dnedic

How do you inspect what is happening then without having ADCs sampling every weight, taking up huge die area?

Self-hosting and homeserver stuff!

I started out when I got a new phone, and didn't know what to do with the old one. One of the ideas I had was a homeserver. Turns out it's not trivial to run Docker even on rooted Android phones, and you need a lot of kernel patching, tweaking and more and it still had issues after that.

The next step was when I figured out I could install postmarketOS on it, and I managed to flash it, SSH into it and set up Nextcloud for our photos and unbound as a recursive DNS for my home network. I thoroughly recommend postmarketOS, and the contributors are amazing as well.

I was however running out of storage, so I ordered an 256GB SD card, and set up mergerFS between it and local storage, worked fine.

After some time however, I got paranoid about having and old device with a LiPo battery constantly being charged in my home, so I decided to get a mini PC from Aliexpress and chucked a 2TB SSD in.

In the meantime, I discovered Immich, which turned out to be much better for photos than Nextcloud, and fell in love with it.

The final thing I added was a miniDLNA service to play my local movies and shows on my LG TV without having to bother with Plex/Jellyfin and reencode anything. Unfortunately, it kept disappearing after roughly 2 days of operation, so I just added a cron job to restart it at 5 AM.

For the time being, I don't need anything more and am turning my attention to other things.

The quote saying that people from 18 to 25 need a safe environment to "explore the world" and "find their purpose" seems very infantile and backward.

First off, it's not realistic at scale and presents a very sheltered worldview. Majority of worlds workforce is between those ages and no automation, nor AI will change this.

Second, even in the first world it's backward because you can also explore the world and find your purpose while working, infact working will teach you much more about the world than any college and you can always decide to get education when you're more mature and better off financially.

It really depends on what you classify under embedded. Everything more expensive than a Padauk or newer than PIC will have hardware support for the stack, and also most likely have the heap provided by newlib.

Bipartite buffers are amazing and criminally underused. For those looking for C and C++ implementations you can check out my libraries: lfbb and lockfee: https://github.com/DNedic/lfbb, https://github.com/DNedic/lockfree (although lockfree contains more data structures as well)

I am not sure how colleges work, but even at reputable universities this happens. Professors who have big fail rates will often be reprimanded for this, regardless of whether it's their fault or not.

Yes, sometimes this genuinely indicates a problem with the professor in question, but often times this leads to a long chain of responsibility passing, where if say a Calculus professor was afraid to have big failure rates, the Fluid Mechanics professor will be in a tough spot where they can either push back and have high failure rates or be more lenient, after that the Aerodynamics professor will inherit the same issue and so on.

Why do you assume electric batteries have to be used?

We've had things like reservoir lakes for hydro plants for a long time. Energy can be stored in more ways than electrical.

If you're interested about practical uses of atomic operations, I wrote lockfree: https://github.com/DNedic/lockfree, a collection of lock-free data structures meant to be readable and both hosted system and embedded friendly.

GNSS satellites talk to base stations on earth to get correction data using the measurements they obtain (including measurements obtained by tracking satellites from the ground). I believe this is what the poster above is reffering to.

RTK is a whole another beast and the meaning of an RTK base station is something else.

Undefined behavior will bite you eventually. Updated the compiler? You have no idea what to expect. Changed lines of code related to that part of code? Again, no idea what to expect.

The boot roms in ESP chips are not particularly small anyway, aside from bootup and flashing functionality, they also embed radio functions for WIFI, Bluetooth, Zigbee/Thread, printf implementations, MD5 functions and more.

Still far more expensive than any electric car including the same Tesla model in China.

Just came back from Shanghai, same situation, BUT from talking to people there are three big reasons for that:

1. The government built the EV charger infrastructure so that even the remotest areas are covered, the chargers have one universal way to use them and you can't go far without seeing one

2. In big cities China made car registration much more expensive for non-EVs

3. Prices of EVs in china are much lower than anywhere in the world

None of these apply for Europe or US, and I am doutbful they can make a swift transition like China can.

I think the electricity savings pale compared to the value of human time saved.

This won't work for:

1. Larger than word size variables

2. Out of order CPUs

3. On multicore CPUs when another core handles the signal

Atomics must be used here for proper synchronization, when they are available. If not, architecture-specific mechanisms should be used.

For what it's worth, my wife just finished her Master's thesis using this suite, she had very little adjustment time needed moving from Word.

Humans can't be forced to evaluate every piece of code 100000 times in an hour. LLMs can.

I wouldn't say this is necessarily a bad thing. The S3 is the most powerful ESP chip to date in terms of raw CPU power and even has AI and DSP instructions, esoteric things like direct CPU GPIO and more, I was kind of suprised to see such a powerhouse used for the Nano.

If you want more than a spsc queue, I've written `lockfree`, a collection of SPSC and MPMC data structures along the same principles the author here used:https://github.com/DNedic/lockfree.

The library is written in standard C++11 (but additional API's for higher C++ versions have been added), uses no dynamic allocation and is configurable so it is both big metal and deeply embedded friendly.

This feels extremely weird to read as someone naiively approaching networking where I actually want to talk to people, not just their professional side.

On the other hand it might even be suboptimal in the long run because when you hire or partner with someone, you don't just hire their knowledge or expirience, you hire their personality, drive and creativity.

The trick is to not even have strings in the binary. Take a look at trice, defmt and logging in ESP-IDF.

Except it's open source and anyone can audit the code any time.

This is already noted in the Queue readme, only the Queue constructs the type, the other 2 data structures are meant for PODs only.

I will take a look at adding support for constructing in-place for the other 2 data structures, but at the moment, they are just for PODs.

True, that would help immensely in creating MPMC data structures, but as these are SPSC there is no problem. Also to clarify, this is only for the indexes, the data members can be anything.

Using these intrinsics or inline assembly would break portability or create situations where platforms have different feature levels, which is not something I intend to do. I want the library to be compatible with everything from a tiny MCU to x86.

You make good points, I will try to rephrase that part of the README to be more accurate and explain benefits not mentioned yet like determinism.

This is why you would use the Ring Buffer or Bipartite Buffer to place 1000 elements at a time, not the queue. Check the documentation for more info.

Thanks a lot, will check this out!

This is a multi producer scenario, the README clearly states that these data structures are only single producer single consumer multi thread/interrupt safe.

I will also add disclaimers to the javadocs comments on methods just to reduce confusion.

Even in single producer single consumer scenarios you need locks for multithreaded/interrupt use if you're not properly using atomics and proper fences.

Could you elaborate on the alleged race conditions? Any advice on reliably testing the race conditions? The problem with adding those is the fact that they will give lots of false negatives and if you rely on them you have a problem.