mastodon
4.6.5
Hacking kernels…
Weekend TODO:
- Catch up reading ECE 108 course notes (~50 pages)
- ditto for ECE 140 course notes (~100 pages)
- Review MATH 119 course notes (5 pages)
- Do more investigation on the ULE scheduler
- Finish BSDCan slides (should’ve finished this last week)
Gotta be a busy weekend…
A new coop student at the FreeBSD Foundation is working on porting ROCm to FreeBSD.
We know that anything that involves the word “VM” (whether it’s virtual memory or virtual machine) is a fun thing to deal with!
I might transfer from computer engineering to mathematical physics (BSc) at uWaterloo. Currently 70% of chance doing it.
After starting second term, I realized that programming as hobby and learning it for grades aren’t the same.
No way Substack added code block right after I moved to Ghost.
I came to wrong university program.
I thought Computer Engineering will be full of fun courses, but the reality is that they are just boring. Now looking back at the curriculum, I find that Mathematical Physics (BSc) is more interesting than CE.
Computer Engineering has tuition of $19K/year. Mathematical Physics only has $9/year. This is ridiculous.
At this point, I would transfer to Mathematical Physics as a regular program. Why not co-op option? Most of the co-op postings on WaterlooWorks aren’t interesting. 70% of them are web frontend work and 90% of them are replaceable by AI. Moreover, as I continue exploring different fields of computing, I found that my only fields of interests are operating systems and microarchitecture. Since job postings for those are mostly external, there is no benefit of using WaterlooWorks to me.
Whether I transfer or not will be decided in a few months, but it’s clear at this point that Computer Engineering as a university program is not for me.
Great article on performance optimization on Swift programs!
I don’t think the Swift language is necessarily slow compared to C (as the article says, it does better than C compiler for SIMD optimization), but it wants programmers to use advanced data structures such as inline array to match or beat C’s performance.
https://www.cocoawithlove.com/blog/matrix-multiplications-swift.html
I really wanted to finish Chapter 4 (Processor design) before leaving for university... but time's up! GG
Swift is the language that I had least issues so far:
Javascript: cjs/ejs, node/browser, npm/pnpm/yarn/bun, nvm/fnm, webpack/vite, pure JS/JSDoc/TS, and much more. Sometimes dev tools become bottlenecks, not the actual code.
Python: Lack of good type checking. Mypy doesn’t always work when combined with other libraries. Performance issue due to GIL. pip3/poetry/uv, setup.py/hatchling/uv-build, black/flake8/ruff, etc. Again, too many dev tools out there.
The issue with JS/Python dev tools is that I need to install multiple programs to do the same job depending on the project I’m working on. If a new project uses different devtools, I need to spend extra time to install and learn them.
Golang: It’s bit off from traditional C-style syntax so I was never motivated to use it. But it’s loved a lot by network and container developers.
Rust: I like the idea, but the learning curve is steep and that’s kinda intimidating. Also I think it gets too much hype than it deserves.
Zig: Still in development. It might be good for embedded programming, but I find that code gets too long in some cases like I/O. Like Rust, I think it’s getting too much hype right now. I would give it a few more years to see how it ends up in 1.0.
C/C++: Header files, macros, and lack of dependency management. For build systems though, I think CMake/Ninja are doing well.
I’m not saying Swift is perfect. It has problems like lack of support for latest release from VSCode extension, broken compiler type check, not enough support for non-Darwin systems, and lack of good toolchain bootstrapping on Linux. When we go further up to macOS/iOS programming, there is Xcode SwiftUI preview crashing all the time.
But when it comes to enjoying the programming itself, Swift is the best language to me. Not only the language design is clean, I never had issues with dev tools as well because there is no dev tools hell like JS and Python.
This is why I would like to see something like FreeBSDKit by Project5BSD more in future.
One downside of Chapter 3 from CS:APP is that it’s based on AMD64 assembly. I know it’s the best ISA for hands-on experience and thus suitable for labs, but the book wastes so many pages on explaining AMD64’s special cases.
For example, any operation that touches lower 32-bit (l) of a register sets upper 32-bits to zeros, and the readers have to keep that on mind while reading the book.
My personal preference would be RISC-V, but it wouldn’t be practical for a course that targets not only CS/CE students but also students from other faculties who wants to take an introductory CS course.
Currently reading Computer Systems: A Programmer's Perspective (aka CS:APP).
Let's see how far I can get before school starts!
Just wrapped up my first co-op term as a Junior Software Developer Intern at the FreeBSD Foundation! Over the past four months, I worked on HMP scheduling (with a patch for the hmp(4) framework that I'll be presenting at BSDCan 2026) and brought LLDB kernel debugging much closer to feature parity with KGDB across FreeBSD's supported architectures — work that will ship in LLVM 23.
Huge thanks to everyone at the Foundation, the FreeBSD Project, and the LLVM Project for their support, mentorship, and thorough code reviews throughout the term. Read the full recap on my blog through the link below.
My work on LLDB’s FreeBSDKernel plugin:
https://minsoo.io/future-of-the-freebsd-kernel-lldb-plugin/
The name FreeBSDKernel will be changed to FreeBSD-core, since it only provides functionalities to examine core dumps. Traditional live kernel debugging is done through gdb-remote plugin with FreeBSD’s gdb stub.