Version: 11.5.1-bp154.1.40
* Wed Jun 23 2021 Jan Engelhardt <jengelh@inai.de>
- Update to release 11.5.1
* Fixed bug that prevented compilation on IBM z
* Sun Jun 20 2021 Jan Engelhardt <jengelh@inai.de>
- Update to release 11.5.0
* Added experimental support for AES-CTR mode as an alternative
to ChaCha20 for Pseudo-Random Number Generation.
* Fri May 07 2021 Jan Engelhardt <jengelh@inai.de>
- Update to release 11.4.4
* Improved Karatsuba code for ZZX and GF2EX (as well as the
non-GMP implementation of ZZ).
* Thu Mar 05 2020 Jan Engelhardt <jengelh@inai.de>
- Update to release 11.4.3
* Added KarMul and KarSqr for ZZ_pX (declared in ZZX.h and
implemented in ZZX.cpp). These are not a part of the
documented interface.
* Mon Nov 04 2019 Jan Engelhardt <jengelh@inai.de>
- Update to release 11.4.1
* Fixed bug in new NTL_EXEC_DIVIDE that could manifest itself
when NTL_THREAD_BOOST=off. Existing code that does not
explicitly use this feature should not be affected by this
bug.
* Fixed some namespace visibility issues in the TLS hack macros.
* Sun Sep 29 2019 Jan Engelhardt <jengelh@inai.de>
- Update to release 11.4.0
* The Schoenhage–Strassen FFT for both ZZ_pX and ZZX is now fully
"thread boosted".
* Sat Aug 24 2019 Jan Engelhardt <jengelh@inai.de>
- Specfile modernization.
* Fri Mar 15 2019 Jan Engelhardt <jengelh@inai.de>
- Update to new upstream release 11.3.2
* Fixed a performance issue in the PowerMod function for the
ZZ class.
* Thu Oct 25 2018 Jan Engelhardt <jengelh@inai.de>
- Update to new upstream release 11.3.1
* Fixed a bug that effected image, kernel, and gauss routines
for Mat<zz_p>. These routines did not behave correctly when
the input matrix was zero. Also improved the mat_lzz_pTest
program.
* Wed Aug 29 2018 Jan Engelhardt <jengelh@inai.de>
- Update to new upstream release 11.3.0
* Performance tuned GF2EX arithmetic. Tuned crossovers for
various algorithms.
* Implemented asymptotocially fast GCD and XGCD for GF2EX,
zz_pEX, and ZZ_pEX.
* Thu Jul 26 2018 jengelh@inai.de
- Update to new upstream release 11.2.1
* The low-level "small-prime" FFT (a.k.a., NTT) was rewritten. It
implements a "truncated" FFT, which can speed up polynomial
multiplication by a factor of two, and which mainly eliminates
"jumps" in the running time at powers of two. The new FFT
routines are in fact a bit faster even at powers of two.
* Improved performance of ZZ mul and sqr on small inputs:
mul speedup: 1 limb: 2.5x; 2 limbs: 1.4x; 3 limbs: 1.3x.
* More efficient implementation of low-level butterfly
operations.
- Remove ntl-automake.diff, add no-static.diff.
* Wed Jun 06 2018 jengelh@inai.de
- Update to new upstream release 11.0.0
* Multithreading is enabled. To get these speedups, you have to
call SetNumThreads.
* Thread boosted all cubic-time operations in mat_ZZ_pE,
mat_lzz_pE, and mat_GF2E. This includes: matrix
multiplication, inversion, determinant, kernel, image, and
solving linear systems.
* Thread boosted RandomPrime, GenPrime, and GenGermainPrime.
* New functions: GetWallTime, VectorRandomWord.
Version: 10.5.0-bp150.1.3
* Sat Sep 30 2017 jengelh@inai.de
- ntl-automake.diff: add missing header file for flint
* Sat Sep 09 2017 jengelh@inai.de
- Update to new upstream release 10.5.0
* Faster linear algebra over ZZ_p. Rewrote mat_ZZ_p routines
inv, solve, determinant, gauss, and kernel to be
thread boosted.
* C++11 support / "move" semantics. Iterators and support for
"range based for loops".
* Sat Apr 29 2017 jengelh@inai.de
- Update description
* Tue Nov 29 2016 jengelh@inai.de
- Update to new upstream release 10.3.0
* Marginally improved performance and crossovers for mat_zz_p
multipliplication.
* Retired the zz_pXAltArgument class, which was used for modular
composition in zz_pX. While this has been in in the documented
interface for a few months, it was flagged as being provisional
and subject to change. In place of zz_pXAltArgument, the class
zz_pXNewArgument was added.
* Implementation of a multi-modular strategy for matrix
multiplication over ZZ_p.
* Thu Oct 20 2016 jengelh@inai.de
- Update to new upstream release 10.1.0
* the classical LIP module is finally thread safe
* Wed Sep 21 2016 jengelh@inai.de
- Update to new upstream release 9.11.0
* Improved the effectiveness of the new, faster ZZ to
zz_p conversion
* Added new routines VectorConv for faster bulk conversion from ZZ
and long to zz_p
* Added new routines VectorRandomBnd (see ZZ.txt) and VectorRandom
for faster bulk random number generation.
