nelsons%psu-cs@uunet.UU.NET (Shannon Nelson) (06/02/88)
I'm doing some basic research for myself, and thought I'd ask the
opinion of 'those in the know'. I know the basics of what shared
memory and message passing architectures are, and I am somewhat
familiar with how they are used, tho' I've had little experience
in programming actual machines.
My questions, then are:
In Your Opinion,
1. Which architecture is easier to program for, and why;
2. Does #1 depend on the application, and why;
3. Who makes the "best" machine of the two architectures, and why.
I realize that there are no cut and dried answers for the above
questions, but I would appriecate your thoughts and experiences.
Send your opinions to me, of course, and I will collect, edit and
summarize.
AdvTHANKSance,
Shannon Nelson
..!tektronix!{psu-cs,reed,ogcvax}!qiclab!nelsons
(503) 643-2214fpst@hubcap.clemson.edu (Steve Stevenson-Moderator) (06/02/88)
Approved: parallel@hubcap.clemson.edu Shannon Nelson <qiclab.UUCP!nelsons%psu-cs@uunet.UU.NET> > 1. Which architecture is easier to program for, and why; This is not a simple question since most folks do not have enough experience on both. Each has obvious advantages, but most of the criteria are performance ones. The operating system folks have the most experience and they probably can't tell you either. Languages are a problem here too. Read Greg Andrews article in the Jan 88 TOPLAS. > 2. Does #1 depend on the application, and why; Undoubtedly. Example: 'cubes can be added to ( FPS T-series goes from 8 to 4096 processors.) If the system is programmed distributively, then needs for memory (for e.g.) are easily met (Ah, hello Turing) by your company's comptroller. Not everything distributes easily - but we don't have the experience yet. > 3. Who makes the "best" machine of the two architectures, and why. I'll take the Fifth Amendment. But I personally think that this is not answerable without - pardon my math background - cost criteria for "best". Steve Stevenson fpst@hubcap.clemson.edu (aka D. E. Stevenson), fpst@clemson.csnet Department of Computer Science, comp.parallel Clemson University, Clemson, SC 29634-1906 (803)656-5880.mabell