1 response

  1. cnxsoft
    December 16, 2013

    I’ve asked some details about the throughput between the host and the FPGA board:

    * Raspberry Pi: “Max theoretical bandwidth for the logi-pi is 3.8MB/s using our protocol over SPI to address the wishbone bus,(16 bit overhead for each transfer). In Practice we got ~3MB/s in user space.”
    * BeagleBone Black: “We have a theoretical bandwidth of 76MB/s and measured up to 69MB/s in user space for transfer of 128KB blocks. Our kernel driver gives access to the whole address space of the FPGA with EDMA but can be re-written to be application specific and get better bandwidth. For now the overhead for switching from user-space to kernel space impacts a lot the performance, thus writing an application specific kernel driver (ours uses EDMA and burst access) with interrupt for transfer generated by the FPGA could get you close to theoretical bandwidth. “

Leave a Reply




Back to top
mobile desktop