1 Copyright 1996, 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008,
2 2009, 2010, 2011, 2012 Free Software Foundation, Inc.
4 Verbatim copying and distribution of this entire article is permitted in any
5 medium, provided this notice is preserved.
8 Changes between GMP version 5.0.4 and 5.0.5
11 * A bug causing AMD 11h processors to be treated like AMD 10h has been fixed.
12 The 11h processors do not correctly handle all 10h (aka K10) instructions,
13 and GMP's use of these instructions results in major miscomputations (not
14 as one would have hoped CPU traps of some 'illegal instruction' sort).
16 * A bug affecting recent Intel Sandy Bridge CPUs resulting in configuration
17 failures has been fixed,
23 * A couple of tests added to the self-check suite.
29 Changes between GMP version 5.0.3 and 5.0.4
32 * Thresholds in mpn_powm_sec for both fat and non-fat builds are now used
33 safely, plugging a one-word buffer overrun introduced in the 5.0.3 release
34 (for non-fat) and a multi-word buffer overrun that existed since 5.0 (for
35 fat). (We have not been able to provoke malign stack smashing in any of
36 the ~100 configurations explored by the GMP nightly builds, but the bug
37 should be assumed to be exploitable.)
39 * Two bugs in multiplication code causing incorrect computation with
40 extremely low probability have been fixed.
42 * A bug in the test suite causing buffer overruns during "make check",
43 sometimes leading to subsequent malloc crashes, has been fixed.
45 * Two bugs in the gcd code have been fixed. They could lead to incorrect
46 results, but for uniformly distributed random operands, the likelihood for
47 that is infinitesimally small. (There was also a third bug, but that was
48 an incorrect ASSERT, which furthermore was not enabled by default.)
50 * A bug affecting 32-bit PowerPC division has been fixed. The bug caused
51 miscomputation for certain divisors in the range 2^32 ... 2^64-1 (about 1
55 * None, except indirectly through recognition of new CPUs, and through better
59 * Some more tests added to the self-check suite.
61 * The AMD "Bulldozer" CPU is now recognised.
67 Changes between GMP version 5.0.2 and 5.0.3
70 * A few minor bugs related to portability fixed.
72 * A slight timing leak of the powm_sec functions have been sealed. (This
73 leak could possibly be used to extract the most significant few bits of the
74 exponent. "Few" here means at most 10.)
76 * The mpz_nextprime function now runs a safer number of pseudo-random prime
79 * A bug in division code possibly causing incorrect computation was fixed.
82 * None, except indirectly through recognition of new CPUs, and through better
86 * New CPUs recognised.
88 * IBM S/390 are now supported in both 31/32-bit and 64-bit mode. (We have
89 not been able to fully test this on any multilib machine, since IBM expired
90 our guest account a few days before our release.)
96 Changes between GMP version 5.0.1 and 5.0.2
99 * Many minor bugs related to portability fixed.
101 * The support for HPPA 2.0N now works, after an assembly bug fix.
103 * A test case type error has been fixed. The symptom of this bug was
104 spurious 'make check' failures.
107 * None, except indirectly through recognition of new CPUs.
110 * Fat builds are now supported for 64-bit x86 processors also under Darwin.
116 Changes between GMP version 5.0.0 and 5.0.1
121 * Fixed crash for huge multiplies when old FFT_TABLE2 type of parameter
122 selection tables' sentinel was smaller than multiplied operands.
124 * The solib numbers now reflect the removal of the documented but preliminary
125 mpn_bdivmod function; we correctly flag incompatibility with GMP 4.3. GMP
126 5.0.0 has this wrong, and should perhaps be uninstalled to avoid confusion.
129 * Multiplication of large numbers has indirectly been sped up through better
130 FFT tuning and processor recognition. Since many operations depend on
131 multiplication, there will be a general speedup.
134 * More Core i3, i5 an Core i7 processor models are recognised.
136 * Fixes and workarounds for Mac OS quirks should make this GMP version build
137 using many of the different versions of "Xcode".
140 * The amount of scratch memory needed for multiplication of huge numbers has
141 been reduced substantially (but is still larger than in GMP 4.3.)
143 * Likewise, the amount of scratch memory needed for division of large numbers
144 has been reduced substantially.
146 * The FFT tuning code of tune/tuneup.c has been completely rewritten, and
147 new, large FFT parameter selection tables are provided for many machines.
149 * Upgraded to the latest autoconf, automake, libtool.
