home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.parallel
- Path: sparky!uunet!destroyer!gatech!hubcap!fpst
- From: cedelle@yin.irisa.fr (Alain Cedell)
- Subject: MPfortran performance question.
- Message-ID: <22JAN199315132248@yin.irisa.fr>
- News-Software: VAX/VMS VNEWS 1.41
- Sender: news@irisa.fr
- Organization: IRISA, Rennes (Fr)
- Date: 22 Jan 1993 15:13 -0700
- Approved: parallel@hubcap.clemson.edu
- Lines: 18
-
- I would be interested to share experience with MPFortran users, especially
- about this point of vector/array performance:
- I'm currently using MPFortran on a DECMPP-12000 (1K processors), and i meet
- many performance problems: The shape of arrays seem affecting the computing
- time. For instance the statement :
- Sb = SUM(Tb,MASK=Mb .eq. 1)
-
- Or even the statement :
- Tb = Mb + 1
-
- Are much more quickly evaluated when Tb and Mb are vectors (array(1:900)), than
- when they have 2 dimensions (array(1:3,1:300)). This for same number of elements
-
- Is this a general rule to always prefer vectors than multi dimension arrays ?
-
- a.c.
- cedelle@irisa.fr
-
-