RE: [Ikvm-developers] RE: ikvm performance

[Jonathan P. <jp...@ny...>]

Jeroen,

>>Interesting idea, but why would you want this? Remember that the is a
>>steep performance penalty for such dynamic access. Wouldn't it be better
>>to just recompile the assembly?

It would be better to recompile the assembly, but I might not always know
which ones need to be recompiled. I guess I could recompile all of the ones
with warnings whenever I had a new version of any of them.

Why not handle missing methods and illegal access errors the same as
classnotfound errors, by using reflection at runtime in case the issues got
resolved since the assembly was compiled? 

You might be able to address the performance issue by generating temporary
stub wrappers that would cache the dynamically loaded type or method once
they were looked up for the next time they were called and invoke them by
reference instead of by name.

These stubs could be used as a map that identifies which classes had issues
that would benefit from static recompilation. You might then be able to
provide some sort of batch option or utility that determines runtime
dependencies on a group of ikvmc compiled assemblies.

Jonathan