This is an automated email from the git hooks/post-receive script. gregoa pushed a change to annotated tag upstream/0.51 in repository libmath-prime-util-perl.
at a7bde13 (tag) tagging 73e672a0e91e7ccfaf56a12420a1a33664c58e35 (commit) replaces upstream/0.50 tagged by gregor herrmann on Sat Jun 27 19:16:46 2015 +0200 - Log ----------------------------------------------------------------- Upstream version 0.51 Dana Jacobsen (1291): Initial commit Implementation Add MemFree object for better memory control Add factoring text cleanup Add some items to do Updates for 32-bit behavior prime_count uses segmented sieve Updates for next release Speed up factoring a smidgeon, prep for 0.03 Make squfof recurse, and fix some issues with it Miller-Rabin and prob_prime New tests and benchmarks factoring benchmarks Split out bitarray and type header files Redo factor loop Factoring is isprime updates More factoring changes pedantic cleanup Next prime overflow, HOLF factoring, compare with Pari Fix next_prime Factoring updates random primes, and no asserts Try another 64-bit detection method Rename some test files Changes for v0.04 Tweak to mulmod use Add pseudoprime test for MR Note need for more pseudoprime testing Make asm x64 only More pseudoprime testing Add perfect square discriminator Overflow for nth_prime, segment prime_count Add math functions Rewrite prime_count for segments Updates from testing, make nth_prime_approx more accurate for large values Allow random_prime to use overridden rand Fix issue with sqrt rounding Small fixes Many more tests Move from mallloc/free to New/Safefree Update for 0.06 Move Test::More to build vs. prereq; comment changes Update for 0.07 Initial scaffolding for threads Accuracy for math functions Remove simple sieve 7/11/13 presieve added 7/11/13 presieve added Add add_factors Add rational Chebyshev approx for Ei Thread safety improve threading Improve threading Add tests for all_factors Segments were in MB instead of KB. Waaaaay too much memory Add PrimeArray tied array Fiddle with Makefile.PL Add tests for primearray, move MemFree, prepare for 0.08 release Get ready for 0.08 release Start work on PP code Pure Perl working on all tests Big speedups for pure perl code, though still too slow threading and PP changes Wasn't correctly selecting win32 Documentation Add documentation PP sieve benchmark, update tests, MR always returns 0 for even input Update timings, fix test plan Add some more implementations Add simple SoA (along with rant) Add some PP benchmarks Style change for is_prime -- no performance difference Speedup for PP is_prime and factor Speedup for powmod/mulmod (helps factor, isprime, etc.) Cleanup random_prime a little Speed up random_ndigit_prime a little for 9+ digits Adjust comments Remove old renamed file Forgot prototype -- could be cause of test failure Turn on bignum support for PP code Lots of bignum support Do a mod before power operations -- avoids fp exception for very large bases on small n Many changes for bignum support Test euler_phi for small values Note some bignum things that should be done Lots of bignum changes, new tests, update version number More bignum fun Benchmark both XS and front end Add BPSW primality test, hooray Miller Rabin returns 0 or 1 only. Export strong Lucas pseudoprime function Add missing test file to MANIFEST Bigint enhancements Doc updates Full bigint support, add -bigint to import list to turn off Merge conflicts? Restore Util.pm from $&^&$ up merge Update for older Perls and bigint Add tests for strong lucas pseudoprimes 3x speedup for bigint factor and primality Redo random primes, add Maurer's algorithm. random_ndigit_prime needs work. Updates for portable bigints and random primes More random prime work More random prime changes Minor updates Tests for random primes, fix for random primes on 5.6.2 Recommend Math::BigInt::GMP and try to use in tests more fun with bigints and random primes Use the MPU:GMP module if we can Fix an issue with random primes Minor GMP related change for performance and doc GMP and bigint stuff fix POD issue Work on factoring a little Factoring speedups Change %lu to use unsigned long Use GMP primes function. Doc tweaks Update TODO, add Math::Big to performance Update performance number with Math::Primality estimate Merge branch 'master' of github.com:danaj/Math-Prime-Util Strip out the prime_count and nth_prime bounds and approx from C code Minor updates for release Tweak native precision factoring sequence Tweaks, Dusart 2010 bounds, documentation overhaul Big documentation update for PP Check main::rand() existance at runtime Turn off threading tests for Cygwin Change exp2 to pow, and incorporate inverse into exponent Wrap inverse into exponent in Riemann Zeta evaluation Bump version number Compile issues, especially MSC Add a prototype, remove an unused variable rearrange math to keep precision in some cases Change tests for next/prev 0..3572 to do 2 tests instead of 7146. Speeds up a lot, and in theory should give us the same debug info bigint / bignum changes Note the changes for today Add racing SQUFOF Racing SQUFOF & other UV factoring changes Turn off racing squfof in main routine Export RiemannZeta function Update versions for next release Call GMP prime_count if available Add prime_set_config, add assume_rh, use Schoenfeld bounds Remove done_testing() Some changes for valgrind -- nothing big Add primorial and pn_primorial Use long doubles for possible better precision Fix issue with string input just a little larger than ~0 Allow trial_primes to be more intelligent instead of always forcing true trial division Better switch between segment and trial methods for very large end values If using the GMP module for primes(), convert the result to BigInts Initial primes.pl implementation Add primes.pl script Add Lucas, Fibonacci, and Palindromic primes, plus speedups Test harness for primes.pl script, comparing to OEIS data Fix up primorial and pn_primorial for bigints Major rewrite of primes.pl, and add more filters Small speedups Next round of primes.pl mods More primes.pl changes, and script test does multiple filters Recommend Math::BigInt::GMP -- it can speed things up a LOT with big numbers Enhancements to primes.pl tests Minor issues with huge inputs Add Lucky primes, make Cuban primes via a generator instead of filter (hugely faster) Speed up powmod a tiny bit Use 2 MR bases for more numbers Rewrite p-1 factoring, enhance racing SQUFOF, switch to racing SQUFOF in factor Simple primality proving added (the GMP code is much better) Add option to ensure all output has gone through primality proving Fix compiler warnings Enable Travis CI Changes for p-1 factoring Add and enhance examples, add bin/factor.pl Add documentation for prime_set_config Fix prime_count bug (added to test suite), add Lehmer prime count Use Lehmer method for big prime counts Comments and a small speedup for Lehmer More Lehmer improvements (faster, lower memory) More Lehmer speedups and memory enhancements Don't run the 64-bit edge case on broken 5.6.2 No timing messages from Lehmer, and reduce threshold since it's faster now Relax constraint on bignum version. It works fine with 0.22 Add AKS primality test Update some examples Skip the AKS tests on 32-bit machines. They're way too slow in PP. Have nth_prime use Lehmer prime count on lower bound. 