Template Parameters - 2023.1 English

The following table lists parameters for the FIR filters on AIE-ML.

Table 53 Parameters Supported by FIR Filters

Parameter Name	Type	Description	Range
TT_DATA	Typename	Data Type	int16, cint16
TT_COEFF	Typename	Coefficient type	int16 cint16
TP_FIR_LEN	Unsigned int	The number of taps	Min - 4, Max - see Maximum FIR Length
TP_SHIFT	Unsigned int	The number of bits to shift unscaled result down by before output.	0 to 61
TP_RND	Unsigned int	Round mode	0 = rnd_floor truncate or floor 1 = rnd_ceil ceiling 2 = rnd_sym_floor floor to zero 3 = rnd_sym_ceil ceil to infinity 8 = rnd_neg_inf round to negative infinity 9 = rnd_pos_inf round to positive infinity 10 = rnd_sym_zero symmetrical to zero 11 = rnd_sym_inf symmetrical to infinity 12 = rnd_conv_even convergent to even 13 = rnd_conv_odd convergent to odd
TP_INPUT_WINDOW_VSIZE	Unsigned int	The number of samples processed by the graph in a single iteration run. For windows, defines the size of input window. For streams, it impacts the number of input samples operated on in a single iteration of the kernel.	Must be a multiple of the 256-bits In addition, must by divisible by: TP_SSR and TP_DECIMATE_FACTOR. No enforced range, but large windows will result in mapper errors due to excessive RAM use, for windowed API implementations.
TP_CASC_LEN	Unsigned int	The number of cascaded kernels to use for this FIR.	1 to 9. Defaults to 1 if not set.
TP_USE_COEFF_RELOAD	Unsigned int	Enable reloadable coefficient feature. An additional ‘coeff’ port will appear on the graph.	0 (no reload), 1 (use reloads). Defaults to 0 if not set.
TP_API	Unsigned int	I/O interface port type	0 = Window 1 = Stream

For a list of template parameters for each FIR variant, see API Reference Overview.

TP_CASC_LEN describes the number of AIE processors to split the operation over, which allows resources to be traded for higher performance. TP_CASC_LEN must be in the range 1 (default) to 9. FIR graph instance creates TP_CASC_LEN kernels. Computation workload of the FIR (defined by its length parameter TP_FIR_LEN) is divided and each kernel in the graph is then assigned a fraction of the workload, i.e. each kernel performs TP_FIR_LEN / TP_CASC_LEN. Kernels are connected with cascade ports, which pass partial accumulation products downstream until last kernel in chain produces the output.

TP_USE_COEFF_RELOAD allows the user to select if runtime coefficient reloading should be used. When defining the parameter:

• 0 = static coefficients, defined in filter constructor

• 1 = reloadable coefficients, passed as argument to runtime function.

  Note

  When used, port port<input> coeff; will be added to the FIR.