IRC logs for #openrisc Thursday, 2014-09-11

--- Log opened Thu Sep 11 00:00:47 2014
stekernolofk: may I suggest that the 'busy' signal is renamed to either 'stall' or invert it and call it 'read_enable'/'write_enable' (re/we)?04:00
stekernto me, busy sounds like the core is busy and can't address a request04:01
stekernI think I'd prefer re/we04:10
olofkstekern: Yes, I would like to invert it too, but I couldn't come up with a good name06:23
olofkI'm thinking of using rdy or ready instead, as in AXI406:26
olofkCould someone throw me some rough speed/area estimates of mor1kx in different configurations?06:44
stekernolofk: but rdy still sounds like an output, not an input07:22
olofkHeavily inspired by AXI4 here07:23
olofkIn which case it's always contra directonal to the data flow07:23
poke53281stekern ran some benchmarks. I think, the speed is comparable to a Pentium 90.07:23
poke53281But this depends on his FPGA07:24
stekernolofk: yes, but the rdy signals there are prepended by the 'signal type' name07:25
olofkIsn't that the case here as well? stream_rdy, stream_data, stream_dv?07:26
olofkI love the smell of bikeshedding in the morning :)07:26
jagadeeshstekern: u pointed me to but i have a trouble in cof file's offset hex address and07:27
stekernno, because stream_data and stream_dv are outputs07:27
olofkAnd stream_rdy is contra directional to that07:27
stekernyes, and that's not how axi4 does it07:27
olofkYes, it does07:27
stekernafaict at least07:27
olofkI've been doing a lot of AXI4 (Stream to be more exact) the last years07:28
stekernok, but then axi4 is backwards ;)07:29
olofkYes, in so many ways, but that's beside the point ;)07:30
stekernmy argument is that stream_rdy sounds like stream_data is ready, not that someone wants to acquire data from it07:32
olofkI see your point07:32
stekernfor the reader, rdy might make sense07:32
stekernbikeshedding yes ;)07:34
stekernjagadeesh: what's the problem?07:34
jagadeeshthe cof file addresses? how do i determine that07:35
jagadeeshthe sof data chooses the starting address of epcs followed by hex data07:36
stekernolofk: speed/area estimates. in the minimal cappuccino configuration around 100MHz on cyclone iv, 200MHz on kintex 7. area, according to sb0 - bloated ;)07:36
olofkCan that run Linux comfortably?07:37
stekernjagadeesh: I don't understand the question, I gave you an example
stekernolofk: that is with the mmu disabled, so no07:39
olofkSo, about 80 on CycloneIV with a standardish config?07:39
stekern85 MHz was the latest figure quartus gave me07:40
stekernjagadeesh: please write your answers here and not in priv07:41
stekernI didn't read the privmsgs07:41
jagadeeshim using this cof file for jic file generation, i dont get the code working. Question: do i have problem with cof file or the bootrom initialization07:44
stekernbut you don't have any addresses defined there07:45
jagadeeshyes i dont know how to detemine the address, for the cof i showed u the sof is at start of address and hex is next to it. this i learned for map file07:46
jagadeeshBut i think the bootrom reads from epsc start address to laod to sdram am i right07:46
jagadeeshTel me where i went wrong07:47
stekernyou just need to define them so there's enough room07:47
jagadeeshThere is no problem with placed of data in epcs i.e 0 to some xxxx for sof and xxxx to xxxx for hex07:48
stekernok, but then you need to just define the address in the bootrom bootloader07:49
jagadeeshah thats where i was stuck07:49
jagadeeshtel me is this right: I mention start address of hex to bootrom so that it writes my hex data to sdram from 0x10007:50
jagadeeshstekern can point to some documentation i dont have proper documents for orpsoc system devlopment07:52
stekernthat's the address that was used in orpsocv207:52
jagadeeshstekern: is there any document for newbies on orpsocv208:00
stekernwell, there's the orpsocv2 documentation in the repo08:01
stekernbut you shouldn't really bother with orpsocv2 other than stealing the bootrom code from it08:02
jagadeeshyes i got stuck with bootrom code i can run app from sdram.. writing to epcs is my problem08:04
stekernyes, but you got the address now, 0x0b000008:07
jagadeeshyes i ll try with this again08:09
olofkstekern: I'm renaming them to data, valid and ready. Both AXI4 and Avalon Stream uses those names08:47
stekernok, I agree with the consistency in that09:11
-!- aburgess_ is now known as aburgess09:15
stekernolofk: is the idea here to just set a default that can be overridden?
