Читать книгу Multi-Processor System-on-Chip 2 - Liliana Andrade - Страница 23

1.3. GFDM algorithm breakdown

Оглавление

Knowledge of algorithm kernel14 requirements with respect to workload is key to determining whether or not the vDSP has enough available compute resources to execute the kernel. This section covers the GFDM algorithm and its processing graph extrapolated pseudo-code.

GFDM processing in the literature is divided into three categories:

1 1) frequency-domain processing (Michailow et al. 2014);

2 2) time-domain processing (Farhang et al. 2016; Matthé et al. 2016);

3 3) Radix-2 cascade processing (Nimr et al. 2018).

We assume (2) time-domain processing for implementation on the vDSP, primarily due to its flexibility and low compute complexity per data point on small workloads.

Multi-Processor System-on-Chip 2

Подняться наверх