IRC logs for #openrisc Tuesday, 2013-08-20

--- Log opened Tue Aug 20 00:00:19 2013
stekernit's perfectly normal that tlb_miss stay X, until the user invalidates the tlb entries02:15
stekern...but not so normal when the mmu isn't included in the "build"...02:39
mor1kx[mor1kx] skristiansson pushed 5 new commits to master:
mor1kxmor1kx/master d1536be Stefan Kristiansson: cappuccino/fetch: assign 0 to mmu output signals when FEATURE_IMMU=="NONE"03:09
mor1kxmor1kx/master 443d973 Stefan Kristiansson: cappuccino/lsu: assign 0 to mmu output signals when FEATURE_DMMU=="NONE"03:09
mor1kxmor1kx/master c0f709a Stefan Kristiansson: cappuccino/fetch: set tlb_reload_ack to 0 by default03:09
stekernolofk: try now03:10
stekernwe really need an automated build system that tries all the possible and impossible combinations of parameters and simulation and synthesis tools03:40
blueCmdstekern: have anybody synced or1k-src lately?09:25
blueCmdfrom upstream i mean09:25
stekernblueCmd: it was a while ago, it's been on my list to do09:43
blueCmdstekern: cool, i had a problem that could sort of be explained by an old toolchain, but I think it was a make -jX dependency problem09:44
blueCmdseems to work now09:44
stekernhas there been major changes in upstream or1k-src lately?09:44
blueCmdI have no idea09:46
blueCmdbinutils changelog doesn't show more than a handful of commits since may09:48
stekernok, I thought you had something special in mind for your bug09:48
blueCmdso probably not09:48
blueCmdstekern: nah, I was just exploring causes09:48
stekernI'd imagine that major changes are target specific09:48
stekernI see09:48
stekern*the majority of changes09:49
stekernso I haven't been worrying so much about that being in sync with upstream as with gcc09:50
blueCmdI can see why09:51
stekernbut it should be done, of course09:51
blueCmdfun thing, I realised that I should probably port glibc instead of eglibc since eglibc syncs from glibc. Read a few mailinglist threads about new archs and they were all pointed to submit to glibc. Copied my work straight from eglibc to glibc and it appears to compile.09:52
stekernone of those days in the near future I (or someone else), will for sure find a gap in their schedule to do it ;)09:52
stekernah, cool!09:52
stekernI reckon you're done with your thesis since you are touching or1k stuff? =)09:53
blueCmdyeah, I'm bored as heck :P09:53
blueCmdSo I figured I might was well do something09:53
stekernwell, congrats on finishing it off!09:55
stekernthe thesis, that is09:55
blueCmdthanks, I still have to present it and probably fix a few things, but the examiner taking ages on reviewing it09:57
stekernand I guess for (e)glibc, it probably makes sense to do development straight against glibc, we don't really have any space constraints in our development environments09:58
blueCmdyes, and anyway even if it's needed to have eglibc AFAIK it will follow automagically09:58
stekernyup, that's what I understood from what you were saying09:59
blueCmdstekern: have you tried building the kernel with the latest gcc lately?10:19
blueCmdkernel/rtmutex.c:676:1: error: inlining failed in call to always_inline ‘rt_mutex_slowtrylock’: indirect function call with a yet undetermined callee10:19
stekernhmm, actually, no10:21
stekernI've been using jonas tree lately, and have just used the or32-linux default there10:22
stekernis that due to a change in the kernel or in gcc?10:23
blueCmdah, yes both jonas and my old 3.6.10 kernel fails to build using my freshly pulled or1k-gcc10:23
stekernok, so in gcc10:23
stekernthat is 4.9.0 you are using, right?10:24
stekernwonder if the problem exist in 4.8.1 too10:24
blueCmdgcc version 4.9.0 20130716 (experimental) (GCC)10:24
blueCmdI can switch branch and test10:25
stekernthat'd be great10:25
stekernI should sync gcc soonish again too, since there is that fix for building a native or1k gcc upstream now10:26
blueCmdI never got that to work, great knowing that works10:28
