IRC logs for #openrisc Tuesday, 2014-01-28

--- Log opened Tue Jan 28 00:00:13 2014
stekernsomeone is giving the ethmac Linux driver some TLC:
jonibothere's always been a lot of acitvity in the ethmac driver08:53
joniboit's used by many others outside of openrisc08:53
jonibonios and microblaze systems (avalon bus) use it a lot08:54
joniboand now xtensa, apparenlty, too08:54
joniboxtensa, fwiw, reminds me a lot of openrisc08:54
stekernyeah, it really shows how something technically inferior can gain success with the right software support08:59
stekernbut that patch set (or the one after the review comments are fixed) should fix some of the issues maxpaln had09:01
maxpalnYep, I was just looking at that - I'll be interested to see how the disabling of GbE advertisement has been achieved.09:12
maxpalnNow i have the datasheet for the PHY on our board I was expecting this to be a register setting in the autonegotiation routine - but maybe there's a more generic way to solve the problem.09:13
jonibostekern: which bit is technically inferior? ethmac or xtensa?09:20
joniboif ethmac, what's superior?09:20
stekernjonibo: ethmac, it has it fair share of issues on rtl level. there are probably several open source MAC implementations out there that are superior, but they lack upstream support in Linux and the ethmac works "well enough".09:26
jonibook... interesting09:26
joniboeasy enough to get support upstream for another driver if there's a better core to use09:26
jonibojsut a matter of writing the driver :)09:26
stekernI know, but the "working well enough" is the barrier for that ;)09:27
jonibothis is a good TODO for an openrisc newbie... add it to the wiki... point them at a better core and ask for a driver09:27
joniboyeah, I know09:27
stekernanother approach is of course fixing up the issues in the ethmac core, which isn't neither a huge task09:28
stekernthe biggest problem is related to clock domain crossings09:28
joniboaha... sounds interesting09:29
PowermaniacAnyone online willing to answer some hopefully not too stupid questions regarding OpenRisc for me? Main question being: Is OpenRISC a completely new architecture or is it slightly based on something like ARM? And if it was manufactured would it be like a microprocessor or a system on a chip or an embedded CPU or workstation/server CPU, or mini mainframe CPU etc?10:59
doomlord_its custom, new i think, when i asked11:00
Powermaniac(Just going by the list of cpu architectures on wikipedia)11:00
doomlord_its not compatible with anything else11:00
doomlord_i get the impression the people behind it want the experience of designing an ISA11:00
PowermaniacOh okay11:01
doomlord_i would have thought you could make it a superset of some other existing machine and get more traction11:01
doomlord_but there's NIH :)11:01
doomlord_i'm always guilty of NIH myself, i can sympathise11:01
PowermaniacWell wouldn't it become closed source then?11:01
doomlord_Not Invented Here11:02
doomlord_= urge to re-invent everything11:02
PowermaniacAhh okay11:02
doomlord_i think you can make things binary compatible, but stealing actual designs is a whole different kettle of fish11:02
PowermaniacYeah I have that same feeling with a project I'm wanting/actually working on at the moment on and off11:03
PowermaniacThe urge to re-invent things that is11:03
PowermaniacWell the only reason you would want to make something binary compatible would be for software compatibility, no?11:03
doomlord_its like a fundemental problem in nature11:03
doomlord_the thing is , often the only way to truly know something is to build it yourself11:04
doomlord_let alone actual ownership, there's a sense of "intellectual" ownership11:04
doomlord_and its a huge part of any engineers motivation to create something that is 'theirs'11:04
PowermaniacWhich if it is aiming at the open source crowd is not too much of a problem if it isn't binary compatible as you can just get C running on it and compile most things from source.11:04
doomlord_yeah if they make a custom ISA, they need to make support in compilers...11:05
doomlord_i dealt with a few risc ISAs in games over the years11:05
doomlord_mips, powerpc, hitachi sh-series11:06
doomlord_not actually done much with arm directly myself11:06
PowermaniacI have to wonder if a company like Freescale offered the guys here or a similar company to work with them and manufacture OpenRISC if they would do it...11:06
doomlord_like whats the point of another risc isa, why not just clone one of those11:06
doomlord_but make an opensource implementation, and maybe have some custom extra instructions11:06
PowermaniacI just like the openness of it all, if that was the only reason they chose to do it this way, that seems like a good enough reason to me11:07
doomlord_heh. next up: open CPU foundry11:07
doomlord_well china have cloned MIPS, and i dont think anyone can sue them over it11:07
PowermaniacWell from what I understand with ARM they license it so you can't actually make an ARM open source processor11:07
doomlord_ok thats a big shame11:07
PowermaniacHaha China doesn't seem to give a shit xD11:07
doomlord_one project that looked interesting to me was a modernday motorola 68k series11:08
doomlord_but it was closed source11:08
PowermaniacHmm okay11:08
doomlord_(but thats not risc)11:08
doomlord_i liked mips11:09
doomlord_powervr are resurrecting it for an soc11:10
PowermaniacI'm sort of new to the whole electronics and actual CPU architecture scene so...11:10
PowermaniacAlthough I've always been rather big into everyday user computer hardware11:10
doomlord_i caught the tail end of the era in which games involved more asm11:10
doomlord_bits of engine code had to be done in asm, knowing all the pipeline hazards etc11:11
PowermaniacOh okay, cool. I've read a lot about how people used to edit games and program there own games as they were in asm11:11
PowermaniacOr well it came with how to code the asm code for the game and you had to manually type in the instructions11:12
doomlord_C plus "intrinsics" has drastically reduced the amount of asm the world needs11:12
doomlord_eg they usually give a customized c compiler with intrinsic functions that represent any special instructions the machine has (like fancy vector arithmetic or cache control)11:13
PowermaniacWell C does seem easier then asm which I guess is a plus(?)11:13
doomlord_yeah ancient 8bit games were exclusively asm11:13
doomlord_yes C has a special sweetspot11:13
doomlord_low level enough to reduce the need for asm, but high level enough to be tolerable to write serious programs  and portable11:13
PowermaniacJust been learning C recently, so yeah...Not sure where I was going to go with that now I think about it xD11:14
doomlord_C started being used in the 16bit era, but that was still a crossover; by 32bits, C was ubiquitous11:14
PowermaniacWould you happen to know if OpenRISC is 32 bits or 64 bits or 16 etc.?11:14
doomlord_no idea11:14
doomlord_i dont know much about it11:14
doomlord_its "an interesting thing" but low down on my list11:15
doomlord_i wish i had time for everything11:15
PowermaniacI'm planning hopefully on sourcing an Altera FPGA to try and build an 'open source' computer with the ORPSoCv311:16
doomlord_wow, nice11:16
PowermaniacHa yeah I know that feeling re: having time for everything11:16
PowermaniacAlthough if I get into uni this year I will probably be using a Xilinx FPGA at uni...11:17
PowermaniacConsidering what the guys here were saying about Xilinx I'm not sure how the uni is going to deal with the new developments of the free software not having the old features11:17
PowermaniacSomething about it changed recently and thus Xilinx is best to avoid at the moment11:17
PowermaniacBut the uni just bought a whole stack of new Xilinx FPGA boards11:18
PowermaniacThe Zedboard11:18
doomlord_i've never messed with fpga11:19
doomlord_i suppose they might be good at things like neural nets11:19
olofk_doomlord_, Powermaniac: OpenRISC is loosely based on MIPS-I and DLX11:19
PowermaniacWell I've sort of have. Was watching other uni students implement some basic features, it was a counter that had made...So nothing fancy11:19
Powermaniacolofk_: Thanks!11:20
PowermaniacIt is currently a single core processor right?11:20
olofk_Powermaniac: You can put as many cores as you can fit in an ASIC or FPGA :)11:20
PowermaniacOh okay cool11:21
olofk_But there is no real system for interconnecting the cores, so basically that would be come a shitload of single cores11:21
PowermaniacSo is anything holding it back speed wise, like if I were to use it as a CPU what year x86 processor would you say it compares to?11:21
olofk_And it has been manyfactured a lot of times over the last ten to fifteen years11:21
PowermaniacYeah I remember seeing Samsung had implemented an OpenRISC well one of you guys told me about it and I went hunting for it11:22
olofk_It's in Zigbee ASICs, in Samsung Digital TVs, in telecom equipment, in some of the Allwinner SoCs for tablets and other places11:22
olofk_The wikipedia lists a few of them
PowermaniacWow one was on the ISS that is pretty cool11:24
olofk_ISS? Where did you see that?11:25
PowermaniacThe wiki11:25
olofk_ahh.. the techedsat11:25
olofk_I didn't know that one was on ISS11:25
Powermaniacolofk_: So you still don't really recommend Xilinx FPGAs right?11:26
maxpalnHopefully not the inspiration for Gravity :-P11:26
PowermaniacAs I still need to look around to find a source here in Australia for Terasic FPGA boards11:26
olofk_Powermaniac: I recommend Altera FPGAs over the competitors (sorry maxpaln) because that toolchain is currently best supported11:27
PowermaniacAhh yeah I thought so11:27
maxpalnthat's ok - we're new to the OPENRisc thing11:27
olofk_I've started working on Xilinx support for ORPSoCv3 however, and it's not a very big task, so don't let that discourage you11:27
olofk_and we will soon have excellent Lattice support too ;)11:27
PowermaniacOh so I could get a Atlys FPGA?11:27
maxpaln:-) of course!11:28
maxpalnas an open source solution porting to different FPGA architectures isn't a massive task - supporting the peripheral set on a specific board is the bigger task but still manageable11:29
Powermaniacmaxpaln: Do you ahve some connection to Xilinx?11:29
olofk_The first OpenRISC board we made at ORSoC had an Atlys FPGA, so it was quite well supported in ORPSoCv2, but the toolchain was a pain to work with and we haven't seen any renewed interest in support for Atlys11:29
olofk_(or MicroSemi as they are called now)11:29
olofk_The OpenRISC in space runs on an Atlys FPGA, btw11:30
maxpalnit's probably one of the biggest advantages of the solution - you aren't forced to follow one set of devices. you can choose any FPGA vendor, an ASIC etc.11:30
maxpalnand the SW shouldn't need to change when the HW is retargeted (although I don't have any first hand experience of this so I'll let the other guys confirm!)11:31
olofk_maxpaln: Yes, that's probably the biggest benefit for consumers of these cores, and one of the prime targets for ORPSoCv3 is to hopefully make it even easier to move between targets11:31
PowermaniacNow I just need to remember if the Atlys board had all the features I wanted to make an open source computer11:31
olofk_Powermaniac: Probably not :)11:31
olofk_Powermaniac: But please send a link if you need some second opinion11:32
maxpalnNot wanting to promote the compeition but that Altera board looked pretty well kitted out with peripherals - it might be a better bet11:32
Powermaniacmaxpaln: Yeah just sourcing Altera boards here seems to be a pain11:33
PowermaniacWell that also have academic pricing*11:33
maxpalnah, well you're dealing with a big company - I have no idea what provision they make for one of purchases. Don't they have an online store or something11:33
maxpalnah, of course - educational discounts are worth a lot :-)11:33
Powermaniacwell that is new the DE1-SoC Board11:34
PowermaniacThere was some issue with the USB ports if I remember correctly on both the DE2-115 and Atlys boards that may make my plan difficult11:38
PowermaniacSomething about them not being a traditional USB port that can accept both data in and out11:39
PowermaniacOh seeing as BlueCmd isn't here do you guys know whether I only need to update to Debian Unstable/Testing to get the kernal upgrade to run the qemu-or32?11:42
stekernolofk_: now you're mixing up the terms, Actel FPGA, not atlys11:45
olofk_oh... sorry11:46
stekernsee what happens when I leave for lunch and stop watching you guys, you have filled several pages of irc logs =)11:46
olofk_And you need to clean up the mess that we left :)11:47
stekernyup ;)11:52
stekerndoomlord_: "new" is relative, the openrisc 1000 ISA is soon to be about 15 years11:54
olofk_oh god. I now see the confusion that I have created :)11:54
PowermaniacWhat do you guys think of the Nexys 4?11:55
olofk_Arithmetics in FPGA is a fucking pain!13:40
olofk_i.e. I suck at it13:41
blueCmdstupid timezones14:32
blueCmdwoke up at 4 AM fully rested14:32
blueCmdolofk_: how do you want the rebased binutils sources?14:35
maxpalnI am bout to start the job I have been putting off for ages :-) I am about reimplement the wishbone interface to the DDR3 module based on the DDR2 interface that is part of the Xilinx Spartan Board code14:48
maxpalnwish me luck!14:48
blueCmdglhf :)14:49
blueCmd*-changed contains the full files after they have been patched and what not, the patches can be found in debian/patches after running dpkg-sources -x on the .dsc file.14:57
blueCmdcombined changelogs are provided for easy 'grep for filenamoutput all authors'14:58
blueCmdmy internet is quickly deteriorating since everybody seems to be waking up at the hotel14:58
olofk_maxpaln: You can also take a look at the wb_sdram_ctrl test bench. It reads, writes and verifies a lot of data with my wishbone BFMs14:59
maxpalnolofk_: thanks - that could be useful15:00
olofk_blueCmd: Great. This should be a good starting point for anyone who wants to analyze who made the OpenRISC-specific changes15:01
blueCmdolofk_: I thought you were the 'anyone' in this case ;)15:36
blueCmdi'll start a google doc and let people fill in and we can get the ball rolling15:37
blueCmdjeremybennett: do you have any idea what kind of release statement we would need from the code authors?15:37
blueCmddo we need to do lawyers and stuff?15:37
stekernblueCmd: the way I have understood it, each one would need to sign a copyright assignment form to FSF15:40
stekernI think you can get the assignment form by sending a request to this address: copyright-clerk@fsf.org15:42
stekernbut I have no hard facts, jeremybennett probably knows better15:42
blueCmdlet's hope itt can be digitally signed15:45
olofk_blueCmd: I took a quick look at the files and quite a lot of the stuff is autogenerated or trivial makefile changes15:46
olofk_Would it be ok to just go by the author names in the source files, or do we need to track down the commits?15:47
blueCmdolofk_: the ocmmits15:48
olofk_ah crap15:48
blueCmdbut the changelog should be good for that I think15:48
olofk_Is that trustworthy?15:48
olofk_I'm worried about having to track git->svn->cvs otherwise15:49
blueCmdI think we have been pretty anal about having people write in the changelog15:50
blueCmdwe don't need everyone to sign for each commit, we only need to find out who comitted15:50
olofk_I tried to do a proper release 3 of or1200. That commit log was anything but clean15:50
blueCmd(and preferably comitted stuff we actually use)15:50
olofk_I found Johan Rydberg, Julius Baxter, Jeremy Bennett and Peter Gavin from just looking at a few files. Anyone familiar with the first guy?15:52
olofk_Or is his stuff already in binutils? Have to make some diffs15:57
blueCmdthe stuff already in binutils is probably all outdated, but that would mean that he has signed the thing I guess15:58
stekernolofk_: johan rydberg wrote what's already upstream, nothing after that17:20
stekernand I can't recall jeremy touching anything in or1k binutils, but gdb and newlib, but I can remember wrong and not that it matters17:50
amsolofk: help!17:55
amsjuliusb: help!17:55
amsolofk, juliusb, I need an FPGA for some retro mini-computer hackery... More exactly, either a CADR running on a FPGA, or something Ivory like.  ... what each of those are -- Lisp says it all.17:59
stekernams: almost any of the half decent to decent fpga devboards are probably sufficient for that18:58
olofkams: Do you have any experience with FSF copyright assignments? We're looking to do a serious attempt of upstreaming binutils22:21
olofkBut we don't like GPL, so we want to use MS-PL instead, but I guess that shouldn't be any problems22:23
stekernI don't understand why people go on about Linux booting quickly, it took me a month to boot this:
--- Log closed Wed Jan 29 00:00:15 2014

Generated by 2.15.2 by Marius Gedminas - find it at!