./math/sollya, Environment and library for safe floating-point code development

[ CVSweb ] [ Homepage ] [ RSS ] [ Required by ] [ Add to tracker ]


Branch: CURRENT, Version: 8.0nb3, Package name: sollya-8.0nb3, Maintainer: pkgsrc-users

Sollya is both a tool environment and a library for safe floating-point
code development. It is particularily targeted to the automatized
implementation of mathematical floating-point libraries (libm). Amongst
other features, it offers a certified infinity (supremum) norm and a
fast Remez algorithm.


Required to run:
[textproc/libxml2] [math/mpfr] [devel/gmp] [math/fplll] [math/mpfi]

Required to build:
[pkgtools/cwrappers]

Master sites:

Filesize: 4321.769 KB

Version history: (Expand)


CVS history: (Expand)


   2023-11-08 14:21:43 by Thomas Klausner | Files touched by this commit (2377)
Log message:
*: recursive bump for icu 74.1
   2023-04-19 10:12:01 by Adam Ciarcinski | Files touched by this commit (2359) | Package updated
Log message:
revbump after textproc/icu update
   2023-01-02 15:07:51 by Havard Eidnes | Files touched by this commit (1)
Log message:
math/sollya: add use of atomic64.mk, as that's required.
   2022-11-23 17:21:30 by Adam Ciarcinski | Files touched by this commit (1878) | Package updated
Log message:
massive revision bump after textproc/icu update
   2022-10-29 01:49:43 by Alexander Nasonov | Files touched by this commit (6)
Log message:
Update math/sollya to version 8.0.

Changes from version 7.0 to 8.0:
================================
Changes of syntax and output:
  * Hexadecimal constants ending on '+' or '-' (e.g., 0x123p-) now
    provoke a syntax error (as it always should have).
  * The match construct on error now matches (it used to return
    error).
  * In certain cases, procedures could not be applied to an empty set
    of arguments because a syntax error occurred before. This is now
    possible.
  * The meaning and use of infinities and signed zeros in intervals
    has been changed (see documentation for details).
  * Input and output of decimal constants is now fully exact in
    rationalmode. In all other instances, a warning message is now
    produced when an internal constant gets printed in decimal and a
    rounding occurs.

Critical bug fixes:
  * Making Sollya compile again with recent versions of fplll (at
    least on the Linux side).

Improvements in internal handling:
  * Improved performance on manipulation of polynomials.

Major bug fixes:
  * Patched a bug in interval erfc.
  * Patched a bug preventing match to work on unary minus.

Major new functionalities:
  * The interpolate command allows for polynomial interpolation with
    and without error (see documentation for details).
  * The bezout command allows for computation of a gcd along with
    matching cofactors (see documentation for details).

Minor new functionalities:
  * The externaldata command now allows generic external library
    symbols to be bound to Sollya objects (see documentation for
    details).
  * Improvements on use of the L'Hopital rule for function
    evaluations.
  * When started as an interactive tool or as a library, Sollya now
    binds the executable's 0th argument to the Sollya identifier
    __program_name.
  * In situations when the user wishes to provide the minimax with
    real coefficients to fpminimax, it is now possible to give a list
    of coefficients instead of the expression of the real minimax (see
    documentation for details).
  * The free mathematical variable can now be unbound with the rename
    command (see documentation for details).
  * The bind command now allows for procedures with an arbitrary
    number of arguments (see documentation for details).
  * The char *sollya_lib_get_help_text(char *topic) function now
    allows for access to the help text of a certain topic.

Others:
  * In certain circumstances, the printing of a list may have had
    quadratic complexity in the past. This has been corrected.
  * If the compiler supports the -flto flag; this option is disabled
    by default though. To enable LTO compilation, use
    --enable-lto-compilation. On certain systems the -flto flag is
    supported by the compiler but LTO compilation fails. In this case,
    it is recommended not to enable LTO compilation.
  * dirtyfindzeros used to include in the list of zeros any point
    where the expression given as input evaluated to NaN. Such points
    are now included only when there is a real suspicion that it could
    be a true zero of the expression.
  * externalprocedure, externalfunction and externaldata entries can
    have cross-dependencies for their symbols. The tool now makes sure
    all symbols are unloaded in the precise inverse order of their
    loading, to allow for such cross-dependencies.
  * chebyshevform command can now be used out of execute() statements.
  * remez command now tries to recognize the situation when the
    optimal error function is identically 0.
  * configure now allows for cross-compilation.

Deprecated functionalities:
  * Deprecated readxml and printxml.
  * Deprecated the proof generation feature of the infnorm command.
  * Deprecated end-elliptic lists ending with a number as a format
    list in fpminimax.

And many bugs fixed.
   2022-10-28 21:59:03 by Alexander Nasonov | Files touched by this commit (1)
Log message:
Update website.
   2022-04-18 21:12:27 by Adam Ciarcinski | Files touched by this commit (1798) | Package updated
Log message:
revbump for textproc/icu update
   2021-12-08 17:07:18 by Adam Ciarcinski | Files touched by this commit (3063)
Log message:
revbump for icu and libffi