Perl 6 - the future is here, just unevenly distributed

IRC log for #perl11, 2016-07-07

| Channels | #perl11 index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:10 mako on strawberry perl for win10 (64bit) perl v5.24.0
03:39 sten1 bulk88 hi
03:39 sten1 have you got some improvements with this bug https://rt.perl.org/Public/Bug/Display.html?id=121274 ?
03:40 sten1 currently I'm getting lots of arenas with lots of available chunks(700000+) for long running perl process with constatly increasing mem usage(3gb+)
03:41 sten1 do you have some solution?
04:05 bulk88 sten1 I think you have a leak
04:05 bulk88 that bug was allocate an insanely large perl array, set every element to 1, free the array, memory doesn't decrease
04:10 sten1 I don't have leak(checked with Internals::DumpArenas), I have daemon that sometime can use lot of memory to process data chunks. aftering dumping all arenas I saw that available chunks is constatly increasing
04:11 sten1 after that I found similar bug filled by you
05:47 sten1 https://github.com/perl11/cperl/issues/178
07:30 travis-ci RPerl build failed. Will Braswell says 'Learning RPerl, Print Operator, Change Dots To Commas'
07:30 travis-ci https://travis-ci.org/wbraswell/rperl/builds/142964809 https://github.com/wbraswell/rperl/compare/ff289f224abc...f4875f83c401
07:35 punter joined #perl11
08:37 willthechill whoopsie
09:35 punter joined #perl11
10:48 travis-ci perl11/cperl#1261 (smoke/gh173-sig2pp - 0bf75d4 : Reini Urban): The build is still failing. https://travis-ci.org/perl11/cperl/builds/143002445
11:49 ashevchuk joined #perl11
12:01 mako joined #perl11
12:04 mako left #perl11
15:47 willthechill joined #perl11
16:00 ashevchuk joined #perl11
16:51 mako joined #perl11
17:55 mako left #perl11
17:55 mako joined #perl11
18:33 punter joined #perl11
19:49 mako joined #perl11
19:54 willthechill joined #perl11
20:10 willthechill punter: howdy!
20:10 willthechill I saw your question yesterday
20:10 willthechill "Is rperl's plan to develop rperl independently to something incompatible of perl5?"
20:10 willthechill "i mean, i wonder how many features rperl is willing to add to its language"
20:11 willthechill RPerl is working TOWARD full Perl 5 compatibility, not AWAY from it!
20:12 willthechill high-magic Perl 5 has so many insane features that I do not intend to add any whatsoever... EXCEPT for those features which relate directly to runtime performance, and in those cases only "adding features" which follow the Perl & C++ family of languages
20:12 willthechill such as adding SSE and GMP support
20:12 willthechill etc.
20:12 willthechill those features can be used in non-compiled (interpreted) RPerl code just like normal Perl
20:13 willthechill bulk88 & mako: what was our decision about "long long" vs "longlong"?
20:13 willthechill and the other Windows stuff?
20:18 travis-ci perl11/cperl#1262 (smoke/gh173-sig2pp - bc12e4d : Reini Urban): The build was fixed. https://travis-ci.org/perl11/cperl/builds/143142876
20:47 bulk88 willthechill read the backscroll
20:47 bulk88 this room is logged
20:56 mako About "long long" to "longlong" I made the decision to modify 'rperltypes::type_integer_errorcheck' to make rperl's complaints silent. But this is not a solution because '$Config{ivtype' is in other functions I didn't modified. The weird thing is I could make rperl working with changing only the error check.
20:58 punter willthechill, hi!
21:00 mako All other functions using '$Config{ivtype}' remain unchanged. And you know what, it did compile! So there has to be a value of 'long long' anywhere but the stuff is working.
21:00 punter Would parsing a huge Apache log file with regular expressions be much faster when done with RPerl?
21:01 punter I believe not, but a friend (sysadmin) asked, so I'm asking you.
21:16 willthechill bulk88: already did read the backscroll
21:17 willthechill and yes I know the room is logged.  :P
21:18 willthechill bulk88: I only know of "long long", what is "longlong"?
21:18 mako punter: Do you have any measurements that could give a hint there's a difference between a program run by perl and run by rperl?
21:18 willthechill mako: thank you for your very good explanation, I will discuss with bulk88 and hopefully find a permanent solution which will solve your issue as well as other people who may have a similar system
21:18 willthechill mako: punter is new like you
21:19 willthechill there are some performance benchmarks already online  http://rperl.org/performance_benchmarks.html
21:19 willthechill there are more performance benchmarks coming soon
21:19 mako willthechill: ok
21:20 willthechill currently missing: pi generator, mandelbrot fractal, julia fractal, and gauss-seidel
21:20 willthechill currently listed: bubble sort, n-body
21:20 willthechill so I have 2 benchmarks published, 4 which run but I have not yet published
21:20 willthechill sorry but it is a lot of work for 1 guy!  that is why I can use all the help I can get.  ;-)
21:21 willthechill pi generator is a few times faster, mandelbrot and julia are insanely faster
21:21 willthechill gauss-seidel is brand new to me so I don't know all the numbers just yet
21:21 willthechill BUT gauss-seidel does run in parallel!
21:21 willthechill and I know I nearly doubled in speed by going from just 1 to 2 cores
21:21 punter mako, I don't have any measurements yet, I haven't read much of the rperl manual