blueCmdwill help in porting debian later :P10:28
stekernI haven't tested the patch that actually went in to fix the problem, but I think it should be good10:29
stekernI had a slightly different patch that fixed the problem in another way, that I've tested10:29
stekernpoke53282 found the same problem in directfb too10:30
stekernit's related to our ABI for var args, I think he has a patch in his gcc that changes the ABI to be more like x86/arm etc10:31
blueCmdyeah I had to hack around that a few times in glibc10:37
blueCmdthat fix might render those hacks obsolete10:38
blueCmdthat would be sweet10:38
blueCmdstekern: 4.8.1 works10:38
stekernhmm, ok, perhaps a momentarily upstream problem then11:04
stekernexactly why I wanted to do that 4.8.1 branch11:05
stekernto have something stable to compare against11:06
blueCmdI want to buy myself a big bad buildserver11:06
blueCmdhah, cool - it worked. running openssh and zsh using glibc12:30
stekernI want to get apache up and running, afaict it's shortcomings in uclibc that prevent me from it14:12
blueCmdI can try to compile it15:39
blueCmdstekern: compiled and running :)16:31
olofkams: Did you just endorse yourself on linkedin?16:33
stekernblueCmd: cool!16:34
blueCmdyes, things seems to be even more stable than I left it in march, so that's good16:38
blueCmdI still cannot use jonas linux due to eth0 not being detected with or1k sim though16:38
blueCmdit is detected with pgavins linux16:38
stekernI don't know what's the problem there is, works perfectly here16:42
stekernboth in or1ksim and on real hw16:43
blueCmdhm, really? odd16:43
blueCmdstekern: you're using the compiled version right?16:43
blueCmd(as in pre-compiled)16:44
blueCmdI should probably talk to jonas about that and solve that once and for all. I need to merge some stuff upstream in the kernel so that would be the nicest way to do it16:50
stekernprecompiled what?16:51
stekernno, the one from github16:51
blueCmdlinux kernel16:51
stekernumm, precompiled linux kernel? now I certainly fell of the wagon ;)16:51
poke53282Ok, have not read the whole chat. But I have a full working patch for the varargs problem.16:52
blueCmdah, I thought you were using a precompiled (by jonas)16:52
poke53282A better one than what I send here a few weeks ago.16:53
blueCmdstekern: if you're compiling it yourself I would love to have your .config in case it's something wrong with mine16:53
poke53282Actually I wanted to test it a little more but had no time. I think I will just put the current patch on the mailing list.16:53
stekernpoke53282: yeah, put it out as an RFC, then at least it's saved for the future17:00
blueCmdstekern: still using or32-linux as the target for linux?17:04
stekern13:22 < stekern> I've been using jonas tree lately, and have just used the or32-linux default there17:15
blueCmdstekern: thanks. I get some weird linking errors with your config, disabling module support fixes it. your config without modules works which is cool - now I have something to work with and debug why mine doesn't. Also, I discovered that I was running on a weird branch.17:20
blueCmdworks as in: i have eth017:20
blueCmdhuh, seems that mine works as well now. maybe it was only the wrong branch that was the problem. oh well, great17:25
olofkstekern: mor1kx works fine now in orpsocv317:30
olofkNow juliusb and stekern. How about slapping on a tag on mor1kx?17:33
stekernolofk: I'm all for it17:52
stekernjuliusb is the one that is slowing the tag-team down =)17:52
stekernI mean, I think we can forget that we "slow down and just push bugfixes for a release"17:53
stekernthat's never going to happen17:53
stekernbetter to just tag a release, and then backport fixes to that when they emerge17:54
stekernthat's my 2 cents17:54
stekern... but we better hurry, because tonight I'll start introduce new bugs with the storebuffer ;)18:20
stekernolofk: thanks for catching those yesterday btw!18:21
olofkThere's an insane mount of weird legacy options in unzip18:36
