IRC logs for #openrisc Tuesday, 2014-09-23

--- Log opened Tue Sep 23 00:00:06 2014
stekern"Die IT-Evolution" has a certain ring to it when read in english04:41
olofkhmm... how can I express a negative value with a bit width? 12'd1772 works, but 12'd-587 doesn't08:08
stekernolofk: -12'd587 ?09:59
sb0_olofk, if you're doing signed arithmetic, use migen.10:53
sb0_the verilog rules are completely messed up and you *will* write bugs.10:53
olofkGahh.. signaltap is driving me crazy11:42
olofkSorry signaltap. I didn't mean that. My bad11:47
mithroolofk / blueCmd: If you added me an owner of the I would happily script up and automated push from svn to github. I already have most of a script which would do that. I can give you access to the machine which runs it too if you wanted.13:53
blueCmdmithro: sure, I'd be happy to do that, but I'd like to review how you do the svn->git conversion first13:54
mithroblueCmd: my plan is the following, use svnsync to pull each of the core repos locally, then use git-svn to create the resulting git repo13:55
blueCmdalso, I'm intrigued how you got around the brokenness of the SVN server, a lot of repositories I tried to migrate just didn't work with that the SVN server either timed out or replied with a HTTP error code (I don't recall the exact code)13:56
blueCmdsvnsync is pretty clever13:56
mithroblueCmd: I think svnsync has been built to deal with crappy links / servers13:56
blueCmdgit-svn isn't what i've been using though, that will not preserve branches and tags right?13:57
mithroblueCmd: as the time you really want to use svnsync is when your on a crappy connection :)13:57
blueCmd(I don't think it's git-svn I've been using anyway)13:57
mithroblueCmd: dunno what you have been using?13:57
mithroblueCmd: have you been using talyor?13:57
blueCmdno. I don't recall the name, give me a sec13:57
blueCmdsvn2git I think13:58
mithroblueCmd: from the readme for svn2git13:59
mithroblueCmd:  It uses git-svn to clone an svn repository and does some clean-up to make sure branches and tags are imported in a meaningful way.13:59
blueCmd~/svn2git/bin/svn2git$1/$1 --verbose --log-window-size=113:59
blueCmdI know it uses git-svn as a backend14:00
mithroblueCmd: okay14:01
mithroblueCmd: seems pretty simple for me to use that instead of git-svn directly14:01
blueCmd with that branch, but that's only because I needed log-window-size to be able to use the crappy server14:01
blueCmdif you're going to sync it, you will probably want to enable the metadata which svn2git disables by default though14:02
blueCmdto not force-push all the time and rewrite everything every push14:02
mithroblueCmd: okay14:02
blueCmdmithro: are you 'mithro' on github?14:03
mithroblueCmd: yeah14:03
blueCmdthere you go14:03
blueCmdplease don't wipe all the data :)14:03
blueCmdalso, thanks for doing this14:04
mithroblueCmd: no worries14:04
mithroblueCmd: Open Source hardware makes me happy, so I want to help make it happen :)14:05
blueCmdwell you're in the right place then :)14:06
mithroblueCmd: my current primary FOSS project is TimVideos (over in #timvideos) - we are working on trying to record and live stream (plus make it interactive) every FOSS event. Part of that has been working on an FOSS hardware capture system for HDMI/DVI14:11
blueCmdmithro: yeah I saw some hdmi2usb thing on your blog14:12
mithroblueCmd: ahh cool, stalked me a bit huh? :P14:13
blueCmdolofk: I think I'm going to tear out the floating point stuff in GCC tonight and see what the damage is14:13
blueCmdmithro: basic background check before I give you commit access :)14:14
mithroblueCmd: yeah, due diligence and all14:15
mithroblueCmd: I haven't figured out a great way to "backup" the opencores website. I tried httrack and it seems to have done an okay job but it is obviously limited by being a webscraper. It also can't access any of the authenticated only sections.14:24
blueCmdmithro: I just used wget I think14:36
sb0_the mixing of temporary/callee-saved registers in the ABI is a (small) mess15:33
blueCmdolofk: so GNU assignment office has confirmed that everybody except Matjaz has completed the assignment process15:49
poke53282sb0_: We know.15:49
blueCmdolofk: I did get an email from Matjaz and I did ping him 10 days ago, so I think he's just stalling15:49
blueCmdnot intentionally15:50
blueCmdso next step is to find a GCC reviewer15:50
