I think I pretty much have it. Buffer overruns are detected and handled. There are still some spikes, so it looks like there might still be some dropped data points. I am not sure how we can handle this. My only thought might be to either use pru1 to do the actual copying to ARM memory (which might be possible, since the pru can access ARM memory). Maybe using the 12K shared memory which would allow larger buffers. Another option would be a kernel driver, which would be able to access pru memory and ARM memory without the overhead of mmap.
There are sometimes PRU SPI timeouts. I don't know what causes that.
I think I pretty much have it. Buffer overruns are detected and handled. There are still some spikes, so it looks like there might still be some dropped data points. I am not sure how we can handle this. My only thought might be to either use pru1 to do the actual copying to ARM memory (which might be possible, since the pru can access ARM memory). Maybe using the 12K shared memory which would allow larger buffers. Another option would be a kernel driver, which would be able to access pru memory and ARM memory without the overhead of mmap.
There are sometimes PRU SPI timeouts. I don't know what causes that.