Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Combined Event Performance Trace for Aix

IP.com Disclosure Number: IPCOM000099951D
Original Publication Date: 1990-Mar-01
Included in the Prior Art Database: 2005-Mar-15
Document File: 1 page(s) / 49K

Publishing Venue

IBM

Related People

Albaugh, VA: AUTHOR

Abstract

Disclosed is a design that permits event and performance tracing of system software execution natively in the AIX* environment. Prior to this implementation, performance tuning on AIX was done primarily by the use of additional hardware and a connection to a separate processor. Key to the design is the efficient availability of a high resolution timer. Such a timer has been architected as a special-purpose register for future RISC processors.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 63% of the total text.

Combined Event Performance Trace for Aix

       Disclosed is a design that permits event and performance
tracing of system software execution natively in the AIX*
environment.  Prior to this implementation, performance tuning on AIX
was done primarily by the use of additional hardware and a connection
to a separate processor.  Key to the design is the efficient
availability of a high resolution timer.  Such a timer has been
architected as a special-purpose register for future RISC processors.

      The trace data collection mechanism consists of a device
driver, trace recording routines, a process that reads the data and a
process that monitors and modifies the trace state.  The device
driver allocates and pins trace buffers during open and configures
the trace for the specified management of those buffers (circular
fill, single sweep, last buffer, or continuous to file).  Trace data
is passed in registers to trace recording routines which record the
data for the event atomically in the buffer along with a time stamp.
The use of routines, rather than inline code, permits the routines to
be different for various trace configurations (for example, the data
can be directed to the bus rather than to memory, or events can be
real time filtered).  When trace is inactive, these routines simply
return.  Activating trace incurs a very small delta to system
overhead such that execution flow is not appreciably altered.  The
reading process reads the trace buffers and...