Its highly dependent on implementation.

https://www.pugetsystems.com/labs/articles/stable-diffusion-performance-professional-gpus/

The experience on Linux is good (use docker otherwise python is dependency hell) but the basic torch based implementations (automatic, comfy) have bad performance. I have not managed to get shark to run on linux, the project is very windows focused and has no documentation for setup besides “run the installer”.

Basically all of the vram trickery in torch is dependent on xformers, which is low-level cuda code and therefore does not work on amd. And has a running project to port it, but it’s currently to incomplete to work.

I wish. Sadly the google play store requires monotonically increasing build numbers, so any option of resetting build numbers after major releases goes out the window.

Bundling my two sata ssds into a single zfs volume, instead of manually moving stuff around between the nvme and two sata ssds. Combined with compression, and for my code folder deduplication it also resulted in a lot more usable space.

That’s an interesting approach. The Traditional way would be to go by game score like the AI Mario Projects. But I can see the value in prioritizing Bullet Avoidance over pure score.

Does your training Environment Model that shooting at enemies (eventually) makes them stop spitting out bullets? I also would assume that total survival time is a part of the score, otherwise the Boss would just be a loosing game score wise.

Bringing more modern tools and features to existing large code bases is “destroying his reputation”? Bjarne and the committee is constantly extending and modernizing a language with code bases older than me. Yes that means the old stuff has to be kept around but that is the price of allowing existing code to migrate gracefully instead of just throwing it out of the window. There is a problem with some missing rails to enforce current and saver techniques but Bjarne is not denying that.

That is the mindset that gives us text editors using 100% cpu to blink a cursor because their css triggers a bug in the web browser they ship to render the text editor.

You can be memory save without shipping a whole browser, but disregarding power and memory efficiency will just make performance gained by hardware evaporate in overhead.

This also works the other way round:

The best programmers won’t be able to fix you clunky mess of bureaucracy by making it digital.