Vadim V. Zhytnikov wrote:
> Valery Pipin writes:
>
> The symptoms you describe is not about speed it is about RAM.
That's right. However the amount of RAM is always finite.
The test shows that I can tune the gcl to calculate the specific task with
the best performance. However this tuning is not universal. SBCL has a bit
lesser speed, however it is more predictable.
> I'm almost 100% sure that gcl runs out of physical memory
> on you tasks. The point is that gcl 2.6 is _indeed_ less
> memory-efficient than sbcl, cmucl and clisp.
> gcl 2.6 has 3-word cons cells while other lisps have
> 2-word cons cells. Not yet released gcl 2.7
> eliminates this this gcl's drawback.
Very good! Is it possible to compile maxima with gcl-cvs?