Hello Philippe

A comment about replacing LTR/BZ combinations with CIJE  -   the compare 
immediate instructions do NOT set the condition code so you need to make 
sure that the BZ is the only instruction testing the condition code.

Also earlier, you made a comment about Execute that you need to review 
these as well - my understanding is that Execute only modifies the 
instruction instream and as such does not attempt to update the 
instruction cache and the subsequent delays that causes.     Of course the 
instruction that is being Executed may be a cause of the problem in 
itself.

If you do have non-rent programs with embedded save areas, then you should 
see a benefit from making sure that the save area is in a different cache 
line - the judicious use of ORG  to place code and data onto specific 
boundaries may help here.   For example,    ORG   ,256    will move the 
location counter to the next 256 byte boundary - you will need to use the 
assembler options GOFF and SECTALGN  for this to work, as well as binding 
the module so that it is loaded onto the correct boundary

Kind Regards

John

Internet: [email protected]
Phone: +61 8 926 18473 (xtn 18473; Tie-Line: 701 8473)




From:   Philippe Cloarec <[email protected]>
To:     [email protected]
Date:   12/11/2016 21:16
Subject:        Re: SIIS "issue" after upgrade to z13 machine.
Sent by:        IBM Mainframe Assembler List 
<[email protected]>



Hi Martin and Rob, 

First thx much for your input.  Here are 3 links which relate the "issue" 
and common code for which we fall into a SIIS scenario:
https://www.google.fr/#q=istream_flash_062606_v4
http://s3-us-west-1.amazonaws.com/watsonwalker/ww/wp-content/uploads/2016/03/06173415/18017-The-Cheryl-and-Frank-zRoadshow.pdf

http://conferences.gse.org.uk/attachments/presentations/ibHo4j_1446285934.pdf


Unfortunately I have more than 1000 pgms to review and handle on a short 
timeframe,  so I will have no time to implement "Baseless  processing". I 
do plan to SAK(Search And Kill) SIIS occurences for scenarios described in 
above links I will find   and implement newer instructions as immediate 
ones  to eliminate memory references and constants in storage as well. 

I will check cases where two instructions can be replaced by one only...to 
use CIJE in place of LTR/BZ combination as for example. 

Since we do talk of CPU cycles savings here I will check for AGI cases and 
their resolution and try to implement instruction grouping as much I can. 

>From my humble point this is a real topic and all z13 sites having old 
productions Batch programs should perform some action.

regards
Philippe 

Reply via email to