Perl 6 - the future is here, just unevenly distributed

IRC log for #moarvm, 2017-04-28

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
01:49 ilbot3 joined #moarvm
01:49 Topic for #moarvm is now https://github.com/moarvm/moarvm | IRC logs at  http://irclog.perlgeek.de/moarvm/today
04:46 KDr2_c joined #moarvm
06:39 domidumont joined #moarvm
06:45 domidumont joined #moarvm
07:43 domidumont joined #moarvm
07:50 TimToady joined #moarvm
08:27 zakharyas joined #moarvm
11:49 lizmat hmmm...notice of build breakage on CentOS6
11:49 lizmat https://github.com/MoarVM/MoarVM/issues/587
11:50 timotimo thanks
13:05 AlexDaniel joined #moarvm
13:43 stmuk_ timotimo: see my comment on that issue
13:59 timotimo stmuk_: can you help me get the rdtscp configure-time probe working?
13:59 timotimo oh
13:59 timotimo i think i know what's wrong
14:01 Geth ¦ MoarVM/configure-telemeh-probe-rdtscp: e7e46332d8 | (Timo Paulssen)++ | src/profiler/telemeh.c
14:01 Geth ¦ MoarVM/configure-telemeh-probe-rdtscp: we always define MVM_RDTSCP, but sometimes with 0 instead of 1.
14:01 Geth ¦ MoarVM/configure-telemeh-probe-rdtscp: review: https://github.com/MoarVM/MoarVM/commit/e7e46332d8
14:05 stmuk_ timotimo: sure
14:05 stmuk_ shall I test now or wait a bit?
14:06 timotimo now
14:07 stmuk_ timotimo: https://gist.github.com/stmuk/5a09e24384984da2865c1dc59ca29fbb
14:07 timotimo interesting
14:08 timotimo i'll reproduce locally by just setting RDTSCP to 0 in the makefile
14:09 timotimo huh
14:10 Geth ¦ MoarVM/configure-telemeh-probe-rdtscp: bcd8b59c81 | (Timo Paulssen)++ | src/profiler/telemeh.c
14:10 Geth ¦ MoarVM/configure-telemeh-probe-rdtscp: give __rdtscp a more proper stub implementation
14:10 Geth ¦ MoarVM/configure-telemeh-probe-rdtscp: review: https://github.com/MoarVM/MoarVM/commit/bcd8b59c81
14:10 timotimo should be bettern ow
14:12 stmuk_ works fine on Centos 6 (I can test on OpenBSD later)
14:13 stmuk_ timotimo++
14:13 timotimo fantastic
14:52 Geth ¦ MoarVM/master: 7 commits pushed by (Timo Paulssen)++
14:52 Geth ¦ MoarVM/master: b77a3b81a3 | enable --telemeh, --no-telemeh, and probe for rdtscp
14:52 Geth ¦ MoarVM/master: 1850cb96d8 | document --no-telemeh and --telemeh
14:52 Geth ¦ MoarVM/master: 4ba8a172ba | include or exclude telemetry log env var in usage/--help
14:52 Geth ¦ MoarVM/master: 387a771d8f | fix copy-pasto, MasterDuke17++
14:52 Geth ¦ MoarVM/master: fdc653db0e | we always define MVM_RDTSCP, but sometimes with 0 instead of 1.
14:52 Geth ¦ MoarVM/master: b373dcdaef | give __rdtscp a more proper stub implementation
14:52 Geth ¦ MoarVM/master: 66c6ddabb5 | Merge branch 'configure-telemeh-probe-rdtscp'
14:52 Geth ¦ MoarVM/master: review: https://github.com/MoarVM/MoarVM/compare/e5b527edbb...66c6ddabb5
14:55 Zoffix How do you pronounce "telemeh"? "t-eh-l-eh-m-eh"?
14:58 stmuk_ tele meh
15:19 timotimo tell-uh-meh
15:21 robertle jsut like telomere
16:00 geekosaur joined #moarvm
16:04 geekosaur joined #moarvm
16:40 domidumont joined #moarvm
16:56 geekosaur joined #moarvm
17:21 timotimo hm. i still don't know how we could cause all other threads to be interrupted reliably so that we can slurp up their profile data
17:22 MasterDuke_ that would be awesome though...
17:23 timotimo i mean, normally when we exit we just have the other threads killed dead in their tracks by the process just ending
17:33 timotimo we could probably abuse the gc signaling mechanism, but instead of doing gc, we do our profile stuff
17:42 timotimo the thing is ... we probably need to GC multiple times inside the process of creating the data structures from the profile data
17:42 timotimo however
17:43 timotimo hm
17:43 timotimo i thought perhaps we can just recursively memcpy stuff away so the gc doesn't touch it any more, but that sounds like recipe for absolute disaster
17:47 AlexDaniel joined #moarvm
18:32 timotimo though of course by turning on gen2-allocatoin-by-default gets us out of "having to gc a bunch" trouble
18:55 samcv good *
18:55 timotimo nood goon
18:55 timotimo that came out strange
18:57 samcv heh
19:42 timotimo the cat really likes the upper left edge of my laptop screen
19:47 lizmat timotimo: perhaps that's the warmest part ?
19:48 * lizmat remembers our cats liking to sleep on top of a monitor
19:48 lizmat but that was in the pre-flatscreen days  :-)
19:52 timotimo heh
19:52 timotimo nah, he's rubbing his flew on it
19:59 geekosaur joined #moarvm
20:10 * Zoffix wonders what a flew of a cat is :)
20:13 timotimo the dictionary spat that word out
20:14 timotimo hm, seems like that's reserved for canines
20:14 Zoffix face?
20:14 Zoffix whiskers?
20:14 timotimo "upper lip"
20:14 timotimo basically "where the whiskers come out, or below there"
