The Brutal Performance Impact From Mitigating The LVI Vulnerability

SMHasher as a hash function quality/speed benchmark saw its mitigated performance at just 10% the original speed pre-mitigations. In particular, having an LFENCE after every load instruction led this massive performance hit while the LFENCE before indirect branches led to a much lower impact.

Or another way to look at it, the cycles per hash increased by nearly 4x for Wyhash.

The results for other hashing functions tested by SMHasher were also quite devastated.

The t1ha0 function with AES + AVX2 usage was one of the most dramatically impacted workloads where having an LFENCE after every load for mitigating LVI led to performance just 6% the original throughput.

The FFTW performance with the complete mitigation led to 8.6% the performance of the Xeon system without any mitigations. Granted, pending any other similar vulnerabilities coming to light or showing more exposure of LVI than is public at this point, most users likely do not need to rebuilding their entire software collection with these assembler flags.