Why is using built-in variables faster than custom variables?

[william_ahk]

One would think that it would be slower for built-in variables to dynamically resolve than storing the reference to a custom variable. But my tests below seem to show that that is not the case. Can someone explain the how behind it?

Code: Select all

#Requires AutoHotkey v2.0
#SingleInstance Force
ListLines 0

; Globals
Tray := A_TrayMenu

; Cases
TestCases := [
	() => (A_TrayMenu),
	() => (Tray),
]
; Test
TestCount := 10000000
TimeRecords := []
BestRecord := 9223372036854775807 ; LLONG_MAX
For Index, TestCase in TestCases {
	QPC(1)
	Loop TestCount
		TestCase()
	Record := QPC()
	If Record < BestRecord
		BestRecord := Record
	TimeRecords.Push(Record)
}

Result := ""
For Record in TimeRecords {
	Result .= Record (Record = BestRecord ? " <-" : "") "`n"
}
Msgbox Result

My results:

4.6398664 <-
4.6715327

[lexikos]

Why do you think that built-in variables would "dynamically resolve" while custom variables would have a reference stored?

They are both resolved to addresses at load time, and although the type of indirection differs, it is not in any significant way.

Only a fraction of the time needed for each iteration of your benchmark is actually used to evaluate the variable reference. It is such an insignificant operation that coincidental factors such as the layout of your script (or AutoHotkey's binary code) in memory can have a greater effect on the performance (e.g. due to CPU cache and other low-level CPU optimizations).