blueCmd(middle step is for me to remove the floating point code)15:50
poke53282sb0: But just remember odd and even, and it is more or less Ok.15:52
sb0_let's see how much bloat, slowness, bugs, design flaws and software toolchain shittiness this riscv thing has...15:58
poke53282I would like change a lot in the OpenRISC specification.16:10
poke53282but then you have to change all code.16:10
poke53282And I think there is someone missing who does the final decision16:14
poke53282someone build even an riscv emulator in Javascript already. :)16:16
poke53282That was fast.16:16
poke53282Well, I hope hear something at orconf about this cpu.16:18
poke53282sb0_: I will definitely take a look at this cpu. But when I take a look at the javascript I have to say, that OpenRISC is currently ahead.16:54
poke53282not even /proc/cpuinfo works.16:59
poke53282But this is from March this year.16:59
poke53282What I like as a someone who is doing numerics is, that they support quad precision floating point as an option. I wish Intel would do it.17:08
stekernALSA device list: #0: 95000000.wb_i2s-ssm2602-hifi17:10
stekernprogress at last17:10
poke53282with your DMA implementation?17:11
stekernnow I just need to make it actually work too...17:11
stekernyes, with the dma driver17:13
poke53282I thought about the boot discussion and the dts compiler. I think, I will try it.17:32
poke53282If there is a way to alter the dts text file before boot it would be great.17:33
olofkmithro: Have you considered showing up at orconf too?17:34
olofkblueCmd: Good news. Have you had time to try removing the fp stuff?17:35
mithroolofk: I assume that orconf is not in Australia?17:36
olofkmithro: Germany isn't that far off17:37
mithroOnly a minimum of 22 Ish hours flying.17:39
PaulfraOSAAolofk: I don't assume you have a sim file for your lx9 implementation?17:39
PaulfraOSAAWhen I try it, it kinda hangs17:39
mithroAre you guys planning on recording the stuff at all?17:39
olofkPaulfraOSAA: Ah, there you are. I had some comments on your wiki, but I completely forgot what it was17:40
PaulfraOSAACool ;)17:40
olofkPaulfraOSAA: What do you mean with sim file?17:40
PaulfraOSAAI probably did something wrong17:40
PaulfraOSAAWait a mo, just setting the or1ksim environment up17:41
olofkaha or1ksim config file?17:41
olofkmithro: Yes, we have recorded the previous years17:41
olofkNo live feeds though17:42
PaulfraOSAAyeah, it kinda complains about reading 0x800000 or something17:42
olofkLive feeds are for sex shows17:42
mithroCool. Considered live streaming?17:42
PaulfraOSAAWhen using the standard sim stuff17:42
PaulfraOSAAI'm still just trying to do the asm stuff17:42
mithroWhere are the previous recording hosted?17:43
olofkhmm.. is that address inside the RAM?17:43
olofkPaulfraOSAA: That address is just outside of an 8MB RAM (which happens to be the default value in or1ksim)17:45
stekernpoke53282: do you have alsautils et al in some nice packages?17:46
olofkMake sure to set the RAM size to 32MB, or whatever is on the lx917:46
PaulfraOSAAyes, and that is also the size of the [slave mem]size= setting17:46
poke53282stekern: yes17:46
poke53282was already there17:47
PaulfraOSAAolofk: anyway I use the pad to 512 option when compiling17:47
poke53282alsa.tar.bz2 and alsa-utils.tar.bz217:47
poke53282I would suggest to use "speaker-test -t sine"17:47
olofkPaulfraOSAA: Hmmm... did I ever get around to fix the memory interface for lx9, or is it still only using block RAM?17:48
mithrowell it's past my bedtime, time for me to sleep.17:48
poke53282stekern: But I removed some config files from the libalsa package. He doesn't seem to need them for my config. Just in case it asks for some missing file. I have all of them :)17:49
stekernah, I might compile it myself at a later stage17:49
olofkmithro: Good night and just ping us if you happen to be in Europe around 11-12 October17:50
stekernI just want to get to a point of failure that I can poke at now17:50
poke53282This is what I remove17:50
PaulfraOSAAolofk: well that shouldn't really matter when I'm trying to run 10 lines of assembler in the simulator, now should it?17:52
poke53282stekern: And these are my config options. Completely harmless17:52
olofkPaulfraOSAA: Can you send the code?18:02
olofkOr the output from or1k-elf-objdump -d if it's just a small asm hack18:03
poke53282stekern: Are these the latest u-boot patches?
