Well, the fixes I put in helped to reduce that by reducing the number
of temporary objects that got created during processing.  If the real
issue was excessive GC cycles, then I fixed that problem back about
the year 2000 before Object Rexx was even open source.

Interestingly, when I was looking at this problem back in 3.2.0, I
made a number of attempts to create the an array object the size that
would be needed for the actual result...and strangely, every single
attempt at doing that resulted in SLOWER performance.  This one
seriously puzzled me!  It turned out to be faster to have the arrays
extend themselves than to use lines() to get an accurate count first.
Probably because the cost of I/O vastly outweighs the other factors.

Rick

On Thu, Jul 3, 2008 at 9:51 AM, Gil Barmwater <[EMAIL PROTECTED]> wrote:
> Rick McGuire wrote:
>> On Thu, Jul 3, 2008 at 9:30 AM, Gil Barmwater <[EMAIL PROTECTED]> wrote:
>>
>>>With all the changes to STREAM in the new release, I was wondering if
>>>the performance issue with ARRAYIN has been addressed as well (I seem to
>>>remember a discussion about the cause having been located but don't know
>>>if it was an easy fix or not).
>>
>>
>> I never really found a root cause for the performance problem.  I did
>> find a few things that made arrayin a little faster, but those changes
>> were in trunk.  I've not had the chance to make any performance
>> measurements yet, so I don't know how things will compare.
>>
> I vaguely remember something about lots of garbage collection cycles
> being triggered by the arrayin processing if that helps??
>
>>
>>>BTW, GREAT JOB on ALL the changes that
>>>everyone has put into this release!  And the "tutorial" on testcase
>>>developemnt has been a great addition that, hopefully, will get us some
>>>more participation from folks that were hesitant to try.  Thanks!
>>
>>
>> Additional participation is going to be necessary if we wish to get
>> this next release out in other than "geologic time".  It's really
>> going to require a pretty thorough testing/beta test cycle given the
>> amount of core code that has changed.  And we haven't even started the
>> process of trying to get this to run on a 64-bit system yet.  In
>> theory, it should be ready to go, but once we do get it running, the
>> better the test suite, the more confidence we'll have that we can
>> release 64-bit versions.
>>
>>   "In theory, there's no difference between theory and practice...
>>   In practice, there's no similarity between theory and practice"
>>
>> Rick
>>
>>
>>
>>>Mark Miesfeld wrote:
>>>
>>>>On Wed, Jul 2, 2008 at 3:35 PM, Rick McGuire <[EMAIL PROTECTED]
>>>><mailto:[EMAIL PROTECTED]>> wrote:
>>>> > Cool!  ...
>>>> > starting showing up now because the calls to native functions are now
>>>> > significantly faster than the previous release.
>>>>
>>>>Speaking of significantly faster, I had meant to draw your attention
>>>>this.  Take a close look at the results of running the entire test suite
>>>>on 3.2.0:
>>>>
>>>>Interpreter: REXX-ooRexx_3.2.0(MT) 6.02 30 Oct 2007
>>>>ooRexxUnit:  2.0.0_3.2.0        ooTest: 1.0.0_3.2.0
>>>>Tests ran:           16589
>>>>Assertions:          537820
>>>>Failures:            2
>>>>  (Known failures:)  42
>>>>Errors:              0
>>>>Exceptions:          0
>>>>Skipped files:       0
>>>>Messages:            0
>>>>File search:        00:00:43.440000
>>>>Suite construction: 00:00:01.485000
>>>>Test execution:     00:02:04.538000
>>>>Total time:         00:02:49.900000
>>>>compared to running the entire suite on trunk:
>>>>
>>>>Interpreter: REXX-ooRexx_3.3.0(MT) 6.02 29 Jun 2008
>>>>ooRexxUnit:  2.0.0_3.2.0        ooTest: 1.0.0_3.2.0
>>>>Tests ran:           16628
>>>>Assertions:          545595
>>>>Failures:            4
>>>>  (Known failures:)  0
>>>>Errors:              0
>>>>Exceptions:          0
>>>>Skipped files:       0
>>>>Messages:            0
>>>>File search:        00:00:03.641000
>>>>Suite construction: 00:00:02.078000
>>>>Test execution:     00:02:16.321000
>>>>Total time:         00:02:22.664000
>>>>
>>>>
>>>>I would hazard a guess that the stream library it /quite /a bit faster
>>>>in 4.0. <grin>  Plus, my trunk build is a debug build.
>>>>
>>>>--
>>>>Mark Miesfeld
>>>>
>>>>
>>>>------------------------------------------------------------------------
>>>>
>>>>-------------------------------------------------------------------------
>>>>Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW!
>>>>Studies have shown that voting for your favorite open source project,
>>>>along with a healthy diet, reduces your potential for chronic lameness
>>>>and boredom. Vote Now at http://www.sourceforge.net/community/cca08
>>>>
>>>>
>>>>------------------------------------------------------------------------
>>>>
>>>>_______________________________________________
>>>>Oorexx-devel mailing list
>>>>Oorexx-devel@lists.sourceforge.net
>>>>https://lists.sourceforge.net/lists/listinfo/oorexx-devel
>>>
>>>--
>>>Gil Barmwater
>>>
>>>
>>>-------------------------------------------------------------------------
>>>Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW!
>>>Studies have shown that voting for your favorite open source project,
>>>along with a healthy diet, reduces your potential for chronic lameness
>>>and boredom. Vote Now at http://www.sourceforge.net/community/cca08
>>>_______________________________________________
>>>Oorexx-devel mailing list
>>>Oorexx-devel@lists.sourceforge.net
>>>https://lists.sourceforge.net/lists/listinfo/oorexx-devel
>>>
>>
>>
>> -------------------------------------------------------------------------
>> Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW!
>> Studies have shown that voting for your favorite open source project,
>> along with a healthy diet, reduces your potential for chronic lameness
>> and boredom. Vote Now at http://www.sourceforge.net/community/cca08
>> _______________________________________________
>> Oorexx-devel mailing list
>> Oorexx-devel@lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/oorexx-devel
>>
>>
>
> --
> Gil Barmwater
>
>
> -------------------------------------------------------------------------
> Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW!
> Studies have shown that voting for your favorite open source project,
> along with a healthy diet, reduces your potential for chronic lameness
> and boredom. Vote Now at http://www.sourceforge.net/community/cca08
> _______________________________________________
> Oorexx-devel mailing list
> Oorexx-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/oorexx-devel
>

-------------------------------------------------------------------------
Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW!
Studies have shown that voting for your favorite open source project,
along with a healthy diet, reduces your potential for chronic lameness
and boredom. Vote Now at http://www.sourceforge.net/community/cca08
_______________________________________________
Oorexx-devel mailing list
Oorexx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/oorexx-devel

Reply via email to