On Thu, 4 Dec 2003, McIlwrath, BK (Brian) wrote:
> Peter W. Draper wrote on 04 December 2003 13:11:
>
> > Brian's changes to HDS have never made it out of RAL, so only
> > he can let us know what state that work is in (the issue was
> > to do with the size of the pointers stored in an HDS file,
> > these limit offsets within an HDS file to around 500Mb, with
> > your full featured NDFs I expect you'd run into this problem
> > quite quickly).
>
> The current status is that I redesigned the HDS file format to cope with
> 64-bit addressing and I have the C routines which interface to this format.
> It then got bound with an effort to develop a "pure C" HDS - with BOTH 32
> and 64-bit FORTRAN wrapper layer.
>
> No-one seemed very enthusiastic about me finishing or releasing this (until
> now!) - but I need to get back to it I see!
>
> It would be be quicker (and safer) to use the new file format, back out the
> "pure C" HDS changes and keep the existing FORTRAN top-level (with INTEGER*8
> where appropriate) as a first step.
Hi Brian,
that sounds like it would be a useful first step to allow the rest of us
to see what the real issues are (although a pure C interface is still very
very tempting, especially when I look at JNIHDS). When you say passing out
INTEGER*8, I presume you mean addresses as well as your new _LONGINT data
type. In that case how does this tie into the CNF_PVAL arrangements, or
have these been superceded by this work?
Cheers,
Peter.
|