The speed tests are a great idea and creating a larger dataset for faster machines was well-warranted. However, I'd noticed in the old dataset that the EPI's byte order was listed as being hard-coded for MSB_FIRST. Would it not make more sense to code them as CPU_NATIVE? We don't care about the actual results and if the CPU is LSB it must 2swap on the fly. Or does the timing routine factor this out?
Craig