Tuesday, 2018-10-09

*** tpb has joined #timvideos00:00
mithroshorne: When you get a moment, can you give me a tip on what I need to add to my gcc branch to try again with -mror ?00:32
*** tac-tics has quit IRC00:57
shornemithro: Yes, basically just add another option to gcc/config/or1k/or1k.opt, you can make a dummy one since we dont implement ror yet.02:10
mithroshorne: Hrm? You haven't actually implement ror?02:11
shornebut, ill create a or1k-next tag that you can foloow. (I rather use that then latest since not all my changes will be or1k branches)02:11
mithroshorne: okay02:12
shornemithro: thats the "beauty" gcc, if you dont implement something it will emulate it02:12
shorne(not so optimally) but it will work02:12
shorneI plan to target those optimizations after upstreaming02:12
shorneAlso, I haven't see a ror workload slow anything down for me yet02:13
shornemithro: btw, I think I figured out the issue with \r\n in verilator emulator02:13
shornepatch coming up (for litex)02:13
mithroshorne: I'm mainly interested in making sure that your upstream or1k works with our workload so we can start using upstream gcc when there is a newrelease02:15
shornemithro: sure, so my patches should be fine, I have implemented all the old hooks I could find02:17
shornemaybe not -mff1..02:17
shornemithro: fyi, https://github.com/enjoy-digital/litex/pull/11602:26
shorneyour pointer was correct02:26
shorneI just had to take some time to understand verilator and the plugins framework in litex02:27
mithroshorne: eekk that is some horrible code :-P02:27
mithroshorne: I don't think you made it _worse_ however....02:28
cr1901_modernIt looks okay to me...02:29
mithroshorne: I think you actually want to just modify02:30
tpbTitle: litex/serial2console.c at 8877dba7e96115ffbc406461919fca0d5aee8b0b · enjoy-digital/litex · GitHub (at github.com)02:30
shornemithro: you might be referring to the hard coded 2048, and ring buffer stuff, its pretty ugly, but makes sense02:31
shorneand its pretty compact so doesnt warrant making it more abstract I think02:31
shornemithro: oh, can the termios flags be used to create a \r?02:32
shorneok, let me look at that, I didnt know too much about termios02:32
mithroI can never remeber what you want02:33
mithro       INLCR  Translate NL to CR on input.02:33
mithro       IGNCR  Ignore carriage return on input.02:33
mithro       ICRNL  Translate carriage return to newline on input (unless IGNCR is set).02:33
shorneI think all of those would break something i.e. Litex BIOS expects '\n',  H2U firmware expects '\r'02:36
shornewe need to get '\r\n' or we need to fix one of the firmwares02:36
shornemithro: Ill see if there is a translate NL to CRNL02:37
mithroNewline translation – if a newline ("\n", ASCII 0x0A) comes in the right, a carriage-return/line-feed combination (CRLF, "\r\n", ASCII 0x0D 0x0A) is sent out the left. While most programs on UNIX accept a bare "\n" as a newline, your terminal needs both – "\r" tells it to move the cursor to the beginning of the line, and "\n" tells it to move to the next line.02:38
shornemithro: I created this tab, I will add it to my push script to push changes there, https://github.com/stffrdhrn/gcc/tree/or1k-next02:40
tpbTitle: GitHub - stffrdhrn/gcc at or1k-next (at github.com)02:40
mithroshorne: ONLCR  = 0000004 // ONLCR Map NL -> CR-NL02:41
shorneThere is ONLCR  (XSI) Map NL to CR-NL on output.02:42
shorneok, let me try that instread02:42
mithroshorne: But I don't think that effects what is sent to the other end?02:42
shorneI think it might?  i.e. termios is controlling my psuedo terminal linux driver and will change was 'read()' returns right?02:43
shorneMight as well try it...02:44
shorne... oh, but it says 'on output'02:46
shornetrying anyway02:46
shornedoesnt work02:47
mithroDunno which way is input or output02:47
shorneinput is what we get from 'read' I guess02:48
shorneoutput is what terminal gets when we do 'printf'02:49
shorneanyway, the patch does work, I could add 'IGNCR  Ignore carriage return on input.' to termios structure02:49
shornemithro: cr1901_modern: adding IGNCR would prevent duplicate CR's in case we get '\r\n' directly from the the terminal02:50
cr1901_modernshorne: I'm on Windows FYI :P, so termios.h doesn't have meaning for me02:51
cr1901_modernI was just commenting that I thought your C code was fine02:52
shornecr1901_modern: got it02:54
