Pending mac VM 3.7.4B1

Tim Rowledge tim at sumeru.stanford.edu
Sun Jun 20 16:14:09 UTC 2004


John M McIntosh <johnmci at smalltalkconsulting.com> wrote:

> I think Tim's work on the primitive call logic has increase sends/sec  
> performance.
> I know he saw a huge performance boost on RISC.
Yup - almost 2.5 times faster on the large benchmarks in the Benchmarks
package off SM. No change in bytecodes/sec since that isn't affected by
changes in prim calling, and a small improvement in the sends/sec test
for similar reasons. But if your time() like calls cost anything
noticeable and you're doing it up to three times for every prim
(including simple arithmetic & atputs) it adds up fast.

I'm still pretty sure that send/return could be several times faster
with work but I'm not about to start spending that sort of time unless
somebody pays.

tim
--
Tim Rowledge, tim at sumeru.stanford.edu, http://sumeru.stanford.edu/tim
Strange OpCodes: PNG: Pass Noxious Gas



More information about the Squeak-dev mailing list