January 2002 - Boost-users - lists.preview.boost.org

object_pool, O(N) destroy?
by kevin_godden＠agilent.com 29 Jan '02

29 Jan '02

Hi, I am considering using boost::object_pool for fast allocation and deallocation of likkle container/token objects. I need to make this as fast as possible, to this end I wonder why object_pool seems to use boost::pool::ordered_free() (O(N)) rather than the faster free() (O(1))? There must be a reason, but I can't seem to find it in the Lib's documentation. I would be interested in hearing if peoples experiences/comments on using object_pool. Thanks for reading! Kevin\

2 1

RE: [Boost-Users] RE: object_pool, O(N) destroy?
by kevin_godden＠agilent.com 29 Jan '02

29 Jan '02

I agree, it does sound like a job for a Policy whatsit. If I get a chance I could try a few things and see if anything works out and let you know. Although, if time constraints work out the way I think they will, I might have to stick with a cut&paste solution! Kevin\ -----Original Message----- From: scleary(a)jerviswebb.com [mailto:scleary@jerviswebb.com] Sent: Tuesday, January 29, 2002 5:13 AM To: Boost-Users(a)yahoogroups.com Cc: shammah(a)voyager.net Subject: [Boost-Users] RE: object_pool, O(N) destroy? > -----Original Message----- > From: hicks [mailto:hicks@kgk.co.jp] > > It sounds like a case where a type trait, identifying the need for the > pool to responsible for deletion, > would be appropriate. Then objects which did not need to be destroyed > (e.g. ptrs, "plain" structs) > could be handled by optimized code. But that's more work than cut and > paste. Actually, the difference is in how the pool is used, not on the type of object being allocated. So it sounds more like a case for a policy-based design. Too bad, I just don't have time to do that right now... :( -Steve Info: <http://www.boost.org> Wiki: <http://www.crystalclearsoftware.com/cgi-bin/boost_wiki/wiki.pl> Unsubscribe: <mailto:boost-users-unsubscribe@yahoogroups.com> Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/

1 0

RE: object_pool, O(N) destroy?
by scleary＠jerviswebb.com 29 Jan '02

29 Jan '02

> -----Original Message----- > From: hicks [mailto:hicks@kgk.co.jp] > > It sounds like a case where a type trait, identifying the need for the > pool to responsible for deletion, > would be appropriate. Then objects which did not need to be destroyed > (e.g. ptrs, "plain" structs) > could be handled by optimized code. But that's more work than cut and > paste. Actually, the difference is in how the pool is used, not on the type of object being allocated. So it sounds more like a case for a policy-based design. Too bad, I just don't have time to do that right now... :( -Steve

1 0

Dependencies for tokenizer?
by Anthony Kolarik 29 Jan '02

29 Jan '02

In order to use the tokenizer classes I end up requiring 47 files, which seems a little steep for my purposes. All I really need is to tokenize strings based on a couple of single char delimiters. Kind of a C++ step above strtok() was what I had in mind. Any ideas about how to get this functionality w/o all the headers? Thanks, -- Tony K.

3 2

RE: [Boost-Users] object_pool, O(N) destroy?
by kevin_godden＠agilent.com 29 Jan '02

29 Jan '02

Thanks for the Info Steve, In this case, I am less interested in the pool's destructor as I _should_ have a 1 to 1 on malloc()ing and destroying the individual objects (if all goes well...). The pool will only be destroyed once during the lifetime of the application so I could probably suffer a O(N^2) pool destructor. I will try what you recommend in your postscript. Thanks again.. Kevin\ -----Original Message----- From: scleary(a)jerviswebb.com [mailto:scleary@jerviswebb.com] Sent: Monday, January 28, 2002 2:10 PM To: Boost-Users(a)yahoogroups.com Cc: shammah(a)voyager.net Subject: RE: [Boost-Users] object_pool, O(N) destroy? > From: kevin_godden(a)agilent.com [mailto:kevin_godden@agilent.com] > > I am considering using boost::object_pool for fast allocation and > deallocation of likkle container/token objects. I need to make this as fast > as possible, to this end I wonder why object_pool seems to use > boost::pool::ordered_free() (O(N)) rather than the faster > free() (O(1))? > > There must be a reason, but I can't seem to find it in the Lib's > documentation. The object_pool keeps its free list ordered because of the requirements of its destructor. Its destructor steps through the free list, calling the destructor for any objects that weren't freed. The current algorithm is O(N) because the free list is ordered; if it wasn't ordered, it would be O(N^2). I'll document this in the code. -Steve P.S. An object_pool without the destructor behaviour may be what you want. Boost.Pool doesn't provide one, but you could just cut & paste the object_pool, remove the destructor, and change ordered_malloc/ordered_free to the regular malloc/free (you also won't need nextof). Just note that if you ever construct something from that pool and don't destruct it, you enter the fun realm of undefined behaviour... :) Info: <http://www.boost.org> Wiki: <http://www.crystalclearsoftware.com/cgi-bin/boost_wiki/wiki.pl> Unsubscribe: <mailto:boost-users-unsubscribe@yahoogroups.com> Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/