21:21 willthechill so yeah, parallel speed works too!  :-D
21:22 willthechill punter: LMPC #45. Thou Shalt Not Use Regular Expressions   http://rperl.org/the_low_magic_perl_commandments.html
21:23 willthechill more specifically, RPerl v1.0 and v2.0 both use the same low-magic grammar, which has a spot reserved for regular expressions but does not currently attempt to support regexes in any way
21:23 punter o
21:24 willthechill BECAUSE the regular expression engine is a separate runtime engine, it is not actually "Perl"
21:24 willthechill and the Perl 5 regex engine is already written in highly hand-optimized C code
21:24 willthechill so it can't go any faster
21:24 punter you plan to add regex support though, right?
21:24 punter even if it won't get faster?
21:24 willthechill yes yes, we will support it, that is why I said we already have a spot for it in the grammar
21:25 willthechill RPerl v3.0 will support medium-magic grammar
21:25 punter I need to match regexes like ...A...K.. with the UNIX dictionary - I guess I can do it with foreach and ifs
21:26 willthechill https://github.com/wbraswell/rperl/blob/master/lib/RPerl/Grammar.eyp#L87-L88
21:26 willthechill those are the 2 lines in the existing RPerl low-magic grammar which are for regular expressions
21:26 willthechill and anywhere else that matches "OP06_REGEX"
21:28 punter goodnight alll
21:28 willthechill punter: yes if you have loops and conditionals and string manipulation (not regexes) then it will be much faster in RPerl
21:28 willthechill okay goodnight my friend, talk to you soon!  :-)
21:28 mako willthechill: Some strange thinking of myself: What if there would be an R&D branch of RPerl (I would call it RPerl93r named after 93r (my user name at github)) where experiments could be done without interferring with the normal development of RPerl?
21:29 mako Hypothetically speaking. This is just in my mind for now.
21:29 willthechill I would say anyone coming up with such an idea would want to have about 100 commits accepted by me into the RPerl "master" branch, or else such a person would not actually have any idea what they are doing
21:29 willthechill RPerl system internals are very complicated
21:30 willthechill there are already 20 or 30 people who have created their own branches of RPerl
21:30 willthechill you are of course free to do so
21:30 willthechill just don't expect to have any success without spending a few months working with me ahead of time
21:30 willthechill unless you are some kind of super-genius!
21:30 willthechill :-D
21:31 willthechill then maybe you should have invented RPerl before me.  ;-)
21:31 willthechill oh and BTW, none of the people who have created their own RPerl forks have actually written any new or "experimental" code.
21:31 willthechill it is simply too high of a learning curve
21:31 willthechill like trying to join P5P
21:32 willthechill well, hopefully not THAT bad, haha!  *ducks*
21:38 mako Ok, now I know I asked the wrong question. How about this: When I want to add something to RPerl like C++'s type inference, how could this be done and could RPerl code look like?
21:42 willthechill you would need to first show me some code examples of what C++ behavior you want
21:42 willthechill in theory we can add any C++ features into RPerl
21:50 mako Something like 'my auto $v = "hello"' would resolve to either 'string v = (const string) "hello"' or 'auto v = (const string) "hello"'. Maybe my thinking is wrong or weird. But shouldn't be this possible.
21:50 willthechill yes that should be possible
21:50 willthechill we already have scalartype
21:50 willthechill and also unknown
21:51 willthechill my scalartype $foo = SOMETHING;
21:51 willthechill my unknown $bar = SOMETHING_ELSE;
21:51 willthechill both scalartype and unknown types are currently placeholders only
21:51 willthechill they do not compile to C++ yet
21:51 willthechill they only run in PERLOPS_PERLTYPES AKA test mode
21:53 willthechill we will support gradual typing in a future RPerl version
21:53 willthechill but personally I won't use it because it will likely not be as fast as static types
21:55 mako I know it's difficult to translate between perl and c++. And I know even bitwise ops can't be translated 1:1. I read Learning RPerl. Many functions used in Perl are in status 'coming soon' when it comes to RPerl.
21:55 mako That won't be easy.
21:58 mako For example 'substr', 'length', 'log', and many more. The question is: How to translate these without loosing the ability to run with standard perl and making it work with intermediate c++
22:02 willthechill substr, length, and log are easy
22:02 willthechill sorry I don't think I understand your question?
22:06 mako Builtin perl functions should run either when Perl or RPerl is in use.
22:08 mako I used 'length' with RPerl. In testmode without compiling it worked very well. But when it came to subcompiling it failed.
22:09 willthechill yes we have not implemented most of the Perl operators
22:10 willthechill I have 1 new RPerl system developer, he has started work on operators, first 'abs'
22:24 mako Good to know
22:26 mako Maybe I should concentrate on making RPerl running on Windows Systems (and on OpenBSD, my ultimate goal even though I deferred this for now)
22:28 mako So Ladys and Gents have a good night. I need to sleep. Best wishes.
22:29 mako left #perl11

| Channels | #perl11 index | Today | | Search | Google Search | Plain-Text | summary