DSP Libraries and Graphical Tools Accelerate Distributed Corner Turn Development
Virtually every radar application requires many FFTs on a set of radar pulse returns. The FFTs are normally performed in fast time dimension (row data) or in the slow time dimension (column data). However, mathematical operations performed on column data are normally very slow as column data is not contiguous in memory. This means that streaming accesses are not available, and SDRAM pages may need to be opened and closed; it is therefore faster to perform a matrix transpose or “corner turn” on the column data then process the slow time dimension FFTs on rows of data. This matrix transpose operation is extremely straightforward if the FFT processing is performed with a single embedded processor. However, real-time constraints usually mandate the use of multi-computers where the data flow and processing are much more complicated. This white paper will show how DSP libraries and graphical tools accelerate this distributed corner turn development.