100x speedup for big numbers. Update documentation Fix prime count issue and make standalone. Add Meissel method. Ready for next release Add 'verbose' config option Have verbose option set GMP verbosity also Increased precision for bignum zeta and R Update MR bases for deterministic tests Use references in primes.pl for performance. Lucky primes take less memory. Move mulmod, powmod, etc. to separate file. A few changes. Let AKS in XS work with larger inputs. Split AKS out of isprime tests. Extend AKS from halfword to fullword. Add jordan_totient, divisor_sum. Slight speedup for euler_phi and moebius. Streamline random prime test, add more testing for nbit and maurer random primes Streamline some expensive tests, move bignum Zeta/R to separate file Changes file has test and zeta changes mentioned Skip 15+ digit tests on broken 5.6.2. Workaround bug in Math::BigInt in extra tests. Fixup 5.6.2, and some li and Ei range cases Tweak MR test, add environment variables to disable XS and GMP Add PP Lehmer prime count, including nth_prime speedup. Fix some no-XS-with-GMP issues. Speed up PP Lehmer by 10-100x at the expense of memory Fixes for various compile / test issues Update for release, and standardize infinity processing Modify tests to handle Win32 infinity also Missed an inf Fix 32-bit issue with lehmer Work around broken 64-bit Perl 5.6.2 Use Math::MPFR if possible for Ei/li/Zeta/R functions. Huge speedup and accuracy gains for BigFloats. Tests for primorial, jordan_totient, divisor_sum Make random_*_prime do bignums when passed Math::BigInt objects Random prime updates Major changes to random primes. Return BigInts for big results on primorial and random primes. Update tests for new functions, and Perl 5.6.2 workarounds Revamp internal rand system for random primes Swap system/user rand Fix issue with Math::BigInt::GMP on 32-bit Win32 Fix some horrible breakage on 32-bit Better R accuracy with multiple calls Remove an extra loop from PP M-R Work around randbits >= 32 on 32-bit system Perl 5.8.2 gets 12345**4 wrong, change to 12345*12345*12345*12345 Don't use last bit of drand48 MSC doesn't use inline. In theory __inline should work. Update timings for prime_count Make sure we don't try to use really old versions of MPFR Very small optimization for PrimeArray. Lighten threads testing for normal case (helps some slower machines), add more for release testing Allow turning off MPFR. Run travis tests on dynamic dependency variations. Install mpfr before Math::MPFR Speed up Travis CI Unroll inner loop of sieve for another 20% speedup Allow primes.pl script to eval simple math expressions Allow an expression to eval to 0 Silence perl critic for expression eval Revamp rand internals yet again. Sadly also a rand API change. Update rand function in test - Win32 32-bit needs mod instead of and Update versions and dates for release Tweak rand setup a little Don't run the long AKS test on any 32-bit platform Fix old Solaris compilation Log some more bignum info to get a clue why old ActiveState's BigInt::Calc barfs Round the perfect power test, note its faults Add random_strong_prime, update to 0.18 Travis test with all features as well as deselections Update MR bases Fix bignum / Calc issues (things turning into BigFloats) Add tests for is_provable_prime Fix more bignum issues Better test coverage More bignum and 5.6 changes Update MR bases Update M-R bases Modify validation to lower overhead Prepare for 0.19 release Speedup for PP AKS, turn off tests on 32-bit machines Change PP perfect square test New perfect power implementation Bump version for release Add function def to example Add circular primes to primes.pl (OEIS 68652, also in Project Euler problem 35) Speedups for Pillai primes Minor tweaks Spelling corrections in documentation Add Euler totient range, speedup divisor_sum Switch to Bytes::Random::Secure New M-R bases Range Moebius function. Fast Mertens function. Ranged mobius function Update documentation to reflect the change to Bytes::Random::Secure Faster Mertens Add Mangoldt function Performance enhancements Testing changes for release Final changes for release Remove Algorithm::Diff dependency, as it isn't used any more Totient and Mobius changes, move factor loop out of XS.xs Switch to pthreads condition variables Squash some compiler warnings Compiler warnings and coverage Put segment mutex back move scripts Turn thread tests on for Cygwin and NetBSD. Add simple RiemannZeta test cases Getting ready for release Optimizations for von Mangoldt function Merge Akron's fix to dates in Changes Slight doc change for moebius/phi/lambda/M Change XS Zeta code Update PP Zeta -- much better now Add consec int lcm, documentation changes Fix return value with consec int lcm and MPU::GMP Add binary search for nth_prime, for inputs > 2e11 Minor updates Add Chebyshev theta and psi functions 10x speedup for divisor_sum Test coverage and small AKS speedup Start to add LMO prime_count Change type of mobius return. Some different ranged algorithms in comments Move totient range to util.c, speed it up a little bit Fix divisor sum on bigints New data structure for phi(x,a): faster prime_count In-place phi(x,a) merge Minor updates for release Fix 32-bit issues Fix segmented euler_phi Some more speedups More lehmer tweaks Set smaller preseive with standalone primesieve Improvements for PP prime_count Minor updates for release Move a test into xt directory Add table sieve, fix off-by-one in Lehmer, change sieve/advanced transition in Lehmer Merge branch 'master' of github.com:danaj/Math-Prime-Util Newest M-R bases Change to unified method for small is_prime, next_prime, prev_prime, prime_count