152 Changes between GMP version 4.3.X and 5.0.0
155 * None (contains the same fixes as release 4.3.2).
158 * Multiplication has been overhauled:
159 (1) Multiplication of larger same size operands has been improved with
160 the addition of two new Toom functions and a new internal function
161 mpn_mulmod_bnm1 (computing U * V mod (B^n-1), B being the word base.
162 This latter function is used for the largest products, waiting for a
163 better Schoenhage-Strassen U * V mod (B^n+1) implementation.
164 (2) Likewise for squaring.
165 (3) Multiplication of different size operands has been improved with the
166 addition of many new Toom function, and by selecting underlying
167 functions better from the main multiply functions.
169 * Division and mod have been overhauled:
170 (1) Plain "schoolbook" division is reimplemented using faster quotient
172 (2) Division Q = N/D, R = N mod D where both the quotient and remainder
173 are needed now runs in time O(M(log(N))). This is an improvement of
175 (3) Division where just the quotient is needed is now O(M(log(Q))) on
177 (4) Modulo operations using Montgomery REDC form now take time O(M(n)).
178 (5) Exact division Q = N/D by means of mpz_divexact has been improved
179 for all sizes, and now runs in time O(M(log(N))).
181 * The function mpz_powm is now faster for all sizes. Its complexity has
182 gone from O(M(n)log(n)m) to O(M(n)m) where n is the size of the modulo
183 argument and m is the size of the exponent. It is also radically
184 faster for even modulus, since it now partially factors such modulus
185 and performs two smaller modexp operations, then uses CRT.
187 * The internal support for multiplication yielding just the lower n limbs
188 has been improved by using Mulders' algorithm.
190 * Computation of inverses, both plain 1/N and 1/N mod B^n have been
191 improved by using well-tuned Newton iterations, and wrap-around
192 multiplication using mpn_mulmod_bnm1.
194 * A new algorithm makes mpz_perfect_power_p asymptotically faster.
196 * The function mpz_remove uses a much faster algorithm, is better tuned,
197 and also benefits from the division improvements.
199 * Intel Atom and VIA Nano specific optimisations.
201 * Plus hundreds of smaller improvements and tweaks!
204 * New mpz function: mpz_powm_sec for side-channel quiet modexp
207 * New mpn functions: mpn_sqr, mpn_and_n, mpn_ior_n, mpn_xor_n, mpn_nand_n,
208 mpn_nior_n, mpn_xnor_n, mpn_andn_n, mpn_iorn_n, mpn_com, mpn_neg,
209 mpn_copyi, mpn_copyd, mpn_zero.
211 * The function mpn_tdiv_qr now allows certain argument overlap.
213 * Support for fat binaries for 64-bit x86 processors has been added.
215 * A new type, mp_bitcnt_t for bignum bit counts, has been introduced.
217 * Support for Windows64 through mingw64 has been added.
219 * The cofactors of mpz_gcdext and mpn_gcdext are now more strictly
220 normalised, returning to how GMP 4.2 worked. (Note that also release
221 4.3.2 has this change.)
224 * The mpn_mul function should no longer be used for squaring,
225 instead use the new mpn_sqr.
227 * The algorithm selection has been improved, the number of thresholds have
228 more than doubled, and the tuning and use of existing thresholds have
231 * The tune/speed program can measure many of new functions.
233 * The mpn_bdivmod function has been removed. We do not consider this an
234 incompatible change, since the function was marked as preliminary.
236 * The testsuite has been enhanced in various ways.
239 Changes between GMP version 4.3.1 and 4.3.2
242 * Fixed bug in mpf_eq.
243 * Fixed overflow issues in mpz_set_str, mpz_inp_str, mpf_set_str, and
245 * Avoid unbounded stack allocation for unbalanced multiplication.
246 * Fixed bug in FFT multiplication.
249 * None, except that proper processor recognition helps affected processors.
252 * Recognise more "Core 2" processor variants.
253 * The cofactors of mpz_gcdext and mpn_gcdext are now more strictly
254 normalised, returning to how GMP 4.2 worked.
257 Changes between GMP version 4.3.0 and 4.3.1
260 * Fixed bug in mpn_gcdext, affecting also mpz_gcdext and mpz_invert.
261 The bug could cause a cofactor to have a leading zero limb, which
262 could lead to crashes or miscomputation later on.
263 * Fixed some minor documentation issues.
269 * Workarounds for various issues with Mac OS X's build tools.
270 * Recognise more IBM "POWER" processor variants.
