(PHP 4, PHP 5, PHP 7)
microtime — Return current Unix timestamp with microseconds
microtime() returns the current Unix timestamp with microseconds. This function is only available on operating systems that support the gettimeofday() system call.
By default, microtime() returns a string in the form "msec sec", where sec is the number of seconds since the Unix epoch (0:00:00 January 1,1970 GMT), and msec measures microseconds that have elapsed since sec and is also expressed in seconds.
If get_as_float is set to TRUE, then
microtime() returns a float, which
represents the current time in seconds since the Unix epoch accurate to the
nearest microsecond.
Example #1 Timing script execution with microtime()
<?php
/**
* Simple function to replicate PHP 5 behaviour
*/
function microtime_float()
{
list($usec, $sec) = explode(" ", microtime());
return ((float)$usec + (float)$sec);
}
$time_start = microtime_float();
// Sleep for a while
usleep(100);
$time_end = microtime_float();
$time = $time_end - $time_start;
echo "Did nothing in $time seconds\n";
?>
Example #2 Timing script execution in PHP 5
<?php
$time_start = microtime(true);
// Sleep for a while
usleep(100);
$time_end = microtime(true);
$time = $time_end - $time_start;
echo "Did nothing in $time seconds\n";
?>
Example #3 microtime() and REQUEST_TIME_FLOAT (as of PHP 5.4.0)
<?php
// Randomize sleeping time
usleep(mt_rand(100, 10000));
// As of PHP 5.4.0, REQUEST_TIME_FLOAT is available in the $_SERVER superglobal array.
// It contains the timestamp of the start of the request with microsecond precision.
$time = microtime(true) - $_SERVER["REQUEST_TIME_FLOAT"];
echo "Did nothing in $time seconds\n";
?>
You can use one variable to check execution $time as follow:
<?php
$time = -microtime(true);
$hash = 0;
for ($i=0; $i < rand(1000,4000); ++$i) {
$hash ^= md5(substr(str_shuffle("0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ"), 0, rand(1,10)));
}
$time += microtime(true);
echo "Hash: $hash iterations:$i time: ",sprintf('%f', $time),PHP_EOL;
?>
A lot of the comments here suggest adding in the following way: (float)$usec + (float)$sec
Make sure you have the float precision high enough as with the default precision of 12, you are only precise to the 0.01 seconds.
Set this in you php.ini file.
precision = 16
Note that the timestamp returned is "with microseconds", not "in microseconds". This is especially good to know if you pass 'true' as the parameter and then calculate the difference between two float values -- the result is already in seconds; it doesn't need to be divided by a million.
All these timing scripts rely on microtime which relies on gettimebyday(2)
This can be inaccurate on servers that run ntp to syncronise the servers
time.
For timing, you should really use clock_gettime(2) with the
CLOCK_MONOTONIC flag set.
This returns REAL WORLD time, not affected by intentional clock drift.
This may seem a bit picky, but I recently saw a server that's clock was an
hour out, and they'd set it to 'drift' to the correct time (clock is speeded
up until it reaches the correct time)
Those sorts of things can make a real impact.
Any solutions, seeing as php doesn't have a hook into clock_gettime?
More info here: http://tinyurl.com/28vxja9
http://blog.habets.pp.se/2010/09/
gettimeofday-should-never-be-used-to-measure-time
Check Program runtime:
<?php
class bwruntime
{
var $timestart;
var $digits;
function bwruntime($digits = "")
{
$this->timestart = explode (' ', microtime());
$this->digits = $digits;
}
function totaltime()
{
$timefinish = explode (' ', microtime());
if($this->digits == ""){
$runtime_float = $timefinish[0] - $this->timestart[0];
}else{
$runtime_float = round(($timefinish[0] - $this->timestart[0]), $this->digits);
}
$runtime = ($timefinish[1] - $this->timestart[1]) + $runtime_float;
return $runtime;
}
}
?>
Need a mini benchmark ?
