[Q] Speed comparison 2.8 vs 3.8
tim Rowledge
tim at rowledge.org
Thu Aug 25 15:31:45 UTC 2005
On 25-Aug-05, at 11:48 AM, Günther Schmidt wrote:
> Thanks all,
>
> that's interesting news. I wasn't aware that there is indeed a problem
> with Morphics performance.
There are some very simplistic and non-scaling algorithms involved in
Morphic displaying and updating. To observe this you need only open a
couple of windows and then add a third; it will typically open fairly
quickly. Now add another dozen. Watch how long the last one takes.
You can also experiment with turning off the deferred displaying to see
just how much gets rendered when a top window is moved or closed. I
strongly suspect that some careful consideration of the algorithms used
might result in dramatic improvements.
Referring back to the comment that the VM is about 10% faster since 2.8
- it depends a lot on what you consider a suitable benchmark. For a
general picture I wouldn't take any notice of anything less than the
full Benchmark package. The trivial bytecode loop + recursive send
tests are not much more meaningful than a slashdot poll. On my machine
the latest VMs are nearly three >times< as fast as the 2.8 ones. I have
no doubt that on some machines the difference might be minimal - it all
depends on compilers and options used, OS version etc etc etc.
tim
--
tim Rowledge; tim at rowledge.org; http://www.rowledge.org/tim
More information about the Squeak-dev
mailing list
|