IRC logs for #openrisc Sunday, 2013-09-29

--- Log opened Sun Sep 29 00:00:17 2013
stekernpoke53282: apart from some minor hacks I had to do, the rest of the X build process went smooth07:14
stekernthere was a couple of places where I had to replace or32 with or1k "manually" in the make script, since autoregen overwrote them (tslib and Mesa)07:15
stekernand I had to manually edit dri_interface.h for it to not include drm.h07:15
stekernpoke53282: but when running 'make demo' it fails, because I don't have ldd and strace and it can't find libGLU and libGL in sysroot07:37
stekernhmm, this kernel doesn't boot on orpsocv3/sockit/mor1kx08:24
stekernjust hangs after "Switched to clocksource openrisc_timer"08:24
-!- olof is now known as olofk08:27
PowermaniacJuliusb: You attended the University of Queensland?How was it?09:44
PowermaniacHeh just being a cyber stalker and finding everyone on Google Plus or Twitter...09:46
juliusbPowermaniac: hi. I did indeed. it was fine. why do you ask?09:47
PowermaniacHanging around hacking forums makes you oddly good at tracking down people on the net (although most aren't hard to find)09:47
juliusbNo, it's not secret that I studied at UQ. I believe it's in the first 2 lines on my home page.09:47
PowermaniacYou have a site?09:47
juliusbI would be more impresed if you found my grades and forwarding address09:48
PowermaniacHeh, I prefer to just find people to make connections09:48
PowermaniacAlthough I know people who could do that...09:48 - it's where we also host the IRC logs so I presume you may have stumbled upon it there09:48
PowermaniacOr so I assume09:48
PowermaniacIs it costly to host a site/and the irc channel?09:49
PowermaniacNever really looked into having my own site09:49
juliusbPowermaniac: I don't host the IRC channel - does that and they are awesome for doing it. hosting a website is pretty dirt cheap these days09:49
Powermaniac(I'm also from/live in Australia see, that is why I was interested in the fact you went to the Uni of Queensland)09:50
PowermaniacReally like this community you have here, really nice/helpful people09:50
juliusboh cool. where abouts?09:52
PowermaniacI'm in South Australia09:52
PowermaniacNorthern Suburbs09:52
PowermaniacThat's as detailed as I'm going to get09:53
juliusbregarding UQ, I wouldn't say they have a stellar VLSI course, it's OK, but I believe they've scaled it back due to fewer people enrolling in it every year. they dropped the advanced subject they used to do, so I'd look elsewhere if you're looking at going to study VLSI09:53
PowermaniacAh okay09:54
PowermaniacWell I'm looking around as that will be a decision I shall have to make in 2 years time09:54
PowermaniacYear 11 now so09:54
PowermaniacWTF for a second there I thought Google Plus was telling me Linus Torvalds had my cousin in his one of his 'circles'09:56
PowermaniacStill not sure whether that was what it was telling me or not09:56
PowermaniacOn the note of university, people have been telling me as a backup plan to try and get into ANU09:57
PowermaniacI'm planning of going the root of electronic engineering, and then from there computational neuroscience09:57
Powermaniacjuliusb: Do you know of IBM's SyNAPSE project, and if so what do you think?10:00
juliusbPowermaniac: ANU could be good. You'd have to live in Canberra then, though.... :-/ but electronic engineering sounds good. tinkering with digital design can only help your understanding of things before you get to uni, and no, never heard of the SyNAPSE project10:02
Powermaniacjuliusb: Never been out of this state so I wouldn't know what Canberra is like anyway...10:03
juliusbolofk: would it be correct to say that no IP core source is stored in orpsoc-cores?10:03
juliusbPowermaniac: I'd reckon Sydney or Melbourne would be a better bet in terms of having a nice life while studying, Canberra is a bit small.10:04
juliusbolofk: actually, it wouldn't, would it? I can see RTL in cores/gpio10:04
Powermaniacjuliusb: Probably have to work on top of it to afford it all anyway. It is looking like I will probably stay here though as it is just easier for now anyway.10:05
olofkjuliusb: We store some RTL directly in orpsoc-cores. I call them "local" cores10:06
olofkMostly small utilities10:06
juliusbolofk: cool, thanks.10:08
Powermaniacolofk: Where are you from, as I haven't encountered the likes of your name very often? (Found you on Google+)10:08
