I normally count() an array, so I wanted to see how empty() would stack up.

<?php
    $test = array();
    $test2 = array();
    for ($x = 0; $x < 1000; $x++)  $test[] = $x;

    $ts = microtime(true);
    for ($x = 0; $x < 100000000; $x++)
    {
        if (count($test))
        {
        }
    }

    echo "Time taken:  " . (microtime(true) - $ts) . " sec\n";
?>

For 100,000,000 comparisons, here are the results against PHP 7.2.16 on my hardware:

count($test):  2.697 sec
count($test2):  2.596 sec
$test === array():  2.579 sec
$test2 === array():  2.552 sec
empty($test):  3.085 sec
empty($test2):  3.113 sec

In short, it doesn't matter what method is used although empty() is actually just ever so slightly slower despite it being a language construct.  YMMV.