- CLN-list - ginac.de

My "effort" to make my mail keep the first place.
by Fabrice Marchant 16 Feb '06

16 Feb '06

> Mister Byron wrote : > I am deleting your message because the date is set far in advance, > in an apparent effort to make your email stand out over all others. What is "apparent" is not always true. I DID not wanted to do this. The date of this machine was false. It moved several years later indeed but I do not know how it happened. I have fixed this as soon I've noticed the problem. I use CLN lib for years and I love it. I'm NOT so stupid to use such a pitiful trick the first time I post to CLN-list ! But I couldn't care less you removed my msg. The important thing is I received interesting informations about the kind of hardware to build my next machine. Hopefully several persons didn't focused on this erroneous date and gently answered clever advices. Fabrice Marchant

2 1

A good PC hardware for CLN ?
by Fabrice Marchant 14 Feb '06

14 Feb '06

Hi ! I used CLN lib on linux debian Woody and try now libcln3 on debian Sarge ( on a Athlon T-Bird 1GHz 400Mo RAM ). I think to buy a more recent PC computer. What kind of PC for CLN - processor INTEL or AMD - to execute speedly integer computations ? Is motherboard worthy ? Thank you for your help. Fabrice

4 3

how to assign cl_I type to structure number
by mahesh ram 06 Feb '06

06 Feb '06

i am r.seetaram, M.Tech, C.S. i have faced the following problem: cl_I val=10; struct node { cl_I data; struct node * link; }; struct node * start=NULL; start = (struct node *) malloc(sizeof(struct node)); start->data = value; // error in this line; --------------------------------- Yahoo! Autos. Looking for a sweet ride? Get pricing, reviews, & more on new and used cars.

3 2

Re: Re: A good PC hardware for CLN ?
by Fabrice Marchant 28 Jan '06

28 Jan '06

> 1. Re: A good PC hardware for CLN ? (Joerg Arndt) > 2. Re: A good PC hardware for CLN ? (Richard B. Kreckel) Thank you very much for your answers. > > > ---------------------------------------------------------------------- > > Message: 1 > Date: Tue, 24 Jan 2006 16:33:40 +0100 > From: Joerg Arndt <jj(a)suse.de> > I invariably recommend AMD64 these days (socket 939). > > My off the shelf 2.2GHz system is on par with a 3.6GHz intel > equivalent (EMT64(?), ex-pen-sive!) for the interesting stuff. System > is 2-3.5 times faster than Athlon at same clock. Memory throughput is > a dream. > > The timings given (by Granlund) are a true eye opener: > http://swox.com/doc/x86-timing.pdf > > Today's intel 64bit cores are a lousy hack at best. > > Alternative architectures (PPC) might be fine but are surely > much more expensive. I'll follow you and the document : AMD64 ! Are dual core useful ? I looked at the prices (France) in this cheap store : http://www.ldlc.com/navigation/cat.html?sscat=140025&constructeur=C00000080… I noticed on this page : - AMD Athlon 64 3700+ - 2.2 GHz, Cache L2 1 Mo Socket 939 0.09 micron - Core San-Diego (version boîte) 259,01 euros and - AMD Athlon 64 X2 Dual-Core 3800+ Socket 939 0.09 micron (version boîte) 338,99 euros. > > Message: 2 > Date: Tue, 24 Jan 2006 22:45:39 +0100 > From: "Richard B. Kreckel" <kreckel(a)ginac.de> > Even without x86-64 Assembler code, the use of 64 bit Digits gives an > very impressive performance gain. If you buy such a machine, make sure > to run int in 64 bit userland if arbitrary precision arithmetic is > important for what you're doing. Please, what does 64 bit userland mean ? Maybe recompiling your .cc sources for 64 bits will improve the speed too ? Does g++3.3 is able to produce such code ? Arbitrary precision arithmetic is a fun stuff and then very important to soften the hard life ! Sorry for my poor english language. When I wrote : "Is the motherboard worthy ?", that didn't had any meaning. I meant : "Is the motherboard important for speed too ?". I wonder if processor cache size is important for integer computations with CLN ? ( I focus on integers. ) And RAM amount ? Dual channel ? Can we forget hard drive perfs, hoping a huge RAM avoids to swap ? Regards Fabrice

3 2

CLN 1.1.11 released
by Richard B. Kreckel 22 Nov '05

22 Nov '05

Good evening, This is to announce the availability of CLN 1.1.11. It's a maintenance release with only one change: All input of numbers is now much speedier (except when the base is a power of two, in which case the performance was already optimal). As always, it is available from <http://www.ginac.de/CLN/>. Enjoy! -richy. -- Richard B. Kreckel <http://www.ginac.de/~kreckel/>