PowermaniacOh and hopefully nobody here minds I've gone around following everyone in here that I can find...>_>10:09
juliusbPowermaniac: the UniSA computer systems engineering course looks pretty good10:09
Powermaniacjuliusb: They are also supposedly the only state with a computational neuroscience 'lab' not sure what to call it exactly10:10
juliusbPowermaniac: yeah, that electrical engineering (computer systems) looks very similar to what I did at UQ, in the final year you don't get as many advanced VLSI courses, but I don't think UQ does that anymore anyway, as there's no industry in australia :( but... doesn't stop you going overseas and contributing to the brain drain :)10:11
Powermaniacjuliusb: Wait there isn't any industry here at all? Hmm didn't know that till now...10:12
juliusbit's got some cool stuff in years 2 and 3, and the final year project looks pretty flexible, you could probably choose somethign which goes in the direction of what you're planning to do after undergrad10:12
juliusbPowermaniac: well, almost no chip design stuff, but there certainly is embedded system development, lots of it10:13
PowermaniacWell I sort of now want to work at IBM for the simple fact they seem the furthest ahead with designing hardware for AI well to replicate the brain on.10:14
PowermaniacSort of the dream job at the moment10:14
Powermaniacjuliusb: You're looking at the UniSA site right not the Uni of Adelaide site?10:15
juliusbyes, unisa10:16
PowermaniacHmm that doesn't have very high requirements compared to Mechatronics that I was looking at earlier in the year10:21
Powermaniac(Refocused half-way through the year)10:21
juliusbolofk: if I don't install the orpsoc tool, can I just set ~/git/orpsoc/bin in my $PATH?10:26
juliusb... and that'll work?10:26
juliusbolofk: and is most of what is on the orpsocv3 wiki page now basically inaccurate?10:35
stekernwow, this is beyond cool:
juliusbmmm, somethign went wrong with my first attempt at buildling:
stekernjuliusb: you should be in build-orpsoc10:48
juliusboh right10:48
stekernor whatever you call your build dir10:48
juliusbi have build/ under orpsoc-cores/10:48
juliusbi have to be in there?10:48
Powermaniacstekern: Is that what I think it is, Mor1k running a 'full' system?10:49
juliusbstekern: ah, I see INSTALL will probably tell me what to do...10:55
stekernjuliusb: you are doing it right, try first, read instructions later ;)10:56
stekernPowermaniac: yes, X running on mor1kx10:57
Powermaniacstekern: That is awesome!10:57
stekernon orpsocv3 on the sockit board ;)10:57
stekernjuliusb: I did exactly the same mistake (it was actually that that olofk told me was uncool), I tried building in the orpsoc-cores directory11:05
stekernwhich I think people should be able to do, but as olofk mentioned yesterday, there's some bug preventing that from work11:06
juliusbi have discovered another bug, which is that if I have ~/git/blahblah as cores_root and ssytems_root it breaks11:07
juliusband I can fix it so it works initially, but then it breaks further on...11:08
juliusbanyway, will sort that, should be simple11:08
stekernyou mean the ~ breaks things?11:08
juliusbyes, we need a os.path.expanduser() somewhere on the path11:08
stekernit works if you have /home/jules?11:09
juliusbwell, on the path from that variable being pulled out of the *_root strings and into the system11:09
stekernI see11:11
PowermaniacSo do you guys also work?11:13
juliusbI have a hard time believing stekern does based on his prolific output of prodigious work, but I'll speak for myself and say yes, I also work11:16
stekernnot on sundays11:17
PowermaniacI was going to say, how do you manage if you don't work, heh.11:17
stekernjuliusb: ah, but admit that it's not just work that intrude on your openrisc hacking12:24
stekernyou just have a more exciting life than I ;)12:25
PowermaniacOh really ~_~12:25
juliusbhaha, true, ever since I joined Mossad my weekends have been taken up with other things12:25
juliusboh...whoops, they told me never to say that in public....12:26
juliusbso, a heads up, i'm doing a fairly big update of the ORPSoC wiki pages on OpenCores12:26
PowermaniacJust googled it...O_O12:26
juliusbbasically moving all of the ORPSoCv2 stuff to its own page and making the ORPSoC page the main one containing info about v312:26
PowermaniacThe Israeli NSA...Basically12:27
