Browse Prior Art Database

Decoding of a Specific Performance Monitor Encoded Request to a Generic Interface Format

IP.com Disclosure Number: IPCOM000118032D
Original Publication Date: 1996-Aug-01
Included in the Prior Art Database: 2005-Mar-31
Document File: 4 page(s) / 120K

Publishing Venue

IBM

Related People

Levine, FE: AUTHOR

Abstract

An algorithm is defined that takes machine specific support and maps it into a Generic Interface format.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 44% of the total text.

Decoding of a Specific Performance Monitor Encoded Request to a Generic
Interface Format

      An algorithm is defined that takes machine specific support
and maps it into a Generic Interface format.

      Based upon the criteria set out in the disclosure entitled
"Generic Performance Monitor Interface Approach," identify an
algorithm which remaps requests processed by the encoding routine
described in the  disclosure entitled "Generic Performance Monitor
Interface Event List Encoding" that are passed as machine specific
input back to the generic  interface support.

      The decode routine takes a machine specific encoded request and
maps it back to a generic request.  This routine uses machine
specific tables to remap the encoded requests back to the generic
request. The  major purpose for providing this support is to identify
exactly what is  supported by the machine specific request.  For
example, the sample time  requested may only be supported as an
approximate value.  This routine  identifies the actual sample time
supported by the machine. Similarly,  the threshold request may only
be approximated by the original encoding  routine.

      The mapping of specific flags from the encoded state is only
done for the first entry of the set of requests and is fairly
straight forward.  The mapping of the encoded list of events back to
a generic list, which includes a "trigger event" and a "correlate
event" is not as  straightforward and is described next.  One area of
difficulty is that  an implementation may allow for the default
selections of events. Under  this circumstance, the default events
may or may not be intended as correlate events.  This algorithm
identifies an event as a correlate event, if the event has a
non-default selection code and is repeated or  if the event is
counted on every pass.  The algorithm could be extended  to provide a
list of events that are counted on each pass.  This algorithm might
be helpful in a situation where only MMCRn values are supplied for
decoding.

      In some processors, the default selection value of zero (0)
does not count any specific item and thus selection fields of zero
may be ignored for reconstructing the original set of requests from
the selection values.  However, there may be some processors, where
the selection values of zero represent default selection items.
Default selection items are expected to be items that are commonly
counted. For  the default selection case, mapping from the selection
values to a list  of items to count may include items that are not
from the original list.  There may be as many as eight new items,
that is, one for each counter.  If the machine specific list was put
together without going through the encoding routine, there could be
more than the original MAX_EVENTS.  Thus, the mapping back to the
original list could contain  more events than the original space
reserved for events.  Instead of making this a...