kuter

I had the pleasure of working with Johannes as part of Google Summer of Code in 2020 and 2021 working on LLVM. He mentored me and many students over the years.

I remember in 2020 he would meet with us at 7AM his time to accommodate every student he was working in different parts of the world (Turkey, Greece, China, Japan iirc), I was the one from Turkey.

I had Coffee with him and his wife few months back. He was so full of life despite everything.

He touched the lives of many people. What a great life lived despite being so short.

Hello HN, I wrote a fuzzer that can generate ISA docs for NVIDIA GPU architectures.

Here are is docs for hopper(sm90a) https://kuterdinel.com/nv_isa/

and here is for 4090 (sm89) https://kuterdinel.com/nv_isa_sm89

I am here to answer your questions if you have any!

It looks like the future for laptops is ARM. I hope Lenovo makes a ThinkPad with this chip, there already is a ThinkPad with a Snapdragon SoC (ThinkPad X13s), but from what I have read it's not very good. I considered getting a MacBook to use with Asahi but some important features are still unsupported (like HDMI output support).

I did something similar for Instagram on android few years ago. The usual methods for bypassing certificate didn't work on Instagram, they were statically linking openssl into a shared library called libcoldstart.so. I Spent some time reading openssl documentation and ended up installing a hook to the function that configured the certificate verification.

In case you are curious. I used Frida for installing hooks to native functions.

The entire thing runs in the browser without needing a backend (unlike godbolt).

Fun challenge: Look at the source code and figure out how it works.

The deal breaker with obsidian for me is that it is not opensource. It's great that it's at least using a markdown derivative format.

I am using Emacs Org Mode and quite happy with it. You can link different files, include images, embed and view LaTeX, encrypt your notes with GPG and much more. I think it will stand the test of time better than Obsidian which is something I care a lot for note taking and journaling.

Few points:

- Not sure if people would want to "talk to" a device in public.

- Anything and more that this device do can be done by a smart phone that people already have.

- I feel like people prefer using a UI interface instead of speech.An example to this behavior might be using delivery apps instead of calling a restaurant.

- Looks too wide to fit into most pockets.

I wonder if the recent departure of Kyle Vogt from Cruise is related. Emmett Shear and Kyle Vogt founded Twitch together.

Took a peek at the models they use. It seems to be a vision transformer encoder decoder architecture with a resent backbone. Looks really good. I had a similar idea of training a model and making a desktop application, but haven't had the opportunity. I wonder how much compute it took to train the model.

I think this paper was the first one to do OCR on LaTeX: http://cs231n.stanford.edu/reports/2017/pdfs/815.pdf The paper describes an Encoder-Decoder architecture with CNN encoder and LSTM based decoder.

I am the author of the article, please let me know if you have questions.

Hello HN, I am graduating in a few months, last summer I was an SWE intern at Google before that I did a Google Summer of Code with the LLVM foundation working on the inter-procedural optimizations. I enjoy taking on deeply technical challenges. Although I have more professional experience towards low level programming. I enjoy experimenting with Deep Learning in my spare time. I am also very interested to learn more about finance when I have time.

Location: Turkey

Remote: Ok

Willing to relocate: Yes (Prefer to)

Technologies: C/C++, LLVM, Compilers, Python, Deep learning(Mainly Tensorflow but Torch is Ok), TS, JS, Rust(Intermediate).

Résumé/CV: Available upon request.

Email: kuterdinel@gmail.com

Blog: kuterdinel.com

This is kinda funny but I was going to try to get it merged to mainline but then realized there was already a similar patch waiting on review.

This my code, large part of it is based on other HP WMI stuff https://github.com/kuterd/hp_omen_led_controler

One project I made for myself was to to wtite a kernel module that could control the led backlight of my laptop keyboard.

I had to first reverse engineer the windows utility that was written in c# and found out that it used wmi commands. After reading the kernel code for a while I realized there where already some code that implemented HP WMI commands.

