andrew@alice.att.com (Andrew Hume) (12/26/90)
following suggestions by tve and kenner, i reran my benchmarks as a sequential stream on one cpu (cpu 7) and all other cpus were executing a busy loop (for(;;);). the results? the average spread decreased from .65% to .46% but more importantly, only one benchmark had a big spread (.91%). (previously, there had been 3: 1.26%, 1.01% and .83%.) i can live with this but frankly it is annoying that there is so much imprecision. andrew@research.att.com