@gottesmm Who owns the benchmark suite?

I wouldn't say that anyone owns it specifically. If this is fixing a real issue, feel free to change. We always fix bugs. Put up a PR and lets see if the fix changes performance?

FYI: @atrick@eeckstein

Oh, I see this is just in the driver. Please put up a PR. I can review this. The change will for sure be tested by a full test.

Comment by Jacob Mao (JIRA)

Hi:
Can I take this one?

I am going to change the code to:

let p = UnsafeMutablePointer(withUnsafeMutablePointer(to: &u) { return $0 })

jacobmao0416 (JIRA User) Sure thing, feel free to assign yourself! Note though that the pointer provided by withUnsafeMutablePointer(to:) is only valid within the scope of the closure passed to it, attempting to escape it is undefined behaviour. What you'll want to do is move the call to proc_pid_rusage inside the closure, e.g:

withUnsafeMutablePointer(to: &u) { p in
  p.withMemoryRebound(to: Optional<rusage_info_t>.self, capacity: 1) { up in
    let _ = proc_pid_rusage(getpid(), RUSAGE_INFO_V4, up)
  }
}

Though while we're here we should also avoid using withMemoryRebound, as the type we're rebinding to has a completely different layout. I believe we should be able to use assumingMemoryBound(to:) instead given we're passing the resulting pointer off to a C API.

Comment by Jacob Mao (JIRA)

@hamishknight, thanks

I have created pr for this:

#28648

Fixed in the above linked PR, thanks jacobmao0416 (JIRA User)!