IRC logs for #openrisc Tuesday, 2014-01-14

--- Log opened Tue Jan 14 00:00:53 2014
PowermaniacHey guys, how's it going?08:33
_franck_web_Powermaniac: hi. It's been a while09:43
PowermaniacSure has09:43
PowermaniacSo how is openrisc going?09:44
PowermaniacOh and I shouldn't be rude, so how are you?09:44
_franck_web_I'm good thanks. What about you ? You were on the way to choose a school right ?09:46
PowermaniacWell I'm planning to go to university this year hopefully, although I'm being stuffed around by the high school I went to last year. But besides that I'm good thanks.09:47
_franck_web_about openrisc, it is quiet here lately09:48
PowermaniacOh okay. Any significant changes or advancements been made?09:48
_franck_web_not that much09:49
PowermaniacDid anyone end up working on getting an actual operating system running on ORPSoCv3?09:50
PowermaniacAs I was wanting to work on it but got caught up with other things/distracted plus I didn't ahve enough experience programming09:50
PowermaniacWhich I'm currently working on learning to program in C09:50
_franck_web_board supported in orpsocv3 (altera de1 and de0_nano) are running Linux09:52
PowermaniacWWell I'm just going to double check but when you say Linux what exactly do you mean, as I might mean something different. I'm meaning say Ubuntu or Debian or something along those lines...09:53
_franck_web_ah ok, so no09:54
PowermaniacAlright. Now wasn't there also someone else wanting to use ORPSoCv3 besides myself to produce an open source computer?09:55
PowermaniacSure there was someone that came after me that wanted the same thing basically09:55
_franck_web_Can't remember exactly, but yes, I think so09:57
PowermaniacOkay as I was going to ask if they are still around how it went...09:59
stekernPowermaniac: blueCmd_ have resurfaced in recent days, he's working on glibc support10:11
Powermaniacstekern: Hey! Okay thanks.10:12
blueCmd_Powermaniac: I actually built the first Debian or1k package yesterday10:16
blueCmd_so things are coming along nicely10:16
PowermaniacWow, cool10:16
blueCmd_at the moment you have to install Linux 'LFS' style (compile everything by hand)10:16
PowermaniacThat's fine10:17
blueCmd_I have scripts to build everything from scratch to a root filesystem with sshd, but that's about it10:17
PowermaniacHmm you have me once again wanting to build that open source computer I had dreams of last year10:18
PowermaniacNow I have to toss up between a new GPU or an FPGA...Hmm10:18
PowermaniacblueCmd_: Does it matter which FPGA board you use?10:18
stekernbuy an FPGA and build the GPU yourself, problem solved? ;)10:19
Powermaniacstekern: Wish it was that easy, I need CUDA support so =\10:19
PowermaniacPlus I was going to use it for gaming on the side too10:19
PowermaniacAlthough my Birthday is in April and I was going to wait for the new series of cards...10:20
PowermaniacFPGA it is!10:20
PowermaniacSo yeah now the question is which FPGA board to get...What would be most suitable for using as a computer...10:22
PowermaniacblueCmd_: Got any recommendations on a suitable FPGA board as I assume you had to make this decision also at some point?10:30
PowermaniacIf you are planning on using it as an open source computer that is10:31
-!- knz_ is now known as knz10:36
blueCmd_Powermaniac: actually no, I'm just using or1ksim10:37
blueCmd_stekern: and get graphics that was mediocre in 1990!10:38
blueCmd_http://www.kickstarter.com/projects/725991125/open-source-graphics-processor-gpu10:38
PowermaniacblueCmd_: You are just using or1ksim O_o10:43
blueCmd_Powermaniac: yup, it's much easier to automate10:43
blueCmd_I want to have hardware but I haven't gotten around to buying anything10:43
PowermaniacDoes everything actually run at a decent speed seeing as it is simulated?10:44
olofk_I remember that or1ksim ran faster than the old Actel based ordb1 boards :)10:45
olofk_But those Actel ProASIC FPGA's were no speed demons. I think they realized later on that having carry logic makes the FPGA a bit more useful10:50
PowermaniacblueCmd_: So would your Debian package actually work on hardware or would that require some extra work?10:55
blueCmd_Powermaniac: they would work just fine11:02
stekernblueCmd_: exactly, the graphics from the 90s were the best, no need to ruin it with fancy stuff ;)11:26
stekernyay: https://lkml.org/lkml/2014/1/14/8911:28
olofk_stekern: Congratulations. Your first code in the kernel?11:29
stekernif you don't count one-liners, yes11:32
stekernalthough it's not strictly in yet11:32
stekernthat ought to attract some chicks, right?11:35
blueCmd_Hah, cool!11:37
blueCmd_stekern: now the value of you saying my name on the internet has risen!11:37
stekernyeah, I'm expecting there'll be enough chicks for you to ;)11:42
_franck_web_stekern: great12:36
PowermaniacYou guys generally recommend Altera FPGA boards right?13:09
PowermaniacAs there is a slight problem cost wise getting a Xilinx board here in Aus is far more convenient13:09
olofk_Powermaniac: Xilinx support isn't in place for orpsocv3 yet. There's also the problem that Xilinx no longer ships a free version of modelsim with simulation models of their hard blocks, which means that system level simulation isn't possible to do13:23
PowermaniacWell damn13:23
PowermaniacWell I'm off, good talking to you all again and thanks for the info. Will have to look into an Altera FPGA board some more and see if I can get one shipped here cheaply or something.14:06
PowermaniacBye!14:07
olofkok, I've done enough to run verilator simulations with an UART19:30
olofkfor or1200-generic19:30
olofkIt's a little hackish though, but I'll clean it up a little and push it. Improvements can come later19:31
olofkCan't boot linux yet though :(19:32
stekernhow come?19:36
_franck_olofk: me too :)19:57
_franck_too bad we spent some time doing the same thing19:59
_franck_olofk: this is what I did: http://pastie.org/8633516 it's not finished yet20:15
_franck_I moved bench/verilog/uart16550_model.v in tb_private_src_files in both [verilog] and [verilator] sections20:15
_franck_so they can have their own version of this file20:16
olofkYeah, that's a good idea20:32
olofkI got around it in another way, but having different tb_private_src_files would have made it smoother20:32
blueCmd_stekern: when you are rebasing upstream, would you mind writing down the commit you are rebasing from?20:35
olofkMy somewhat dirty solution was to add uart_transceiver.v from mmuart in the milkymist repo. I hooked that up to a process that just $display's the received byte20:35
blueCmd_I'm creating diffs against upstream and I'm sort of guessing from the contents of the changelogs20:35
olofkBut being able to have different tb_private_src_files would allow me to have different test bench toplevels for icarus/modelsim and verilator20:36
olofkRight now it's just an ifdef20:36
_franck_indeed20:37
_franck_olofk: I let you work on this, I get back to my or1ksim to RTL bridge20:51
stekernblueCmd_: you mean or1k-src?20:56
blueCmd_yeah21:13
olofk_franck_: I'll have some stuff to commit, but they contain no changes to orpsoc, so your patch for tb_private_src_files is welcome21:15
olofkBut the or1ksim to RTL bridge sounds more exciting :)21:21
olofk_franck_: I want to add your fancy libs feature to modelsim vpi. Should I do it during compilation or linking?21:31
_franck_well, I guess linking21:32
olofkLinking I guess21:32
olofk:)21:32
olofkI also realized that I haven't added -std=c99 to the gcc argument list. Should that be default, or do we want is a configurable parameter?21:33
olofkI'm not up to date on what is the current standard21:36
_franck_I would say configurable. but it means add a vpi_module['cflags'] so you can put it as default21:37
olofkcflags makes sense21:38
olofkReading the gcc man now. Is really // comments a C++ feature? I've been using that for as long as I remember21:38
_franck_I used this flag when compiling or1kelfloader with gcc. Then I used iverilog-vpi and I guess it is the default here21:39
_franck_:)21:39
_franck_you won't put // in any linux source code :)21:40
olofk // is much better. /* can get you in trouble when you want to divide something with a pointer value :)21:41
_franck_:)21:41
olofkBut if they want their stone age comment styles, they won't get any of my code21:42
* olofk is bitter for not having any code in the kernel21:42
olofkAdding libs support for modelsim was dead easy. I must say that I'm proud of how easy it is to add features to the orpsoc code base21:43
_franck_I start to understand how it works after I wrote many patches :) I should start learning python21:44
olofkYeah. You're really starting learning python the hard way :)21:44
_franck_I adding a feature to add this kind of thing:21:45
_franck_[scripts]21:45
_franck_pre_run_scripts  = bench/start.sh21:45
_franck_post_run_scripts = bench/post.sh21:45
olofkYes! Great! That's also something I've been looking forward to21:46
_franck_so we can start any other thing before and after with run the simulation (like openocd or any program talking to the vpi module)21:46
olofkOr run some code generator21:47
_franck_also pre_build and post _build21:47
olofkI think it could be nice to export some environment variables when we run the scripts so they can know where the source directories and simulation directories are21:48
* _franck_ need to find out how to do that21:48
olofkos.environ or something like that21:49
_franck_ok, thanks21:49
olofkWe don't need to figure it out right away, but I think it will be helpful in some cases21:49
olofkHmm... it's probably better to add the environment variables directly to subprocess.check_call instead21:51
olofkIt's just to add a dictionary with the environment mappings in a parameter called env21:52
stekernblueCmd_: yeah, that might have been a good idea. I think the next time that repo is synced it should be split up and rebased upon real git repos though (as per the discussion with jeremy the other day)22:01
olofkJust pushed some fixes to orpsoc/orpdov-votrd22:01
olofkorpsoc-votrd is the name of the new repo. I got bored with calling it orpsoc-cores22:02
blueCmd_stekern: +122:03
blueCmd_stekern: or just upstream it22:03
blueCmd_:D22:03
_franck_olofk: I had to fix this: wb_intercon/wb_mux.v:95: Signal unoptimizable: Feedback to clock or circular logic: v->wb_intercon0.wb_mux_or1200_d.slave_sel_int22:04
_franck_did you see that too with verilator ?22:04
stekernyeah that too...22:05
_franck_you should push a fix because the current core can't be verilated22:09
blueCmd_look at that: binutils-or1k-linux-gnu_2.24-2.1_amd64.deb22:16
olofk_franck_: Hmm.. didn't see that. There is a lot of warnings in the wb_intercon stuff however. Probably should clean that up22:26
olofkWhich version of verilator are you using?22:26
_franck_Verilator 3.805 2010/07/10 rev verilator_3_80522:26
_franck_pretty old22:27
olofkCould be that. I'm using 3.8.50 here22:31
olofk_franck_: Could you try to remove wb_intercon from the cache? I remember that I had to rewrite the handling of slave_sel_int because the original code didn't work in verilator22:36
olofkJust to make sure that you don't have an old version in your cache22:37
olofkGot to sleep now22:37
_franck_I'm not ready to do that22:37
_franck_ok, good night22:37
_franck_I mean not ready to test verilator write now. I'm still working for 20 min before I go to bed ;)22:38
--- Log closed Wed Jan 15 00:00:54 2014

Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!