Mikael Brockman <mikael / phubuh.org> writes:

> I wrote a silly benchmark to test the speed of callcc as the stack
> grows.
> 
> The graphs[1] suggests that callcc performance and stack speed aren't
> linearly proportional.  If that's right, why?  I thought the only
> time-consuming part of callcc was the stack copy.  Maybe the stack copy
> isn't linear?  If so, why?
> 
> 1: http://www.phubuh.org/callcc-graph.png for 0..1000,
>    http://www.phubuh.org/callcc-graph-from-100.png for 100..1000

Duh!  I was plotting the rate of callcc calls, instead of the speed.  Of
course, the latter *is* linear.