Projects for GSoC
Sergey
kornburn at yandex.ru
Mon Jan 20 11:02:17 UTC 2025
On Monday, 20 January 2025 at 08:45:03 UTC, RazvanN wrote:
> Hello everyone,
>
> We are now working on applying to GSoC
Just some foreign comment from community-member.
Of course I don't know how to do better and more properly, but
decided just share my 5cents on potential projects:
Check projects of other langs from the recent years. Try to
identify some patterns. Try to understand/hack the logic of the
GSoC committee - what they prefer.
From my own (very limited) analysis, I would assume GSoC
interested more in interesting, innovative, fresh and hot
(khm-ML/AI-khm) topics, that could be potentially useful for wide
audience of open source/science/cutting-edge technologies
participants.
Remove old list of "abandoned" (this list doesn't look very well)
project ideas and create new fresh list with updated and
confirmed mentors.
Google most probably is not interested in internal things of D
compiler and infrastructure itself. So maybe consider projects
that use D (compiler/std lib/ecosystem) as a tool that can bring
value for wider number of users.
Also those projects should be interesting for students and have
some potential application. Maybe forget more about individual
compiler stuff - and try to think about more general stuff.
Let me give some examples:
Fixing bugs / docs inside D compiler - bad. Relevant only for D -
bad as well. No one else care about that (outside very limited D
community).
Fix some code for GC in D - bad, but try to implement new
approach in GC research inside D compiler, that in case good
results could be implemented in other solutions - good.
Maybe focus more on ecosystem and interop, than internal only-D
relevant tasks.
For example, make good and "official" JVM interop module (based
on previous work and great JNI library from ARSD) - could be also
good, because it gives some potential for Android (Google
product) ecosystem.
Some langs in 2023/2024 made some numerical/AI
libraries/projects, but since Ilya left D community maybe DLF
will need to find new mentors for Mir/GPU/DlangScience projects..
Maybe revised PyD update and code improvement could be an example
of good project (Google loves Python as well - and similar
projects of Python interop were in the past)
Try to point out on strong parts of D, but keep in mind and be
careful with complexity estimation.
Ability to "write fast, read fast, and run fast" - is good. So
maybe consider to think about the project that 1 student could
make in some limited time, but the task/project should not be
"too hard". Maybe have several levels of complexity in the list.
Consider to have popular projects/sub-eco-systems in the list (to
show the audience) - like Inochi2D, Auburn Sounds, Vibe.d (many
repo orgs with a lot of stars in GH)
General CS tasks could be also interesting. Like "general
research about ranges" could be potentially helpful for C++ and
Carbon (and now even for Go).
But lack of proper data structures in the lang is limiting this
topic.
Maybe try to think in a product way as well.
Web: any web standards Google can like. Oauth library, async HTTP
client, WASM support, WebRTC/WebGL, HTTP2/3..
Google has several pictures libs: maybe JPEG XL improvements for
Gamut. Or WebP support..
Google is doing ARM chips(Axion): so port of DMD to ARM(with
description of fast compilation step) and maybe addition of
NEON/SVE2 instructions to the intel-intrinsics package projects..
Google is doing quantum chips(Willow): so the project of Timon of
lang for quantum computing, implemented in D..
Maybe something for ChomeOS: like flawless Dart interop, where
Dart is doing interface part and D backend/computations..
More information about the Digitalmars-d
mailing list