| 20 Nov 2025 |
 Randy Eckenrode | Anyway, I’ll see what from Linux Darwin doesn’t need now that it does need Vulkan stuff. If it doesn’t make things too much worse, I may try to combine them. Or not. I’ll look at it tonight. | 13:06:50 |
 K900 | Feel free | 13:07:21 |
| K900 | Though I'd expect that it'll be more trouble than it's worth | 13:07:37 |
| K900 | Even now that the Linux expression is considerably more sane | 13:07:47 |
| Randy Eckenrode | There’s no hurry. Unless we want to provide an unstable Mesa on DarwinC KosmicKrisp won’t be available until Mesa 26.0. | 13:09:24 |
| Randy Eckenrode | * | 13:09:36 |
| K900 | Uhh what | 13:09:58 |
| K900 | 25.3 has it, no? | 13:10:02 |
| Randy Eckenrode | No. Compare
https://gitlab.freedesktop.org/mesa/mesa/-/tree/25.3 to https://gitlab.freedesktop.org/mesa/mesa/-/tree/main. The kosmickrisp driver is only present in the latter. | 14:13:14 |
| Randy Eckenrode | * No. Compare https://gitlab.freedesktop.org/mesa/mesa/-/tree/25.3 to https://gitlab.freedesktop.org/mesa/mesa/-/tree/main. The kosmickrisp driver is only present in the latter. | 14:13:25 |
| Randy Eckenrode | According to https://www.phoronix.com/news/KosmicKrisp-Merged-Mesa-26.0, it’s targeting the 26.0 release next year. | 14:14:14 |
| K900 | Hm OK | 14:15:37 |
| K900 | I thought it made 25.3 | 14:15:40 |
| K900 | Then we don't need to rush | 14:15:45 |
| Randy Eckenrode | I'm tracking HEAD in my work. There’s still work being done. | 14:16:39 |
| Randy Eckenrode | According to Alyssa, feature parity with MoltenVK will happen, which means we may be able to use it by default in the future. | 14:17:30 |
| Randy Eckenrode | Timing-wise, it works out really well for 26.11. | 14:17:54 |
| Randy Eckenrode | KosmicKrisp requires at least macOS 15 and doesn’t support GPUs other than Apple Silicon. We’ll be bumping the minimum to 15.0 and dropping Intel support at that time. | 14:18:52 |
| Randy Eckenrode | There won’t be a need for elaborate mechanisms to seap the GPU driver. | 14:19:18 |
| Randy Eckenrode | * | 14:19:26 |
| Randy Eckenrode | * | 14:19:38 |
| Randy Eckenrode | RIP Wine on Darwin though. | 14:19:46 |
| Randy Eckenrode | Apparently FEX doesn’t support 16k pages, and upstream doesn’t seem interested in supporting it. | 14:20:19 |
| Randy Eckenrode | If Asahi maintains patches, we may be able to use those, but I don’t know. | 14:20:55 |
 Alyssa Ross | Asahi runs FEX in a 4K page VM | 14:22:34 |
| Randy Eckenrode | I thought FEX was the VM-less solution. | 14:23:00 |
| Alyssa Ross | No, FEX is the x86_64→aarch64 part | 14:23:21 |
| Randy Eckenrode | Or is it just less of one compared to the previous implementation? | 14:23:24 |
| Alyssa Ross | the VM is the 16K→4K part | 14:23:28 |
| Randy Eckenrode | Ah. I guess we need to hope CodeWeavers does the work. | 14:23:55 |
