karolherbst 🐧 🦀

Linux Graphics Developer working at Red Hat
Nouveau and stuff
Implemented OpenCL in Rust for fun

Aggressive 🦀🦀🦀 marketing

🏳️‍🌈🏳️‍⚧️ All are welcomed.

Nazis, $hitcoin cultists, Right-Libertarians and other fascists not welcomed.

Private account, please direct all business inquiries to: https://twitter.com/karolherbst

All photos posted are CC-BY 4.0

#mesa
#nouveau
#nvk
#opencl
#rust #rustlang
#rusticl

#photography

Public Key

npub1cpdz8xz2ye2s2496de040yg587u0hvfdledm8tjqzf6wqpfe7s7sncumus

NIP-05 Address

karolherbst_at_chaos.social@mostr.pub

Profile Code

nprofile1qqsvqk3rnp9zv4g92jaxuh6hjy2rlw8mkyklukan4eqpya8qq5ulg0gpz4mhxue69uhhyetvv9ujumt0wd68ytnsw43qz9nhwden5te0wfjkccte9ec8y6tdv9kzumn9ws53qvr3

Publishing to

relay.mostr.pub

relay.primal.net

Author Public Key

npub1cpdz8xz2ye2s2496de040yg587u0hvfdledm8tjqzf6wqpfe7s7sncumus

Show more details

Last Notes

2024-10-12 17:53:58

- reply

@npub1wly…c4mu some people assume it's a saving face strategy before quitting LLM AI.
I doubt they'll remove the acceleration structures from their hardware, because it's useful in other applications (e.g. camera post-processing or face detection)

2024-10-12 03:28:22

- reply

@npub15qv…e3a8 yeah, that's the perfect analogy. It's a magic trick and most people don't see the trick.

2024-10-12 03:01:37

“we found no evidence of formal reasoning in language models"
yep, checks out.

2024-10-08 12:07:44

- reply

@npub1da2…p79u I wished I could join your streams more often, though this week I'm in NA, so I might be able to if you have one scheduled for this weekend or so :D

2024-10-08 12:05:42

Nice.. my flight to Montreal is late 🙃

2024-10-08 08:59:09

- reply

@npub1ec6…3zj6 NOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO 😭 😭 😭

2024-10-07 19:51:58

- reply

@npub1xwv…xwwt when you have to tell something interested about yourself and need to come up with something

2024-10-06 20:57:41

Can people please stop with boomer humor in videos?
They literally turn great videos into absolutely shitty ones.

2024-10-05 09:10:18

people engaging in voter shaming generally show that they have no clue about lived experiences of marginalized folks.
It's really sad, because people engaging in voter shaming tell more about themselves than they are aware of.
It's often classicistic, but also ignores that for some folks it's literally dangerous to go voting.

2024-10-03 17:42:20

- reply

@npub1532…5ffy they need to be 🦖 nuggies!

2024-10-03 12:20:13

It's kinda sad to see that when people talk about conflicts they are emotionally or personally impacted by, that every "side" lives in their own detached reality and it's just causing people to talk by each other.
And it's entirely understandable, but I still wished that people would be able to acknowledge at least that much and accept that there is no objective reality and that we should strive to learn how the reality of others look like.

2024-10-01 20:29:30

- reply

@npub1f8e…v2mv Firefox added that so that people complain on twitter so the ads reach an even wider audience!

2024-09-25 09:37:07

- reply

@npub1pfq…m4an I'd claim that 95% of the "technical aspects" being brought up are in fact only opinions. And it would help everybody if developers reflect on those things a lot more.

2024-09-25 09:36:29

- reply

@npub1pfq…m4an Yeah, I won't deny that, it's just that those also often don't spark the amount of discussions you see with other things.
Most developers are convinced by facts, but what happens a lot is that some developers perceive something as a fact which actually isn't, but still claim it's all technical and a huge disconnect happens as a result.
Then both sides are "purely technical" and just claim the other side is just wrong or something along those lines.

2024-09-25 09:32:15

- reply

@npub1uqg…uza4 yeah, I think that's what I tried to address with the very last part. Once you get very deep down into the facts you can create an objective view on the problem, using benchmarks, etc.. but even that is hard, because what benchmarks are the one that matter? What trade-off are you willing to make? Is code quality more important or absolute performance?
Even there you can stray away from "technical" pretty fast.
The more theoretical (math) you get the easier it's to be objective though.

