xref: /netbsd-src/external/lgpl3/gmp/dist/NEWS (revision f14316bcbc544b96a93e884bc5c2b15fd60e22ae)
1Copyright 1996, 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008,
22009, 2010, 2011, 2012, 2013 Free Software Foundation, Inc.
3
4Verbatim copying and distribution of this entire article is permitted in any
5medium, provided this notice is preserved.
6
7
8Changes between GMP version 5.1.2 and 5.1.3
9
10  BUGS FIXED
11  * The internal functions mpn_sbpi1_div_qr_sec mpn_sbpi1_div_r_sec could
12    compute garbage with a low probability.  They are now rewritten, and the
13    test code has been improved.
14
15  * A bug in the ia64 implementation of mpn_divrem_2, clobbering some
16    callee-save registers, has been fixed. This is an internal
17    function, with the bug manifesting itself as miscomputation in,
18    e.g., mpn_sqrtrem.
19
20  * The documentation now correctly says 'const' for input arguments.
21
22  SPEEDUPS
23  * None.
24
25  FEATURES
26  * None.
27
28  MISC
29  * None.
30
31Changes between GMP version 5.1.1 and 5.1.2
32
33  BUGS FIXED
34  * A bug in mpz_powm_ui triggered by base arguments of at least 15000 decimal
35    digits or mod arguments of at least 7500 decimal digits has been fixed.
36
37  * A AMD Bulldozer specific bug affecting the 64-bit Windows ABI has been
38    fixed.  This bug was in a key function (mpn_mul_1) and made both Bulldozer
39    specific builds and fat builds run on Bulldozer completely non-functional.
40
41  SPEEDUPS
42  * None.
43
44  FEATURES
45  * None.
46
47  MISC
48  * Fixes and generalisations to the test suite.
49
50  * Minor portability enhancements.
51
52
53Changes between GMP version 5.1.0 and 5.1.1
54
55  BUGS FIXED
56  * On Windows 64-bit, an error causing link errors about
57    __gmp_binvert_limb_table has been fixed.
58
59  * Aarch64 alias ARM64 support now works.
60
61  * A possible buffer overrun in mpz_ior has been fixed.
62
63  * A rare sign flip in mpz_remove has been fixed.
64
65  * A bug causing problems with mpf numbers with absolute value >= 2^31 has
66    been fixed.
67
68  * Several bugs in mini-gmp have been fixed.
69
70  * A bug caused by automake, related to the 'distcheck' target, has been fixed
71    by upgrading the automake used for GMP release engineering.
72
73  SPEEDUPS
74  * None.
75
76  FEATURES
77  * Preliminary support for the x32 ABI under x86-64.
78
79  MISC
80  * The mini-gmp testsuite now tests the entire set of functions.
81
82  * Various improvements of the GMP testsuite.
83
84
85Changes between GMP version 5.0.* and 5.1.0
86
87  BUGS FIXED
88  * When reading a C++ number (like mpz_class) in an istream reaches the end
89    of the stream, the eofbit is now set.
90
91  * The result sign of mpz_rootrem's remainder is now always correct.
92
93  * The mpz_remove function now handles negative divisors.
94
95  * Contains all fixes from release 5.0.5.
96
97  SPEEDUPS
98  * The n-factorial and n-over-k functions have been reimplemented for great
99    speedups for small and large operands.
100
101  * New subquadratic algorithm for the Kronecker/Jacobi/Legendre symbol.
102
103  * Major speedup for ARM, in particular ARM Cortex-A9 and A15, thanks to broad
104    assembly support.
105
106  * Significant speedup for POWER6 and POWER7 thanks to improved assembly.
107
108  * The performance under M$ Windows' 64-bit ABI has been greatly improved
109    thanks to complete assembly support.
110
111  * Minor speed improvements of many functions and for many platforms.
112
113  FEATURES
114  * Many new CPUs recognised.
115
116  * New functions for multi-factorials, and primorial: mpz_2fac_ui,
117    mpz_mfac_uiui and mpz_primorial_ui.
118
119  * The mpz_powm_sec function now uses side-channel silent division for
120    converting into Montgomery residues.
121
122  * The fat binary mechanism is now more robust in its CPU recognition.
123
124  MISC
125  * Inclusion of assembly code is now controlled by the configure options
126    --enable-assembly and --disable-assembly.  The "none" CPU target is gone.
127
128  * In C++, the conversions mpq_class->mpz_class, mpf_class->mpz_class and
129    mpf_class->mpq_class are now explicit.
130
131  * Includes "mini-gmp", a small, portable, but less efficient, implementation
132    of a subset of GMP's mpn and mpz interfaces. Used in GMP bootstrap, but it
133    can also be bundled with applications as a fallback when the real GMP
134    library is unavailable.
135
136  * The ABIs under AIX are no longer called aix32 and aix64, but mode64 and 32.
137    This is more consistent with other powerpc systems.
138
139  * The coverage of the testsuite has been improved, using the lcov tool.  See
140    also http://gmplib.org/devel/lcov/.
141
142  * It is now possible to compile GMP using a C++ compiler.
143
144  * K&R C compilers are no longer supported.
145
146  * The BSD MP compatibility functions have been removed.
147
148
149Changes between GMP version 5.0.4 and 5.0.5
150
151  BUGS FIXED
152  * A bug causing AMD 11h processors to be treated like AMD 10h has been fixed.
153    The 11h processors do not correctly handle all 10h (aka K10) instructions,
154    and GMP's use of these instructions results in major miscomputations (not
155    as one would have hoped CPU traps of some 'illegal instruction' sort).
156
157  * A bug affecting recent Intel Sandy Bridge CPUs resulting in configuration
158    failures has been fixed.
159
160  SPEEDUPS
161  * None.
162
163  FEATURES
164  * A couple of tests added to the self-check suite.
165
166  MISC
167  * None.
168
169
170Changes between GMP version 5.0.3 and 5.0.4
171
172  BUGS FIXED
173  * Thresholds in mpn_powm_sec for both fat and non-fat builds are now used
174    safely, plugging a one-word buffer overrun introduced in the 5.0.3 release
175    (for non-fat) and a multi-word buffer overrun that existed since 5.0 (for
176    fat).  (We have not been able to provoke malign stack smashing in any of
177    the ~100 configurations explored by the GMP nightly builds, but the bug
178    should be assumed to be exploitable.)
179
180  * Two bugs in multiplication code causing incorrect computation with
181    extremely low probability have been fixed.
182
183  * A bug in the test suite causing buffer overruns during "make check",
184    sometimes leading to subsequent malloc crashes, has been fixed.
185
186  * Two bugs in the gcd code have been fixed.  They could lead to incorrect
187    results, but for uniformly distributed random operands, the likelihood for
188    that is infinitesimally small.  (There was also a third bug, but that was
189    an incorrect ASSERT, which furthermore was not enabled by default.)
190
191  * A bug affecting 32-bit PowerPC division has been fixed.  The bug caused
192    miscomputation for certain divisors in the range 2^32 ... 2^64-1 (about 1
193    in 2^30 of these).
194
195  SPEEDUPS
196  * None, except indirectly through recognition of new CPUs, and through better
197    tuning parameters.
198
199  FEATURES
200  * Some more tests added to the self-check suite.
201
202  * The AMD "Bulldozer" CPU is now recognised.
203
204  MISC
205  * None.
206
207
208Changes between GMP version 5.0.2 and 5.0.3
209
210  BUGS FIXED
211  * A few minor bugs related to portability fixed.
212
213  * A slight timing leak of the powm_sec functions have been sealed.  (This
214    leak could possibly be used to extract the most significant few bits of the
215    exponent.  "Few" here means at most 10.)
216
217  * The mpz_nextprime function now runs a safer number of pseudo-random prime
218    tests.
219
220  * A bug in division code possibly causing incorrect computation was fixed.
221
222  SPEEDUPS
223  * None, except indirectly through recognition of new CPUs, and through better
224    tuning parameters.
225
226  FEATURES
227  * New CPUs recognised.
228
229  * IBM S/390 are now supported in both 31/32-bit and 64-bit mode.  (We have
230    not been able to fully test this on any multilib machine, since IBM expired
231    our guest account a few days before our release.)
232
233  MISC
234  * None.
235
236
237Changes between GMP version 5.0.1 and 5.0.2
238
239  BUGS FIXED
240  * Many minor bugs related to portability fixed.
241
242  * The support for HPPA 2.0N now works, after an assembly bug fix.
243
244  * A test case type error has been fixed.  The symptom of this bug was
245    spurious 'make check' failures.
