pugscode.org/ | nopaste: sial.org/pbot/perl6 | pugs: [~] <m oo se> (or rakudo:, kp6:, elf: etc.) (or perl6: for all) | irclog: irc.pugscode.org/
Set by Tene on 29 July 2008.
cognominal_ sln : this not pir but sonething that would be nice would be able to have different actions files for a given grammar.pg. Because one may want to compile source code or just beautify it. That would imply modify the code of dev/mk_processor_shell.pl or and/or of the called libraries 00:02
ruoso @tell pmurias, I think at some point we could publish it to CPAN, but I'm not sure on how to do it... 00:35
lambdabot Consider it noted.
ruoso is the irc log down, or is it just me? 00:36
lambdabot ruoso: You have 1 new message. '/msg lambdabot @messages' to read it.
ruoso moritz_, are you around? 01:05
pugs_svn r21903 | ruoso++ | [smop] first implementation of the default block signature bind code. It does not work yet, but is a great progress anyway. 01:25
pugs_svn r21904 | lwall++ | [STD] speed up parser by 33% by using P5 regex instead of TRE for ltm backoff matches 04:32
cognominal_ ltm? 04:34
audreyt for downloading a pugs tarball, git.pugscode.org/ now provides a "download" link
lambdabot Title: audreyt's pugs at master — GitHub
Auzon cognominal_: ltm is longest token matching
cognominal_ there is a few TLAs that I have trouble to get in my head :( 04:35
Auzon yeah, that one is pretty specific. I only recognized it since I heard a lot of pain being spent in getting LTM to work correctly :) 04:36
cognominal_ ...and github is a nice mac interface to see changes 04:39
I meant gitnub!! 04:40
pugs_svn r21905 | lwall++ | [Cursor] buglet from incomplete transition on semantics 04:48
dduncan so is Pugs on Git now, or is that GitHub thing just a mirror? 06:41
I mean, which is the official repository now? 06:42
(I haven't resynced in months, so sort of out of the loop ... svn/svk has been more trouble on my system lately) 06:43
my current work is just my own projects atm, which use Git
audreyt it's just a mirror 07:14
svn is still official
but you can track/clone with git now
dduncan can I push commits using git too and have those forward to the official repo? 07:42
anyway, if people can agree, I suggest putting the official version on git 07:44
dduncan in some respects I think doing so would make it easier for people to contribute, since git is much simpler to install and its repositories seem more reliable and compact 07:45
I'm not specifically recommending that host, just the tool chain 07:47
fullermd People? Agree on a version control system? Sure, that should be no problem... :p 07:49
dduncan I don't mean everyone, just a critical mass 07:52
or alternately just one or a few of the right individuals 07:53
Alias_ git is not easier 07:54
git is easier to be disconnected with, and to do work without permission of others
Working WITH full trust on a common codebase is harder
dduncan just installing the program I found to be easier 07:55
Alias_ apt-get install ...
Anyways, gotta pack
Leaving hotel in 5 minutes
fullermd There's a really bad joke in there somewhere about "if you used git" and "packing"... 07:56
dduncan its written in C and is self-contained
while using the svn stack needed a number of other non-trivial prerequisites 07:57
with git I could just download the tarball, and do configure, make, make test, make install
as for the working with trust thing, that's a social issue not a tool issue 08:01
Alias_ sure it's a tool issue 08:05
Now I need to make contortions in order to push to the main repository
I can't just go "commit"
Alias_ ok, heading to hackathon 08:10
pmurias ruoso: hi 09:19
ruoso: re CPAN I have never released anything on it yet (but i'm intrested in trying), the biggest problem i think is that there are a lot of stuff we used that are not CPANed yet like STD and elf 09:22
ruoso pmurias, indeed.. the dependencies are the most problematic issue 09:38
pmurias ruoso: from the reaction of the audience at my talk i fear that getting perl5 stackless won't be easy 09:44
ruoso: are you blocking on anything now? 09:50
ruoso pmurias, how was the talk?
