Perl 6 - the future is here, just unevenly distributed

IRC log for #moarvm, 2016-11-14

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

All times shown according to UTC.

Time Nick Message
06:19 domidumont joined #moarvm
06:22 domidumont joined #moarvm
06:30 JimmyZ hmm, seems that call mkdir and then call readdir will get a error
06:30 JimmyZ rm -rf test && ./perl6-m -e 'use nqp; nqp::mkdir("test", 0o777) unless "test".IO.e; my $io := nqp::opendir("test"); say nqp::nextfiledir($io);'
06:30 JimmyZ Failed to read dirhandle: 2
06:32 JimmyZ this only happens when make an un-existed dir and then read it
06:45 dalek MoarVM: a94e744 | (Jimmy Zhuo)++ | src/io/dirops.c:
06:45 dalek MoarVM: fixed readdir, so it won't check old errno
06:45 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/a94e744b44
06:52 dalek MoarVM/even-moar-jit: a70f9fb | brrt++ | src/jit/linear_scan.c:
06:52 dalek MoarVM/even-moar-jit: Implement register assignment
06:52 dalek MoarVM/even-moar-jit:
06:52 dalek MoarVM/even-moar-jit: This almost enables the use of the linear scan algorithm.
06:52 dalek MoarVM/even-moar-jit: review: https://github.com/MoarVM/MoarVM/commit/a70f9fb496
07:10 brrt joined #moarvm
07:16 brrt \o
07:17 JimmyZ o/
07:17 brrt reproting on progress: going well, but I've still got problems with (referencing) pseudotiles
07:17 brrt because everything is accessed by index
07:23 domidumont joined #moarvm
07:36 domidumont joined #moarvm
07:40 domidumont joined #moarvm
07:53 brrt i'll get into more detail on that later, have to run now
08:14 domidumont joined #moarvm
08:22 zakharyas joined #moarvm
09:34 domidumont joined #moarvm
09:41 domidumont joined #moarvm
10:04 domidumont joined #moarvm
11:04 domidumont joined #moarvm
12:02 dalek MoarVM: 20c8591 | FROGGS++ | src/platform/posix/time.c:
12:02 dalek MoarVM: work around clock_gettime issue on OSX (#438)
12:02 dalek MoarVM:
12:02 dalek MoarVM: This resolves issue #437. Tested on pre- and post OSX 10.12 / Xcode 8.
12:02 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/20c8591ad7
13:26 domidumont joined #moarvm
13:35 domidumont joined #moarvm
14:04 domidumont joined #moarvm
14:14 domidumont joined #moarvm
14:15 stmuk FROGGS++
14:16 stmuk apple--
14:41 _longines joined #moarvm
16:37 FROGGS joined #moarvm
16:37 FROGGS o/
17:04 japhb joined #moarvm
17:56 lizmat joined #moarvm
17:58 FROGGS btw, I've got now access to an AIX@powerpc machine now
17:58 FROGGS though, libatomic-ops and dyncall dont support it
17:58 FROGGS and libffi isnt installed
18:01 timotimo oof
18:05 FROGGS what do we actually need libatomic-ops for?
18:06 timotimo MVM_ADD, MVM_CAS
18:06 FROGGS which means?
18:06 timotimo well, CAS is Compare-And-Swap
18:07 jnthn Atomic operations and memory barriers, in short.
18:07 jnthn Providing access to CPU primitives that we use to implement various concurrent data structures.
18:16 FROGGS hmmm, thanks :o)
18:18 pyrimidine joined #moarvm
18:25 lizmat jnthn: any thoughts on the parse-base addition as a sub ?
18:41 pyrimidine joined #moarvm
18:45 zakharyas joined #moarvm
18:48 pyrimidine joined #moarvm
18:48 domidumont joined #moarvm
18:52 * timotimo first read that as "parse-based addition" and was quite confused
18:56 FROGGS was able to build libffi on aix... with some hacks :S
18:58 lizmat timotimo: well, I thought we weren't supposed to just add built-in subs in 6.c ?
18:58 lizmat or exposed classes?
18:59 lizmat we were going to use a "6.d" setting for this, weren't we ?
18:59 vendethiel joined #moarvm
18:59 lizmat I mean, personally I would be fine with adding stuff and call it 6.d at some point in the (near) future
19:00 timotimo right, i suppose it'd have to be unavailable if you "use v6.c"
19:00 pyrimidine joined #moarvm
