An optimized libcint branch for X86 platform
version 6.2.0 2026-10-16
(GPL licensed) qcint is a branch of libcint library. It provides exactly the same APIs as libcint. However, the code is optimized against SSE3, AVX, AVX2 and AVX-512 instructions. On x86_64 platform, qcint can be 5 ~ 50% faster than libcint. Please refer to libcint for more details of the features and installation instructions.
Like libcint, qcint can be built with 64-bit integers in the API
(-DI8=1), which makes FINT (the type of the atm, bas, shls and
dims arrays, natm, nbas and the return values) int64_t instead of
int, e.g. to match Fortran codes compiled with 8-byte default integers.
Callers must then pass 64-bit integer arrays; the library is not ABI
compatible with a default build. -DCACHE_SIZE_I8=1 instead makes only
CACHE_SIZE_T, the return type of the integral functions (which return the
required cache size when out is NULL), int64_t; it is implied by
-DI8=1.
Qiming Sun osirpt.sun@gmail.com