Latency with v1.11-rc3 REPL over SSH

I don’t know, but I doubt that was the problem. You wouldn’t feel optimizations for newer, vs targeting oldest 386. What you may feel is recompilations, or too much optimizations, triggering such.

Sys.CPU_NAME gives you the LLVM target

julia> Sys.CPU_NAME
"skylake"
1 Like

They seem to be correctly identified. I get Cascade Lake for the Xeon and znver2 for the Epyc.

1 Like