Linux 7.4 Slated To Land The AVX-512 Optimized xor_gen For A Big Performance Win
Curated from Phoronix
If you manage storage infrastructure on modern hardware, this kernel update is a quiet but significant win. For years, software RAID performance has been bottlenecked by parity calculation overhead, particularly when using standard x86 instructions. By optimizing the xor_gen function specifically for AVX-512, the kernel now leverages the wide vector units found in recent AMD Zen 4 and Intel server CPUs. This is not just a theoretical improvement; it translates directly into higher throughput for RAID5 and RAID6 arrays without requiring hardware RAID controllers. For practitioners, the takeaway is straightforward: if you are running Linux 7.4 on supported hardware, verify that your RAID drivers are utilizing the new code paths. You should expect measurable I/O improvements in parity-heavy workloads simply by upgrading the kernel, eliminating the need for expensive hardware offloading.
Over the past few months there have been patches floating around for an AVX-512 optimized xor_gen() for the Linux kernel for that functuon used by the Linux software RAID code. It's been shhowing very nice performance gains with this function being used for generating and validating parity blocks such as for RAID5 and RAID6.
— Phoronix