00:18 nebuchad` joined 00:21 xiaomiao joined 00:30 bonsaikitten joined 00:31 japhb_ joined, nwc10_ joined 00:38 TimToady joined 01:48 ilbot3 joined 02:31 ilbot3 joined 04:10 vendethiel joined 06:07 FROGGS[mobile] joined 06:20 retupmoca joined, FROGGS joined 07:20 zakharyas joined 07:54 Ven joined
arnsholt FROGGS: IIRC structs should have the same alignment as the element with the strictest (that is, largest) alignment requirement 08:48
FROGGS hmmm 08:50
makes sense I guess
that means we need to add and alignment property to MVMCStructREPRData 08:51
(and CUnion and CPPStruct) 09:00
arnsholt Yup
FROGGS testing now... 09:01
dalek arVM: e8c92d5 | FROGGS++ | src/6model/reprs/C (6 files):
handle alignment of inlined structs/unions, arnsholt++
 09:15
FROGGS no regressions
arnsholt Excellent. Pretty sure it's the spec-correct way to handle it 09:18
FROGGS arnsholt: I am continuing to work on libxml2 anyway, so I'll probably notice if something is handled incorrectly 09:22
arnsholt Very cool. Libxml is likely a very useful binding to have, too!
FROGGS yeah, I need it :o)
(for $work)
then I "just" need to port stuff like XML::Compile 09:23
arnsholt Yeah, libxml is more infrastructure than directly usable
FROGGS I still hate xml anyway 09:25
but yuo can't get around it often enough sadly
you*
the good thing is that I usually forget that things are xml underneath... like when building b2b interfaces 09:26
10:20 ShimmerFairy joined 10:43 Ven joined 10:55 colomon joined 12:25 Ven joined
[Coke] moarvm release soon? 12:47
13:31 FROGGS joined
timotimo will be AFK for a long time in a few 13:37
FROGGS a long time? 13:38
:S
timotimo uh 13:40
just multiple hours
FROGGS phew 13:44
:o)
so it is more like C long (seconds) instead of bigint long 13:45
14:10 FROGGS_ joined
TimToady
.oO(a century is multiple hours)
 14:44
14:51 tokuhiro_ joined
timotimo I'll be back before Christmas 15:02
japhb I notice you didn't say what year ... 15:14
16:53 hoelzro joined
FROGGS :D 16:58
timotimo japhb what say you? push the p6-only benchmarks? 17:32
japhb timotimo: Yes, sorry, noticed you'd asked about that, and I hadn't gotten a chance to reply yet 18:11
timotimo: It's perfectly OK to make language-specific benchmarks, especially if the thing you're testing doesn't make much sense in the other languages (e.g. variations of flattening behavior in NQP) 18:12
timotimo: Also, ISTR you were complaining about time taken to run the tests. Remember that there are several kinds of filtering available to decide which tests to run. You can easily choose a selection of test tags to run, instead of all of them -- and you can tune down the run counts and such to get faster total testing time. 18:14
[Coke] wonders if there's anything we can do to speed up running profile.htmls 18:48
18:59 dalek joined
timotimo japhb: i always want to run everything :) 19:00
19:02 psch joined 19:09 lizmat joined, [Coke] joined 19:14 dalek joined 19:19 Util joined 19:23 brrt joined
brrt \o 19:23
what dynasm does with the mrm-byte-marking is weird 19:36
incomprehensibly weird 19:37
FROGGS :/ 19:43
[Coke] profile call graph page taking 2.3G of memory in chrome and 112.1% of cpu. 19:50
timotimo ugh 19:51
and our QJson we use for the qt profiler capitulates at a surprisingly low file size, too
brrt wow 19:52
[Coke] been trying to show me that page for minutes now.
brrt my suspicion is that this bit was always broken with regards to variable registers 19:53
AAAARGH 19:54
FROGGS ó.ò
[Coke] 2.9GB... 19:55
19:55 vendethiel joined
brrt that's actually quite small 19:55
[Coke] brrt: was that to me? 19:56
19:58 tokuhiro_ joined
brrt no, that was to dynasm 20:10
japhb timotimo: Perhaps then the answer is to reorder the tests, so you can get "canary data" quickly. Remember that you can split up your timing into multiple timeall runs (`bench time`), and analyse all the collected benchmark data together as a pool. 20:24
timotimo: So you can tag (or otherwise select) a small subset of the tests to run first, and the inverse to run second. Then as soon as the first timing run is complete, you can analyze it while you spawn off the remainder to collect all the "less important" test info, and finally analyze the whole batch. 20:26
brrt it seems i have some success.... 20:30
timotimo brrt: glad to hear it! 20:31
20:32 colomon joined
brrt dalek won't report on it, so: github.com/MoarVM/dynasm/commit/a3...d5842b9878 20:41
github.com/bdw/lab/blob/master/c/d...ample.dasc and this test case is now handled correctly 20:42
i'm ... reasonably confident this is reasonably less brittle
the original dynasm modrm dance was insanity itself 20:43
FROGGS brrt: what you are doing is scary :o) 20:44
brrt is mad 20:47
i have reported on this before i started
FROGGS aye
timotimo Exercise 4-12. Adapt the ideas of Cthulhu() to write a recursive version of the Forbidden Song of Hali; that is, to unravel the fibres of reality and allow the icy liquid darkness of Carcosa to devour your mind. 20:48
Exercise 4-13. Write a function reverse(s) which reverses the string s by turning the mind inside out, converting madness into reality and opening the door to allow the Old Ones to creep forth once more from their sunken crypt beyond time.
20:48 Peter_R joined
brrt fixedone! 20:51
exercise 4-14: write x86 assembly
:-P
brrt afk
timotimo :) 20:53
psch missed the opportunity to bring in an exercise 4.bleemteen :/
woulda fit with all the madness :P 20:54
timotimo i don't know what that is a reference to :| 20:55
psch timotimo: www.strangehorizons.com/2000/200011...mber.shtml 20:56
pretty good story
not amazing, but good
timotimo cute 21:05
FROGGS psch: nice story :o) 21:25
21:38 colomon joined 21:59 tokuhiro_ joined

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!