Performance Questions and a little need for some education.

[Bob Deskin]

I checked my recollection with development. DEFINEs in QUIZ do require a 
memory storage area (buffer) because they get evaluated in the input 
phase. Note that if we output the defined item then it also becomes part 
of the sort record.

DEFINEs in QTP (other than DEFINEs with PARM) do not need a buffer. The 
expression is stored of course, as it is in QUIZ, but the result of the 
expression processing is stored in the target area rather than in a DEFINE 
buffer. TEMP items have a buffer.

SO one could argue that if a DEFINE and a TEMP are evaluated at the same 
time and moved into an output record at the same time, the DEFINE should 
be slightly faster since the expression evaluation is moved directly into 
the target whereas the TEMP is evaluated, moved into the TEMP buffer, and 
then into the target.

By the way, isn't it Itanium (as opposed to Titanium)?

Bob



From:
John Stires <jstires at EBDS.com>
To:
"powerh-l at lists.sowder.com" <powerh-l at lists.sowder.com>
Date:
2009-08-18 11:01 AM
Subject:
Performance Questions and a little need for some education.
Sent by:
powerh-l-bounces+bob.deskin=ca.ibm.com at lists.sowder.com



Thanks for the input.  I may need just a little more to put my mind to 
rest on this, though,
 
I have run similar bench tests before and the results have always been 
consistent.  This is the first time I have seen them vary.  The only thing 
I see different in this case, is the Titanium machine and a little later 
version of PowerHouse.  I am wondering if there is some tuning that is 
done on the Titanium machine that is unique to cause this kind of results. 
 The other possible cause might be that COGNOS has refined QTP to manage 
its baggage much better than it used to do.  As it is, the results on the 
ALPHA  are consistent to what I have seen in the past.  It is just on the 
Titanium that I am seeing such a difference.
 
In this case the output file is just a simple subfile.  There are no 
SORTs, or PARMs.  The QUIZ has only GETSYSTEMVALs to make up for an 
extract request in the QTP used only to set 3 Global TEMPs and 
GETSYSEMVALs are  only evaluated once in QUIZ.
 
In the past, I have never seen QTP out-perform QUIZ for simply extracting 
data and dumping it to a subfile, especially where there is no totaling or 
summarizing going included.  In this case the data is getting dumped into 
a subfile which is then renamed after the extract is complete.
 
My understanding is that DEFINES are done in registers while TEMPs use 
memory, which easily explains why DEFINEs have to be reevaluated and TEMPs 
do not.  All things being equal, and no DEFINE cascading, DEFINES would be 
faster.  Register access much faster than general memory.  It is only when 
you have multiple references, the cascading, that DEFINEs that the “fast” 
DEFINEs have to reevaluate and slow performance.  Register values are only 
transient.
 
QTP, as you say is robust and carries a tremendous amount flexibility, but 
with that, come a lot of baggage that is loaded whether it is used or not. 
 Whereas QUIZ does not carry such a large amount of baggage and, generally 
gets through files pretty awful fast by comparison for simple extracts. 
 
Let me show a little more of my lack of knowledge and ask what (IMHO) is?
 
Thanks,
John-- 
= = = = = = = = = = = = = = = = = = = = = = = = = = = =
Mailing list: powerh-l at lists.sowder.com
Subscribe: 'subscribe' in message body to 
powerh-l-request at lists.sowder.com
Unsubscribe: 'unsubscribe &lt;password&gt;' in message body to 
powerh-l-request at lists.sowder.com
http://lists.sowder.com/mailman/listinfo/powerh-l
This list is closed, thus to post to the list you must be a subscriber.
Add 'site:lists.sowder.com powerh-l' to your search terms to search the 
list archive at Google.