273 Changes between GMP version 4.2.X and 4.3.0
276 * Fixed bug in mpz_perfect_power_p with recognition of negative perfect
277 powers that can be written both as an even and odd power.
278 * We might accidentally have added bugs since there is a large amount of
279 new code in this release.
282 * Vastly improved assembly code for x86-64 processors from AMD and Intel.
283 * Major improvements also for many other processor families, such as
284 Alpha, PowerPC, and Itanium.
285 * New sub-quadratic mpn_gcd and mpn_gcdext, as well as improved basecase
287 * The multiply FFT code has been slightly improved.
288 * Balanced multiplication now uses 4-way Toom in addition to schoolbook,
289 Karatsuba, 3-way Toom, and FFT.
290 * Unbalanced multiplication has been vastly improved.
291 * Improved schoolbook division by means of faster quotient approximation.
292 * Several new algorithms for division and mod by single limbs, giving
294 * Improved nth root computations.
295 * The mpz_nextprime function uses sieving and is much faster.
296 * Countless minor tweaks.
299 * Updated support for fat binaries for x86_32 include current processors
300 * Lots of new mpn internal interfaces. Some of them will become public
301 in a future GMP release.
302 * Support for the 32-bit ABI under x86-apple-darwin.
303 * x86 CPU recognition code should now default better for future
305 * The experimental nails feature does not work in this release, but
306 it might be re-enabled in the future.
309 * The gmp_version variable now always contains three parts. For this
310 release, it is "4.3.0".
313 Changes between GMP version 4.2.3 and 4.2.4
316 * Fix bug with parsing exponent '+' sign in mpf.
317 * Fix an allocation bug in mpf_set_str, also affecting mpf_init_set_str, and
321 * None, except that proper processor recognition helps affected processors.
324 * Recognize new AMD processors.
327 Changes between GMP version 4.2.2 and 4.2.3
330 * Fix x86 CPU recognition code to properly identify recent AMD and Intel
332 * The >> operator of the C++ wrapper gmpxx.h now does floor rounding, not
334 * Inline semantics now follow the C99 standard, and works with recent GCC
336 * C++ bitwise logical operations work for more types.
337 * For C++, gmp.h now includes cstdio, improving compiler compatibility.
338 * Bases > 36 now work properly in mpf_set_str.
341 * None, except that proper processor recognition helps affected processors.
344 * The allocation functions now detect overflow of the mpz_t type. This means
345 that overflow will now cause an abort, except when the allocation
346 computation itself overflows. (Such overflow can probably only happen in
347 powering functions; we will detect powering overflow in the future.)
350 Changes between GMP version 4.2.1 and 4.2.2
352 * License is now LGPL version 3.
355 * Shared library numbers corrected for libcxx.
356 * Fixed serious bug in gmpxx.h where a=a+b*c would generate garbage.
357 Note that this only affects C++ programs.
358 * Fix crash in mpz_set_d for arguments with large negative exponent.
359 * Fix 32-bit ABI bug with Itanium assembly for popcount and hamdist.
360 * Fix assembly syntax problem for powerpc-ibm-aix with AIX native assembler.
361 * Fix problems with x86 --enable-fat, where the compiler where told to
362 generate code for the build machine, not plain i386 code as it should.
363 * Improved recognition of powerpc systems wrt Altivec/VMX capability.
364 * Misc minor fixes, mainly workarounds for compiler/assembler bugs.
367 * "Core 2" and Pentium 4 processors, running in 64-bit mode will get a
368 slight boost as they are now specifically recognized.
371 * New support for x86_64-solaris
372 * New, rudimentary support for x86-apple-darwin and x86_64-apple-darwin.
373 (Please see http://gmplib.org/macos.html for more information.)
376 Changes between GMP version 4.2 and 4.2.1
379 * Shared library numbers corrected.
380 * Broken support for 32-bit AIX fixed.
384 * Exact division (mpz_divexact) now falls back to plain division for large
388 * Support for some new systems.
391 Changes between GMP version 4.1.4 and 4.2
394 * Minor bug fixes and code generalizations.
395 * Expanded and improved test suite.
398 * Many minor optimizations, too many to mention here.
399 * Division now always subquadratic.
400 * Computation of n-factorial much faster.
401 * Added basic x86-64 assembly code.
402 * Floating-point output is now subquadratic for all bases.
403 * FFT multiply code now about 25% faster.
404 * Toom3 multiply code faster.
407 * Much improved configure.
408 * Workarounds for many more compiler bugs.
