00:40
acathla has quit [Ping timeout: 260 seconds]
00:47
<
GitHub55 >
artiq/new-py2llvm c21387d whitequark: compiler.embedding: support calling methods marked as @kernel.
00:47
<
whitequark >
ugh. that turned out to have to be extremely contrived in order to work
00:48
acathla has joined #m-labs
01:00
ylamarre has quit [Quit: ylamarre]
01:52
<
sb0 >
rjo, so how did you debug it?
02:03
<
GitHub95 >
artiq/master ebfd8c9 Sebastien Bourdeauducq: language/environment: fix set_* with parent
02:03
<
GitHub95 >
artiq/master c7d9bb7 Sebastien Bourdeauducq: Merge branch 'master' of github.com:m-labs/artiq
02:16
<
sb0 >
yes, FUD is one rtio clock long
02:54
travis-ci has joined #m-labs
02:54
<
travis-ci >
m-labs/artiq#447 (master - c7d9bb7 : Sebastien Bourdeauducq): The build passed.
02:54
travis-ci has left #m-labs [#m-labs]
03:27
_florent__ has joined #m-labs
03:27
_florent_ has quit [Ping timeout: 246 seconds]
03:28
_florent__ is now known as
_florent_
04:30
larsc has quit [Remote host closed the connection]
05:36
<
whitequark >
what's a good way to serialize execution of python processes?
05:36
<
whitequark >
lit tries to parallelize tests, which obviously breaks if i try to use serial communication
05:36
<
whitequark >
actually, nevermind, i should not use lit for that.
05:44
acathla has quit [Changing host]
05:44
acathla has joined #m-labs
05:51
<
GitHub153 >
artiq/new-py2llvm d80be48 whitequark: Allow clearing core device log buffer....
05:51
<
GitHub153 >
artiq/new-py2llvm 7c1abb2 whitequark: compiler.embedding: test all diagnostics....
05:51
<
GitHub153 >
artiq/new-py2llvm 13e612c whitequark: Fix tests.
06:00
<
whitequark >
sb0, is that a better diagnostic than artiq.transforms.tools.NotASTRepresentable? :]
06:14
<
GitHub194 >
artiq/new-py2llvm 6b55e3b whitequark: compiler.embedding: use qualified name when embedding methods.
06:42
FabM has quit [Remote host closed the connection]
06:43
FabM has joined #m-labs
06:44
<
GitHub160 >
artiq/new-py2llvm d473d58 whitequark: artiq_{compile,run}: adapt to new compiler.
06:53
<
rjo >
clocking the rtio domain phase locked w.r.t the dds synclk worked reliably (~100 trials). rtio domain at 125mhz, dds at 2.4ghz, synclk thus at 100mhz. that tests all kinds of fud-synclk alignments. and most of them meet setup hold specification.
06:58
<
rjo >
but the setup was a bit ragged. i would want to redo it to be sure. and really test 1e9 trials with feedback to the fpga. we could do something pseudo-digital with a short (~1us) phase pulse at zero ftw. then match the amplitude to ttl and feed it into the core device.
06:58
<
rjo >
and raghu did a lot of the work.
07:11
<
GitHub2 >
artiq/new-py2llvm ed236eb whitequark: artiq_run: pretty-print diagnostics.
07:11
<
GitHub2 >
artiq/new-py2llvm 9936768 whitequark: ARTIQIRGenerator: fix non-nullary method calls.
07:11
<
GitHub2 >
artiq/new-py2llvm 9605e82 whitequark: coredevice.ttl: update for new int semantics.
07:14
larsc has joined #m-labs
07:23
<
GitHub7 >
artiq/new-py2llvm 677cc69 whitequark: compiler.embedding: actually use qualified name when embedding methods....
07:23
<
GitHub7 >
artiq/new-py2llvm edf33f1 whitequark: compiler.targets: dump module signature with ARTIQ_DUMP_SIG=1.
07:23
<
GitHub7 >
artiq/new-py2llvm 6b8ef8c whitequark: artiq_run: use "artiq_run_" as user code module prefix, not "file_import_".
07:51
<
GitHub198 >
artiq/new-py2llvm ac92aab whitequark: Fix default argument fiasco.
