[Date Prev][Date Next][Thread Prev][Thread Next][Author Index][Date Index][Thread Index]
- To: lukka@xxxxxx
- Subject: Mono quote
- From: Benja Fallenstein <b.fallenstein@xxxxxx>
- Date: Tue, 2 Apr 2002 11:34:57 +0300
- Cc: zzdev@xxxxxxxxxx
here's the quote from the Mono page I mentioned when we were talking about
processing Java bytecode. Re-reading it, I do not claim to understand what
exactly it's supposed to say.
The CIL has some advantages over the Java byte code: it is really an
intermediate representation and there are a number of restrictions on how you
can emit CIL code that simplify creating better JIT engines.
For example, on the CIL, the stack is not really an abstraction available for
the code generator to use at will. Rather, it is a way of creating a postfix
representation of the parsed tree. At any given call point or return point,
the contents of the stack are expected to contain the same object types
independently of how the instruction was reached.
I still hold that we shouldn't go into that anytime soon; there's too much
effort for too little gain. The main speedhog, vstream cells, would be helped
better by unifying Cells and Steppers internally; see my other mail about