| octavius | hi lkcl, read you got a new laptop. Which version of debian did you install? I'm thinking of re-purposing mum's old macbook air at some point | 12:13 |
|---|---|---|
| lkcl | debian/11. it was absolute hell | 12:18 |
| lkcl | it was one of the worst experiences of installing debian, in over 22 years. | 12:19 |
| lkcl | https://twitter.com/lkcl/status/1547565205386776578 | 12:19 |
| lkcl | the worst part is: given the choice i am forced to make the exact same choice again, because the alternatives are even more of a waste of my time both short and long term | 12:20 |
| octavius | Oh no | 12:22 |
| lkcl | wicd - the only non-KDE non-Gnome non-network-manager-integrated WIFI management GUI - has been removed from debian/11. | 12:23 |
| octavius | This system I'm running on uses systemd, but that's because I didn't have a clue as to how to switch to sysvinit. | 12:23 |
| octavius | Why has wicd been removed, not maintained? | 12:23 |
| lkcl | because it uses python 2.7 | 12:23 |
| lkcl | systemd is perfectly fine if you are fine with pottering's pathological behaviour | 12:24 |
| octavius | I'm not, but I'm also lazy XD | 12:26 |
| lkcl | https://cve.mitre.org/cgi-bin/cvekey.cgi?keyword=systemd | 12:26 |
| octavius | I did use to use arch and gentoo, but after working with you for a year I became content with debian | 12:26 |
| octavius | I think I'll try to do a sysvinit install for the spare laptop, since it's not needed for work atm anyway | 12:27 |
| lkcl | the outward consequence of pottering's pathological behaviour indirectly results in the CVEs you see ther | 12:27 |
| lkcl | you'll find it goes: | 12:27 |
| lkcl | "do you really want to remove kde, gnome, pulseaudio, xserver-xorg-core, firefox, chrome, gtkwave, .... .... ....?" | 12:28 |
| lkcl | Y/N | 12:28 |
| lkcl | yes it's really that bad. | 12:29 |
| octavius | One of my biggest annoyances is that since systemd became the de-facto standard, there are few tutorials/materials for managing services with other init systems. The learning curve is stupid, given that the simpler init systems are basically symlinks and shell scripts | 12:29 |
| lkcl | as i said in the link above i had to download xserver-xorg-core (apt-get build-dep xserver-xorg-core; apt-get source xerver-xorg-core) | 12:29 |
| lkcl | then hand-edit debian/rules.flags | 12:29 |
| octavius | for example I know system-ctl is the normal command people use. Whereas sysvinit took me far longer to find | 12:30 |
| octavius | Yeah, I remember gentoo being 'fun' to deal with those systemd deps | 12:30 |
| lkcl | there isn't one. you just do /etc/init.d/{name-of-service} stop/start/restart | 12:30 |
| octavius | Ah, I was thinking of the devuan one | 12:31 |
| octavius | ok | 12:31 |
| lkcl | as all of this stuff has been in place for 20+ years it's not actually that hard. | 12:31 |
| octavius | Perhaps I just need to go over it (which I will for the install) and write a blog post | 12:31 |
| lkcl | every /etc/init.d script is a bash script. | 12:32 |
| lkcl | how hard does bash have to be in order for the /etc/init.d scripts to be "un-understandable"? | 12:32 |
| lkcl | you get a problem you can't work out you just kill the daemon | 12:32 |
| octavius | huh, these scripts in init.d are still in a systemd system, are they used in tandem? | 12:33 |
| lkcl | it *doesn't* f*****g well get restarted under "managed" control, forcing you to desperately search for the command | 12:33 |
| lkcl | i don't know, and please don't take this the wrong way: i don't care and i'm not interested in finding out, either | 12:34 |
| octavius | sure | 12:34 |
| lkcl | :) | 12:34 |
| octavius | We got more interesting problems to deal with | 12:34 |
| lkcl | mhm | 12:34 |
| octavius | also added the env var to tas-yagle | 12:34 |
| lkcl | brilliant | 12:35 |
| octavius | Will update the dev script and test in a chroot | 12:36 |
| * lkcl melting my brain with subvl stepping in a precise-interruptible (FSM) fashion | 16:20 | |
| lkcl | can't just do "for i in range(VL): for j in range(SUBVL):" | 16:21 |
| lkcl | it has to be state-based: | 16:21 |
| lkcl | init(): SVSTATE.srcstep = dststep = substep = 0 | 16:21 |
| lkcl | step(): | 16:21 |
| lkcl | if substep < subvl: substep += 1 | 16:21 |
| lkcl | else: | 16:21 |
| lkcl | substep = 0; srcstep += 1; dststep += 1 | 16:22 |
| lkcl | if srcstep == vl: END | 16:22 |
| lkcl | where Pack/Unpack will REVERSE the ordering, ah ha ahahahahaha | 16:22 |
| lkcl | when Pack is enabled, srcstep gets incremented first (inner loop) and subvl gets incremented last (outer loop) | 16:23 |
| midnight | what are the tools for gate-level design these days? | 17:06 |
| octavius | midnight: That's what libre-soc uses: http://coriolis.lip6.fr/ | 17:07 |
| octavius | lkcl, how's the weather where you are? My car read 39C outside temp XD | 17:08 |
| midnight | octavius: Thank you. | 17:20 |
| lkcl | octavius, i'm trying not to open the door(s) to find out | 17:31 |
| lkcl | octavius, coriolis2 is a gate *layout* program, converting gate-level design or RTL automatically into GDS-II | 17:31 |
| lkcl | midnight, do you mean like schematics? | 17:32 |