Add xt test to do exhaustive small is/next/prev tests Update for release Documentation tweaks Speed up p-1 stage 2 New internal macro to loop a..b using primary sieve Forgot to skip one of the tests with broken 5.6.2 Update for primesieve 4.2 Documentation update, including a giant SEE ALSO section Tweak factoring and factor tests based on coverage analysis Documentation changes Speed up rsqufof a little, use for 19 dig semiprimes Allow turning on/off gnufactor or pari Travis-CI change in installation order Add large bases to pseudoprime tests Move the a=1 pbrent into XS.xs instead of factor.h Move to next version number Documentation updates Fix: Was checking wrong BigInt variable and loading a second backend. Primality verification (needs documentation) Remove Data::Dump used for debugging Add tests for verify_prime Work on pure perl factoring Add primality certificates, elliptic curve start Fix misspelled proof name PP: add simple ECM factoring and BLS75 primality proof Documentation formatting Projective ECM for pure Perl Primality proof updates Add GMP-ECPP verification example Tweak projective EC point Fixup for 5.6.2 Allow factor.pl to parse many MPU functions Move a couple of the bigger primality proofs to RELEASE_TESTING -- slow without GMP Fixes for final release Fix spelling error (RT 84782) Speedups for pure perl factoring of small numbers Fix filename of ecpp verification, add sage version Fix error with big aks inputs (order would roll over) Must compare n and q as bigints primality certificate edge cases Convenient primality proof random test Merge branch 'master' of github.com:danaj/Math-Prime-Util Allow BLS75 theorem 7 n-1 proofs pure perl factoring tweaks Merge branch 'master' of github.com:danaj/Math-Prime-Util Small speedup for composites in is_prob_prime/is_prime Increase coverage Remove dead code Add spelling test and do some spelling adjustments Add more test coverage Let primes.pl accept +N as a second argument More test coverage More verbose/error output consistency for verify_prime. Tests. Use XS initial validator Make is_prime go to is_prob_prime after initial tests New 3-base solution Adjust tests to work around Perl 5.6.2 quirks Push full primality test into PP from Util.pm Invert XS/Perl relation for is_prime, is_prob_prime, next_prime, prev_prime Make sure test works on 32-bit platforms Rewrite PP Lucas code. About same speed, but simpler. Updates for release, including docuemntation and direct C->MPU:GMP calls Minor updates for release More efficient ranged Moebius code Add forprimes multicall Backward compatibility changes around multicall Fix memory leak in forprimes. Document it. Fix Win32 threading breakage Update primearray and add a simple iterator Document the prime iterator tests for forprimes and iterator Update for release Fix some multicall portability issues (5.8.2) Add new files to MANIFEST ranged moebius was depending on char being signed. Fix. Thanks Debian testers. segmented sieve infrastructure, and have forprimes use it. Speed up primality proof tests on non-64-bit machines Speed up primality proof tests on non-64-bit machines Add strong Lucas test Don't use MULTICALL yet -- memory leak Add two more probable prime tests Add standard Lucas test Transition some functions to forprimes Fixes for 5.6 (sigh) Make sure non-XS can use forprimes properly Update some benchmarks and examples Code comments about primality testing and the non-exported F-U test Fix prime_count_approx being really slow for > 10^36 without Math::MPFR Workaround for MSVC compiler Try to work around old Activestate crud Add Lucas sequence and have all 3 Lucas tests use it Change to submod and mulsubmod for better portability Fix a refcount GMP callback issue, and add test for it Remove unused variables, set reproducible random results in a benchmark Add F-U compositeness test, and tighted ECPP certificate conditions Use Data::Dump::dumpf Add lucas_sequence Add PP code and simple test for frobenius_underwood probable prime test Documentation updates Add braindead p+1 factoring Speedup for P=1,Q=-1 Lucas sequence, which is half the standard/Strong Lucas test cases Make sure n is odd to get exact results from lucas_sequence Update Lucas tests Almost extra strong lucas prps Cleanup is_prob_prime Rewrite some performance comparison text Turn on gcc's uint128 types if we have a new enough version Update some benchmarks Update todo list Update p+1 factoring, add to tests Combine lots of functions with common signatures fixup AES Lucas test Merge pull request #2 from dsteinbrunner/patch-1 Clean up Lucas, update PP is_prime, remove an test func that got put in XS.xs, add documentation Changes file follows CPAN::Changes::Spec format Not all gcc targets support uint128, e.g. Atom Fix case of forprimes with begin = 6 returning 5, and add test signed chars fix for defined-out LMO function Tested multicall with 5.8, 5.10, 5.16, 5.17s, 5.18 -- looks good so just use it. Add cpanfile is_prime is faster now, so SophieGermain is over 2x faster INCOMPLETE BROKEN : Switching to new text proofs, part 1 Add primality proving submodule Switch to new text proofs, part 2 Add standalone verifier (from MPU::GMP) Small grammar updates Add x86_64 asm for addmod from Kruppa Add more limitations on __int128 use Add proof documentation to TODO list Update primality certificate documentation 10-15% speedup for ranged totient internals Patch up silly MPU::GMP cert mistake Updates for random_nbit_prime and documentation Add random_proven_prime, modify random_nbit_prime Add carmichael_lambda; use F&T A2 for random_nbit_prime Turn off F&T A2, need to make sure top bit is set Update for new M-R base from Panasiuk and Izykowski Remove code for F&T A2 -- needs fundamental change to have nth bit set Add BLS75 paper to references list Add a couple more CPAN mirrors, as the Travis-CI one has lost data Redo mirrors Small updates for prime array Add output comments to primesarray bench Update a couple examples Speedup Maurer slightly by going straight to is_prob_prime with GMP Use MPU::GMP::pn_primorial if possible Add ifs to powmod initial mod; add verbose to factor.pl Merge branch 'master' of github.com:danaj/Math-Prime-Util Allow parallel phi sum by using critical section around phicache Add Pk_2 function, simplify some code Add speedup for divisor count Merge branch 'master' of github.com:danaj/Math-Prime-Util Revised mulmod x86_64 asm, work with Wojciech Izykowski Merge branch 'master' of https://github.com/danaj/Math-Prime-Util