- GiNaC-devel - ginac.de

Re: [GiNaC-cvs] ginac flags.h mul.cpp power.cpp
by Richard B. Kreckel 05 Nov '07

05 Nov '07

CVS Service wrote: > Update of /home/cvs/GiNaC/ginac by vollinga > > Modified Files: > Tag: ginac_1-4 > flags.h mul.cpp power.cpp > Log Message: > Synced to HEAD: > - This helps mul::expand() and friends to recognize objects which have > no indices at all [Sheplyakov]. Chris, We are curious what can be done about the performance regression one of your patches for dummy index renaming introduced. The seemingly innocuous patch has been identified by Jens and Alexei: <http://www.ginac.de/pipermail/ginac-list/2007-September/001215.html>. At first, the above log message suggested to me that the problem has been dealt with. But that is not so: The program is still extremely slow. In fact, Alexei's patches are entirely unrelated. Since neither an explanation of what's happening nor a fix has been suggested, we would really appreciate some expert insight by you. Thanks in advance -richy. -- Richard B. Kreckel <http://www.ginac.de/~kreckel/>

4 5

Restore the speed of polynomial operations.
by Alexei Sheplyakov 18 Sep '07

18 Sep '07

Hello! These patches commit 0633ed8082961673eedc092689e06fa39d6bc322 Author: Jens Vollinga <vollinga(a)thep.physik.uni-mainz.de> Date: Thu May 19 18:10:40 2005 +0000 Fixed bug in expanding expressions containing dummy indices. [V.Kisil] (see <http://www.ginac.de/pipermail/ginac-cvs/2005-May/001586.html>) commit 6b1b11264fe4872132fef8cb8b22cb983ba01736 Author: Chris Dams <Chris.Dams(a)mi.infn.it> Date: Thu Jul 13 20:02:38 2006 +0000 Needed to call get_all_dummy_indices_safely more often. (see http://www.ginac.de/pipermail/ginac-cvs/2006-July/001780.html) commit 4cf53676d7823ec3c55a90df7723f7f1492a95b2 Author: Chris Dams <Chris.Dams(a)mi.infn.it> Date: Mon Jul 24 16:47:43 2006 +0000 Improved dummy index renaming. (see http://www.ginac.de/pipermail/ginac-cvs/2006-July/001782.html) made polynomial operations (in particular expand()) substantially slower. For example, Fateman's polynomial expanding benchark degraded from 32sec. to 80sec. My own code which uses multivariate GCD (which in turn uses .expand() extensively) become 4x slower. I'll post 4 patches which address the issue in reply to this message. Comments, questions, suggestions, etc. are wellcome. Best regards, Alexei -- All science is either physics or stamp collecting.

3 12

Volunteer to continue the work on NTL factorisation bindings
by Remco Bloemen 06 Sep '07

06 Sep '07

I have created a simple factorisation for polynomials in Q[x] with real roots (source attached, based on [2]) but it's really slow. Creating a reasonable implementation would require much work. An NTL based implementation would be very powerfull and feasible, as said in [1]. I would like to volunteer to continue this work. I have the following questions: 1) Where can I obtain the source of this binding? (or should I start from the code in gTybalt, or from scratch?) 2) Is there any chance such a patch will be included? (this would imply dependecy on NTL) 2a) And if it is compile-time optional? 2b) And if it also included integration of rational functions? 2c) And if it also included integration of trancedental functions? 3) Do you have any related advice, hints or pointers? [1] http://www.ginac.de/pipermail/ginac-devel/2004-August/000695.html [2] http://yacas.sourceforge.net/Algochapter3.html

