> Just because somebody makes an implementation decision, doesn't mean > it's the best way to do things. :-) My background leads me to believe > that this is unnecessary. As I haven't dug into the differences all the > heavily, I could be 100% wrong. My gut feeling is that using two VMs was > just the easiest way to do this. Based on what little I know about the > differences between v9 and v8, I think Gnash can handle both in the same > VM.
The new VM uses a completely different bytecode format (similar to JVM one btw) which is a "typed bytecode" (all class structures with types are present inside the bytecode). This enable agressive optimizations in the JIT but the bytecode needs first to be "checked" in order to ensure that it's correct. The FP9/AS3/AVM2 APIs are also completely different from the AS2 ones, so it is needed to entirely rewrite the engine/vm bindings and add the new F9 features such as a completely new event model. You can compare the performances of FP8 vs FP9 by compiling using haXe (http://haxe.org). The differences of API are also documented there : http://haxe.org/api Best, Nicolas _______________________________________________ Gnash mailing list [email protected] http://lists.gnu.org/mailman/listinfo/gnash
