[parsec-users] nThreads in Swaption

Jim Dempsey jim at quickthreadprogramming.com
Mon Mar 29 11:25:26 EDT 2010


Don't you mean

  nThreads = nSwaptions;

I think the test data files use small numbers of items (Swaptions in this
case). And in which case, on large systems, you may have more threads than
pieces for the threads to work on.

In the fluidanimate program you will find other restrictions such as number
of threads being power of 2, and which presents a problem on 3 and 6 core
systems. BTW, the QuickThread implementation of fluidanimate is not
sensitive to power of 2.

Jim 

-----Original Message-----
From: parsec-users-bounces at lists.cs.princeton.edu
[mailto:parsec-users-bounces at lists.cs.princeton.edu] On Behalf Of Paul Keir
Sent: Monday, March 29, 2010 9:57 AM
To: parsec-users at lists.cs.princeton.edu
Subject: [parsec-users] nThreads in Swaption

Hi all,

There's a check near the start of the Swaption benchmark (in
HJM_Securities.cpp) which changes the number if threads if there are two
many wrt. the number of swaptions. Now I don't know what a swaption is, but
if more swaptions means more work, shouldn't this be removed in a
benchmarking context? Here it is:

if(nSwaptions < nThreads) {
  nSwaptions = nThreads;
}

Cheers,
Graham


The University of Glasgow, charity number SC004401
_______________________________________________
parsec-users mailing list
parsec-users at lists.cs.princeton.edu
https://lists.cs.princeton.edu/mailman/listinfo/parsec-users



More information about the parsec-users mailing list