#asahi · 2022-07-23 13:14
janrinzeaccording to Wikipedia https://en.wikipedia.org/wiki/Page_(computer_memory)#Multiple_page_sizes we can have 64k pages on ARMv7#asahi · 2022-07-23 13:10
janrinzeif this was on Intel or AMD the effects may differ quite a bit i.r.t. M1.#asahi · 2022-07-23 13:09
janrinzethen on which system was this test performed?#asahi · 2022-07-23 13:08
janrinzej`ey: on which M1 was this done? M1,M1 pro, M1 max and M1 ultra might all behave quite different in this respect..#asahi · 2022-07-23 13:06
janrinzeit's not a big jump in performance from 16k to 64k pages. This is sort of similar to cache sizes. above a certain point doubling cache size will only gain a little performance.#asahi · 2022-07-23 13:03
janrinzethe benefit to goto 64k pages is there but only if you're not going to do swapping.#asahi · 2022-07-23 13:02
janrinzej`ey: RAM size in MB i guess#asahi · 2022-07-23 12:58
janrinzekov: that's interesting. memory allocation might have something to do with that. malloc performance on 16k pages might work more efficient. which makes me wonder what would happen when using 64k pages. Since these setings are all 'legacy' from times when memory was measured in MB's rather than GB's.. #asahi · 2022-07-23 12:18
janrinzejannau: your u-boot fix is exactly what i was trying to do with u-boot. thanks!! #asahi · 2022-07-23 12:01
janrinzegreat guy, friendly and very smart.#asahi · 2022-07-23 12:00
janrinzehe's busy on the compilers for ARM, i.i.r.c.#asahi · 2022-07-23 11:59
janrinzeif you see him say hi from me. haven't spoken to him in decennia.#asahi · 2022-07-23 11:57
janrinzeknow Wilco D?#asahi · 2022-07-23 11:57
janrinzej`ey: your work at ARM, right?#asahi · 2022-07-23 11:54
janrinzeon the topic of the 4k page patches, there was a mention of a performance penalty. which parts are hurt most by that?#asahi · 2022-07-23 11:51
janrinzeand installing an OS on them is way more painful than asahi ;-)#asahi · 2022-07-23 11:49
janrinzethey're NVidia systems. embedded and fanless.#asahi · 2022-07-23 11:49
janrinzethese 1Gb need to be point-to-point unfortunately#asahi · 2022-07-23 11:48
janrinzewe require multiple 1Gbit network adapters on those. The mini could do that over USB3. When the mac studio supports USB3 it might actually become a viable option.#asahi · 2022-07-23 11:46
janrinzeI know that asahi linux isn't production ready so i won't be advising to migrate our production systems to mac studio's yet ;-)#asahi · 2022-07-23 11:44
janrinzeat work we have arm64 in production systems, I like the binary compatibility. The drivers that i installed on those systems for hardware we have apparently work 1on1 on the mac mini. which is definitely a plus when developing. So i can build/run/test on asahi linux as if it was on the production systems.. (but waay faster ;-) )#asahi · 2022-07-23 11:40
janrinzei understand.#asahi · 2022-07-23 11:37
janrinzenot many will use asahi linux on a mac studio.. right? I guess most have a mac book pro/air or a mac mini.#asahi · 2022-07-23 11:07
janrinzeChaosPrincess: sounds good to me.. #asahi · 2022-07-23 11:05
janrinzeif the firmware is compressed in the .kext that might be a bit more difficult of course.#asahi · 2022-07-23 11:04
janrinzedottedmag: i don't have a hypervisor setup but i think that getting the signature and finding the firmware in the .kext could be a way for the installer to extract the firmware from the .kext#asahi · 2022-07-23 11:02
janrinzewe could capture a signature like the first X bytes and locate it in the .kext using that?#asahi · 2022-07-23 11:01
janrinzeusing the hypervisor, could we intercept the firmware write?#asahi · 2022-07-23 11:00
janrinzethe USB is Apple proprietary, right?#asahi · 2022-07-23 10:59
janrinzeah.. and on boot the firmware won't get loaded because the osx kernel doesn't get run..#asahi · 2022-07-23 10:58
janrinzehmm.. interesting choice of Apple.#asahi · 2022-07-23 10:57
janrinzeaha. and not yet for the USB. i see.#asahi · 2022-07-23 10:57
janrinzethe USB firmware is different then?#asahi · 2022-07-23 10:56
janrinzewe already use apple firmware for several components, right?#asahi · 2022-07-23 10:52
janrinzejannau: is that different from the mac mini?#asahi · 2022-07-23 10:51
janrinzeso it only will use high speed.#asahi · 2022-07-23 10:50
janrinzeusb 11-1: new high-speed USB device number 2 using xhci-hcd#asahi · 2022-07-23 10:49
janrinzeprobably just showing us that a certain flag is set.. right?#asahi · 2022-07-23 10:48
janrinzedmesg mentiones Host supports USB 3.1 Enhanced SuperSpeed#asahi · 2022-07-23 10:45
janrinzeOn the mac mini the USB A ports are USB 3 and i was hoping we could have the same with the mac studio.#asahi · 2022-07-23 10:45
janrinzeit's been asked here probably several times.. the USB A ports on the mac studio, do we know how to enable/use them?#asahi · 2022-07-20 22:07
janrinze:-D#asahi · 2022-07-20 22:04
janrinzeclover[m]: I use Geekbench-5.4.4-LinuxARMPreview#asahi · 2022-07-20 22:03
janrinzeclover[m]: have you ran Geekbench 5 on that lenovo yet?#asahi · 2022-07-20 21:22
janrinzeTimer frequency is 250Hz do we want 1000Hz? these systems are plenty fast..#asahi · 2022-07-20 21:21
janrinzewas using LXDE for a long time after fvwm2 and now migrated to LXQT#asahi · 2022-07-20 21:21
janrinzeNVidia kernel driver blobs usually would crash on PREEMPT#asahi · 2022-07-20 21:20
janrinzeit's also very sensitive to badly written drivers..#asahi · 2022-07-20 21:19
janrinzein my experience having a full PREEMPT kernel is on ARM very responsive#asahi · 2022-07-20 21:17
janrinzei keep forgetting to set the option to have /proc/config.gz it's so darn useful when you're testin different kernels..