Move primality functions to new file. Use Monty routines Primorial primes print a bit faster standardize variable names Add znorder function for multiplicative order Fix sluggish validation of AES increment for common cases Add Monty math for Frob-Underwood, AES Lucas, and one case for standard Lucas Divisor sum with integer second arg Speedups for divisor_sum Montgomery reduction for all Lucas tests Perl 5.6 fixes Revamp input processing to accept bigints Cleanup Tweak sieve to make more efficient for large bases Add prime iterator object Test for edge cases, fix some near-maxuv issues Use mpu method to get new oo iterator Minor cleanup ECPP verification: floor(2*sqrt(n)) instead of 2*floor(sqrt(n)) Speedup for random_nbit_prime Make better comparisons for next/prev test Updates for release Add stopword for newest spellchecker Remove trailing whitespace Speedup lucas sequence => speedup BPSW test Dynamically load BRS; PP random_nbit_prime speedup Merge branch 'master' of github.com:danaj/Math-Prime-Util Fix pure perl lucas sequence with Q=-1 Redo benchmark section Add miller_rabin_random; add primality testing opinions to documentation Make factor consistent in scalar context Merge branch 'master' of github.com:danaj/Math-Prime-Util Cache random function for less overhead Make sure strtoull is used for 64-bit when possible Adjust some benchmarks Merge branch 'master' of https://github.com/danaj/Math-Prime-Util Remove all mirrors from travis config Simplify some of the randomness loop code Make Test::Warn a suggested dependency -- no longer required Only use TEST_REQUIRES with new enough MM Merge branch 'master' of github.com:danaj/Math-Prime-Util delete TEST_REQUIRES after merging Update Changes Example for abundant, deficient, perfect numbers Tiny PP factor speedup Iterator object tests and fixes for next/prev prime behavior near native boundary Update iterator object value_to_i* Documentation update Updated LMO and Lehmer internals, switch to LMO Silence compiler warnings on 32-bit Little updates before release fix some prime count performance and openmp cases Add XS routine for ([p1,p2,p3],[e1,e2,e3]) return from factor Lucas sequence should return a bigint if arg is a bigint Add liousville, factor_exp, partitions. all_factor includes 1,n. Internally use factor_exp in many places. Fix tests for all_factor, add liouville stopword Update partitions function Document partitions, tweak implementation for a little more speed Tests for partition() Add tests for liouville function Add tests for factor_exp and liouville Update version numbers Comment changes for part Zeta updates Zeta accuracy test in xt/ Keep BigInt accuracy for Riemann Zeta/R Speedups for pure perl testing pondering pedantic perlcritic palaver exp_mangoldt in XS by default Improve test coverage Fix PP is_pseudoprime; Fix test issues Use Lehmer for small prime counts (below 8000 million), LMO for large Tweaks to LMO performance Hash->Scalar. Add peek() and __iter__() Update with some more iterator examples Segment chebyshev_theta/phi, slight performance for ranged totient Tiny efficiency for Mertens Make sure large primorials don't overflow sigma_0 Make sure we don't overflow large divisor counts on 32-bit machines Change C99 comments that snuck in Quote 0 in two tests, 5.8.2 use bigint treats 0 = NaN Make all tests run with 5.6.2 (with lots of skipping) Remove unused feature requirement Revise prime count table code, speedup 3.6M to 6M update for release Update for release Remove 64-bit test for 32-bit machines Update to next version number Switch small primes and lpf arrays to uint32_t Wrap large constant in UVCONST Add the extended prime count test For 64-bit, primei must be a UV instead of uint32_t auto bigint, round 1 Tests hit by Math::BigInt::GMP bug. Work around. Next try at consistent type upgrades for nth_prime_upper/lower Add commentary to measure_zeta_accuracy test Use XS prime count for small approx/lower/upper Changes for 32-bit 32-bit fixes Extended LMO algorithm Put icbrt in util.h Move phi_total to sieve struct Adjust M factor: save a little memory, gain a bit of performance Update version number Just skip 64-bit testing on 5.6.2 Remove two tests obsoleted by new BigInt behavior Add nth prime extra test Give the M adjustment a steeper slope Change nth_prime to use inverse Li+Corr instead of inverse R Restructure step 7 phi summation to streamline inner loop Protect against overflow Documentation updates Back off expensive PP prime count testing unless extended testing Thanks TravisCI -- need to count for the extra test when skipping Final updates for release Add some missing files to manifest Microsoft VC isn't C99, sigh fix LMO previous prime (sieve) when very near 32-bit limit Measured mu+lpf methods, adjusted comments A fix for prime counts > 18446744056529682436 Comment changes Ensure M=alpha*n3 doesn't underflow Documentation change Add some 5.6.2 workarounds Microsoft compiler workarounds 32-bit fixes, documentation updates Speedup for Legendre Pi (use small phi) LMO phi sieve is UV-word based instead of 32-bit-word bases Documentation updates; popcnt updates for LMO Remove trailing white space Have 32-bit use parallel bit count instead of 8-bit table Documentation updates _divisor_list in XS, use it. Add simple forcomposites. first round of changes for factor(1) Second round of factor(0),factor(1) changes typo in all_factors alias Add fordivisors, forcomposites forcomposites using sieve Fix some compiler warnings forcomposites UVMAX edge cases Mapes => table for small phi Revert back to SvREFCNT_inc for old Perls Lehmer update cache settings New sieve code for finding segment composite inc/masks Performance changes for prime sieve Fix spelling error (RT 91562) Fix macro comparison, new benchmarks Line too long in synopsis Try to do fast word counts on 64-bit machines Some performance tweaks Remove dead code; convert count to segment start/next/end Convert segment_primes to segment start/next/end Speedup for basic sieve walking Fixup for earlier sieve count change Switch to primesieve 5.0 Add sign multiply to caching phi for a=0,1 Don't use GCC's builting popcount unless we have asm. It can generate awful code. Small comment changes Use faster bit counting for sieve part of nth prime Fix _XS_legendre_phi for new table phi Use presieve/test for large segment bases. Much faster and less memory. Wrap isqrt in #define so only included when needed Switch znorder to factoring Carmichael Lambda New