home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.transputer
- Path: sparky!uunet!inmos!titan.inmos.co.uk!news
- From: conor@lion.inmos.co.uk (Conor O'Neill)
- Subject: Re: On-chip RAM using D7205.- Help!
- Message-ID: <1993Jan7.121850.1489@titan.inmos.co.uk>
- Sender: news@titan.inmos.co.uk
- Organization: INMOS Limited, Bristol, UK
- References: <1993Jan06.190526.42203@dcc.uchile.cl>
- Date: Thu, 7 Jan 1993 12:18:50 GMT
- Lines: 37
-
- In article <1993Jan06.190526.42203@dcc.uchile.cl> maurro@inf.utfsm.cl (Mauricio Gomez Calderon) writes:
- >Hi...
- > In order to accelerate a numerical calculation, I'm trying to put
- > small code in the fast on-chip RAM, using
- >
- > SET processor (order.code := -1)
- > MAP logical.processor ONTO processor
- >
- > and the corresponding /RE option on the occonf.
- >
- > It works... but don't increases the speed!!!
- >
- > The small code is a small program that perform 1000000 iterations
- > changing only 5 REAL64 variables. This code is put into five transputers
- > connected like a ring. Communications take place only 3 times over the
- > network.
- >
- > What's wrong???
-
- Try looking at the file produced by the collector's -p option.
- This tells you where the code and data etc has been placed for each processor.
- This will show you exactly what effect the various changes you make are
- having to the memory layout.
- In general, it is better to place workspace on-chip, because transputers
- make more accesses to data than to code. This is why this is the
- default behaviour of the configurer.
-
- If your example really is as simple as you describe, you will probably
- find that both code and data are already on-chip, so changing their order
- will have no effect.
-
- Hope this helps.
-
- ---
- Conor O'Neill, Software Group, INMOS Ltd., UK.
- UK: conor@inmos.co.uk US: conor@inmos.com
- "It's state-of-the-art" "But it doesn't work!" "That is the state-of-the-art".
-