Use microtime with this (very smart) benchmark function :
mixed mini_bench_to(array timelist[, return_array=false])
return a mini bench result
-the timelist first key must be 'start'
-default return a resume string, or array if return_array= true :
'total_time' (ms) in first row
details (purcent) in next row
example :
<?php
unset($t); // if previous used
//-- zone to bench
$t['start'] = microtime(true);
$tab_test=array(1,2,3,4,5,6,7,8);
$fact=1;
$t['init_values'] = microtime(true);
foreach ($tab_test as $key=>$value)
{
$fact=$fact*$value;
}
$t['loop_fact'] = microtime(true);
echo "fact = ".$fact."\n";
//-- END zone to bench
echo "---- string result----\n";
$str_result_bench=mini_bench_to($t);
echo $str_result_bench; // string return
echo "---- tab result----\n";
$tab_result_bench=mini_bench_to($t,true);
echo var_export($tab_result_bench,true);
?>
this example return:
---- string result----
total time : 0.0141 ms
start -> init_values : 51.1 %
init_values -> loop_fact : 48.9 %
---- tab result----
array (
'total_time' => 0.0141,
'start -> init_values' => 51.1,
'init_values -> loop_fact' => 48.9,
)
The function to include :
<?php
function mini_bench_to($arg_t, $arg_ra=false)
{
$tttime=round((end($arg_t)-$arg_t['start'])*1000,4);
if ($arg_ra) $ar_aff['total_time']=$tttime;
else $aff="total time : ".$tttime."ms\n";
$prv_cle='start';
$prv_val=$arg_t['start'];
foreach ($arg_t as $cle=>$val)
{
if($cle!='start')
{
$prcnt_t=round(((round(($val-$prv_val)*1000,4)/$tttime)*100),1);
if ($arg_ra) $ar_aff[$prv_cle.' -> '.$cle]=$prcnt_t;
$aff.=$prv_cle.' -> '.$cle.' : '.$prcnt_t." %\n";
$prv_val=$val;
$prv_cle=$cle;
}
}
if ($arg_ra) return $ar_aff;
return $aff;
}
?>
Of the methods I've seen here, and thought up myself, to convert microtime() output into a numerical value, the microtime_float() one shown in the documentation proper(using explode,list,float,+) is the slowest in terms of runtime.
I implemented the various methods, ran each in a tight loop 1,000,000 times, and compared runtimes (and output). I did this 10 times to make sure there wasn't a problem of other things putting a load spike on the server. I'll admit I didn't take into account martijn at vanderlee dot com's comments on testing accuracy, but as I figured the looping code etc would be the same, and this was only meant as a relative comparison, it should not be necessary.
The above method took on average 5.7151877 seconds, while a method using substr and simply adding strings with . took on average 3.0144226 seconds. rsalazar at innox dot com dot mx's method using preg_replace used on average 4.1819633 seconds. This shows that there are indeed differences, but for normal use noone is going to notice it.
Note that the substr method mentioned isn't quite the one given anonymously below, but one I made based on it:
<?php
$time=microtime();
$timeval=substr($time,11).substr($time,1,9);
?>
Also worth noting is that the microtime_float() method gets faster, and no less accurate, if the (float) conversions are taken out and the variables are simply added together.
Any of the methods that used + or array_sum ended up rounding the result to 2 digits after the decimal point, while (most of) the ones using preg_replace or substr and . kept all the digits.
For accurate timing, since floating-point arithmetic would lose precision, I stored microtime results as-is and calculated time difference with this function:
<?php
function microtime_used($before,$after) {
return (substr($after,11)-substr($before,11))
+(substr($after,0,9)-substr($before,0,9));
}
?>
For further information, the script itself, etc, see http://edorfaus.xepher.net/div/convert-method-test.php
Rather than using the list() function, etc. I have found the following code to be a bit cleaner and simpler:
<?php
$theTime = array_sum( explode( ' ' , microtime() ) );
echo $theTime;
# Displays "1212018372.3366"
?>
Get date time with milliseconds
$micro_date = microtime();
$date_array = explode(" ",$micro_date);
$date = date("Y-m-d H:i:s",$date_array[1]);
echo "Date: $date:" . $date_array[0];
Test accuracy by running it in a loop.
