Add GPU GSoC project (#2221)

Also slightly improve the "All projects" page, which rendered badly.

jsoc/allprojects.md

@@ -1,7 +1,5 @@
-## View all GSoC/JSoC Projects
+# View all GSoC/JSoC Projects
 
 This page is designed to improve discoverability of projects. You can, for example, search this page for specific keywords and find all of the relevant projects.
 
-## Projects
-
 {{ all_gsoc_projects }}

jsoc/gsoc/MLJ.md

@@ -1,13 +1,12 @@
-# MLJ.jl Projects – Summer of Code
+@def mintoclevel = 2
+
+# MLJ.jl Projects - Summer of Code
 
 [MLJ](https://github.com/alan-turing-institute/MLJ.jl) is a machine
 learning framework for Julia aiming to provide a convenient way to use
 and combine a multitude of tools and models available in the Julia
 ML/Stats ecosystem.
 
-
-### List of projects
-
 MLJ is released under the MIT license.
 
 \toc

jsoc/gsoc/compiler.md

@@ -73,20 +73,3 @@ BoF calendar invite][threadcall] on the Julia Language Public Events calendar.
 
 **Recommended Skills**: Varies by project, but generally some multi-threading and C experience is needed\
 **Contact:** [Jameson Nash](https://github.com/vtjnash)
-
-
-## Automation of testing / performance benchmarking (350 hours)
-
-The Nanosoldier.jl project (and related <https://github.com/JuliaCI/BaseBenchmarks.jl>) tests for
-performance impacts of some changes. However, there remains many areas that are not covered (such as
-compile time) while other areas are over-covered (greatly increasing the duration of the test for no
-benefit) and some tests may not be configured appropriately for statistical power. Furthermore, the
-current reports are very primitive and can only do a basic pair-wise comparison, while graphs and
-other interactive tooling would be more valuable. Thus, there would be many great projects for a
-summer contributor to tackle here!
-
-**Expected Outcomes**: Improvement of Julia's automated testing/benchmarking framework.
-**Skills**: Interest in and/or experience with CI systems.
-**Difficulty**: Medium
-
-**Contact:** [Jameson Nash](https://github.com/vtjnash), [Tim Besard](https://github.com/maleadt)

jsoc/gsoc/contractionorder.md

@@ -1,4 +1,4 @@
-# Tensor network contraction order optimization and visualization
+# Tensor network contraction order optimization and visualization – Summer of Code
 
 [OMEinsum.jl](https://github.com/under-Peter/OMEinsum.jl) is a pure Julia package for tensor network computation,
  which has been used in various projects, including

jsoc/gsoc/documenter.md

@@ -1,4 +1,4 @@
-# Documentation tooling
+# Documentation tooling – Summer of Code
 
 ## Documenter.jl
 

jsoc/gsoc/gpu.md

@@ -0,0 +1,28 @@
+# GPU Projects - Summer of Code
+
+[JuliaGPU](https://github.com/JuliaGPU) provides a suite of packages for programming GPUs in Julia. We have support for AMD, NVIDIA and Intel GPUs through various backends, unified by high-level array abstractions and a common programming model based on kernel programming.
+
+## Improving GPU Stack Portability
+
+**Difficulty:** Medium
+
+**Duration:** 175 or 350 hours (the scope of functionality to port can be adjusted accordingly)
+
+**Description:** The Julia GPU stack consists of several layers, from low-level vendor-specific packages like CUDA.jl to high-level abstractions like GPUArrays.jl. While the high-level packages aim to be vendor-agnostic, many optimized operations are still implemented in vendor-specific ways. This project aims to improve portability by moving these implementations to GPUArrays.jl using KernelAbstractions.jl.
+
+The project will involve:
+- Identifying vendor-specific kernel implementations in packages like CUDA.jl
+- Porting these kernels to KernelAbstractions.jl in GPUArrays.jl
+- Improving KernelAbstractions.jl where needed to support these kernels
+- Ensuring performance remains competitive with vendor-specific implementations
+- Adding tests to verify correctness across different GPU backends
+
+**Required Skills:**
+- Experience with Julia programming
+- Familiarity with GPU programming concepts
+- Experience with GPU programming in Julia is a plus
+- Understanding of performance optimization
+
+**Expected Results:** A set of optimized GPU kernels in GPUArrays.jl that are vendor-agnostic and performant across different GPU backends. This will improve the portability of the Julia GPU stack and make it easier to support new GPU architectures.
+
+**Mentors:** [Tim Besard](https://github.com/maleadt), [Valentin Churavy](https://github.com/vchuravy)

jsoc/gsoc/graphics.md

@@ -3,7 +3,7 @@
 Removed reference to this file on the main projects page for summer 2021 since they weren't updated.
 -->
 
