Hmm. I haven't a lot of time at the moment to gather the numbers discussed on the list and in JIRA. A glance through some relevant issues (VELOCITY-606, VELOCITY-607, VELOCITY-570, VELOCITY-545, VELOCITY-223), i see that at one point Jarkko saw a 25% speed up in some cases and drastic memory improvements in a macro heavy test (down to 5MB from 80MB). I ran some tests myself and recall similar numbers, though i never wrote them down. I do know that the biggest speed-ups were for high concurrency usage and greatest memory reductions were for velocimacro heavy templates. I also know that we fixed some memory and speed issues specifically during tokenizing of templates.
Of course these are all variable depending on the type of usage (caching, many/few macros, jdk5+ vs jdk4, etc). You definitely get more speed with jdk5, especially if you turn off the limit on the cache size (see VELOCITY-648). If you're stuck on jdk1.4, using cached templates with few macros under light load, you are not likely to see much benefit. If you're on jdk1.5+, use a lot of macros, or have high load, you should definitely notice big improvements. And more than just faster and lighter, performance is simply better as there were a number of concurrency bugs fixed and at least one memory leak as well. :) On Wed, Dec 31, 2008 at 10:22 AM, Ray Auge <[email protected]> wrote: > Hey All, Nathan, > > We recently upgraded Liferay to Velocity 1.6.1 which was as simple as > dropping in the new jar without a single code change. Thanks for that. > > In my JIRA ticket for the upgrade I stated: > > > "Version 1.6.1 significantly increases performance of the > parser, improves concurrent access to the introspection cache, > and many more." > > > Now I'm getting questions like: > > > "Do we know how much faster this is?" > > > Anyone have any (including rough, or even very general) figures on this? > > Thanks, > > -- > Raymond Augé > Senior Software Engineer > Liferay, Inc. > Enterprise. Open Source. For Life. > > --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
