When people use BW in their performance models, they don’t use only 1 bandwidth, but whatever combination of bandwidth makes sense for the _memory access pattern_.

So if you are always accessing the same word, the first acces runs at DRAM BW, and subsequent ones at L1 BW, and any meaningful performance model will take that into account.

I use a git repo, markdown documents, and one daemon on each device, to sync and share notes across all my devices.

I have yet to find something that beats this in portability, flexibility, note quality (from markdown to interactive jupyter notebooks), file support (all files), math support (latex formulas for markdown), graph support (mermaid), presentation support (reveal.js),...

Beer Webcam I’ve ever had.

It is saying that worse candidates get jobs over better candidates because there is an artificial minority quota that must be achieved.

I helped fill a PhD position in Germany, it took 6 months to fill, because the first 19 properly qualified candidates were men. The 20th was a women, their qualifications were significantly worse than most men, but she got the job because of the quotas set by the German research foundation.

I never helped with filling a position again and this was one of the contributing factors for my decision against staying in academia.

It’s just not the default.

If I want to work on N windows at a time, i3 handles their tiling semi-automatically. All windows that I’m not currently using are pushed into different workspaces.

The Rust development and specification happens all in the open on GitHub.

There is no single company controlling its future or direction.

At the end of the day, buying a Computer is a tradeoff. A lot of people would very happily tradeoff freedom for other values if the value proposition is good.

That’s the main difference in these languages features.

Saying that Rust generics are „weaker“ than C++ templates is technically incorrect, since they are actuall „stronger“.

Using these terms to mean something else is definitely „weird“.

I mean, I get that you can’t enable these on AMD 6000 series because the card doesn’t support these well.

Spending a week/year working around ROCm would already cost you 5k$ plus the opportunity cost. For a whole team that’s a money sink.

It seems this is only for shaders, but if that’s the case one has been able to use shaders with Rust via GLSL since forever.