07:51
<
whitequark >
I want to find whoever came up with the default argument semantics and force them to write Python every single day
07:51
<
whitequark >
as a punishment
07:53
<
GitHub54 >
artiq/master 241b765 Yann Sionneau: manual: fix typo in package name
07:53
<
GitHub14 >
artiq/new-py2llvm 6a29775 whitequark: compiler.types.Type: more useful __str__.
08:06
<
GitHub150 >
artiq/new-py2llvm 9fd25a1 whitequark: LLVMIRGenerator: fix syscall emission.
08:06
<
GitHub150 >
artiq/new-py2llvm 30cdb20 whitequark: log.c: fix warnings.
08:06
<
GitHub150 >
artiq/new-py2llvm 98bb570 whitequark: log.c: fix off-by-one error.
08:07
aeris_ has joined #m-labs
08:24
<
GitHub3 >
artiq/new-py2llvm 83ebb99 whitequark: transforms.Inferencer: improve attribute unification diagnostic.
08:24
<
GitHub3 >
artiq/new-py2llvm c621b1f whitequark: compiler: handle language.core.int during embedding.
08:31
<
whitequark >
alright, the LED example now works unmodified if I change now_mu() to now_init() in ttl.py (since there's no now_mu() yet...)
08:31
<
whitequark >
but all the embedding stuff works correctly.
08:33
<
GitHub114 >
artiq/master a36a50b Sebastien Bourdeauducq: comm_tcp: add connect timeout
08:33
<
GitHub114 >
artiq/master 031d8d0 Sebastien Bourdeauducq: doc/manual: add note about bidirectional TTL in tutorial
08:33
<
GitHub114 >
artiq/master 1cace1c Sebastien Bourdeauducq: style
08:33
<
whitequark >
seems to generate massive amounts of assembly for simple functions for some reason
08:33
<
whitequark >
ohhh, it's the loads from closures + spills necessary to free up registers for those
08:34
<
whitequark >
should get amortized for large functions, I think
08:41
<
sb0 >
whitequark, cool!
08:41
<
sb0 >
the now_init stuff should be quite straightforward, I think
08:41
<
whitequark >
it's a part of interleaving...
08:42
<
sb0 >
re. NotASTRepresentable. were you referring to Joe's issue?
08:43
<
sb0 >
we should be able to use Scannable on the device (but that is a lower priority than interleaving)
08:43
<
sb0 >
well, not Scannable, the objects it creates
08:44
<
sb0 >
rjo, how can it work if the rtio clock is faster than synclk?
08:44
travis-ci has joined #m-labs
08:44
<
travis-ci >
m-labs/artiq#448 (master - 241b765 : Yann Sionneau): The build passed.
08:44
travis-ci has left #m-labs [#m-labs]
08:44
<
sb0 >
the dds code assumes that the rtio clock has the same frequency as synclk, and is phase-locked
08:45
<
whitequark >
sb0: there are no generators at the moment... hrm
08:54
<
GitHub148 >
artiq/new-py2llvm 72823cf whitequark: test.{coredevice,coredevice_vs_host} → test.coredevice.{rtio,portability}.
09:26
travis-ci has joined #m-labs
09:26
<
travis-ci >
m-labs/artiq#449 (master - 1cace1c : Sebastien Bourdeauducq): The build passed.
09:26
travis-ci has left #m-labs [#m-labs]
09:35
fengling has quit [Ping timeout: 245 seconds]
09:42
fengling has joined #m-labs
09:56
<
GitHub90 >
artiq/new-py2llvm 79af228 whitequark: ksupport: provide abort() to the kernel library.
09:56
<
GitHub90 >
artiq/new-py2llvm 5916c10 whitequark: ARTIQException: replace ARTIQ library path with <artiq> in tracebacks.
10:00
<
whitequark >
fucking amazing
10:00
<
whitequark >
we can't convert lambdas to RPCs
10:03
<
whitequark >
because there's no way to get column number and thus distinguish between two lambdas on the same line
10:09
<
whitequark >
python's inspect.getsource does ''.join(lines)
10:09
<
whitequark >
becaue ~~python is garbage~~
10:15
<
GitHub115 >
artiq/new-py2llvm b35051c whitequark: compiler.embedding: handle calls into lambdas (poorly).
