Pike-devel November 2003

pike-devel@lists.lysator.liu.se

7 participants
35 discussions

alloca usage
by nisse＠lysator.liu.se 20 Nov '03

20 Nov '03

It seems that the default configuration of GMP, with WANT_TMP_ALLOCA, allocates arbitrarily large temporaries on the stack. For example mpn_tdiv_qr allocates temporary storage proportional to the size of the input. Fredrik Hübinette (Pike author) reports that his pike program that calculates a million decimals of Pi crashes in gmp with a default configuration of gmp-4.1.2, but works fine if gmp is reconfigured with --enable-alloca=malloc-reentrant. The problem is made worse by linux threads (and perhaps threads on other systems as well), which limits stack space to only 4 MB. (See http://fredrik.hubbe.net/hacker/viewtopic.php?t=26). Perhaps it would be a good idea to define TMP_ALLOC as something like #define TMP_ALLOC(size) \ ((size) < PAGESIZE*2) ? alloca((size)) : heap_alloc((size), &tmp_marker) Then for small inputs, we get only a single comparison of extra overhead, while for large inputs, the allocation overhead should be small compared to the cost of the actual computation. Due to the way the "red zone" stack extension mechanism works, I don't think it's reliable to allocate more than a few pages of data on the stack at a time. If you allocate too many pages you may get a segmentation violation of you access the pages in an unfortunate order. Regards, /Niels

1 0

Strange GMP coredumps
by Fredrik (Naranek) Hubinette (Real Build Master) ＠ Pike (-) developers forum 20 Nov '03

20 Nov '03

I've figured out why Pike dumps core when using very large numbers in GMP. See this link for more info: http://fredrik.hubbe.net/hacker/viewtopic.php?t=26

3 5

Strange GMP coredumps
by Johan Sundstr�m (f�ngar heffaklumpar) ＠ Pike (-) developers forum 19 Nov '03

19 Nov '03

(For Linux systems, at least.)

1 0

machine.h
by Mirar ＠ Pike developers forum 19 Nov '03

19 Nov '03

The errors in this build, http://pike.ida.liu.se/generated/pikefarm/7.5/1446_351/compilelog.txt >> ...global.h:87:21: machine.h: No such file or directory is that a common error? Any clues? machine.h *is* in the build root where it should be.

2 5

support my pike presentation on irc
by Martin Baehr 17 Nov '03

17 Nov '03

hi, today the 18th from about 8pm to 10pm european time il'll be giving a pike presentation to the linux user group in essen. as usual i'd like to introduce the pike community via a live chat. so if you are around during that time probably somewhere in the middle. (9pm europe, 3pm US-eastern, 9am NZ) i'd welcome you to join #pike on irc.caudium.net hope to see you there, greetings, martin. -- interested in doing pike programming, sTeam/caudium/pike/roxen training, sTeam/caudium/roxen and/or unix system administration anywhere in the world. -- pike programmer working and travelling in europe open-steam.org unix system- bahai.or.at iaeste.(tuwien.ac|or).at administrator (stuts|black.linux-m68k).org is.(schon.org|root.at) Martin Bähr http://www.iaeste.or.at/~mbaehr/

1 0

Re: xenofarm proxy patch
by Marcus Comstedt (ACROSS) (Hail Ilpalazzo!) ＠ Pike (-) developers forum 16 Nov '03

16 Nov '03

This is the same way as in Pike, btw. :-)

2 1

xenofarm proxy
by Martin Baehr 16 Nov '03

16 Nov '03

hi, i just tried running xenofarm on one of my machines in vienna, then i noticed that there is a problem because i am behind a http proxy. setting $http_proxy does not seem to be enough, and grepping through the source i could not find any reference to proxy support either. what does it take to get xenofarm to run behind a http proxy? greetings, martin.

2 7

Re: Can I get detailed documents about Pike languages design
by david chen 14 Nov '03

14 Nov '03

>From: "Martin Stjernholm, Roxen IS @ Pike developers forum" <10353(a)lyskom.lysator.liu.se> >To: pike-devel(a)lists.lysator.liu.se >Subject: Can I get detailed documents about Pike languages design >Date: Thu, 13 Nov 2003 05:30:01 +0100 (CET) > >Nice to hear. Unfortunately there isn't much in the way of core docs, >so you're pretty much left with investigating the source and the >commentary in it. I have at least added a manifest for the files in >the src directory now (in 7.5 - please check it out from the anonymous >cvs) so that you'll get a little help in finding your way there. > >Anyway, the compiler is fairly straightforward in principle (two pass, >lexer, LALR parser, abstract syntax tree, tree optimizer, opcode >generator, peephole optimizer) so provided you know standard compiler >theory it's mostly a matter of learning where the stuff is and what >it's called. The exceptions to that are mainly how the resolver works >and what's done in the different passes, I suspect. You can always ask >questions here; at least I will do my best to answer. > >Do you have anything specific in mind that you want to do? > > / Martin Stjernholm, Roxen IS > Thank you very much. I have not special idea about Pike design yet, because I am not familiar with the design principle of Pike, which is my current task to be done. In the furture, I hope to do some optimization on Pike. I did some interprocedural analysis and optimization for academic research. But I think this method is not adapt to Pike, which is complex, results in lower compiling effeciency, and mainly applied to compiled languages. Maybe from other open interpreted languages, we can get some inspiration to do it. David _________________________________________________________________ 与联机的朋友进行交流，请使用 MSN Messenger: http://messenger.msn.com/cn

