January 2009 - PPL-devel - BUGSENG Mailman

Bug in C++ interface (comparison between rational and double)
by Abramo Bagnara 08 Feb '09

08 Feb '09

The comparison between mpz_class and double works fine with +inf and -inf, while the comparison between mpq_class and double gives a floating point exception, due to inappropriate mpq_set_d call in eval specialization. I see two way to fix this: 1) add support for comparison between mpq and double in C interface and use that in gmpxx.h 2) add infinity (and nan) check in comparison eval specializations in gmpxx.h IMHO 1) gives a more complete solution. If you want we're willing to write the (trivial) patch. -- Abramo Bagnara Opera Unica Phone: +39.0546.656023 Via Borghesi, 16 48014 Castel Bolognese (RA) - Italy

4 5

Re: [PPL-devel] mpz_addmul_ui with ui = 1
by Roberto Bagnara 06 Feb '09

06 Feb '09

Jim White wrote: > I have a loop that spends 99+% of its time executing these two statements: > > mpz_addmul_ui(BB, B, a); > mpz_swap (B, BB); > > To my surprise, it runs 25% faster if I replace it with: > > if (a==1) mpz_add (BB, B, BB); > else mpz_addmul_ui(BB, B, a); > mpz_swap(B, BB); > > It's an RCF (regular continued-fraction) algorithm, so a=1 about 41.5% of the time. > > No big deal, but perhaps in future revs mpz_addmul_ui and its sister function might auto-detect this case and reroute the call? In the Parma Polyhedra Library, we found out that, for mpz_mul(), checking the source arguments against 0 gives rise to significant speedups in some algorithms. We thought (but we are not sure) that this is indeed a feature, and not a limitation: since only the client code can know about the frequence of special cases, let the client code take the decision. The alternative would be that GMP does a lot of checks that are possibly useless, perhaps duplicating work that has already been done by the client code. I am glad you raised the issue here and I hope Torbjorn will correct me if I misinterpreted the GMP's policies in this respect. All the best, Roberto -- Prof. Roberto Bagnara Computer Science Group Department of Mathematics, University of Parma, Italy http://www.cs.unipr.it/~bagnara/ mailto:bagnara@cs.unipr.it

7 16

PPL 0.10 testsuite failures on powerpc-linux
by Janis Johnson 02 Feb '09

02 Feb '09

I'm trying to build the Parma Polyhedra Library on powerpc-linux so I can use it for Graphite support in GCC. I've tried a lot of things and my builds normally succeed but I get lots of testsuite failures: in Box, 60 tests with -DBOX_INSTANCES=db_r_oc in Octagonal_Shape, 58 tests with -DOCTAGONAL_SHAPE_INSTANCE=double in BD_Shape, 59 tests with -DBD_SHAPE_INSTANCE=double Most of the failures in the double tests are due to infinite values for the second argument to less-than functions, which end up at __gmp_binary_less::eval which calls __gmp_binary_less::eval which invokes DOUBLE_NAN_INF_ACTION which calls __gmp_invalid_operation for the infinity. It looks as if the infinity should be detected and handled earlier by functions in ppl-0.10/src/check-ext.inlines.hh. I'll attach a build script for one of the various builds I've done that get these errors. In this case I start with GCC 4.3.2, build GMP 4.2.4 as only static libraries, and then build PPL as only static libraries. I get the same testsuite failures when starting with GCC mainline and building GMP and PPL as shared libaries, but run into other problems with earlier versions of GCC or the default libgmp on my test system. Am I doing something obviously wrong? If not, can you offer any suggestions on how to find out what's going wrong, short of learning to understand the PPL and GMP sources? I tried using GMP 4.1.4, which doesn't use DOUBLE_NAN_INF_ACTION, but ran into various PPL configure problems. One is that some configure checks end up using /usr/include/gmpxx.h instead of the one specified for --with-libgmp, and another is that the options to link libgmpxx end up including libgcc_s.so and libm.so even though I'm trying to build and use only static libraries. If there are known issues with those configure issues I'd like to know about them as well. Janis Johnson IBM Linux Technology Center

2 4

ppl/interfaces/OCaml Makefile.am
by Enea Zaffanella 30 Jan '09

30 Jan '09

CVSROOT: /cvs/ppl Module name: ppl Changes by: zaffanella(a)cs.unipr.it 2009-01-30 12:24:28 Modified files: interfaces/OCaml: Makefile.am Log message: The parallel build problem should now be solved. Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/interfaces/OCaml/Makefile.am.…

