This is a performance benchmark comparing PHP SplStack, SplQueue, and plain array performance for insert and read/remove operations.

PHP SplStack vs SplQueue vs Array

<?php

namespace PhpBenchmarks;

use SplStack;
use SplQueue;

const NUM_TEST_ITEMS = 100000;
const DATA_SIZE = 32; //Bytes.

$data = [];

$stack = new SplStack();
$queue = new SplQueue();
$array = [];

foreach (range(1, NUM_TEST_ITEMS) as $i) {
    $data[$i] = random_bytes(DATA_SIZE);
}

echo 'Testing SplStack (Push) ... ';

$start = microtime(true);

foreach (range(1, NUM_TEST_ITEMS) as $i) {
    $stack->push($data[$i]);
}

$stop = microtime(true);

echo number_format(NUM_TEST_ITEMS) . ' items in ' . (string) ($stop - $start) . ' seconds.' . "\n";

echo 'Testing Array (Push) ... ';

$start = microtime(true);

foreach (range(1, NUM_TEST_ITEMS) as $i) {
    $array[] = $data[$i];
}

$stop = microtime(true);

echo number_format(NUM_TEST_ITEMS) . ' items in ' . (string) ($stop - $start) . ' seconds.' . "\n";

echo "\n";

echo 'Testing SplStack (Pop) ... ';

$start = microtime(true);

foreach (range(1, NUM_TEST_ITEMS) as $i) {
    $stack->pop();
}

$stop = microtime(true);

echo number_format(NUM_TEST_ITEMS) . ' items in ' . (string) ($stop - $start) . ' seconds.' . "\n";

echo 'Testing Array (Pop) ... ';

$start = microtime(true);

foreach (range(1, NUM_TEST_ITEMS) as $i) {
    array_pop($array);
}

$stop = microtime(true);

echo number_format(NUM_TEST_ITEMS) . ' items in ' . (string) ($stop - $start) . ' seconds.' . "\n";

echo "\n";

echo 'Testing SplQueue (Push) ... ';

$start = microtime(true);

foreach (range(1, NUM_TEST_ITEMS) as $i) {
    $queue->push($data[$i]);
}

$stop = microtime(true);

echo number_format(NUM_TEST_ITEMS) . ' items in ' . (string) ($stop - $start) . ' seconds.' . "\n";

echo 'Testing Array (Push) ... ';

$start = microtime(true);

foreach (range(1, NUM_TEST_ITEMS) as $i) {
    $array[] = $data[$i];
}

$stop = microtime(true);

echo number_format(NUM_TEST_ITEMS) . ' items in ' . (string) ($stop - $start) . ' seconds.' . "\n";

echo "\n";

echo 'Testing SplQueue (Shift) ... ';

$start = microtime(true);

foreach (range(1, NUM_TEST_ITEMS) as $i) {
    $queue->shift();
}

$stop = microtime(true);

echo number_format(NUM_TEST_ITEMS) . ' items in ' . (string) ($stop - $start) . ' seconds.' . "\n";

echo 'Testing Array (Shift) ... ';

$start = microtime(true);

foreach (range(1, NUM_TEST_ITEMS) as $i) {
    array_shift($array);
}

$stop = microtime(true);

echo number_format(NUM_TEST_ITEMS) . ' items in ' . (string) ($stop - $start) . ' seconds.' . "\n";

echo "\n";

Results:

Testing SplStack (Push) ... 100,000 items in 0.012847900390625 seconds.
Testing Array (Push) ... 100,000 items in 0.0077648162841797 seconds.

Testing SplStack (Pop) ... 100,000 items in 0.0076310634613037 seconds.
Testing Array (Pop) ... 100,000 items in 0.0084760189056396 seconds.

Testing SplQueue (Push) ... 100,000 items in 0.011080026626587 seconds.
Testing Array (Push) ... 100,000 items in 0.0070550441741943 seconds.

Testing SplQueue (Shift) ... 100,000 items in 0.007843017578125 seconds.
Testing Array (Shift) ... 100,000 items in 14.495656967163 seconds.

Curiously, plain arrays perform better while inserting, but worse when read/removing versus the SplStack. As suspected, plain array shift (an O(N) operation because it has to reindex the array) is by far slower than the SplQueue shift (an O(1) operation since the implementation is a pointer machine and only involves reassigning pointers).