2 1

1_26_0: regex/thread on W32 build via jam?
by Bruno.Voigt＠ic3s.de 28 Jan '02

28 Jan '02

Hi! For my statically linked application (w32/sol8/linux) I was up to now using only the boost regex library. I build the regex lib the oldstyle way going to the build path and calling nmake -f vc7.mak, etc. In the project settings I included the regex/build/vc7 path as an additional lib search path. The buld step works as expected this way. boost::thread does provide only the newstyle jam build way. So I downloaded the boost::winbuild stuff and tried to build the whole boost via jam -sBOOST_ROOT=. -sTOOLS=msvc. It shows some errors for some pytho stuff but builds the regex and thread lib part. So I get for example a path boost_1_26_0\libs\regex\build\bin\libboost_regex.lib\msvc\release\runtime-link-dynamic with a libboost_regex.lib But how do I have to change my project settings for a static executable? I just changed the the lib search path to this new monster path and it does not work The linker is missing the vc7-re300dm.lib then. And how to link the boost thread stuff? Is it supported for statically linked apps at all? Any suggestions on how to make use of the jam build stuff in the right way ae really appreciated. TIA, Bruno PS: will there be a working suncc-tools.jam file in the next release? -- Bruno.Voigt(a)ic3s.de # IC3S AG # Tel. +49.4106.655.105 # Fax. +49.4106.655.55 http://www.ic3s.de # Mobile +49.700.686.00.686 [Non-text portions of this message have been removed]

2 1

RE: [Boost-Users] object_pool, O(N) destroy?
by scleary＠jerviswebb.com 28 Jan '02

28 Jan '02

> From: kevin_godden(a)agilent.com [mailto:kevin_godden@agilent.com] > > I am considering using boost::object_pool for fast allocation and > deallocation of likkle container/token objects. I need to make this as fast > as possible, to this end I wonder why object_pool seems to use > boost::pool::ordered_free() (O(N)) rather than the faster > free() (O(1))? > > There must be a reason, but I can't seem to find it in the Lib's > documentation. The object_pool keeps its free list ordered because of the requirements of its destructor. Its destructor steps through the free list, calling the destructor for any objects that weren't freed. The current algorithm is O(N) because the free list is ordered; if it wasn't ordered, it would be O(N^2). I'll document this in the code. -Steve P.S. An object_pool without the destructor behaviour may be what you want. Boost.Pool doesn't provide one, but you could just cut & paste the object_pool, remove the destructor, and change ordered_malloc/ordered_free to the regular malloc/free (you also won't need nextof). Just note that if you ever construct something from that pool and don't destruct it, you enter the fun realm of undefined behaviour... :)

1 0

MacCVS Pro hints
by voidampersand 27 Jan '02

27 Jan '02

I'm trying to get with the program here, so I just got the boost sources from CVS, using MacCVS Pro. The main thing that tripped me up was the instructions on the SourceForge page have command lines that are word wrapped. Anyway, here are some MacCVS session settings that work for me: Remote Host Information Server Hostname: cvs.boost.sourceforge.net CVS Root: /cvsroot/boost CVS User Name: anonymous CVS Password: [+] Save Password Also, in "File Translation and Encoding" you will want to add these mappings before you check out: .hpp TEXT CWIE .cxx TEXT CWIE .jam TEXT CWIE .pdf TEXT CARO MacCVS Pro is available at <http://sourceforge.net/projects/maccvspro/>. Regards, Tom

2 1

Re: [Boost-Users] memcmp, strlen, strcmp not found in string.h?
by Steven Woolgar 27 Jan '02

27 Jan '02

> My question: Why aren't they in STLPort's string.h? > Is it a voluntary omission, and if so, why? > Should I talk to the guys who created Boost instead? > Should I be asking this on the developer's mailing list?? =) If I am not mistaken, the standard C functions would be in the the standard C include files, not the C++ files. So: strlen would be in <cstring> the string class would be in <string> Note that the new include style does not have the .h at the end. HTH Woolie

1 0

Suggestion -- Boost Building
by Duane Murphy 26 Jan '02

26 Jan '02

It seems a lot of questions come up on the list about "building" the boost libraries. I wonder how many of these people realize that most of the functionality of the Boost "libraries" is completely encapsulated in the header files? There are significant portions like regex and the graph library (?) that do require compiling, but much of it does not. Is there anything that can be done to make it more clear that compiling is often not necessary? I'm just trying to relieve some stress from people who may very well not require compiling. And I have compassion for those that do and are having trouble. I'm not sure exactly what can be done, but it might be worth some thought. ...Duane

1 0