19:00 [Coke] wrong channel?
19:00 lizmat [Coke]: perhaps
19:09 FROGGS not bad, not bad:
19:09 FROGGS compiling src/core/loadbytecode.o
19:09 FROGGS compiling src/math/num.o
19:09 FROGGS src/math/num.c:9:1: error: initializer element is not constant
19:09 pyrimidine joined #moarvm
19:10 jnthn lizmat: Will try and catch up with perl6-dev once I'm done cooking/eating dinner :)
19:11 jnthn But yeah, we could do with the 6.d setting added
19:11 jnthn nine++ did have a branch that did it, iirc
19:12 jnthn And while it's arguable we just don't have the userbase size to worry a lot about it at the moment, we should get into the habbit and figure out the quirks/issues before that point.
19:15 pyrimidine joined #moarvm
19:17 timotimo jnthn: would you spend a minute looking at the multidimarray_view branch - it'd probably be enough to look at flat_index and the struct - and tell me if it's a sane direction to go in?
19:31 FROGGS uhhh, moarvm compiles further...
19:32 nine The branch is called language_versions
19:43 FROGGS compiling 3rdparty/libuv/src/unix/linux-core.o
19:43 FROGGS 3rdparty/libuv/src/unix/linux-core.c:33:23: fatal error: sys/prctl.h: No such file or directory
19:43 FROGGS compilation terminated.
19:43 * FROGGS cries
19:43 domidumont joined #moarvm
19:55 * FROGGS uncries
20:23 TimToady m: say v6.d after v6.d.a
20:23 camelia rakudo-moar c196af: OUTPUT«True␤»
20:27 lizmat TimToady: so that's wrong, right ?
20:28 TimToady no, that's correct
20:28 TimToady alphas come before .0, .1, .2...
20:29 lizmat because 6.d would give you the latest 6.d ?
20:30 TimToady 6.d is supposed to be short for 6.d.0+ I thought
20:31 TimToady if we changed that, it has fled my brane
20:32 TimToady which happens frequently enough :)
20:34 ggoebel joined #moarvm
20:43 TimToady I guess smartmatching is a little more forgiving
21:12 jnthn Certainly as I recollect it, the idea was we'd have 6.d.ALPHA or 6.d.a or such to be our "get a preview of what 6.d will be"
21:13 jnthn So when we introduce the 6.d setting, for now you'll use 6.d.a to get at it.
21:13 * jnthn makes an attempt to backlog #perl6-dev :)
21:16 pyrimidine joined #moarvm
21:25 pyrimidine joined #moarvm
21:37 pyrimidine joined #moarvm
21:43 pyrimidine joined #moarvm
21:51 japhb timotimo: re: API for ndarray-view: Perl 5 had a particularly powerful one called PDL (pdl.perl.org).  It does pretty cool things like allowing you to do a multidimensional slice, swap coordinates, etc. extremely cheaply.
21:54 timotimo japhb: yeah, i'm hoping what i'm implementing can let you do that, too
21:54 timotimo though how am i supposed to know if it's cheap until i implement it fully and we look at jitting all the ndarray stuff0
21:55 timotimo if i'm not mistaken, my implementation will allow to go from any amount of dimensions to any other (as long as the number of values doesn't exceed the target array)
21:55 jnthn timotimo: will take a look later this week...really tired today :(
21:55 timotimo it'll allow any dimension to be mapped to any target dimension, and to map it forwards, backwards, or with full-integer steps in either dierction
21:56 timotimo and also it'll allow you to access an array of 64bit ints as many more 8bit ints, for example
21:56 timotimo your move, PDL
23:46 yfkj joined #moarvm
23:47 yfkj timotimo: are you implementing it column-major order ?
23:47 yfkj it's the order of Fortran, R, OpenGL and MATLAB
23:49 timotimo ndarrays have already been implemented with C order, i believe
23:49 timotimo but using views you can store your array in column-major order and have a C-style view onto the data
23:51 yfkj mainly thinking that having OpenGL, ATLAS, LAPACK binding would be cool ;)

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