While doing some experiments on using microtime()'s output for an entropy generator I found that its microsecond value was always quantified to the nearest hundreds (i.e. the number ends with 00), which affected the randomness of the entropy generator. This output pattern was consistent on three separate machines, running OpenBSD, Mac OS X and Windows.
The solution was to instead use gettimeofday()'s output, as its usec value followed no quantifiable pattern on any of the three test platforms.
I like to use bcmath for it
<?php
function micro_time() {
$temp = explode(" ", microtime());
return bcadd($temp[0], $temp[1], 6);
}
$time_start = micro_time();
sleep(1);
$time_stop = micro_time();
$time_overall = bcsub($time_stop, $time_start, 6);
echo "Execution time - $time_overall Seconds";
?>
I have been getting negative values substracting a later microtime(true) call from an earlier microtime(true) call on Windows with PHP 5.3.8
Produces negative values
------------------------------
for($i = 0; $i<100; $i++) {
$x = microtime(true);
//short calculation
$y = microtime(true);
echo ($y-$x) . "\n"; // <--- mostly negatives
}
Calling usleep(1) seems to work
---------------------------------------
for($i = 0; $i<100; $i++) {
$x = microtime(true);
//short calculation
usleep(1);
$y = microtime(true);
echo ($y-$x) . "\n"; // <--- fixed now
}
I use this for measure duration of script execution. This function should be defined (and of couse first call made) as soon as possible.
<?php
/**
* get execution time in seconds at current point of call in seconds
* @return float Execution time at this point of call
*/
function get_execution_time()
{
static $microtime_start = null;
if($microtime_start === null)
{
$microtime_start = microtime(true);
return 0.0;
}
return microtime(true) - $microtime_start;
}
get_execution_time();
?>
However it is true that result depends of gettimeofday() call. ([jamie at bishopston dot net] wrote this & I can confirm)
If system time change, result of this function can be unpredictable (much greater or less than zero).
Microtime as float in a functional way
<?php
array_reduce(
explode(" ", microtime),
function($time, $microtime) {
return $microtime + (float) $time;
},
0.0
);
?>
To know how much time a code need to execute, I use my own class which I defined like this :
<?php
class RenderTime{
const PRECISION_SECOND = 0;
const PRECISION_MILLISECOND = 1;
const PRECISION_MICROSECOND = 2;
private $start;
private $end;
public function __construct(){
$this->start = null;
$this->end = null;
}
public function start(){
$this->start = microtime(true);
$this->end = null;
}
public function end(){
$this->end = microtime(true);
}
/**
* This function return the time the code use to process
* @param $precision the precision wanted, with const. second, millisecond and microsecond available (default PRECISION_SECOND)
* @param $floatingPrecision the number of numbers after the floating point (default 0)
* @param $showUnit precise if the unit should be returned (default true)
* @return the render time in the precision asked. Note that the precision is ±0.5 the precision (eq. 5s is at least 4.5s and at most 5.5s) <br/>
* The code have an error about 2 or 3µs (time to execute the end function)
*/
public function getRenderTime($precision = self::PRECISION_SECOND, $floatingPrecision = 0, $showUnit = true){
$test = is_int($precision) && $precision >= self::PRECISION_SECOND && $precision <= self::PRECISION_MICROSECOND &&
is_float($this->start) && is_float($this->end) && $this->start <= $this->end &&
is_int($floatingPrecision) && $floatingPrecision >= 0 &&
is_bool($showUnit);
if($test){
$duration = round(($this->end - $this->start) * 10 ** ($precision * 3), $floatingPrecision);
if($showUnit)
return $duration.' '.self::getUnit($precision);
else
return $duration;
}else{
return 'Can\'t return the render time';
}
}
private static function getUnit($precision){
switch($precision){
case self::PRECISION_SECOND :
return 's';
case self::PRECISION_MILLISECOND :
return 'ms';
case self::PRECISION_MICROSECOND :
return 'µs';
default :
return '(no unit)';
}
}
}
?>
If you want to use it, it's quiet simple:
<?php
$render_time = new RenderTime();
$render_time->start();
for($i = 0; $i < 2 ** 24; $i++){}
$render_time->end();
echo 'Time to render: '.$render_time->getRenderTime(RenderTime::PRECISION_MILLISECOND);
?>
It should output somthing like this : 'Time to render: 3029 ms'
Very useful for quick optimisation
Following are tested under PHP 5.6.2:
microtime() returns a string in the form "msec.sec" not "msec sec".
