| 4 Jun 2023 |
 kvark | also getting annoyed by RenderDoc's shader debugger
I realize this is super cool tech and all - self-written shader executor. And that NVidia doesn't have shader debugging in NSight.
But you can't really trust what it produces. It doesn't match the hardware. It's plainly wrong from what I see. | 01:43:31 |
| kvark | i.e. ImageSampleExplicitLod selecting from the last LOD (1x1 size), but the sampled value appears to be from LOD 0. Pretty confident this is just RenderDoc's bug. | 01:44:27 |
| kvark | aha, I take it back, it was my sampler erroneously configured for LOD range 0-0 😅
Can't trust anything, including my own code! | 01:54:38 |
| kvark | at the same time, that could be attributed to Ash not having the right defaults | 01:56:17 |
 Ralith | ash's defaults are very deliberately not trying to have any particular semantics for any field save sType | 06:46:52 |
| Ralith | that would be a huge maintenance liability | 06:47:01 |
| kvark | I agree it makes sense. It's just becoming a bit confusing with regards to the builder pattern. When using a builder, it's not always clear which pieces can be reasonably omitted, and which are required to be set. | 07:40:36 |
| 5 Jun 2023 |
 cwfitzgerald | bitflags 2 is great | 03:30:41 |
| cwfitzgerald | wgpu-info's code is so much simpler | 03:30:46 |
| cwfitzgerald | 
Download image.png | 03:34:13 |
| cwfitzgerald | not bad for a 400 line file | 03:34:17 |
| kvark | how did it help? | 03:45:35 |
| cwfitzgerald | Iteration and name extraction are now first class, so I removed the hacks we had for that | 03:47:51 |
| cwfitzgerald | Also they're is a bitflags trait now, which we had our own mini copy of | 03:48:29 |
| kvark | Since the trait is not required to use a bitflag, I wonder if they are doing something similar to https://crates.io/crates/hidden-trait | 03:52:42 |
| cwfitzgerald | The methods exist both natively and as part of the trait | 03:54:13 |
| cwfitzgerald | so I think they just create both methods, likely having the native ones call the trait | 03:54:24 |
| kvark | that's exactly what hidden-trait does as well, yeah | 04:53:13 |
| kvark changed the room topic to "Off-topoc" from "NixOS migration". | 20:25:26 |
| kvark changed the room topic to "Off-topic" from "Off-topoc". | 20:25:32 |
| cwfitzgerald | off-tupac | 20:25:53 |
| kvark | Big fat off-topic request: if you have talented friends, who know C++, computer science, physics, rendering, ML, whatever, and they care about environment and want to work on cutting edge technology with interesting problems, please send them my way! Tesla Autopilot Simulation needs talent right now, we are growing.
Note: I assume this doesn't apply to readers of this channel directly, because you folks are all brilliant and are already doing cool stuff in Rust. | 20:27:33 |
| 7 Jun 2023 |
 m4b | wgpu on vision pro when? :D | 03:48:36 |
 @heavyrain266:mozilla.org | At least it's worth bothering with Apple's new linker introduced in Sonoma | 19:21:14 |
| @heavyrain266:mozilla.org | Compilation on macOS is horribly slow | 19:21:26 |
| @heavyrain266:mozilla.org | * Compilation and cross-compilation on macOS is horribly slow | 19:21:39 |
| @heavyrain266:mozilla.org changed their profile picture. | 19:45:53 |
| 8 Jun 2023 |
| kvark | Ralith: tried Nickel for Nix packages yet? I wonder if it’s getting there | 15:29:41 |
| Ralith | In reply to @kvark:matrix.org
Ralith: tried Nickel for Nix packages yet? I wonder if it’s getting there nope; at a glance, nickel doesn't look much less idiosyncratic than Nix, so I'm not entirely sure I see the point | 16:50:53 |
| cwfitzgerald | Man, the workflow of “write what you want in a comment then let github copilot fill it in” is wild. Very helpful for things like shell scripts where I know basically nothing | 22:25:01 |
