# GPU Unit Trimester Meeting

## Agenda:

- Major engineering tasks for upcoming 3 months (Tour de Table):
- how many man-hours do you plan for each? {50, 100, 150, 200, 250, 300}
- Rommel:
- sea-ice thd performance improvements: 100+
- integration of sea-ice thd: 50-
- porting the ICE modules that remain between the modules that have already been ported (adv,rdgrft,thd): 100+
- Okke:
- Christian: First, I believe this meeting is useful, and even leaving BSC I would like to assist online. Could it be possible? If so, the tasks are:
- Classes: 100
- Research (UAB-Growth collaboration with UAB groups): 100
- Research2 (BSC-Continue paper CAMP): 100
- Andrew:
- CAMP test suite: 50
- Solver updates/changes: 100+
- reformatting: 100+

- Retrospective review of major engineering tasks for past 3 months (Tour de Table):
-
- Rommel:
- The plan was:
- integration of new rdgrft: 50-
- sea-ice thd: managed memory: 100+
- sea-ice thd: separate memory: 100+
- Okke:
- The plan was:
- EDITO: porting into GPU: 1) icedyn_adv (pra); 2) LBC; 100h
- EERIE: 50+
- Christian:
- The plan was:
- Validation tests: MONARCH-CAMP: documentation 50-;
- Validation tests: MONARCH-CAMP: discussion of test scope: 50
- Validation tests: MONARCH-CAMP: learn AUTOMONARCH validation routine and adapt to: 50+
- Andrew:
- The plan was:
- Aerosol unit tests: 50+
- Testing and debugging of aerosols on GPU: 50
- Test automation: 100+
- what was real time for previously planned tasks?

- Rommel:
- integration of new rdgrft: 50-
- sea-ice thd: managed memory: 100-
- sea-ice thd: separate memory: 50+
- GPU porting tooling: 50+
- Christian:
- Documentation 50.
- Test scope: 50.
- learn AUTOMONARCH validation routine: 0.
- But not fully integrated validation tests, just documented how to replicate
- Andrew:
- CAMP developments (including aerosol tests, but also test-suite redesign): 100
- MONARCH (advection optimisation, running Christian's CAMP-MONARCH tests): 50-
- Other (testing camp with single CPU process per GPU, running physics code): 50--
- why? what brings us back? what can we improve?
- Christian:
- why: Emerged unexpected higuer priority project (GANANA) and requirement of transfer knowledge due to leaving BSC.
- Andrew:
- Spent about 2 weeks working on a last-minute BLUEFUSE project proposal
- Still waiting for viable test cases for most of the aerosol reactions
- Task priority changing quite frequently
- Need to meet with Oriol (AC) more often

- Personal development:
- what each engineer thinks he has to learn to improve performance?
- how we can implement this?
- free discussion on courses/materials/books recommendations

- Tools to learn:
- pick one tool and do a quick overview:
- (???) Gitlab runners? (Christian comment: No progress on the BSC side, so nothing to comment)

- Papers and talks:
- which of complete tasks contribute to papers?
- (Okke) DEODE can lead to energy profiling paper
- (Christian) GPU MONARCH-CAMP will contribute to a paper
- (Alexey) Challenges in ocean model porting to GPUs
- (Andrew) Currently writing a paper with Hannah on a methodology for scientific software modernisation
- ...
- what experience we may convert to talks?
- (Andrew) Planning to give a department talk on collaborative development and Git, need to talk with DDT