409 * Temporary allocations are now made on the stack only if small.
410 * New systems supported: HPPA-2.0 gcc, IA-64 HP-UX, PowerPC-64 Darwin,
412 * New i386 fat binaries, selecting optimised code at runtime (--enable-fat).
413 * New build option: --enable-profiling=instrument.
414 * New memory function: mp_get_memory_functions.
415 * New Mersenne Twister random numbers: gmp_randinit_mt, also now used for
416 gmp_randinit_default.
417 * New random functions: gmp_randinit_set, gmp_urandomb_ui, gmp_urandomm_ui.
418 * New integer functions: mpz_combit, mpz_rootrem.
419 * gmp_printf etc new type "M" for mp_limb_t.
420 * gmp_scanf and friends now accept C99 hex floats.
421 * Numeric input and output can now be in bases up to 62.
422 * Comparisons mpz_cmp_d, mpz_cmpabs_d, mpf_cmp_d recognise infinities.
423 * Conversions mpz_get_d, mpq_get_d, mpf_get_d truncate towards zero,
424 previously their behaviour was unspecified.
425 * Fixes for overflow issues with operands >= 2^31 bits.
428 * mpfr is gone, and will from now on be released only separately. Please see
432 Changes between GMP version 4.1.3 and 4.1.4
434 * Bug fix to FFT multiplication code (crash for huge operands).
435 * Bug fix to mpf_sub (miscomputation).
436 * Support for powerpc64-gnu-linux.
437 * Better support for AMD64 in 32-bit mode.
438 * Upwardly binary compatible with 4.1.3, 4.1.2, 4.1.1, 4.1, 4.0.1, 4.0,
442 Changes between GMP version 4.1.2 and 4.1.3
444 * Bug fix for FFT multiplication code (miscomputation).
445 * Bug fix to K6 assembly code for gcd.
446 * Bug fix to IA-64 assembly code for population count.
447 * Portability improvements, most notably functional AMD64 support.
448 * mpz_export allows NULL for countp parameter.
449 * Many minor bug fixes.
450 * mpz_export allows NULL for countp parameter.
451 * Upwardly binary compatible with 4.1.2, 4.1.1, 4.1, 4.0.1, 4.0, and 3.x
455 Changes between GMP version 4.1.1 and 4.1.2
460 Changes between GMP version 4.1 and 4.1.1
463 * New systems supported: NetBSD and OpenBSD sparc64.
466 Changes between GMP version 4.0.1 and 4.1
469 * Speed improvements.
470 * Upwardly binary compatible with 4.0, 4.0.1, and 3.x versions.
471 * Asymptotically fast conversion to/from strings (mpz, mpq, mpn levels), but
472 also major speed improvements for tiny operands.
473 * mpn_get_str parameter restrictions relaxed.
474 * Major speed improvements for HPPA 2.0 systems.
475 * Major speed improvements for UltraSPARC systems.
476 * Major speed improvements for IA-64 systems (but still sub-optimal code).
477 * Extended test suite.
478 * mpfr is back, with many bug fixes and portability improvements.
479 * New function: mpz_ui_sub.
480 * New functions: mpz_export, mpz_import.
481 * Optimization for nth root functions (mpz_root, mpz_perfect_power_p).
482 * Optimization for extended gcd (mpz_gcdext, mpz_invert, mpn_gcdext).
483 * Generalized low-level number format, reserving a `nails' part of each
484 limb. (Please note that this is really experimental; some functions
485 are likely to compute garbage when nails are enabled.)
486 * Nails-enabled Alpha 21264 assembly code, allowing up to 75% better
487 performance. (Use --enable-nails=4 to enable it.)
490 Changes between GMP version 4.0 and 4.0.1
495 Changes between GMP version 3.1.1 and 4.0
498 * Speed improvements.
499 * Upwardly binary compatible with 3.x versions.
500 * New CPU support: IA-64, Pentium 4.
501 * Improved CPU support: 21264, Cray vector systems.
502 * Support for all MIPS ABIs: o32, n32, 64.
503 * New systems supported: Darwin, SCO, Windows DLLs.
504 * New divide-and-conquer square root algorithm.
505 * New algorithms chapter in the manual.
506 * New malloc reentrant temporary memory method.
507 * New C++ class interface by Gerardo Ballabio (beta).
508 * Revamped configure, featuring ABI selection.
509 * Speed improvements for mpz_powm and mpz_powm_ui (mainly affecting small
511 * mpz_perfect_power_p now properly recognizes 0, 1, and negative perfect
513 * mpz_hamdist now supports negative operands.
