FLINT 3.4

FLINT 3.4.0 (2025-11-25)#

Contributors: Albin Ahlbäck (AA), Andrii Yanovets (AY), Claus Fieker (CF), David Lowry-Duda (DLD), Dimitri Lesnoff (DL), Daniel Schultz (DS), Doug Torrance (DT), David Walker (DW), Éric Schost (ES), Fredrik Johansson (FJ), Guillem Blanco (GB), Joel Dahne (JD), Jean Kieffer (JK), kac3pro (K3), Lars Göttgens (LG), Mathieu Gouttenoire (MG), Mael Hostettler (MH), Marc Mezzarobba (MM), Pascal Molin (PM), Ricardo Buring (RB), Remi Prebet (RP), user202729 (U2), Vincent Neiger (VN).

• Features

  • Add gr_series as a public module (contains both the gr_series and gr_series_mod types) (FJ).
  • Add directed rounding support for nfloat (FJ).
  • Matrix permanents (gr_mat_permanent, fmpz_mat_permanent and fmpq_mat_permanent) (FJ).
  • Add the fmpz_mod_mpoly_q module for multivariate rational functions over a prime field (AY).
  • Gröbner basis computation for fmpz_mod_mpoly (AY).
  • Add nmod_poly_minimal_irreducible and fq_nmod_ctx_init_minimal_weight_ui for generating irreducible polynomials with minimal weight for any word-size prime p and degree n, and use these by default to construct a finite extension field when a Conway polynomial is not available (FJ).
  • Fast multipoint evaluation and interpolation on geometrically spaced points for nmod_poly (ES, VN, MH).
  • Add a debug context for generics (gr_ctx_init_debug) (FJ).
  • Add _nmod_vec_invert (VN).
  • Add gr_ctx_ngens and gr_ctx_gen_name (MM).
  • Preconditioned modular multiplication for nmod_poly (FJ).
  • Add uniformly random functions for various modular types (including _nmod_vec_rand, nmod_mat_rand, fmpz_mod_mat_rand) (DL).
  • Add Kronecker substitution multiplication for gr_poly (FJ).
  • Add gr_poly_mullow_complex_reorder (FJ).
  • Add _gr_mpoly_normalise to remove zero coefficients (RB).
  • Add gr_mpoly_derivative and gr_mpoly_integral (RB).
  • Add gr_derivative_gen, with implementations for polynomial rings (RB).
  • Add gr_poly_mullow_classical as a standalone function (FJ).
  • Support truth_t in flint_printf (FJ).
  • Add delta, eta parameters to arb_mat_spd_lll_reduce and arb_mat_spd_is_lll_reduced (JK).
  • QR and LQ factorization of matrices (gr_mat_qr, gr_mat_lq) (FJ).
  • Add gr_mat_is_orthogonal (FJ).
  • Add gr_mat_randtest_orthogonal (FJ).
  • Add fmpq_poly_discriminant (K3).
  • Add gr_mat_is_row_lll_reduced_naive (FJ).
  • Add generic Newton solver for series solutions of linear systems of ODEs with rational function coefficients (gr_mat_gr_poly_solve_lode_newton) (RB).
  • Add nfloat_complex_sqrt and nfloat_complex_rsqrt (FJ).
  • Generate power series contexts in gr_ctx_init_random (FJ).
  • Functions for converting a gr_poly between the standard monomial basis and a Newton basis (FJ).
  • Polynomial interpolation for gr_poly (Newton basis and asymptotically fast) (FJ).
  • Add gr_poly_product_roots (FJ).
  • Add fmpq_poly_interpolate_fmpq_vec (RP).
  • Add error handling to fmpz_poly_interpolate_fmpz_vec (FJ).
  • Add fmpz_poly_interpolate_exact for faster exact interpolation without error handling (FJ).
  • Add gr_ctx_is_rational_vector_space, gr_ctx_is_real_vector_space, gr_ctx_is_complex_vector_space (FJ).

• Bug fixes

  • Fix erroneous bounds in _fmpz_mat_bound_ovals_of_cassini, which would lead fmpz_mat_minpoly to compute wrong results for some matrices (CF).
  • Fix fmpq_poly_resultant_modular_div, fmpz_poly_resultant_modular_div and nf_elem_norm_div with non-monic polynomials (CF).
  • Fix initial value selection in acb_poly_find_roots preventing convergence for certain polynomials with coefficients in geometric progression (FJ).
  • Fix gr_re and gr_conj for nfloat_complex (FJ).
  • Fix memory corruption bug in nmod_mpoly_divides_heap_threaded (FJ, DS).
  • Fix and test edge cases in fmpz_set_str (DW).
  • Enable accidently disabled gr_poly test (MM).
  • Consistently throw FLINT_DIVZERO when trying to divide by zero (LG).
  • Fix bug in gr_poly_div_series_basecase: this would previously return GR_DOMAIN in some cases where the quotient actually exists (FJ).
  • Correct semantics for generic series: operations on gr_series should no longer create enclosures for elements that don't exist (FJ).
  • Fix gr_poly_shift_equivalent in positive characteristic (MM).
  • Fix a memory leak in nmod_mpoly_gcd and improve test coverage of this function (FJ).

• Performance

  • Use nfloat, arf and arb arithmetic instead of mpf, mpfr and fmpq in fmpz_lll, greatly speeding up multiprecision LLL (FJ).
  • Use SQUFOF in fmpz_factor to speed up factorisations of cofactors in the 65-80 bit range (FJ).
  • Reduce multiplication depth in _gr_poly_compose_axnc (U2).
  • Improve precision in arb_set_str, making it more likely to round to nearest (DW).
  • Minor speedup for fmpz_poly_compose (U2).
  • Speed up n_is_prime by replacing the BPSW test with smarter strong probable prime tests and other optimisations (FJ).
  • Speed up n_factor_one_line and n_factor_SQUFOF by improving the square root computation (FJ).
  • Speed up n_factor by optimising trial division, doing earlier primality testing, and using the one line factor algorithm for larger bit sizes (FJ).
  • When possible speed up nmod_poly multiplication via fft_small by a factor two for tiny moduli by packing a linear polynomial into each coefficient (FJ).
  • Add AVX2 version of nmod_vec_dot_half (VN).
  • Improve speed and accuracy for arb_min and arb_max with non-overlapping input (JD).
  • Speed up _fmpz_vec_scalar_divexact_si, _fmpz_vec_scalar_divexact_ui and small-divisor _fmpz_vec_scalar_divexact_fmpz (FJ).
  • Make multivariate polynomial multiplication via nested gr_poly over various ground types asymptotically fast (FJ).
  • Speed up _nmod_poly_divrem_newton_n_preinv for sparse moduli by handling these specially (FJ).
  • Incorporate trial division in nmod_poly_is_irreducible when the prime is small (FJ).
  • Speed up nmod_poly modular composition by using rectangular splitting instead of Horner for the third step of Brent-Kung (FJ).
  • Faster preconditioned multiplication and polynomial GCD for mpn_mod (FJ).
  • Improve fmpq_poly_interpolate_fmpz_vec by incorporating an asymptotically fast multimodular algorithm and other improvements (RP).
  • Improve fmpz_poly_interpolate_fmpz_vec by incorporating an asymptotically fast multimodular algorithm and other improvements (FJ).
  • Make various gr_poly functions more robust or performant when dealing with inexact representations of zero (FJ).

• Build system, general maintenance

  • Fix issues with profiling code on Clang (AA).
  • Mark many internal functions as static (AA).
  • Compile with -Wmissing-prototypes by default (AA).
  • Fix many compiler warnings (AA).
  • Fix acb_theta_all and acb_theta_one disappearing from the shared library (AA).
  • Add V Makefile variable for verbose builds (DT).
  • Add link-time optimization configure option (MG).
  • Fix situations where the compiler runs out of registers for inline assembly on x86 (AA).
  • Fix static linkage of d_randtest (AA).
  • Remove the obsolete mpfr_vec and mpfr_mat modules (FJ).
  • Refactor fmpz_lll to reduce code size and improve clarity (FJ).
  • Refactor power series code (FJ).
  • Add FLINT_NORETURN to function pointer argument (AA).
  • Test msolve in CI (VN).
  • Corrections and improvements to the documentation (AA, DL, DLD, GB, LG, U2).

• Example programs

  • Add example program to compute pi using AGM iteration (FJ).
  • Add example program for computing coefficients of modular forms (PM).

@misc{flint-doc-history-flint3-v3-4,
  author       = {{The FLINT team}},
  title        = {{FLINT 3.4}},
  howpublished = {FLINT: Fast Library for Number Theory, documentation, version 3.7.0-dev},
  year         = {2026},
  url          = {https://flintlib.org/doc/history/flint3/v3-4.html},
  note         = {Generated 2026-10-20}
}