
Most exciting for end-users with libgcrypt 1.12 is adding an VAES/AVX-512 accelerated implementation for the Advanced Encryption Standard (AES). With this new AVX-512/VAES implementation, AES-OCB "Offset Codebook" is around two times faster on AMD Zen 5 processors thanks to their AVX-512 support across the board from consumer CPUs through server platforms.
Libgcrypt 1.12 also adds AVX2 and AVX-512 code paths for improved CRC performance. There are also some optimizations for avoiding AVX-512/AVX2/SSSE3 use for single-block processing with ChaCha20 or for Blake2. Plus there are some RISC-V Vector implementations for different crypto optimizations and other enhancements.
Libgcrypt 1.12 also brings Dilithium (ML-DSA) support.
Downloads and more details on this important cryptographic library update via the GnuPG mailing list .