php.net |  support |  documentation |  report a bug |  advanced search |  search howto |  statistics |  random bug |  login
Bug #73070 random_bytes: SandyBridge faster than IvyBridge
Submitted: 2016-09-13 15:09 UTC Modified: -
From: spam2 at rhsoft dot net Assigned:
Status: Open Package: Performance problem
PHP Version: 7.0.11RC1 OS: Linux
Private report: No CVE-ID: None
View Developer Edit
Welcome! If you don't have a Git account, you can't do anything here.
If you reported this bug, you can edit this bug over here.
(description)
Block user comment
Status: Assign to:
Package:
Bug Type:
Summary:
From: spam2 at rhsoft dot net
New email:
PHP Version: OS:

 

 [2016-09-13 15:09 UTC] spam2 at rhsoft dot net
Description:
------------
today i changed our password-generator to random_bytes() and started a test checking also for collisions - on two machines completly different timings - while i expected the one generation newer SandyBridge faster in fact it is many times slower running with large loops

interesting fact that this is true for just "random_bytes" as well as combined with "array_key_exists" which becomes even much more slower and that for most normal websites the newer machine wins as expected - but not for the cases below
_________________________________________________

random_bytes() alone:

Intel(R) Core(TM) i7-2600 CPU @ 3.40GHz
SECONDS: 185
GENERATED: 50000000
RANDOM/SEC: 270270

Intel(R) Core(TM) i7-3770 CPU @ 3.40GHz
SECONDS: 297
GENERATED: 50000000
RANDOM/SEC: 168350
_________________________________________________

combined with array_key_exists()

Intel(R) Core(TM) i7-2600 CPU @ 3.40GHz
SECONDS: 21
COLLISIONS: 0
GENERATED: 3000000
RANDOM/SEC: 142857

Intel(R) Core(TM) i7-3770 CPU @ 3.40GHz
SECONDS: 101
COLLISIONS: 0
GENERATED: 3000000
RANDOM/SEC: 29702

Test script:
---------------
<?php
 if(!function_exists('random_bytes'))
 {
  exit("'random_bytes' not available\n");
 }
 $start = time();
 for($count = 1; $count < 50000000; $count++)
 {
  $x = random_bytes(10);
  echo $count . "\n";
 }
 echo "\n";
 $runtime = time() - $start;
 echo 'SECONDS: ' . $runtime . "\n";
 echo 'GENERATED: ' . $count . "\n";
 if($runtime > 0)
 {
  echo 'RANDOM/SEC: ' . intval($count / $runtime) . "\n";
 }
?>

<?php
 if(!function_exists('random_bytes'))
 {
  exit("'random_bytes' not available\n");
 }
 $start = time();
 $collissions = 0;
 for($count = 1; $count < 3000000; $count++)
 {
  $pwd = random_bytes(10);
  $hash = sha1($pwd);
  echo $count . ': ' . $hash . "\n";
  if(array_key_exists($hash, $used))
  {
   $collissions++;
  }
  $generated++;
  $used[$hash] = 1;
 }
 echo "\n";
 $runtime = time() - $start;
 echo 'SECONDS: ' . $runtime . "\n";
 echo 'COLLISIONS: ' . $collissions . "\n";
 echo 'GENERATED: ' . $count . "\n";
 if($runtime > 0)
 {
  echo 'RANDOM/SEC: ' . intval($count / $runtime) . "\n";
 }
?>

Expected result:
----------------
both at least compareable similar fast and in doubt the newer CPU faster

Actual result:
--------------
the newer CPU is way slower and it's not /dev/urandom given the results below and that 'haveged' is running on both machines and on the newer one which even supports the 'RDRAND' instruction also 'rngd'

[root@srv-rhsoft:~]$ dd if=/dev/urandom of=/dev/zero bs=1M count=512
521142272 bytes (521 MB, 497 MiB) copied, 24,0227 s, 21,7 MB/s
512+0 Datensätze ein
512+0 Datensätze aus
536870912 bytes (537 MB, 512 MiB) copied, 24,736 s, 21,7 MB/s


Patches

Pull Requests

 
PHP Copyright © 2001-2024 The PHP Group
All rights reserved.
Last updated: Wed Dec 04 03:01:30 2024 UTC