Friday, 2018-10-12

*** tpb has joined #timvideos00:00
*** mcarden has joined #timvideos00:00
*** Guest54781 has left #timvideos00:14
*** Toba has joined #timvideos00:15
*** tac-tics has quit IRC01:13
*** shivm28[m] has quit IRC01:37
*** shivm28[m] has joined #timvideos01:38
shornemithro: for the verilator conda package, I am thinking we dont want to distribute the conda build environment g++ package01:52
shorneThe packages we install like libftdi-dev , libusb-1.0-0-dev, we use with the host gcc (i.e. build-essential)01:54
shorneMaybe I am wrong01:56
*** shorne has quit IRC02:50
*** shorne[ic] has joined #timvideos02:53
*** shorne has joined #timvideos02:56
*** sb0 has quit IRC03:24
*** sb0 has joined #timvideos03:24
*** rohitksingh_work has joined #timvideos03:55
*** [d__d] has quit IRC04:11
*** paddatrapper has quit IRC04:12
*** paddatrapper has joined #timvideos04:12
*** [d__d] has joined #timvideos04:14
*** Kripton has quit IRC04:15
*** Kripton has joined #timvideos04:20
*** Kripton has quit IRC05:49
*** Kripton has joined #timvideos05:49
cr1901_moderndaveshah: Just for reference, how badly bugged are the hard IPs in up5k?07:48
daveshahcr1901_modern: I think it's mostly the documentation that is broken07:58
daveshahBut there are rumours of silicon bugs too07:58
daveshahThey also have a quite overcomplicated interface07:58
cr1901_moderndaveshah: Current status: https://imgur.com/a/pjsDaSC07:59
tpbTitle: Imgur: The magic of the Internet (at imgur.com)07:59
cr1901_modernI'm going to have to prob disable multipliers for now (or cache... blargh)07:59
daveshahWhat is the utilisation?08:00
cr1901_modern4840 out of 528008:00
cr1901_modernIt may actually route this design before timeout, but it's a bit unfair to expect a user to wait 30 mins08:01
daveshahYes, but it is a tradeoff08:01
daveshahThis is a high utilisation design also probably with a lot of relatively big buses around the place08:02
cr1901_modernThere was an issue inferring DSP multipliers on ice40?08:02
daveshahYes, Yosys doesn't support that at all08:02
cr1901_modernI know, but is it because the DSPs are fundamentally incompatible w/ e.g. Xilinx multipliers, or just nobody has done it?08:02
daveshahBecause Yosys doesn't really have a nice framework for doing it08:03
daveshahAnd for the iCE40 we didn't really care because it only has 8 of them08:03
daveshahWe will do it now that ECP5 is coming along08:04
cr1901_modernI might hack lm32 to add an instantiated up5k multiplier impl for now08:04
cr1901_modernSo I can keep the cache around08:04
daveshahThe iCE40 multipliers will need a bit of logic around them as they can't switch signedness on the fly08:05
cr1901_modern(The main problem w/ disabling multipliers is _florent_ reported that it broke things)08:06
*** tac-tics has joined #timvideos08:07
cr1901_modernmithro: https://github.com/cr1901/litex-buildenv/commit/450a31f559fb8c54ee1ab039df12beba43f841a5 Lots of progress made tonight. Bad news- lm32.lite doesn't route on up5k... the pnr tool runs for 30 minutes before giving up.08:49
tpbTitle: platforms: Add ICE40UP5K-B-EVN platform, and base.py and bridge.py ta… · cr1901/[email protected] · GitHub (at github.com)08:49
cr1901_modernhttp://ix.io/1oYN08:51
cr1901_moderndaveshah: What does "ERROR: Failed to re-route arc 2 of net lm32_cpu.instruction_unit.instruction_d[5]." mean in English?08:52
daveshahBasically it has managed to paint itself into a corner, such that one connection of a net is totally unroutable for some reason08:54
cr1901_modernSo why doesn't it do more ripup?08:55
cr1901_modernwithout knowing nextpnr's internals... feels weird that it does all that ripup and then decides "f*** it, I'm not doing this anymore"08:56
daveshahistr there is a maximum rip up limit08:58
daveshahBut I don't know the current router algorithm in detail08:58
cr1901_modernWell it was just a curiosity09:00
cr1901_modernI would've figured it would've given up sooner :P09:00
daveshahMy understanding is that the penalty for a rip up increases with time09:02
daveshahRip ups need to be encouraged to deal with suboptimal initial routing, but there must be something to stop it getting stuck in a loop09:02
cr1901_modernHmmm09:03
* cr1901_modern gracefully bows out tonight09:04
cr1901_modernI was actually productive- quite a bit so. I think I earned a break.09:04
cr1901_modernBtw mithro: This was downright awful to do :P https://github.com/m-labs/migen/pull/14509:04
cr1901_modernNot looking forward to the other lattice boards09:05
*** mauz555 has joined #timvideos09:38
*** maks_ has joined #timvideos10:32
*** maks_ has quit IRC11:03
*** mauz555 has quit IRC11:26
*** mauz555 has joined #timvideos11:26
*** mauz555 has quit IRC11:30
*** mauz555 has joined #timvideos11:31
*** rohitksingh_work has quit IRC12:36
*** mauz555 has quit IRC13:07
*** mauz555 has joined #timvideos13:13
*** rohitksingh has joined #timvideos13:15
mithrorohitksingh: ping?13:59
rohitksinghmithro: pong!14:00
