ZSH Performance: variable vs captured subshell output

var_subshell_perf.zsh

#!/usr/bin/env zsh

test1() {
	test_var=$1
}

test2() {
	print -n $1
}

test4() {
	eval "$1=\$2"
}

zmodload zsh/datetime

typeset -g test_var test_var4
float +E time1 time2 time3 time4

time1=$EPOCHREALTIME
for i in {1..9999}; do
	test1 $i
	print $test_var > /dev/null
done
(( time1 = EPOCHREALTIME - time1 ))

time2=$EPOCHREALTIME
for i in {1..9999}; do
	print $(test2 $i) > /dev/null
done
(( time2 = EPOCHREALTIME - time2 ))

time3=$EPOCHREALTIME
print "$(
	for i in {1..9999}; do
		test2 $i
	done
)" > /dev/null
(( time3 = EPOCHREALTIME - time3 ))

time4=$EPOCHREALTIME
for i in {1..9999}; do
	test4 test_var4 $i
	print $test_var4 > /dev/null
done
(( time4 = EPOCHREALTIME - time4 ))

print "variable: $time1"
print "subshell: $time2"
print "all-in-one-subshell: $time3"
print "eval-var: $time4"

# Results:
# variable: 0.22123885154724121
# subshell: 9.453078031539917
# all-in-one-subshell: 0.13457703590393066
# eval-var: 0.2532649040222168