AVX-512/AVX10 & VAES Optimized AES-GCM Implementation Lands In Linux 6.11
Exciting with the crypto updates for Linux 6.11 are merging the AVX-512/AVX10 and VAES optimized implementation of AES-GCM. The new AES-GCM implementation can score up to 162% faster performance with modern Intel and AMD processors. Eric Biggers of Google has been working on this new tuned AES Galois/Counter Mode implementation after recently upstreaming some nice AES-XTS optimized implementations for modern Intel/AMD processors.
The performance from rewriting the AES-NI implementation and the new VAES and AVX-512/AVX10 AES-GCM implementation is great especially for AMD Zen 4 and newer Intel Xeon servers with AVX-512:
The crypto updates for Linux 6.11 also include x25519 support on ppc64le, Allwinner H616 crypto support is added to the sun8i-ce driver, DMA support in the STM32 driver, and Exynos 850 hardware random number generator support.
More details on these exciting changes via the crypto pull request that has already been merged to Linux 6.11 Git.
4 Comments
