Dorado bytecodes per second
Andres Valloud
sqrmax at optonline.net
Sat Apr 30 20:53:48 UTC 2005
Hello...
AR> That's not to say we should throw away speed but the patterns
AR> really have changed. For example, we use "foo isNil" without even
AR> realizing that doing that is some 30 times slower than the
AR> equivalent "foo == nil".
A community unwilling to send a message? This is Smalltalk, right?
The focus should be more along the lines of "why do we need to check
these things?". Every check like that is expensive, regardless of
whether it executes fast or not. It's even faster not to need to
check at all --- to remove the need for ifTrue:/ifFalse:.
That usually requires short methods, clean code, crystal clear design,
enough classes to model the special cases without having to inline
ifTrue:/ifFalse: everywhere... and especially, no more
spaghetti-interlaced frameworks.
In my opinion, these are things that, collectively, we have failed to
keep in mind. And we will keep paying the price, with interests,
until we change.
--
Best regards,
Andres mailto:sqrmax at optonline.net
More information about the Squeak-dev
mailing list
|