Not sure if this is possible - but I hesitate to attach my quad\'s since it will tie them up for a few months. If CPDN was able to make the app multithreaded, I\'d love to download one WU and have it done in 25% of the time.

Is that something BOINC has to add support for first or something climateprediction could do?

-Sangamon

---

Team Starfire World BOINC

Both.
But it\'s unlikely to happen for several reasons.

However, there is an option in General preferences, (in your Account page on the server), to set the number of processors used.
So you could set it for just 1 or 2, and this is how many models would run.

And BOINC is designed to allow other programs to use the resouces as needed, although this is dependenat on having sufficient ram so that the swap file isn\'t constantly being used, which slows down the response time of BOINC/science apps to a request for the resources.

I am only doing one CPDN unit on my dual processor HT (4 processing threads), while the other 3 threads stay busy with other projects.

---

Both. But it\'s unlikely to happen for several reasons.

That\'s too bad. Hope it happens sooner than you anticipate. Thanks for the response.

-S

Both.
But it\'s unlikely to happen for several reasons.

However, there is an option in General preferences, (in your Account page on the server), to set the number of processors used.
So you could set it for just 1 or 2, and this is how many models would run.

And BOINC is designed to allow other programs to use the resouces as needed, although this is dependenat on having sufficient ram so that the swap file isn\'t constantly being used, which slows down the response time of BOINC/science apps to a request for the resources.

That is unfortunate. With so many of the new processors having multiple cores, the possibilities for more complex models would open up. Is the problem just a lack of programmers, or is there a research-design reason why this can\'t be done?

Both.
The programs are the ones used by the UK Met Office for weather/climate modelling/forecasting. Over a million lines of Fortran, developed over several decades by numerous people to run a supercomputers.

The OS used by desktop computers has a totally different \'philosophy\' to that of supercomputers, even though both have multiple processors.

The models are divided into \'cells\' vertically and horizontally, and the starting point of one, depends on the calculated results of the previous.

The first 4 sections in the science pages here give more detail.

Not sure if this is possible - but I hesitate to attach my quad\'s since it will tie them up for a few months. If CPDN was able to make the app multithreaded, I\'d love to download one WU and have it done in 25% of the time.
...

-Sangamon

I\'m assuming you have a Q6600 or QX6700 - the Core2 should be able to do it in around 2 months. If you don\'t want to tie up the entire machine, why not have one or two cores work on CPDN while the others work on Rosetta?

---

I'm a volunteer and my views are my own.
News and Announcements and FAQ

Both.
The programs are the ones used by the UK Met Office for weather/climate modelling/forecasting. Over a million lines of Fortran, developed over several decades by numerous people to run [on] supercomputers.

The OS used by desktop computers has a totally different \'philosophy\' to that of supercomputers, even though both have multiple processors.

I\'m not sure how many of us crunching volunteers appreciate how complex a million lines of Fortran are. They were phasing out Fortran when I got my computer science degreee; I learned Pascal and C instead. Yes, I agree that supercomputer architectures can be quite different than x86 standards. Getting CPDN to run on WIN32 was an impressive feat.

I suppose in the long-term, being able to multi-thread such a large simulation would be of considerable gain to the researchers. There have been some remarkable advances in Beowolf clusters, Intel\'s new \"tile\" design, and the prevalence of multi-core CPU\'s on the desktop, just to name a few. This is the future, and I hope that other experiments can muster the time and resources to build their simulations to use this new hardware.

I just purchased dual Athlon MP 2600 CPUs, so I\'m hoping to see a vast improvement in the speed once I install the new chips. The large cache (512KB) per chip should also help. I read a technical paper on the Athlon MP chips, and I really like what they\'ve done with TLB and MOESI.