John, do you have Sophie running on this thing yet?
Brad Fuller
brad at bradfuller.com
Sun Jan 21 17:16:48 UTC 2007
Aaron Reichow wrote:
> Brad,
>
> Newer Squeak is slower on any machine- you just don't notice when the
> machine is sufficiently fast enough. It is dependent on the CPU speed of
> the machine you're using. When you are running a 200-600 MHz ARM CPU,
> which don't have floating points units, you definitely notice the
> difference between the versions of Squeak, feeling a steady decline in
> responsiveness form 2.4 (the smallest version I've an image sitting
> around for) up to 3.9. Mind you, when I say "newer Squeaks are slower"
> I'm referring to Morphic and how it feels, not the result of any
> benchmark I've run. It is the result of changes- both good and bloat- in
> Morphic, not something completely fundamental. MVC feels identical in
> any Squeak image I've used on a slower machine.
>
> I'm not sure what makes it slower specifically- all I know is that there
> has been a steady march of slowdown for each version of Squeak
> released. I don't know enough about Morphic and the changes to Morphic
> between each version to tell you why, but I imagine a lot of it is
> various improvements to Morphic over the years. A lot of these changes
> could probably be refactored and performance improved, but since that
> isn't really my area of interest of expertise I simply use an older
> version of Squeak, like 3.2, a version for which I have no problems
> porting my code forward to newer Squeak releases and generally for which
> I don't have problems getting newer code to run with a few tweaks. Some
> people assume 3.2 is ancient, but aside some visual aspects looking
> nicer in 3.7-3.9, it is pretty much the same Squeak we all know and
> love. :)
Thanks for the review, Aaron. From what I read, this is a common
criticism of the newer squeaks. If Morphic is the sole culprit, or a
major contributor (I understand you are reporting observations, not hard
tests) then that gives more weight to the Morphic team to test - and if
the affirm your observations, refactor. I know that has got to be one
tough job.
--
brad fuller
www.bradfuller.com
More information about the Squeak-dev
mailing list
|