Andy Wardley wrote: >Nope, I'm still not seeing the slowdown that you. Admittedly my benchmark >is rather primitive. Can you elaborate more on what you're doing in your >benchmark to see if we can identified where the problem might lie? > I redid my benchmarks today to make sure I am doing everything consistently. First I downloaded TT 2.06 and developer releases 2.06a-d and applied Simon's compiled templates patch to all of them. I then configured, made, and installed each one in turn. To test each version I first deleted the files in the template cache and then ran the following command twice:
time for i in `seq 1 50`; do wget -O - 'http://localhost/bztemplate/buglist.cgi' > /dev/null; done During the runs I didn't do anything else on my workstation (although various programs were running in the background). After each run I recorded the "real" time reported by the "time" command, which according to the man page represents "the elapsed real time between invocation and termination." For this test I ran with the XS stash and COMPILE_DIR options enabled. Here are my results as reported by "time": First Run Second Run TT 2.06 1m38.511s 1m37.677s TT 2.06a 1m51.425s 1m50.924s TT 2.06b 1m51.605s 1m51.586s TT 2.06c 1m51.779s 1m49.323s TT 2.06d 2m50.544s 2m47.156s -myk more info about the script: The "buglist.cgi" script I am using in the test is part of Bugzilla. It does not use mod_perl and thus instantiates a new Template object with each request. The pre-templatized version of the script, the templatization bug report, and the templatization patch (without the XS stash and COMPILE_DIR options) are all available online: http://lxr.mozilla.org/mozilla/source/webtools/bugzilla/buglist.cgi?raw=1 http://bugzilla.mozilla.org/show_bug.cgi?id=103778 http://bugzilla.mozilla.org/attachment.cgi?id=68119&action=view
