Changes between GMP version 5.0.1 and 5.0.2
BUGS FIXED
* Many minor bugs related to portability fixed.
* The support for HPPA 2.0N now works, after an assembly bug fix.
* A test case type error has been fixed. The symptom of this bug
was spurious 'make check' failures.
SPEEDUPS
* None, except indirectly through recognition of new CPUs.
FEATURES
* Fat builds are now supported for 64-bit x86 processors also under Darwin.
MISC
* None.
Unfortunately, with recent gcc, the x86 version of mpn fails to build on
Solaris. Add it to Darwin on the list of platforms for which we build a
generic mpn instead.
Changes in GMP 5.0.1
BUGS FIXED
- Fat builds fixed.
- Fixed crash for huge multiplies when old FFT_TABLE2 type of parameter selection tables' sentinel was smaller than multiplied
operands.
- The solib numbers now reflect the removal of the documented but preliminary mpn_bdivmod function; we correctly flag
incompatibility with GMP 4.3. GMP 5.0.0 has this wrong, and should perhaps be uninstalled to avoid confusion.
SPEEDUPS
- Multiplication of large numbers has indirectly been sped up through better FFT tuning and processor recognition. Since many
operations depend on multiplication, there will be a general speedup.
FEATURES
- More Core i3, i5 an Core i7 processor models are recognised.
- Fixes and workarounds for Mac OS quirks should make this GMP version build using many of the different versions of "Xcode".
MISC
- The amount of scratch memory needed for multiplication of huge numbers have been reduced substantially (but is still larger
than in GMP 4.3.)
- Likewise, the amount of scratch memory needed for division of large numbers have been reduced substantially.
- The FFT tuning code of tune/tuneup.c has been completely rewritten, and new, large FFT parameter selection tables are provided
for many machines.
- Upgraded to the latest autoconf, automake, libtool.
Changes in GMP 5.0.0
BUGS FIXED
- None (contains the same fixes as release 4.3.2).
SPEEDUPS
- Multiplication has been overhauled:
1. Multiplication of larger same size operands has been improved with the addition of two new Toom functions and a new
internal function mpn_mulmod_bnm1 (computing U * V mod (B^n-1), B being the word base. This latter function is used for the
largest products, waiting for a better Schoenhage-Strassen U * V mod (B^n+1) implementation.
2. Likewise for squaring.
3. Multiplication of different size operands has been improved with the addition of many new Toom function, and by selecting
underlying functions better from the main multiply functions.
- Division and mod have been overhauled:
1. Plain "schoolbook" division is reimplemented using faster quotient approximation.
2. Division Q = N/D, R = N mod D where both the quotient and remainder are needed now runs in time O(M(log(N))). This is an
improvement of a factor log(log(N))
3. Division where just the quotient is needed is now O(M(log(Q))) on average.
4. Modulo operations using Montgomery REDC form now take time O(M(n)).
5. Exact division Q = N/D by means of mpz_divexact has been improved for all sizes, and now runs in time O(M(log(N))).
- The function mpz_powm is now faster for all sizes. Its complexity has gone from O(M(n)log(n)m) to O(M(n)m) where n is the size
of the modulo argument and m is the size of the exponent. It is also radically faster for even modulus, since it now partially
factors such modulus and performs two smaller modexp operations, then uses CRT.
- The internal support for multiplication yielding just the lower n limbs has been improved by using Mulders' algorithm.
- Computation of inverses, both plain 1/N and 1/N mod B^n have been improved by using well-tuned Newton iterations, and
wrap-around multiplication using mpn_mulmod_bnm1.
- A new algorithm makes mpz_perfect_power_p asymptotically faster.
- The function mpz_remove uses a much faster algorithm, is better tuned, and also benefits from the division improvements.
- Intel Atom and VIA Nano specific optimisations.
- Plus hundreds of smaller improvements and tweaks!
FEATURES
- New mpz function: mpz_powm_sec for side-channel quiet modexp computations.
