I have a software offering that Dave Cole markets. You can go to his
website, and take a look at z/XPF. It competes in a niche market on z/OS,
application profiling. What makes mine different from the other guys in this
arena is the approach I use to get the profiling data. Instead of
establishing tasks or running srb's into an address space, my code scrapes
the system trace table for entries that are related to the target address
space. This approach provides for a variety of statistical data about
application performance that the other products know nothing about.
Although I have been working on this since late in 2005, it is really still
in its infancy as to what it eventually will become.
One of my current customers requested the ability to communicate to my
profiler that the application had passed a specific point in their code...a
trace point. Since my code already has the ability to pick up SVC entries
from the trace table, this seemed the most logical approach. Hence, a do
nothing SVC whose sole purpose was to communicate to existing code that an
event had occurred. (This is a shorthand version of what is involved). A
user would then be able to bracket pieces of their code, and get a count and
elapsed time data for how often the code went thru the trace point, and the
elapsed time between one trace point and the next, which formed a pair of
brackets in their code.
I have abandoned the SVC approach, and am now working on the code to use
PTRACE instead. PTRACE will function in any environment, which makes it
more appealing. But, you only get 5 words of trace data with every trace
entry created using this approach. Means I will have to process 3 trace
records to get the trace data I want. The SVC approach generated two trace
entries, SVC enter and SVC return. Tradeoffs in everything.
I asked the original question, just because the manuals seemed ambiguous
as to the correct course. Peter Relson's comments on having the right data
in a dump when someone is looking at it is spot on. I didn't think of that
at all until he mentioned it. Tunnel vision.
--Dave Day
----- Original Message -----
From: "John Gilmore" <[email protected]>
Newsgroups: bit.listserv.ibm-main
To: <[email protected]>
Sent: Tuesday, January 03, 2012 12:36 PM
Subject: Re: Question on adding an SVC routine dynamically to a running
system
Tom Marchant wrote:
| What's wrong with just issuing a TIME macro to get the time entry?
Notionally, this would greatly reduce system overheads; but if elapsed
time is what is required, what's wrong with a pair of STCKEs? Using
them would reduce these overheads to zero.
It may be, however, that we don't fully understand the problem. BD
knows hows to use the TIME macro.
John Gilmore, Ashland, MA 01721 - USA
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