246
247  SPEEDUPS
248  * None, except indirectly through recognition of new CPUs.
249
250  FEATURES
251  * Fat builds are now supported for 64-bit x86 processors also under Darwin.
252
253  MISC
254  * None.
255
256
257Changes between GMP version 5.0.0 and 5.0.1
258
259  BUGS FIXED
260  * Fat builds fixed.
261
262  * Fixed crash for huge multiplies when old FFT_TABLE2 type of parameter
263    selection tables' sentinel was smaller than multiplied operands.
264
265  * The solib numbers now reflect the removal of the documented but preliminary
266    mpn_bdivmod function; we correctly flag incompatibility with GMP 4.3.  GMP
267    5.0.0 has this wrong, and should perhaps be uninstalled to avoid confusion.
268
269  SPEEDUPS
270  * Multiplication of large numbers has indirectly been sped up through better
271    FFT tuning and processor recognition.  Since many operations depend on
272    multiplication, there will be a general speedup.
273
274  FEATURES
275  * More Core i3, i5 an Core i7 processor models are recognised.
276
277  * Fixes and workarounds for Mac OS quirks should make this GMP version build
278    using many of the different versions of "Xcode".
279
280  MISC
281  * The amount of scratch memory needed for multiplication of huge numbers has
282    been reduced substantially (but is still larger than in GMP 4.3.)
283
284  * Likewise, the amount of scratch memory needed for division of large numbers
285    has been reduced substantially.
286
287  * The FFT tuning code of tune/tuneup.c has been completely rewritten, and
288    new, large FFT parameter selection tables are provided for many machines.
289
290  * Upgraded to the latest autoconf, automake, libtool.
291
292
293Changes between GMP version 4.3.X and 5.0.0
294
295  BUGS FIXED
296  * None (contains the same fixes as release 4.3.2).
297
298  SPEEDUPS
299  * Multiplication has been overhauled:
300    (1) Multiplication of larger same size operands has been improved with
301        the addition of two new Toom functions and a new internal function
302        mpn_mulmod_bnm1 (computing U * V mod (B^n-1), B being the word base.
303        This latter function is used for the largest products, waiting for a
304        better Schoenhage-Strassen U * V mod (B^n+1) implementation.
305    (2) Likewise for squaring.
306    (3) Multiplication of different size operands has been improved with the
307        addition of many new Toom function, and by selecting underlying
308        functions better from the main multiply functions.
309
310  * Division and mod have been overhauled:
311    (1) Plain "schoolbook" division is reimplemented using faster quotient
312        approximation.
313    (2) Division Q = N/D, R = N mod D where both the quotient and remainder
314        are needed now runs in time O(M(log(N))).  This is an improvement of
315        a factor log(log(N))
316    (3) Division where just the quotient is needed is now O(M(log(Q))) on
317        average.
318    (4) Modulo operations using Montgomery REDC form now take time O(M(n)).
319    (5) Exact division Q = N/D by means of mpz_divexact has been improved
320        for all sizes, and now runs in time O(M(log(N))).
321
322  * The function mpz_powm is now faster for all sizes.  Its complexity has
323    gone from O(M(n)log(n)m) to O(M(n)m) where n is the size of the modulo
324    argument and m is the size of the exponent.  It is also radically
325    faster for even modulus, since it now partially factors such modulus
326    and performs two smaller modexp operations, then uses CRT.
327
328  * The internal support for multiplication yielding just the lower n limbs
329    has been improved by using Mulders' algorithm.
330
331  * Computation of inverses, both plain 1/N and 1/N mod B^n have been
332    improved by using well-tuned Newton iterations, and wrap-around
333    multiplication using mpn_mulmod_bnm1.
334
335  * A new algorithm makes mpz_perfect_power_p asymptotically faster.
336
337  * The function mpz_remove uses a much faster algorithm, is better tuned,
338    and also benefits from the division improvements.
339
340  * Intel Atom and VIA Nano specific optimisations.
341
342  * Plus hundreds of smaller improvements and tweaks!
343
344  FEATURES
345  * New mpz function: mpz_powm_sec for side-channel quiet modexp
346    computations.
