 |
BorlandTalk.com Borland discussion newsgroups
|
| View previous topic :: View next topic |
| Author |
Message |
David Novo Guest
|
Posted: Thu Dec 08, 2005 7:22 pm Post subject: speed penalties |
|
|
Hello,
It seems that "better design" seems to involve creating lots of little
objects that help with different things. Instead of hard coding stuff,
and creating descendants with large inheritance trees, you encapsulate
the variation within objects.
I am just wondering, is there a potential of speed penalties in this
regard. I can forsee that even for basic things, we are going to be
creating many more objects/interfaces and passing them around and then
freeing them.
Is this a realistic issue that comes up and we should be aware of it,
or for all intents and purposes it doesn't happen?
Thanks,
Dave
|
|
| Back to top |
|
 |
Iman L Crawford Guest
|
Posted: Thu Dec 08, 2005 8:35 pm Post subject: Re: speed penalties |
|
|
"David Novo" <dave (AT) xxx (DOT) denovosoftware.xyz.com> wrote in news:439895fc$1
@newsgroups.borland.com:
| Quote: | Is this a realistic issue that comes up and we should be aware of it,
or for all intents and purposes it doesn't happen?
|
I think this is a valid concern. When it becomes a concern will most
likely be application specific.
--
Iman
|
|
| Back to top |
|
 |
Bob Dawson Guest
|
Posted: Thu Dec 08, 2005 10:58 pm Post subject: Re: speed penalties |
|
|
"David Novo" wrote
| Quote: |
It seems that "better design" seems to involve creating lots of little
{...] |
In general I'd agree with that, and not just at the object level. I find
myself looking suspiciously at any method that doesn't fit in a normal edit
window.
| Quote: | I am just wondering, is there a potential of speed penalties in this
regard.
|
Potentially, but not generally. At the speed of today's processors, basic
memory allocation and release operations are pretty fast. And the new memory
manager in D2006 is supposed to be better specifically with this pattern of
memory usage (also multi-threading. IIRC).
bobD
|
|
| Back to top |
|
 |
Jim Cooper Guest
|
Posted: Thu Dec 08, 2005 11:12 pm Post subject: Re: speed penalties |
|
|
| Quote: | Is this a realistic issue that comes up and we should be aware of it,
|
Yes, to a point. However, may programmers suffer from premature optimisation,
spending ages speeding up code that just doesn't matter.
Write the code first, profile afterwards. You may be surprised what
inefficiencies you can get away with :-)
The upside of many small methods (the number of classes is less important for
this) is that profiling then tends to be more specific. It's almost always
easier to work out how to speed up 5 lines of code than 50.
Cheers,
Jim Cooper
__________________________________________
Jim Cooper [email]jcooper (AT) tabdee (DOT) ltd.uk[/email]
Skype : jim.cooper
Tabdee Ltd http://www.tabdee.ltd.uk
TurboSync - Connecting Delphi to your Palm
__________________________________________
|
|
| Back to top |
|
 |
Gerrit Beuze Guest
|
Posted: Fri Dec 09, 2005 7:37 am Post subject: Re: speed penalties |
|
|
| Quote: | Is this a realistic issue that comes up and we should be aware of it,
or for all intents and purposes it doesn't happen?
|
Yes, it is an issue, but the performance issue usually appear in places where you did not expect it.
My experience is that one should focus on the correct design - and yes, that in some cases involves
lots of little objects. And only if there's a performance problem use a profiler to find bottlenecks.
I have more than once found that I was worried about performance after some re-design
that involved yet more small objects - but no customer ever complained.
For example: when profiling ModelMaker Code Explorer it turned out that
I could gain lots in places where I had never expected it. And it was quite easy to
find a different approach, leaving the design intact.
For a good profiler: check AutomatedQA's AQTime. It does profiling without
instrumenting source code + comes with lots of different profilers.
Gerrit Beuze
ModelMaker Tools
|
|
| Back to top |
|
 |
David Novo Guest
|
Posted: Sat Dec 10, 2005 12:10 am Post subject: Re: speed penalties |
|
|
Thanks for all the comments! We already do use AQTime, which I agree
with Gerrit is an excellent product, but probably not as often as we
should.
Then again, we don't refactor as often as we should...
update the documentation as often as we should....
make test cases as often as we should.....
take vacation as often as we should :-)
|
|
| Back to top |
|
 |
|
|
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
|
|