PaulfraOSAAolofk: It's at the wiki now, do you need the url again?18:10
poke53282stekern: I see, the patches are in the mainline.18:17
stekernpoke53282: yes, everything openrisc related should be in mainline18:19
stekernthere are a couple of drivers that aren't though...18:19
stekernmy drivers are starting to fail in a way that makes sense now18:21
stekerni.e. it tries to run my stub functions to setup DMA transactions18:23
poke53282You can still try to use my code which I pasted two days ago. Then of course you have to handle the transfer without dma, or you have to program the dma device directly in your sound driver.18:30
olofkPaulfraOSAA: Got the address, but don't see any code18:31
PaulfraOSAAIn the bottom?18:31
blueCmdolofk: still @ work18:31
olofkPaulfraOSAA: I can run it in or1ksim, but nothing exciting happens when I do18:39
PaulfraOSAASorry was away soldering19:03
PaulfraOSAAHow do you run it?19:04
poke53282stekern: cat /proc/cpuinfo doesn't work any longer when I take the smp branch but without smp. I didn't test with smp.19:05
poke53282of course could be my error somewhere.19:11
poke53282I will tell you19:11
blueCmdolofk: stekern:
blueCmdseems to be working as far as I can see20:17
blueCmdplease test20:18
PaulfraOSAAHmm, If I say I lost my password on opencores, get a reset password and put in another username, It seems I can access any account I want. I just logged into some poor bloke from Poland's account using that method20:25
blueCmdPaulfraOSAA: haha, awesome20:26
PaulfraOSAAHow do I get the or1k-sim to debug from adress 0x10020:33
blueCmdstekern: what the f* are we going to do about the tests for gcc?22:23
blueCmdwe want 100% test passing or disabled for everything C and C++ I guess22:23
poke53282blueCmd: Just give me one gcc or g++ testsuite problem that fails.  Then I will investigate this.  Ok, I have already two from the last three weeks. But they are hard to investigate. Maybe you have a important one that fails.22:36
blueCmdpoke53282: is what I have, but a lot of those are just test harness failures so verify that they indeed do fail before digging too deep22:38
blueCmdbut as the path says, that's from jul-25 as well - but I don't know if any real work has been done since22:39
poke53282well, we upgraded to 4.9.122:39
poke53282or at least I did22:39
blueCmdright, but I don't think that will change much22:39
poke53282me too22:39
blueCmd has some annotations of some older problems that still exists22:40
blueCmdI tried automating the testsuite runs, and I got pretty far but since I got timeouts/lockups in qemu's network interface it's hard to run22:43
blueCmdanyway, time to sleep22:45
poke53282gcc.c-torture/execute <= this one was already successful executed some time ago.22:54
poke53282and the memcmp.c code works22:54
poke53282this is the first error in your list22:54
poke53282by the way. My network works flawless.22:55
poke53282and those on the uclibc website are also outdated.22:59
--- Log closed Wed Sep 24 00:00:07 2014

Generated by 2.15.2 by Marius Gedminas - find it at!