ID:               48403
 Updated by:       [email protected]
 Reported By:      gregor at hostgis dot com
-Status:           Open
+Status:           Feedback
 Bug Type:         Performance problem
 Operating System: Linux
 PHP Version:      5.2.9
 New Comment:

Works fine for me. You should really use 2 decimals for the time. I got
0.044s without PHP, 0.04s with PHP..


Previous Comments:
------------------------------------------------------------------------

[2009-05-27 07:37:57] gregor at hostgis dot com

Description:
------------
Commands executed via shell_exec() take a lot longer to run than when
run directly from the Unix shell. In the example code below, a task that
takes 0.5-0.7 seconds takes 5 seconds.

This effect is reliable, and a given task's "how much longer it takes
than it should" ratio seems constant. In the simple case below, it takes
8-10 times longer than it should. For a specific more complex test case
at our side, the task consistently takes 14 seconds from shell and 83
seconds from shell_exec()

This effect is apparently identical for the entire family of
process-handling functions: passthru, exec, shell_exec, and popen all
show the exact same "slowness multipliers" as described above.

This report may be similar to http://bugs.php.net/bug.php?id=15746
However, this is for Linux systems and not specific to Windows NT or
NETPBM.

Reproduce code:
---------------
<?php

// gdal_translate converts GIS data
// in this case, simply copy a GeoTIFF with no transforms
$command = 'gdal_translate /tmp/in.tif /tmp/out.tif';

$t0 = microtime(true);
exec($command);
$t1 = microtime(true);

printf("%.1f seconds", $t1 - $t0);
?>

Expected result:
----------------
When run from the shell, this takes 0.5 to 0.7 seconds.

Actual result:
--------------
This takes 5.0-5.5 seconds.


------------------------------------------------------------------------


-- 
Edit this bug report at http://bugs.php.net/?id=48403&edit=1

Reply via email to