GTF HOOK issues a MONITOR CALL instruction which is handled by the program check FLIH.
Warren Brown -------------- Original message from Bill Fairchild <[email protected]>: -------------- From: IBM Mainframe Discussion List [mailto:[email protected]] On Behalf Of > Tom Marchant > Sent: Wednesday, March 11, 2009 9:19 AM > To: [email protected] > Subject: Re: How much does GTF trace impact system throughput? > > Bill Fairchild wrote: > > > >The GTF hook causes a program interrupt > > Really? I'm surprised. It's not the first time. > > Tom Marchant > > > It's documented in the PoOps book. Check out MC (Monitor Call). If you are > not tracing any events with GTF or any other user of the Monitor Call trace > facility, then all the various MC instructions scattered around IBM's code > are executed as NOP instructions. There are 16 possible classes of events, > such as I/O, SVC, program interrupts, etc. If you have enabled tracing for > event class X, then wherever there is a MC instruction that specifies event > class X you will get a program interrupt with interrupt code X'0040' when you > go through that code. Program Interrupt FLIH checks the interrupt code; if it > says the PI was > caused by a MC instruction, then you go off into MC interrupt handling code > which checks for filtering, etc., as I wrote before. I built a MC trace > facility application once before, and it was relatively easy. The tracing > routines that are entered are covered by a SETFRR before the MC code calls > the > tracing routine, so I only rarely crashed my test system.! > The only real problem I had was in making it work while GTF was also running. > This was back in 1987 before IBM enhanced the MC trace facility so that > multiple > MC applications could run simultaneously. GTF is but one user of the Monitor > Call trace facility. I don't know of any others except the one I built into > TMON/MVS, but there is nothing in the MC architecture to prevent a large > number > of different users of this trace facility from all running simultaneously and > tracing the same events. > > Bill Fairchild > Rocket Software > > ---------------------------------------------------------------------- > For IBM-MAIN subscribe / signoff / archive access instructions, > send email to [email protected] with the message: GET IBM-MAIN INFO > Search the archives at http://bama.ua.edu/archives/ibm-main.html ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
