Post-Link Optimization for Linux on POWER
A post-link optimization utility for the POWER architecture that optimizes an executable program or a shared library, based on its run-time profile.
Date Posted: June 30, 2004
|
|
 |
|
|
Update: July 21, 2008
New version includes support for programs built by GCC4.3 (expected to be the basis of forthcoming RHEL6 and SLES11); Power 6-related optimizations; and many bug fixes. See details in forum and man page.
What is Post-Link Optimization for Linux on POWER?
Post-Link Optimization for Linux® on POWER™, also known as FDPR-Pro, is a performance-tuning utility used to improve the execution time and the real memory utilization of user-level application programs, based on their run-time profiles.
How does it work?
The tool optimizes the executable image of a program by collecting information on the behavior of the program while the program is used for some typical workload. It then re-analyzes the program (together with the collected profile), applies global optimizations (including program restructuring), and creates a new version of the program that is optimized for that workload. The new program generated by the optimizer typically runs faster and uses less real memory than the original program.
Example of a typical use: A database application could be profiled while a representative stream of requests and transactions is being performed. With the profile available, the Post-Link Optimization tool analyzes the program and creates an optimized version of it, maximizing its performance for the given run-time profile. For example, a large database server, whose large execution memory exceeds cache size, can be restructured so that its effective size, for the representative workload, will fit into the cache and thus execute faster.
Because the post-link optimization comes at the end of the development process, adding it to an existing development build of an application does not disrupt the existing build. The compile-and-link stage proceeds as-is, with the post-link optimization added as the last step.
|
|
|
| | About the technology author(s):
Yaakov Yaari, Ph.D., works with Code Optimization Technologies in the IBM® Haifa Research Lab. He holds several patents primarily in the area of computer architecture. His interests are in computer architecture, performance analysis, and compilation tools. Dr. Yaari has been with IBM since 2001.
Moshe Klausner works with Code Optimization Technologies in the IBM Haifa Research Lab. His interests include post-link optimization algorithms and flight simulators. Mr. Klausner has been with IBM for several years; previously he worked in the Isreali Air-Force in the area of flight simulators.
Alex Kogan works with Code Optimization Technologies in the IBM Haifa Research Lab. His interests are in code optimization techniques and algorithms.
Other contributors include Guy Bashkansky, Gad Haber, Roy Levin, Maxim Gurevich, and Vadim Eisenberg.
Linux is a registered trademark of Linus Torvalds in the United States, other countries, or both.
IBM and POWER are trademarks of IBM Corporation in the United States, other countries, or both.
Other company, product, or service names may be trademarks or service marks of others.
|
|
|
| |