514 * mpz_jacobi now accepts non-positive denominators.
515 * mpz_powm now supports negative exponents.
516 * mpn_mul_1 operand overlap requirements relaxed.
517 * Float input and output uses locale specific decimal point where available.
518 * New gmp_printf, gmp_scanf and related functions.
519 * New division functions: mpz_cdiv_q_2exp, mpz_cdiv_r_2exp, mpz_divexact_ui.
520 * New divisibility tests: mpz_divisible_p, mpz_divisible_ui_p,
521 mpz_divisible_2exp_p, mpz_congruent_p, mpz_congruent_ui_p,
522 mpz_congruent_2exp_p.
523 * New Fibonacci function: mpz_fib2_ui.
524 * New Lucas number functions: mpz_lucnum_ui, mpz_lucnum2_ui.
525 * Other new integer functions: mpz_cmp_d, mpz_cmpabs_d, mpz_get_d_2exp,
526 mpz_init2, mpz_kronecker, mpz_lcm_ui, mpz_realloc2.
527 * New rational I/O: mpq_get_str, mpq_inp_str, mpq_out_str, mpq_set_str.
528 * Other new rational functions: mpq_abs, mpq_cmp_si, mpq_div_2exp,
529 mpq_mul_2exp, mpq_set_f.
530 * New float tests: mpf_integer_p, mpf_fits_sint_p, mpf_fits_slong_p,
531 mpf_fits_sshort_p, mpf_fits_uint_p, mpf_fits_ulong_p, mpf_fits_ushort_p.
532 * Other new float functions: mpf_cmp_d, mpf_get_default_prec, mpf_get_si,
533 mpf_get_ui, mpf_get_d_2exp.
534 * New random functions: gmp_randinit_default, gmp_randinit_lc_2exp_size.
535 * New demo expression string parser (see demos/expr).
536 * New preliminary perl interface (see demos/perl).
537 * Tuned algorithm thresholds for many more CPUs.
540 Changes between GMP version 3.1 and 3.1.1
542 * Bug fixes for division (rare), mpf_get_str, FFT, and miscellaneous minor
546 Changes between GMP version 3.0 and 3.1
549 * Improved `make check' running more tests.
550 * Tuned algorithm cutoff points for many machines. This will improve speed for
551 a lot of operations, in some cases by a large amount.
552 * Major speed improvements: Alpha 21264.
553 * Some speed improvements: Cray vector computers, AMD K6 and Athlon, Intel P5
554 and Pentium Pro/II/III.
555 * The mpf_get_prec function now works as it did in GMP 2.
556 * New utilities for auto-tuning and speed measuring.
557 * Multiplication now optionally uses FFT for very large operands. (To enable
558 it, pass --enable-fft to configure.)
559 * Support for new systems: Solaris running on x86, FreeBSD 5, HP-UX 11, Cray
560 vector computers, Rhapsody, Nextstep/Openstep, MacOS.
561 * Support for shared libraries on 32-bit HPPA.
562 * New integer functions: mpz_mul_si, mpz_odd_p, mpz_even_p.
563 * New Kronecker symbol functions: mpz_kronecker_si, mpz_kronecker_ui,
564 mpz_si_kronecker, mpz_ui_kronecker.
565 * New rational functions: mpq_out_str, mpq_swap.
566 * New float functions: mpf_swap.
567 * New mpn functions: mpn_divexact_by3c, mpn_tdiv_qr.
568 * New EXPERIMENTAL function layer for accurate floating-point arithmetic, mpfr.
569 To try it, pass --enable-mpfr to configure. See the mpfr subdirectory for
570 more information; it is not documented in the main GMP manual.
573 Changes between GMP version 3.0 and 3.0.1
575 * Memory leaks in gmp_randinit and mpz_probab_prime_p fixed.
576 * Documentation for gmp_randinit fixed. Misc documentation errors fixed.
579 Changes between GMP version 2.0 and 3.0
581 * Source level compatibility with past releases (except mpn_gcd).
583 * Much improved speed thanks to both host independent and host dependent
585 * Switch to autoconf/automake/libtool.
586 * Support for building libgmp as a shared library.
587 * Multiplication and squaring using 3-way Toom-Cook.
588 * Division using the Burnikel-Ziegler method.
589 * New functions computing binomial coefficients: mpz_bin_ui, mpz_bin_uiui.
590 * New function computing Fibonacci numbers: mpz_fib_ui.