1 0

ppl/tests/Box Makefile.am simplifyusingcontext1.cc
by Enea Zaffanella 28 Jan '09

28 Jan '09

CVSROOT: /cvs/ppl Module name: ppl Changes by: zaffanella(a)cs.unipr.it 2009-01-28 14:00:27 Modified files: tests/Box : Makefile.am Added files: tests/Box : simplifyusingcontext1.cc Log message: Added test file for method Box::simplify_using_cntext_assign(). Tests test01() and test02() marked as failing, since they depend on method Interval::empty_intersection_assign(), currently provided with a dummy implementation. Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/tests/Box/simplifyusingcontex… http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/tests/Box/Makefile.am.diff?cv…

1 0

ppl/src Box.templates.hh
by Enea Zaffanella 28 Jan '09

28 Jan '09

CVSROOT: /cvs/ppl Module name: ppl Changes by: zaffanella(a)cs.unipr.it 2009-01-28 13:56:52 Modified files: src : Box.templates.hh Log message: Drafted an implementation for Box::simplify_using_context_assign(). Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/src/Box.templates.hh.diff?cvs…

1 0

ppl/src Interval.templates.hh
by Enea Zaffanella 28 Jan '09

28 Jan '09

CVSROOT: /cvs/ppl Module name: ppl Changes by: zaffanella(a)cs.unipr.it 2009-01-28 13:56:10 Modified files: src : Interval.templates.hh Log message: Added dummy implementation for method empty_intersection_assign(); corrected partial implementation of method simplify_using_context_assign(). Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/src/Interval.templates.hh.dif…

1 0

ppl/src Interval.defs.hh
by Enea Zaffanella 28 Jan '09

28 Jan '09

CVSROOT: /cvs/ppl Module name: ppl Changes by: zaffanella(a)cs.unipr.it 2009-01-28 10:11:38 Modified files: src : Interval.defs.hh Log message: Added specification for new Interval method template <typename From> typename Enable_If<Is_Interval<From>::value, void>::type empty_intersection_assign(const From& y); to be used in the implementation of Box::simplify_using_context_assign. Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/src/Interval.defs.hh.diff?cvs…

1 0

AC_C_BIGENDIAN does not work well with AX_PREFIX_CONFIG_H
by Roberto Bagnara 27 Jan '09

27 Jan '09

Hi there, we have been bitten by the following problem. Autoconf implements AC_C_BIGENDIAN so that it defines WORDS_BIGENDIAN this way: /* Define WORDS_BIGENDIAN to 1 if your processor stores words with the most significant byte first (like Motorola and SPARC, unlike Intel). */ #if defined AC_APPLE_UNIVERSAL_BUILD # if defined __BIG_ENDIAN__ # define WORDS_BIGENDIAN 1 # endif #else # ifndef WORDS_BIGENDIAN /* # undef WORDS_BIGENDIAN */ # endif #endif The spaces between `#' and `define' completely spoil AX_PREFIX_CONFIG_H (which otherwise works very well on everything else). Since 'PPL' is our prefix, we want the symbol PPL_WORDS_BIGENDIAN to be defined (and not WORDS_BIGENDIAN). Now, I don't believe this is a bug of AX_PREFIX_CONFIG_H. In fact we use that feature: by leaving spaces between the `#' and the macro name we "protect" some macros from renaming. And, of course, it is not a bug of Autoconf, since that spaces are normally immaterial. Still we have a problem. And no workaround because the obvious one does not work (i.e., it does not undefine WORDS_BIGENDIAN): AH_BOTTOM([ #ifdef WORDS_BIGENDIAN # define PPL_WORDS_BIGENDIAN # undef WORDS_BIGENDIAN #endif ]) Perhaps this (commenting out all #undef's) is a limitation of AX_PREFIX_CONFIG_H. Any idea how we could solve this problem? Thanks, Roberto -- Prof. Roberto Bagnara Computer Science Group Department of Mathematics, University of Parma, Italy http://www.cs.unipr.it/~bagnara/ mailto:bagnara@cs.unipr.it

1 0

ppl configure.ac
by Roberto Bagnara 27 Jan '09

27 Jan '09

CVSROOT: /cvs/ppl Module name: ppl Changes by: bagnara(a)cs.unipr.it 2009-01-27 21:25:28 Modified files: . : configure.ac Log message: Partially work around a bug of AX_PREFIX_CONFIG_H whereby WORDS_BIGENDIAN is not prefixed. Version number bumped. Patches: http://www.cs.unipr.it/cgi-bin/cvsweb.cgi/ppl/configure.ac.diff?cvsroot=ppl…

1 0