linda performance

[Haibin Jiu]

General question: Does anyone have a compilation on Linda's performance?
Esp. that of C-Linda? On any system...

Specific question: We seem to have been unable to run more than 4 or 5
C-Linda processes on a LAN of Sun SPARCstations? Even
the simplest programs -- some taken from a paper by
the guy who started Linda and some other dude on parallel
programming -- lock up if more than 4 Suns are involved.
Any comments?

Thanks for any useful info.

Bugs

[David Kaminsky]

ji...@husc10.harvard.edu (Haibin Jiu) writes:

> General question: Does anyone have a compilation on Linda's performance?
> Esp. that of C-Linda? On any system...

Numbers abound. "Experience with Linda" (Bjornson et. al.,
YALEU/DCS/TR-866) has data on codes doing matrix multiplication, LU
decomposition, parallel sequence comparison, fluid dynamics and more.
Linda tech reports can be obtained from Chris Hatchell
(hatc...@cs.yale.edu).

> Specific question: We seem to have been unable to run more than 4 or 5
> C-Linda processes on a LAN of Sun SPARCstations? Even
> the simplest programs -- some taken from a paper by
> the guy who started Linda and some other dude on parallel
> programming -- lock up if more than 4 Suns are involved.
> Any comments?

Without knowing more about your problem, there is not much I
can say other than to note that Linda programs routinely run on much
larger collections of machines.

BTW, with respect to a comment from an earlier posting, SCA's
next release (v2.4.7, in use now at Yale) supports multiple processes
per node.

David