On Wed, Jan 02, 2008 at 10:06:43PM -0800, Darren New wrote:

Indeed, the CIL has a number of features that are designed to increase the performance over the JVM. For example, at any given point in the bytecode execution, the CIL requires there to be the same number and types of items on the data stack (if I recall correctly), which apparently makes it easier to generate better JIT code by allowing some of those locations to be assigned to registers, for example.

There are a couple of these decisions that bubble up into the languages as
well.  A given structured block of data has to be well organized from the
perspective of the garbage collector.  I was kind of frustrated to find out
that you can have arbitrary byte arrays inline in a class.  They have to be
pointers to separate heap objects.  It mostly comes to play when trying to
match an external data structure.

Dave

--
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-lpsg

Reply via email to