GitLab

Thanks to Jan Engelhardt <jengelh@inai.de> for figuring out the API
of fplll-5.

* configure.ac: If fplll-4 is not found, check for fplll-5.
* src/lll.cpp (lll_fplll): Add a version working with fplll-5, and simplify
  the code for fplll-4.

This is necessary since terminal codes for colours may appear in the output.

* scripts/cmh-classpol.sh.in: Define GP_OPTS to unset the colours and use
  it everywhere gp is called.

32 bit machines may choke on addressing 8GB of stack, while smaller sizes
may not be enough on large examples.

This partially reverts commit 370d5ebe.

Reported by Jan Engelhardt <jengelh@inai.de>.

* scripts/cmh-classpol.sh.in: Drop fixing the gp stack size with "-s",
  and drop the "-f" parameter to invocations of gp.

* configure.ac: Replace gforge mailing list by Andreas's.

Increasing the initial stack size from 500k to 4M is enough for
the example [296, 346, 27265].

* src/lll.cpp (lll_pari): Increase the PARI stack.

* src/quadratic_field.c (cmh_halfgcd): Modify to also return the second
  component of the PARI ghalfgcd function, and only one column of the
  first component.
  (quadratic_number_recognize): Adapt.

This is a trivial adaptation: The function is now called ghalfgcd, and
it returns a two-element vector with the 2x2 transformation matrix and
additionally the two consecutive remainders.

* src/quadratic_field.c (cmh_halfgcd): Update.

This is a follow-up to commit b15df9ff,
which could lead to mpz_mul_2exp being called with a negative exponent.

* src/quadratic_field.c (quadratic_number_recognize): Abort when the
  precision is lower than the exponent of the number.

* README: Add as symbolic link to README.md.

using a half-gcd; activated if an as yet unreleased version of PARI/GP
is available.

For [296, 346, 27265], the time for recognising coefficients drops from about
0.12s to 0.08s with FPLLL-4, and from 0.75s to 0.35s with PARI.

* src/quadratic_field.c (cmh_halfgcd): New function.
  (quadratic_number_recognize): Precondition the matrix.

* configure.ac: If --with-pari is given, modify not only the include, but
  also the library path.

* configure.ac: Change a message.

* configure.ac: Check explicitly for "fplll/matrix.h". Version 5 of fplll
  has moved this header, and we do not support its new api.
Signed-off-by: Andreas Enge <andreas.enge@inria.fr>

of 2^Xp from the last column. The time for [296, 346, 27265] decreases
by about a third. At the same time, drop the negligible factor s from the
scaled unit matrix and replace divisions by powers of 2 by right shifts;
this has no effect on the running time, while preserving correctness
of the result.

* src/quadratic_field.c (quadratic_number_recognize): Modify the matrix
  passed to LLL.

* examples/bench-theta.m, examples/theta.m, examples/thetatool.c:
  Remove files.

* .gitignore: Add config/test-driver.

* src/quadratic_field.h, src/quadratic_field.c (quadratic_number_set,
  quadratic_number_swap, quadratic_number_mul_z,
  quadratic_number_mul_z_normalize): Remove unused functions.
  (quadratic_number_normalize): Let the function work in place.
  (quadratic_number_recognize): Use macros (which do no longer clash with
  macros from PARI or FPLLL).
* src/macros.h (fget_z): New macro.

* src/recognize_quadratic.cpp: Rename to...
* src/lll.cpp: ...this.
  (recognize_quadratic): Move functionality...
* src/quadratic_field.c (quadratic_number_recognize): ...here.
* src/recognize_quadratic.h: Rename to...
* src/lll.h: ...this. Export cmh_lll.
* src/Makefile.am: Update the changed file names.
* src/quadratic_field.c: Update the changed name of the included file.

* src/recognize_quadratic.cpp (recognize_quadratic_pari,
  recognize_quadratic_fplll): Remove functions.
  (lll_pari, lll_fplll): New functions encapsulating the LLL algorithm
  on matrices with mpz_t coefficients.
  (recognize_quadratic): Put all the logic here.

* src/macros.h (min, max): Rename to CMH_MIN and CMH_MAX.
* src/misc.c, src/newton.c: Replace the macro calls.

* configure.ac: Update version number.

* src/macros.h: Replace "GMP_RNDN" by "MPFR_RNDN". Add fset_z.

* src/macros.h: Remove tabs and shuffle macros.

The commit yields a wrong polynomial for [D, A, B] = [157, 3, 13].

This reverts commit e77b5aca.

* .gitignore: Add files created by the test.

* tests/Makefile.am, tests/test.sh, tests/example.pol: New files.
* Makefile.am (SUBDIRS): Add tests.
* configure.ac (AC_CONFIG_FILES): Add tests/Makefile.

of them with the currently known denominator instead of a fixed factor
depending only on the number of the polynomial.

* src/cm2.c (reconstruction_prepare_for_coeff): Multiply coefficient by
  state->denom_lcm instead of state->premult.
  (recognise_quadratic_coefficients): Set state->denom_lcm instead of
  state->premult.
* rsrc/cm2.h (reconstruction_status_s): Drop the premult field.

GP scripts.

* configure.ac: Increase minimal PARI/GP version to 2.9.
* scripts/shimura.gp: Drop Vecrev_compat function and use Vecrec instead.

* .gitignore: Add src/cm2-mpi.

* src/recognize_quadratic.cpp (recognize_quadratic_pari): Define the function
  only when it is used.

* configure.ac: When FPLLL is found, set HAVE_FPLLL in config.h; otherwise
  just emit a warning instead of an error.
* src/recognize_quadratic.cpp (recognize_quadratic): Call
  recognize_quadratic_pari instead of recognize_quadratic_fplll if FPLLL
  is not available.

* src/recognize_quadratic.cpp (recognize_quadratic_pari): New function.
  (recognize_quadratic_fplll): Rename from...
  (recognize_quadratic): ...this. Rewrite this function to call FPLLL,
  since PARI is much slower.
* src/recognize_quadratic.h: Include parsepari.h.
* src/Makefile.am: Link with libparsepari.

* configure.ac: Modify the check for PARI to not only check for the header
  file, but to also try to link a program, and add the linker flag.

* src/parsepari.c (mpz_set_GEN): New function.
  (mpz_get_GEN): Renamed from mpz_get_Z and made public.
* src/parsepari.h: Declare the two public functions.

* configure.ac: Drop AM_PROG_CC_C_O, obsolete since automake 1.14.

* src/cmh.h: Correct endianness in the theta function labelling.

* config/checkversion.sh: Remove file.
* configure.ac: Substitute AC_PACKAGE_VERSION for CMH_LONG_VERSION.
* src/version.c.in: Drop commented out line.

* .gitignore: New file.