MVSFORUMS.com

tattva · Beginner Joined: 02 Feb 2005 Posts: 97 Topics: 36

Hi all ,

i need to replace the below piece of code with a more efficient code in order to improve the system performance. let me know ur suggestions

PERFORM
VARYING index1 FROM 1 BY 1
UNTIL index1 > MAX-COUNT
IF CUST-NUM(index1)
= CUST-NUM(index2)
AND dia-num(index1)
>= '40000'
AND dia-num(index1)
<= '49999'
STATEMENTS...
......
END-PERFORM

this is consuming a lot of CPU time and needs to be tuned.

Could you guys help me out...

TIA
--------
tattva

kolusu · Posted: Wed Feb 02, 2005 8:16 am Post subject:

Tatva,

How are INDEX1 & INDEX2 defined?

Using indexes to address a table is more efficient than using subscripts since the index already contains the displacement from the start of the table and does not have to be calculated at run time. Subscripts, on the other hand, contain an occurrence number that must be converted to a displacement value at run time before it can be used. When using subscripts to address a table, use a binary (COMP) signed data item with eight or fewer digits (for example, using PICTURE S9(8 ) COMP for the data item). This will allow fullword arithmetic to be used during the calculations. Additionally, in some cases, using four or fewer digits for the data item may also offer some added reduction in CPU time since halfword arithmetic can be used.

Also check this link for a detailed explanation on COBOL for MVS & VM Performance Tuning

http://www-1.ibm.com/support/docview.wss?org=SW&doc=7001476&aid=1

Hope this helps...

Cheers

Kolusu
_________________
Kolusu
www.linkedin.com/in/kolusu

tattva · Beginner Joined: 02 Feb 2005 Posts: 97 Topics: 36

Hi kolusu

Thanks for ur reply. But i have a couple of more ques here .

1) can i some how replace the exisiting code with Binary seacrh
2) is there any performace based differnce in searching a table by using
a) perform varying
b) search

as far as i know both a & b can be used to do a sequential search.correct me if i am wrong here!!!

TIA
------
Tattva

programmer1 · Beginner Joined: 18 Feb 2004 Posts: 138 Topics: 14

I have observed SEARCH performing better than the Perform Varying loop.
_________________
Regards,
Programmer

Bithead · Posted: Wed Feb 02, 2005 9:58 am Post subject:

SEARCH ALL provides the best performance. If there are duplicate CUST-NUM entries, try a SEARCH ALL then decrease the index until the CUST-NUM is less than the required value, then increase it by 1.

powerhawk · Posted: Wed Feb 02, 2005 10:54 am Post subject:

It's no big performance difference between SEARCH and SEARCH ALL if you have a small table (10-50 entries). If it's possible to put the most search values in the first entries you can get better performance with SEARCH than with SEARCH ALL. With bigger tables SEARCH ALL is much faster. If the table have 1000 entries SEARCH ALL is 83% faster in Enterprise Cobol.
To get best performance you should:
- Use SEARCH ALL
- Use OCCURS DEPENDING ON and declare the indexes with INDEXED BY.
- Of course sort the table, if you don't SEARCH ALL will fail.

powerhawk · Posted: Wed Feb 02, 2005 10:59 am Post subject:

I guess you can get a small speed improvement by moving CUST-NUM(index2) to a field declared outside the table, like W-CUST-NUM, and change the test in the SEARCH statement to: WHEN CUST-NUM(index1) = W-CUST-NUM...

powerhawk · Posted: Thu Feb 03, 2005 2:28 am Post subject:

Here is sample code:

tattva · Beginner Joined: 02 Feb 2005 Posts: 97 Topics: 36

Hi Power,

thanks for the reply...

But if i have to replace it with ur logic.. then what about the third condition..
AND dia-num(index1)
>= '40000'
AND dia-num(index1)
<= '49999'

this cannot be taken care in Binary search .. right

Correct me if i am wrong here!!

Thnx
-------
Tattva

powerhawk · Posted: Thu Feb 03, 2005 6:43 am Post subject:

Sorry I don't know. I would fix a table sorted in ascending order on both fields and try. If it's possible you can also avoid to load all items with dia-num outside 4000-4999, or create a work table and move all items with dia-num from 4000 to 4999 from the original table to the work table.

Bithead · Posted: Thu Feb 03, 2005 10:42 am Post subject:

Tattva,

Please show all the code in the PERFORM?

slade · Posted: Thu Feb 03, 2005 8:56 pm Post subject:

Another question is how are these 2 tables populated? Granted I don't know the full picture, but I get the feeling this might be better solved using a "file match" algorithm.
_________________
Regards, Jack.

"A problem well stated is a problem half solved" -- Charles F. Kettering

tattva · Beginner Joined: 02 Feb 2005 Posts: 97 Topics: 36

hi guys,

Thanks for all your responses.. i was able to solve this.. But am facing some other Problem

Here's the problem

say am now doing a binary search on a table entries of which are loaded from a cursor..

1) should all the keys on which i am searching the table be present in the ORDER by clause ?

TIA
------
Tattva

kolusu · Posted: Fri Feb 04, 2005 9:49 am Post subject:

slade · Posted: Sat Feb 05, 2005 2:00 pm Post subject:

Hi Tattva,

I'm not sure I understand the problem stmt you provided. You want to use SEARCH ALL using CUST-NUM as the search arg.

Then you compare CUST-NUM(IDX1) equal to CUST-NUM(IDX2). So far as I know you can't have dupe keys in a SEARCH ALL tbl. It'll work but the results are iffy. I'm assuming that CUST-NUM(IDX2) is referring to a different instance of the same data field in the same table as CUST-NUM(IDX1), since they both reference the same unqualified dataname. Is that correct?

BTW, how is the value of IDX2 determined?

All that aside, I guess you're trying to do something like this, as illustrated in Powerhawk's code (but that WHEN clause bothers me) Confused

: