| vkoskiv | Trying to get linux 0.01 compiled | 00:15 |
|---|---|---|
| vkoskiv | Almost there, hunting down link errors. Lots of missing symbols. | 00:15 |
| vkoskiv | I did end up commenting out some parts to deal with later, and I replaced a few things with builtins. I really just want to get it to compile so I can hop around the code with helix :D | 00:16 |
| minute | funky | 00:25 |
| vkoskiv | :29 | 00:25 |
| vkoskiv | Whoops. Basically already there, just some global symbols with _ prefix that I need to fix | 00:26 |
| vkoskiv | I also started writing a script to convert a C codebase to 8.3 FAT filenames, while also fixing references in includes :D | 00:26 |
| vkoskiv | I want to build my raytracer under DOS+DJGPP | 00:27 |
| - eibachd (QUIT: Ping timeout: 246 seconds) (~eibachd@p200300dcf7353e00481b860ed5c57452.dip0.t-ipconnect.de) | 01:00 | |
| + eibachd (~eibachd@p200300dcf7353e00e3949f789e8e79f8.dip0.t-ipconnect.de) | 01:01 | |
| vkoskiv | All the objects link now, next up is this funky final build step | 01:01 |
| vkoskiv | Maybe I got far enough as is | 01:01 |
| vkoskiv | LOL, it doesn't build boot/boot.o because my username isn't the same as linus's | 01:05 |
| vkoskiv | It just calls `ls -l` on a file, and then cuts a hard-coded range from the output | 01:06 |
| - skipwich (QUIT: Ping timeout: 276 seconds) (~skipwich@user/skipwich) | 01:06 | |
| vkoskiv | Fixed, now I just need to translate boot/boot.s from whatever funky ancient assembly syntax this is to something more modern | 01:14 |
| vkoskiv | That's for tomorrow then, I guess. Getting a bit late. | 01:14 |
| + skipwich (~skipwich@user/skipwich) | 01:19 | |
| vkoskiv | Ah, apparently as86 is still available | 01:21 |
| vkoskiv | I might get this thing fully built! | 01:21 |
| - cobra (QUIT: Quit: ZNC 1.8.2 - https://znc.in) (~cobra@user/Cobra) | 01:46 | |
| + cobra (~cobra@user/Cobra) | 01:55 | |
| + nocko (~nock@user/nocko) | 02:05 | |
| - nsc (QUIT: Ping timeout: 268 seconds) (~nicolas@170-97-142-46.pool.kielnet.net) | 03:19 | |
| + nsc (~nicolas@90-98-142-46.pool.kielnet.net) | 03:20 | |
| - Ar|stote|is (QUIT: Quit: https://quassel-irc.org - Chat comfortably. Anywhere.) (~linx@149.210.8.201) | 03:21 | |
| + Ar|stote|is (~linx@149.210.8.201) | 03:22 | |
| - mjw (QUIT: Ping timeout: 276 seconds) (~mjw@gnu.wildebeest.org) | 03:38 | |
| + vagrantc (~vagrant@2600:3c01:e000:21:7:77:0:20) | 04:20 | |
| - Kooda (QUIT: Remote host closed the connection) (~kooda@natsu.upyum.com) | 06:21 | |
| + Kooda (~kooda@natsu.upyum.com) | 06:24 | |
| - eibachd (QUIT: Ping timeout: 246 seconds) (~eibachd@p200300dcf7353e00e3949f789e8e79f8.dip0.t-ipconnect.de) | 07:11 | |
| + eibachd (~eibachd@2a01:599:306:d8f3:7a4c:f300:c113:7105) | 07:11 | |
| josch | minute: some more a311d feedback. Since the screen and battery of my old thinkpad broke, it has been in the shelf and i have used the reform. Today, I *really* needed an x86_64 system as i was investigating an issue that is not reproducible on arm64, so i revived the system far enough to run the testsuite of one of my projects on it. Turns out, that a311d is *faster* executing that testsuite (4 hours 11 | 07:56 |
| josch | minutes) than my old thinkpad with an intel core i5 5300u (4 hours 35 minutes) -- so th a311d is an *upgrade* for me! o0 | 07:56 |
| - vagrantc (QUIT: Quit: leaving) (~vagrant@2600:3c01:e000:21:7:77:0:20) | 08:15 | |
| - yankcrim- (QUIT: Ping timeout: 260 seconds) (~nick@gw.tetromino.io) | 09:22 | |
| + mjw (~mjw@gnu.wildebeest.org) | 09:45 | |
| minute | josch: nice!! | 10:15 |
| josch | now we just need proper suspend (and not just reducing power consumption by half like imx8mq for those for whom that's working) and the reform becomes something i'd get for my partner once their current laptop breaks. | 10:22 |
| minute | i tried suspend only like once on a311d. it did something (on serial) but didn't fully work. more investigation needed | 10:24 |
| josch | so much to do, so little time! :D | 10:28 |
| minute | if someone is interested in NPU stuff (in the a311d, and i think also in imx8mp), support is upstreamed to mesa now https://blog.tomeuvizoso.net/2024/01/etnaviv-npu-update-15-we-are-upstream.html?m=1 | 10:28 |
| minute | josch: yeah currently the two main things i'm working on are finishing the first rk3588 adapter and managing the pocket reform assembly&shipment | 10:29 |
| josch | rk3588 is also super exciting i think :) | 10:29 |
| josch | there is also some good news from my side that i think i didn't mention yet? | 10:30 |
| minute | oh? | 10:30 |
| josch | we got gobject-introspection stuff to cross-build in Debian | 10:30 |
| josch | so now the cross-building of gstreamer in reform-debian-packages can be done in 4 minutes instead of 1.5 hours | 10:30 |
| josch | i have a gobject-introspection branch in the repo that does that | 10:30 |
| minute | oh wooow | 10:30 |
| minute | that's awesome | 10:31 |
| [tj] | that is an amazing reduction, what did you change? | 10:31 |
| josch | [tj]: cross-compiling anything that involves gobject-introspection requires running foreign-architecture stuff (g-ir-scanner) as that is how the .gir files get created. The underlying problem is, that upstream g-i does not support creating the gir files for architectures that are not the native ones. | 10:33 |
| josch | [tj]: other distributions than debian like yocto, void linux or ptx run g-ir-scanner under qemu to lift this limitation | 10:34 |
| josch | this means that the process is still broken when trying to bootstrap architectures for which qemu support is not yet available but it does help in all cases where we want to cross-build for established architectures | 10:35 |
| josch | in fall last year, Helmut Grohne (our Debian cross-building expert) got together with Simon McVittie (the gobject-introspection maintainer) in a Debian sprint in Cambridge and they sat together and worked on running gir stuff under qemu when cross-building automatically as it is already done for all the embedded distros out there | 10:36 |
| josch | after that was working, i only needed to patch imath and meson to cooperate | 10:37 |
| josch | i already filed all the patches with these projects and am waiting for those to get merged and/or new releases with the merged patches to happen | 10:37 |
| minute | really cool | 10:39 |
| josch | well... on the other hand, all this will soon not be needed anymore :D | 10:41 |
| josch | because at some point this year there will be another gstreamer release which incorporates the things we patch stable gstreamer with right now :) | 10:42 |
| minute | heh :D | 10:46 |
| [tj] | so the change is to do the cross platform part inside a qemu-user instance? | 10:51 |
| [tj] | how did it work before when it was slow? | 10:51 |
| josch | [tj]: it was slow because we did the whole build inside qemu | 10:52 |
| [tj] | ah ha | 10:52 |
| [tj] | that is a great improvement, qemu-user is a wonderful thing | 10:52 |
| [tj] | freebsd uses it for doing ports cross builds in some cases | 10:52 |
| - mjw (QUIT: Ping timeout: 264 seconds) (~mjw@gnu.wildebeest.org) | 10:59 | |
| - buckket (QUIT: Quit: buckket) (~buckket@vps.buckket.org) | 12:34 | |
| + buckket (~buckket@vps.buckket.org) | 12:34 | |
| - amk (QUIT: Remote host closed the connection) (~amk@user/amk) | 12:52 | |
| + amk (~amk@user/amk) | 12:56 | |
| - amk (QUIT: Remote host closed the connection) (~amk@user/amk) | 13:04 | |
| + amk (~amk@user/amk) | 13:04 | |
| - amk (QUIT: Remote host closed the connection) (~amk@user/amk) | 13:11 | |
| + amk (~amk@user/amk) | 13:12 | |
| - amk (QUIT: Remote host closed the connection) (~amk@user/amk) | 13:20 | |
| + amk (~amk@user/amk) | 13:20 | |
| - amk (QUIT: Remote host closed the connection) (~amk@user/amk) | 13:24 | |
| + amk (~amk@user/amk) | 13:24 | |
| - amk (QUIT: Remote host closed the connection) (~amk@user/amk) | 13:43 | |
| + amk (~amk@user/amk) | 13:44 | |
| - bkeys (QUIT: Read error: Connection reset by peer) (~Thunderbi@45.134.140.153) | 14:21 | |
| + mjw (~mjw@gnu.wildebeest.org) | 15:10 | |
| + bkeys (~Thunderbi@45.134.140.153) | 15:14 | |
| - robin (QUIT: Remote host closed the connection) (~robin@user/terpri) | 15:36 | |
| + robin_ (~robin@user/terpri) | 15:36 | |
| - robin_ (QUIT: Remote host closed the connection) (~robin@user/terpri) | 15:46 | |
| + robin_ (~robin@user/terpri) | 15:46 | |
| bkeys | minute: Just asking I know you are busy, I'd like to order the display adapter ASAP; I have a lengthy vacation I'd like to take the Reform on with cm4 | 16:00 |
| minute | bkeys: remind me of your situation... a311d not working? | 16:30 |
| bkeys | I have the cm4 and I only ordered the a311d with the cm4 adapter | 16:55 |
| bkeys | You said I need a display adapter for my cm4 to show up on the screen | 16:55 |
| bkeys | You said you were gonna upload the display adapter as a separate part in the store in January (we had this conversation in late December) | 16:56 |
| minute | bkeys: yeah i remember that. but you need to use the pi cm4 in your vacation, yes? just ooc what's the downside of a311d vs it? | 17:01 |
| - Ar|stote|is (QUIT: Ping timeout: 264 seconds) (~linx@149.210.8.201) | 17:05 | |
| + Ar|stote|is (~linx@149.210.21.255) | 17:09 | |
| - eibachd (QUIT: Read error: Connection reset by peer) (~eibachd@2a01:599:306:d8f3:7a4c:f300:c113:7105) | 17:29 | |
| + eibachd (~eibachd@p200300dcf7353e00f5614e3497c6408e.dip0.t-ipconnect.de) | 17:30 | |
| - eibachd (QUIT: Ping timeout: 246 seconds) (~eibachd@p200300dcf7353e00f5614e3497c6408e.dip0.t-ipconnect.de) | 18:19 | |
| + eibachd (~eibachd@2a01:599:306:d8f3:7a4c:f300:c113:7105) | 18:20 | |
| + vagrantc (~vagrant@2600:3c01:e000:21:7:77:0:20) | 18:39 | |
| - eibachd (QUIT: Ping timeout: 256 seconds) (~eibachd@2a01:599:306:d8f3:7a4c:f300:c113:7105) | 19:01 | |
| + eibachd (~eibachd@2a01:599:302:1adc:1368:31e4:1143:85d0) | 19:13 | |
| - eibachd (QUIT: Read error: Connection reset by peer) (~eibachd@2a01:599:302:1adc:1368:31e4:1143:85d0) | 19:18 | |
| + eibachd (~eibachd@p200300dcf7353e002320bafd2adfeb5b.dip0.t-ipconnect.de) | 19:19 | |
| - eibachd (QUIT: Ping timeout: 264 seconds) (~eibachd@p200300dcf7353e002320bafd2adfeb5b.dip0.t-ipconnect.de) | 19:23 | |
| + eibachd (~eibachd@p200300dcf7353e00d60f808c3fd74c29.dip0.t-ipconnect.de) | 19:25 | |
| - GNUmoon (QUIT: Remote host closed the connection) (~GNUmoon@gateway/tor-sasl/gnumoon) | 20:30 | |
| minute | rk3588 is getting closer to reality https://mastodon.social/@mntmn/111818314280393711 | 20:51 |
| minute | also, > please recommend your favorite linux-supported SDIO/UART WiFi/BT module now! | 20:52 |
| sevan | is it likely there would be 32GB RAM support? | 20:52 |
| - vagrantc (QUIT: Quit: leaving) (~vagrant@2600:3c01:e000:21:7:77:0:20) | 20:52 | |
| minute | sevan: yes | 20:53 |
| sevan | <3 | 20:53 |
| josch | minute: i don't understand the need for selecting a wifi/bt module. Can not everybody install their own into the mini pci-e? | 20:53 |
| sevan | I'm broke & excited! | 20:53 |
| sknebel | josch: wifi/bt on the SoM/adapter works better with the Pocket I guess | 20:57 |
| josch | oooh this is for the pocket! | 20:58 |
| minute | josch: wifi yes, bt no | 20:58 |
| minute | it's not primarily for pocket. firstly for reform | 20:59 |
| minute | but also what if you could put 2 ssds ~and~ wifi in your reform ;) | 20:59 |
| josch | indeed... freeing up the mini-pcie would be kinda sweet | 20:59 |
| josch | minute: one of the benefits of ls1028a was the larger address space and thus you made the experiments with gpu cards -- does rk3588 offer similar possibilities? | 21:00 |
| minute | josch: maybe, i still have to try that. the large number of pcie lanes suggests it... | 21:15 |
| minute | another option would be to only offer BT on the module (via uart) and wifi via mpcie... | 21:15 |
| minute | wouldn't work for pocket thooo | 21:15 |
| josch | i think having wifi/bt on the cpu module adapter is a creative idea i think! I only ever thought of the cpu adapters as physical adapters but of course there is a lot of real estate for more components on there as well... | 21:24 |
| josch | the iCore-3588Q looks tiny -- would there be enough area for offering easy access for things that the rk3588 offers but will not get connected? would there be enough space for a mini-pcie wifi card next to the iCore-3588Q? | 21:26 |
| josch | the m.2 connector is a bit smaller than the mini-pcie connector | 21:27 |
| josch | maybe the m.2 laird module would fit and make the whole wifi/bt thing modular and replacable as well? | 21:27 |
| minute | josch: it's bigger than the cm4! | 21:30 |
| josch | oh then i'm misled by the large mounting holes in the photos :D | 21:31 |
| minute | around 6.5 x 5cm | 21:31 |
| minute | yeah it has large holes and the SoC is _huge_ | 21:31 |
| minute | all those pcie lanes... | 21:31 |
| minute | :D | 21:31 |
| minute | the soc is like 24x24mm | 21:32 |
| josch | oh wow o0 | 21:32 |
| minute | ls1028a is like 16x16mm | 21:33 |
| josch | knowing that the thing in the middle is a square inch large puts things into perspective :) | 21:33 |
| josch | anyways, i'm super excited to see how well that platform will work in the end -- looking forward to your future updates on this! :) | 21:39 |
| minute | cool :3 i'm also really excited about this one | 21:44 |
| - eibachd (QUIT: Ping timeout: 256 seconds) (~eibachd@p200300dcf7353e00d60f808c3fd74c29.dip0.t-ipconnect.de) | 21:45 | |
| josch | was there any further process with snapdragon 8 gen3? | 21:46 |
| + eibachd (~eibachd@p200300dcf7353e000322003c9aa7e6a8.dip0.t-ipconnect.de) | 21:46 | |
| vkoskiv | I'm doing some software archeology, and stumbled on lsmem(1) and chmem(8). I've been reading the man pages for a while now, I still don't grok what the purpose of these is :D | 21:57 |
| vkoskiv | "ranges of memory"? "online status"? Why is a range exactly 128MB, and what happens if I use chmem to set some to offline? I got the sense that this may be for hot-swapping memory in a server or something, but I have no idea. | 21:58 |
| sknebel | vkoskiv: jup, thats hotplug support. the interface to https://www.kernel.org/doc/html/latest/admin-guide/mm/memory-hotplug.html | 22:01 |
| vkoskiv | For context, I'm compiling Linux 0.01 for fun and stumbled upon a chmem command in the main Makefile | 22:01 |
| sknebel | and 128MB just happens to be the block size on x86 (not clear to me if thats a hardware thing or just a kernel implementation choice) | 22:01 |
| sknebel | at least thats the context where I've seen those before | 22:02 |
| vkoskiv | I have to assume this requires hardware support too, I don't think consumer RAM is built to be removed at runtime :D | 22:04 |
| sknebel | yeah, although I have no idea what actually happens if you try on a consumer system :D | 22:05 |
| vkoskiv | I do have a random laptop I wouldn't mind trying this on for the heck of it | 22:16 |
| vkoskiv | An asus or acer I pulled out of the garbage some years ago. | 22:17 |
| minute | josch: snapdragon 8 gen 3 i don't have any access to. but still considering to do qcs6490 | 22:17 |
| josch | i guess in the end the platform with better mainline support wins | 22:22 |
| minute | yeah | 22:30 |
| minute | that is so | 22:30 |
| - mjw (QUIT: Ping timeout: 260 seconds) (~mjw@gnu.wildebeest.org) | 23:36 | |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!