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
futarisIRCcloud has quit [Quit: Connection closed for inactivity]
<sb0> hartytp, who is AUTProgram?
<sb0> hartytp, this has already been suggested and discussed, and we're OK with that, but it is not funded
<sb0> whitequark, now that flterm support has been removed from the bootloader, and that the sdram support in the bootloader and misoc have diverged, it is a pain to bypass the dysfunctional sayma flash on marmelada's board ...
<sb0> whitequark, in general 1) things like the bootloader should be designed assuming that every single component on the target board is going to malfunction at some point and needs a bypass so that progress can continue to be made 2) replacing something is done only after it has feature and issue parity with the old solution
<whitequark> (1) makes sense but in practice (2) just makes sure that nothing can ever be replaced with anything else without wasting a huge amount of time, or ever
<whitequark> if you mean that the rust bootloader should support SFL, then just say that...
<sb0> whitequark, yes it should support SFL
<sb0> whitequark, I don't see how it "wastes" more time than plowing ahead with the replacement and then firefighting
<sb0> (2) basically is planning things carefully
<sb0> whitequark, when you replaced the compiler, there was a period where it worked significantly worse than mine and I had to explain why. ditto smoltcp, on a much worse scale. now the bootloader. it would be a pity not to see a pattern and learn lessons there.
<sb0> whitequark, and this is not a criticism of the replacements themselves, which are clear improvements compared what a was there before, but of the careless way in which they are made, which causes a lot of practical problems and pain and frustration to users
<sb0> whitequark, if you look at SED, which is one major replacement I made, you can see that it was 1) discussed beforehand with stakeholders 2) implemented in a branch 3) merged only after it passed unittests and was tested by different people
<sb0> and I have raised the SFL issue earlier, but since there are so many things going on I didn't press on
rohitksingh_work has joined #m-labs
<whitequark> SFL is just one thing the misoc bootloader supports. it has a lot more functionality that no one has missed.
<sb0> yet? I use the mr/mw commands regularly
<sb0> I just haven't written much relevant gateware lately
<sb0> netboot can be dropped/delayed, or replaced with something better
<sb0> also this whole thing is a distraction from the camera driver and the allaki failures
<sb0> the camera driver and allaki are funded and visible to users, the bootloader replacement is neither
ncl has quit [Read error: Connection reset by peer]
ncl has joined #m-labs
rohitksingh_work has quit [Ping timeout: 260 seconds]
hartytp has joined #m-labs
<hartytp> sbo: one of our students
<hartytp> Not asking for anything from you right now
<hartytp> And know this isn't funded
<hartytp> Just giving you a heads up that we're doing it
<hartytp> If you want us to push upstream when we're done then we will otherwise we will just use locally
<hartytp> Your call
hartytp has quit [Ping timeout: 260 seconds]
<GitHub-m-labs> [artiq] enjoy-digital commented on issue #967: It's working on a simple design (Nexys Video / SERWB overa HDMI cable and the two HDMI TX/RX ports / A master and a Slave in the same design). I'm doing more testing on that (P&R is only 30 seconds, so it's easier), cleaning up things, and i'll test/integrate in ARTIQ. https://github.com/m-labs/artiq/issues/967#issuecomment-379163341
<whitequark> sb0: I was finally able get some work done today
<whitequark> currently waiting for another build... that should be all I think
<GitHub160> [llvm-or1k] whitequark pushed 1 new commit to artiq-6.0: https://github.com/m-labs/llvm-or1k/commit/b534f6535c95065a1d1d67330fa0ea2ad4710d4b
<GitHub160> llvm-or1k/artiq-6.0 b534f65 whitequark: [SelectionDAG] Fix return calling convention in expansion of ?MULO....
<whitequark> bb-m-labs: force build --props=package=llvm-or1k conda-all
<bb-m-labs> build #110 forced
<bb-m-labs> I'll give a shout when the build finishes
rqou has quit [Ping timeout: 276 seconds]
rqou has joined #m-labs
<whitequark> sb0: I've seen you upstreamed some kasli patches, can you tell me the status of that?
<sb0> whitequark, I have reduced the cpu clock frequency, and the plan is to move to a faster speed grade in the next hardware revision
<sb0> current kasli uses the original settings but at 95% of the speed
<sb0> next one will have 100% (or more)
<whitequark> sb0: hm
<whitequark> so the LLVM patch wasn't necessary after all?
<whitequark> or can we go to 100% with the patch?
