#asahi · 2021-09-29 12:14
sorear3d graphics and mpeg decode are handled in different hardware blocks on PCs too, but there are cultural differences in whether people use the term "GPU" in a wide or narrow sense#asahi · 2021-09-29 08:59
sorearbinary translation won't sandbag graphics workloads (unless you have precompiled shaders you're binary translating, but that's very rare outside console games)#asahi · 2021-09-28 16:58
sorearis there any metadata that can be extracted from calls and/or reading memory, that could be checksummed as a consistency check?#asahi · 2021-09-27 14:55
sorearcan more be shared with userland SEP support on macOS for ssh etc?#asahi · 2021-09-25 09:44
sorearalso, the latency doesn't matter because you can do this asynchronously?#asahi · 2021-09-25 09:41
sorearby the time you've done a UI interaction to wake up the phone, open the camera, taken a few dozen preview frames, autofocused, run the neural network noise reducer, etc, etc, etc, the final JPEG encode before storing the image is lost in the noise#asahi · 2021-09-25 09:37
sorearbecause cameras are expected to do video?#asahi · 2021-09-25 09:36
sorearI can't see how this makes any sense except for video as VinDuv first suggested#asahi · 2021-09-25 09:35
sorearon average over the life of your phone, how many still images per second have you taken?#asahi · 2021-09-22 05:58
soreari was under the impression there was also a kernel backtracer which didn't use that and just printed all of the code addresses in the stack, which if it were possible to do much better than that on optimized code without symbols someone would have done so already#asahi · 2021-09-22 05:55
sorearis there anything that makes the hv version of this problem materially different from the kernel's relatively mature version of the unwinder problem?#asahi · 2021-09-20 18:59
sorearkettenis: 2pi = 1?#asahi · 2021-09-20 18:56
sorearthat's something you've only been able to memorize in the last 2 years#asahi · 2021-09-14 18:15
sorearare we expecting a summary later in non-video form?#asahi · 2021-09-12 06:01
sorearsomething like that#asahi · 2021-09-12 05:38
sorearcan the fuses be used to brick the soc during usage or do they have a "lock out fuse writes" fuse#asahi · 2021-08-21 18:21
sorearanaloq: video decoding is not part of what is being discussed as "gpu"#asahi · 2021-08-20 01:31
sorearcitizen1[m]: https://asahilinux.org/about/#is-this-a-linux-distribution#asahi · 2021-08-20 01:11
sorearlinux supports the arm instruction set but a platform is more than an instruction set. it's true that 80-90% of the work is already done for us#asahi · 2021-08-18 18:27
sorearwasn't there discussion earlier of not even needing to change partition sizes for that, because iboot is apfs-subvol-aware?#asahi · 2021-08-18 18:24
sorearconversely, this setup means that you can nuke the asahi partition and reinstall it from scratch through 1TR without touching the distro rootfs or the distro-created files in the ESP?#asahi · 2021-08-18 18:19
sorearso you'd have an apfs volume with boot firmware, m1n1, and u-boot, and u-boot would then load the final kernel from a Linux filesystem via USB or NVMe?#asahi · 2021-08-18 18:15
soreardoes apple silicon use GPT?#asahi · 2021-08-14 12:19
sorearthank#asahi · 2021-08-14 11:51
sorearfocus for code or for words?#asahi · 2021-08-14 11:45
soreari think the same argument could be used to suppress all of the discussion of DCP progress#asahi · 2021-08-14 11:38
sorearoverall very nice piece, my desire for a clear statement of the current status of usb/pcie/spi/nvme etc remains unsated but that's not on any of you#asahi · 2021-08-14 11:13
sorearEOF#asahi · 2021-08-14 11:09
sorear"However, we do have access to the raw display controller registers too" not an editing comment but I'm wondering how we found this out, given that we can't trace DCP#asahi · 2021-08-14 11:02
sorearthere might be a small lie in the first bit because an informal "GPU" also includes video codecs, which I don't think are part of DCP or AGX, but I'm not sure what the best way around that is#asahi · 2021-08-14 11:01
sorearthinking something like: But what most people think of as a “GPU” is actually two completely distinct pieces of hardware: the GPU proper, which is in charge of rendering images to pixels in memory, and the display controller, which is in charge of sending those pixels from memory to the display.#asahi · 2021-08-14 11:00
sorearBut what most people think of as a “GPU” is actually two completely distinct pieces of hardware: the GPU proper, which is in charge of rendering images, and the display controller, which is in charge of sending those rendered images to the display. <-- this is a little clumsy both for the "most people think" bit and the vagueness of "rendered images"#asahi · 2021-08-13 19:54
sorearadding an option to the system preferences is "making it less usable"?#asahi · 2021-08-13 16:06
sorearDFU is what lets you boot over USB from another computer even if 100% of the flash is zeroed#asahi · 2021-08-10 06:21
sorearif user-provided shaders could tell it to access arbitrary memory in the device address space, and if DART was the only line of defense against that, it would quack a lot like an untrusted device (vfio/VM PCI passthrough being the other interesting case but that's much further down the road)#asahi · 2021-08-10 06:19
sorearso AGX doesn't need to be treated by DART as an "untrusted device" using the terminology of the thread, but only because AGX is out of scope for DART entirely?#asahi · 2021-08-10 06:17
sorearquestion about the split iommu pagesize thing. Does AGX require the IOMMU to be configured per-process to control what memory each shader has access to, or is that handled somewhere else in the stack? (My relevant knowledge is limited to one AMD HSA slide deck a few years ago)#asahi · 2021-07-30 11:37
sorear(what part of that is new?)#asahi · 2021-07-07 21:53
sorearnot really _that_ many places to look#asahi · 2021-07-07 21:52
sorearout of 20+ communities I was in half went to OFTC (which I've been on for years since all the kernel porting stuff was here), half went to libera, and one had the admins just decide to shut down IRC in favor of matrix without going through their established RFC process#asahi · 2021-07-01 12:50
sorear"bridge terminating on SIGTERM"#asahi · 2021-06-11 01:26
sorearkrbtgt: UMIP but this is getting very offtopic#asahi · 2021-06-10 16:23
sorearm68k and x86 started out with a CurrentEL mechanism but got rid of it later, a breaking change in the 68020’s case. they’re also saying about as directly as possible “popek/goldberg considered harmful”#asahi · 2021-06-10 15:43
sorearthat had to have been very contentious#asahi · 2021-06-10 15:38
sorearCurrentEL is readable at EL0? ew#asahi · 2021-06-10 15:05
sorearI need to look into how nested virt works on arm, I take it it’s not just a matter of trapping and emulating EL2-only registers#asahi · 2021-06-10 14:58
sorearor was this about malware not being able to use the architectural EL2#asahi · 2021-06-10 14:52
sorearWhat about Arm makes non-architectural extra hypervisors more difficult than other architectures?#asahi · 2021-06-02 20:07
sorearfor half a second i thought this was a new apple silicon-based playstation, they've migrated architectures enough already#asahi · 2021-06-02 05:39
sorearwhat arm platform has the best support to emulate? :p