Eugene Kuleshov wrote: > In my subjective tests, the StackMap calculation overhead is much > higher then performance gain you'll get from the new verifier. This is > actually by design of the new verifier, so pre-verification step when > StackMap is generated runs off line and can take long, but > verification at run time is really quick because it uses pre- > calculated StackMap structures. > The only chance to avoid that is to not use COMPUTE_FRAMES (skip the > data flow analysis algorithm) and generate StackMap together with the > bytecode if you know your code structure and type hierarchies.
So if I follow you correctly, COMPUTE_FRAMES should not be used for runtime code unless we're willing to swallow it being much more costly than just allowing the normal verification process to handle that code at runtime. That's good information; we don't generate code frequently enough at runtime to have noticed a performance bottleneck in classloading, but every little bit helps; I'll probably turn this off for JITted methods generated at runtime. - Charlie --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "JVM Languages" group. To post to this group, send email to jvm-languages@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/jvm-languages?hl=en -~----------~----~----~----~------~----~------~--~---
