[racket] Writing parallel racket benchmarks

One question out of curiosity. In the parallel Mandelbrot benchmark on
the language shootout page, languages like C, C++, Ada, Java 7 have 100%
CPU utilization on all cores. Racket has 94% on all cores, whereas other
languages values are sometimes much lower and differ from core to core.

What influences these values? Why does Racket achieve only 94% usage in
a tight loop with futures? Don't get me wrong, 94% is close enough and
Racket generally fares very well in the shootout. I'm just wondering
where the 94 comes from. Do these numbers converge to 100% when the
benchmark is run longer?


Best,


Erich



On Sun, 2011-09-25 at 21:21 -0400, Sam Tobin-Hochstadt wrote:
> http://scriptstoprograms.wordpress.com/2011/09/25/benchmarketing/