factor_exp function Add kronecker, totient, carmichael_lambda, znprimroot Updated validate_int; znprimroot, carmichael_lambda, kronecker Add kronecker, znprimroot; XSify carmichael_lambda Move znorder to XS; add tests for znprimroot Fix PP test issues Huge speedup for znprimroot with non-cyclic input Allow string numbers to have a leading + sign Make sure PP znprimroot handles non-cyclic nputs quickly Speedups for Legendre/Meissel/Lehmer/LMOS old prime count routines Add tests for kronecker Merge pull request #4 from bulk88/master Merge branch 'master' of github.com:danaj/Math-Prime-Util Add moebius/exp_mangoldt to util.c Shuffle functions between XS/Util/PP; significant XS changes Cleanup compiler warnings Change back to dTHX for now Hacky workaround to for MPU:GMP not understanding leading plusses Remove redundancy in inner loop of forprimes/forcomposites/fordivisors Merge pull request #5 from bulk88/master Move divisor_sum to XS; more XS interface cleanup Wrap large constants in UVCONST for 64-bit Perl on 32-bit machines More XS changes, plus small optimizations from bulk88 Workaround for 5.6.2 PP Merge XS changes from bulk88 Simplify my_svuv/my_sviv Merge branch 'master' of https://github.com/bulk88/Math-Prime-Util into bulk88-pr6 Modifications to previous merge Stop checking return of New/Renew and its propagations Don't compile Legendre/Meissel/Lehmer/LMOS prime counts by default Move factor_exp and divisors to XS->Perl, merge with factor Merge pull request #7 from bulk88/master Merge euler_phi and moebius XS More updates from bulk88 suggestions Next set of XS changes Input validation XS tweaks from bulk88 Fixes for 5.6.2 Use constant Input validation and better 5.6.2 support Fix threading argument Work around RT 49569 (Math::BigInt::FastCalc) Add Kronecker tests Add bigint tests for new functions Add tests for legendre_phi PP legendre phi wasn't making enough primes via public API nth_prime for legendre phi, not next_prime Add test for t/81-bignum.t no-clobber Update version number and date Move prime_count, nth_prime to XS More XS->Perl changes. Add constant SV* returns for -1,0,1,2. Modify const SV for threads Another case of RT71548 Change sieve macro advance to not read ahead (valgrind) Change comment to match current number of array items Add gcd function. Fix forcomposites. Fix test for GMP function definition. Refactor forprimes Use memcpy doubling -- fewer calls needed Go back to 200M threshold vs. 2M for looking in cache use constant with hash. Saves a tiny bit of startup time Removed old SQUFOF code. Faster is_perfect_square. Streamline trial division pre-factoring. Tweaks to SQUFOF, pbrent, and strategy Speedup for palindromic primes Make is_bpsw_prime a first class function (though currently not exported) is_perfect_square doesn't use isqrt mapes -> tablephi. Caching legendre phi Speed up some testing. Move znorder to PP.pm Add lcm. Better PP legendre_phi. Prune PP documentation. Try to use Pari in benchmark, but silently remove if not available Performance tweaks (small) for bigints Performance updates for no-GMP, focusing on making test suite run faster. Right number of extra tests, remove references to old _XS_next_prime First cut znlog Merge pull request #8 from bulk88/master Remove flags on get_cv Back off internal PP prime count in test, to help out very slow machines Double speed of AKS for small numbers on slow-modulo machines. poly_mod_mul checks for overflow in all terms are summed Make sure we can use as an int Adjust use of constant to allow backward compatibility Merge branch 'master' of github.com:danaj/Math-Prime-Util Small aks speedup Move PP AKS to only run with EXTENDED_TESTING. It's horribly slow, a function nobody should practically use for anything, and they would run the C or GMP version anyway. Reduce some variable scoping Trim PP testing Use long double interface for the other three FP functions decl after code fix When doing XS bigint downgrade on old Perl, set the string Perl 5.6.2 workarounds for bigint downgrade Performance for PP, and a few pre-5.8 64-bit workarounds More tests run on 5.6.2 64-bit now Trim some tests, convert some proofs to new cert format, more tests work with pre-5.8 64-bit Turn off default irand in tests that don't care znlog tests and PP placeholder Switch more funcs to long double, and return results as NV Add clz and ctz More clz,ctz,log2floor changes Use new clz/ctz/log2floor Fix arguments for non-public *_factor functions Add ppport.h finally Merge bulk88 XS changes. Add ppport.h use. spacing small performance increase for next_prime Add simple (and likely buggy) Pollard Rho DLP Call Math::BigInt importer only if it hasn't been loaded already Changes for Math::BigInt::Pari (it doesn't play well with others) Remove the old -nobigint flag. Better random ndigit for some old examples. Honor rounds in Fermat Fix stupid cpp typo Push some constants into a shared header file Use constants.h. Simplify prev/next prime. Simplify chebyshev functions. Results in smaller object files. ranged totient faster and uses less memory Adjust PP testing to reflect some functions now being in XS Rearrange factor main loop remove _is_prime7, remove _XS_ header from a couple C functions Minor changes to examples Fix MSC support for ctz/clz, hopefully Simple Pari comparator (random values for functions in Pari and MPU) Pari's nth prime doesn't want 0 Move jordan_totient to XS->PP from Util znorder and znlog adjustments for edge cases Updated Pari compare test LogarithmicIntegral using bignum and no-MPFR returns (1) better results, (2) rounded to a sane number of digits signed/unsigned warning Pari compare test done Adjust comments Adjust MANIFEST Make sure logl(constants) are only done once Simplify next_prime / prev_prime BigInt was upgrading to float in some cases Documentation updates Merge branch 'master' of https://github.com/danaj/Math-Prime-Util Adjust factor recipe, 3-5% faster for large inputs. Adjust primes sub to reflect new segment sieve behavior Simplify primes() Fix 32-bit unsigned long issues Load PP and Math::BigInt only when used. More work needed No validation means we weren't stripping weirdness from input bit counts. Move some functions from Util.pm to util.c. Faster, and reduces startup bloat. Start work on separate PP front end Put back some PP validation, as it does bigint magic Bump version number More movement to PP front end More startup measurements More moving functions out of Util Clean up some random / proven primes Add some