TELEMAC-MASCARET Forum: Using graphical card processors as parallelization method (GPU) (1/1)

Welcome, Guest

TOPIC: Using graphical card processors as parallelization method (GPU)

Using graphical card processors as parallelization method (GPU) 10 years 10 months ago #11502

pilou1253
OFFLINE
openTELEMAC Guru
Posts: 584
Thank you received: 106

Hi all!

I've seen that that DHI (MIKE) has implemented a new tool enabling to use the cores of a high performance graphical card (GPU) for their 2D hydraulic module.

This seems to be very efficient, with speed-up factors up to ca 60. Furthermore, this method seems very attractive on an economical point of view compared to HPC since the type of graphical card needed costs around $1000.

Is it something that could be implemented within telemac in the future?

Thanks in advance and happy new year to everyone!

Regards
PL

The administrator has disabled public write access.

Using graphical card processors as parallelization method (GPU) 10 years 10 months ago #11521

jmhervouet

Hello,

Yes we are thinking of that. However so far you need to rewrite your code into a different language and we are reluctant to do so and to become dependent on proprietary software or machines. Preliminary tests show that the speed-up is not 60 but rather 10 for our applications. Most probably all the computers will have such processors in the future, hopefully without the current drawbacks. We remain convinced that parallelism has a far better speed-up potential.

With best regards,

Jean-Michel Hervouet

The administrator has disabled public write access.