sb0 changed the topic of #m-labs to: ARTIQ, Migen, MiSoC, Mixxeo & other M-Labs projects :: fka #milkymist :: Logs http://irclog.whitequark.org/m-labs
_whitelogger has joined #m-labs
<sb0> whitequark, if we have a CPU with separate memories for data and code, can rust/llvm target it? e.g. reading the program memory would not be allowed
<ohsix> dont' know much about rust/llvm in particular, but there are other targets where the 'memories' are the same but need special instructions to access and they manage fine, i think that implies very different memories would work
<ohsix> err, the bit that followed doesn't apply to rust, just llvm
<mithro> sb0: that might be useful from verifying that my stuff is mostly valid - I'm on a yak shave of adding ac97 decoding to sigrok at the moment
<sb0> whitequark, where would you buy m6 stainless steel screws (for cf35 vacuum flanges) of various lengths?
<sb0> silver plated if not expensive, otherwise i can just use the mos2
<sb0> I could also cut those I have to the desired length but that sounds messy
<mithro> sb0: Did anyone ever port the m1 soc to somewhat run on the mixxeo (even if just very hackily)?
<sb0> I want good quality ones that won't break (removing any broken part stuck in my slightly expensive cf35 cube sounds annoying) so not so keen on trying random hardware store stuff
<sb0> mithro, no, mixxeo always used misoc
<mithro> sb0: Yeah - I thought as much
<GitHub193> [sinara] marmeladapk pushed 1 new commit to master: https://github.com/m-labs/sinara/commit/bd01541494d3a419dfa12c1cfc38648d017e85a5
<GitHub193> sinara/master bd01541 marmeladapk: Updated power supplies schematic, added leds for each voltage...
rohitksingh_work has joined #m-labs
kaalia has quit [Ping timeout: 246 seconds]
<GitHub84> [migen] sbourdeauducq pushed 1 new commit to master: https://git.io/v5cN3
<GitHub84> migen/master 2e53295 Sebastien Bourdeauducq: sayma_rtm: add pulldowns on SPI MISO
<bb-m-labs> build #185 of migen is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/migen/builds/185
<whitequark> sb0: not allowed to read code at all? hmmmm let me think about it
<whitequark> yes, I think that could be reasonably done
<whitequark> I don't think the OR1K target actually reads any code, it doesn't have "constant islands" or something
<whitequark> sb0: regarding SS screws, do you mean cutting by sawing off? normal bolt/screw cutters fail if you use them with SS
<whitequark> anyway, silver plated is probably taobao, these are special stuff, and regular ones you can get at reclamation st in one of the zillion hardware shops
<whitequark> it's pretty hard to fuck up an SS screw...
<GitHub3> [smoltcp] whitequark commented on issue #34: Okay, so I think we have some confusion with this PR.... https://git.io/v5cpx
<GitHub164> [smoltcp] whitequark commented on issue #34: Okay, so I think we have some confusion with this PR.... https://git.io/v5cpx
<GitHub182> [smoltcp] whitequark pushed 1 new commit to master: https://git.io/v5chZ
<GitHub182> smoltcp/master 614323b whitequark: README: add some notable omissions.
<travis-ci> m-labs/smoltcp#185 (master - 614323b : whitequark): The build passed.
<GitHub164> [artiq] sbourdeauducq commented on issue #648: @cjbe Do you still want this? https://github.com/m-labs/artiq/issues/648#issuecomment-325908652
<GitHub80> [artiq] cjbe commented on issue #648: I still think it would be very useful, but have not come up with a tidy implementation. Happy to close this issue if you want. https://github.com/m-labs/artiq/issues/648#issuecomment-325910241
<GitHub72> [artiq] sbourdeauducq closed issue #648: Absolute imports of repository modules https://github.com/m-labs/artiq/issues/648
<GitHub117> [artiq] sbourdeauducq commented on issue #648: Yes. I'm not sure if there is any tidy solution that doesn't introduce a problem at least as bad as the initial one. https://github.com/m-labs/artiq/issues/648#issuecomment-325910651
<_florent_> sb0: for the delay-finding, no it's not handling all the cases right now (since you wanted to use phase_detector I haven't spent too much time on it), but if you are now fine with this method, I'll make sure we are covering all the cases and add testbenches for that.
