This is the mail archive of the
systemtap@sourceware.org
mailing list for the systemtap project.
Re: coverage database support
- From: William Cohen <wcohen at redhat dot com>
- To: William Cohen <wcohen at redhat dot com>
- Cc: SystemTAP <systemtap at sources dot redhat dot com>
- Date: Sat, 07 Jul 2007 17:55:42 -0400
- Subject: Re: coverage database support
- References: <468E6C00.9060405@redhat.com>
I have been working on optimizing the performance of the coverage database code.
The removed column has been removed from the table, and the type has been
changed from a string to an enums.
I also found that the collect_derivation_chain was just adding additional
elements to the vector rather than creating a new vector from scratch each time.
This was corrected reducing the runtime. A correction has been checked in CVS.
However, the killer test is testsuite/semok/twenty.stp. This creates a probe for
every possible location. For the fc6 machine I am doing the experiments on.
there about 523,000 probes. Each one of those probes would have an entry in the
table. The coverage code appears to be inserting about 450 entries per MINUTE
into the table. This would take about 38 hours to create coverage information
for this one test and generate about 2G table (which is manageable size in sqlite).
Definitely, need to improve the speed of the insertion into the table. Would
like to factor out the file names and store, have the filenames as numbers in
the main table, and have a separate table(s) store the map between number and
file name.
-Will