Anthony Hannan wrote:

> Pre-compiling a whole Squeak image to native code would not be better
than just-in-time 
> compiling primarily because native code takes up a lot more space than
bytecodes.  It is 
> better to only compile bytecodes that are needed.  

	I also like true JIT. The term 'pre-JIT' is kind of funny too;
but in the lack of a JIT, this may be a second best. In 'Back To The
Future', quote:

	'However, useful as it is for debugging, the Squeak virtual
machine running on top of Smalltalk is orders of magnitude too slow for
useful work: running in Squeak itself, the Smalltalk version of the
Squeak virtual machine is roughly 450 times slower than the C version. '

	BTW, I recall Ian posted a request for help to crash his J4. Did
anyone participate and know what's the current status of Ian's J4 ?

> Each bytecode requires an average of 3 native instructions, 
> yielding about 12 bytes per bytecode [1]. 
> Compiling all bytecodes would create a huge image

	Not if we have an efficient module system ;-)

> that could cause performance problems on machines with limited memory.

> In fact in some cases when the cpu cache is sufficiently small 
> it is faster to never compile to native code and always interpret
bytecodes [1].

	This study was published in 1997. The hardware scenario in 2002
may be quite different from that at the time the study was carried out.

	FWIW, I think .NET Framework is delivered pre-JITted

	Cheers,

	PhiHo.