juliusbPowermaniac: there was a high profile case recently of an Australian who worked for Mossad and eventually ended his own life in prison after he was jailed for screwing up a few major operations they had12:27
PowermaniacHoly shit12:28
PowermaniacReminds me how I've been thinking of asking my cousin about the NSA (He works for Google) thinking it probably isn't a good idea...12:29
juliusbGoogle "Prisoner X"12:29
PowermaniacHmm that is kind of creepy12:30
PowermaniacDare I ask you what you think about all this NSA news?12:31
PowermaniacI personally suspected it, but the proof sort of has made me try and change how I do a lot of things, slowly but surely trying to go entirely open source and learn more so I can monitor what things are doing12:32
PowermaniacI wonder if Snowden has any information on Prisoner X, as the Israeli government and US government are pretty close..12:33
juliusbI'm not surprised, and strangely I'm not too bothered by it, but it does seem like something which could be tempting to abuse if one worked there.12:34
knzjuliusb: which version of orpsoc are we going to do a hands-on with next weekend?12:34
PowermaniacI don't quite get why the guys in #OpenRISC aren't bothered by it12:35
juliusbPowermaniac: there's little mystery around that now - several leaks basically allowed journalists to put the whole story together and there was a good Foreign Correspondent episode on it.  Anyway this is all getting quite off topic now. I'll leave it there.12:36
juliusbknz: 3!12:36
PowermaniacOh so you guys don't like it when I get off topic. Was wondering that but hadn't yet brought it up.12:37
juliusbbrief tangents, no probs but there's probably a limit, don't ask me to define it though12:37
PowermaniacFair enough12:38
knzjuliusb: thx :)12:39
Powermaniacknz: Are you a professor, well would you classify yourself as one?13:26
Powermaniacknz: Besides that I was also going to ask your opinion on the book Practical Electronics for Inventors, so what do you think about it?13:27
PowermaniacWell forget it, 1. it's sort of off-topic anyway. 2. I'm out for the night.13:47
PowermaniacGoodnight everyone13:47
stekernjuliusb: orpsoc wiki page update, sounds great14:42
juliusbstekern: once you build the de0_nano image with ORPSoCv3, how did you download it to the board?14:48
stekernyes, with quartus_pgm14:50
juliusbok, just checking there's no way of doing via orpsoc just yet...14:52
stekernthere's a patch by franck pending to add that funcrionality14:52
juliusbah right14:52
stekernthat's what olofk was talking about that he wa,ted to check before doing a release14:53
juliusboh we can manually do it for now with v3.114:53
stekernI also wanted the option to chose cable, but perhaps it just me that have several boards connected at the same time14:54
juliusbquite possibly :)14:55
juliusbI'm wondering about where the best place to put the workshop materials14:59
juliusbmaybe do up a nice page on the OpenCores wiki, or use the github orpsoc wiki?14:59
stekernI'd say on the orconf wiki page15:00
stekernwell on a subpage perhaps15:00
juliusbOK, but in that case I'll make it a more generic page which isn't ORCONF-specific, and is just a general "ORPSoC(v3) on DE0 Nano"15:01
juliusbspecifically, "ORPSoC(v3.1) on DE0 Nano (as seen on TV rebroadcasting of ORCONF)"15:01
knzjuliusb: that would be great :)15:04
stekernah, yeah that make sense15:05
juliusbhmmm, franck's openOCD doesn't build for me15:06
juliusbjust junk like or1k_jtag_adv.c:132:18: error: variable 'node' set but not used [-Werror=unused-but-set-variable]15:07
* juliusb unleashes --disable-werror15:13
stekernhe probably have an older gcc not erroring out on those15:13
juliusbya, likely15:14
juliusbso we _really_ need to do in orpsocv3 is allow the config in the board config file to be turned into `defines so we can get rid of the orpsoc-defines.v file!15:15
juliusband im sure we can eventually get some gui thing which can edit those config files15:16
juliusb... and am I right in thinking that the interconnect in this de0 nano is now autogenerated?15:17
stekernyes, the config is in data/wb_intercon.conf15:19
juliusbso so cool!15:19
stekernand the generator is incores/wb_intercon15:20
stekernyes, olofk did a great job with that!15:20
stekernwith board config, you mean the .core/.system file?15:22
juliusbwell, something shoudld allow us to control the precise config of the system, right?15:22
