INFRA Signal 370
AVX-512 xor_gen optimization shows up to 43% speedup for Linux RAID on Zen 5
↗
Illustration only Photo by Aaron McLean on Unsplash
In June, Google engineer Eric Biggers introduced an AVX-512 xor_gen optimization for Linux RAID that later achieved up to a 43% performance gain on an AMD Ryzen 9 9950X processor.
The optimization can reduce CPU time spent on the xor_gen step in Linux RAID, improving overall storage throughput on modern CPUs. It shows that AVX-512 instructions can deliver measurable benefits for software RAID workloads without altering the RAID algorithm. Engineers evaluating performance upgrades can consider enabling AVX-512 paths for similar compute-intensive storage functions.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
Google engineer Eric Biggers posted an AVX-512 xor_gen optimization for Linux RAID in June.
The initial version yielded up to a 41% improvement on an AMD Ryzen 9 9950X (Zen 5) processor.
A revised patch increased the reported gain to up to 43% for the same xor_gen() function.
THE READ
What the cluster adds up to.
On processors lacking AVX-512, the optimization cannot be used, so no gain is realized.
If a RAID configuration does not invoke the xor_gen function, the optimization provides no benefit.
The advantage is limited to workloads where xor_gen constitutes a significant portion of CPU time.
Thus the optimization is most effective on AVX-512 capable CPUs running software-bound RAID workloads that rely on xor_gen.
Written by elseif from the cluster below · checked for specifics the sources never containedTHE CLUSTER