Rasmus Lerdorf on PHP Performance

[t31os]

I've just watched this video today by chance and found it very interesting, Rasmus uses WordPress in alot of his demonstrations and it's interesting to see his suggestions on optimizing code for performance.

http://about.digg.com/blog/rasmus-le...hp-performance

Video is just over an hour long, but if you're interested in both PHP and WordPress it's an interesting watch.

[Ryan]

He didn't seem to impressed about the WordPress code quality.

Interesting to hear that using Hip Hop makes very little difference to site performance. I assumed it made a huge difference, maybe 100% performance improvement, but his tests indicated that the difference was very minor.

[mfields]

@Ryan, I think he illustrated that HipHop did in fact make a huge difference.

[Ryan]

I guess that depends on how you define "huge". I was expecting much bigger performance increases than that.

[Otto]

I've had discussions about this with FB people. HipHop is clever, but the thing is, with modern systems and with some of the enhancements made by the PHP team, it doesn't really help a whole heck of a lot to compile to native code. It does indeed help, but unless you're running a site with the traffic of FB, your initial performance improvements will be better if you focus elsewhere.

WP does need a lot of optimization. Once we switch to PHP 5.x, we can do those way more easily.

Like everything else, performance is relative to your specific situation.

[Ryan]

Good to know that HipHop isn't worth worrying about for us unwashed masses. I'd rather not have another thing to worry about before launching a site.

Some of the WordPress issues mentioned in that presentation seem like they were presented as errors in the software, but they're more like intentional features that have been added in full knowledge that there will be a slight performance hit. Things like doing checks for which database type to use etc.

[t31os]

I thought the presentation was rather interesting and that WordPress did quite good on the whole (if it were poorly written i'd have expected more obvious optimization examples in the presentation).

Rasmus also did a very similar presentation at DrupalCon and it didn't do quite as well as WordPress, though not miles apart, you can find the video for that here.

http://vimeo.com/14445227

Or alternatively here
http://drupalradar.com/video-rasmus-...note-drupalcon

[mfields]

@Ryan. Maybe I'm wrong (or just making an assumption) but it appeared to me that Rasmus made a couple optimizations before running Hip Hop. These optimizations had a rather larege impact on performance. I think that maybe this is why Hip Hop appears to have had little impact on the optimization routine. Just guessing here, but it reminds me of zipping a .jpeg file -> Sure you'll save a few k but a jpeg is already optimized - so zipping it doesn't really save much. However, if you zip a .tiff - you'll save a ton of space.

Really don't know if this is the best analogy here all I would like to communicate is that it would be interesting to see what Hip Hop would do if applied as the only optimization.

One of the parts that I found really interesting was Hip Hop's ability to recognize flaws that allude the php parser.

[hakre]

Big Thanks for sharing! Nice pick, still watching ... .