347
348  * New mpn functions: mpn_sqr, mpn_and_n, mpn_ior_n, mpn_xor_n, mpn_nand_n,
349    mpn_nior_n, mpn_xnor_n, mpn_andn_n, mpn_iorn_n, mpn_com, mpn_neg,
350    mpn_copyi, mpn_copyd, mpn_zero.
351
352  * The function mpn_tdiv_qr now allows certain argument overlap.
353
354  * Support for fat binaries for 64-bit x86 processors has been added.
355
356  * A new type, mp_bitcnt_t for bignum bit counts, has been introduced.
357
358  * Support for Windows64 through mingw64 has been added.
359
360  * The cofactors of mpz_gcdext and mpn_gcdext are now more strictly
361    normalised, returning to how GMP 4.2 worked.  (Note that also release
362    4.3.2 has this change.)
363
364  MISC
365  * The mpn_mul function should no longer be used for squaring,
366    instead use the new mpn_sqr.
367
368  * The algorithm selection has been improved, the number of thresholds have
369    more than doubled, and the tuning and use of existing thresholds have
370    been improved.
371
372  * The tune/speed program can measure many of new functions.
373
374  * The mpn_bdivmod function has been removed.  We do not consider this an
375    incompatible change, since the function was marked as preliminary.
376
377  * The testsuite has been enhanced in various ways.
378
379
380Changes between GMP version 4.3.1 and 4.3.2
381
382  Bugs:
383  * Fixed bug in mpf_eq.
384  * Fixed overflow issues in mpz_set_str, mpz_inp_str, mpf_set_str, and
385    mpf_get_str.
386  * Avoid unbounded stack allocation for unbalanced multiplication.
387  * Fixed bug in FFT multiplication.
388
389  Speedups:
390  * None, except that proper processor recognition helps affected processors.
391
392  Features:
393  * Recognise more "Core 2" processor variants.
394  * The cofactors of mpz_gcdext and mpn_gcdext are now more strictly
395    normalised, returning to how GMP 4.2 worked.
396
397
398Changes between GMP version 4.3.0 and 4.3.1
399
400  Bugs:
401  * Fixed bug in mpn_gcdext, affecting also mpz_gcdext and mpz_invert.
402    The bug could cause a cofactor to have a leading zero limb, which
403    could lead to crashes or miscomputation later on.
404  * Fixed some minor documentation issues.
405
406  Speedups:
407  * None.
408
409  Features:
410  * Workarounds for various issues with Mac OS X's build tools.
411  * Recognise more IBM "POWER" processor variants.
412
413
414Changes between GMP version 4.2.X and 4.3.0
415
416  Bugs:
417  * Fixed bug in mpz_perfect_power_p with recognition of negative perfect
418    powers that can be written both as an even and odd power.
419  * We might accidentally have added bugs since there is a large amount of
420    new code in this release.
421
422  Speedups:
423  * Vastly improved assembly code for x86-64 processors from AMD and Intel.
424  * Major improvements also for many other processor families, such as
425    Alpha, PowerPC, and Itanium.
426  * New sub-quadratic mpn_gcd and mpn_gcdext, as well as improved basecase
427    gcd code.
428  * The multiply FFT code has been slightly improved.
429  * Balanced multiplication now uses 4-way Toom in addition to schoolbook,
430    Karatsuba, 3-way Toom, and FFT.
431  * Unbalanced multiplication has been vastly improved.
432  * Improved schoolbook division by means of faster quotient approximation.
433  * Several new algorithms for division and mod by single limbs, giving
434    many-fold speedups.
435  * Improved nth root computations.
436  * The mpz_nextprime function uses sieving and is much faster.
437  * Countless minor tweaks.
438
439  Features:
440  * Updated support for fat binaries for x86_32 include current processors
441  * Lots of new mpn internal interfaces.  Some of them will become public
442    in a future GMP release.
443  * Support for the 32-bit ABI under x86-apple-darwin.
444  * x86 CPU recognition code should now default better for future
445    processors.
446  * The experimental nails feature does not work in this release, but
447    it might be re-enabled in the future.
448
449  Misc:
450  * The gmp_version variable now always contains three parts.  For this
451    release, it is "4.3.0".
452
453
454Changes between GMP version 4.2.3 and 4.2.4
455
456  Bugs:
457  * Fix bug with parsing exponent '+' sign in mpf.
458  * Fix an allocation bug in mpf_set_str, also affecting mpf_init_set_str, and
459    mpf_inp_str.
