Valentin Ignatev

There are many ways to kill performance in a multithreaded program. My favorite one is false sharing, because you can get 10x performance just by moving some bytes around. So let's look into it.

00:00 - When multiple threads is not even faster
04:23 - CPU L1 cache primer
09:40 - Getting ALL the performance
15:35 - Looking at disassembler
20:29 - How to find cache line address
22:38 - false sharing in disassembly
25:00 - closing words

Mario is one of the most reasonable voices in AI. We need more people like him

So, I'm very new to this, so please tell me - do they actually nerf current models right before releasing new one to fake benchmarks, or it's some conspiracy psyop?

I personally haven't notice any difference in the past couple weeks.

My pet peeve with a lot of modern open source is that it's made with clear intention of raising VC money or hopefully get bought out by a big tech without explicitly saying it, or straight up trying to downplay/deny it when asked.

Wanting to make money out of software is good and noble goal, but projects like this tend to accumulate a lot of non-organic hype, and then they are enshittified (if bought), or stagnate (if not), really quick, forcing users who don't understand the game to churn.

Which is, again, open source, no warranty etc etc, but it used to be that projects with thousands of github stars mostly meant mature and stable maintenance, but now everything is fake and inflated, and you need to see through all the astroturf.

procedure is called rgba
it returns argb
It's actually bgra in memory order (because we are on little endian machine).

Got I love graphics programming!