gma2 at CORNELL.EDU (Gregory M. Acland) wrote:
>The pedigrees involved are canine, and do not have huge numbers of
>individuals. Probably somewhere in the order of 30 to 50 individuals per
>family. The loops are unavoidable, however, as they are prexisting rather
>than planned pedigrees. Something like 6 to 10 loops per pedigree is
>typical.
>Does changing Maxloops to something like 10 increase the compute time
>drastically for pedigrees of this sort of size?
I don't think changing maxloop itself involves much (if any) cost -
it's how many loops there actually are in the pedigrees which matter.
6-10 loops could produce a _major_ slowdown.
>I am currently running the Linkage package on a DOS machine (100 Mhz 486,
>16 MB RAM), and don't have ready access to a Unix system.
>What source programs do I need to compile to yield a version of the Linkage
>package customized to deal with this many loops. What are the current
>versions that should be used? Is Borland Turbo/Pascal still the compiler of
>choice? Are there other compile options that should be considered at the
>same time?
>Alternatively, can Fastlink deal with loops at all, and if so, can it deal
>with this many loops. If so, can I obtain a DOS-executable version with
>the capacity to handle this many loops?
Use FASTLINK, which can certainly handle loops. As the source is C you
can recompile with gcc which is free and uses all available memory
(i.e. is not limited to 640K). I still suspect speed considerations
may render including all these loops impractical.
Dave Curtis (dcurtis at hgmp.mrc.ac.uk)
Institute of Psychiatry, London
http://www.iop.bpmf.ac.uk/home/depts/psychmed/general/dcurtis/dcurtis.htm