<sb0> we can't do 100% reliably on all targets with the patch.
<whitequark> ok, I see
<whitequark> that's annoying
<sb0> yeah welcome to FPGA world
<GitHub135> [llvm-or1k] whitequark pushed 1 new commit to artiq-6.0: https://github.com/m-labs/llvm-or1k/commit/3f6ef875b16491e13e536a0503dcbf3b904935db
<GitHub135> llvm-or1k/artiq-6.0 3f6ef87 whitequark: [OR1K] Add missing initializer.
<whitequark> well I'm not complaining
<bb-m-labs> build #383 of conda-lin64 is complete: Exception [exception interrupted] Build details are at http://buildbot.m-labs.hk/builders/conda-lin64/builds/383
<bb-m-labs> build #229 of conda-win64 is complete: Exception [exception interrupted] Build details are at http://buildbot.m-labs.hk/builders/conda-win64/builds/229
<bb-m-labs> build #110 of conda-all is complete: Failure [failed] Build details are at http://buildbot.m-labs.hk/builders/conda-all/builds/110
<whitequark> sigh
<GitHub159> [llvm-or1k] whitequark pushed 1 new commit to artiq-6.0: https://github.com/m-labs/llvm-or1k/commit/a5472fd63d8c0fc35fa4c50d3a05c4c82e6ef4e8
<GitHub159> llvm-or1k/artiq-6.0 a5472fd Eugene Zelenko: Add missing header for InstructionCombining.cpp, in order to export LLVMInitializeInstCombine as extern "C". Fixes PR35947....
<whitequark> bb-m-labs: force build --props=llvm-or1k conda-all
<bb-m-labs> Something bad happened (see logs)
<whitequark> bb-m-labs: force build --props=package=llvm-or1k conda-all
<bb-m-labs> build #111 forced
<bb-m-labs> I'll give a shout when the build finishes
<whitequark> okay, we're at LLVM 6.
<whitequark> or rather, will be once the builds finish
<whitequark> tests pass...
<whitequark> allaki now.
<whitequark> sb0: oh also when building packages now I'm enabling clang and rust on windows
<whitequark> any objections to that?
<whitequark> it means being able to install artiq-dev on windows.
<sb0> whitequark, that's okay, but not a priority
<whitequark> already done, it was copying a few lines...
<bb-m-labs> build #384 of conda-lin64 is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/conda-lin64/builds/384
<bb-m-labs> build #230 of conda-win64 is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/conda-win64/builds/230
<bb-m-labs> build #111 of conda-all is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/conda-all/builds/111
<GitHub> [conda-recipes] whitequark pushed 1 new commit to master: https://github.com/m-labs/conda-recipes/commit/d9f6128730bc1f6a5c5c17949a4fc6eac11eb87f
<GitHub> conda-recipes/master d9f6128 whitequark: llvmlite-artiq: update to 0.23.0 (LLVM 6.0.0).
<whitequark> bb-m-labs: force build --props=package=llvmlite-artiq conda-all
<bb-m-labs> build forced [ETA 59m15s]
<bb-m-labs> I'll give a shout when the build finishes
<bb-m-labs> build #231 of conda-win64 is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/conda-win64/builds/231
<bb-m-labs> build #385 of conda-lin64 is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/conda-lin64/builds/385
<bb-m-labs> build #112 of conda-all is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/conda-all/builds/112
rohitksingh_work has joined #m-labs
futarisIRCcloud has joined #m-labs
rohitksingh_work has quit [Read error: Connection reset by peer]
<GitHub> [conda-recipes] whitequark pushed 1 new commit to master: https://github.com/m-labs/conda-recipes/commit/d2e641b5be40064199bd05923eafa627736a7551
<GitHub> conda-recipes/master d2e641b whitequark: llvm-or1k: build tools on Windows.
<whitequark> bb-m-labs: force build --props=package=llvm-or1k conda-win64
<bb-m-labs> build forced [ETA 32m06s]
<bb-m-labs> I'll give a shout when the build finishes
<GitHub16> [rust] whitequark created artiq-1.25.0 (+4424 new commits): https://github.com/m-labs/rust/compare/cf49d6b080bb^...82c97b2c0536
<GitHub16> rust/artiq-1.25.0 4e80530 Scott Olson: Sort uses.
<GitHub16> rust/artiq-1.25.0 fbec376 Scott Olson: Add more tests. Remove unecessary attributes.