- New mpn functions: mpn_sqr, mpn_and_n, mpn_ior_n, mpn_xor_n, mpn_nand_n, mpn_nior_n, mpn_xnor_n, mpn_andn_n, mpn_iorn_n,
mpn_com, mpn_neg, mpn_copyi, mpn_copyd, mpn_zero.
- The function mpn_tdiv_qr now allows certain argument overlap.
- Support for fat binaries for 64-bit x86 processors has been added.
- A new type, mp_bitcnt_t for bignum bit counts, has been introduced.
- Support for Windows64 through mingw64 has been added.
- The cofactors of mpz_gcdext and mpn_gcdext are now more strictly normalised, returning to how GMP 4.2 worked. (Note that also
release 4.3.2 has this change.)
MISC
- The mpn_mul function should no longer be used for squaring, instead use the new mpn_sqr.
- The algorithm selection has been improved, the number of thresholds have more than doubled, and the tuning and use of existing
thresholds have been improved.
- The tune/speed program can measure many of new functions.
- The mpn_bdivmod function has been removed. We do not consider this an incompatible change, since the function was marked as
preliminary.
- The testsuite has been enhanced in various ways.
Changes in GMP 4.3.2
Bugs:
- Fixed bug in mpf_eq.
- Fixed overflow issues in mpz_set_str, mpz_inp_str, mpf_set_str, and mpf_get_str.
- Avoid unbounded stack allocation for unbalanced multiplication.
- Fixed bug in FFT multiplication.
Speedups:
- None, except that improved processor recognition helps affected processors.
Features:
- Recognise more "Core 2" processor variants.
- The cofactors of mpz_gcdext and mpn_gcdext are now more strictly normalised, returning to how GMP 4.2 worked.
also affects some files.
Changes between GMP version 4.3.0 and 4.3.1
Bugs:
* Fixed bug in mpn_gcdext, affecting also mpz_gcdext and mpz_invert.
The bug could cause a cofactor to have a leading zero limb, which
could lead to crashes or miscomputation later on.
* Fixed some minor documentation issues.
Features:
* Workarounds for various issues with Mac OS X's build tools.
* Recognise more IBM "POWER" processor variants.
Changes between GMP version 4.2.X and 4.3.0
Bugs:
* Fixed bug in mpz_perfect_power_p with recognition of negative perfect
powers that can be written both as an even and odd power.
* We might accidentally have added bugs since there is a large amount of
new code in this release.
Speedups:
* Vastly improved assembly code for x86-64 processors from AMD and Intel.
* Major improvements also for many other processor families, such as
Alpha, PowerPC, and Itanium.
* New sub-quadratic mpn_gcd and mpn_gcdext, as well as improved basecase
gcd code.
* The multiply FFT code has been slightly improved.
* Balanced multiplication now uses 4-way Toom in addition to schoolbook,
Karatsuba, 3-way Toom, and FFT.
* Unbalanced multiplication has been vastly improved.
* Improved schoolbook division by means of faster quotient approximation.
* Several new algorithms for division and mod by single limbs, giving
many-fold speedups.
* Improved nth root computations.
* The mpz_nextprime function uses sieving and is much faster.
* Countless minor tweaks.
Features:
* Updated support for fat binaries for x86_32 include current processors
* Lots of new mpn internal interfaces. Some of them will become public
in a future GMP release.
* Support for the 32-bit ABI under x86-apple-darwin.
* x86 CPU recognition code should now default better for future
processors.
* The experimental nails feature does not work in this release, but
it might be re-enabled in the future.
Misc:
* The gmp_version variable now always contains three parts. For this
release, it is "4.3.0".
This changes the buildlink3.mk files to use an include guard for the
recursive include. The use of BUILDLINK_DEPTH, BUILDLINK_DEPENDS,
BUILDLINK_PACKAGES and BUILDLINK_ORDER is handled by a single new
variable BUILDLINK_TREE. Each buildlink3.mk file adds a pair of
enter/exit marker, which can be used to reconstruct the tree and
to determine first level includes. Avoiding := for large variables
(BUILDLINK_ORDER) speeds up parse time as += has linear complexity.