20:14 Zoffix ah
20:15 Zoffix I looked at the first def in dictionary and it was something about tubes and fire boilers :)
20:15 timotimo which also causes the corner of the laptop lid ends up inside the mouth one time out of three
20:24 timotimo MasterDuke_: could you give me that little set of sql queries again? the ones that you made to give the same results as the html profiler?
20:25 MasterDuke_ timotimo: i put some examples in the PR comments https://github.com/perl6/nqp/pull/354. let me know if those don't get you what you need
20:26 timotimo ah, *there* it was
20:43 * timotimo is briefly fascinated by zmodem
21:32 geekosaur joined #moarvm
22:01 timotimo i ran a profile of a thing and it's totally killed my puter
22:03 MasterDuke_ creating the profile did? or accessing it?
22:04 timotimo creating
22:04 timotimo tried to build it with the sql mode. i'm not sure if it even reached the perl6 part of the code, though
22:05 timotimo wonder when my system will be resurrected ...
22:06 timotimo we might really want to turn on gen2-by-default allocation during the call graph dumping
22:07 japhb timotimo: Ragnarok?
22:08 MasterDuke_ should probably batch writing out the sql lines. it's simpler not to, but our IO is slow
22:09 timotimo have you profiled that? :)
22:10 MasterDuke_ heh, yes and no. i've timed the json and the sql output, and json is faster
22:10 timotimo mhm
22:10 MasterDuke_ for the exact same profile (i.e., i hacked it to write both)
22:12 timotimo i'd like to use my desktop again please
22:13 MasterDuke_ i guess i can't be sure the reason to_sql is slower is because of how it does IO, but that's my guess
22:15 timotimo i wonder why this particular profile is so explosive
22:18 MasterDuke_ what's the perl6 code?
22:19 timotimo it's my brownian tree implementation off of rosettacode
22:21 MasterDuke_ ha, taking longer than the 25s mentioned in the description?
22:22 timotimo depends on what size and particle count you're using
22:22 timotimo my run was 55 seconds
22:22 timotimo and since then it's been trying to build the profile
22:24 MasterDuke_ is the output file being written to?
22:24 timotimo how am i supposed to find out?
22:25 MasterDuke_ it's getting bigger/has more lines?
22:25 timotimo well, yeah, but how am i supposed to find out how big it is?
22:26 timotimo i'm telling you i can not use the device
22:26 timotimo at. all.
22:26 timotimo no mouse cursor movement, no ssh connection
22:26 timotimo the clock hasn't updated since 23:56, now it's 00:26
22:26 MasterDuke_ oh!
22:27 MasterDuke_ that...does make it harder...
22:28 timotimo a little bit
22:28 timotimo i might kill the machine
22:28 timotimo if i can, that is
22:29 timotimo i pressed a few combinations and pushed the power button
22:29 timotimo don't think anything's happening
22:30 timotimo i held the power button and it did a hard shut-off
22:31 MasterDuke_ hate when i have to do that
22:32 timotimo yeah
22:32 timotimo not sure how to make it unnecessary
22:34 timotimo hm
22:34 timotimo the numbers in the calls table are reaching 80k
22:34 timotimo when i have only size=10 and particlenum=30
22:35 MasterDuke_ seems a lot
22:35 timotimo i think so, too
22:41 timotimo rec_depth goes up to 280
22:43 timotimo i have not written any sql in far too long
22:54 timotimo MasterDuke_: since the callees table was renamed calls, maybe we should rename callee_id to call_id?
22:55 timotimo or just id, like routines has a column named "id"
22:55 timotimo huh
22:56 timotimo why does allocations have both a caller_id and a call_id?
22:56 timotimo sorry, callee_id
23:02 MasterDuke_ heh. yeah, i wasn't sure at first which was needed, so i just added both (caller and callee to allocations)
23:02 MasterDuke_ and yeah, renaming callee_id would probably make sense
23:04 timotimo DBIish is not happy with me
23:05 timotimo yeah, user error
23:09 timotimo poor dot
23:09 timotimo has to handle 80k nodes for this particular document
23:10 timotimo Error: <stdin>: syntax error in line 83954 near ''
23:14 MasterDuke_ in the sql output?
23:15 timotimo no, in my dot transformation of it
23:16 timotimo it was some bad escaping or something
23:24 geekosaur joined #moarvm
23:27 timotimo well, it's still trying to render
23:27 timotimo not sure if it's layouting or what
23:29 timotimo it shouldn't be terribly hard to do for dot, the graph should be perfectly tree-like
23:41 timotimo i'll suspend my computer so i can sleep (it's noisy), but i'll leave dot running so it'll maybe finish tomorrow %)
23:42 MasterDuke_ i've used dot a bunch before, but always on tiny graphs, no idea how well it handles large things

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary