1 Changes between MP version 3.1 and 3.1.1
3 * Bug fixes for division (rare), mpf_get_str, FFT, and miscellaneous minor
6 Changes between MP version 3.0 and 3.1
9 * Improved `make check' running more tests.
10 * Tuned algorithm cutoff points for many machines. This will improve speed for
11 a lot of operations, in some cases by a large amount.
12 * Major speed improvments: Alpha 21264.
13 * Some speed improvments: Cray vector computers, AMD K6 and Athlon, Intel P5
14 and Pentium Pro/II/III.
15 * The mpf_get_prec function now works as it did in GMP 2.
16 * New utilities for auto-tuning and speed measuring.
17 * Multiplication now optionally uses FFT for very large operands. (To enable
18 it, pass --enable-fft to configure.)
19 * Support for new systems: Solaris running on x86, FreeBSD 5, HP-UX 11, Cray
20 vector computers, Rhapsody, Nextstep/Openstep, MacOS.
21 * Support for shared libraries on 32-bit HPPA.
22 * New integer functions: mpz_mul_si, mpz_odd_p, mpz_even_p.
23 * New Kronecker symbol functions: mpz_kronecker_si, mpz_kronecker_ui,
24 mpz_si_kronecker, mpz_ui_kronecker.
25 * New rational functions: mpq_out_str, mpq_swap.
26 * New float functions: mpf_swap.
27 * New mpn functions: mpn_divexact_by3c, mpn_tdiv_qr.
28 * New EXPERIMENTAL function layer for accurate floating-point arithmetic, mpfr.
29 To try it, pass --enable-mpfr to configure. See the mpfr subdirectory for
30 more information; it is not documented in the main GMP manual.
32 Changes between MP version 3.0 and 3.0.1
34 * Memory leaks in gmp_randinit and mpz_probab_prime_p fixed.
35 * Documentation for gmp_randinit fixed. Misc documentation errors fixed.
37 Changes between MP version 2.0 and 3.0
39 * Source level compatibility with past releases (except mpn_gcd).
41 * Much improved speed thanks to both host independent and host dependent
43 * Switch to autoconf/automake/libtool.
44 * Support for building libgmp as a shared library.
45 * Multiplication and squaring using 3-way Toom-Cook.
46 * Division using the Burnikel-Ziegler method.
47 * New functions computing binomial coefficients: mpz_bin_ui, mpz_bin_uiui.
48 * New function computing Fibonacci numbers: mpz_fib_ui.
49 * New random number generators: mpf_urandomb, mpz_rrandomb, mpz_urandomb,
50 mpz_urandomm, gmp_randclear, gmp_randinit, gmp_randinit_lc_2exp, gmp_randseed,
52 * New function for quickly extracting limbs: mpz_getlimbn.
53 * New functions performing integer size tests: mpz_fits_sint_p,
54 mpz_fits_slong_p, mpz_fits_sshort_p, mpz_fits_uint_p, mpz_fits_ulong_p,
56 * New mpf functions: mpf_ceil, mpf_floor, mpf_pow_ui, mpf_trunc.
57 * New mpq function: mpq_set_d.
58 * New mpz functions: mpz_addmul_ui, mpz_cmpabs, mpz_cmpabs_ui, mpz_lcm,
59 mpz_nextprime, mpz_perfect_power_p, mpz_remove, mpz_root, mpz_swap,
60 mpz_tdiv_ui, mpz_tstbit, mpz_xor.
61 * New mpn function: mpn_divexact_by3.
62 * New CPU support: DEC Alpha 21264, AMD K6 and Athlon, HPPA 2.0 and 64,
63 Intel Pentium Pro and Pentium-II/III, Sparc 64, PowerPC 64.
64 * Almost 10 times faster mpz_invert and mpn_gcdext.
65 * The interface of mpn_gcd has changed.
66 * Better support for MIPS R4x000 and R5000 under Irix 6.
67 * Improved support for SPARCv8 and SPARCv9 processors.
69 Changes between MP version 2.0 and 2.0.2
73 Changes between MP version 1.3.2 and 2.0
75 * Division routines in the mpz class have changed. There are three classes of
76 functions, that rounds the quotient to -infinity, 0, and +infinity,
77 respectively. The first class of functions have names that begin with
78 mpz_fdiv (f is short for floor), the second class' names begin with mpz_tdiv
79 (t is short for trunc), and the third class' names begin with mpz_cdiv (c is
82 The old division routines beginning with mpz_m are similar to the new
83 mpz_fdiv, with the exception that some of the new functions return useful
86 The old function names can still be used. All the old functions names will
87 now do floor division, not trunc division as some of them used to. This was
88 changed to make the functions more compatible with common mathematical
91 The mpz_mod and mpz_mod_ui functions now compute the mathematical mod
92 function. I.e., the sign of the 2nd argument is ignored.
94 * The mpq assignment functions do not canonicalize their results. A new
95 function, mpq_canonicalize must be called by the user if the result is not
96 known to be canonical.
97 * The mpn functions are now documented. These functions are intended for
98 very time critical applications, or applications that need full control over
99 memory allocation. Note that the mpn interface is irregular and hard to
101 * New functions for arbitrary precision floating point arithmetic. Names
102 begin with `mpf_'. Associated type mpf_t.
103 * New and improved mpz functions, including much faster GCD, fast exact
104 division (mpz_divexact), bit scan (mpz_scan0 and mpz_scan1), and number
105 theoretical functions like Jacobi (mpz_jacobi) and multiplicative inverse
107 * New variable types (mpz_t and mpq_t) are available that makes syntax of
108 mpz and mpq calls nicer (no need for & before variables). The MP_INT and
109 MP_RAT types are still available for compatibility.
110 * Uses GNU configure. This makes it possible to choose target architecture
111 and CPU variant, and to compile into a separate object directory.
112 * Carefully optimized assembly for important inner loops. Support for DEC
113 Alpha, Amd 29000, HPPA 1.0 and 1.1, Intel Pentium and generic x86, Intel
114 i960, Motorola MC68000, MC68020, MC88100, and MC88110, Motorola/IBM
115 PowerPC, National NS32000, IBM POWER, MIPS R3000, R4000, SPARCv7,
116 SuperSPARC, generic SPARCv8, and DEC VAX. Some support also for ARM,
117 Clipper, IBM ROMP (RT), and Pyramid AP/XP.
118 * Faster. Thanks to the assembler code, new algorithms, and general tuning.
119 In particular, the speed on machines without GCC is improved.
120 * Support for machines without alloca.
121 * Now under the LGPL.
123 INCOMPATIBILITIES BETWEEN GMP 1 AND GMP 2
125 * mpq assignment functions do not canonicalize their results.
126 * mpz division functions round differently.
127 * mpz mod functions now really compute mod.
128 * mpz_powm and mpz_powm_ui now really use mod for reduction.