shornemithro: I will leave the patch as is for now I am not sure who else uses verilator02:54
mithroshorne: I'll leave _florent_ to merge it as I can't test his platform02:55
cr1901_modernI use iverilog mostly, personally. Easier to set up IME02:55
shornecr1901_modern: up until now I used iverilog only as well, its very easy to setup02:56
shornemithro: did you have any luck with the new verilator 4.004 packages?  I am trying now...02:57
mithroshorne: They should be in conda now....02:58
mithroshorne: https://anaconda.org/TimVideos/verilator/files02:58
shornemithro: yes, I saw them, Also my sim build passes CI too02:59
shorneits not showing (green) because I cancelled all the rest I guess02:59
mithroSomething went wrong there...03:00
mithromake[2]: /tmp/really-really-really-really-really-really-really-really-really-really-really-really-really-long-path/conda/conda-bld/verilator_1538956858781/_build_env/bin/x86_64-conda_cos6-linux-gnu-c++: Command not found03:01
mithroDoesn't look like it _should_ have succeeded :-P03:01
shorneneed to actually check logs03:02
mithroDoes bring up the fact that verilator should be pulling in the compiler...03:02
shorneit should be there from build-essentials though?03:03
mithroshorne: Yes but it's trying to use the conda compiler that it was compiled with03:17
shornemithro: I see... I just reproduced on my desktop using verilator frmo conda03:18
mithroshorne: Just add a test to our conda build and building locally03:18
shornemithro: where is your conda build?  I see the conda repo, but wondering where you manage all the sources03:20
mithroshorne: https://github.com/timvideos/conda-hdmi2usb-packages ?03:20
tpbTitle: GitHub - timvideos/conda-hdmi2usb-packages: Conda build recipes for the toolchains needed by LiteX / MiSoC firmware (at github.com)03:20
mithroshorne: Building locally first03:21
shornemithro: ok, good, I didnt know, let me see if its on the wiki, I may add it there03:28
shornehopefully my random docs are helpful03:28
shorneFYI, now that sim is kind of working I am trying to run linux on it03:29
shornehmm, bios is not going into 'netboot'03:48
shorneI guess because of bootorder, its hitting the flash boot first, need to probably disable that for net targets03:50
*** rohitksingh_work has joined #timvideos03:53
shornesomething is wrong with the networking under verilator I think, packets are getting send but look wrong04:10
mithroshorne: https://github.com/timvideos/conda-hdmi2usb-packages/pull/5704:11
tpbTitle: verilator: Fix issue with hardcoded path in Verilator makefile. by mithro · Pull Request #57 · timvideos/conda-hdmi2usb-packages · GitHub (at github.com)04:11
cr1901_modernshorne: Asking this sincerely- is it possible the data is being recv'd reverse as intended04:11
cr1901_modernI've had that happen before04:12
cr1901_modern"ship it, fix it in the driver"04:12
shornecr1901_modern: I am thinking that... but maybe its something else taking over the port...04:13
shorne13:05:09.078240 IP > 0*- [0q] 3/0/0 (Cache flush) TXT "", (Cache flush) PTR lianli.local., (Cache flush) SRV lianli.local.:22 0 0 (122)04:13
shornethe is something real, but why is it coming from (the simulator ip and tap interface)04:14
shornemaybe getting reversed?04:14
mithroshorne: We used 6069 as a secondary tftp ?04:15
mithroshorne: Dunno if anything from https://github.com/timvideos/litex-buildenv/wiki/Networking helps?04:17
tpbTitle: Networking · timvideos/litex-buildenv Wiki · GitHub (at github.com)04:17
mithroshorne: going to walk home while I wait for verilator to build on travis04:18
shornemithro: ok, I checked it out04:20
shornemithro: it looks ok, one comment on how the test actually works, also I noticed you commented some stuff04:20
shornemithro: Right, the port should be 6069, this mdns stuff is 535304:21
shornethe mdns packets are probably something on my network04:22
shorneIm going to go for a walk then go home too04:22
*** CarlFK has quit IRC05:34
*** sb0 has quit IRC05:54
mithroshorne: Merged the pull request, Verilator should be updated in about an hour06:01
*** futarisIRCcloud has quit IRC06:01
*** CuriousLearner has quit IRC06:14
*** CuriousLearner has joined #timvideos06:20
shornemithro: cool, I am debugging the sim ethernet stuff, I think once that is running and build is working, we would want to merge the sim patches.  I dont think it hurts much06:47
