AVX-Accelerated SM3 Secure Hashing Queued For Linux 5.18

Last year I wrote about Alibaba wanting to contribute AVX-optimized SM3 hashing for the Linux kernel to speed up that Chinese hashing algorithm that has use-cases similar to that of SHA256 while having similar security and performance. That work is now queued up into the crypto subsystem's "-next" branch for Linux 5.18.

SM3 is part of the Chinese Commercial Cryptography Suite and while there has been support for this algorithm within the Linux kernel - including within Arm's TrustZone CrpytoCell driver - there hasn't been any x86_64 optimized version. The work from Alibaba is optimized for modern Intel/AMD CPUs with Advanced Vector Extensions (AVX) and can dramatically speed-up the hashing process.

Last week the code was merged into cryptodev for creating an SM3 standalone library within the kernel and then introducing the AVX optimized code.

Alibaba's Tianjia Zhang is reporting up to a 38% increase in performance with this AVX code on a Skylake system. Look for this in Linux 5.18 should you have a use for SM3 not fulfilled by the likes of SHA256.