mithrorohitksingh: If you add the mimas a7 to https://github.com/timvideos/litex-buildenv-udev/blob/master/98-hdmi2usb-dev-boards.rules it should get the correct permissions, ignored by modem manager and a bunch of extra aliases14:00
tpbTitle: litex-buildenv-udev/98-hdmi2usb-dev-boards.rules at master · timvideos/litex-buildenv-udev · GitHub (at github.com)14:00
rohitksinghmithro: oh okay, I'll add it there below arty14:01
mithrorohitksingh: You might need to add a 97-hdmi2usb-dev-boards-ftdi.rules or something to do the ftdi modprobe thingy earlier14:02
rohitksinghmithro: okay, let me push the updates so you can review them14:03
rohitksinghmithro: can you please check now https://github.com/timvideos/litex-buildenv-udev/pull/1/files14:25
tpbTitle: Add Mimas A7 rules by rohitk-singh · Pull Request #1 · timvideos/litex-buildenv-udev · GitHub (at github.com)14:25
mithroLGTM!14:28
*** Amirag has joined #timvideos14:29
*** Amirag has quit IRC14:31
*** somepotatoro has joined #timvideos14:40
felix_is there any abstraction for the polarity of lvds pairs (p/n swapped or not swapped) in litex?14:43
felix_oh and is there a lightweight example project where litex/lite* is included as submodule? neither the "just check out some version and install it" nor using litex-buildenv seems to me to be the best way to do that for photonsdi14:45
*** somepotatoro has quit IRC14:46
felix_btw: is there another channel for litex/lite* or is this the right channel to discuss things related to that?14:48
xobsfelix_: this seems to be a good channel for litex discussion.14:54
felix_ok :)14:54
xobsfelix_: also, you can try https://github.com/xobs/lxbuildenv it's somewhat experimental but it's what I use.14:55
tpbTitle: GitHub - xobs/lxbuildenv: Simplified environment for litex (at github.com)14:55
xobslxbuildenv is my take on the problem. You add "import lxbuildenv" to the top of your program, and it checks out modules and patches up the environment as necessary.14:57
felix_thx; i'll have a closer look later. from a brief look it seems though that it does much more than i need15:00
felix_basically i want to have lite* as submodules and make sure that those get used; the project is already created manually15:00
rohitksinghmithro: you have openocd patch also!15:01
*** dwellereE has joined #timvideos15:05
*** dwellereE has quit IRC15:14
*** disposablecT has joined #timvideos15:16
*** disposablecT has quit IRC15:28
*** shakygatormH has joined #timvideos15:36
*** tracernzlg has joined #timvideos15:39
*** twkVn has joined #timvideos15:42
*** shakygatormH has quit IRC15:42
*** AndyWojozp has joined #timvideos15:44
*** twkVn has quit IRC15:46
*** platvoetenOu has joined #timvideos15:46
*** platvoetenOu has quit IRC15:46
*** AndyWojozp has quit IRC15:47
*** tracernzlg has quit IRC15:50
mithrorohitksingh: Just waiting for travis to say it's okay15:50
*** a1cypher has joined #timvideos15:53
*** a1cypher has quit IRC15:55
*** rohitksingh has quit IRC16:05
*** orrcLJ has joined #timvideos16:23
*** orrcLJ has quit IRC16:24
*** rohitksingh has joined #timvideos16:38
rohitksinghmithro: travis build has passed. Please do vote on the upstream openocd patch also16:47
mithrorohitksingh: merged!16:52
mithroxobs: are you able to test the latest litex-buildenv toolchain?16:53
rohitksinghmithro: awesome! thank you so much! up next is HDMI2USB-litex-firmware16:54
xobsmithro: not until Monday.  I take it a lot has changed?16:54
mithroIt sound no longer have the vsyscall issue16:54
xobsOh yay!  I'll see if I can't try it sooner.16:54
mithroS/sound/should/16:54
mithroxobs: also it is now GCC 8.216:55
mithroAnd there is RISC-V toolchain to16:55
xobsFancy!  I think Bunnie wanted a riscv gcc 8.2 for some reason.16:55
mithroPlus Yosys and nextpnr if you want to test out ice40 without needing to install any vendor tools!16:56
*** NightAPt has joined #timvideos17:04
*** hreinnbeckPV has joined #timvideos17:05
*** NightAPt has quit IRC17:06
*** hreinnbeckPV has quit IRC17:12
*** Guest26093 has joined #timvideos17:13
*** tac-tics_ has joined #timvideos17:15
*** Guest26093 has quit IRC17:19
*** PaulFertserwV has joined #timvideos17:21
*** PaulFertserwV has quit IRC17:26
*** TheAssassin has quit IRC18:17
*** mauz555 has quit IRC18:17
*** Kydontk has joined #timvideos18:20
*** Kydontk has quit IRC18:21
*** TheAssassin has joined #timvideos18:23
*** FennyeH has joined #timvideos18:41
*** rohitksingh has quit IRC19:17
*** rohitksingh has joined #timvideos19:18
*** gthm_tq has joined #timvideos19:47
*** gthm_tq has quit IRC19:49
*** rohitksingh has quit IRC19:50
*** jbetauG has joined #timvideos19:57
*** jbetauG has quit IRC19:58
*** yambo has joined #timvideos20:31
*** GNU\colossuspT has joined #timvideos20:37
*** sttic has joined #timvideos20:40
*** lionelmK has joined #timvideos20:40
*** yambo has quit IRC20:42
*** GNU\colossuspT has quit IRC20:43
*** tac-tics_ has quit IRC20:46
*** sttic has quit IRC20:47
*** tac-tics_ has joined #timvideos20:56
*** lexano has quit IRC23:42
*** lexano has joined #timvideos23:54

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