more terms to integer zeta table. Speeds up prime_count_approx by 1.5x Small RiemannR speedup Some TODO items are done znlog update Remove the empty function call entirely for non-threads Update some examples Update examples and examples/README Add 1 to skipped 5.6.2 tests Move benchmarks to separate bench directory Adjust yafu tests Move last tests out of examples, into bench and xt Add Porter sequence example Update Changes Remove unused popcnt on 32-bit Minor documentation update Add some Project Euler examples Merge branch 'master' of github.com:danaj/Math-Prime-Util First round: add is_perfect_power, is_perfect_square, is_perfect_cube Add todo about perfect power. Going to change name. Merge branch 'master' of github.com:danaj/Math-Prime-Util Change to is_power, still needs work is_power updates PP is_power with negative arguments Update some PE examples Trimming suggested by bulk88 (croak after New, move memset to just the needed branch) Wrap dlp_prho debugging vomit inside #ifdef DEBUG, as nobody else wants to see that detail Code movement Update PE 357 example Fix 5.6.2 issue with is_power euler_phi(0,n) had wrong [1] when n >= 100 Add inverse totient example PP is_power(n,a) has to deal with 0 return unlike the split XS code Better is_power Merge with master Typo in bigint section More or less initial pbrent based on mulmod speed Add two more PE examples Add perfect power test in factoring Small speedup for is_power(x,2) Add a PE example Only test for squares and cubes in factor. Lowers cost. Add xt ispower test and some more PE examples Slight speedup Small totient speedup Revised example of Math::Factor::XS::matches Add PE example Remove unused line Remove old test and print comment Spelling error: Dedekind Small change to PP znprimroot Doc updates Update version number Updates for AKS Small AKS update Try harder to use lgamma Add some threshold returns to protect vs. messed up rounding modes Custom lgamma so improved AKS works on all platforms ActiveState's gcc can't find builtins. Pre-2005 compiler? Allow threading failures with diagnostics No MS intrinsics with MSC clang ispower test and 5.6.2 borkness Change NaN generation method Speedup for znprimroot Don't use Win32 intrins on WinCE fix sign warning in new znprimroot code 32-bit with Math::BigInt::GMP causes havoc because of its RT71548 Allow some author tests to run better on 32-bit Perl Partitions uses Math::BigInt only when needed logl -> log Bump version Don't use Math::BigInt::Pari for threading tests. Math::Pari + threads = bad juju PP jordan totient, be more aggressive about BigInt args. ActiveState weirdness? Add float.h to includes Bump version number Update PP F-U primality test Don't die in t/04 if platform doesn't have NaN (some activestate) Speedup for exp_mangoldt Documentation updates Speedup exp_mangoldt for large numbers -- primality + power detection instead of factoring Shawe-Taylor random provable primes. No documentation. Shawe-Taylor random proven primes from FIPS 186-4 Don't let ST random prime give up so easily random ST primes streamline the 32-bit case, allow 2-bit to return 2 Update benchmark time Add coverage areas Add twin_prime_count Add twin_prime_count_approx give twin_prime_count a much denser table (92 vs. 5) Expand twin prime table Add nth_twin_prime Merge branch 'master' of github.com:danaj/Math-Prime-Util Add PP nth_twin_prime AKS comments Small nth_twin_prime speedup. xt test for twin_prime_count Add xt nth_twin_prime test Documentation tweaks Add new functions to tests Add PP twin_prime_count_approx Concerns about each Move a big hash to closure instead of sub local Improve test coverage Trim some PP tests for performance Inverse Riemann R for nth_prime_approx Add nth_twin_prime_approx, tighten twin_prime_count_approx Finish nth_twin_prime_approx Make sure nth_twin_prime goes to PP when the argument will overflow the C code Tighten nth_prime_{upper,lower} Edge case for legendre_phi speedups for legendre_phi with very large a Add one more big-a optimization to legendre_phi Add slightly faster ranged totient for start = 0 Fix an old random_prime() bigint case. No effect on proven, ndigit, or nbit random primes. More optimizations for euler_phi(0,n) Version comparison, make no warnings with test versions Don't do test for ST random prime without new Digest::SHA Fix typos Jordan totient even number shortcut Update precalc and timings improve twin prime approximations Add PP requirement for benchmark Better AKS testing, change to znorder from naive order Add numseqs example file, first version Add last sequences to numseqs examples Add PythagoreanHypots to numseqs example add valuation function croak with sane message when lucas_sequence given n < 2 Add lucas_sequence tests New monty modinverse, nice primality speedup Update benchmark Improved code for mont_prod64 Use 64-bit Montgomery math in primality tests even for 32-bit inputs Full trial divide of small factor inputs Another couple percent speedup for very small factor inputs Fix forcomposite with endpoint past max prime forcomposites, edge cases and performance Speedup for znprimroot Replace complicated invmod with simpler version that works properly for all n Bump version number Add invmod znorder arguments in PP code invmod edge cases, add tests is_pseudoprime / is_strong_pseudoprime Compiler warnings Simple forpart Update to forpart forpart now does restricted partitions forpart better interface, add docs, PP, tests vecsum() Some forpart restriction speedups Validate input in PPFE for gcd, lcm, vecsum Add binomial(n,k) More aggressive binomial reduction forpart edge cases Make MPFR free cache after eint init entire array to ensure good behavior in n=0 case Extend binomial to negative arguments Full Kronenburg extension for binomial Work around old Math::BigInt bug Add Catalan sequence using binomial Move some examples to vecsum from List::Util::sum Allow GMP binomial Fix for old Math::BigInt and negative args binomial documentation update Add MephistoWaltz, PisanoPeriod, ProthNumbers Add native integer code to PP binomial whitespace and timings BM and BBS examples Allow better optimization for invmod and large modulus Add binomial, valuation, and invmod Updates for release PP next_prime much faster to return first bigint prime. Huge speedup for xt/primes-edgecases.pl vecsum edge case: fix and add test Improve DLP (gcd before invmod) Further DLP rho improvements (speed up znlog) Add BSGS for znlog Add PP BSGS Use naive slot allocator in XS