Using that as a foundation it was quite easy to write my own kernel module that could change the led colors. Then I tried to integrate it with the linux kernel's led API. Since the LEDs where RGB it was a little different to integrate but I managed to find some examples in some playstation joystick driver code.

This project was a lot of fun, but I would say that you should be careful not to break your hardware. I have broken devices even from the user space by pushing it beyond it's limits.

GCC has a book recommendation page [0] I had been using some of these books to self study.

[0] https://gcc.gnu.org/wiki/ListOfCompilerBooks

Frame Format

   The IP datagram is printed, on a small scroll of paper, in
   hexadecimal, with each octet separated by whitestuff and blackstuff.
   The scroll of paper is wrapped around one leg of the avian carrier.
   A band of duct tape is used to secure the datagram's edges.  The
   bandwidth is limited to the leg length.  The MTU is variable, and
   paradoxically, generally increases with increased carrier age.  A
   typical MTU is 256 milligrams.  Some datagram padding may be needed.

   Upon receipt, the duct tape is removed and the paper copy of the
   datagram is optically scanned into a electronically transmittable
   form.

I was expecting this to be a April Fools RFC [0]. I was disappointed when I checked the date.

[0]https://en.wikipedia.org/wiki/April_Fools%27_Day_Request_for...

naive investors ... probably.

For the purpose of one to many type of live streaming you would probably want to use HLS.

Twitch uses it's own transcoding system. Here is a interesting read from their engineering blog [0]

[0] https://blog.twitch.tv/en/2017/10/10/live-video-transmuxing-...

I don't think GDB is too bad. Most people probably don't use watch expressions, conditional breakpoints etc.

There was a incident in Turkey (31 march 2015) that caused a country wide blackout. From what I understand the west side and the east side loose sync. The west side had under supply and the east side had over supply.

Very interesting. Here are some papers that they have listed [0]. They are using computer vision to detect charged particles that hit camera image sensors. [0]https://wipac.wisc.edu/deco/research

This probably just splits B+ tree pages. I suggested something similar for using sqlite with static web pages see [0]

Distributed search engines are not new, many companies tried to make them [1].

I think the main challenges of distributed search engines are not technical. I don't think there isn't much financial incentive to create/maintain/seed them.

There isn't much advantage to using them either (if a alternative is available). I think the only viable use case for them is for indexing pirated torrents.

Maybe a foundation could use this type of technology to lower their expenses. Like a charity version of Google

[0]https://news.ycombinator.com/item?id=25843786

[1]https://en.wikipedia.org/wiki/Distributed_search_engine

Visually they look similar.

Some of the supported messaging apps doesn't allow automation or provide APIs. I appreciate the work that went into supporting those. But I don't think it would be possible to have consistent support for some of those chat apps. Things like API changes or just getting you server ips banned would disrupt service.

Reminds me of the Japanese art Kintsugi/kintsukuroi which is the name for repairing broken pottery with lacquer and gold silver or platinum powder.

https://en.wikipedia.org/wiki/Kintsugi

I haven't looked at the code but this sounds cool. There is also something called Kademlia which is a distributed hash table for keeping track of seeds for torrents. https://en.wikipedia.org/wiki/Kademlia

For anyone interested there is something called fuzzing that uses usually code coverage based heuristics to generate data to find bugs.

For example LLVM's lib fuzzer uses instrumentation to track code coverage and mutates data to find invalid behaviour.

https://llvm.org/docs/LibFuzzer.html

It uses a compiler pass to insert code to branch points functions calls etc. I think it uses genetic algorithms to increase coverage by changing the data.

There are others that work in similar ways one of them is. https://github.com/google/AFL

The clients would just send ajax queries with range headers to get the parts of the database that they want.

It is going to be read only.

Locking wouldn't be problem because in the eyes of the sqlite running in the browser it would have it's own read only copy of the database.