591 * New random number generators: mpf_urandomb, mpz_rrandomb, mpz_urandomb,
592 mpz_urandomm, gmp_randclear, gmp_randinit, gmp_randinit_lc_2exp,
593 gmp_randseed, gmp_randseed_ui.
594 * New function for quickly extracting limbs: mpz_getlimbn.
595 * New functions performing integer size tests: mpz_fits_sint_p,
596 mpz_fits_slong_p, mpz_fits_sshort_p, mpz_fits_uint_p, mpz_fits_ulong_p,
598 * New mpf functions: mpf_ceil, mpf_floor, mpf_pow_ui, mpf_trunc.
599 * New mpq function: mpq_set_d.
600 * New mpz functions: mpz_addmul_ui, mpz_cmpabs, mpz_cmpabs_ui, mpz_lcm,
601 mpz_nextprime, mpz_perfect_power_p, mpz_remove, mpz_root, mpz_swap,
602 mpz_tdiv_ui, mpz_tstbit, mpz_xor.
603 * New mpn function: mpn_divexact_by3.
604 * New CPU support: DEC Alpha 21264, AMD K6 and Athlon, HPPA 2.0 and 64,
605 Intel Pentium Pro and Pentium-II/III, Sparc 64, PowerPC 64.
606 * Almost 10 times faster mpz_invert and mpn_gcdext.
607 * The interface of mpn_gcd has changed.
608 * Better support for MIPS R4x000 and R5000 under Irix 6.
609 * Improved support for SPARCv8 and SPARCv9 processors.
612 Changes between GMP version 2.0 and 2.0.2
617 Changes between GMP version 1.3.2 and 2.0
619 * Division routines in the mpz class have changed. There are three classes of
620 functions, that rounds the quotient to -infinity, 0, and +infinity,
621 respectively. The first class of functions have names that begin with
622 mpz_fdiv (f is short for floor), the second class' names begin with mpz_tdiv
623 (t is short for trunc), and the third class' names begin with mpz_cdiv (c is
626 The old division routines beginning with mpz_m are similar to the new
627 mpz_fdiv, with the exception that some of the new functions return useful
630 The old function names can still be used. All the old functions names will
631 now do floor division, not trunc division as some of them used to. This was
632 changed to make the functions more compatible with common mathematical
635 The mpz_mod and mpz_mod_ui functions now compute the mathematical mod
636 function. I.e., the sign of the 2nd argument is ignored.
638 * The mpq assignment functions do not canonicalize their results. A new
639 function, mpq_canonicalize must be called by the user if the result is not
640 known to be canonical.
641 * The mpn functions are now documented. These functions are intended for
642 very time critical applications, or applications that need full control over
643 memory allocation. Note that the mpn interface is irregular and hard to
645 * New functions for arbitrary precision floating point arithmetic. Names
646 begin with `mpf_'. Associated type mpf_t.
647 * New and improved mpz functions, including much faster GCD, fast exact
648 division (mpz_divexact), bit scan (mpz_scan0 and mpz_scan1), and number
649 theoretical functions like Jacobi (mpz_jacobi) and multiplicative inverse
651 * New variable types (mpz_t and mpq_t) are available that makes syntax of
652 mpz and mpq calls nicer (no need for & before variables). The MP_INT and
653 MP_RAT types are still available for compatibility.
654 * Uses GNU configure. This makes it possible to choose target architecture
655 and CPU variant, and to compile into a separate object directory.
656 * Carefully optimized assembly for important inner loops. Support for DEC
657 Alpha, Amd 29000, HPPA 1.0 and 1.1, Intel Pentium and generic x86, Intel
658 i960, Motorola MC68000, MC68020, MC88100, and MC88110, Motorola/IBM
659 PowerPC, National NS32000, IBM POWER, MIPS R3000, R4000, SPARCv7,
660 SuperSPARC, generic SPARCv8, and DEC VAX. Some support also for ARM,
661 Clipper, IBM ROMP (RT), and Pyramid AP/XP.
662 * Faster. Thanks to the assembler code, new algorithms, and general tuning.
663 In particular, the speed on machines without GCC is improved.
664 * Support for machines without alloca.
665 * Now under the LGPL.
667 INCOMPATIBILITIES BETWEEN GMP 1 AND GMP 2
669 * mpq assignment functions do not canonicalize their results.
670 * mpz division functions round differently.
671 * mpz mod functions now really compute mod.
672 * mpz_powm and mpz_powm_ui now really use mod for reduction.