10:15
<
GitHub115 >
artiq/new-py2llvm bcba86d whitequark: compiler.embedding: handle errors during initial kernel call.
10:15
<
GitHub115 >
artiq/new-py2llvm 37811f6 whitequark: ARTIQIRGenerator: support comparisons against None.
10:16
<
whitequark >
ah, you can't annotate return type of lambda anyway
10:16
<
whitequark >
so the problem sort of fixes itself
10:25
<
GitHub152 >
artiq/new-py2llvm cbd903a whitequark: compiler.embedding: add tests for quote serialization.
10:27
_florent_ has quit [*.net *.split]
10:27
attie has quit [*.net *.split]
10:27
early has quit [*.net *.split]
10:27
jaeckel has quit [*.net *.split]
10:27
MiW has quit [*.net *.split]
10:27
awallin_ has quit [*.net *.split]
10:27
siruf has quit [*.net *.split]
10:27
mindrunner has quit [*.net *.split]
10:29
<
GitHub80 >
artiq/new-py2llvm 2124ff9 whitequark: Fix tests.
10:29
<
whitequark >
alright, that makes me sure enough it's not going to break randomly.
10:29
awallin_ has joined #m-labs
10:29
siruf has joined #m-labs
10:29
mindrunner has joined #m-labs
10:29
<
whitequark >
time to figure out interleaving... whoops, it's 0530
10:30
_florent_ has joined #m-labs
10:30
attie has joined #m-labs
10:30
early has joined #m-labs
10:30
jaeckel has joined #m-labs
10:30
MiW has joined #m-labs
10:30
_florent_ has quit [*.net *.split]
10:30
attie has quit [*.net *.split]
10:30
early has quit [*.net *.split]
10:30
jaeckel has quit [*.net *.split]
10:30
MiW has quit [*.net *.split]
10:32
_florent_ has joined #m-labs
10:32
early has joined #m-labs
10:32
jaeckel has joined #m-labs
10:32
attie has joined #m-labs
10:32
MiW has joined #m-labs
11:30
fengling has quit [Quit: WeeChat 1.2]
11:44
aeris_ has quit [Ping timeout: 246 seconds]
12:20
aeris_ has joined #m-labs
15:51
<
GitHub87 >
artiq/master 2056831 Yann Sionneau: manual: add install instructions for pygit2
15:53
aeris_ has quit [Ping timeout: 272 seconds]
16:01
<
GitHub17 >
artiq/master fc1c7c2 Yann Sionneau: setup.py: automate installation of llvmlite_artiq dependency
16:20
<
rjo >
sb0: i said "most of them".
16:25
sb0 has quit [Quit: Leaving]
16:50
travis-ci has joined #m-labs
16:50
<
travis-ci >
m-labs/artiq#450 (master - 2056831 : Yann Sionneau): The build passed.
16:50
travis-ci has left #m-labs [#m-labs]
17:27
travis-ci has joined #m-labs
17:27
<
travis-ci >
m-labs/artiq#451 (master - fc1c7c2 : Yann Sionneau): The build has errored.
17:27
travis-ci has left #m-labs [#m-labs]
18:09
<
ysionneau >
whitequark: humm when llvmlite_artiq is automatically fetched as an artiq dependency (when doing python setup.py develop --user from artiq root dir) it does install llvmlite_artiq ... but without the libllvmlite.so file
18:09
<
ysionneau >
it is not even built
18:09
<
ysionneau >
it seems to only run python setup.py bdist_egg from the llvmlite root dir and it does not run build_ext target
18:10
<
ysionneau >
I'm absolutely not a setuptools rock star
18:11
<
ysionneau >
on the other hand the setup.py of llvmlite already has several subclassing of setuptools commands so ...
18:13
<
ysionneau >
hmm I could add a comment to explain why I do this
18:16
kaalia has joined #m-labs
21:37
ylamarre has joined #m-labs
22:08
fengling has joined #m-labs
22:47
travis-ci has joined #m-labs
22:47
<
travis-ci >
m-labs/artiq#451 (master - fc1c7c2 : Yann Sionneau): The build passed.
22:47
travis-ci has left #m-labs [#m-labs]
22:59
fengling has quit [Ping timeout: 245 seconds]