2024-09-25 09:12:05

I'm confused by the narrative of differentiating between social and technical problems
Most technical problems are based on personal opinions and therefore are just that: opinions
So if you want to discuss a technical problem with others you already face different opinions, which in itself is a social problem to figure out what's agreeable or desired by everybody involved
The IT space looks so technical, but everything is so much based on anecdotes, opinions and feelings, and rarely on facts

2024-09-25 09:12:05

2024-09-23 21:28:19

*sigh*

2024-09-02 08:27:45

It's frustrating, everything. And I don't mean the election result. That was entirely expected.
What's frustrating is, that most people apparently don't know why the Nazis get into power. That outlawing the party without fixing structural problems (poverty, wealth inequality, east-west relation, etc..) does nothing.
Saying that media is at fault for inviting Nazis, is also saying you think people are too stupid for the discourse, it's an inherently anti-democratic take.
1/x

2024-09-01 19:32:45

what's the absolute preferred way of installing cosmic through dnf on fedora?
Using the "ryanabx/cosmic-epoch" copr or something else?

2024-08-25 20:37:49

users demanding things from FOSS developers are the same as people shouting "the customer is always right" at retail store workers.

2024-08-17 15:13:38

As people seem to discuss Code of Conduct stuff again, please keep in mind that decisions are normally not purely made based on specific "wrong doings" or "having the wrong opinion".
What matters as well is how one behaves if confronted about things and if a constructive and reasonable discussion is possible with the person in question.
Reasonable people are entirely fine with having those discussions and we can together find a path forward and figure something out working for everybody.

2024-08-11 10:00:42

I wished we'd had a way in #rust to say "usize is at least u32 big", because that would allow me to get rid of a loooooooot of "as" statements in #rusticl

2024-08-11 10:00:42

I wished we'd had a way in #rust to say "usize is at least u32 big", because that would allow me to get rid of a loooooooot of "as" statements in #rusticl

2024-08-09 17:13:03

R.I.P. AMD

2024-08-09 10:22:58

I wished Ubuntu would have chosen to ship the absolute latest LTS branch or to allow switching the kernel release mid ubuntu release, but I guess this is a bit better over the old policy as this makes backporting patches a little easier.

2024-08-01 15:39:39

I have no idea who is the source, but...
https://assets.chaos.social/media_attachments/files/112/887/509/489/956/512/original/4c48a7636b8eb9a6.png

2024-07-01 19:21:39

protect yourself from all openssh vulnerabilities with this little trick:
systemctl stop sshd; exit

2024-06-26 10:46:44

Soo. the initial set of #rusticl enablement patches for the raspberry pi4 and newer are merged and will be part of mesa 24.2 (https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25362)
It doesn't support profiling and images yet, but it's stable enough to pass around 97% of the OpenCL CTS
Of course there will be bugs, so any kind of testing will be appreciated.
In the meantime I'll try to get images supported.

2024-06-19 14:11:12

The answer to the question of "do I use a floating point arithmetic for currency" was, is and always will be: No
Please just don't do this to yourself.

2024-06-16 13:46:59

Did anybody started to collect the "dos" and "don'ts" of job applications with suggestions on how to trick AIs?

2024-06-13 18:40:18

everything goes according to plan *hehehehehehehehehehhheheheheh*
tldr: "NVIDIA has been exploring ways to better support the effort for an
upstream kernel mode driver"
https://lore.kernel.org/dri-devel/20240613170211.88779-1-bskeggs@nvidia.com/T/#m17464be2ff30be41d75b214a10102ad60b708812

2024-05-19 08:31:37

git shouldn't disclose any email addresses or rather, we should just remove email addresses altogether from git

2024-05-13 15:05:17

Is there a good way in #rustlang to prevent stackoverflows with long dependency chains of Arced objects?
Basically I need to prevent something like this from happening: https://godbolt.org/z/s9aTfeeaT
And no, `Weak` can't be used here as in my case I need to access the dependency reliably.

2024-05-09 12:28:20

I should play Hades 2....