D
dbeavon
Guest
I was comparing to c#.net which is also a "managed" runtime (CLR) and is somewhat analogous to the AVM. As I understand, the primary difference is that c# is typically jitted into native instructions at some point along the way. So that prevents the same statement from being re-interpreted repeatedly. My original question assumed that the re-interpretation of the same r-code in a loop was the overhead that is primarily responsible for the performance difference between the CLR and the AVM. But as Gus points out, the interpretation of the method invocation cannot be the *only* thing that accounts for the performance difference. There is almost certainly a lot of functionality that is different, in addition to the interpretation of the r-code. However the PDSOE profiler doesn't allow us to know what is actually happening for those ~600 ms. Most programmers (ABL or otherwise) who are writing code would not account for the overhead cost of the method invocation itself, since we assume it is essentially free. But that doesn't seem to be the case here.
Continue reading...
Continue reading...