460
461  Speedups:
462  * None, except that proper processor recognition helps affected processors.
463
464  Features:
465  * Recognize new AMD processors.
466
467
468Changes between GMP version 4.2.2 and 4.2.3
469
470  Bugs:
471  * Fix x86 CPU recognition code to properly identify recent AMD and Intel
472    64-bit processors.
473  * The >> operator of the C++ wrapper gmpxx.h now does floor rounding, not
474    truncation.
475  * Inline semantics now follow the C99 standard, and works with recent GCC
476    releases.
477  * C++ bitwise logical operations work for more types.
478  * For C++, gmp.h now includes cstdio, improving compiler compatibility.
479  * Bases > 36 now work properly in mpf_set_str.
480
481  Speedups:
482  * None, except that proper processor recognition helps affected processors.
483
484  Features:
485  * The allocation functions now detect overflow of the mpz_t type.  This means
486    that overflow will now cause an abort, except when the allocation
487    computation itself overflows.  (Such overflow can probably only happen in
488    powering functions; we will detect powering overflow in the future.)
489
490
491Changes between GMP version 4.2.1 and 4.2.2
492
493  * License is now LGPL version 3.
494
495  Bugs:
496  * Shared library numbers corrected for libcxx.
497  * Fixed serious bug in gmpxx.h where a=a+b*c would generate garbage.
498    Note that this only affects C++ programs.
499  * Fix crash in mpz_set_d for arguments with large negative exponent.
500  * Fix 32-bit ABI bug with Itanium assembly for popcount and hamdist.
501  * Fix assembly syntax problem for powerpc-ibm-aix with AIX native assembler.
502  * Fix problems with x86 --enable-fat, where the compiler where told to
503    generate code for the build machine, not plain i386 code as it should.
504  * Improved recognition of powerpc systems wrt Altivec/VMX capability.
505  * Misc minor fixes, mainly workarounds for compiler/assembler bugs.
506
507  Speedups:
508  * "Core 2" and Pentium 4 processors, running in 64-bit mode will get a
509     slight boost as they are now specifically recognized.
510
511  Features:
512  * New support for x86_64-solaris
513  * New, rudimentary support for x86-apple-darwin and x86_64-apple-darwin.
514    (Please see http://gmplib.org/macos.html for more information.)
515
516
517Changes between GMP version 4.2 and 4.2.1
518
519  Bugs:
520  * Shared library numbers corrected.
521  * Broken support for 32-bit AIX fixed.
522  * Misc minor fixes.
523
524  Speedups:
525  * Exact division (mpz_divexact) now falls back to plain division for large
526    operands.
527
528  Features:
529  * Support for some new systems.
530
531
532Changes between GMP version 4.1.4 and 4.2
533
534  Bugs:
535  * Minor bug fixes and code generalizations.
536  * Expanded and improved test suite.
537
538  Speedups:
539  * Many minor optimizations, too many to mention here.
540  * Division now always subquadratic.
541  * Computation of n-factorial much faster.
542  * Added basic x86-64 assembly code.
543  * Floating-point output is now subquadratic for all bases.
544  * FFT multiply code now about 25% faster.
545  * Toom3 multiply code faster.
546
547  Features:
548  * Much improved configure.
549  * Workarounds for many more compiler bugs.
550  * Temporary allocations are now made on the stack only if small.
551  * New systems supported: HPPA-2.0 gcc, IA-64 HP-UX, PowerPC-64 Darwin,
552    Sparc64 GNU/Linux.
553  * New i386 fat binaries, selecting optimised code at runtime (--enable-fat).
554  * New build option: --enable-profiling=instrument.
555  * New memory function: mp_get_memory_functions.
556  * New Mersenne Twister random numbers: gmp_randinit_mt, also now used for
557    gmp_randinit_default.
558  * New random functions: gmp_randinit_set, gmp_urandomb_ui, gmp_urandomm_ui.
559  * New integer functions: mpz_combit, mpz_rootrem.
560  * gmp_printf etc new type "M" for mp_limb_t.
561  * gmp_scanf and friends now accept C99 hex floats.
562  * Numeric input and output can now be in bases up to 62.
563  * Comparisons mpz_cmp_d, mpz_cmpabs_d, mpf_cmp_d recognise infinities.
