Parallel Computing Question

S. A. Modena samodena at csemail.cropsci.ncsu.edu
Sat Apr 17 05:24:16 EST 1993


In article <9304132221.AA23039 at pi18.arc.umn.edu> eichler at pi18.arc.umn.edu (Rogene Eichler) writes:

>>    1. What presently is available (or seen coming in the near
>>       future) on parallel machines that would aid molecular biology
>>       research?
>
>There is a program called BigNet written for the Cray, CM200, and CM5

Of course, there are other machines that might be considered....the
Maspar machine has an excellent $-per-CPU-minute ratio when kept busy on
very large problems....I don't happen to know it for the Convex products
but I'll bet it's good also.

>
>Much of the problem with parallel algorithms is that they are machine 
>specific. There have been attempts at standardization of the parallel
>languages (F90, Fortran D, CMFortran), but communication calls are 
>far from standard......
>.......
>In general, if you have a problem that requires a significant amount
>of interprocessor communication, a serial machine might give you better
>performance. 
>							-Rogene

Let's put this the other way around: if one has an algorithm which is
purely synchronous (each processor in the machine is executing the same
line of FORTRAN at the same moment), then you may be home free!  The
implementation of the algorithm on your machine-of-choice may be greatly
simplified (and I know of at least one major sequence classification
algorithm that falls into this category).  

So to expand Rogene's point, when you decide to implement *new* algorithms
(can't be a copy-cat forever), what new knowledge-skills-software tools
will be required...what is the learning curve....what will be the likely
structural nature of algorithms of interest to you and your
colleagues....is massively parallel supercomputing the most fruitful path
for your interests in computational molbio?

For some categories of algorithms, massively parallel is the way to go!
And for other categories, it may be the pits.  :^)  [ I'm thinking of the
contrast in two crop process modeling presentations a few years ago....the
guy that grew FORTRAN "supercomputed" corn had a gut-wrenching experience,
whereas the guy that threw away COMAX (FORTRAN Cotton Model) and grew cotton
in LISP from first principles thought it was a breeze to do. ]

Steve
---
+------------------------------------------------------------------+
|     In person:  Steve Modena     AB4EL                           |
|     On phone:   (919) 515-5328                                   |
|     At e-mail:  nmodena at unity.ncsu.edu                           | 
|                 samodena at csemail.cropsci.ncsu.edu                |
|                 [ either email address is read each day ]        |
|     By snail:   Crop Sci Dept, Box 7620, NCSU, Raleigh, NC 27695 |
+------------------------------------------------------------------+
         Lighten UP!  It's just a computer doing that to you.    (c)
OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO
         



More information about the Comp-bio mailing list