gcc -fopt-info-vec-missed and Clang's -Rpass-missed=loop-vectorize exist as dedicated compiler flags precisely because auto-vectorization silently fails far more often than people expect, and codebases like FFmpeg and OpenCV hand-check this output routinely to confirm a hot loop actually got AVX instructions instead of staying scalar. The gap between a naive loop and one that vectorizes cleanly with __restrict__ and OpenMP SIMD pragmas is routinely 4-8x, which is why numerical library authors treat 'did it vectorize' as a mandatory code-review question, not an assumption.