564  * Conversions mpz_get_d, mpq_get_d, mpf_get_d truncate towards zero,
565    previously their behaviour was unspecified.
566  * Fixes for overflow issues with operands >= 2^31 bits.
567
568  Caveats:
569  * mpfr is gone, and will from now on be released only separately.  Please see
570    www.mpfr.org.
571
572
573Changes between GMP version 4.1.3 and 4.1.4
574
575* Bug fix to FFT multiplication code (crash for huge operands).
576* Bug fix to mpf_sub (miscomputation).
577* Support for powerpc64-gnu-linux.
578* Better support for AMD64 in 32-bit mode.
579* Upwardly binary compatible with 4.1.3, 4.1.2, 4.1.1, 4.1, 4.0.1, 4.0,
580  and 3.x versions.
581
582
583Changes between GMP version 4.1.2 and 4.1.3
584
585* Bug fix for FFT multiplication code (miscomputation).
586* Bug fix to K6 assembly code for gcd.
587* Bug fix to IA-64 assembly code for population count.
588* Portability improvements, most notably functional AMD64 support.
589* mpz_export allows NULL for countp parameter.
590* Many minor bug fixes.
591* mpz_export allows NULL for countp parameter.
592* Upwardly binary compatible with 4.1.2, 4.1.1, 4.1, 4.0.1, 4.0, and 3.x
593  versions.
594
595
596Changes between GMP version 4.1.1 and 4.1.2
597
598* Bug fixes.
599
600
601Changes between GMP version 4.1 and 4.1.1
602
603* Bug fixes.
604* New systems supported: NetBSD and OpenBSD sparc64.
605
606
607Changes between GMP version 4.0.1 and 4.1
608
609* Bug fixes.
610* Speed improvements.
611* Upwardly binary compatible with 4.0, 4.0.1, and 3.x versions.
612* Asymptotically fast conversion to/from strings (mpz, mpq, mpn levels), but
613  also major speed improvements for tiny operands.
614* mpn_get_str parameter restrictions relaxed.
615* Major speed improvements for HPPA 2.0 systems.
616* Major speed improvements for UltraSPARC systems.
617* Major speed improvements for IA-64 systems (but still sub-optimal code).
618* Extended test suite.
619* mpfr is back, with many bug fixes and portability improvements.
620* New function: mpz_ui_sub.
621* New functions: mpz_export, mpz_import.
622* Optimization for nth root functions (mpz_root, mpz_perfect_power_p).
623* Optimization for extended gcd (mpz_gcdext, mpz_invert, mpn_gcdext).
624* Generalized low-level number format, reserving a `nails' part of each
625  limb.  (Please note that this is really experimental; some functions
626  are likely to compute garbage when nails are enabled.)
627* Nails-enabled Alpha 21264 assembly code, allowing up to 75% better
628  performance.  (Use --enable-nails=4 to enable it.)
629
630
631Changes between GMP version 4.0 and 4.0.1
632
633* Bug fixes.
634
635
636Changes between GMP version 3.1.1 and 4.0
637
638* Bug fixes.
639* Speed improvements.
640* Upwardly binary compatible with 3.x versions.
641* New CPU support: IA-64, Pentium 4.
642* Improved CPU support: 21264, Cray vector systems.
643* Support for all MIPS ABIs: o32, n32, 64.
644* New systems supported: Darwin, SCO, Windows DLLs.
645* New divide-and-conquer square root algorithm.
646* New algorithms chapter in the manual.
647* New malloc reentrant temporary memory method.
648* New C++ class interface by Gerardo Ballabio (beta).
649* Revamped configure, featuring ABI selection.
650* Speed improvements for mpz_powm and mpz_powm_ui (mainly affecting small
651  operands).
652* mpz_perfect_power_p now properly recognizes 0, 1, and negative perfect
653  powers.
654* mpz_hamdist now supports negative operands.
655* mpz_jacobi now accepts non-positive denominators.
656* mpz_powm now supports negative exponents.
657* mpn_mul_1 operand overlap requirements relaxed.
658* Float input and output uses locale specific decimal point where available.
659* New gmp_printf, gmp_scanf and related functions.
660* New division functions: mpz_cdiv_q_2exp, mpz_cdiv_r_2exp, mpz_divexact_ui.