olofkAwesome! github allows downloading as .tar.gz as well. Bye bye zip18:41
stekernheh, of course18:42
stekernyou can do git archive in your local repo too18:43
poke53282zip was never good. Unfortunatelz it is still being used. Even by me to distribute my software. I can't expect that everyone knows to handle .rar or .tar.bz2 files.18:43
stekernyeah, in $dayjob it's zip files for me too18:44
poke53282The worst part was the encryption option.18:44
poke53282Just for one I tried one decryption software and was amazed that I got a 10 digit password within 2 hours on a slow computer. This is already 13 years ago.18:46
poke53282Just for fun ....18:47
stekernpoke53282: you have those extra exports in rcS only, right?18:49
stekernthe SDL_ ones I mean18:49
poke53282Oh, probably I have them twice. I tried several positions and only one worked. Then I forgot to remove them. I have also one .profile and one .bashrc file18:50
stekernok, that explains it... it didn't work in rcS for me neither18:51
stekernbut grepping took to long ;)18:51
poke53282Probably I tried .bashrc first. That didn't work. Then I tried rcS. That didn't work. Then I looked up the busybox doc. ;)\18:52
stekernthat's the engineering spirit, don't look in the docs until everything else have failed18:55
olofkTime to add license headers to orpsocv3. What to choose...?19:02
olofkI'll probably go with GPLv3.19:03
olofkThis is for the python files19:03
jeremybennettolofk: Historically we have used LGPL variants for hardware and GPL for software (most recently GPLv3).19:04
stekernpersonally, I'm not a huge fan of gplv3, but it's your call19:04
jeremybennettLGPL is a really bad choice for hardware, but we are stuck with it by history. See this blog post for some helpful info
jeremybennettThe CERN OHL upgrade is not yet released, so for a copyleft style protection of hardware, CC BY-SA 3.0 is recommended.19:06
olofkThis is pure python software, so I'll go with GPLv3. Now I'm just trying to find a nice standard template to use.19:07
stekernI think our biggest dilemma is that, what are open cores really, software or hardware?19:07
stekernI'd say neither19:08
stekernit's no more hardware than a firmware burnt into a ROM19:09
stekernso any license that would be suitable for such firmware should be as suitable for us19:10
stekernthe question is, is there a license suitable for firmware?19:11
olofkNo, the question is what the hell I should write at the top in my pyton files for orpsocv3 :P19:12
stekernyeah, yeah, I get it, you haven't been productive enough and want to increase the KLOCS a bit ;)19:12
olofkbusted :(19:14
poke53282In my opinion put a license file and add some part in the readme file. Don't add a header in every file with the license. This is in my opinion not necessary and leads to increased pulse everytime I have to scroll down two pages to get the first useful information.19:16
stekernhey, excercise is healthy you know19:18
olofkYeah, I'm thinking about leaving the files as they are. There seems to be a lot of people who think that just adding a project-wide license is enough19:19
olofkNow should it be called LICENSE or COPYING?19:20
poke53282Does not matter. Both is obvious enough19:23
stekerntrying to tick this store buffer thing of my TODO-list just added two new items19:33
stekernfirst incarnation will of course just be a simple FIFO type of thing, but some time I'd like to make it combine < 32-bit stores and out of order19:34
stekernrelatively simple would probably be to let it fill up with a burst length too, and keep a flag if they are "burstable"19:36
olofkI just officially released orpsoc 3.019:51
olofkAs soon as I figure out where to put my tar ball, I will celebrate with a mailing list announcement and a tweet19:51
stekernthat's what I call a party!19:56
olofkopencores only handles .odt, pdf and doc files19:56
stekernspeaking of which, our 8 year old had his birthday party today, what a chaos...19:57