stekerni.e. MAX_BURST_LEN can be != 2**FIFO_AW in some cases?09:19
sb0olofk, yes. bloated. as in a few times the size of LM32, with similar functionality and performance.09:23
stekerna few times is exaggerating, it's not even twice the size10:17
stekernolofk: you are driving wbm_cti_o from two different blocks: and
heshamstekern: The simple approach of copying bootrom.v with hard-coded boot program did not work. Not sure the problem is with spi, uart, or the way I generate the mcs file. I think I noticed something says the UART won't work with some core.10:45
jagadeeshhesham: me too trying for the same but with de0_nano for orpsocv210:54
stekernhesham: what did you put in the mcs?10:55
heshamjagadeesh: I was able to run orpsocv2 with bare-metal hello and u-boot for Atlys10:55
heshamstekern: Bare-metal hello (worked with orpsocv2)10:55
heshambitgen -w -g StartUpClk:CClk orpsoc_top.ncd orpsoc_top_spi.bit10:55
jagadeeshyes can u do it with writing the fpga configuration and sw to flash10:55
heshampromgen -spi -p mcs -w -o orpsoc_top_spi.mcs -s 16384 -u 0 orpsoc_top_spi.bit -data_file up 1c0000 hello-bsw.bin10:56
heshamjagadeesh: Yes10:56
jagadeeshgreat i'm stuck here with my de0_nano board10:57
jagadeeshstekern suggested me to checkout the bootrom file10:58
jagadeeshi have some addressing issues in copying the data form flash to sdram10:58
heshamjagadeesh: When I used orpsocv2, the command to get it write was so easy (make orpsoc.mcs BOOTLOADER_BIN=/path/to/ur/bootloader)10:58
jagadeeshyeah did u try this link
jagadeeshI think u got it right11:00
heshamjagadeesh: I did not see this link before but I used another tutorial that is similar to this link. However, I used hello and u-boot instead of the LED program there, and it works.11:11
jagadeeshhesham: r u trying with fusesoc11:13
heshamYes, but was not lucky until now..11:14
jagadeeshdid u try copying the rom.v along with the generated bootrom.v that we use in orpsocv211:15
jagadeeshin orpsocv2 rom.v holds the bootcode11:16
jagadeesholofk was trying some bootloader for fusesoc11:17
jagadeeshthe first thing i do if i got this right is a good document11:24
heshamIn orpsocv2, bootrom.v also included in rom.v, I did the same with fusesoc, but it did not work.11:25
jagadeeshhave anyone tried orpsoc with de0 nano applications running from epcs12:36
jagadeeshsomeone refer me any link12:36
jagadeesha simple led blink example would help12:37
olofkstekern: I had some ideas that it could be more efficient to allow different max burst lengths and FIFO depth, but now I only think that will make things more complicated17:11
olofkstekern: Nice catch with wbm_cti_o17:12
olofkOn the way back from FPGA world now. Seems like documentation is the number one issue people have with this project17:14
stekernolofk: quartus did the catching ;)17:23
mor1kx[mor1kx] skristiansson pushed 2 new commits to master:
mor1kxmor1kx/master bcbcde3 Stefan Kristiansson: simple_dpram_sclk: change "TRUE"/"FALSE" parameter logic to 1/017:24
mor1kxmor1kx/master 6394a80 Stefan Kristiansson: simple_dpram_sclk: add read enable logic...17:24
stekernmor1kx_rf_ram, be afraid, be very afraid. Your days are counted17:25
olofkThank you Quartus then ;)17:25
jagadeeshstekern:i got leds up with epcs flash in de0_nano... thank u very much17:56