661* New divisibility tests: mpz_divisible_p, mpz_divisible_ui_p,
662  mpz_divisible_2exp_p, mpz_congruent_p, mpz_congruent_ui_p,
663  mpz_congruent_2exp_p.
664* New Fibonacci function: mpz_fib2_ui.
665* New Lucas number functions: mpz_lucnum_ui, mpz_lucnum2_ui.
666* Other new integer functions: mpz_cmp_d, mpz_cmpabs_d, mpz_get_d_2exp,
667  mpz_init2, mpz_kronecker, mpz_lcm_ui, mpz_realloc2.
668* New rational I/O: mpq_get_str, mpq_inp_str, mpq_out_str, mpq_set_str.
669* Other new rational functions: mpq_abs, mpq_cmp_si, mpq_div_2exp,
670  mpq_mul_2exp, mpq_set_f.
671* New float tests: mpf_integer_p, mpf_fits_sint_p, mpf_fits_slong_p,
672  mpf_fits_sshort_p, mpf_fits_uint_p, mpf_fits_ulong_p, mpf_fits_ushort_p.
673* Other new float functions: mpf_cmp_d, mpf_get_default_prec, mpf_get_si,
674  mpf_get_ui, mpf_get_d_2exp.
675* New random functions: gmp_randinit_default, gmp_randinit_lc_2exp_size.
676* New demo expression string parser (see demos/expr).
677* New preliminary perl interface (see demos/perl).
678* Tuned algorithm thresholds for many more CPUs.
679
680
681Changes between GMP version 3.1 and 3.1.1
682
683* Bug fixes for division (rare), mpf_get_str, FFT, and miscellaneous minor
684  things.
685
686
687Changes between GMP version 3.0 and 3.1
688
689* Bug fixes.
690* Improved `make check' running more tests.
691* Tuned algorithm cutoff points for many machines.  This will improve speed for
692  a lot of operations, in some cases by a large amount.
693* Major speed improvements: Alpha 21264.
694* Some speed improvements: Cray vector computers, AMD K6 and Athlon, Intel P5
695  and Pentium Pro/II/III.
696* The mpf_get_prec function now works as it did in GMP 2.
697* New utilities for auto-tuning and speed measuring.
698* Multiplication now optionally uses FFT for very large operands.  (To enable
699  it, pass --enable-fft to configure.)
700* Support for new systems: Solaris running on x86, FreeBSD 5, HP-UX 11, Cray
701  vector computers, Rhapsody, Nextstep/Openstep, MacOS.
702* Support for shared libraries on 32-bit HPPA.
703* New integer functions: mpz_mul_si, mpz_odd_p, mpz_even_p.
704* New Kronecker symbol functions: mpz_kronecker_si, mpz_kronecker_ui,
705  mpz_si_kronecker, mpz_ui_kronecker.
706* New rational functions: mpq_out_str, mpq_swap.
707* New float functions: mpf_swap.
708* New mpn functions: mpn_divexact_by3c, mpn_tdiv_qr.
709* New EXPERIMENTAL function layer for accurate floating-point arithmetic, mpfr.
710  To try it, pass --enable-mpfr to configure.  See the mpfr subdirectory for
711  more information; it is not documented in the main GMP manual.
712
713
714Changes between GMP version 3.0 and 3.0.1
715
716* Memory leaks in gmp_randinit and mpz_probab_prime_p fixed.
717* Documentation for gmp_randinit fixed.  Misc documentation errors fixed.
718
719
720Changes between GMP version 2.0 and 3.0
721
722* Source level compatibility with past releases (except mpn_gcd).
723* Bug fixes.
724* Much improved speed thanks to both host independent and host dependent
725  optimizations.
726* Switch to autoconf/automake/libtool.
727* Support for building libgmp as a shared library.
728* Multiplication and squaring using 3-way Toom-Cook.
729* Division using the Burnikel-Ziegler method.
730* New functions computing binomial coefficients: mpz_bin_ui, mpz_bin_uiui.
731* New function computing Fibonacci numbers: mpz_fib_ui.
732* New random number generators: mpf_urandomb, mpz_rrandomb, mpz_urandomb,
733  mpz_urandomm, gmp_randclear, gmp_randinit, gmp_randinit_lc_2exp,
734  gmp_randseed, gmp_randseed_ui.
735* New function for quickly extracting limbs: mpz_getlimbn.
