Parrot 6.4.0 "Double-eyed Fig Parrot" | parrot.org/ | Log: irclog.perlgeek.de/parrot | #parrotsketch meeting Tuesday 19:30 UTC
Set by moderator on 25 May 2014.
00:05 basiliscos joined 04:28 rurban joined 07:24 woosley left, FROGGS joined 07:35 basiliscos joined 09:28 FROGGS joined 09:47 Psyche^ joined 10:07 cosimo joined 10:19 Chirag joined
Chirag My new branch with a few more WB optimizations : github.com/ZYROz/parrot/tree/ZYROz...ig3-gh1069 10:20
11:51 rurban joined
rurban Chirag: Could you seperate in the new branch optimizations from bugfixes into 2 seperate commits? class, mappedbytearray have bugfixes 12:31
orderedhashiterator looks wrong
dalek rrot/rurban/pmc2_orig3-gh1069: e5e24bd | ZYROz++ | src/pmc/ (20 files):
[pmc] WB optimizations
 12:50
rrot/rurban/pmc2_orig3-gh1069: 3037bf3 | rurban++ | src/pmc/ (3 files):
[pmc] Fixup some wrong WB optimizations
13:07 Chirag joined 13:14 rurban joined
rurban_ Chirag: Could you seperate in the new branch optimizations from bugfixes into 2 seperate commits? class, mappedbytearray have bugfixes 13:15
Chirag sure! 13:16
can you please see my replies to the comments 13:17
rurban_ done 13:28
dalek rrot/rurban/pmc2_orig3-gh1069: 5c43af2 | rurban++ | src/pmc/structview.pmc:
[pmc] fixup structview.set_pmc_keyed. revert to original
 13:29
Chirag two commits done.. 13:40
rurban_ mappedbytearray.destroy also looks like a bugfix 13:42
nope, your rebase looks a bit odd. 13:43
I'll do it. I'm faster
There are also new trailing whitespace errors 13:44
done 13:49
waiting for the smoke test 13:51
Chirag you merged all the the 20 files? 13:52
rurban_ I fixed up your 3 commits into 2
and fixed the whitespace
looks good
yes 13:53
Chirag oh .. thanks.. :)
rurban_ you really need to run perl t/codingstd/trailing_space.t before 13:56
or prove t/codingstd/trailing_space.t 13:57
Chirag yeah.. I missed that
rurban_ which editor are you using?
Chirag i keep switching between emacs and gedit .. dont really find emacs comfortable 13:58
rurban_ You need some time to get a proper emacs setup 13:59
on the web are several nice .emacs files. we have parrot mode files 14:00
Chirag oh.. I will look into them.. even eclipse seemed tedious with parrot..
rurban_ (add-to-list 'auto-mode-alist '("\\\\.nqp$" . cperl-mode)) 14:02
;(setq cperl-hairy t) ;(setq cperl-hairy nil)
(autoload 'pir-mode "pir-mode" nil t)
(add-to-list 'auto-mode-alist '("\\\\.pir$" . pir-mode))
(autoload 'pasm-mode "pasm-mode" nil t)
(add-to-list 'auto-mode-alist '("\\\\.pasm$" . pasm-mode))
(load-library "parrot")
(setq-default indent-tabs-mode nil)
Chirag thanks! 14:03
umm.. will be back in a while.. 14:04
rurban_ I'm doing some benchmarks now 14:05
dalek rrot/rurban/pmc2_orig3-gh1069: a6ee7f4 | ZYROz++ | src/pmc/ (17 files):
[pmc] More WB optimizations

Signed-off-by: Reini Urban <rurban@cpanel.net>		 14:08
rrot/rurban/pmc2_orig3-gh1069: 1c0872a | ZYROz++ | src/pmc/ (2 files):
[pmc] Fixed WB Bugs