jagadeeshstekern:now processing with fusesoc for the same17:56
stekernjagadeesh: \o/17:57
jagadeeshstekern:the mistake i did was not reading the user-guide properly.... any way i got it working thank you again17:57
jagadeeshi will make a clean document or a blog post on this....17:58
stekernno problems, you did the lion's share of it anyway17:58
olofkjagadeesh: Cool. Blog posts about this is very welcome. Just tell me if you want some proof-reading17:59
jagadeesholofk: yeah sure ill frame the contents and ask u if i need help.......... thank you all for your help18:01
PaulfraOSAAolofk, thank you for an interesting presentation today18:13
stekernPaulfraOSAA: he made some sound here a moment ago18:21
PaulfraOSAAOk, he mentioned a or1k version starting with m today, just trying to google it and failing18:24
olofkPaulfraOSAA: pong!18:26
olofkTest bench is paying off. Starting to catch quality bugs now. stekern, be prepared for a push of my stream writer later tonight18:27
* stekern is waiting anticipatedly19:02 soon as I catch the bug in my fifo_reader19:04
olofk...which I have been trying to do for some time19:05
olofkI'll push it anyway19:05
mor1kx[mor1kx] skristiansson pushed 2 new commits to master:
mor1kxmor1kx/master 0454b5b Stefan Kristiansson: store_buffer: remove redundant fifo_raddr signal...19:08
mor1kxmor1kx/master 2440fab Stefan Kristiansson: store_buffer: use re logic of ram instead of registering fifo_dout...19:08
olofkHmm.. looks like it works now19:13
olofkAnd regression tests for fifo still works as well19:13
olofkYep. fifo and wb_streamer pushed now19:16
olofkThat was a good train journey19:16
stekernok, let me take it for a spin19:18
olofkLooks like someone else is working on a SoC building tool19:19
olofkCompetition is bad. I will start spreading lies about his project19:20
stekernolofk: you still have the duplicate wbm_cti assignment19:52
mor1kx[mor1kx] skristiansson pushed 1 new commit to master:
mor1kxmor1kx/master 9abd671 Stefan Kristiansson: rf: use mor1kx_simple_dpram_sclk instead of rf_ram...20:33
olofkstekern: Doh! Forgot to check wb_stream_ctrl.v. Was concentrating on the other direction today20:41
olofkFixed now (I think)20:41
poke53281*Frontier: Elite II* runs. And it's even playable!21:00
_franck__stekern: (SoC building tool),OpenRISC,0,553221:01
-!- _franck__ is now known as _franck_21:01
poke53281olofk:    "cd /usr/share/games/frontier" and then ./frontier21:02
poke53281The resolution is still 320x200. But I will change that.21:02
olofkpoke53281: That's so fucking cool :)21:19
stekernbut that makes me wonder why there's a too21:41
poke53281olofk: Interesting is the conversion history: David Braben's brain -> Atari ST Assembler -> Atari ST 68k binary -> Atari ST disassembly -> C-Code -> openrisc binary.21:48
poke53281and the binary has now a size of almost 5MB.22:06
poke53281"You may ....  give exact copies of the original evaluation SOFTWARE to anyone; and distribute the22:11
poke53281evaluation SOFTWARE in its unmodified form via electronic means"22:11
poke53281stekern, olofk: Do you think, that this conversion can be called "unmodified"? ;)22:11
--- Log closed Fri Sep 12 00:00:48 2014

Generated by 2.15.2 by Marius Gedminas - find it at!