3 3

[PATCH] improved CLN C++ code output
by Alexei Sheplyakov 05 Sep '07

05 Sep '07

* small integers are printed as int (not as cl_I) * no more ugly x+-y output * powers always printed using cln::expt function. Please note: this patch is for development branch only, since it breaks binary compatibility (due to power::do_print_csrc_cl_N). I'll post patches for GiNaC 1.[34] later. --- ginac/add.cpp | 34 +++++++++++++++------------------- ginac/numeric.cpp | 51 ++++++++++++++++++++++++++++++++++++++++++++++----- ginac/power.cpp | 35 +++++++++++++++++++++-------------- ginac/power.h | 1 + 4 files changed, 83 insertions(+), 38 deletions(-) diff --git a/ginac/add.cpp b/ginac/add.cpp index b37ab79..98e8c79 100644 --- a/ginac/add.cpp +++ b/ginac/add.cpp @@ -23,6 +23,7 @@ #include <iostream> #include <stdexcept> #include <limits> +#include <string> #include "add.h" #include "mul.h" @@ -172,37 +173,32 @@ void add::do_print_csrc(const print_csrc & c, unsigned level) const if (precedence() <= level) c.s << "("; - // Print arguments, separated by "+" + // Print arguments, separated by "+" or "-" epvector::const_iterator it = seq.begin(), itend = seq.end(); + char separator = ' '; while (it != itend) { - // If the coefficient is -1, it is replaced by a single minus sign - if (it->coeff.is_equal(_ex1)) { + // If the coefficient is negative, separator is "-" + if (it->coeff.is_equal(_ex_1) || + ex_to<numeric>(it->coeff).numer().is_equal(*_num_1_p)) + separator = '-'; + c.s << separator; + if (it->coeff.is_equal(_ex1) || it->coeff.is_equal(_ex_1)) { it->rest.print(c, precedence()); - } else if (it->coeff.is_equal(_ex_1)) { - c.s << "-"; + } else if (ex_to<numeric>(it->coeff).numer().is_equal(*_num1_p) || + ex_to<numeric>(it->coeff).numer().is_equal(*_num_1_p)) + { it->rest.print(c, precedence()); - } else if (ex_to<numeric>(it->coeff).numer().is_equal(*_num1_p)) { - it->rest.print(c, precedence()); - c.s << "/"; - ex_to<numeric>(it->coeff).denom().print(c, precedence()); - } else if (ex_to<numeric>(it->coeff).numer().is_equal(*_num_1_p)) { - c.s << "-"; - it->rest.print(c, precedence()); - c.s << "/"; + c.s << '/'; ex_to<numeric>(it->coeff).denom().print(c, precedence()); } else { it->coeff.print(c, precedence()); - c.s << "*"; + c.s << '*'; it->rest.print(c, precedence()); } - // Separator is "+", except if the following expression would have a leading minus sign or the sign is sitting in parenthesis (as in a ctor) ++it; - if (it != itend - && (is_a<print_csrc_cl_N>(c) || !it->coeff.info(info_flags::real) // sign inside ctor arguments - || !(it->coeff.info(info_flags::negative) || (it->coeff.is_equal(*_num1_p) && is_exactly_a<numeric>(it->rest) && it->rest.info(info_flags::negative))))) - c.s << "+"; + separator = '+'; } if (!overall_coeff.is_zero()) { diff --git a/ginac/numeric.cpp b/ginac/numeric.cpp index ef61ce7..74f0e50 100644 --- a/ginac/numeric.cpp +++ b/ginac/numeric.cpp @@ -399,6 +399,38 @@ static void print_real_csrc(const print_context & c, const cln::cl_R & x) } } +template<typename T1, typename T2> +static inline bool coerce(T1& dst, const T2& arg); + +/** + * @brief Check if CLN integer can be converted into int + * + * @sa http://www.ginac.de/pipermail/cln-list/2006-October/000248.html + */ +template<> static inline bool +coerce<int, cln::cl_I>(int& dst, const cln::cl_I& arg) { + static const cln::cl_I cl_max_int = + (cln::cl_I)(long)(std::numeric_limits<int>::max()); + static const cln::cl_I cl_min_int = + (cln::cl_I)(long)(std::numeric_limits<int>::min()); + if ((arg >= cl_min_int) && (arg <= cl_max_int)) { + dst = cl_I_to_int(arg); + return true; + } + return false; +} + +template<> static inline bool +coerce<unsigned int, cln::cl_I>(unsigned int& dst, const cln::cl_I& arg) { + static const cln::cl_I cl_max_uint = + (cln::cl_I)(unsigned long)(std::numeric_limits<unsigned int>::max()); + if ((! minusp(arg)) && (arg <= cl_max_uint)) { + dst = cl_I_to_uint(arg); + return true; + } + return false; +} + /** Helper function to print real number in C++ source format using cl_N types. * * @see numeric::print() */ @@ -406,11 +438,20 @@ static void print_real_cl_N(const print_context & c, const cln::cl_R & x) { if (cln::instanceof(x, cln::cl_I_ring)) { - // Integer number - c.s << "cln::cl_I(\""; - print_real_number(c, x); - c.s << "\")"; - + int dst; + // fixnum + if (coerce(dst, cln::the<cln::cl_I>(x))) { + // can be converted to native int + if (dst < 0) + c.s << "(-" << dst << ")"; + else + c.s << dst; + } else { + // bignum + c.s << "cln::cl_I(\""; + print_real_number(c, x); + c.s << "\")"; + } } else if (cln::instanceof(x, cln::cl_RA_ring)) { // Rational number diff --git a/ginac/power.cpp b/ginac/power.cpp index faa0d05..68bec71 100644 --- a/ginac/power.cpp +++ b/ginac/power.cpp @@ -50,7 +50,8 @@ GINAC_IMPLEMENT_REGISTERED_CLASS_OPT(power, basic, print_func<print_latex>(&power::do_print_latex). print_func<print_csrc>(&power::do_print_csrc). print_func<print_python>(&power::do_print_python). - print_func<print_python_repr>(&power::do_print_python_repr)) + print_func<print_python_repr>(&power::do_print_python_repr). + print_func<print_csrc_cl_N>(&power::do_print_csrc_cl_N)) typedef std::vector<int> intvector; @@ -162,6 +163,21 @@ static void print_sym_pow(const print_context & c, const symbol &x, int exp) } } +void power::do_print_csrc_cl_N(const print_csrc_cl_N& c, unsigned level) const +{ + if (exponent.is_equal(_ex_1)) { + c.s << "recip("; + basis.print(c); + c.s << ')'; + return; + } + c.s << "expt("; + basis.print(c); + c.s << ", "; + exponent.print(c); + c.s << ')'; +} + void power::do_print_csrc(const print_csrc & c, unsigned level) const { // Integer powers of symbols are printed in a special, optimized way @@ -172,29 +188,20 @@ void power::do_print_csrc(const print_csrc & c, unsigned level) const c.s << '('; else { exp = -exp; - if (is_a<print_csrc_cl_N>(c)) - c.s << "recip("; - else - c.s << "1.0/("; + c.s << "1.0/("; } print_sym_pow(c, ex_to<symbol>(basis), exp); c.s << ')'; // <expr>^-1 is printed as "1.0/<expr>" or with the recip() function of CLN } else if (exponent.is_equal(_ex_1)) { - if (is_a<print_csrc_cl_N>(c)) - c.s << "recip("; - else - c.s << "1.0/("; + c.s << "1.0/("; basis.print(c); c.s << ')'; - // Otherwise, use the pow() or expt() (CLN) functions + // Otherwise, use the pow() function } else { - if (is_a<print_csrc_cl_N>(c)) - c.s << "expt("; - else - c.s << "pow("; + c.s << "pow("; basis.print(c); c.s << ','; exponent.print(c); diff --git a/ginac/power.h b/ginac/power.h index f5e65d6..3375040 100644 --- a/ginac/power.h +++ b/ginac/power.h @@ -88,6 +88,7 @@ protected: void do_print_csrc(const print_csrc & c, unsigned level) const; void do_print_python(const print_python & c, unsigned level) const; void do_print_python_repr(const print_python_repr & c, unsigned level) const; + void do_print_csrc_cl_N(const print_csrc_cl_N & c, unsigned level) const; ex expand_add(const add & a, int n, unsigned options) const; ex expand_add_2(const add & a, unsigned options) const; -- 1.4.4.4 Best regards, Alexei -- All science is either physics or stamp collecting.