juliusbi mean, the exact config at synthesis15:22
juliusbinstead of storing that in verilog15:23
juliusbwe should have the source in the config file15:23
juliusbs/the config file/a config file/15:23
stekernhmm, yeah, but I don't think we have the same information in any toher file atm15:24
stekernit could of course be , instead of the .v15:25
juliusbsure, i'm saying long term, we should migrate away from having the info in orpsoc-defines.v and probably just autogenerate that file\15:25
stekernthat would make it easier to export the config in other formats15:27
juliusbthen you can control generation of IO constraints and the like based on whether the core was enabled, what its parameters were, etc.15:28
stekernlike an .dts for instance15:28
juliusbhad to do it a gross way in Make previously :-/ basically process the verilog, turn it all into environment variables, then use the shell environment to do the conditional statements.15:28
stekerni at least purged the orpsoc-params.v in de0 nano15:32
juliusbthank you15:33
stekernah, while I remember, if you want to do your fancy "tap demo", you'll need to add spi15:40
juliusbah right15:40
juliusbwe have UART, right?15:40
juliusbUART and GPIO flashing LEDs could be enough15:40
stekernI thought adding SPI would be a good chance for you to quickly get acquintez with the system though ;)15:42
juliusb... if I have time :0/15:43
juliusbi should do, though15:43
juliusbI have all week to finish the tutorial, just need to get some initial tool chain set up instructions to people15:44
juliusbok, if I use the new tool chain you put together stekern it dies looking for shared object....?!15:46
juliusbI have libgmp10 and libgmp-dev installed on my system15:47
stekerncan't copy-paste from this phone, will look later15:48
juliusb/opt/gcc-or1k-elf-4.8.1-x86/bin/../libexec/gcc/or1k-elf/4.8.1/cc1: error while loading shared libraries: cannot open shared object file: No such file or directory15:49
stekernso, what are your libgmp called?15:50
stekernI did the x86 on a not-so-up-to-date debian machine15:50
juliusbthat's cool, I can give it a go buildling the x86 one15:52
stekernuse the 4.8.1 branch of gcc then15:53
juliusbis there anything special you do to compile it so it works well on other people's machines?15:54
stekernI have no clue ( as you just noticed) =p15:55
juliusbhaha, well, not sure you could do anything abou tthat15:55
poke53282*yawn* *stretch* Good Morning! :) I will take a look when I am really awake stekern.16:20
stekernpoke53282: no hurry, everything got sorted out, I've got X running on my boards ;)16:34
stekernah, well, actually, if you could give a hint about the GL(U) stuff, that'd be great16:35
stekernagain, a superb job putting together those scripts and hunting down bugs along the way to make it possible!16:36
stekernand good morning ;)16:38
poke53282make mesa; make glu17:02
poke53282should be no problem.17:03
poke53282Thanks stekern.17:03
poke53282Is glxgears working?17:03
poke53282I don't think you need glu for that.17:04
poke53282And I think you can skip the whole tslib stuff. I thought you have a mouse.17:04
poke53282In the picture you send a few hours back I think the mouse didn`t work so far. :)17:07
poke53282olofk, juliusb, stekern, knz, powermaniac, jonibo, blueCmd and whoever reads this: At the moment I am planning a second image for my emulator with some development software. Basically gcc and make. I am searching for nice and funny stuff to give the users to compile and explore that is all together not larger than 1MB. At the moment I have  helloworld.c and fbdemo.c. I am thinking of some bsdgames or something from the obsfucated code contes17:15
knzpoke53282: a simple sudoku solver17:15
knzpoke53282: a mandelbrot set computation17:15
knzpoke53282: figlet / banner17:16
knzpoke53282: xlock with all the animations17:17
poke53282Nothing with X. This will be too large. gcc and X together. Then the user has to download 20MB.17:18
poke53282That are all nice ideas.17:22
poke53282But with this programs I can never fill 1MB ;)17:22
_franck_juliusb: openocd from the official repo doesn't compile ? It has been tested by the gerrit build server on 5 different hosts/compilers17:36
stekernpoke53282: I thought I tested 'make glu'... oviously I was hallucinating17:54
stekernand I use tslib on the de0 nano, so I wanted that anyway17:54
stekernhaven't tested on that board, and no mouse (or keyboard) on the sockit yet17:54
