[comp.parallel] Parallel Benchmarking Programs

steve@mullian.ee.mu.OZ.AU (Steve Mabbs) (05/31/90)

Can anyone provide pointers to benchmarking software for parallel
systems?  I've written a tool called 'ptrace' for the Sequent Symmetry
which single-steps through processes (watching creation of children
and also tracing them), reporting statistics on possible memory contention,
etc.  What I need are some programs which are considered to be psuedo-
standard tests  e.g. is there a parallel grep, tar or dhrystone for the
Sequent?

Thanks,

	Steve

---------------
Stephen A. Mabbs, Dept of Electrical and Electronic Engineering
University of Melbourne		Phone:	office    (room 3.10): +61 3 344 6714
PARKVILLE VIC 3052			vlsi lab  (room 3.16): +61 3 344 7436
AUSTRALIA			Fax: +61 3 344 6678

peter@sigma21.oz (Peter Farrell) (06/05/90)

In article <9168@hubcap.clemson.edu>, steve@mullian.ee.mu.OZ.AU (Steve Mabbs) writes:
> Can anyone provide pointers to benchmarking software for parallel
> systems?  I've written a tool called 'ptrace' for the Sequent Symmetry
> which single-steps through processes (watching creation of children
> and also tracing them), reporting statistics on possible memory contention,
> etc.  What I need are some programs which are considered to be psuedo-
> standard tests  e.g. is there a parallel grep, tar or dhrystone for the
> Sequent?
> 

You might consider a few of the standard packages, eg sh, make and dump.
These are definitely parallelised. I think tar is also. If you have
access to one of the FORTRAN compilers you could try some linpack
runs. These have been parallelised using the fortran pre-processor,
though I don't know how this would fit in with your ptrace fn. 

Several years ago, I did a quick hack on dhrystone to parallelise it
but it was very crude and followed no standards.

I hope this helps.