primecount-8.7

The C formula in Gourdon’s prime-counting algorithm now runs approximately twice as fast, improving primecount’s overall performance by up to 20%. The improved C algorithm is described in section 6 of Easy-Special-Leaves.pdf. Thanks to @ChopperLin the memory usage has been reduced by about 32% in the range [6.5e22, 2.5e26].

ChangeLog

• primecount now requires C++14 or later (only for building, linking still works with C++11).
• nth_prime_sieve.cpp: Fix rare OpenMP multi-threading bug.
• Fix GCC/Clang -Wundef warnings.
• PiTable.cpp: Add multi-threaded primes vector initialization, #134.
• FactorTableD.hpp: Reduce Gourdon D memory using 16-bit prime factor ordinals, #133.
• primecount.pc.in: Add missing libatomic and libquadmath dependencies, #132.
• primecountConfig.cmake.in: Only find OpenMP when required.
• ci.yml: Add pkg-config and CMake find_package() CI tests.
• D_arm_sve.hpp: Fix GCC ARM SVE failed auto-vectorization, #130.
• Fix get_max_x() for x > 10^31.
• CmdOptions.cpp: Detect too many input numbers.
• P2.cpp: Fix potential 64-bit integer overflow.
• phi.cpp: Fix phi(x, a) for x >= 4611686065672028281 which requires primes > INT32_MAX.
• lmo/*.cpp: Fix int32_t primes overflow, #129.
• api.cpp: Fix nth_prime() error message.
• api.cpp: Fix 128-bit nth_prime() for negative inputs.
• main.cpp: Validate 128-bit to 64-bit integer conversions.
• LoadBalancerS2.cpp: Fix Clang scaling issue.
• AC.cpp: Improve clustered easy leaves algorithm using Gourdon's inversion equality.
• AC.cpp: Implement segmented C1 algorithm to improve cache efficiency, #128.
• AC.cpp: Use multi-threading to initialize the libdivide vector.
• util.cpp: Retune Gourdon alpha factors due to improved C algorithm.
• test.cpp: Improve test coverage using alternate alpha_z values.
• doc/Easy-Special-Leaves.pdf: Add description of new C algorithm.
• doc/alpha-factor-gourdon.pdf: Retune Gourdon alpha factors.