stekernI have no access to pin-headers nor usb from the usb side, so I'm thinking about adding a "ps/2 emu" core17:55
stekernjust a register that can be written over the wishbone bus17:56
stekernand then have a server running on the arm core that will write values into that that clients write17:56
stekernthat would be pretty flexible I think, I could forward the input from some USB-hid device for example17:57
stekernor have the client on my workstation17:57
stekernactually, I could even have the server running on or1k17:58
stekernat least on the boards with ethernet17:59
stekernwhich makes me wonder again if something like this already exist...17:59
stekernshould be simple enough to write one myself though17:59
stekerns/from the usb side/from the fpga side18:00
stekernI could also just compile x11vnc and get the mouse/keyboard input from that18:01
stekernbut then it would be x-specific18:01
stekernnow I got mouse working (on the atlys board):
stekernI had to add '--enable-kdrive-mouse' to the xorg-server config18:51
knzpoke53282: typespeed (typing game), aalib with its demos19:01
knzpoke53282: if you want to turn your 1MB into something *useful* I would suggest ppp/pppd, slip, netcat, nmap, minicom19:02
knzthat would make it a nice hacking station19:02
knzand telnet and ssh of course19:03
poke53282knr: No network available. And this special image is supposed for compiling stuff. And not for networking. These tools can be part of another image.19:06
poke53282typespeed is a nice idea19:06
poke53282and aalib too :)19:07
poke53282stekern: Great. Looks fantastic. One question. If you close the programs is Xfbdev crashing. What is the speed of glxgearsÉ19:07
poke53282stekern: xtern should work too. But my emulator is missing keyboard input. So I couldn't test it.19:22
poke53282You can try this. At the moment it is crashing after some time, before the output. So I am curious if this works for you19:25
stekernglxgears are running slooow, a couple of fps only20:13
stekernand yes, closing programs make Xfbdev die20:15
poke53282a couple of fps is much better than I have20:16
stekernmake[1]: *** No rule to make target `prepare'.  Stop.20:18
stekernah, it's not going into that dir20:19
stekerncd src/$@; make CC=$(CC) USE_X86_ASM=N STRIP=or1k-linux-uclibc-strip prepare V=1 tyr-qwcl20:24
poke53282stekern: One question for the git expert. You know that my repository contains binary files and they could have a size up 10MB. Every change increases my repository by a few MB. I don't want this. Can I exclude those files from commits but still keep a directory with those files in the repository? I know that contradicts in some way the philosophy from git.20:57
poke53282I am afraid that my repository will have a size above 1GB in the future. The problem is that the website gh-pages is done via git.21:02
stekernumm, yeah, but if you exclude them from the commits... they are not in the commit21:04
stekernnot sure what you meant21:04
poke53282I want to change a binary file in the repository without committing.21:05
stekernI don't think that's possible, but if you're not interested in the history, you can of course just rewrite it everytime and add that binary in the last commit21:07
poke53282I want that the repository has a folder which is ignored but contains files.21:07
stekernmmm, but you *have* to commit it some time have it in the repository21:08
poke53282That's the problem. gh-pages works like this.21:09
stekernbut if you do what I said above, you kind of get that21:09
stekernso basically every commit, you ammend the last commit and remove the binary from list of added files and commit that21:10
stekernand then add it to your latest commit and commit that21:11
olofkstekern, juliusb : I added the pgm patch from _franck_, so you should be able to do orpsoc pgm de0_nano --whatever-options with orpsoc 3.121:12
* stekern deletes his script21:13
poke53282Thanks stekern. I try it.21:13
stekernnow I just have to add qsys support so I can delete my script21:14
stekernpoke53282: so is it just suppose to build tyr-qwcl?21:19
juliusb_franck_: yeah, it was
juliusbmy gcc is gcc (Ubuntu/Linaro 4.7.3-1ubuntu1) 4.7.321:34
poke53282I build all of them.21:37
juliusbstekern: looking good man21:37
poke53282Forgot the correct name21:38
--- Log closed Mon Sep 30 00:00:18 2013

Generated by 2.15.2 by Marius Gedminas - find it at!