We are performing small fft using DMA in pipelined mode.
accel_cfft_small_pipe(fft_device, 1.0, n_points);
We are using 32 Byte aligned INPUT and OUTPUT buffer in SDRAM.
FFTA engine gets stuck after few iteration (N_POINTS 512 complex_float). Sometimes it executes 10000 iterations and sometime only 50 iterations.
When we run the same program for 256 complex N_POINTS, it is running smoothly for long durations lakhs of iteration tested.
What could be the reason behind this w.r.t. N_POINTS configuration.
Is there any memory limitation on FFTA for small pipelined FFT?
What all things are to be taken care while input/output buffer and FFTA's DMA config?
Moving to CrossCore Embedded Studio and Add-ins
Can you please try "ComplexFFTPipelined" example in the SC5xx BSP package.
You can download "ADSP-SC5XX Board Support Package" from the below link www.analog.com/.../EVAL-ADSP-SC589.html
After installation, "ComplexFFTPipelined" example available in the below path:<Installation path>\ADSP-SC5xx_EZ-KIT_Lite-Rel2.0.2\ADSP-SC5xx_EZ-KIT\Examples\drivers\fft
As you are facing issues when INPUT and OUTPUT buffer in SDRAM, can you please let us know if you are facing the same issue when buffers are located in Internal L1 memory.