home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!wupost!cs.utexas.edu!swrinde!network.ucsd.edu!news.acns.nwu.edu!nucsrl!ddsw1!inland!cmkrnl!jeh
- From: jeh@cmkrnl.com
- Newsgroups: vmsnet.internals
- Subject: Re: monitor locations (for processor modes)
- Message-ID: <1992Jul27.211317.605@cmkrnl.com>
- Date: 28 Jul 92 04:13:17 GMT
- References: <7520241@MVB.SAIC.COM>
- Organization: Kernel Mode Consulting, San Diego, CA
- Lines: 31
-
- In article <7520241@MVB.SAIC.COM>, "Bob George" <BOB@aeneas.ims.disa.mil> writes:
- > I've been looking through the manuals, but haven't seen it yet. Does
- > anyone know the symbolic system locations that monitor uses to determine
- > the time spent in each of the processor modes? Thanks,
- > bob
-
- Sure. They're in the CPU$ data structure. It's a series of seven longwords,
- the first of which is called CPU$L_KERNEL. The others don't appear to have
- distinct names as they are accessed mostly by clever uses of index registers
- with values found in the "access mode" and "interrupt stack" fields of the PSL.
- (see the section titled "hardware clock interrupts" in TIMESCHDL.LIS in
- facility SYS in the listings kit.) However their contents are in order as
- described in the description of this structure in the Device Support Reference
- Manual. Their values represent numbers of 10 millisecond clock ticks since
- system boot.
-
- on an SMP (multi-CPU) machine, you have to look at all the CPU$ structures to
- get the totals for the node.
-
- btw, "idle" time as recorded in the CPU$ structure is a subset of time spent on
- the interrupt stack. ie you need to subtract the idle time (which is, roughly,
- the time spent with no process current) from the recorded interrupt stack time
- in order to get the true time spent on the interrupt stack. Once you've done
- that, the seven values added together will add up to the time (or at least
- the number of hardware clock interrupts recorded) since system boot.
-
- --- Jamie Hanrahan, Kernel Mode Consulting, San Diego CA
- uucp 'g' protocol guru and release coordinator, VMSnet (DECUS uucp) W.G., and
- Chair, VMS Programming and Internals Working Group, U.S. DECUS VAX Systems SIG
- Internet: jeh@cmkrnl.com, hanrahan@eisner.decus.org, or jeh@crash.cts.com
- Uucp: ...{crash,eisner,uunet}!cmkrnl!jeh
-