asym and fir_resampler only support int16 data and int16 coeff type combination on AIE-ML device. 4. Cfloat data type is not supported on AIE-ML device. |