Browse Prior Art Database

Op Code Sampling

IP.com Disclosure Number: IPCOM000035760D
Original Publication Date: 1989-Aug-01
Included in the Prior Art Database: 2005-Jan-28
Document File: 2 page(s) / 13K

Publishing Venue

IBM

Related People

Westcott, DW: AUTHOR [+2]

Abstract

Disclosed is a system for collecting data about the execution of instruction streams in a parallel processing environment that does not significantly degrade execution time or change the executed instruction stream at a small circuit cost.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

Page 1 of 2

Op Code Sampling

Disclosed is a system for collecting data about the execution of instruction streams in a parallel processing environment that does not significantly degrade execution time or change the executed instruction stream at a small circuit cost.

Parallel processing can divide software tasks into several subtasks which execute in multiple processors at the same time. In some instances, one or more of the subtasks results may be discarded as a result of the outcome of another subtask. As a result of this discarding, the simple monitoring of instructions in each of the parallel processors cannot be used to draw conclusions about the software task since both used and discarded instructions are included in a simple monitoring scheme.

High performance processors execute instructions at a rate of 10 or more times faster than the fastest available recording mediums (tape or disk). This requires any collection of data about processor instructions streams to collect data at a high speed and read it out at a slow speed, resulting in only a sampling of the available data. A Random-Access Memory (RAM) can be used for this purpose.

Simply taking a software task and executing it in a non-parallel environment, be it a single processor or serialization of multiple processors, changes the stream of instructions executed in the processor. The input and output demands on the instruction stream change since they occur less frequently, changing the effective instruction stream executed. Running the software on a single processor, or a serialized set of processors, and collecting the instruction stream executed would not be indicative of the instruction stream executed on parallel processors. It would also degrade the exe...