<sb0> _florent_, ok. is the phase detector issue on the ultrascale or artix side? the ultrascale phase detector would be nice to have for drtio over rtm/amc
<_florent_> sb0: it's on the ultrascale side
<sb0> of course
<_florent_> sb0: I got an answer on xilinx forum but haven't tested it yet
<_florent_> sb0: the issue is that you want to be able to have the 90° phase shift between the master and slave serdes
<_florent_> sb0: when you use it in counter mode (ie you specify the number of taps to do 90°phase shift and not the time), it's working fine, init ok, reload ok
<_florent_> sb0: when you use it in time mode, init is working fine, but when you reload it's set to 0
<sb0> ah it's a "phase shift" on the input data, via idelay
<sb0> what is the difference between "init" and "reload"?
<_florent_> sb0: and the things with ultrascale is that you taps have variables delays
<sb0> because of no calibration like spartan6? or is it something else?
<_florent_> sb0: init: after reset, reload, you set the load pin of the serdes
<_florent_> sb0: tap delay can vary from 2.5 to 15 ps...
<sb0> they had cleaned that up in 7-series, and they fucked it again?
<_florent_> sb0: yes I found 7-series a lot better on that
<larsc> I believe 7-series also had these uncalibrated delays, they just didn't expose them
<larsc> (in the documentation)
<larsc> the uncalibrated delays give you a finer resolution at the cost that they are uncalibrated
<larsc> e.g. 7-series had 32 calibrated tabs, ultra scale has 512 uncalibrated tabs
<whitequark> sb0: item (c) done
<whitequark> ok, let's see how this fares against ARTIQ
<GitHub113> [smoltcp] whitequark pushed 2 new commits to master: https://git.io/v5CnZ
<GitHub113> smoltcp/master cd82aa6 whitequark: Rework TCP retransmit logic to be much more robust....
<GitHub113> smoltcp/master 3f69e40 whitequark: test
<whitequark> shit
<GitHub98> [smoltcp] whitequark force-pushed master from cd82aa6 to 79d8470: https://git.io/vMLjV
<GitHub98> smoltcp/master 79d8470 whitequark: Rework TCP retransmit logic to be much more robust....
<travis-ci> m-labs/smoltcp#186 (master - cd82aa6 : whitequark): The build has errored.
<travis-ci> m-labs/smoltcp#187 (master - 79d8470 : whitequark): The build passed.
<GitHub184> [smoltcp] whitequark pushed 1 new commit to master: https://git.io/v5Ccc
<GitHub184> smoltcp/master 1ece71a whitequark: Return from EthernetInterface::poll() on errors, don't swallow them....
<whitequark> sb0: okay, the TCP issue should be fixed, but fixing it exposed a different problem, I'm getting checksum errors now
<whitequark> let's see why this happens
karltheawesome has quit [Quit: Connection closed for inactivity]
rohitksingh_wor1 has joined #m-labs
rohitksingh_work has quit [Ping timeout: 248 seconds]
early has quit [*.net *.split]
Gurty has quit [*.net *.split]
cr1901_modern has quit [*.net *.split]
forrestv has quit [*.net *.split]
cr1901_modern has joined #m-labs
early has joined #m-labs
folkert has quit [Ping timeout: 240 seconds]
folkert has joined #m-labs
forrestv has joined #m-labs
Gurty has joined #m-labs
rqou_ has joined #m-labs
rqou has quit [Read error: Connection reset by peer]
rqou_ is now known as rqou
folkert has quit [Ping timeout: 240 seconds]
folkert has joined #m-labs
rohitksingh_wor1 has quit [Read error: Connection reset by peer]
<GitHub32> [artiq] enjoy-digital pushed 2 new commits to sinara: https://github.com/m-labs/artiq/compare/9ba50098a83e...32ca51faee17
<GitHub32> artiq/sinara 32ca51f Florent Kermarrec: gateware/targets/sayma_amc_standalone/rtm: use new serwb modules
<GitHub32> artiq/sinara 41d57d6 Florent Kermarrec: gateware/serwb: SERWBPLL, SERWBPHY, SERWBCore and add checks in delay finding to verify the sampling window
rohitksingh has joined #m-labs
<whitequark> aha, figured
<GitHub188> smoltcp/master 280ddde whitequark: Fix Ipv4Packet::payload{,_mut}() returning overly long buffers....