-#  Graphic Projects – Summer of Code
+# Graphic Projects – Summer of Code
 
 ## Makie
 

jsoc/gsoc/juliaconstraints.md

@@ -1,4 +1,4 @@
-# Constraint Programming in Julia
+# Constraint Programming in Julia – Summer of Code
 
 [JuliaConstraints](https://juliaconstraints.github.io/) is an organization supporting packages for Constraint Programming in Julia. Although it is independent of it, it aims for a tight integration with JuMP.jl over time. For a detailed overview of basic Constraint Programming in Julia, please have a look at our video from JuliaCon 2021 [Put some constraints into your life with JuliaCon(straints)](https://youtu.be/G4siuvNMj0c).
 

jsoc/gsoc/juliagenai.md

@@ -1,4 +1,4 @@
-# JuliaGenAI Projects
+# JuliaGenAI Projects – Summer of Code
 
 ![JuliaGenAI Logo](https://github.com/JuliaGenAI/juliagenai.org/blob/main/assets/logos/logo-256.png?raw=true)
 
@@ -68,7 +68,7 @@ Julia stands out as a high-performance language that's essential yet underrepres
 **Project Goals and Deliverables:**
 
 1. **Knowledge Base Expansion:** Grow the AIHelpMe.jl knowledge base to include comprehensive, up-to-date resources from critical Julia ecosystems such as the Julia documentation site, DataFrames, Makie, Plots/StatsPlots, the Tidier-verse, SciML, and more. See [Github Issue](https://github.com/svilupp/AIHelpMe.jl/issues/3) for more details. This expansion is crucial for enriching the context and accuracy of AI-generated responses related to Julia programming.
-   
+
 2. **Performance Tuning:** Achieve at least a 10% improvement in accuracy and relevance on a golden Q&A dataset, refining the AIHelpMe.jl Q&A pipeline for enhanced performance.
 
 ### Project 4: Enhancing Julia's AI Ecosystem with ColBERT v2 for Efficient Document Retrieval
@@ -147,7 +147,7 @@ As a pivotal resource for the Julia community, the [Julia LLM Leaderboard](https
 
 ### Project 7: Counterfactuals for LLMs (*Model Explainability* and *Generative AI*)
 
-**Project Overview:** This project aims to extend the functionality of [CounterfactualExplanations.jl](https://github.com/JuliaTrustworthyAI/CounterfactualExplanations.jl) to Large Language Models (LLMs). As a backbone for this, support for computing feature attributions for LLMs will also need to be implemented. The project will contribute to both [Taija](https://github.com/JuliaTrustworthyAI) and [JuliaGenAI](https://github.com/JuliaGenAI). 
+**Project Overview:** This project aims to extend the functionality of [CounterfactualExplanations.jl](https://github.com/JuliaTrustworthyAI/CounterfactualExplanations.jl) to Large Language Models (LLMs). As a backbone for this, support for computing feature attributions for LLMs will also need to be implemented. The project will contribute to both [Taija](https://github.com/JuliaTrustworthyAI) and [JuliaGenAI](https://github.com/JuliaGenAI).
 
 **Mentor:** [Jan Siml](https://github.com/svilupp) (JuliaGenAI) and [Patrick Altmeyer](https://github.com/pat-alt) (Taija)
 
@@ -170,6 +170,6 @@ As a pivotal resource for the Julia community, the [Julia LLM Leaderboard](https
 
 We'd love to hear your ideas and discuss potential projects with you.
 
-Probably the easiest way is to join our [JuliaLang Slack](https://julialang.org/slack/) and join the `#generative-ai` channel. 
+Probably the easiest way is to join our [JuliaLang Slack](https://julialang.org/slack/) and join the `#generative-ai` channel.
 You can also reach out to us on [Julia Zulip](https://julialang.zulipchat.com/#narrow/stream/423470-generative-ai) or post a GitHub Issue on our website [JuliaGenAI](https://github.com/JuliaGenAI/juliagenai.org).