So in Example #1, this line should be
list ($usec, $sec) = explode(".", $e);
I wrote a class which allows to time execution of php scripts very easily. You can even time multiple separate blocks and then get the total time at the end.
The class is pretty long so I won't post it here, but here's the link:
http://codeaid.net/php/calculate-script-execution-time-(php-class)
An quick usage example:
<?php
for ($i = 0; $i < 100; $i++) {
// calculate the time it takes to run fx()
Timer::start();
fx();
Timer::stop();
fy();
// calculate the time it takes to run fz()
Timer::start();
fz();
Timer::stop();
}
print Timer::get();
?>
Want to make sure your script doesn't time out on large updates? Say you're doing a poor mans cron job or working with large amounts of data in loop and the user begins to hang too long. Instead of them giving up and leaving you can break the loop when it reaches a set time and still display data.
Here's a simple handy dandy loop timer rounded to second.
<?php
$starttime = round(microtime(true));
$totaltime = 0;
$maxtime = 5; //seconds
for ($i = 1; $i <= 20; $i++) { /// 20 loops 1 sec each
usleep(1000000);
if($totaltime < $maxtime){
$currenttime = round(microtime(true));
$totaltime = $totaltime + ($currenttime -$starttime);
//update and continue
echo $i.'<br>';
}else{
// stop updating or display not updated data.
break;
}
}
?>
<?php
class StopWatch {
public $total;
public $time;
public function __construct() {
$this->total = $this->time = microtime(true);
}
public function clock() {
return -$this->time + ($this->time = microtime(true));
}
public function elapsed() {
return microtime(true) - $this->total;
}
public function reset() {
$this->total=$this->time=microtime(true);
}
}
$stopwatch = new StopWatch();
usleep(1000000);
echo "Checkpoint 1: ".$stopwatch->clock()." seconds<br />";
usleep(1000000);
echo "Checkpoint 2: ".$stopwatch->clock()." seconds<br />";
echo "Total Elapsed: ".$stopwatch->elapsed()." seconds<br />";
?>
This function is very useful for putting a start and end point in your page to find out where is the delay.
<?php
$start = microtime(true);
// My source code here
$end = microtime(true);
echo $end."-".$start."=".($end - $start). " seconds";
?>
If you try this example above (without any source code between the start and the end point). You will get an ugly value, something like:
1212690530.4132-1212690530.4132=8.1062316894531E-6 seconds
You will wonder why you get this because both numbers seem to be equal. Well this is because there is a hidden precision that we are not able to see.
To solve this problem I made a new function:
<?php
function my_microtime($precision = 4)
{
return round(microtime(true),4);
}
$start = microtime(true);
// My source code here
$end = microtime(true);
echo $end."-".$start."=".substr(($end - $start),0,5). " seconds";
?>
It would be useful if they add another parameter for precision in this function or at least another boolean that will not include the hidden precision.
You can read more about the hidden precision in http://php.net/float
Hi, I made a function to get the generation page time :
<?php
function gentime() {
static $a;
if($a == 0) $a = microtime(true);
else return (string)(microtime(true)-$a);
}
?>
(You can add a round() to the return value if you want)
Use :
<?php
# you should include your libraries/conf files here (including the gentime function)
gentime();
# your source code here
echo 'Generated in '.gentime().' seconds.'
?>
Observed in php 5.3.3 on CentOS 6 and php 5.3.14 on Amazon Linux 2012.03:
microtime()'s string value is in the format "0.dddddd00 ddddd...". That is, there's room for 10-ns precision that isn't being used, and the "msec" value is in decimal, not a literal number of microseconds (and *especially* not milliseconds, which is what msec usually stands for) as the documentation might imply.
I wanted to find out whether echo would be quicker in small chunks or one large chunk to test the theory mentioned in the previous post. The following experiment shows that there is no significant performance difference, in terms of execution time elapsed, between the two methods of using echo. I ran two test cases, one with a string that is 100000 bytes long and another with a string length of 1000000. The source code follows below.
