IRC logs for #openrisc Thursday, 2014-01-16

--- Log opened Thu Jan 16 00:00:56 2014
stekernblueCmd_: have 4.8.2 been released now?04:01
Powermaniac...were one of you guys at 30C3?06:03
PowermaniacAs I just heard this:
stekernPowermaniac: I don't know if anyone was, but how did you drew that conclusion from that?06:45
PowermaniacWell, someone mentioning they have developed a debian toolchain for FPGAs...06:45
PowermaniacSounded like what BlueCmd_ is doing06:45
stekernno, the guy said he have created debian packages for the altera tools06:46
stekerni.e. quartus.deb06:46
PowermaniacOh okay, my bad06:46
olofk_blueCmd_: I realized that it was cross-compiler stuff just after I asked that. Still, that's even better as a lot of people are having problems with building the toolchain06:48
olofk_blueCmd_: When are you going to do the ebuild for Gentoo? ;)06:48
olofk_blueCmd_: Are the cross-compile debs working well enough to be recommended to users?08:07
blueCmd_olofk_: I haven't installed them yet so I wouldn't know :P09:06
blueCmd_olofk_: gentoo smentoo, what will you do? do emerge world on an or1k? it will complete around the year 224109:06
olofk_blueCmd_: Well, given the available or1k software on gentoo it will be a small world, so I'll probably be finished sooner :)09:09
blueCmd_olofk_: also currently they are built against unstable debian only since I need quite recent sources09:09
blueCmd_time to go to work, afk for a while09:10
olofk_Keep us posted on the status. This is great for people who want to quickly get into OpenRISC development.09:10
olofk_IIRC jonibo tried to do something similar a few years ago. I think I installed some stuff from his Ubuntu PPA09:11
jonibodebian has nice infrastructure for building cross-compilers, iff your arch is upstream09:12
joniboif I recall correctly, the cross-compiler debs weren't impossible to do, but a nightmare to maintain09:13
blueCmd_a _lot_ of work has been done in debian with this for the last year09:51
blueCmd_I don't know if you recall me pushing for including or1k in config.guess and stuff like that, and that really paid of09:51
blueCmd_arm64 has pushed a lot of cool cross compiler features to Debian09:52
blueCmd_I'm being 'mentored' by a guy who works with making arm64 work on debian, so it's all good.09:52
jonibocool... sounds like you're definitely on the good path!  keep up the good work!09:57
blueCmd_thanks! yes I'm hopeful :)09:59
maxpalnout of interest, has anyone implemented a GMII MAC for the ORSOC?10:23
olofk_maxpaln: I know that this one (,ethernet_tri_mode) has been downloaded a lot. I have no idea how good it works or how much tech-specific stuff it contains, but I've been interested in trying it out for orpsocv310:29
olofk_The benefit of adding GMII to the classic ethmac on the other hand would be that it is battle-proven (although not bug-free) and has had a linux driver for a long time10:30
olofk_But it's probably worth doing a bit of refactoring first. There are some shady areas in the RTL10:31
maxpalnthanks olofk_10:46
maxpalnGiven our Versa board contains a GbE PHY I think it will be useful to get a GMII MAC implemented - it will certainly maximise the functoinality of the board10:47
maxpalnWhen I get to this I'll take a look at the project you highlighted and see how it looks10:47
stekernolofk_: looks like you have a lot of fun ahead of you reviewing those latest orpsoc pull requests...10:51
olofk_stekern: Yes. That will take some time10:52
olofk_But it looks like interesting stuff in there. Bug fixes and Verilator additions10:52
olofk_Just need to sort it out10:52
olofk_Thankfully I have my git support line here to assist me in my rebasing and patch mangling :)10:53
stekernhehe, I would suggest you nudge the initial patch mangling to that Jose guy ;)10:53
_franck_web_I have orpsoc and verilator working quite good. I'm almost ready to submit this to you olofk_11:00
_franck_web_or1k-elf-loader code is not copied anymore11:01
_franck_web_I don't know how it's going to fit with Jose's work11:02
stekernyeah, you might get some conflicts in the .svn entries I suppose11:04
* _franck_web_ is going to hurry up before he needs to rebase his work :)11:06
olofk_Don't worry _franck_web_ I'll pull in your stuff first. The other stuff will need a lot of love anyway11:30
_franck_web_olofk_: how do you set VERILATOR_ROOT on your machine ?12:38
_franck_web_the way you use it in the current code doesn't work for me12:38
