Actually, one of my students, Nathan Hanish, at the University of Alabama in
Huntsville added profiling to Kaffe's JIT for the IA-32 architecure this
spring. We haven't released it yet. We are hoping we could incorporate it
into the new code generator that is suppose to be out soon.
We will have a poster at Java Grande (June 12-13) in San Francisco that
describes it in more detail. The profiling makes use of inline code that
reads the Pentium timestamp register. Below is the abstract for the poster
presentation:
Profiling Java Programs via High-Resolution Timing Measurements
Nathan M. Hanish
William Cohen
Department of Electrical and Computer Engineering
University of Alabama in Huntsville
Huntsville, AL 35816
Abstract
Given the interest in high-performance Java such as Hotspot
[4][1], high-resolution, low-overhead time measurements are
required for accurate profiling of the executing program. These
time measurements allow programmers of applications and runtime
systems to identify the code that should be examined for possible
performance improvements. Modifications were made to the Kaffe
Java Virtual Machine's (JVM) Just-In-Time (JIT) translator [5] to
generate instructions to read the Pentium timestamp register [2]
and determine the amount of time spent in each method and the
number of times that each method was invoked. The overhead of the
additional instrumentation was relatively low, typically about 2%.
This high-resolution profiling enhanced Kaffe JVM was used to
profile a number of benchmark programs and several possible
performance improvement were identified for Kaffe's Java Virtual
Machine. The profiling tool showed approximately one quarter of
the methods were invoked only once during the execution of the
programs and that there was an overhead of 1.5 microseconds per
method invocation to check for the stack overflow. The profiling
tool identified class initializers, such as
/java/lang/System<clinit>, contributed over 300 milliseconds to
the runtime of each program. Additionally, the tool was used to
identify possible improvements to Pizza [3], a Java to bytecode
translator.
References
[1] Ping Huang, Selecting better-performing alternative code
using run-time profiling feedback,
http://www.ai.mit.edu/projects/reinventing_computing/pa-
pers/theses/pshuang/meng/main.html, Masters Thesis,
Massachusetts Institute of Technology, June 1995.
[2] Intel, Intel Architecture Optimization Manual, Mt.
Prospect, Illinois, order number 242816-003, 1997.
[3] Martin Odersky and Philip Wadler, ``Pizza into Java:
Translating theory into practice,'' Proc. 24th ACM
Symposium on Principles of Programming Languages, Paris,
France, January 1997.
[4] Sun, THE JAVA HOTSPOTTM PERFORMANCE ENGINE ARCHITECTURE,
http://www.javasoft.com/products/hotspot/whitepaper.htm-
l, April 1999.
[5] Transvirtual, Kaffe,
http://www.transvirtual.com/kaffe.html, May 1999.
>
>
> On Sat, 5 Jun 1999, Samarjeet S Tomar wrote:
>
> | Is there a Java profiler for kaffe ?
>
> I don't believe Kaffe supports the "-prof" command-line option yet.
>
> Alan
>
>
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
William Cohen [EMAIL PROTECTED]
Assistant Professor
Electrical and Computer Engineering
College of Engineering
University of Alabama in Huntsville voice: 256/890-6830
Huntsville, AL 35899 fax: 256/890-6803
ECE Homepage: http://www.eb.uah.edu/ece/ecehome.html
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