ruoso pmurias, I'm trying to implement the default block signature 09:53
but I'm getting "reading empty register 21"
ruoso pmurias, re p5 stackless: nothingmuch was pretty much confident that it is possible... I had talked with p5p people about it, and they were very much hopeless too 10:00
but... in nothingmuch we trust ;)
nothingmuch was I? ;-) 10:02
pmurias (network cabel got loose) 10:02
nothingmuch re stackless p5: 10:03
it'd need fairly major refactoring of pp*
i don't htink it's easy
or even possible in the short term
what i do think is that it's possible to make it compatible
basically #ifdefs to make vTHX delimitable 10:04
pmurias ruoso: i run out of slides during the talk and had to demonstrate how smop work now, but it wasn't as bad as i feared, and i can't really tell how other people liked it
nothingmuch and a runops that fiddles PL_stack_base on every op 10:04
setting it up correctly
i think that's enough but it's going to be veeery slow and clunky
pmurias we can always embed kurilla ;) 10:05
* kurila
nothingmuch to make this change actually useful it will probably require creating static variants of all the perl opcodes
that take register args
and use an arity table
basically converting things from push/enter to eval/apply 10:06
none of this is possible in the short term though, IMHO
ruoso but even if p5 can't get stackless... as long as we have a unique smop instance for each runops recursion... I think we're fine... 10:07
nothingmuch *nod*
anyway, back to real $work ;-) 10:08
oh
the llvm plan
in case anyone is interested:
building a statically compiled flattenning of an optree is only about 10% faster with llvmgcc
(baiscally hijacking a CV and upgrading it to an XSUB)
surprisingly with plain gcc it's actually 40% *slower* than runops 10:09
so bascially llvm is much smarter than gcc, but still not smart enough to make simple stack operation static
pmurias nothingmuch: the signature stuff you did with clkao was awesome 10:12
ruoso nothingmuch, llvm might be usefull in smop as soon as we have a bootstrapped interpreter...
clkao appears in a puff or orange smoke 10:13
nothingmuch: i am cleaning up my b changes and putting them into b::utils for the moment
ruoso I had this insight of building a type for each signature with the code that runs the BIND and the ACCEPTS code for that spefic signature...
nothingmuch clkao++
i'm working on making MXSD prophet ready =)
ruoso: "type" as in C struct? 10:14
ruoso nothingmuch, type as SMOP__Object type
nothingmuch ah
i thought you were talking about Data::Bind for a sec
ruoso this type then holds jit compiled code for the bind and accepts for that specific signature 10:15
pmurias nothingmuch: re not smart enough, one can write custom optimisers for llvm you know, and i can't see why should't we use Perl 6 for them
nothingmuch pmurias: the amount of effort I can invest in p5/llvm mashups is very limited 10:16
so writing an LLVM optimizer pass is beyond the scope currently
nothingmuch currently I think it's much easier to approach this from the perl side 10:17
since the C code for the perl internals is so macro heavy
and relies on so mch implicit state 10:18
nothingmuch that even if I were writing a p5 specific LLVM pass it would probably be too hard 10:18
pmurias i should get round to reading that assembly tutorial, and trying to just save the C stack soon 10:19
pugs_svn r21906 | pmurias++ | [mold] an error is raised if a NULL is returned from a method call 10:31
pmurias ruoso: i think the error is the "something returns a NULL" scenario again, i should now be doing stuff i put off before the yapc, if something is definitly an mold error i'll do it now otherwise i'll get back to it later if it's not fixed by then 10:38
pugs_svn r21907 | pmurias++ | [mold] the content of registers is printed after a NULL is returned from a method call 12:17
pugs_svn r21908 | ruoso++ | [smop] prints current mold frame position when aborting. 14:09
pugs_svn r21909 | ruoso++ | [smop] some more diagnostic information on mold errors... few NULL-returns caught... 14:23
r21910 | ruoso++ | [smop] test/27 have a part of the test passing! 14:25
pugs_svn r21911 | ruoso++ | [smop] native int implements infix:> 14:40
pugs_svn r21912 | ruoso++ | [smop] test 27 almost passing... only thing to fix is an extra scalar that is getting on the way for some reason... 15:03
ruoso pmurias, if you have some time, could you please try to help me with test 27? 15:04
pugs_svn r21913 | ruoso++ | [smop] YAY! Default Block Signature Works!!!!!!! 15:07
ruoso pmurias, now we have another good example of a mold code 15:08
pmurias, and a very nice example on how slime and mold can coexist... take a look at test/27 and at src/smop_s1p_defaultblocksignature.sm0p
pmurias ruoso: i had network problems, do you still need help? 15:32
ruoso fortunally, no... ;)
but take a look at test/27 and the defaultblocksignature implementation.... 15:33
it's an even better use case for mold than the other...
