[Jscheme-user] -server and performance benchmarks

[Ken A. <kan...@bb...>]

I think that both Tim and i measured that using generic functions is about a factor of 3 slower than invoking a JScheme Primitive.  Rusty and i have been using java -server in our application and i just did some simple tests that the overhead is between 1.2 and 1.5 which is pretty acceptable.  Generally, our application runs 30% faster with -server.

So, maybe with -server a purely reflective JScheme woun'd be so bad.

It looks like one issue is that Invoke methods have 2 boolean flags, maybe by partial evaluation we can improve the -client version.

k