Browse Prior Art Database

Missing symbol detection and reporting for performance analysis

IP.com Disclosure Number: IPCOM000013229D
Original Publication Date: 2000-Mar-01
Included in the Prior Art Database: 2003-Jun-18
Document File: 1 page(s) / 36K

Publishing Venue

IBM

Abstract

One of the primary problems with the information used by tprof and mtrace is the availability of symbolic information identifying internal subroutines in the system under test. Specifically, the names of the static subroutines or subroutines internal to a procedure are not typically embedded in the executable file itself. In fact the names may not even available at the link edit step unless special compile and link options are used. The avoidance of this problem is well known in the industry in the shipping of .dbg or map files that contain the required information. However, when this information is used by the performance analysis tools, the information as to whether or not all of the required symbolic information is being used is not readily available. The outcome of this lack of information is that the reports provide incorrect information: Subroutine A externalized

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 69% of the total text.

Page 1 of 1

Missing symbol detection and reporting for performance analysis

One of the primary problems with the information used by tprof and mtrace is the availability of symbolic information identifying internal subroutines in the system under test. Specifically, the names of the static subroutines or subroutines internal to a procedure are not typically embedded in the executable file itself. In fact the names may not even available at the link edit step unless special compile and link options are used. The avoidance of this problem is well known in the industry in the shipping of .dbg or map files that contain the required information. However, when this information is used by the performance analysis tools, the information as to whether or not all of the required symbolic information is being used is not readily available. The outcome of this lack of information is that the reports provide incorrect information: Subroutine A - externalized

.

.

Subroutine B - static-not externalized and not identified

.

.

Time in subroutine B is reported as time in Subroutine A

Disclosed is an approach to identify the occurrence of this condition so that when this occurs the reports indicate that this condition has occurred.

During the "gathering of symbols phase," we have typically have the executables available for analysis. This analysis is typically done off-line, e.g., not during the performance run. However, just prior to or just after the performance run, the executables are typi...