Can anyone account for this remarkable increase in speed? I would think that the incorporation of block closures would tend to slow, rather than increase the speed of the interpreter. Is something missing or counted differently, or should we immediately dump the present VM for a way more efficient machine?
On Wednesday, October 23, 2002, at 01:41 AM, Anthony Hannan wrote:
Hello Squeakers,
I'm pleased to announce that I'm finally finished with the latest version of VI4. You can get it from its usual place at http://minnow.cc.gatech.edu/squeak/VI4. It has been rewritten and cleaned up and should be very stable. Once other image format changes are incorporated (if any) and once it is approved for a future Jitter, it should be ready for prime time.
The Interpreter still operates directly on the process stack object, but hooks have been provided so the process stack can be internalized by the VM if desired. See the aboutVI4 method comment for more details.
0 tinyBenchmarks reports a 30% increase in bytecode speed and a 75% increase in send speed.
Enjoy! Anthony
"Andrew C. Greenberg" werdna@mucow.com is claimed by the authorities to have written:
Can anyone account for this remarkable increase in speed?
Yup. Better send/return performance due to having a linearized stack. No copying of arguments. Eliot showed how dramatic the effect of this can be in BrouHaHa way back in 87. PS & HPS do even better.
Personally I feel there are some more improvements needing to be completed before adopting this new code but it is very promising. Anthony has done a very good job and we owe him a big round of beer (or other appropriate drug of choice).
tim
squeak-dev@lists.squeakfoundation.org