Real Floating-Point Filter - 2022.2 English

Vitis Tutorials: AI Engine Development

Document ID
XD100
Release Date
2022-12-01
Version
2022.2 English

In the example, the filter has 16 coefficients which do not fit within a 256-bit register. The register must be updated in the middle of the computation.

For data storage a small 512-bit register is used. It is decomposed in two 256-bit parts: W0, W1.

  • First iteration

    • Part W0 is loaded with first 8 samples (0…7)

    • Part W1 with the next 8 samples (8…15)

    • Part W0 with the following ones (16…23)

  • Second iteration

    • Part W0 : 8…15

    • Part W1 : 16…23

    • Part W0 : 24…31