Jarle Brinchmann writes:
>
> Hi Chris,
>
> It seems sufficient (to me) to merely change
> pdl_grow in pdlhash.c.  This at least seems to
> work for me now with up to 4Gb pdls (after which
> my machine grinds to a halt).

At the least one would need to change pdl_grow()
and pdl_malloc() as far as the allocation goes.
Then the pdl structure would need to have nvals
declared as STRLEN to support possible char
piddles of large size.

> I merely changed the int's to STRLENs and put a
> couple of casts in. Seems ok but I haven't had
> a chance to test it extensively as my computer
> has 4Gb memory and I have other things to do :)

I know the feeling.

> However PDL passes all tests with this small
> modification and since the modification is all
> within one subroutine I believe it should be fine.

Unfortunately, I'm certain that the pdl test suite
does not include any tests for large data sizes
(otherwise this would have been reported as a
fail already).

> PS: I am not checking this into the repository for
> now - it needs some independent testing.

Definitely.  We don't want to have PDL silently
given bad results for large data sizes.  That is
where it would be the hardest to detect as a
user.

Cheers,
Chris
_______________________________________________
Perldl mailing list
[email protected]
http://mailman.jach.hawaii.edu/mailman/listinfo/perldl

Reply via email to