| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Active Research
Forum: Africa Rainfall Project
Thread: The Weather Company is using GPUs to predict localized rainfall, can we? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 7
|
|
| Author |
|
|
Dayle Diamond
Senior Cruncher Joined: Jan 31, 2013 Post Count: 452 Status: Offline Project Badges: 
|
https://developer.nvidia.com/blog/new-gpu-acc...tically-improves-accuracy
These GPUs are offering hourly forecasts on 3 square kilometer blocks of land in the developing world. We're using data from The Weather Company, to do something similar to TWC's efforts around the world, but we're only running on CPUs. The programming standard they mention, OpenACC, is already compatible with Intel, AMD and NVIDIA graphics cards. https://en.wikipedia.org/wiki/OpenACC Perhaps the Africa Rainfall Project should be the next GPU-enabled project. |
|||
|
||||
|
caitilarkin
Former World Community Grid Admin USA Joined: Nov 4, 2015 Post Count: 331 Status: Offline Project Badges: 
|
Hi, Dayle Diamond. Here's the Q & A from our Help section regarding enabling GPUs for World Community Grid projects.
Will you make GPU processing available for other research projects? Not all algorithms are suited to general purpose computing on GPUs and may not receive performance improvements. Additionally, general purpose computing on a GPU is a complex process, requiring numerous modifications and optimizations to the underlying algorithms of an application to realize performance increases. We will work with the research scientists to make projects available on GPU if feasible and applicable to advancing the science application. However, it is up to the research scientists to enable their application to utilize GPU technology. |
||
|
|
Dayle Diamond
Senior Cruncher Joined: Jan 31, 2013 Post Count: 452 Status: Offline Project Badges:
|
Yeah, not all algorithms are suitable for GPU computing.
But this project uses data that NVIDIA specifically says is compatible with GPU computing. So isn't it worth asking the research scientists to take another look at the algorithm before the first experiment completes in 2023? |
||
|
|
caitilarkin
Former World Community Grid Admin USA Joined: Nov 4, 2015 Post Count: 331 Status: Offline Project Badges:
|
The Africa Rainfall Project runs on a platform called Weather Research and Forecasting (WRF). Because WRF was not originally created for distributed computing, the World Community Grid tech team (a small group) had to create a new version. This took quite a bit of time just to create and test for CPU.
Creating a new version that would use GPU power is a decision to be made by the project's scientists (also a small team ), who are concentrating on data analysis, data storage, and data sharing. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Since ARP uses the WRF program, they would have to completely scrap what they are doing and redesign using the other program. All of that would probably take as long as sticking with the current process. Find a GPU version of WRF and they may be willing to look at it but even that would be a long shot.
|
||
|
|
Dayle Diamond
Senior Cruncher Joined: Jan 31, 2013 Post Count: 452 Status: Offline Project Badges:
|
Okay Entity, challenge accepted!
----------------------------------------It looks like this company in Colorado has successfully run WRF work on GPUs. https://www2.cisl.ucar.edu/sites/default/file...0WRF%20on%20the%20GPU.pdf Their speedups aren't as dramatic as with OpenPandemics (5x to 7x faster), but it still seems promising to me. It's true that coming up with a modified, distributed GPU version would be highly time consuming, but it could also allow for much more efficient follow-up research if the ARP researchers want to run similar experiments with different parameters. [Edit 1 times, last edit by Dayle Diamond at Apr 9, 2021 7:18:04 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Excellent! Your points are well taken but it isn't going to be worth it for this particular project. Maybe a follow-on project perhaps.
|
||
|
|
|