On the arm version of macOS, the first time an x86 process is run it is converted to arm code. Each following time the process is run the arm version should be used from a cache to avoid the conversion process.
However, if I use execvp to start the process, I believe the arm conversion happens every time. The progress takes 3+ seconds to run when it should take less than 0.02 seconds. If I use std::system to run the process it is fast.
Is there a way to get execvp to use the Rosetta cache, or is there a similar function I can use?