<GitHub16> rust/artiq-1.25.0 cf49d6b Scott Olson: Add the interpreter from my rustc branch and hook it up to CompileController.
rohitksingh has joined #m-labs
<GitHub173> [rust] whitequark force-pushed artiq-1.25.0 from 82c97b2 to e7e8cec: https://github.com/m-labs/rust/commits/artiq-1.25.0
<GitHub173> rust/artiq-1.25.0 01ebd0b whitequark: Add OR1K support.
<GitHub173> rust/artiq-1.25.0 58ced4b whitequark: Add #[cfg(target_family="none")] support to libpanic_*.
<GitHub173> rust/artiq-1.25.0 c5ccc88 whitequark: Remove LLVM submodules.
<sb0> _florent_, the sayma should work again on the lab computer now, btw
<_florent_> sb0: ok, it was no longer working?
<sb0> you said jtag was broken due to ctrl-c on openocd
<sb0> or did usbreset work?
<_florent_> ah, yes usbreset worked
<_florent_> i'm using sayma3
<whitequark> there's no other sayma online atm
<whitequark> but I can bring up sayma1 if desired
<_florent_> whitequark: i saw that, that's fine with only sayma
<_florent_> sayma3
<whitequark> sb0: btw can we get internet in the lab any faster?
<whitequark> the build server wastes a lot of time downloading code at 500KB/s because conda can't cache git submodules
<sb0> 500KB is limited by the other endpoint I think
<whitequark> 500-1000KB/s
<sb0> the lab connection should do 1M
<whitequark> it's much faster at my home
<whitequark> might not go to 1M because of roundtrips in the protocol
FabM has quit [Quit: ChatZilla 0.9.93 [Firefox 52.7.2/20180316222652]]
rohitksingh has quit [Quit: Leaving.]
rohitksingh has joined #m-labs
<whitequark> bb-m-labs: force build --props=package=llvmlite conda-win64
<bb-m-labs> The build has been queued, I'll give a shout when it starts
rohitksingh has quit [Quit: Leaving.]
rohitksingh has joined #m-labs
rohitksingh has quit [Client Quit]
rohitksingh has joined #m-labs
<whitequark> sb0: uploaded 13579 of 300858Kb: 4.51% ETA: 76.7 minutes
<whitequark> this is bullshit
<whitequark> it literally takes longer to upload the package than to build it!
<sb0> and is that from the lab connection?
<whitequark> yes
<whitequark> I'm now going to rebuild everything with shared libs because it will take less time
<whitequark> which is absurd
<bb-m-labs> build #232 of conda-win64 is complete: Exception [exception interrupted] Build details are at http://buildbot.m-labs.hk/builders/conda-win64/builds/232
<bb-m-labs> build forced [ETA 32m06s]
<bb-m-labs> I'll give a shout when the build finishes
<bb-m-labs> build #233 of conda-win64 is complete: Exception [exception conda_build_output] Build details are at http://buildbot.m-labs.hk/builders/conda-win64/builds/233
<sb0> can you collect evidence and send a complaint to WT&T? it should be faster than that
<whitequark> speedtest.net says they give us symmetric 10/10
<sb0> yes, that's what the contract says too
<whitequark> so this is some kind of adverse interaction with anaconda.org
<whitequark> well, one silver lining of going to shared libraries is less rebuilds of dependents (rust, llvmlite) for llvm changes
<whitequark> I was getting seriously tired of rebuilding six packages for every minor change in LLVM
<whitequark> and in theory this shouldn't even break anything.
<whitequark> in theory.
<whitequark> sigh.
<GitHub> [conda-recipes] whitequark pushed 2 new commits to master: https://github.com/m-labs/conda-recipes/compare/08b106b4b6eb...caf8614adef8
<GitHub> conda-recipes/master caf8614 whitequark: llvm-or1k: build as shared libraries.
<GitHub> conda-recipes/master 29a46a5 whitequark: llvm-or1k: harmonize build scripts.
<whitequark> bb-m-labs: force build --props=package=llvm-or1k conda-all
<bb-m-labs> build forced [ETA 32m11s]
<bb-m-labs> I'll give a shout when the build finishes
<whitequark> maybe I should just put the build server at my home
<whitequark> that would solve this particular problem, at least.
