LLVM Clang 14 Lands an “Amazing” Performance Optimization

phoronix.com

27 pointsdmmalam3 comments

Wow, that's incredible! 12% gain in runtime perf of Postgres benchmarks is no joke

Also check out the LLVM patch, it's got a C code example explaining what is going on: https://reviews.llvm.org/D113289

Reading that chart literally, it's saying the the Read-Write "Average Latency" is +12% with the load hoisting applied.

Isn't latency something we'd want to reduce, not raise?

I think that's part of the benchmark's name.