olofkhahaha. I can imagine. I will probably experience it myself in 7 years and 8 months19:58
stekernbe afraid, be very afraid....19:59
stekernI had to handle the first half an hour myself too, it was a long time ago since I longed so much for my wife to come home20:00
olofkjuliusb: It looks like you have uploaded stuff to the opencores FTP. Can you upload orpsoc-3.0 for me?20:06
_franck_olofk: why don't you use openrisc github for your orpsocv3 ? and use tags...20:26
olofk_franck_: I still want a tar ball that has passed make dist20:27
stekernolofk: don't you have the login/pass to the ftp?20:28
olofkstekern: It was probably sent to my orsoc mail account20:28
stekernthat's why I never use company e-mail address for anything else than strictly work things20:30
olofkstekern: I learned my lesson20:31
stekernbut yeah, in that case it was probably not your choice though20:31
* _franck_ is porting his altera de1 board to orpsocv320:33
olofk_franck_: That is awesome! Please tell me if you need any help20:35
olofkstekern: Thanks :)20:37
olofkHmm. doesn't firefox have a built-in ftp uploader?20:38
stekernno idea, I use ftp, ncftp or krusader if I fancy something guiy20:40
olofkTurned out that I had neither installed, so I used a firefox plugin instead20:43
olofkGentoo considers an FTP client to be bloat ware :)20:44
stekernthat's what I like about krusader (and total commander when I'm forced to use windows), you get a lot of that "Bloatware" bundled in one package20:47
_franck_one should create a script for RTL source code...20:50
olofkAs long as you drop the .pl extension, I'm ok with that :)20:50
olofkI'm also against checkpatch.vba20:50
olofkOh well. That's it for tonight. I'm expecting to release a 3.0.1 as soon as someone tries to use it20:52
_franck_I'm a pshycho-tab/space/ident-path20:52
stekernolofk: you'd love this "application"
stekernI had to use that to calculate some ntc curves a while ago, it's basically a whole program embedded into excel...20:53
stekern_franck_: whatcha looking at?20:54
stekernI'm a bit allergic to trailing whitespaces myself, after turning on global-white-space in emacs as default20:55
_franck_orpsoc_top.v there is so much space/tab mix20:55
stekernyeah, that's a bit messy I think20:56
stekerninconsistently messy20:57
stekernwe've got the mixed space/tab thing in mor1kx too, but that's just what the default verilog mode in emacs chose to do20:57
stekernit's probably derived from the gnu coding style20:58
stekernbut that's consistently messy ;)20:59
_franck_that's what I thought, it looks like gnu coding style20:59
stekerni.e. it follow the rules that the emacs verilog mode set up21:00
_franck_anyway, my board won't run tonight, there is too much spaces to replace by hand ;)21:01
stekernI'd personally prefer tabs only, but it's nothing I'm going to push juliusb into using ;)21:02
stekernI think almost any coding style is acceptable, as long as it is consistently followed21:03
_franck_+1 for tabs, size 8 like the kernel coding style21:03
stekernah, I think this is it for tonight for me too21:12
stekernI got a nice little store buffer module layed out and halfly hooked up and it seems to do the things right so far21:13
stekernjust need to connect it so it actually performs the write out on the bus and stall the pipeline when it's full21:13
stekernand stall loads when it's not empty21:14
_franck_a store buffer is something you put between the writeback stage and the memory ?21:20
tintenammaeCould anyone help tell me how/if stdin is set up when running c programs (no linux) on orpsoc altera dev board? Like stdout sends characters to the serial port, not sure about stdin?21:29
tintenammaewould be very grateful if someone could give/point me in the direction of some information on this?21:56
olofkstekern: I was going to sleep, but then I decided to play DOTT instead. Bad decision :)22:36
--- Log closed Wed Aug 21 00:00:21 2013

Generated by 2.15.2 by Marius Gedminas - find it at!