BSGS for nice speedup Push DLP Rho ahead of BSGS 16M Fix PP 0-return cases (e.g. a=1) More znlog speedups 8-bit at a time inverse replacing 2-bit. Switch to unrolled montgomery inverse Alternate RiemannR calculation Alternate RiemannR, fix bignum Li BSGS no longer requires g to be invertible Add CRT to util, add Pohlig-Hellman, use PH in znlog Remove unused variables Add some tests from Pari bug trackers; fix PP znlog with a=0 RiemannZeta and RiemannR updates Add gcdext and chinese, neither are finished yet Finish up gcdext Finish chinese Load Math::BigFloat when required in lehmer Streamline UV setting in forprimes hot loop Merge branch 'master' of github.com:danaj/Math-Prime-Util Fiddle with sieve/partial-sieve crossover Bump segment sizes up. Will want dynamic adjustment Dynamic segment sieve sizes; remove partial sieve Update csrand example Latest Montgomery product change from W. Izykowski Update performance numbers Fix type change in forprimes Update primality times; document weird closure-in-block case Updates to PP next/prev_prime,is_prime7,chebyshev_psi/theta Add release tests for documentation examples Revise PP sieve parsing, uses less memory Small changes for compilation PrimeArray examples added to t/33-examples.t work around bug in 5.20.0 Partial sieve + test for small windows Skip znlog bigint test if not using GMP backend or EXTENDED_TESTING Speed PP pbrent and znlog znorder edge case; work around the usual Math::BigInt::GMP bug range and bigint edge cases Make sure examples test works with 32-bit Spelling Add one more digit to Zeta with MPFR More workarounds for borked 64-bit Perl nth_twin_prime_approx upper limit PP factor methods should skip basic factor when possible reftyped needs to handle negative numbers Back off random strong prime from 190 to 140 bits Use GMP gcdext and vecsum if available RiemannR with huge input should return inf weirdness with vecsum and the GMP hash PP chinese: push lcm*s into bigint loop. Bump strong prime back to 180 bits in test, for 5.6 Add xt/chinese test, fix 64-bit chinese, PP addmod Speed up nth_twin_prime Update release version and date Update upper bound in comment Add MCE and serial versions of fibprime PP optimize for primes with small arguments SternDiatomic added Add tests for chinese with large inputs PP ranged moebius fencepost fix Moebius should try calling GMP directly Add GMP (via PP) processing options for euler_phi and jordan_totient znorder via PP; sort divisors with qsort reftyped should accept undef Fixup some synopsis examples csrand must have bits > 64 Add Micali-Schnorr CSPRNG to examples/csrand Merge branch 'master' of https://github.com/danaj/Math-Prime-Util Add Math::GMP version of CSPRNG example script Comment note that all the commented out tests have passed LMO Many synopsis test pass. Make sure BBS uses at least 3500 bits Add foroddcomposites Remove deprecated: miller_rabin and all_factors Merge branch 'master' of https://github.com/danaj/Math-Prime-Util Very slight speedup for carmichael_lambda small factor and carmichael_lambda speedup TODO alternate prefix sums for LMO Add twin_primes() Simplify PP twin_primes() and make more efficient Tests for twin primes Fix forpart with amax=>0 Add test for forpart with amax 0 Small speedup for twin prime counting Update description Update comment Speed up for Emirps Add use_primeinc config option for random primes Use faster ranged totient when input is < 100 instead of = 0 is_perfect_square uses isqrt Merge the 3 extra-strong-Lucas param selections into one static function Merge branch 'master' of github.com:danaj/Math-Prime-Util HOLF premult; 10% speedup for 49 to 64 bit factoring Update version number Hacks to PP chinese for 5.6.2 Merge branch 'master' of github.com:danaj/Math-Prime-Util Comment about older perl 5.8s Add link to TODO Comment for vecsum Merge branch 'master' of github.com:danaj/Math-Prime-Util Change znorder from Das to Cohen Tweak factoring parameters Fix AIX XLC long double issues Speed up isprime for small inputs on no-montgomery-math machines Tested List::Gen Add forcomb Add ntheory module alias Add forperm Add Algorithm::Combinatorics to see-also section Change from Heap's permutation to lexicographical PP versions of forcomb and forperm Add factorial; add tests for forcomb, forperm, factorial Add ecm_factor as option, finish adding is_bpsw_prime Update version number Test ecm in simple factoring test Version to 0.44_001 Fixes for Win32 and older MPUGMP Redo PP forcomb loop, PP code returns one empty array for k=0. Unused additional Li implementaiton, may be useful for the PP code later forperm 0 and forcomb n.0 return an one empty list Fix spelling error in references Add Fibonacci pseudoprime test to lucas sequences test Only a few Fibonacci pseudoprimes if PP without GMP Update TODO Add basic Frobenius primality test Fix XS call-through for Frobenius Add Perrin pseudoprime, Frobenius has arguments crude PP lambertw function Documentation for Perrin and Frobenius Speedup for divisor_sum(n,k) with large k Add lambertw Cleanup for release tests Add PP Perrin and Frobenius primality tests Tests for more pseudoprimes and lambertw grammar fix in docs documentation for ntheory; provide modules Update docs Upper/lower bounds were having problems with bignums -- fix Last change ignored non-Math::BigFloats Add vecmin and vecmax functions Internal start on bernfrac Add bernfrac and bernreal Tests for bernfrac and bernreal Add caching of small values (under 100) to bernfrac Add Pi; look for better factorials in backend; add bernreal precision Update for Pi Update Pi Add Pi test Compiler warnings Tests for Pi Release tests Add simple tests for ntheory alias Add Pi to reference card Add vecprod Small changes for release Update version number Long double changes Update version number Fix forcomb with k=0, reading out of array bounds More fiddling with NaN, guessing at ActiveState Update version number Add MPUGMP to dep list automatically Fix memory leak in segment primes Redo bernfrac to make sure we load bigint before bigfloat simple PP stirling numbers Add Stirling numbers Workaround broken Math::BigInt::GMP, call GMP stirling Tests and docs for stirling Add PPFE front end for stirling Add simple product tree Edge cases for new PP vecprod Update version number Update Change date Another nan/infinity change Put Carp use in correct package Changes to ZetaBigFloat Add Pi*B_n variant of zeta calculation Speedup forpart with large nmin Documentation fixes Add prime gap