segall@paul.rutgers.edu (Ed Segall) (09/13/90)
Upon rereading my posting from yesterday ("Request for IPSC
Applications or Benchmarks"), I realize that my wording was slightly
ambiguous. I am not looking for Linda programs. Rather, I need
programs that perform reasonably on the IPSC/2, regardless of what
language they are written in.
Actually, I would appreciate any application performance information
you can supply, even if it comes without code, and even if it doesn't
seem "reasonable".
The original message:
> Request for IPSC applications/benchmarks
>
>
> Hello,
>
> I am doing some development work in Linda on an IPSC/2. In order to
> objectively evaluate the performance of programs I write, I need to
> compare them with programs that are generally accepted to have good
> performance on this machine.
>
> Does anyone out there have a program that performs well (not
> necessarily outstandingly) for its application area and for this
> machine? If it's not too large, I would like to rewrite it in Linda
> and compare performance. I need either source code that I can compile
> (this is preferable, so that I can make sure I'm comparing apples to
> apples), or performance information along with an algorithm, with
> enough information that I can reproduce the algorithm in Linda and
> extrapolate the performance results to whatever machine configuation I
> run it on.
>
> Lacking either of the above, I would appreciate at least minimal
> information, such as "We achieved XXX MFLOPS on a 32 processor machine
> without vector units, doing Gaussian Elimination with partial
> pivoting", so that I can at least make a meaningful comparison with
> something I write.
Thank you,
Edward Segall
--
uucp: ...{anywhere}!rutgers!caip.rutgers.edu!segall
arpa: segall@caip.rutgers.edu