[m-dev.] Re: benchmark for trailing & coroutining

[Peter Schachte]

On Fri, 26 Sep 1997, Fergus Henderson wrote:

> This was on kryten, using `-O6 --intermodule-optimization'.
> 
> program 	grade		user  sys   real    %cpu	slowdown vs asm_fast
> ------- 	-----           ---   ---   ----    ----
> cqueens	asm_fast	0.01u 0.05s 0:00.10 60.0%	 1.0
> cqueens	asm_fast.tr	0.01u 0.05s 0:00.10 60.0%	 1.0	~0
> cqueens	asm_fast.gc	0.06u 0.21s 0:00.33 81.8%	 3.30	230%
> cqueens	asm_fast.gc.tr	0.12u 0.20s 0:00.34 94.1%	 3.40	230% + ~10%

Interesting.  Any idea why adding .tr to asm_fast adds no overhead, while
adding .tr to asm_fast.gc doubles user time (but still has little effect on
real time)? 

Maybe you should try a bit larger N, maybe 12 or 14, to get larger numbers? 
I'm always suspicious of benchmark numbers where the differences are a few
100ths of a second.

-Peter Schachte      URL:  http://www.cs.mu.oz.au/~pets/
pets at cs.mu.OZ.AU     PGP:  finger pets at 128.250.37.150 for key
    Do insects spend hours demammaling their programs?