Intel Macs, OpenLaszlo 3.2, and fast fastness

Two interesting things happened last week here at Laszlo headquarters: one, OpenLaszlo 3.2 was released. It includes lots of good new stuff, but the release notes gloss over this change: the script compiler is about five times faster(!) than it was in 3.1. Why? We ported the script compiler from Jython to Java. (and by "we", I mostly mean "Tucker")

I figured that the Laszlo Mail codebase would make a decent real-world benchmark. So I compiled the LZX piece of Laszlo Mail from scratch, with an empty cache, on my Thinkpad T42 (1.7ghz, 1 gig RAM) Here's the compile times:

  • OpenLaszlo 3.1: 95 seconds
  • OpenLaszlo 3.2: 34 seconds

That's pretty flipping good.

Right. So the second thing that happened is that we got a shiny new MacBook. Today I compiled LaszloMail on a co-worker's desktop G5, and the shiny Intel thing. Ready?

  • G5 (dual 2ghz, 1 gig RAM), OpenLaszlo 3.2: 32 seconds
  • MacBook Pro (dual core 2ghz, 2 gig RAM), OpenLaszlo 3.2: 17 seconds

Hot damn!

So if you were avoiding a Powerbook because it seemed too slow for OpenLaszlo work, avoid no more.

8 Responses to “Intel Macs, OpenLaszlo 3.2, and fast fastness”

  1. Jeff Schiller Says:

    Cool. What version of swf was that for?

    Also, can you share any figures on how fast the current speed is when compiling to the DHTML target? I realize that the DHTML capabilities are not available to the layman yet, so if you could shed some light on when a preview version of this capability would be available, that would rock also.

  2. Jeff Schiller Says:

    By the way, when submitting the above comment, I got shot off to a sparse page that wanted me to enter a captcha code. Unfortunately, all that was there was the phrase “captcha_img” so there may be something wrong with your captcha…

  3. gse Says:

    Hm… not sure what’s up with that, Jeff. We’ll look into it.

  4. Jeff Schiller Says:

    And now my first comment has gone missing. Basically I asked what you internal builds compiling to a DHTML target had and when we could expect the DHTML version of LZX to hit the streets.

  5. Chris Says:

    So basically Intel x86 chips are faster than PPC chips of the same clock speed. A big, resounding “duh” comes to mind.

  6. jgrandy Says:

    Chris - um, no. The more likely answer is that Intel/Sun have put many more man-years of effort into optimizing the JVM and the Java JIT compiler for x86 than have IBM/Motorola/Sun for the PPC.

    Another example of this is the relative performance of the x86 and PPC gcc backends. Similar clock speeds, similar bus speed, similar disk throughput and RAM bandwidth. x86 gcc is just much faster than PPC. Again, many more man-years of optimization.

    One of the benefits of 85% market share.

  7. jgrandy Says:

    And Jeff - look for an announcement soon on DHTML. We’re readying a “technology snapshot” that has most of our ducks in a row so that we can share development tasks with the contributor community. The first release that is robust enough for application development won’t be available until later in the summer.

  8. jbardi Says:

    jgrandy - um, yes. Check benchmarks on more than jvm and gcc. The core duo out beats the G5 (even dual chip scenario) in nearly every aspect. Its simply more efficient and runs a heck of a lot cooler… a problem IBM just could never manage to correct. I admit, the dual core G5’s are nice, but try and find that speed in a portable like the MacBook without going Intel. Ahh yes, coding on openlaszlo in blazing speeds while going 75 MPH down the Interstate (as a passenger of course) on a MacBook Pro.. imagine.. two core duos or a single quad duo intel.. right around the corner. I don’t think you can cram your G5 with portable generator in your car very comfortably. Besides, the heat on the Interstate would reak havoc on that overbearing G5 anyway.

Copyright © 2005-2010 Laszlo Systems, Inc.