1 2

[PATCH] Add self-contained example of a GiNaC class.
by Alexei Sheplyakov 05 Sep '07

05 Sep '07

Hello! Some people claim that extending GiNaC is "unnecessary complex" (see e.g. http://www.ginac.de/pipermail/ginac-list/2006-July/000882.html) So I think it is good idea to provide self-contained, minimalistic example of a GiNaC class. The actual code is copy/pasted from the tutorial. --- doc/examples/Makefile.am | 3 +- doc/examples/mystring.cpp | 104 +++++++++++++++++++++++++++++++++++++++++++++ doc/tutorial/ginac.texi | 3 +- 3 files changed, 108 insertions(+), 2 deletions(-) diff --git a/doc/examples/Makefile.am b/doc/examples/Makefile.am index 5ee2c5d..e63da35 100644 --- a/doc/examples/Makefile.am +++ b/doc/examples/Makefile.am @@ -1,6 +1,7 @@ ## Process this file with automake to produce Makefile.in -EXFILES = archive1.cpp compile1.cpp compile2.cpp compile3.cpp lanczos.cpp +EXFILES = archive1.cpp compile1.cpp compile2.cpp compile3.cpp lanczos.cpp \ + mystring.cpp TEXI = ginac-examples.texi diff --git a/doc/examples/mystring.cpp b/doc/examples/mystring.cpp new file mode 100644 index 0000000..cf0d8a2 --- /dev/null +++ b/doc/examples/mystring.cpp @@ -0,0 +1,104 @@ +/** + * @file mystring.cpp Example of extending GiNaC: writing new classes + */ +#include <iostream> +#include <string> +#include <stdexcept> +using namespace std; + +#include <ginac/ginac.h> +using namespace GiNaC; + +class mystring : public basic +{ + GINAC_DECLARE_REGISTERED_CLASS(mystring, basic) + +public: + mystring(const string &s); + mystring(const char *s); + ex eval(int level) const; + +private: + string str; + +protected: + void do_print(const print_context &c, unsigned level = 0) const; + +}; + +GINAC_IMPLEMENT_REGISTERED_CLASS_OPT(mystring, basic, + print_func<print_context>(&mystring::do_print)) + +// ctors +mystring::mystring() : inherited(&mystring::tinfo_static) { } +mystring::mystring(const string &s) : inherited(&mystring::tinfo_static), str(s) { } +mystring::mystring(const char *s) : inherited(&mystring::tinfo_static), str(s) { } + +// comparison +int mystring::compare_same_type(const basic &other) const +{ + const mystring &o = static_cast<const mystring &>(other); + int cmpval = str.compare(o.str); + if (cmpval == 0) + return 0; + else if (cmpval < 0) + return -1; + else + return 1; +} + +// archiving/unarchiving +mystring::mystring(const archive_node &n, lst &sym_lst) : inherited(n, sym_lst) +{ + n.find_string("string", str); +} + +void mystring::archive(archive_node &n) const +{ + inherited::archive(n); + n.add_string("string", str); +} + +ex mystring::unarchive(const archive_node &n, lst &sym_lst) +{ + return (new mystring(n, sym_lst))->setflag(status_flags::dynallocated); +} + +// printing +void mystring::do_print(const print_context &c, unsigned level) const +{ + // print_context::s is a reference to an ostream + c.s << '\"' << str << '\"'; +} + +/** + * evaluation: all strings automatically converted to lowercase with + * non-alphabetic characters stripped, and empty strings removed + */ +ex mystring::eval(int level) const +{ + string new_str; + for (size_t i=0; i<str.length(); i++) { + char c = str[i]; + if (c >= 'A' && c <= 'Z') + new_str += tolower(c); + else if (c >= 'a' && c <= 'z') + new_str += c; + } + + if (new_str.length() == 0) + return 0; + else + return mystring(new_str).hold(); +} + +int main(int argc, char** argv) +{ + ex e = mystring("Hello, world!"); + cout << is_a<mystring>(e) << endl; + cout << ex_to<basic>(e).class_name() << endl; + cout << e << endl; + ex another = pow(mystring("One string"), 2*sin(Pi-mystring("Another string"))); + cout << another << endl; + return 0; +} diff --git a/doc/tutorial/ginac.texi b/doc/tutorial/ginac.texi index a6befad..b2038db 100644 --- a/doc/tutorial/ginac.texi +++ b/doc/tutorial/ginac.texi @@ -7823,7 +7823,8 @@ Now we will start implementing a new class @code{mystring} that allows placing character strings in algebraic expressions (this is not very useful, but it's just an example). This class will be a direct subclass of @code{basic}. You can use this sample implementation as a starting point -for your own classes. +for your own classes @footnote{The self-contained source for this example is +included in GiNaC, see the @file{doc/examples/mystring.cpp} file.}. The code snippets given here assume that you have included some header files as follows: -- 1.4.4.4 Best regards, Alexei -- All science is either physics or stamp collecting.

