Harlan is a high level language for general purpose GPU
Re: Harlan is a high level language for general purpose GPU
Unfortunately, no. RNA-World uses existing software with thousands of lines of codes. Rewriting this code in Harlan would require huge amounts of man power. Besides, it was shown for the current software using GPUs is not beneficial because of the limited CPU to GPU bandwidth. In the future we plan to add another software to RNA-World which will benefit from GPUs. Stay tuned.
- Michael H.W. Weber
- Vereinsvorstand
- Beiträge: 22435
- Registriert: 07.01.2002 01:00
- Wohnort: Marpurk
- Kontaktdaten:
Re: Harlan is a high level language for general purpose GPU
Thanks for the info/URL.
Michael.
Fördern, kooperieren und konstruieren statt fordern, konkurrieren und konsumieren.
http://signature.statseb.fr I: Kaputte Seite A
http://signature.statseb.fr II: Kaputte Seite B
http://signature.statseb.fr I: Kaputte Seite A
http://signature.statseb.fr II: Kaputte Seite B
Re: Harlan is a high level language for general purpose GPU
It seems that the best solution would be to update to Infernal 1.1rc4. The resulting hundred-fold performance increase would solve the problem with 1000+ hour tasks that never checkpoint and also result in a tremendous increase in the effective research capability of RNA World.
- Michael H.W. Weber
- Vereinsvorstand
- Beiträge: 22435
- Registriert: 07.01.2002 01:00
- Wohnort: Marpurk
- Kontaktdaten:
Re: Harlan is a high level language for general purpose GPU
Well, this is true only if you want to run the program in default mode. And default mode misses details.jon b. hat geschrieben:It seems that the best solution would be to update to Infernal 1.1rc4. The resulting hundred-fold performance increase would solve the problem with 1000+ hour tasks that never checkpoint and also result in a tremendous increase in the effective research capability of RNA World.
Even the program version we use currently only deploys tasks with extreme run times when we set out to go for these details.
But all the issues related to long run times will be history soon, because checkpointing tests are coming up this week.
Michael.
Fördern, kooperieren und konstruieren statt fordern, konkurrieren und konsumieren.
http://signature.statseb.fr I: Kaputte Seite A
http://signature.statseb.fr II: Kaputte Seite B
http://signature.statseb.fr I: Kaputte Seite A
http://signature.statseb.fr II: Kaputte Seite B