Comment on Jochen's long stack traces. The difference must be in how our languages expect the call site to resolve. In my case I compile all of the target methods to match the callsite stack structure. So the fast path adds no additional manipulations ( binds etc ) between the callsite and the target. I put all of the parameter binds and mods needed for the fallback into the fallback path. Here I bind the selector, callsite and other items needed to look up the target. I also have to add some mutations to the test method to extract the field I use to select the target with. This seems redundant to me as each GWT does it over. As an aside I am also able to write most of my fallback in Java so I don't have that many lambdaForms there either. One simplification I have used so far is that all of my Smalltalk objects are a single Java class. Not sure if that is good or bad yet.
thanks for sharing mark
_______________________________________________ mlvm-dev mailing list mlvm-dev@openjdk.java.net http://mail.openjdk.java.net/mailman/listinfo/mlvm-dev