2 1

-DNOT_INSTALLED
by Mirar ＠ Pike developers forum 14 Nov '03

14 Nov '03

There's lots of red dots in pikefarm for 7.5 now. Anyone working on it? | Compiling /home/xenofarm/xenoclient/pike-7.5/orchid.mirar.org/buildtmp/Pike7.5-20031113-045850/src/post_modules/GL/top.c | /home/xenofarm/xenoclient/pike-7.5/orchid.mirar.org/buildtmp/Pike7.5-20031113-045850/build/linux-2.6.0-test9-i686/pike -DNOT_INSTALLED -DPRECOMPILED_SEARCH_MORE -m/home/xenofarm/xenoclient/pike-7.5/orchid.mirar.org/buildtmp/Pike7.5-20031113-045850/build/linux-2.6.0-test9-i686/master.pike /home/xenofarm/xenoclient/pike-7.5/orchid.mirar.org/buildtmp/Pike7.5-20031113-045850/src/post_modules/GL/gen.pike < /home/xenofarm/xenoclient/pike-7.5/orchid.mirar.org/buildtmp/Pike7.5-20031113-045850/src/post_modules/GL/auto.c.in > /home/xenofarm/xenoclient/pike-7.5/orchid.mirar.org/buildtmp/Pike7.5-20031113-045850/src/post_modules/GL/auto.c | Unknown options -DNOT_INSTALLED.

3 3

Re: pike external module documentation
by Martin Stjernholm, Roxen IS ＠ Pike developers forum 13 Nov '03

13 Nov '03

> The core modules are either built with a pike glue file, or if one > isn't provided, a two line glue file is automatically generated. /.../ Sorry, but that's not correct. You must be looking at a pike built with static modules. They get a glue file to appear in the right spot in the namespace, but dynamic ones don't need that. > True, and something that I shouldn't have let slip. It was copied from > somewhere within the pike tree, so somewhere there's a testsuite that > has behavior that might not be so good test-wise. /.../ Some of the standard modules, e.g. Gmp, do that because they are optional, so the main testsuite still works when they are disabled. But that doesn't apply to third party modules. / Martin Stjernholm, Roxen IS Previous text: >2003-11-10 23:52: >Subject: Re: pike external module documentation >-------------------------------------------------------------------- >Thanks for the feedback... I will try to get the changes you suggested >incorporated. > >> Is it possible to build the docs for the module only, without merging >> it with the system docs? > >I hadn't thought of this, but it should be possible, though right now >the build assumes the modref would be installed in the system >documentation directory pikedir/docs/modref. Again, this should be >relatively simple to accomplish. > >> What about pike only modules? Should be trivial, but it'd still be >> nice if they too are covered by the standard tools so that they can be >> handled the same way. Maybe it works to provide a standard Makefile >> for that case, so that not even autoconf is required. > >I checked some changes in about 2 weeks ago that should fix this. I've >actually been testing on a pike-only module, and it seems to work just >fine with a copy of pike from CVS. > >> Doesn't work for modules with no pike layer. Also, the check works >> better if it's a file which is very specific for the module. > >That choice of filename was really just a suggestion; I assumed that >the text following it would clarify that it would be wise to choose a >relatively unique filename (i think i wrote that; but will check.) > >> Btw, you seem to assume that there always should be a pike glue file >> like that. Is it really necessary? > >The core modules are either built with a pike glue file, or if one >isn't provided, a two line glue file is automatically generated. I was >merely following convention; whether it's a good or necessary >convention I'll leave to others. > >> It should be enough with "VPATH=@srcdir@". (A while ago I cleaned up a >> lot of vpaths like that in the module tree - seems to have been a bad >> case of cut'n'paste back to very old code.) >> > >I probably had that source file for a while, so i will fix this. > >> A bit odd example. In the testsuite for a specific module, isn't it >> better to assume it exists? If there has been some major problem so >> that the module isn't installed at all, one wouldn't expect it to >> silently pass the testsuite. > >True, and something that I shouldn't have let slip. It was copied from >somewhere within the pike tree, so somewhere there's a testsuite that >has behavior that might not be so good test-wise. I was mostly >interested in getting the framework together. I agree that I might have >picked a better set of tests, but was in a bit of a hurry when i did >it. It actually might be nice for someone with actual knowledge of the >testsuite system to put together a few paragraphs about it. > > > > / Brevbäraren >

3 8

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

Pike-devel November 2003