#asahi-gpu · 2022-10-02 10:54
linaThe current uAPI is just a demo, so all of that is going away entirely.#asahi-gpu · 2022-10-02 10:53
linaI had a long discussion with jekstrand and alyssa about uAPI design, so I have a good plan I think... but I really want to see it proven before we jump to upstreaming!#asahi-gpu · 2022-10-02 10:52
linaBut that's okay, it can live downstream for a while!#asahi-gpu · 2022-10-02 10:52
linaIt's also why I don't think this driver is going to be upstreamed any time soon, because rushing it makes no sense. I want to see GL 3.1 and OpenCL and at least some Vulkan running before we even attempt it.#asahi-gpu · 2022-10-02 10:51
linaYes and not really, which is why it's important to get this right the first time!#asahi-gpu · 2022-10-02 10:46
linaElla[m]: Thank you for working on this! Having a Vulkan implementation (even if incomplete) is going to be super useful to make sure we don't mess up the real uAPI design ^^#asahi-gpu · 2022-10-02 10:45
linaYou need to flip bit 0 to 0 for non-GL Z clip (I need to move that into the uAPI ASAP...)#asahi-gpu · 2022-10-02 10:45
linahttps://github.com/AsahiLinux/linux/blob/e92bd6fc55c71382d76d8950face426d292e79c1/drivers/gpu/drm/asahi/render.rs#L152#asahi-gpu · 2022-10-02 10:45
linaThe uAPI needs a full revamp to really support Vulkan, but I guess you should be able to test basic rendering jobs as is? Just keep in mind the Z clipping is hardcoded to OpenGL mode right now...#asahi-gpu · 2022-10-02 10:40
linaBy the way, we have at least one other user running the crazy it-works branch. I told her to just make a laundry list of what apps are broken, so we can go through it at our leisure as we get to fixing things.#asahi-gpu · 2022-10-02 10:39
linaIt being stable is no surprise though, it's a great way to clear cached state#asahi-gpu · 2022-10-02 10:39
linaYou mean as smooth as it does? Thank Apple's awesome PM ^^#asahi-gpu · 2022-09-30 14:09
linakarolherbst: ^^#asahi-gpu · 2022-09-30 12:26
linaAlso I don't know what DCP will do on laptops either...#asahi-gpu · 2022-09-30 12:25
linaMy DT data is hardcoded for 8GB machines, so it won't work for 16GB machines, and I'm not 100% confident it's going to work on anything but the mac mini either!#asahi-gpu · 2022-09-29 05:23
linahttps://twitter.com/LinaAsahi/status/1575345368018997248#asahi-gpu · 2022-09-29 05:23
linaIt works!!!! With a super cursed hack but it works!!!!#asahi-gpu · 2022-09-24 09:03
linaThank you <3#asahi-gpu · 2022-09-24 09:03
lina^^;;#asahi-gpu · 2022-09-21 19:41
linajannau: It's okay! I'm rebooting for every test anyway so... ^^#asahi-gpu · 2022-09-17 07:27
linaBut the GPU has a whole bunch of per-SoC parameters, so I don't think it's practical to try to make it generic across SoCs with just DT changes...#asahi-gpu · 2022-09-17 07:27
linaI think for DCP it makes sense to keep that in the DT, since I think the DCP interface is mostly the same for every SoC, at any given firmware version?#asahi-gpu · 2022-09-17 07:26
linaThat's different from the GPU though. For the GPU, the driver just maps everything ahead of time and tells the firmware what the virtual addresses are.#asahi-gpu · 2022-09-17 06:04
linaRight now it's hardcoded like this: https://github.com/AsahiLinux/linux/blob/gpu/rust-wip/drivers/gpu/drm/asahi/hw/t8103.rs#L23#asahi-gpu · 2022-09-17 06:04
linaNote that the driver doesn't use any of those mappings, it only directly uses the firmware coprocessor mailbox/start stuff, and the GPU registers directly in special cases (those would be in the DT as regular reg entries)#asahi-gpu · 2022-09-17 06:02
linaAnd since currently we don't mess with the device tree based on firmware in the bootloader, that's probably an argument *not* to include it? Since it would require more complex firmware -> DT mapping in m1n1#asahi-gpu · 2022-09-17 06:01
linaThis can also somewhat change between firmware versions, but the driver also needs explicit support for each firmware version so being flexible there also doesn't help#asahi-gpu · 2022-09-17 06:01
linaThis is different per SoC. Since I think the driver will need explicit support for each SoC anyway, I'm tempted to just hardcode it in the driver as part of the per-SoC data... but do you think it should be in the device tree instead?#asahi-gpu · 2022-09-17 06:00
linaIn terms of data, it's a sparse list of <phys_start, len, readonly> tuples#asahi-gpu · 2022-09-17 05:59
linarobher: I have a question about reg mappings. The GPU driver needs to map a bunch of miscellaneous hardware into the GPU firmware page tables (not just the GPU itself, also random temperature sensors, power management, and interrupt related register ranges)#asahi-gpu · 2022-09-16 18:05
linaIt's completely different hardware from Intel so it doesn't make any sense to think you'd have any of the problems you had on Intel!#asahi-gpu · 2022-09-16 17:53
linaAnd that also doesn't have anything to do with the GPU, which is what this channel is about...#asahi-gpu · 2022-09-16 17:52
linaI'm pretty sure DCP manages memory bandwidth by itself properly...#asahi-gpu · 2022-09-16 17:52
linabluetail: no offense, but none of what you are talking about any sense and also extrapolating from Intel systems to Apple Silicon doesn't make sense either.#asahi-gpu · 2022-09-15 18:06
linaI think that makes sense if it's just going into sysfs, yeah...#asahi-gpu · 2022-09-15 18:06
linaSo it does support arbitrary floats, and even arbitrary invalid strings, since it doesn't try to parse them ^^#asahi-gpu · 2022-09-15 18:05
linaSee iio_show_mount_matrix()#asahi-gpu · 2022-09-15 18:05
linaLooks like that just gets exposed via sysfs, the kernel never parses it?#asahi-gpu · 2022-09-15 17:53
linaI guess since /bits/ is already a thing, float vs. double would have to be specified that way anyway, so if it isn't an explicit keyword then anything with a . would be parsed as a float of whatever the current bitness#asahi-gpu · 2022-09-15 17:48
linaCould also be a keyword prefix in the style of `/bits/ 64`, like `/float/ 32 <1.0>` or just `/f32/ <1.0>`#asahi-gpu · 2022-09-15 17:47
linaWhat syntax would you prefer? C-style, like 1.0f? (Trailing f means f32, I don't need f64 but C style would be decimal without the f for that)#asahi-gpu · 2022-09-15 17:42
linaThe code could just do atof of course#asahi-gpu · 2022-09-15 17:42
linaI think you would just define the usual grammar for float syntax and then the code would have to parse it?#asahi-gpu · 2022-09-15 17:40
linaI might just implement it then and see how it goes#asahi-gpu · 2022-09-15 17:40
linaWell, at least knowing you think it's okay helps ^^#asahi-gpu · 2022-09-15 17:39
linaAh! Sorry, I didn't know.#asahi-gpu · 2022-09-15 17:34
lina(Well, the firmware in this case, but to the kernel it's all the same)#asahi-gpu · 2022-09-15 17:33
lina"Drivers don't use float" is great and all but when the hardware itself gives you an interface with floats... ^^;#asahi-gpu · 2022-09-15 17:33
linaI don't know whether a minimal NIH implementation would be better than just pulling in math-emu?