Benchmarking reporting bogus numbers in Fedora Core

[GoogleCodeExporter]

Running standard benchmarking scripts.

We are getting negative values and extremely large numbers (much longer times 
than the full page time).

This is either related to getrusage() not returning microtimes or bcmath not 
being installed on the server.

Original issue reported on code.google.com by tyc...@gmail.com on 1 Oct 2010 at 4:28

[GoogleCodeExporter]

This is actually two related bugs.

When yum install php-bcmath (and server restart), benchmarking is much better 
(but there is still division by zero).

So this means that my bcmath substitute functions are buggy.

However the ratios and division by zeros appear, which means that the 
getrusage() not handling micro values is also buggy. :-(

In the meantime, I'll use hte walltime instead of the rtime and use php-bcmath()

Original comment by tyc...@gmail.com on 1 Oct 2010 at 4:32