Opened 4 years ago
Closed 2 years ago
#2368 closed Task (fixed)
HPC-10_mcastril_HPDAonlineDiagGPU
| Reported by: | mcastril | Owned by: | mcastril |
|---|---|---|---|
| Priority: | low | Milestone: | Unscheduled |
| Component: | DIA | Version: | trunk |
| Severity: | minor | Keywords: | diagnostics, GPU, offload, performance, HPC, |
| Cc: |
Description
High performance data analytics solutions aiming at tackling the online diagnostics of the NEMO model will be explored as complementary components in the model diagnostics software eco-system. Online techniques leveraging fast (low latency and real-time) data analytics approaches (e.g. on fat nodes) will be evaluated in real cluster environments. In particular, an interface of NEMO to the High Performance Data Analitics (HPDA) framework will be designed and implemented for online diagnostics.
The rationale of this activity is to improve the NEMO computational performance by executing the computations for diagnostics on GPU.
Workplan action
Commit History (3)
| Changeset | Author | Time | ChangeLog |
|---|---|---|---|
| 14690 | jchanut | 2021-04-12T09:53:52+02:00 | #2368, Add AGRIF zoom coincident with North and South boundaries in python script |
| 14675 | ldebreu | 2021-04-06T14:44:31+02:00 | AGFdomcfg: corrections for east-west cyclic grids, #2368 |
| 14674 | ldebreu | 2021-04-06T14:36:08+02:00 | AGFdomcfg: corrections for east-west cyclic grids, #2368 |
Change History (6)
comment:1 Changed 2 years ago by ldebreu
comment:2 Changed 2 years ago by ldebreu
In 14675:
comment:3 Changed 2 years ago by jchanut
In 14690:
comment:4 Changed 2 years ago by nemo
- Milestone changed from IMMERSE 2020 to Unscheduled
comment:5 Changed 2 years ago by mcastril
- wp_comment changed from As first step, the portability of NEMO diagnostic calculations to GPUs has been analyzed, exploring how to adapt these regions from the current MPI implementation to the CUDA paradigm. A toy model has been created to perform preliminary tests, that were done using the dia_hsb diagnostic. The code itself is executed 50x faster than in a single CPU but the data transfer to and from GPU is the main bottleneck. We are working on the asynchronous strategy in order to hide all communications among GPU/CPU. We also plan to increase the efficiency of the overall solution, by mitigating the impact of the offloaded data and extending our approach to the rest of the diagnostics. to As first step, the portability of NEMO diagnostic calculations to GPUs has been analyzed, exploring how to adapt these regions from the current MPI implementation to the CUDA paradigm. A toy model has been created to perform preliminary tests, that were done using the dia_hsb diagnostic. The code itself is executed 50x faster than in a single CPU but the data transfer to and from GPU is the main bottleneck.We are working on the asynchronous strategy in order to hide all communications among GPU/CPU. We also plan to increase the efficiency of the overall solution, by mitigating the impact of the offloaded data and extending our approach to the rest of the diagnostics.
To follow on #2662
comment:6 Changed 2 years ago by mcastril
- Resolution set to fixed
- Status changed from new to closed
In 14674: