IRC logs for #openrisc Wednesday, 2014-09-10

--- Log opened Wed Sep 10 00:00:45 2014
poke53281olofk: you played it?02:58
stekernolofk: ok, I'll take a look03:16
poke532819 PM: Just my computer, my compiler, my emulator, the Elite 2 source code and me.04:10
poke53281And a Star Trek episode in the background04:10
poke532812.5 hours time04:11
stekernlucky bastard ;)04:17
heshamDoes the current Atlys system (built from fusesoc) supports bootloader (u-boot?) like orpsocv2?06:11
olofkpoke53281: Yes, I've played it a lot. I remember when I had bought it, I panicked when I realized it contained millions of different planets, and I wasn't sure if it would fit on my small harddrive. When the box arrived it contained a single DD formatted disk. I was so damn impressed by that. Probably part of the reason why I got so interested in embedded and working with limited resources06:54
olofkMy father played it even more though. He printed out a matrix of A4 papers that he put up on the wall to write down all his trade routes and where he had placed drilling towers. That made me lose interest since I knew I never would be that into the game :)06:55
stekernhah, I have a similar memory with my father, we competed playing this game:
olofkstekern: We have so much in common. Still got that one downstairs :)07:01
stekernand when the game ends the "map" where you fail is showing for a minute or so, so I rebuilt it with matches to solve it in peace07:01
olofkMy favorite game and watch07:01
stekernyeah, it's probably the best game&watch game they made07:01
olofkI hope you didn't try to recreate one of the moving maps ;)07:02
stekernhaha, no07:04
stekernbut it was pretty silly that it was a dual-screen game, the only thing the upper screen was used for was the animiation where the bomber-dude jumps down to the underground07:05
olofkTrue. But the marketing worked. I probably got it because I thought it would be twice as good as a single-screen game07:05
stekernsame here07:05
stekernI hesitated a bit since there was no fire button though07:06
olofkGod, that thought seems familiar07:09
stekernthe things that go on inside the head of an 8 year old kid ;)07:21
olofkWindows, I will soon kill you07:56
olofkMaybe I shouldn't try to connect to exotic hardware from inside of a VM, but I'm really scared of putting windows on this computer07:57
olofkIf only I could recompile this crapplication for Linux07:58
stekernolofk: can't you symlink a complete core?10:08
stekernI'm trying out your wb_streamer, and I checked it out in ../../wb_streamer realative to orpsoc-cores/cores10:09
stekernand then tried to symlink it there, but fusesoc bitches that it doesn't exist10:10
olofkJust add it to cores_root in your fusesoc.conf10:19
olofkcores_root = /path/to/orpsoc-cores /path/to/other/repos/parent10:20
olofkSymlinks should work though, but I haven't tested it, so there might be bugs10:20
olofkAnd as a sidenote, systems_root isn't needed anymore so you just need a cores_root entry to the orpsoc-cores root instead of one to /cores and one to /systems10:21
olofkYou need a quite recent fusesoc for this though10:24
stekernoh, but that's more annoying than symlinking it10:26
stekernbecause I have millions and millions of different build dirs10:26
olofkNot sure I understand the problem10:28
olofk(of course symlinks should work. Just not seeing what you're seeing)10:28
stekernthe problem is that I have seperate fusesoc.conf in each build dir10:30
olofkah ok.10:30
stekernbut all use the same orpsoc-cores10:30
olofkYou can put one in ~/.config/fusesoc/fusesoc.conf then10:30
olofkAnd only put one in $pwd if you want to override that one10:30
stekernyes, or I can symlink it and bitch about it not working in #openrisc ;)10:30
stekernand then just copy the whole dir into orpsoc-cores10:31
stekernbut git submodule support would really make sense for such 'fusesoc-aware' cores10:33
olofkNot sure what the problem is, but I'm using os.isdir() in some places. Might be that it fails on symlinks10:33
olofkAbsolutely. I've changed my mind on that one and plan to add that before the next release10:34
olofkBut I want to keep it out of the tree for now anyway. I have a separate staging area for stuff like this10:35
olofkso I have cores_root = ~/path/to/orpsoc-cores ~/path/to/staging10:35
olofkDoes anyone have experience of building with wxwidgets?10:46
stekernolofk: is wb_stream_ctrl.v the top file?11:17
olofkNo, for your purposes wb_stream_writer.v would be the toplevel11:18
olofkI haven't implemented the wb config interface yet, so wb_stream_reader_cfg.v is just hardcoded atm11:18
olofksorry s/reader/writer11:21
stekernI will be too lazy to fix my orpsoc_top.v later...11:23
stekernwill the "final" version have the intention that they are used individually?11:24
stekernor will there be a wb_streamer.v that pull everything together?11:24
olofkYes, I'm not sure they should be in the same repo at all. Looks like they won't share any code11:24
stekernwell, IMO, two files can be in the same repo even though they don't happen to share any code ;)11:25
olofkYeah, they are small enough for that11:25
stekernolof 'one repo per file' kindgren ;)11:26
olofkAnd maybe it would make sense to have a combined top-level. Haven't thought enough about that yet11:26
stekernI can make a wb_streamer.v in my system/sockit dir in the meantime11:27
olofkIt makes it look like I'm doing a lot more code if I have many repos :)11:27
stekernwell, I think it makes sense to only have one config slave if you are using both the reader and writer11:28
olofkYeah, too much wires to connect on the top-level otherwise11:32
olofkActually, the nicest solution would probably be to have a proper DMA component and only have slave interfaces on the stream components11:36
stekernyeah. but you need some bus interface and arbitration between the components then11:39
stekernso that could as well be wb?11:40
stekernok, 'reader' and 'writer' refer to the 'stream'11:40
olofkI renamed it yesterday, because I thought that made the most sense11:41
stekernyeah, I think either will confuse you ;)11:41
olofkNo question about that11:41
stekernbut the name says 'stream writer', so I agree that the current convention probably makes most sense11:42
olofkThat's what I figured too11:42
olofkIf I want to kill someone slowly I will let them work with string conversions in C++ with wxwidgets11:56
sb0just built mor1kx again... still bloated12:57
sb0in fact it gained 21 LUTs since last time12:58
PowermaniacHowdy, been a while any of you guys heard of this project: ?13:09
stekernsb0: yeah, I haven't received any de-bloating patches ;)13:56
stekernjoking aside, it's still on my todo list13:57
olofkMaybe I should buy up all the tickets to orconf and sell them on the black market. I could make millions from this14:38
olofkgahh!! Software must be the most stupid fucking idea ever16:06
stekernno, analog electronics must be the most stupid fucking idea ever...16:32
stekernmy brain can't grasp more than 2 states16:32
olofkTrue. Everyone knows that it's just crap between 0 and 116:33
stekernand everything is just imprecise guess-work16:33
olofkwoohoo! Looks like I actually made stuff compile16:38
olofk...speaking of imprecise guess-work16:48
stekernwoohoo! Looks like I got a LED to light up!17:22
stekern...and it's a crazy powerled, so I'm halfblind now17:23
olofkhmm.. I just realized that we have a proper DMA if we connect the stream interfaces of my reader and writer17:34
stekernthat's what I said (vaguely) before17:36
heshamstekern: Does your u-boot work with mor1kx and fusesoc for Atlys ?18:37
stekernlast I checked, yes. but it was a while ago18:41
stekernbut if you experience problems, I can take a look18:44
heshamI created .mcs file from orpsoc_top.bit, and then used promgen to embed the u-boot.bin, but it does not work.18:50
heshamI can investigate sometime to get some knowledge about mor1kx RTL, so you may just pass some docs and/or hints for me to start from.18:52
heshamBTW, the orpsocv2 works fine on my Atlys, with both bare-metal hello, and u-boot, however it does not work with the vmlinux image (that ran on or1ksim fine)18:59
stekernah.. there's no flash loader in the fusesoc/orpsoc-cores atlys board port19:00
stekernI just realised I never answered your question about that19:00
heshamnm, Can I help with that?19:01
jagadeeshhi how can i load fusesoc/de0_nano system along with application in ecps flash19:02
stekernyes, blueCmd_ already started something for the de0 nano board I think19:03
stekernbut the quick-n-dirty solution would be to just take the bootrom .v from orpsocv2 and use that19:03
jagadeeshcan i get a little more info of doing that19:04
heshamSo, if it's that easy, why not added so far?19:04
stekernbecause we don't want the dirty solution ;)19:05
heshamMakes sense ;)19:07
stekernthe 'proper' solution use olofk's wb_ram and preloads it with the bootcode19:07
jagadeeshis it like the bootcode in wb_ram copies the code in flash to sdram19:08
stekernand we of course want the bootloader to be available as source too (and an option to choose what bootloader to use), not just a 'binary' blob dumped in there19:08
stekernjagadeesh: yes, this was what the old orpsocv2 bootrom did19:08
heshamIn that case, the bootcode can be u-boot?19:09
jagadeesh u-boot does the same is it19:09
jagadeeshthen we should have to place the u-boot in wb_ram am i right19:10
stekernwell, u-boot is pretty large, but if you have enough SRAM/ROM space to share, basically yes19:11
stekern(on the atlys/de0 nano boards, you don't)19:12
heshamDo not have enough Flash space or what?19:13
jagadeeshdo not have enough sram/rom in fpga19:13
heshamThe SPI flash size is 16M for Atlys19:14
jagadeeshthe bootcode is something that resides in the processor in our case the fpga, am i right?19:15
-!- edward__ is now known as funfunctor19:51
jagadeeshwriting to epcs flash not successful, how can i find the offset address of the cof file21:15
jagadeeshthe sof takes place at starting of the flash followed by software hex21:16
jagadeeshdoes the problem lies in hex offset address or should i tel the bootrom.s to read from the starting address of the epcs flash hex content21:17
olofkstekern: I actually started a or1k-bootloaders core that I figured could contain some general-purpose bootloaders21:19
olofkFor de0 nano, I want to blink some LEDs, load from flash or execute from 0x100 depending on the state of the dip switches21:20
olofkI think that could be useful on more boards21:21
jagadeeshyeah me too want the same21:21
olofkBlinking LEDs by default is a great way to get some feedback21:21
olofkjagadeesh: Unfortunately we're not there quite yet. For some reason, the preloaded memories seem to be empty when I build the FPGA image21:22
jagadeeshis there any help i could get21:22
olofkjagadeesh: Yes. If you could figure out why the hell the memories aren't preloaded, that would be great21:22
olofkde0_nano in orpsoc-cores already uses wb_ram with a bootloader called clear_r3_and_jump_to_0x100. I used signal tap to see what the CPU reads from the bootloader memory, but it seems to be empty21:24
olofkIt would be interesting to see if it works on Xilinx devices21:24
olofkI'm not sure if I have done something really stupid, or just that Quartus fails to understand the coding style. Both are equally possible :)21:25
jagadeeshas far as bootloader im new21:26
olofkGot to sleep now. Going to FPGA World in Copenhagen early tomorrow21:26
olofkWhy the hell did I sign up to talk at both places? :/21:27
jagadeeshme too its 3 am here ;-)21:28
jagadeeshhave a good sleep21:28
jagadeeshi was trying to use the old bootloaders21:28
blueCmd_stekern: (since jagadeesh has left): has some stuff related to that22:29
blueCmd_JIC files and what-ever-they-are-named22:29
heshamblueCmd_: Should not simple_spi.v and fifo4.v exist there?
heshamIt should not, it's checked out from opencores repo23:45
--- Log closed Thu Sep 11 00:00:47 2014

Generated by 2.15.2 by Marius Gedminas - find it at!