<GitHub188> [smoltcp] whitequark pushed 1 new commit to master: https://git.io/v5CXy
<GitHub49> [artiq] enjoy-digital pushed 1 new commit to sinara: https://github.com/m-labs/artiq/commit/96502330077d8ac62d040a2cca4d660185a9ab8e
<GitHub49> artiq/sinara 9650233 Florent Kermarrec: gateware/serwb: change serdes clock domain to serwb_serdes
<GitHub57> [smoltcp] whitequark pushed 1 new commit to master: https://git.io/v5CD8
<GitHub57> smoltcp/master a137126 whitequark: Move the TCP receive window clamping hack downwards in stack....
<travis-ci> m-labs/smoltcp#190 (master - a137126 : whitequark): The build passed.
<rjo> whitequark: Is one of the Saymas powered?
<GitHub104> [smoltcp] whitequark force-pushed master from a137126 to b927085: https://git.io/vMLjV
<GitHub104> smoltcp/master b927085 whitequark: Move the TCP receive window clamping hack downwards in stack....
<travis-ci> m-labs/smoltcp#191 (master - b927085 : whitequark): The build passed.
<whitequark> rjo: not yet, still fixing smoltcp, as you can see
<rjo> whitequark: ack. both are critically important. i can make do for another day or two without sayma. what's your schedule?
<whitequark> rjo: I'm making final touches
<whitequark> there are no blatant issues with smoltcp that I can see
<whitequark> rjo: *headdesk* I've just realized
<whitequark> the reason my throughput readings are slow is because my phone's internet is slow
<whitequark> not because smoltcp is
<whitequark> I'm too used to 100 Mbps internet... two whole days of waste
<whitequark> well not waste, it was useful refactoring, but we could've shipped earlier
<whitequark> sb0: ping
<rjo> whitequark: first world problems... ;)
<whitequark> rjo: technically that's not true
<whitequark> first world aka the US has 100 Mbps internet hardly anywhere
<whitequark> so it would be more like second world problems
<rjo> whitequark: i am (pain)fully aware. being a first world country depends on the category.
<whitequark> oh right.
<rjo> whitequark: telcos here have forever been pushing their stupid "vectoring" (of copper).
<whitequark> vectoring?
<rjo> just some DSP tech for DSL. some ideas are similar to MIMO for wireless.
<whitequark> hmm, weird
<whitequark> I got smoltcp working, but artiq_coreconfig write seems to be broken
<whitequark> these two facts do not seem related
<whitequark> ah no they are
<whitequark> some sort of contract violation on the higher layer, sec
<GitHub16> [artiq] whitequark pushed 2 new commits to master: https://github.com/m-labs/artiq/compare/c0eb2ad0b757...39ecbc0d689d
<GitHub16> artiq/master 39ecbc0 whitequark: firmware: update smoltcp.
<GitHub16> artiq/master 2231b16 whitequark: firmware: reduce ethmac maximum burst size by one....
<GitHub190> [artiq] whitequark commented on issue #685: This is finally fixed! :confetti_ball: ... https://github.com/m-labs/artiq/issues/685#issuecomment-326020886
<GitHub32> [artiq] dhslichter commented on issue #685: This is nice to have the improvement, thanks @whitequark. However, this is still very slow in the grand scheme of things, especially if we are going to larger Sinara setups. It seems that we should be aiming to go at least an order of magnitude faster (if not more) once we are working with Metlino cards driving multiple uTCA crates etc -- complex experiments, with all the attendant parameterized wavefor
<GitHub117> [artiq] whitequark pushed 1 new commit to master: https://github.com/m-labs/artiq/commit/f26e698f31e32067ddd84aea7c6c2c075120ebd3
<GitHub117> artiq/master f26e698 whitequark: Revert "firmware: reduce ethmac maximum burst size by one."...
