<font size=5 color=blueyellow>**GPU Programming Workshop -- Planning Meeting**</font>

**Contents of this documents and quicklinks**：
[TOC]

## The 1st planning meeting (June 5)

### Links for the last workshop in 2023
- Lesson material: https://enccs.github.io/gpu-programming/
- Agenda for workshop in 2023: https://hackmd.io/@yonglei/gpu-programming-2023-schedule

### Event page
- Event page: https://hackmd.io/@yonglei/gpu-programming-2024-event

### Date

Last year it was a two-day workshop, this year we prefer it to be a three-day workshop (Tue-Thu or Wed-Fri, 09:00-16:00)

| Week | Date | Poll |
| :--: | :----------: | :------: |
| 38 | Sept. 16-20 | - |
| 44 | Oct.28-Nov.1 | -oo |
| 46 | Nov. 11-15 | ooooo |
| 48 | Nov. 25-29 | oo |
| 50 | Dec. 9-13 | oooo |
| 51 | Dec. 16-20 | ooo |

**==Nov. 12-14 (09:00-16:00 CET) for workshop==**

* [LT] Stepas is available from week 46 (Nov 11).
* [CSC] We are organizing a on-site training at CSC on November 27-29 (80-90 % sure), with backup dates around Dec 9-13.
* [SC] Supercomputing is Nov 17-22
* [CR] CodeRefinery workshop Sep 10-12 & 17-19.

### Instructors and helpers
- Yonglei, LiU/ENCCS
- Thor, RISE/ENCCS
- Stepas, VU/NCC LT
- WeiLi, Liu/ENCCS
- Hicham, UiB/NRIS
- Cristian, CSC
- Dhanya (NRIS/UiB)
- Jonathan, PDC/KTH
- QiangLi, Liu/ENCCS
- Andreea, RoNCC

### HPC resources
- Last year we used LUMI for hand-on exercises
- This year (hope we can renew training projects on LUMI and Leonardo Booster)
- ==AMD GPU on LUMI==
- Nvidia GPU on Leonardo Booster
- At least demonstrations on Booster can be made available by those instructors who have initiative to show Nvidia usage

### Target number of participants
- ==60==

### Lesson Materials and Agenda
- Materials: https://enccs.github.io/gpu-programming/
- Agenda for workshop in 2023
- https://hackmd.io/@yonglei/gpu-programming-2023-schedule

| Time | Contents | (Two) Instructors |
| :----: | :----------: | :-----------: |
| 15 min | Welcome | Yonglei, Thor |
| 20 min | Why GPUs? | |
| 30 min | GPU hardware and software ecosystem | |
| 20 min | What problems fit to GPU? | |
| 40 min | GPU programming concepts | |
| 40 min | Introduction to GPU programming models | |
| | |
| 60 min | Directive-based models (OpenACC, OpenMP) | , (Stepas) |
| 70 min | Portable kernel-based models (Kokkos, OpenCL, SYCL, C++ stdpar, etc.) | |
| | |
| 75 min | Non-portable kernel-based models (CUDA, HIP) | |
| 60 min | Non-portable kernel-based models (Exercises) | |
| | |
| 50 min | High-level language support | Yonglei, |
| 50 min | Multi-GPU programming with MPI | |
| 40 min | Recommendations and discussions | |
| | |
| 60 min | Preparing code for GPU porting | |
| 90 min | Problem examples | Stepas, |
| 10 min | Wrap-up | |

**new:**

* Hip python: https://rocm.docs.amd.com/projects/hip-python/en/latest/
- work on Nvidia (Leonardo booster)?

### Updating lesson materials
- Updating materials?
- before next meeting, we all try to spend some time on reviewing the material and raise issues on GitHub for any suggestions and new ideas
- users feedback from workshop last year.
- More hand-on exercises?
- exercises for Julia and Python from the webinar `Practical intro to GPU programming in Python and Julia`
- Shall we include a **`code challenge`** session on the third day afternoon?
- A (**mentimeter**) quiz in the last 10 minutes for the morning/afternoon session?
- example: https://www.mentimeter.com/app/presentation/al11qosot4j1juefzawfoij1hcf7ipua/oi79r6wmw63o
- frequency?
- A cheat sheet for this workshop?
- Yup!

### Misc
- Shall we include writing the best-practice-guide as part of the workshop preparations?
- Should we include some specific general usecases?
- i think more use cases would be useful
- Doodle for next meeting (Aug. 12-31).

---
---

## The 2nd planning meeting (Aug. xx) -- Lesson Materials

:::danger
:::