[antlr-interest] Exception tests eat performance?

[Ney, Richard]

>-----Original Message-----
>From: Terence Parr [mailto:parrt at cs.usfca.edu] 
>Sent: Wednesday, January 07, 2004 10:20 PM
>To: antlr-interest at yahoogroups.com
>Subject: Re: [antlr-interest] Exception tests eat performance?
>
>
>On Wednesday, January 7, 2004, at 08:30  PM, Ney, Richard wrote:
>
>> Hey Karl,
>>
>> If you are looking at performance issues one item you might look at
is
>> how often are you using the tree parser?
>>
>> The tree parser in the general nature of how it works (i.e. walking a
>> linked list of tree nodes) is a major performance killer. This is due

>> to
>> the problem of the next node not being in the CPU's data cache thus
>> requiring a trip to main memory for the data.
>
>Hi Richard,
>
>I'm not sure you can compare a hand-built class of unknown structure to

>the regular tree parsers.  The algorithm is linear, but possibly with a

>large constant in front.  I doubt the CPU cache is the big issue,  
>particularly if you do multiple walks...the entire tree might fit in a

>cache if it's small enough.
>
>It is a good point to bring up thought that walking the entire tree  
>with the parser is sometimes slower than you need it to be...irregular

>walking can be a big win as you say.
>
>Ter
>
>> In my component I was initially using the tree parser to do a lot of
my
>> runtime operations and was facing major performance issues. I changed

>> my
>> application (which is in C#) to generate a runtime class to handle
the
>> operations with major performance improvements.

I don't know if it was implied but I felt I should specify that the
custom class is generated with the Tree Parser walking the same tree
structure that I use to walk for data processing. So I do take a certain
cost up front generating the class but successive runs run extremely
fast. So it is also leans more towards you evaluation Ter, the CPU cache
may play a small part but the custom class versus invoking the tree
parser is the major part of the performance improvement.

-Richard

>>
>> Don't know if this helps you but I figured I'd mention it.
>>
>> -Richard
>>
>>
----------------------------------------------------------------------- 
>> -
>> ----
>> Richard Ney                                     Aspect Communications
>> Principal Software Engineer
>>
>> <http://www.aspect.com>                         Main:  408.325.2200
>> <mailto:richard.ney at aspect.com>            	SJ Office: 408.325.2464
>> Home Office: 916.797.9602
>>
----------------------------------------------------------------------- 
>> -
>> ----
>>
>>
>> The Three Laws of Infernal Dynamics:
>> 1. An object in motion will always be headed in the wrong direction.
>> 2. An object at rest will always be in the wrong place.
>> 3. The energy required to change either of these states will always
be
>> more than you wish to expend, but never so much as to make the task
>> appear prospectively impossible.
>>
>>
>>
>>
>> -----Original Message-----
>> From: Karl Meissner [mailto:meissnersd at yahoo.com]
>> Sent: Wednesday, January 07, 2004 11:40 AM
>> To: antlr-interest at yahoogroups.com
>> Subject: Re: [antlr-interest] Exception tests eat performance?
>>
>>
>>> But fine, I used returned values in my rules all the time too.
>>> The traditional alternatives to a return value are
>>>
>>> 1) reference parameters
>>
>> They don't work for primitives in Java unfortunately.
>
> Of couse.
>
> ResultInfo would be a object.
>
> whew!
> did not mean to start a flame war.
>
> btw Terr,  I like antlr a lot.  Thanks.
> I am using it in product that is close to shipping.
> So I am starting to think about speedups.
>
> Good luck
>
> Karl
>
>
> __________________________________
> Do you Yahoo!?
> Yahoo! Hotjobs: Enter the "Signing Bonus" Sweepstakes
> http://hotjobs.sweepstakes.yahoo.com/signingbonus
>
>
>
> Yahoo! Groups Links
>
> To visit your group on the web, go to:
>  http://groups.yahoo.com/group/antlr-interest/
>
> To unsubscribe from this group, send an email to:
>  antlr-interest-unsubscribe at yahoogroups.com
>
> Your use of Yahoo! Groups is subject to:
>  http://docs.yahoo.com/info/terms/
>
>
>
>
>
> Yahoo! Groups Links
>
> To visit your group on the web, go to:
>  http://groups.yahoo.com/group/antlr-interest/
>
> To unsubscribe from this group, send an email to:
>  antlr-interest-unsubscribe at yahoogroups.com
>
> Your use of Yahoo! Groups is subject to:
>  http://docs.yahoo.com/info/terms/
>
>
>
--
Professor Comp. Sci., University of San Francisco
Creator, ANTLR Parser Generator, http://www.antlr.org
Co-founder, http://www.jguru.com
Co-founder, http://www.knowspam.net enjoy email again!
Co-founder, http://www.peerscope.com link sharing, pure-n-simple




 

Yahoo! Groups Links

To visit your group on the web, go to:
 http://groups.yahoo.com/group/antlr-interest/

To unsubscribe from this group, send an email to:
 antlr-interest-unsubscribe at yahoogroups.com

Your use of Yahoo! Groups is subject to:
 http://docs.yahoo.com/info/terms/ 



 

Yahoo! Groups Links

To visit your group on the web, go to:
 http://groups.yahoo.com/group/antlr-interest/

To unsubscribe from this group, send an email to:
 antlr-interest-unsubscribe at yahoogroups.com

Your use of Yahoo! Groups is subject to:
 http://docs.yahoo.com/info/terms/