Hacker News Reader

Comment by pixelesque | Hacker News Reader

I had DeepSeek 4 Pro do a very good job yesterday of loop-unrolling and SIMD-ifying (both SSE/AVX and ARM Neon versions) some very old scalar Col3f image processing and resampling/resizing C++ code I hadn't touched in 12 years or so.

It also wrote some unit tests that validated the kernel sampling weights, and wrote some Jupyter notebooks to go along with the kernel algorithms as comparisons.

It's not just web dev... It helps (a lot in some cases) if you ask very specific things rather than just "make this vague thing", but I'm more and more coming round to the conclusion it is now a useful dev tool (until two months ago I was a sceptic).

View on HN

> image processing and resampling/resizing C++ code I hadn't touched in 12 years or so.

So an unimportant, personal project. People seem to extrapolate being able to do something cool into being able to do useful work, which is what this whole discussion is about.

It's an example of useful work. 11 years ago, I would have been able to use that code in production for the company I worked at at the time, and the conversion would have been very useful.

I'm also using Github Copilot at work, but I can't as easily talk about what I'm doing there.

What do you consider useful work?

This is still rather basic stuff to be absolutely honest with you.

The tests and notebooks maybe: not the unrolling and re-architecting to use masking SIMD-ification. I used to do SIMD stuff a lot over the last 17 years, and it would have taken me probably a day to do that (convert 7 algorithms, all with support for various wrapping/mirroring/clamping extent-handling, so around 800 raw source file lines of templated C++) to use SIMD - and well.

Then it probably would have taken me at least half a day to think about what tests to write and to make them.

It did it in around 5 minutes. I had to check it, and it wasn't perfect, but it compiled, ran, and was very close.

Sorry, but this is just not a complex or hard task in image processing. Image resizing is a staple algorithm as common as bread and butter.

You could convert ~800 lines of scalar float processing templated C++ code that used mixins for wrapping logic and float/double to well-optimised SIMD code for two different ISAs (without using wrapper class helpers, it wrote the full intrinsics) in under 5 minutes and write tests for them at the same time?

It's the time compression I'm impressed with.

This is a sifferent point than what the thread started with (speed vs. correct output for hard problems). Yours stands, but it changed the direction of the discussion as I understand it. Hence the pushback.

Yeah I expect it to be helpful in some basic image processing stuff, but it will quickly fail once you try to implement research papers on the topic.

I can specifically and directly state that is false.

Feel free to share specifics.

Reply on news.ycombinator.com