verifier For non-objects that overflow UV, do string vecmin/vecmax Speedup for PP divisors Changes for 5.6.2 Trim some time consuming tests Rewrite PP FU test, work around 5.6.2 segfault Small PP speedup Revert prime count approx test change Simplify a bit of zeta bigfloat Speedup for PP factoring: revised trial factor, small input Rho, more initial trial factoring Simplify divisors from factors Doc change to verify-primegaps example Improve test coverage PP: new trial factor, new primorial, update pbrent, update factor Update factor recipe to match ::PP Bump version number Allow different accuracy in ZetaBigFloat RiemannR Fix new Zeta accuracy for certain values divisor_sum changes Make is_pseudoprime act like is_strong_pseudoprime doc change for ntheory vecsum future note and PP shortcut Adjust Changes file Add hammingweight Release testing issues Add vecreduce function popcnt function for 32-bit Update Changes for release Rename popcount to fix NetBSD misbehavior is_power takes optional third argument Update is_power latest change Speedup for AKS with larger-than-half-word inputs Add AKS change to Changes Documentation fixes from Paul Underwood Merge branch 'master' of github.com:danaj/Math-Prime-Util Standalone compilation Use bfloor instead of bint. bint wasn't in pre-2014 Math::BigInt Bump version number standalone factor accepts multiple numbers per line next_prime / prev_prime do GMP/_to_bigint instead of PP code lucas_sequence via XS Add validation / bigint conversion to PP prev_prime and next_prime Move unexported ecm_factor out of root PM Move verify_cert old format conversion into PrimalityProving.pm Make sure bigint next/prev prime returns the input object type Test value needs to be 32-bit safe Allow negative n in is_power Don't use each on array in test primes.pl can take oone argument (implied start = 0) Faster factoring Merge branch 'master' of github.com:danaj/Math-Prime-Util Move uint128_t into ptypes.h Faster AKS on some machines Add is_mersenne_prime Use GMP lucas sequence if we are in PP is_mersenne_prime Fixup 5.8.x bugs Add is_mersenne_prime to ntheory doc Add large Perrin pseudoprime Fix negative is_power for defined a, still need a=0 case Fix ispower with negative input, add some tests is_mersenne_prime change to table for known values Remove unused variables Skip some expensive tests without GMP Move is_provable_prime to XS Update for release Initial lucasu and lucasv lucasu and lucasv Merge branch 'master' of github.com:danaj/Math-Prime-Util Add lucasu/lucasv to Changes file Revise with some new functions Obnoxious Q=-1 speedup for lucasu/lucasv, need to find better solution Add Axler (2014) prime count bounds Comment changes in prime_count_approx comparisons Fix typo in extended test for mersenne primes Small speedup for p-1 factor Typo in example Typo in example comment Very small pbrent performance improvement Merge branch 'master' of github.com:danaj/Math-Prime-Util Compiler warnings More pbrent for general factor search Retweak of recently changed factor parameters Catch some overflow conditions in lucasu/lucasv ispower with neg arguments, workaround Math::BigInt::Pari/old-perl bug Merge branch 'master' of github.com:danaj/Math-Prime-Util Update version number Adjust PPFE for negative is_power inputs primes.pl use lucasu/lucasv instead of hand calc Update version number Use Axler (2013) nth_prime bounds Add harmreal and harmfrac to PP as dev (needs work) harmfrac and harmreal for Harmonic numbers Speed up harmreal Allow prime count approx to work with string arguments harmreal will use MPFR if possible, and push some of its code to PP Fix a PP issue with nth_prime Use inverse prime count for nth prime in XS code Very simple vecextract Add Harmonic numbers nth_prime_lower/upper adjustments Start add sqrtint Ramanujan primes Tests and PP for ramanujan primes Update for sqrtint Adjust documentation New prime count bounds Add commented out Kulsha li code Update last checked Mersenne number znlog updates znlog use simple trial up to 10k (from 2k) detect cycles in dlp_trial, look at order for trial decisions Add single base hashed M-R for 2^32 primality Merge with main Fix for PP nth_prime on bigints Add documentation and PPFE for vecextract vecextract allow bigint and array ref second argument add sqrtl definitions Add missing threads macro use Add nth_ramanujan_prime Add is_ramanujan_prime Add new ramanujan functions to change list vecextract tests More efficient Ramanujan primes Speed up is_ramanujan_prime Remove unused function use Update last double-checked Mersenne exponent One more level of ramanujan prime upper Update for release Fix for old MPFR; faster PP sieving Add is_catalan_pseudoprime spelling, doc fixes for is_catalan_pseudoprime add base argument to foreach sieve macro lucas_seq with even n fixed for XS Even-n lucas_sequence fix for PP Use Aebi & Cairns result for semiprime catalan pseudoprimes Speedup for is_perrin_pseudoprime Fix global cleanup assert with MemFree object For memfree, just return if mutexes are not init. New code for Perrin pseudoprime pretests Allow squares to be Frobenius pseudoprimes Initial sum_primes and print_primes Fix compilation warnings for recent changes segments for sum_primes and print_primes Notice of sum_primes, print_primes functions Updates for sum_primes and print_primes including docs Comments about file handles for printing Update display less often, reduce timing precision displayed Add Khashin's Frobenius test Add some misc xt programs Fix undefined symbol Add PP version of Khashin Frobenius test Add a reset to prho and pbrent so they have better success Small optimization to Khashin code Update Mersenne limit with new GIMPS double check limit Update for release Changes date David Steinbrunner (2): typo fix typo fixes bulk88 (8): restore Win32 support remove dTHX usage don't set global SP to itself after vcallsubn various XSUB optimizations more XS tweaks non-XSUB tweaks fix declaration after code and redefinition warning for Win32 XS tweaks gregor herrmann (1): Imported Upstream version 0.51 ----------------------------------------------------------------------- No new revisions were added by this update. -- Alioth's /usr/local/bin/git-commit-notice on /srv/git.debian.org/git/pkg-perl/packages/libmath-prime-util-perl.git _______________________________________________ Pkg-perl-cvs-commits mailing list Pkg-perl-cvs-commits@lists.alioth.debian.org http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-perl-cvs-commits