[Newbie] Plot problem (Feature suggestion)

Hello, Ray!

I came to this idea while looking at gcc/as dos windows splashing all 
over my screen (I was running win98 that day). There's certanly a 
difference between multiple calls to integrand and multiple calls to 
integration routine. In plotting the second case occurs. Your tests were 
correct, except that you could also compare the total time of 
translation+compilation+integration. So, either a global switch or a 
compilation result caching will help. Maybe Stavros would clarify his 
idea? (Memoization etc?)

On the other thought, when one uses plotting of quad'ed function it 
means an integrand is a two-variabled function. When calling integration 
routine one of the function's arguments is bound to a value already, 
i.e. *integrands are all different across plotting* and I don't see what 
can be effectively cached here :(

--
Andrei Zorine

Raymond Toy wrote:
> 
> That would imply that quad_qagi would be faster without compiling than
> with.  I think it was done that way on the assumption that the
> integrand would be called many times to perform the integration, so
> the function is compiled.  If that's not really necessary, why not
> just remove the compilation completely?
> 
> I think I measured runtime with and without compiling, and compiling
> was faster.  But perhaps that was true for my tests but not in
> general.
> 
> Ray
> 
>