1 0

[PATCH] INSTALL: do not mention nonexisting autogen.sh
by Alexei Sheplyakov 05 Sep '07

05 Sep '07

That file was removed a long time ago, so... --- INSTALL | 5 ++--- 1 files changed, 2 insertions(+), 3 deletions(-) diff --git a/INSTALL b/INSTALL index 2c60d2d..1fda652 100644 --- a/INSTALL +++ b/INSTALL @@ -60,9 +60,8 @@ library works correctly you can use $ make check -The "configure" script (and "autogen.sh", which invokes "configure") can be -given a number of options to enable and disable various features. For a -complete list, type: +The "configure" script can be given a number of options to enable and +disable various features. For a complete list, type: $ ./configure --help -- 1.4.4.4 Best regards, Alexei -- All science is either physics or stamp collecting.

1 0

GiNaC release 1.4.0
by Jens Vollinga 04 Sep '07

04 Sep '07

Hello everybody, GiNaC 1.4.0 is out and available. The changes are: - New tinfo mechanism. - Removed rtt from class symbol. - Faster archiving by adding a map from strings to indices in the atoms vector. - Improved unarchiving: algorithms order N instead of order N^2. - Added compile function, excompiler class. - Added exset type. - Added step function to GiNaCs built-in functions. - Added is_polynomial() method. - Added real_part() and imag_part() methods. - Added matrix::is_zero_matrix() method. - Added evalm() method to pseries class. - Improved dummy index renaming. - Many improvements for class clifford. - New flag for positive symbols/numbers, added possymbol class. - Added programming examples in doc/examples *Note*: the tinfo mechanism has changed from fixed magic numbers to dynamically assigned tinfo-ids. If have written your own GiNaC classes, you need to adjust your code! As always, this release can be downloaded from http://www.ginac.de/Download.html Best wishes, Jens

2 5

GiNaC release 1.3.8
by Jens Vollinga 28 Aug '07

28 Aug '07

Hello everybody, GiNaC 1.3.8 is out and available. The changes are: - Drop support of ginac-config and ginac.m4. Please use pkg-config instead. - atan2(y,x) branch cut correction. - Fixed bug in series expansion. - Additional transformations for mul and power. - Clifford units now honor representation labels. - Correct csrc output for idx and fderivative. - Improved handling of convergence transformationis for Li/G. - Fix compilation issues with prereleases of GCC 4.3. - info_flags::expanded added. - input_parser also accepts pow(). - texinfo.tex removed from package. As always, this release can be downloaded from http://www.ginac.de/Download.html Best wishes, Jens

1 0

Release?
by Vladimir Kisil 28 Aug '07

28 Aug '07

Jens Vollinga <vollinga at physik.uni-wuppertal.de> Mon Aug 27 10:59:40 CEST 2007: > I'd like to do a 1.3.8 release today and (probably) a 1.4 release > tomorrow. If somebody has some patches in the making or wants certain > issues/problems to be resolved first, I will delay that plan, of course. Is file ginac/compiler.h from the Alexei's patch in the tree? Best wishes, Vladimir -- Vladimir V. Kisil email: kisilv(a)maths.leeds.ac.uk -- www: http://maths.leeds.ac.uk/~kisilv/

2 1

Release?
by Jens Vollinga 27 Aug '07

27 Aug '07

Hi, I'd like to do a 1.3.8 release today and (probably) a 1.4 release tomorrow. If somebody has some patches in the making or wants certain issues/problems to be resolved first, I will delay that plan, of course. Regards, Jens

1 0