The include guard reduces system time by avoiding reading files over and
over again. For complex packages this reduces both %user and %sys time to
half of the former time.
this seems to make the most sense for bulk builds to me.
Should any other platforms fail, we can make a note of it
and exclude them along with Darwin.
Also, add two distribution patches from upstream:
[2008-11-09]
When calling mpf_set_str (perhaps indirectly via
mpf_init_set_str or mpf_inp_str, or via the C++ interface) with the
argument for the base set to 0, any exponent will be ignored.
[2008-11-08]
The mpf_eq function sometimes compares too few bits, not
just too many (the latter is documented).
This might lead to precision loss. When the experimental
--enable-nails feature is enabled at the same time --enable-cxx is
enabled, make check fails. This failure is actually due to bugs
in tests/cxx/t-prec.cc, which makes it use mpf_eq incorrectly.
This patch makes mpf_eq compare the right number of bits,
neither too few, nor to many.
The patch also fixes the test case, and documentation.
(not strictly necessary, but I figured I'd keep in line with
their recommendations)
PKGREVISION++
Changes between GMP version 4.2.3 and 4.2.4
Bugs:
* Fix bug with parsing exponent '+' sign in mpf.
* Fix an allocation bug in mpf_set_str, also affecting mpf_init_set_str, and
mpf_inp_str.
Speedups:
* None, except that proper processor recognition helps affected processors.
Features:
* Recognize new AMD processors.
Changes between GMP version 4.2.2 and 4.2.3:
Bugs:
* Fix x86 CPU recognition code to properly identify recent AMD and Intel
64-bit processors.
* The >> operator of the C++ wrapper gmpxx.h now does floor rounding, not
truncation.
* Inline semantics now follow the C99 standard, and works with recent GCC
releases.
* C++ bitwise logical operations work for more types.
* For C++, gmp.h now includes cstdio, improving compiler compatibility.
* Bases > 36 now work properly in mpf_set_str.
Speedups:
* None, except that proper processor recognition helps affected processors.
Features:
* The allocation functions now detect overflow of the mpz_t type. This means
that overflow will now cause an abort, except when the allocation
computation itself overflows. (Such overflow can probably only happen in
powering functions; we will detect powering overflow in the future.)
When using mpf_set_str, mpf_init_set_str, or mpf_inp_str
with a base > 36, the supplied base will actually be ignored,
and the exponent 0 will be supplanted. [2007-12-10]
All tests pass on NetBSD/amd64 4.99.58.
Bump revision.
to build a "fat" binary on (${MACHINE_ARCH} == "i386" && ${OPSYS} != "Darwin").
This isn't enabled by default now, but it might be useful for build builds,
as it uses cpuid at runtime to choose the most appropriate assembler code.
Insofar as I can tell, no BUILDLINK_A[BP]I_DEPENDS bump is necessary.
NOTE: Support for Darwin-*-i386 and Darwin-*-x86_64 has been improved,
see below. Thus, I recommend that we test building the assembler code
on Darwin at some point.
Changes between GMP version 4.2.1 and 4.2.2:
* License is now LGPL version 3.
Bugs:
* Shared library numbers corrected for libcxx.
* Fixed serious bug in gmpxx.h where a=a+b*c would generate garbage.
Note that this only affects C++ programs.
* Fix crash in mpz_set_d for arguments with large negative exponent.
* Fix 32-bit ABI bug with Itanium assembly for popcount and hamdist.
* Fix assembly syntax problem for powerpc-ibm-aix with AIX
native assembler.
* Fix problems with x86 --enable-fat, where the compiler where told to
generate code for the build machine, not plain i386 code as it should.
* Improved recognition of powerpc systems wrt Altivec/VMX capability.
* Misc minor fixes, mainly workarounds for compiler/assembler bugs.
Speedups:
* "Core 2" and Pentium 4 processors, running in 64-bit mode will get a
slight boost as they are now specifically recognized.
Features:
* New support for x86_64-solaris
* New, rudimentary support for x86-apple-darwin and x86_64-apple-darwin.
(Please see http://gmplib.org/macos.html for more information.)
Changes between GMP version 4.2 and 4.2.1
Bugs:
* Shared library numbers corrected.
* Broken support for 32-bit AIX fixed.
* Misc minor fixes.
Speedups:
* Exact division (mpz_divexact) now falls back to plain division for large
operands.
Features:
* Support for some new systems.
Changes between GMP version 4.1.4 and 4.2
Bugs:
* Minor bug fixes and code generalizations.
* Expanded and improved test suite.
Speedups:
* Many minor optimizations, too many to mention here.
* Division now always subquadratic.
* Computation of n-factorial much faster.
* Added basic x86-64 assembly code.
* Floating-point output is now subquadratic for all bases.
* FFT multiply code now about 25% faster.
* Toom3 multiply code faster.
Features:
* Much improved configure.
* Workarounds for many more compiler bugs.
* Temporary allocations are now made on the stack only if small.
* New systems supported: HPPA-2.0 gcc, IA-64 HP-UX, PowerPC-64 Darwin,
Sparc64 GNU/Linux.
* New i386 fat binaries, selecting optimised code at runtime (--enable-fat).
* New build option: --enable-profiling=instrument.
* New memory function: mp_get_memory_functions.
* New Mersenne Twister random numbers: gmp_randinit_mt, also now used for
gmp_randinit_default.
* New random functions: gmp_randinit_set, gmp_urandomb_ui, gmp_urandomm_ui.
* New integer functions: mpz_combit, mpz_rootrem.
* gmp_printf etc new type "M" for mp_limb_t.
* gmp_scanf and friends now accept C99 hex floats.
* Numeric input and output can now be in bases up to 62.
* Comparisons mpz_cmp_d, mpz_cmpabs_d, mpf_cmp_d recognise infinities.
* Conversions mpz_get_d, mpq_get_d, mpf_get_d truncate towards zero,
previously their behaviour was unspecified.
* Fixes for overflow issues with operands >= 2^31 bits.
Caveats:
* mpfr is gone, and will from now on be released only separately. Please see
www.mpfr.org.
and add a new helper target and script, "show-buildlink3", that outputs
a listing of the buildlink3.mk files included as well as the depth at
which they are included.
For example, "make show-buildlink3" in fonts/Xft2 displays:
zlib
fontconfig
iconv
zlib
freetype2
expat
freetype2
Xrender
renderproto
RECOMMENDED is removed. It becomes ABI_DEPENDS.
BUILDLINK_RECOMMENDED.foo becomes BUILDLINK_ABI_DEPENDS.foo.
BUILDLINK_DEPENDS.foo becomes BUILDLINK_API_DEPENDS.foo.
BUILDLINK_DEPENDS does not change.
IGNORE_RECOMMENDED (which defaulted to "no") becomes USE_ABI_DEPENDS
which defaults to "yes".
Added to obsolete.mk checking for IGNORE_RECOMMENDED.
I did not manually go through and fix any aesthetic tab/spacing issues.
I have tested the above patch on DragonFly building and packaging
subversion and pkglint and their many dependencies.
I have also tested USE_ABI_DEPENDS=no on my NetBSD workstation (where I
have used IGNORE_RECOMMENDED for a long time). I have been an active user
of IGNORE_RECOMMENDED since it was available.
As suggested, I removed the documentation sentences suggesting bumping for
"security" issues.
As discussed on tech-pkg.
I will commit to revbump, pkglint, pkg_install, createbuildlink separately.
Note that if you use wip, it will fail! I will commit to pkgsrc-wip
later (within day).
developer is officially maintaining the package.
The rationale for changing this from "tech-pkg" to "pkgsrc-users" is
that it implies that any user can try to maintain the package (by
submitting patches to the mailing list). Since the folks most likely
to care about the package are the folks that want to use it or are
already using it, this would leverage the energy of users who aren't
developers.
