> What’s the impact of these changes? Our tac 2.0 release is up to 3x faster than the GNU Coreutils version of tac if benchmarking with speculative execution mitigations disabled. If the mitigations are not enabled, then our tac will be even faster than that as we are using memory-mapped files (with specifically mmap-friendly access patterns) to avoid the repeated syscalls to refill the buffers, which are considerably slower with mitigations enabled.

This says it’s comparing disabled mitigations against “not enabled” mitigations, which means the same thing. Presumably should be “not disabled”.