cem@intelca.UUCP (Chuck McManis) (05/01/84)
How about this, say three bench mark algorithims were selected as a "standard" and sent to all manufacturers of microprocessors to hand over to their favorite hacker to hand-code in assembly code. In addition each manufacturer was allowed how come up with their own benchmark algorithim to give to the others to implement similarly. Then all of these benchmarks are run on the individual systems (paying close attention to memory speed, wait states, and the possibility that the average user would really spend $12,000 to implement a 20 processor parallel pipeline that runs at 30MHz and uses 40ns CMOS memories :-) ) Then compare the results from the Start to the Halt instruction in terms of speed. And maybe we could come up with a benchmark that everyone would believe. Naw, they'd never go for it. --Chuck