Re: Slowup in 2.1.54

[Paul Thomas]

John,

I think that you are right about this being an allocator effect.   Up to 
precisely 8192 words, the growth in time taken by the little code is 
linear.  After that, it goes over abruptly to quadratic time dependence. 
The former indicates that the system is able to to reuse the previous 
allocation and then, beyond 64kbyte, x has to be moved, which gives the 
quadratice dependence.

Paul

John W. Eaton wrote:

> On 19-Feb-2004, Paul Thomas <address@hidden> wrote:
>
> | Octave does this sometimes; eg. j = 
> | 1e4; tic;x = []; for i = 1:j; x(i) = i; endfor; toc  changes its spots 
> | altogether when j>1e4 (I haven't checked in the source but I presume 
> | that a reserve of about 8e4 bytes is being allocated?).
>
> No, there is no reserve allocation.  I think in this case, performance
> degrades rapidly because you are repeatedly allocating and copying an
> array that includes just one more element each time.  Doing that 10000
> times is bad, and it may also be that none of the previous allocations
> can be reused, so that it ends up requiring a lot more than 80000
> bytes in the end.  (The released chunks of memory are not the right
> size unless the allocator is able to combine the them into something
> that is useful, and there is no guarantee that it can.)
>
> jwe
>
>
>