736* New functions performing integer size tests: mpz_fits_sint_p,
737  mpz_fits_slong_p, mpz_fits_sshort_p, mpz_fits_uint_p, mpz_fits_ulong_p,
738  mpz_fits_ushort_p.
739* New mpf functions: mpf_ceil, mpf_floor, mpf_pow_ui, mpf_trunc.
740* New mpq function: mpq_set_d.
741* New mpz functions: mpz_addmul_ui, mpz_cmpabs, mpz_cmpabs_ui, mpz_lcm,
742  mpz_nextprime, mpz_perfect_power_p, mpz_remove, mpz_root, mpz_swap,
743  mpz_tdiv_ui, mpz_tstbit, mpz_xor.
744* New mpn function: mpn_divexact_by3.
745* New CPU support: DEC Alpha 21264, AMD K6 and Athlon, HPPA 2.0 and 64,
746  Intel Pentium Pro and Pentium-II/III, Sparc 64, PowerPC 64.
747* Almost 10 times faster mpz_invert and mpn_gcdext.
748* The interface of mpn_gcd has changed.
749* Better support for MIPS R4x000 and R5000 under Irix 6.
750* Improved support for SPARCv8 and SPARCv9 processors.
751
752
753Changes between GMP version 2.0 and 2.0.2
754
755* Many bug fixes.
756
757
758Changes between GMP version 1.3.2 and 2.0
759
760* Division routines in the mpz class have changed.  There are three classes of
761  functions, that rounds the quotient to -infinity, 0, and +infinity,
762  respectively.  The first class of functions have names that begin with
763  mpz_fdiv (f is short for floor), the second class' names begin with mpz_tdiv
764  (t is short for trunc), and the third class' names begin with mpz_cdiv (c is
765  short for ceil).
766
767  The old division routines beginning with mpz_m are similar to the new
768  mpz_fdiv, with the exception that some of the new functions return useful
769  values.
770
771  The old function names can still be used.  All the old functions names will
772  now do floor division, not trunc division as some of them used to.  This was
773  changed to make the functions more compatible with common mathematical
774  practice.
775
776  The mpz_mod and mpz_mod_ui functions now compute the mathematical mod
777  function.  I.e., the sign of the 2nd argument is ignored.
778
779* The mpq assignment functions do not canonicalize their results.  A new
780  function, mpq_canonicalize must be called by the user if the result is not
781  known to be canonical.
782* The mpn functions are now documented.  These functions are intended for
783  very time critical applications, or applications that need full control over
784  memory allocation.  Note that the mpn interface is irregular and hard to
785  use.
786* New functions for arbitrary precision floating point arithmetic.  Names
787  begin with `mpf_'.  Associated type mpf_t.
788* New and improved mpz functions, including much faster GCD, fast exact
789  division (mpz_divexact), bit scan (mpz_scan0 and mpz_scan1), and number
790  theoretical functions like Jacobi (mpz_jacobi) and multiplicative inverse
791  (mpz_invert).
792* New variable types (mpz_t and mpq_t) are available that makes syntax of
793  mpz and mpq calls nicer (no need for & before variables).  The MP_INT and
794  MP_RAT types are still available for compatibility.
795* Uses GNU configure.  This makes it possible to choose target architecture
796  and CPU variant, and to compile into a separate object directory.
797* Carefully optimized assembly for important inner loops.  Support for DEC
798  Alpha, Amd 29000, HPPA 1.0 and 1.1, Intel Pentium and generic x86, Intel
799  i960, Motorola MC68000, MC68020, MC88100, and MC88110, Motorola/IBM
800  PowerPC, National NS32000, IBM POWER, MIPS R3000, R4000, SPARCv7,
801  SuperSPARC, generic SPARCv8, and DEC VAX.  Some support also for ARM,
802  Clipper, IBM ROMP (RT), and Pyramid AP/XP.
803* Faster.  Thanks to the assembler code, new algorithms, and general tuning.
804  In particular, the speed on machines without GCC is improved.
805* Support for machines without alloca.
806* Now under the LGPL.
807
808INCOMPATIBILITIES BETWEEN GMP 1 AND GMP 2
809
810* mpq assignment functions do not canonicalize their results.
811* mpz division functions round differently.
812* mpz mod functions now really compute mod.
813* mpz_powm and mpz_powm_ui now really use mod for reduction.
814