Signed-off-by: Reini Urban <rurban@cpanel.net>
rurban_ 1.29% slower, but my machine is too loaded now 14:21
nqp is also slower 14:33
14:50 bighugedog joined
rurban_ destroy should be all :no_wb 15:22
yes, with :no_wb we are good again 15:32
dalek rrot/rurban/pmc2_orig3-gh1069: 58b5777 | rurban++ | src/pmc/ (26 files):
[pmc] destroy is not :write. mark as :no_wb

This fixes the parrot and nqp benchmarks numbers		 15:33
rrot/rurban/pmc2_orig3-gh1069: 00cd232 | rurban++ | / (4 files):
[pmc] Optimize dynpmc WB and testlib/nci_pmc.in

mark non-writers as :no_wb		 15:55
rrot/rurban/pmc2_orig3-gh1069: 197be0d | rurban++ | / (4 files):
t/codingstd/copyright.t: allow year-year,year

copyrights are only defendable on actual changes. overly broad ranges are not accurate. allow adding single years, or multiple ranges. See eg. select.pmc: 2011,2014, or file.pmc: 2001-2011,2014,
Fixes GH #1074		 16:10
16:19 FROGGS joined 16:23 rurban joined 16:43 bluescreen joined 17:09 basiliscos joined 18:36 Chirag joined
Chirag So, do we have any additional improvement in speed?? 18:44
rurban_ My machine is too busy right now. Looks like same speed 18:53
I added :no_wb to all destroy and mark so it should be a bit faster 18:54
Chirag yes.. i saw..
rurban_ I just finished fixing the perl6 spectests 18:55
6.4.0, master, rurban/pmc2_orig3-gh1069: 10.430843965, 10.595508065, 10.272068754 19:02
I guess it was the destroy which hit us on master
do you have more? I'm ready to merge now 19:03
we also found some unicode issues with perl6
Chirag no.. I didnt really touch METHODS 19:04
and those commits included all the 95 pmcs 19:05
we touched 10.052022194 .. its not that fast 19:06
rurban_ 10.05 was with gcc-4.9. I fell back to 4.8 19:07
Chirag oh
rurban_ gcc-4.9 is MUCH better
and -O3 should also be safe for parrot. But not for nqp
Chirag what is -03 ? 19:09
rurban_ gcc -O3 19:10
parrot only goes up to -O2 and -O2 is actually slower with perl6 currently
gcc goes up to -O6 I believe 19:11
Chirag these are like versions?
rurban_ no, optimization levels. each including several dozen or more specific compiler optimizations 19:20
Chirag ohk.. 19:21
rurban_ parrot defaults to a cc -O2 on --optimize, but should be safe up to -O3, as our cage tests are pretty good. (no undefined behavior)
when you compile nqp you'll see a lot of cc warnings, signed vs unsigned, which could be problematic with -O3. But still better than perl5 19:22
Chirag hmm.. 19:26
Coke "unicode issues with perl6" == issues in parrot's handling of icu that perl6 exposed? or something ele? 19:33
*else? 19:34
nevermind, caught up in #perl6
rurban_ parrot (i.e. icu) I believe 19:35
19:52 bluescreen joined 19:53 bluescreen joined 19:55 bluescreen joined 19:56 bluescreen joined 20:41 basiliscos joined 22:28 Khisanth joined 22:45 Khisanth joined
dalek rrot/icu-namealias-gh1075: 7492590 | rurban++ | VERSION:
bump VERSION to 6.5.0
 23:40
rrot/icu-namealias-gh1075: 1830239 | rurban++ | / (4 files):
[ops] find_codepoint: search more u_charFromName choices

icu stores character names in 4 tables: U_UNICODE_CHAR_NAME, U_EXTENDED_CHAR_NAME, U_CHAR_NAME_ALIAS, U_UNICODE_10_CHAR_NAME but does not allow to search for all name aliases (at least until 52) for alias-only names, like for empty-named control characters. See [GH #1075]
All perl6 control character names, like
  "LINE FEED (LF)" "CARRIAGE RETURN (CR)" "NULL" "NEXT LINE (NEL)" "CHARACTER TABULATION"
will fail.

Powered by Cro and the Raku® Programming Language.

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

Thank you!