3 ms·

It's parallel. Notably, Geohot's focus has been on improving ROCm directly while this project focuses more on improving the downstream tooling.

xiphias2 3y ago
He is doing full stack, and is selling tinybox (a PC with 6 AMD GPUs in it). As he went through the hard work of debugging the driver and getting AMD to care about it, I was expecting to see him in the acknowledgements.
- pokeypokes 3y ago
  Didn't he just find/report a bug in the multi-gpu case? Then have a melt-down after an AMD engineer sent him a fix a couple of days later? I think you are overstating his contributions, in fact it's not clear to me he made any.
  - kanwisher 3y ago
    He literally had a phone call with CEO of AMD, and got them to organize people to fix the problems. He has been whipping AMD into fixing their drivers
    - pokeypokes 3y ago
      Do you work at AMD and therefore have some kind of insight that we don't? Based on my own experience it's much more likely the call was just a marketing exercise, completely orthogonal to actual development. Gfx drivers are complicated, long term projects. They're not something that get fixed with a phone call. He found a bug, and made a bunch of noise about it. No need to make it more than it was.
  - jereees 3y ago
    He shows the git patch fix delivered by AMD engineers. The root cause is not explained and the change itself is about 2-3 SLoC — his frustration, what you call meltdown, is not about the interaction with AMD rather how troublesome it was to get AMD’s demo scripts to run on AMD’s chips.
    - pokeypokes 3y ago
      Feeling frustration about something not working is a normal part of software development. Reporting a bug, or choosing not to use that software are normal too. Making a public video trashing the work of 100s, using the language he did, is a meltdown. I've worked on public software before, and it's left me with an extremely low opinion of this kind of behaviour, and the kind of people that show it.
pokeypokes 3y ago
What contributions did he make to ROCm?
- cosmojg 3y ago
  In your words, "He found a bug, and made a bunch of noise about it."

xiphias2 3y ago

He is doing full stack, and is selling tinybox (a PC with 6 AMD GPUs in it). As he went through the hard work of debugging the driver and getting AMD to care about it, I was expecting to see him in the acknowledgements.

pokeypokes 3y ago
Didn't he just find/report a bug in the multi-gpu case? Then have a melt-down after an AMD engineer sent him a fix a couple of days later? I think you are overstating his contributions, in fact it's not clear to me he made any.
- kanwisher 3y ago
  He literally had a phone call with CEO of AMD, and got them to organize people to fix the problems. He has been whipping AMD into fixing their drivers
  - pokeypokes 3y ago
    Do you work at AMD and therefore have some kind of insight that we don't? Based on my own experience it's much more likely the call was just a marketing exercise, completely orthogonal to actual development. Gfx drivers are complicated, long term projects. They're not something that get fixed with a phone call. He found a bug, and made a bunch of noise about it. No need to make it more than it was.
- jereees 3y ago
  He shows the git patch fix delivered by AMD engineers. The root cause is not explained and the change itself is about 2-3 SLoC — his frustration, what you call meltdown, is not about the interaction with AMD rather how troublesome it was to get AMD’s demo scripts to run on AMD’s chips.
  - pokeypokes 3y ago
    Feeling frustration about something not working is a normal part of software development. Reporting a bug, or choosing not to use that software are normal too. Making a public video trashing the work of 100s, using the language he did, is a meltdown. I've worked on public software before, and it's left me with an extremely low opinion of this kind of behaviour, and the kind of people that show it.

pokeypokes 3y ago

What contributions did he make to ROCm?

cosmojg 3y ago
In your words, "He found a bug, and made a bunch of noise about it."