<GitHub-m-labs> [artiq] cjbe opened issue #980: Dashboard causes 'asyncio:Fatal write error on socket transport' occasionally https://github.com/m-labs/artiq/issues/980
<bb-m-labs> build #386 of conda-lin64 is complete: Failure [failed conda_build] Build details are at http://buildbot.m-labs.hk/builders/conda-lin64/builds/386
<bb-m-labs> build #234 of conda-win64 is complete: Exception [exception interrupted] Build details are at http://buildbot.m-labs.hk/builders/conda-win64/builds/234
<bb-m-labs> build #113 of conda-all is complete: Failure [failed] Build details are at http://buildbot.m-labs.hk/builders/conda-all/builds/113
<GitHub> [conda-recipes] whitequark pushed 1 new commit to master: https://github.com/m-labs/conda-recipes/commit/da12be4c52c6badc9f33b83d2ee1e5f71991bdab
<GitHub> conda-recipes/master da12be4 whitequark: llvm-or1k: don't build clang docs or tests.
<whitequark> bb-m-labs: force build --props=package=llvm-or1k conda-all
<bb-m-labs> build forced [ETA 32m11s]
<bb-m-labs> I'll give a shout when the build finishes
<whitequark> _florent_: sb0: can I use sayma?
<sb0> whitequark, sure. use the lockfile
<whitequark> it's locked.
<sb0> by whom?
<whitequark> _florent_
<bb-m-labs> build #387 of conda-lin64 is complete: Failure [failed conda_build] Build details are at http://buildbot.m-labs.hk/builders/conda-lin64/builds/387
<bb-m-labs> build #235 of conda-win64 is complete: Exception [exception interrupted] Build details are at http://buildbot.m-labs.hk/builders/conda-win64/builds/235
<bb-m-labs> build #114 of conda-all is complete: Failure [failed] Build details are at http://buildbot.m-labs.hk/builders/conda-all/builds/114
rohitksingh has quit [Quit: Leaving.]
rohitksingh has joined #m-labs
<GitHub123> [llvm-or1k] whitequark pushed 1 new commit to artiq-6.0: https://github.com/m-labs/llvm-or1k/commit/527aa86b578da5dfb9cf4510b71f0f46a11249f7
<GitHub123> llvm-or1k/artiq-6.0 527aa86 whitequark: [CMake] Only build LLVMTestingSupport if LLVM_INCLUDE_TESTS=ON....
<whitequark> bb-m-labs: force build --props=package=llvm-or1k conda-all
<bb-m-labs> build forced [ETA 32m11s]
<bb-m-labs> I'll give a shout when the build finishes
<bb-m-labs> build #236 of conda-win64 is complete: Failure [failed conda_build] Build details are at http://buildbot.m-labs.hk/builders/conda-win64/builds/236
<bb-m-labs> build #388 of conda-lin64 is complete: Exception [exception interrupted] Build details are at http://buildbot.m-labs.hk/builders/conda-lin64/builds/388
<bb-m-labs> build #115 of conda-all is complete: Failure [failed] Build details are at http://buildbot.m-labs.hk/builders/conda-all/builds/115
<GitHub> conda-recipes/master f5c1686 whitequark: llvm-or1k: build as a *single* shared library, not split ones.
<GitHub> [conda-recipes] whitequark pushed 1 new commit to master: https://github.com/m-labs/conda-recipes/commit/f5c1686fb6ea68029ea257b53919c80e2cbfac49
<whitequark> bb-m-labs: force build --props=package=llvm-or1k conda-all
<bb-m-labs> build forced [ETA 32m11s]
<bb-m-labs> I'll give a shout when the build finishes
<whitequark> okay, guess I can't work on allaki today...
rohitksingh has quit [Quit: Leaving.]
ncl has quit [Ping timeout: 268 seconds]
<bb-m-labs> build #389 of conda-lin64 is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/conda-lin64/builds/389
<_florent_> whitequark: i'd like to have access to a sayma to test serwb
<_florent_> whitequark: if you want to use sayma3, can you power on the other sayma?
<_florent_> whitequark: but working together on two different sayma will probably be a bit difficult since we need to power off/power on to workadound the 1.8v issue
<_florent_> whitequark: i need one board the next 3-4 hours i think, after that it should be good
X-Scale has quit [Ping timeout: 240 seconds]
mumptai has joined #m-labs
X-Scale has joined #m-labs
<bb-m-labs> build #237 of conda-win64 is complete: Failure [failed conda_build] Build details are at http://buildbot.m-labs.hk/builders/conda-win64/builds/237
<bb-m-labs> build #116 of conda-all is complete: Failure [failed] Build details are at http://buildbot.m-labs.hk/builders/conda-all/builds/116
<whitequark> _florent_: well it's 1:34 here, I already left the lab or I couldn't have returned on MTR
<whitequark> the problem is that I need specifically sayma3
<whitequark> because either sayma1 or sayma1's RTM are broken
<whitequark> so I can't debug allaki on those
<_florent_> whitequark: yes i was aware it was very late for you
<whitequark> I mean, I would've stayed in the lab overnight if I had access to Sayma
<_florent_> whitequark: i try to finish today
<whitequark> thanks!
