vifl@hou2f.UUCP (M.MEKETON) (06/01/84)
I have run some of the benchmarks published by The Computer Center on APL*PLUS/PC w/o 8087 and compared them to their results. In general they are comparable, however I now believe they used version 2.6 of APL*PLUS/PC while I used version 3.0, which did speed up things, especially in matrix divide. My guess is that different timers will get different results, and may also cause some variance in the benchmark. My timer looped N times (N is typically 100), performed a 'quad'WA (which forces a gargabe collection), does a timestamp, executes the expression, does another timestamp, calculates the differences, and loops back. The answer reported was the average time. One of the points of the benchmark was to show the significance of the 8087 chip. Creative Computing has been publishing some benchmarks on accuracy of floating point operations, and computers that used the 8087 (such as the IBM PC and Grid) scored the best in terms of accuracy. While the test was terrible (it only tested square roots and squaring), it does hint at the power of the 8087. Too bad that the 68000 doesn't support the 8087 (except in the IBM XT/370, where they used modified 68000 & 8087 chips). Together, they could make a powerful combination. Marc S. Meketon hou2f!vifl