Does your glibc already have ld.so --list-diagnostics? It would be good to know the x86.cpu_features.shared_cache_size and x86.cpu_features.non_temporal_threshold values.
I suspect you are on a machine with a large core count, so the heuristic tunes things down to a low non-temporal threshold.
Does your glibc already have ld.so --list-diagnostics? It would be good to know the x86.cpu_ features. shared_ cache_size and x86.cpu_ features. non_temporal_ threshold values.
I suspect you are on a machine with a large core count, so the heuristic tunes things down to a low non-temporal threshold.