<whitequark> okay, building LLVM as one large dylib reduced the size to 89M from 250M on Linux and to "build failure" on Windows
<whitequark> I don't know if a dog has Buddha-nature but conda definitely has Sayma-nature
ncl has joined #m-labs
<whitequark> oh.
<whitequark> it's 80M because it built everything as static libraries AND a shared library.
<whitequark> this means that llvmlite will be statically linked so it's good I guess
<GitHub> [conda-recipes] whitequark pushed 1 new commit to master: https://github.com/m-labs/conda-recipes/commit/014cf458426057c3369bbcd641d341236686f5f3
<GitHub> conda-recipes/master 014cf45 whitequark: llvm-or1k: LLVM_LINK_LLVM_DYLIB seems kinda broken on Windows.
<whitequark> bb-m-labs: force build --props=package=llvm-or1k conda-win64
<bb-m-labs> build forced [ETA 32m06s]
<bb-m-labs> I'll give a shout when the build finishes
<whitequark> bb-m-labs: force build --props=package=llvmlite conda-lin64
<bb-m-labs> build forced [ETA 40m32s]
<bb-m-labs> I'll give a shout when the build finishes
<bb-m-labs> build #390 of conda-lin64 is complete: Exception [exception conda_build_output] Build details are at http://buildbot.m-labs.hk/builders/conda-lin64/builds/390
<whitequark> bb-m-labs: force build --props=package=llvmlite-artiq conda-lin64
<bb-m-labs> build forced [ETA 40m32s]
<bb-m-labs> I'll give a shout when the build finishes
<bb-m-labs> build #391 of conda-lin64 is complete: Failure [failed conda_build] Build details are at http://buildbot.m-labs.hk/builders/conda-lin64/builds/391
<GitHub2> [llvmlite] whitequark pushed 1 new commit to artiq-6.0: https://github.com/m-labs/llvmlite/commit/8529e300a5a992e3bcb0fd8e2e56f134d2a6e3fa
<GitHub2> llvmlite/artiq-6.0 8529e30 whitequark: Ignore libLLVM.so when building bindings.
<whitequark> bb-m-labs: force build --props=package=llvmlite-artiq conda-lin64
<bb-m-labs> build forced [ETA 40m32s]
<bb-m-labs> I'll give a shout when the build finishes
<bb-m-labs> build #392 of conda-lin64 is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/conda-lin64/builds/392
<GitHub> [conda-recipes] whitequark pushed 1 new commit to master: https://github.com/m-labs/conda-recipes/commit/fbc056b39b8fc67d5f570d8eb6661fbc17e4ee61
<GitHub> conda-recipes/master fbc056b whitequark: llvmlite-artiq: add missing runtime dependencies.
<GitHub> [conda-recipes] whitequark force-pushed master from fbc056b to 014cf45: https://github.com/m-labs/conda-recipes/commits/master
<whitequark> pfahaha that's still 272MB on Windows
<whitequark> because LLVM can't link tools to libLLVM.so I guess
<whitequark> what a trash fire
<whitequark> and no it's not the ISP, it's the S3's bad connectivity and anaconda's poor choice of storage provider who's at fault
<bb-m-labs> build #238 of conda-win64 is complete: Success [build successful] Build details are at http://buildbot.m-labs.hk/builders/conda-win64/builds/238
<GitHub-m-labs> [artiq] hartytp commented on issue #967: @sbourdeauducq With the caveat that I'm not aware of the precise details of any of the relevant contracts, and with apologies if this is treading on peoples' toes here, but...... https://github.com/m-labs/artiq/issues/967#issuecomment-378239721
<whitequark> bb-m-labs: force build --props=package=llvmlite-artiq conda-win64
<bb-m-labs> build forced [ETA 1h34m44s]
<bb-m-labs> I'll give a shout when the build finishes
<bb-m-labs> build #239 of conda-win64 is complete: Failure [failed conda_build] Build details are at http://buildbot.m-labs.hk/builders/conda-win64/builds/239