* Sun Jul 31 2016 jengelh@inai.de
- Update to new upstream release 9.10.0
* Conversions from ZZ to zz_p are now faster, thanks to
preconditioning. Among other things, the CRT-based ZZX
multiplication code is also a bit faster as a result.
* The BasicThreadPool class now guarantees that exec_range
assigns the current thread first=0, and exec_index assigns the
current thread index=0. This makes it easy for a thread to tell
whether of not it is the current thread, which can be
convienient for some applications.
* Fine tuned the interface for SmartPtr and UniquePtr a bit,
including the ability to attach an explicit deleter policy,
which (among other things) makes it easier to implement the
PIMPL pattern using these classes. Unfortunately, some of these
changes introduced some minor backward incompatibilities (but I
doubt anyone will even notice).
* Introduced a new class CopiedPtr, which has a similar interface
to UniquePtr, but which allows copy and assignment. This class
is meant to replace the OptionalVal class, whose use is now
discouraged.
* Sun Jun 19 2016 jengelh@inai.de
- Update to new upstream release 9.9.1
* Introduced ll_type and related routines which perform a
restricted set of operations on a long-long-like type. It can
be implemented via inline asm, and is a cleaner interface and
sometimes faster. On x86-64/gcc platforms, the assembly code
version is used and gives a modest speed boost.
* Introduced PreconditionedRemainder class for faster reduction
of a ZZ modulo a fixed long. This is intended to make Chinese
Remaindering type computations faster. (For the time being,
this is an undocumented feature.)
* Allow p.move(q), where p is a UniquePtr<T>, q is a
UniquePtr<Y>, and Y* converts to T*.
* Added elts() method to UniqueArray and AlignedArray (for
compatibility with Vec class)
* Added get() and release() methods to OptionalVal
* Sun Apr 17 2016 jengelh@inai.de
- Update to new upstream release 9.7.0
* Changes to mat_lzz_p: improved performance of mul, inv, solve
and deterministic routines
* NTLThreadPool is no longer directly accessible: new access
functions are provided
* Got rid of method SplitProblems, and made a more general/abstract
class PartitionInfo
* Fri Mar 11 2016 jengelh@inai.de
- Update to new upstream release 9.6.4
* A new modular composition implemention for zz_pX. This makes
modular composition up to 3x faster, depending on several
factors.
* Improved performance for polynomial factoring over zz_pX using
CanZass, using the improved modular composition routine (above)
and better choice of baby step / giant step parameters. This
leads to a 1.1x to 1.8x speedup, depending on several factors.
* Sat Nov 14 2015 jengelh@inai.de
- Update to new upstream release 9.6.2
* Performance tuning: ZZ_pX and zz_pX keep getting faster
* Upgrade to pseudo-random number generation: replaced the
underlying PRG with Chacha20 (replacing RC4) and the underlying
key-derivation function with a function based on HMAC-SHA256
(replacing an MD5-based function). The new routines are faster
and more secure.
* Wed Jun 24 2015 jengelh@inai.de
- Update to new upstream release 9.2.0
* Performance improvements to zz_pX and Vec<zz_p>.
* Performance improvements to ZZX: implemented asymptotically fast
CRT code for HomMul and more cache-friendly logic.
* Completed the transition away from floating-point arithmetic for
the implementation of single-precision modular arithmetic.
* Fri Mar 27 2015 jengelh@inai.de
- Update to new upstream release 9.0.0
* The interface to the single-precision modular arithmetic
routines has been modified slightly: In particular, on 64-bit
x86/GCC platforms, single precision moduli can now be up to 60
bits, rather than 50 bits. Using larger moduli speeds up a number
of things, like ZZ_pX arithmetic, as fewer primes need to be used
in Chinese Remaindering steps.
* Thu Feb 12 2015 jengelh@inai.de
- Update ntl-automake.diff: add a missing backslash causing not all
headers to be installed. Support mingw a bit better, too.
Support non-C++11 systems.
* Tue Feb 03 2015 jengelh@inai.de
- Update to new upstream release 8.1.2
* Added support for "user defined" FFT primes for zz_p.
* Added explicit constructors corresponding to promotions.
* Thread safety, requires C++11 concurrency.
* Fri Apr 18 2014 jengelh@inai.de
- Update to new upstream release 6.1.0
* Replaced the old template-like macros for vectors, matrices, and
pairs with true template classes: Vec<T>, Mat<T>, and Pair<S,T>.
* There are many new conversions provided.
* Improve speed for single-precision FFT
* Added support for "user defined" FFT primes for <tt>zz_p</tt>.
* Sun Dec 09 2012 jengelh@inai.de
- Ensure that only NTL_ defined appear in NTL/config.h
- Set RPM group according to rpmlint
* Thu Sep 15 2011 jengelh@medozas.de
- Initial package (version 5.5.2) for build.opensuse.org