and now we're ready to port Code to Mold 15:36
pugs_svn r21914 | pmurias++ | [smop] changed STORE to infix:<=> in test 27 15:40
pmurias i'll change use m0ld to q:m0ld " 15:41
s/"/{...}/
pmurias ruoso: you need m0ld {...} and sub {...} support? 15:48
ruoso pmurias, q:m0ld has the advantage of allowing custom mold variable name 16:07
pmurias, I think sub-molds are probably going to be used by the s1p compiler 16:09
so I'm not sure we want to use it directly in m0ld
pmurias, one cool project now is the signature compiler 16:10
for s1p
we need to be able to compile a signature to a equivalent BIND and ACCEPTS code... 16:11
pmurias ruoso: i think we could implement the Signature in perl6 16:18
ruoso pmurias, the Signature high level type will be implemented in Perl 6 and compiled by s1p 16:27
but before that we need signatures to make other things work 16:28
and that we can compile to mold code for the BIND and ACCEPTS call
pmurias the default signature (*@_,*%_) is enough to make most things work 16:53
ruoso pmurias, in fact... we could have a smop_s1p_defaultroutinesignature type 16:54
just like defaultblocksignature 16:55
pmurias, want to take that to implement?
www.perlfoundation.org/perl6/index....ementation 16:57
pmurias i'm switching to q:* now 16:58
ruoso pmurias, :)
pmurias, but do you think the modelling in the wiki for the code object is ok? 17:01
ruoso shower & 17:02
pmurias the default signature for subs is different (*@_,*%_) 17:03
pugs_svn r21915 | pmurias++ | [smop] switched use * blocks to q:* 17:06
ruoso pmurias, yes... that is for routines in general... 17:08
pmurias and back should be set 17:12
but otherwise i think it's correct
i have some $workish things to finish so i'll try to do only blocking smop stuff for a couple of days 17:17
pmurias @tell mncharity what do you think of releasing elf to CPAN? 20:33
lambdabot Consider it noted.
pugs_svn r21916 | pmurias++ | [smop] converted remaining STORE to assignment 20:48
rindolf Hi pmurias 20:49
pmurias rindolf: hi 20:54
rindolf pmurias: what's up?
pmurias: BTW, where are you from?
pmurias rindolf: Poland
rindolf pmurias: ah. 20:55
pmurias: my grandfather was Polish in origin.
pmurias rindolf: returned from YAPC::EU yesterday
rindolf pmurias: ah, any highlights?
rindolf pmurias: and my grandmother was born in Vilnius. 20:55
pmurias rindolf: i gave my first talk 20:58
rindolf pmurias: how did it go? 21:00
pmurias dunno, i don't read people very well
rindolf pmurias: what was the talk about? 21:01
pmurias smop 21:02
rindolf: did you manage to compile it yet?
rindolf pmurias: I couldn't because it requires ruby-1.9.x
pmurias i'll try downporting it to ruby-1.8 21:03
rindolf pmurias: thanks.
pmurias rindolf: do you have ruby-1.9 in your distro? 21:05
rindolf pmurias: no. 21:06
pmurias the README seems to imply it's something more than a simple bug which caused the requirement, but i'll still try feebly 21:08
rindolf: no luck with that, there were some minor syntax issues i can change, but i don't know how to replace the String#force_encoding method 21:17
rindolf pmurias: I see. 21:19
pugs_svn r21917 | Auzon++ | [gsoc_spectest] added some feed tasks to TASKS. Moved feed.t into spec, and added tests. (added 16 tests) 23:32

Powered by Cro and the Raku® Programming Language.

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

Thank you!