Libgcrypt 1.12.3 Brings Intel SM3 / SM4 / SHA512 Optimizations

Werner Koch announced today the release of Libgcrypt 1.12.3 with various bug fixes and more exciting are the performance improvements. Libgcrypt as this cryptographic library developed as part of the GnuPG project is enjoying faster performance now, especially when making use of newer Intel CPUs like Arrow Lake and Lunar Lake. Libgcrypt 1.12.3 adds an Intel SM3 extension implementation as well as an Intel SHA512 extension implementation, both of which are supported by Arrow Lake, Lunar Lake, and newer. There is also now an Intel SM4 accelerated AVX2 and AVX-512 implementations in Libgcrypt. Some other performance work in this new Libgcrypt update is for using unaligned vector memory accesses on RISC-V where possible, speeding up sntrup761, allowing internal users to skip fast random poll for ciphers and hashes, and other optimizations. More details on these fixes and performance optimizations in Libgcrypt 1.12.3 can be found via the mailing list announcement. Unfortunately the project's dev.gnupg.org continues to suffer from a denial of service attack that may hamper your ability to access assets on their server.