<?php
function echobig($string, $bufferSize = 8192)
{
$splitString = str_split($string, $bufferSize);
foreach($splitString as $chunk)
echo $chunk;
}
global $dat;
$dat = "";
function testit()
{
global $dat;
$data = "";
for( $a = 0; $a <= 1000000; $a += 1 ) $data .= "a";
$u1= microtime(true);
echobig( $data );
$u2= microtime(true);
echo $data;
$u3= microtime(true);
$diff = $u2 - $u1;
$diff2= $u3 - $u2;
$dat .= "$diff2 $diff $u2 $u1\r\n";
}
$i = 0;
while( $i < 5 )
{
testit(); $i += 1;
}
global $dat;
$fp = fopen( "../Data/results.txt", "w" );
fwrite( $fp, $dat );
fclose( $fp );
?>
You can run the above experiment yourself or you can look at my test data. I got some test data here: http://www.codesplunk.com/examples/echo.html
But note that the default 'precision' setting of PHP* - which is used when a float is converted to a stringy format by echo()ing, casting or json_encode()ing etc - is not enough to hold the six digit accuracy of microtime(true).
Out of the box, microtime(true) will echo something like:
1377611450.1234
Which is obviously less than microsecond accuracy. You'll probably want to bump the 'precision' setting up to 16 which will echo something like:
1377611450.123456
*Internally* it will be accurate to the six digits even with the default 'precision', but a lot of things (ie. NoSQL databases) are moving to all-text representations these days so it becomes a bit more important.
* 14 at the time of writing
Interesting quirk (tested in PHP 5.0.3): You can get very wacky results from microtime when it is called in the destructor of an object at the end of a script. These times vary enormously and can be in the *past*, when compared to microtime calls in the body of the script.
As a case example, I played with a timer object that measured microtime when it was created at the start of the script, and measured microtime again at the end of the script using __destruct(); and then printed the total execution time (end time - start time) at the bottom of the page. On short scripts, this would often give a negative time!
This quirk does not appear if microtime is measured with an automatic shutdown function (using <?PHP register_shutdown_function('myfunc') ?>. Incidentally, the automatic shutdown functions are called after output buffers are flushed but before object destructors are called.
This little function comes in handy if you want a single integer when your server doesn't have php >= 5.0
It returns seconds passed unix epoch to the microsecond. Or microseconds since unix epoch.
<?php
//A hack for PHP < 5.0
function utime($inms){
$utime = preg_match("/^(.*?) (.*?)$/", microtime(), $match);
$utime = $match[2] + $match[1];
if($inms){
$utime *= 1000000;
}
return $utime;
}
//Example:
print utime();
//Returns:
//1156127104.746352 Seconds
//Example two:
print utime(1);
//Returns:
//1156127104746352 Microseconds
?>
This function allows you to easily calculate time difference between two points in time without losing the precision.
<?php
/** Calculate a precise time difference.
@param string $start result of microtime()
@param string $end result of microtime(); if NULL/FALSE/0/'' then it's now
@return flat difference in seconds, calculated with minimum precision loss
*/
function microtime_diff( $start, $end=NULL ) {
if( !$end ) {
$end= microtime();
}
list($start_usec, $start_sec) = explode(" ", $start);
list($end_usec, $end_sec) = explode(" ", $end);
$diff_sec= intval($end_sec) - intval($start_sec);
$diff_usec= floatval($end_usec) - floatval($start_usec);
return floatval( $diff_sec ) + $diff_usec;
}
?>
that related to the prescedence of the . operator
<?php
# normal prescedence
$a = "te" . 1 - 2 . "st"; # ((te . 1) - 2 ) . st ==> -2st
$b = "te" . 1 + 2 . "st"; # ((te . 1) + 2 ) . st ==> 2st
# modified prescedence
$c = "te" . ( 1 - 2 ) . "st"; # te . ( 1 - 2 ) . st ==> te-1st
$d = "te" . ( 1 + 2 ) . "st"; # te . ( 1 + 2 ) . st ==> te3st
var_dump( $a, $b );
var_dump( $c, $d );
