Complexity and starting over on the JVM (ideas)
Paul D. Fernhout
pdfernhout at kurtz-fernhout.com
Sun Feb 10 22:51:19 UTC 2008
Still, there are a few ways to do the VM (I'll probably code the core by hand).
What I more want to do is support easily optimizing bottlenecks in, say, 5%
of a Squeak application's code in a cross-platform way.
I'm willing to tolerate a slow Squeak on the JVM (but faster than Dan's :-)
if the tools are otherwise compelling to amplify personal or group
creativity, especially if I know I *could* make it go faster if I really
needed it to (by sacrificing some of the dynamic nature, like translating
large and larger parts of a performance critical application to Scala/JVM or
Java). But *most* Squeak applications might never need that.
Igor Stasenko wrote:
> Hmm, are you saying about translating from one language to another, or
> you speaking about translating VM code?
> The only things, which can improve performance is to translate code to
> bare metal (as Exupery does). Translation to another language (or to
> another VM) is much less effective.
> Don't forget, that any VM/compiler is designed to be optimal for a
> target language, so even if you try as hard as you can, you can't run
> smalltalk with high speeds if your VM is not specialized to run
More information about the Squeak-dev