1 0

CLN 1.1.10 released
by Richard B. Kreckel 23 Oct '05

23 Oct '05

Good evening, This is to announce the availability of CLN 1.1.10. It's a maintenance release that fixes some outstanding issues, most notabliy compilation failures on some new versions of GCC. To quote the NEWS file: Implementation changes * Removed the vptr of modular integer and univariate polynomial classes in order to fix some crashes in that sector of CLN. Code using this needs to be recompiled. * Many more operator signatures make the integration of CLN types with float/double easier. Other changes * Several compilation fixes. * Made it possible to cross-compile CLN. On another note, the GiNaC website has finally moved to a new site. Along with it, CLN's CVS repository and mailing list have moved, too. Without this, most probably there wouldn't be a release yet. If you were using CVS, please check out a new tree. And, please, make sure you use the cln-list(a)ginac.de address, and not one with uni-mainz.de in it. If unsure: the web page has all the relevant information updated. Happy hacking -richy. -- Richard B. Kreckel <http://www.ginac.de/~kreckel/>

1 0

We are moving...
by Richard B. Kreckel 08 Oct '05

08 Oct '05

Dear all, Due to ongoing technical problems at Mainz University (*), it was decided to move GiNaC's and CLN's web server / CVS server / mailing lists to a rented server. This should allow us to continue communicating via mailing lists, doing development and updating the web pages, things that have become difficult during the last 10 weeks or so. We are going to move the DNS entries ASAP. The hosted machine's IP-address is going to be 81.169.130.167. We'll certainly see some outage while the transfer occurs, but hopefully not for long time. Meanwhile, you can all make sure the address you use as mailing list addresses is ...(a)ginac.de, not ...(a)thep.physik.uni-mainz.de or ...(a)zino.physik.uni-mainz.de because these will cease to work. We apologize for the inconveninece -richy. (*) Case in point: right now the web server is dead. -- Richard B. Kreckel <http://www.ginac.de/~kreckel/>

1 0

DLL for mingw?
by David Bateman 05 Oct '05

05 Oct '05

I'm trying to build a DLL of CLN for mingw. I absolutely must have a DLL as my use of CLN is with octave (www.octave.org) and with the lack of symlinks under windows if I use a static library, CLN with be linked in to all of the seperate octave functions, and the amount of disk used with be enormous... However, I have a problem of many many resolved symbols in CLN when linking the DLL. I imagine the same problem must exist under cygwin as well. A sample of the type of error message I'm getting are exam.o(.text+0x133): In function `Z41__static_initialization_and_destruction_0ii': exam.cc:50: undefined reference to `global constructors keyed to cl_module__cl_prin_globals__firstglobalfun' exam_I.o(.text+0x96):e:/octave/build/bld/cln-1.1.9/tests/: undefined reference to `cln::cl_I::cl_I(char const*)' exam_RA.o(.text+0x96):e:/octave/build/bld/cln-1.1.9/tests/: undefined reference to `cln::cl_RA::cl_RA(char const*)' etc... Though I have hundreds of such unresolved symbols... Is there something basic that I'm missing? Or are these really unresolved and the problem only appears in the windows builds? Regards David -- David Bateman David.Bateman(a)motorola.com Motorola Labs - Paris +33 1 69 35 48 04 (Ph) Parc Les Algorithmes, Commune de St Aubin +33 1 69 35 77 01 (Fax) 91193 Gif-Sur-Yvette FRANCE The information contained in this communication has been classified as: [x] General Business Information [ ] Motorola Internal Use Only [ ] Motorola Confidential Proprietary

1 0

Types in cl_float
by Isidro Cachadiña Gutiérrez 12 Sep '05

12 Sep '05

Hi: I miss a cl_float(long double x, cl_F ) definition to make the code below to compile. The code compiles when A is defined as double. ----------------------------------------------------------------------------- #include <iostream> #include <cln/float.h> #include <cln/float_io.h> using namespace cln; using namespace std; int main(int argc, char **argv) { long double A; A=1; cl_F a; float_format_t precision; precision=float_format(100); a=cl_float(A,precision); cerr << a << endl; } --------------------------------------------------------------------------------------- -- *************************************************************************** * Dr. Isidro Cachadiña Gutiérrez * * Departamento de Física * * Facultad de Ciencias * * Universidad de Extremadura * * 06071 Badajoz ( SPAIN ) * * email: icacha(a)unex.es * * Teléfono: +34 924 289 300 Ext. 6826 Fax: +34 924 289 651 * *************************************************************************** * Usuario Linux: 8569 * * Para clave pública GnuPG: http://onsager.unex.es/firma.pub.asc * ***************************************************************************

1 0

CLN on MINGW
by David Robinow 23 Aug '05

23 Aug '05

The following patch allows building on Windows under MSYS. make check succeeds. --- /c/home/cln_cvs/cln/src/base/random/cl_random_from.cc 2004-06-23 17:11:21.000000000 -0400 +++ cl_random_from.cc 2005-08-23 11:31:13.938737700 -0400 @@ -50,6 +50,14 @@ return seed_lo + tmsbuf.tms_utime + tmsbuf.tms_stime; } +#elif defined(__MINGW32__) +#include <sys/timeb.h> +inline uint32 get_seed (void) +{ + struct timeb tm; + ftime(&tm); + return cln::highlow32(tm.time,tm.millitm * 1000); +} #endif #endif

1 0