On 2009-11-20, at 16:39, Chris Kohlhardt wrote: > After a good amount of work, we've managed to get our application completely > migrated to OL4.6.1 and SWF9. > > Thank you very much to everyone involved in making the SWF9 runtime a > reality. The performance of Gliffy is so much faster now, it's almost > unbelievable. We're entering QA next week, and we expect to release SWF9 > Gliffy in mid December.
Hey Chris, Congratulations! And, thanks for the praise. And yes, swf9 does seem to give some pretty amazing speedups. (Kudos to Adobe for the swf9 runtime!) Your experience mirrors others: in general there is some pain in adapting to both the changes that we had to make in OpenLaszlo to accommodate a runtime with strict typechecking, and to the new runtime; but the result is a very perceptible performance increase (due in a large part to the compile-time optimizations that static typing permits). > One thing we noticed is that compilation of SWF9 is a lot slower. After > some digging, we were able to speed things up by: > - setting compiler.swf9.incremental=true in lps.properties > - allocating at least 2GB of memory to the tomcat instance running the lps > - moving developers to a pure 64bit OS (Clint moved to Windows 7 after a > long stint with XP) Funny you should ask. We are trying to get funding right now to work on this problem. We are pretty sure we know where the dead bears are, we just need to find the resources to build the sledges and haul them out. We know that the slowdown is due to some large internal memory structures in the compiler that need remodularization. Because of them, you need a lot of RAM and to give Java plenty of heap, or you end up getting killed by paging or GC overhead. P T Withington OpenLaszlo.ORG/Laszlo Systems