shorneI have things on my todo,  qemu args/upstreaming, linux upstreaming, linux ethernet support, linux gateware info driver06:48
*** sb0 has joined #timvideos07:57
shorneok, some progress on debugging ethernet stuff, I can see arp packets getting out, but my server is not replying, I "should" not have firewall rules blocking it08:35
_florent_shorne: do you have a way to check if the arp packet is valid?08:36
_florent_shorne: we did some changes recently to support little endian (risc-v)08:37
_florent_shorne: i tested with lm32 (big) and risc-v (little)08:37
_florent_shorne: i assume you are testing with mor1kx08:38
_florent_are you able to do a quick test with another cpu (lm32 or vexriscv?)08:38
_florent_also be sure to use latest liteeth08:39
shorne_florent_: it looks ok (I can see it in tcpdump)08:44
shorneit decodes correctly08:44
shornelet me try is lm3208:45
shorne_florent_: need to download conda toolchains...08:47
shornemight not be able to complete today,  Also, I am using litex-buildenv, so switching to latest liteeth/litex might break things and take time08:48
shorneill give it a try too08:48
shorne... give it a try though08:48
shorneNeed to start cooking dinner now08:48
_florent_shorne: i can also try to reproduce, which toolchain should i use for mor1kx?09:25
*** sb0 has quit IRC12:01
*** samsagaz has quit IRC12:09
*** rohitksingh_work has quit IRC12:33
*** sb0 has joined #timvideos13:02
*** CarlFK has joined #timvideos13:05
*** ChanServ sets mode: +v CarlFK13:05
*** rohitksingh has joined #timvideos13:14
*** samsagaz has joined #timvideos14:12
*** CarlFK has quit IRC14:12
*** CarlFK has joined #timvideos14:34
*** ChanServ sets mode: +v CarlFK14:34
*** Kripton has quit IRC14:39
*** rohitksingh has quit IRC14:39
*** Kripton has joined #timvideos14:42
*** Kripton has quit IRC14:47
*** CarlFK has quit IRC14:53
*** Kripton has joined #timvideos14:54
*** rohitksingh has joined #timvideos15:00
*** samsagaz has quit IRC15:12
*** CarlFK has joined #timvideos15:16
*** ChanServ sets mode: +v CarlFK15:16
*** rohitksingh has quit IRC17:20
*** CarlFK has quit IRC17:28
*** tac-tics has joined #timvideos17:43
*** rohitksingh has joined #timvideos18:06
*** rohitksingh has quit IRC18:35
*** ewen has joined #timvideos22:55
ewenmithro: Is there a FPGA Miniconf (and/or pre-LCA2019 FPGA sprint in Christchurch) at LCA2019 in Jan 2019?  The TimVideos survey sent out a few weeks ago implied there was, but maybe that was copied from the survey for pre-LCA2018?22:57
ewenmithro: https://lca2019.linux.org.au/programme/miniconfs/ doesn't seem to show a FPGA Miniconf...22:57
tpbTitle: linux.conf.au 2019 | MiniconfsTwitter (at lca2019.linux.org.au)22:57
*** CarlFK has joined #timvideos22:58
*** ChanServ sets mode: +v CarlFK22:58
mithroewen: There is no FPGA miniconf at LCA201923:12
ewenmithro: Thanks, that's what I'd guessed.23:12
ewenmithro: I'm trying to figure out how early to go down for LCA2019.23:12
ewenLeaning towards the Wednesday before rather than the Monday before now.23:13
ewenmithro: Are you planning on being in Christchurch for LCA2019?  If so are you planning on arriving earlier in the week before?23:14
mithroewen: I'm speaking23:15
ewenSo that's probably a "yes" to attending :-)23:15
tpbTitle: linux.conf.au 2019 | Presentation: SymbiFlow - The next generation FOSS FPGA toolchainTwitter (at lca2019.linux.org.au)23:16
ewenmithro: BTW, they stopped doing Markdown formatting of the abstracts (just plain text) so you might want to tweak that first link.23:17
ewenmithro: Is it likely there'll be a "FPGA/FuPy" sprint *all week* before LCA2019, like 2018?  Or just maybe some stuff at conference/the weekend before.23:18
ewen(I'm needing to book accommodation now, which will fix when I'm travelling.)23:18
mithroewen: I guess I should think about that sometime soon23:53
mithroshorne: Your newest gcc for or1k is in conda23:53
mithroshorne: and it seems to at least compile our firmware :-P23:53
ewenmithro: I'm likely to book accommodation (and maybe travel) this week, possibly today.  Currently I think I'll be in Christchurch Wed 16 Jan - Sat 26 Jan (2019)23:56
ewenmithro: If you want to have a sprint for longer than that I *might* be able to adjust accommodation earlier later on.  But my most likely accommodation choice doesn't seem very communicative :-(23:57

Generated by irclog2html.py 2.13.1 by Marius Gedminas - find it at mg.pov.lt!