| 8 Mar 2025 |
 eyJhb | 48 vcpu for 0.5 USD/h at Hetzner. Could be worth it | 16:42:12 |
| eyJhb | I wonder how fast that can build the kernel | 16:42:21 |
| eyJhb | CCX63 | 16:42:32 |
| eyJhb | https://www.hetzner.com/cloud/ | 16:42:39 |
 woobilicious | You're better off just creating a minimal kernel to build, or not using nix to build the kernel if you can avoid it. | 16:46:27 |
| woobilicious | eyJhb: if you can use a VM: https://github.com/YellowOnion/nixos-shell-kernel-dev | 16:53:14 |
| woobilicious | I used this for bcachefs debugging, should be able to adapt it to your use case. | 16:54:17 |
| eyJhb | In reply to @woobilicious:matrix.org
I used this for bcachefs debugging, should be able to adapt it to your use case. I'm not sure how nice that's when I'm debugging a VR game 😅 | 16:55:46 |
| woobilicious | ohh yeah vfio might be outside the use case lol | 16:56:25 |
| eyJhb | It's a real fun bug to test. Because it requires to put on the headset and stuff. :D | 17:14:09 |
| eyJhb | Surely non-vr games might suffer too, but I haven't found any. | 17:14:27 |
| eyJhb | Also, full disclosure. I finished the game.. mostly doing it because I don't want to deal with it in the future. | 17:14:47 |
 K900 | Well the real benefit of this would be that it actually identifies the issue | 17:27:29 |
| K900 | Or at least hopefully it identifies the issue | 17:27:36 |
| K900 | So it can be fixed for everyone | 17:27:43 |
| eyJhb | Yup. But I'm just so certain that somebody else must have encountered it. I feel like I'm just doing what somebody else reported. | 17:28:14 |
| K900 | Your average gamer doesn't know what a bisect is | 17:28:57 |
| eyJhb | If anybody wants to sanity check that this hasn't been reported before, please do. Basically horrible stuttering/lag in VR Half-Life Alyx after 6.9 (maybe), but surely it just manifests as bad performance in some games. | 17:29:01 |
| K900 | It's very likely that even if it was reported, it wasn't rootcaused | 17:29:08 |
| eyJhb | In reply to @k900:0upti.me
Your average gamer doesn't know what a bisect is But they would complain somewhere , right? | 17:29:09 |
| K900 | Maybe like on a Proton bug tracker | 17:29:26 |
| K900 | Among a million other bugs | 17:29:30 |
| K900 | With no clear reproduction | 17:29:38 |
| eyJhb | Hopefully it gets back to a SINGLE commit, and a revert of that will fix it. That's the best case. Unsure who to contact if/when a specific commit/change is found. But that's a bridge we'll (all of us on the journey with me) when we get to it :) | 17:31:56 |
| K900 | The point of contact would be gitlab.fdo/drm/amd | 17:32:46 |
| eyJhb | I haven't been to bad bisecting today. I'm sick anyways, so I can't really sit too much in front of the PC. So it has mostly been
- Put on a build
- Go to the living room and watch random TV
- Wait for notification on build
- Test it
- Repeat
| 17:33:10 |
| eyJhb | In reply to @k900:0upti.me
The point of contact would be gitlab.fdo/drm/amd Perfect. :D | 17:33:26 |
| eyJhb | The 48 core VPS still takes about 13 min til build the kernel | 19:03:15 |
| eyJhb | However, I discovered a friend of mine has a server with 64 cores. But most likely they aren't that powerful, but any of those solutions are faster than the current one. | 19:05:20 |
| K900 | It's diminishing returns after ~15 minutes yeah | 19:10:08 |
