Does anybody have any real life numbers of the latency and/or throughput associated with moving large amounts of data using the IDMA channels rather than direct memcpy()? Some information online suggests that the IDMA is noticeably slower than the CPU memcpy() and that it does significantly impact the other CPM functionality. I guess the MPC860 was really bad in this regard but the MPC82xx was supposed to be much better. Is anybody using IDMA on their systems (besides the PCI9 workaround?)
FWIW, I am moving around 1 to 4MB/s of data from an FPGA to local memory on a custom MPC8260 board. Thanks, Jeff Angielski The PTR Group