<GitHub175> [artiq] whitequark commented on issue #685: @dhslichter I agree. This is actually slower than the results I've presented earlier in this thread.... https://github.com/m-labs/artiq/issues/685#issuecomment-326026523
<bb-m-labs> build #758 of artiq-board is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq-board/builds/758
<GitHub139> [artiq] dhslichter commented on issue #685: Sounds good @whitequark, and I am glad that you have managed to improve the reliability here (I do recall your reporting better numbers before, but with consistency/reliability issues). Definitely seems like tuning the TCP stack to optimize throughput will be a more straightforward task. Anyway, my main message is "good work!" :) https://github.com/m-labs/artiq/issues/685#issuecomment-326028894
<bb-m-labs> build #557 of artiq-win64-test is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq-win64-test/builds/557
<bb-m-labs> build #1658 of artiq is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq/builds/1658
<whitequark> aha
<whitequark> figured out why it's slow
<GitHub112> [artiq] enjoy-digital pushed 1 new commit to sinara: https://github.com/m-labs/artiq/commit/660f9856ec1fa4e1ab16b32988acd4c9720ced75
<GitHub112> artiq/sinara 660f985 Florent Kermarrec: gateware/serwb: add test for phy initialization
<bb-m-labs> build #759 of artiq-board is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq-board/builds/759
<bb-m-labs> build #558 of artiq-win64-test is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq-win64-test/builds/558
<bb-m-labs> build #1659 of artiq is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/artiq/builds/1659
rohitksingh has quit [Quit: Leaving.]
<GitHub188> [smoltcp] steffengy opened issue #36: Hardware Checksum Calculation https://git.io/v5WRQ
mumptai has joined #m-labs
<travis-ci> batonius/smoltcp#51 (master - b927085 : whitequark): The build passed.
<GitHub65> [smoltcp] whitequark commented on issue #36: Sure. Do you have any particular hardware in mind? Show me the datasheet and I'll think about exposing this capability. https://git.io/v5Wa7
<GitHub81> [smoltcp] steffengy commented on issue #36: STM32F2xxx/F4xxx (F4 is just a superset with more functionality basically, ethernet is afaik the same):... https://git.io/v5WV8
<GitHub173> [smoltcp] steffengy commented on issue #36: STM32F2xxx/F4xxx (F4 is just a superset with more functionality basically, ethernet is afaik the same):... https://git.io/v5WV8
<GitHub41> [smoltcp] steffengy commented on issue #36: STM32F2xxx/F4xxx (F4 is just a superset with more functionality basically, ethernet is afaik the same):... https://git.io/v5WV8
<GitHub146> [smoltcp] whitequark pushed 1 new commit to master: https://git.io/v5WHB
<GitHub146> smoltcp/master 3226770 whitequark: Fix the TCP SEQ acceptability check....
<travis-ci> m-labs/smoltcp#192 (master - 3226770 : whitequark): The build passed.
<GitHub39> [smoltcp] batonius commented on issue #34: No, it's ok, I should have thought it through before making the PR.... https://git.io/v5W5R
<travis-ci> batonius/smoltcp#52 (expandable_ring_buffer - e346dd9 : Egor Karavaev): The build passed.
<travis-ci> batonius/smoltcp#54 (expandable_ring_buffer - 54aa56a : Egor Karavaev): The build passed.
<travis-ci> batonius/smoltcp#53 (master - 3226770 : whitequark): The build passed.
<GitHub54> [smoltcp] whitequark force-pushed master from 3226770 to 3ab58b0: https://git.io/vMLjV
<GitHub54> smoltcp/master 3ab58b0 whitequark: Fix the TCP SEQ acceptability check....
Gurty has quit [Ping timeout: 252 seconds]
<travis-ci> batonius/smoltcp#55 (master - 3ab58b0 : whitequark): The build passed.
<travis-ci> batonius/smoltcp#56 (expandable_ring_buffer - 2605e00 : Egor Karavaev): The build passed.
Gurty has joined #m-labs
Gurty has quit [Changing host]
Gurty has joined #m-labs
<GitHub175> [smoltcp] whitequark pushed 2 new commits to master: https://git.io/v5WhW
<GitHub175> smoltcp/master 21282fd whitequark: Send a TCP ACK after window increases from zero to non-zero.
<GitHub175> smoltcp/master 86a05f1 whitequark: More rigorously treat the TcpSocket::remote_last_ack field....
<travis-ci> m-labs/smoltcp#197 (master - 21282fd : whitequark): The build passed.
mumptai has quit [Remote host closed the connection]
<whitequark> rjo: tcptrace is very awkward to use.
<whitequark> the interface is overly contrived, and xplot is completely unable to legibly render on hidpi displays
<whitequark> wireshark does it nicely though