| lkcl | octavius, i mean, you *could* use coriolis2 to actually perform the direct programmatic (python) creation of the netlists and the cells | 17:32 |
| lkcl | that's actually been done a number of times in coriolis2, for various reason (e.g. IO pads is the obvious one) | 17:33 |
| midnight | lkcl: I am looking specifically for raw, logic-gate arrangement design software-- no VLSI/VHDL/etc. Like.. AND gate here, NOR there, etc. | 17:33 |
| lkcl | geda. and such | 17:33 |
| lkcl | they're schematic layouts in which you will find IEEE-91 ANSI symbols | 17:34 |
| midnight | I build a truth table, I hand-optimize it with karnaugh maps or whatever, I arrange the gates on a grid or otherwise, maybe enclose in a glyph/package-looking icon for placement into other projects, etc. | 17:34 |
| lkcl | you likely want geda, kicad, etc. | 17:34 |
| lkcl | there are many more like that - we don't use any of them for this project | 17:35 |
| midnight | lkcl: There's some way to make such work useful in a libre-soc context? | 17:35 |
| lkcl | typically what i've done is done hand-drawn diagrams then back-converted them to HDL | 17:35 |
| lkcl | there were some online website gate-level design systems discussed here... err.... 18-24 months ago? | 17:36 |
| midnight | Hrm. If I designed a superior ALU for example, or a faster instruction decoder, getting it into a personal libresoc experiment would thus require some finagling in other words. Hey speaking of, you ever heard of the old-school https://en.wikipedia.org/wiki/Open_Graphics_Project ? I don't suppose any of that effort would be transferrable would it? | 17:38 |
| midnight | I have one of those cards still. | 17:39 |
| lkcl | yes, basically, although if it's in a machine-readable file-format then it could be converted programmatically (we do a LOT of compiler technology / language-translation) | 17:39 |
| lkcl | ha, ahh i remember that | 17:39 |
| lkcl | and Frank's not-GPL-GPU | 17:39 |
| lkcl | it was actually "the words of the GPL License with an additional non-free non-commercial clause tacked on the end" | 17:40 |
| lkcl | whoops :) | 17:40 |
| midnight | Was that in the OGD1 source somewhere? lol really brutal. | 17:41 |
| lkcl | https://jbush001.github.io/2016/07/24/gplgpu-walkthrough.html | 17:41 |
| lkcl | Frank Bruno was one of the developers of Plan9 (i think) | 17:42 |
| lkcl | ah interesting there's no mention there of the additional non-commercial clause https://github.com/asicguy/gplgpu | 17:43 |
| lkcl | which i definitely recall seeing somewhere on his website | 17:44 |
| midnight | lkcl: hrmmm.. I just compared the license file in that github with the gpl-v3 and looks like just formatting changes? Maybe he corrected that as a final send-off. | 17:47 |
| midnight | That would be nice of him. | 17:47 |
| lkcl | i honestly don't know, that sounds likely | 17:50 |
| lkcl | istrc it was GPLv2 when i investigated (5+ years ago?) | 17:51 |
| midnight | Anyway, really really glad this project exists. I'm usually pretty quiet but I'm a huge fan of the goals. Also Raptor's support is obv. helpful in differentiating the project from the unfortunately-common traps out there. So, seriously, thank you. When this becomes a product I can stuff in my computer room or into a rack, I'm going to buy so darn many, :-D | 17:52 |
| midnight | Oh, the kickstarter says it was planned to be under the lgpl. | 17:53 |
| lkcl | we don't use kickstarter - crowdsupply runs ethically-hosted campaigns | 17:54 |
| midnight | Indeed. | 17:54 |
| midnight | Oh! I meant the gplgpu thing linked above. | 17:54 |
| lkcl | they also don't "abandon" you. Joshua's great | 17:54 |
| midnight | I love crowdsupply stuff. | 17:55 |
| lkcl | well it helps also if you appreciate this has been going... er... almost 4 years? | 17:55 |
| midnight | https://www.kickstarter.com/projects/725991125/open-source-graphics-processor-gpu | 17:55 |
| midnight | Yes, I'm aware, and also patient. Since the pace is reasonable, I figured perhaps gate-level stuff might be feasible, hence my above question. | 17:55 |
| lkcl | yeah i did the LD/ST Computational Unit entirely gate-level, wrote it out 15 times | 17:56 |
| lkcl | niice. wow, 2014. | 17:56 |
| lkcl | ahh that _was_ frank bruno. | 17:57 |
| lkcl | Number9 Graphics, not Plan9 | 17:57 |
| midnight | Neat! | 18:13 |
| lkcl | https://libre-soc.org/3d_gpu/architecture/ | 18:28 |
| lkcl | where the hell is it :) | 18:29 |
| lkcl | ah. https://libre-soc.org/3d_gpu/architecture/6600scoreboard/ | 18:29 |
| lkcl | there we go https://libre-soc.org/3d_gpu/ld_st_comp_unit.jpg | 18:29 |
| lkcl | and the resultant HDL https://git.libre-soc.org/?p=soc.git;a=blob;f=src/soc/experiment/compldst_multi.py;hb=HEAD | 18:30 |
| lkcl | which is now slightly different as the HDL is about the 20th revision | 18:31 |
| lkcl | midnight, have you read "Design of a Computer" by James Thornton? | 21:54 |
| lkcl | http://archive.computerhistory.org/resources/text/CDC/cdc.6600.thornton.design_of_a_computer_the_control_data_6600.1970.102630394.pdf | 21:54 |
| lkcl | around p126-p127 is the scoreboard | 21:54 |
| lkcl | the world's first supercomputing OoO system, predating Tomasulo by over 5 years. | 21:55 |
Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!