_franck_web_this is what I have avec ./configure && make && make install verilator12:38
olofk_I have a local installation of verilator and just run export VERILATOR_ROOT=/path/to/my_local_install before starting orpsoc12:40
olofk_So I just do ./configure && make12:40
olofk_ahh.. I see12:41
olofk_ok, so to support normal verilator installations we should probably just see if we can find it in $PATH first and run it directly in that case12:42
olofk__franck_web_: So what do you have to change to run your verilator from orpsoc?12:45
_franck_web_AFAIR, if I set VERILATOR_ROOT=/usr/local/ , verilator starts but it cannot find the include dir which is in VERILATOR_ROOT=/usr/local/share/verilator/13:06
olofk_I'll read the installation instructions again and see how it's supposed to be done13:26
olofk_If you run make install for verilator it will hardcode the currently set value of VERILATOR_ROOT into the binary13:31
olofk_So first make sure that VERILATOR_ROOT is unset when you run make install13:33
olofk_Add /usr/local/bin to $PATH and make sure that VERILATOR_ROOT is unset13:33
olofk_How the hell should orpsoc detect if it's installed or just make'd?13:34
olofk_Maybe like this:13:35
olofk_1. Check if verilator is found in path. Run it13:35
olofk_2. else check if VERILATOR_ROOT is set, Run $VERILATOR_ROOT/bin/verilator13:36
olofk_Still need to come up with some helpful text if it fails13:36
olofk_Maybe orpsocv2 solved this in a better way. Should probably check that13:37
xlrohey, I got a quick question: does the mor1kx cappuccino pipeline version work without dcache icache and no mmus, somewhere in the doc it says $i is currently not optional13:40
_franck_web_olofk_: plus I ran verilator_bin and not verilator, it might have something to do with VERILATOR_ROOT also13:40
olofk_verilator is some sort of shell script, I think13:41
xlrothe latest version on git actually has all these set to NONE, which would be great if that works, since we want to plug our own memory architecture outside13:42
olofk_xlro: Sounds like something stekern would know13:51
xlroor juliusb_ I assume :)13:52
xlroolofk_: how are things going with orpsocv3? I still haven't ported my simulation environment, should look into it soon13:53
olofk_xlro: It's going well, but I can't spend as much time as I had wished on it13:54
olofk__franck_web_ has put in a lot of great stuff recently too13:54
olofk_building with Quartus works and simulating with modelsim and icarus works fine. verilator works ok, but there are some useful improvements coming in now13:55
xlrocool, I will mainly use it to simulate with modelsim13:56
xlroI'm especially interested in doing simulation with the embedded openrisc core on gatelevel for accurate timing & power, with the rest outside of orpsoc still behavioural13:58
olofk_Hmm.. interesting use case. There are a few ways that orpsoc could help you there14:00
olofk_1. Use orpsoc to pull together the SoC from the various repos, compile VPI stuff and generate the simulation scripts14:01
stekernxlro: it's possible to disable caches and mmu, both at compile time and runtime14:03
olofk_2. Run everything from orpsoc and use the fancy new feature (_franck_web_ is working on that) to run an external tool to do the gatelevel synthesis of openrisc just prior to building the simulation model14:03
xlrocurrently I have a hacked/modified version of the testbenches/monitor that comes with v2, which works with the cpu on gatelevel (mainly removed references to internal names of the cpu entity that are not available anymore, used by special NOPs etc)14:03
xlrosounds good olofk_14:03
xlrothanks stekern, yeah Im especially interested at compile time, so they do not get instantiated14:04
xlrodoes the spram instantiation in mor1kx_icache somehow get optimized away if FEATURE_INSTRUCTIONCACHE = "NONE" ? the FSM should go directly into BYPASS state and stay there, but does the memory actually not get instantiated?14:21
stekernthe whole icache module is conditionally instantiated by FEATURE_INSTRUCTIONCACHE14:38
stekernumm, but what version of mor1kx are you looking at, there are no FSMs with BYPASS states anymore14:41
xlroI just realized, I think I have not the latest version, mixed up the repos, I think I cloned the one from julius14:42
xlrowhich might be older14:43
xlroindeed I had this one, where its not conditionally instantiated:
xlrothanks alot for the headsup14:45
stekernyou're welcome ;)14:47
xlrojust to make sure ;) that's the latest?
olofk_franck_: What was the problem you tried to solve with the os.path.isdir stuff? Just want to know so that I can test my changes to your patch19:59
_franck_it's when you set core_root to a non exsiting path20:19
stekernjonibo: I'm trying to get my head around the syscall problems in uclibc, something close to mark salter's patches have now been applied, but there is still one question mark20:39
stekerndo I understand things correctly that that patch is "generic" for archs that don't define __ARCH_WANT_SYSCALL_DEPRECATED?20:41
stekernbecause __NR_fork is only defined when _ARCH_WANT_SYSCALL_DEPRECATED is defined20:42
stekernactually, there is something committed to mainline uclibc that is close to that patch, but it seems wrong to me since vfork still need __NR_fork to be defined for the clone implementation to be picked up21:03
stekernis what I'm speaking about21:08
stekernit has a nice syntax error too, a missing ';' after return -1, which makes my believe that mr Markos haven't really tested the patch stronger...21:09
olofk_franck_: I have done some improvements on your first commit now. There are some other things I should clean up in coremanager but I leave them for now21:12
_franck_ok great21:13
olofkPushed it now. Will have to do the other two commits tomorrow21:15
_franck_olofk: you could have added your signed-off here, you did 80% of the work....21:17
stekernah, now I see what has happened it's a merge conflict that's gone terribly wrong...21:17
olofk_franck_: How do I add a signed-off? Is that when I run git commit --amend ?21:19
olofkPushing the env patch too since that was so simple, but I need to look closer at some things in the last patch before pushing that21:19
olofkstekern: Please don't talk about merge conflicts that has gone terribly wrong. I was just about to go to bed, and now I probably won't be able to sleep21:20
_franck_yes I think --amend --signnoff does the trick21:21
olofkMy computer on the other hand will still be on to recompile libreoffice for the third fucking time in a few weeks21:21
blueCmd_olofk you stalker! :)21:24
olofkblueCmd_: Yeah I know. I spend way too much time on linkedin :)21:25
blueCmd_olofk: I got jealous about you having openrisc on your profile21:26
blueCmd_gonna adda that21:26
olofkblueCmd_: Yeah, bring it on. Let the world know what we're up to :)21:27
_franck_how can I do if I want my own git repo master branch to look *exactly* like another git repo branch21:29
olofk_franck_: Isn't that just clone?21:30
blueCmd__franck_: do you have commits in it?21:30
_franck_I could delete my repo on github and fork it agoin :)21:31
_franck_blueCmd_: yes I want to erase all history, just a raw copy21:31
blueCmd_I usually do:21:31
blueCmd_git checkout other_upstream/branch21:31
blueCmd_git push github other_thing_that_i_call_this_branch21:31
olofkAbout two years ago I removed all the commits from when I realized that my first makefile-based structure wasn't very good21:32
blueCmd_olofk: hah21:32
olofkThat's apparently very forbidden to do in git methodology :)21:33
stekern_franck_: git --reset hard another_git_repo_branch21:33
stekernblueCmd_: I'm starting to think u is for unbelievable ugly...21:35
_franck_seems like it did it21:35
blueCmd_stekern: 'u'?21:36
stekernnot as in 'you' ;)21:36
blueCmd_stekern: yeah, I read it that way first, but you never use 'u' = 'you' so I didn't think so :D21:37
blueCmd_pro tip: never rebuild GCC in debian without DEB_BUILD_OPTIONS=nocheck21:37
blueCmd_unless you have unlimited amounts of time21:37
blueCmd_stekern: hah!21:38
stekernblueCmd_: it was u that start calling me ugly ;)21:38
blueCmd_stekern: told you!21:38
blueCmd_english, how does it work21:38
stekernbeats me21:39
blueCmd_has the next openrisc conference been scheduled?21:40
stekernnope, but there was some talk about maybe having it in munich during the last conference21:43
stekernbut nothing concrete21:44
--- Log closed Fri Jan 17 00:00:57 2014

Generated by 2.15.2 by Marius Gedminas - find it at!