>>>>> "Nathan" == Nathan Meyers <[EMAIL PROTECTED]> writes:
Nathan> Mo DeJong wrote:
>> On Mon, 18 Dec 2000, Joseph Shraibman wrote:
>>
>> > In at least a few instances that I know of, jikes produces
>> > slower code. But I don't think javac is a really great
>> > optimizing compiler either.
>>
>> That does not matter. Java code should be optimized by
>> the JIT not the Java compiler.
That's not entirely correct, there are some optimizations that are too
expensive to be done at runtime.
Nathan> I used to believe that until I tried, some months back, a
Nathan> side-by-side comparison of some benchmark code compiled by
Nathan> the old Blackdown JDK1.2.2 javac and a Sun-supplied javac
Nathan> on Solaris. On the same JVM, the code generated by the Sun
Nathan> compiler ran faster and utilized multi-threading well
Nathan> while the code generated by the Blackdown compiler kept
Nathan> just one CPU busy.
That's strange. Did you compare the generated code? I bet it was more
or less the same.
Nathan> I was told (by Sun) that the Blackdown JDK1.2.2 javac was
Nathan> based on an ancient Sun snapshot
Huh? Javac is written in Java: our 1.2.2-javac is the same as Sun's
javac in their 1.2.2.
1.3 comes with a new javac but that does not explain the differences
described above, javac is not a parallelizing compiler.
Nathan> which was the closest I ever got to an explanation of
Nathan> results that were, to say the least, wildly
Nathan> counterintuitive.
Juergen
--
Juergen Kreileder, Blackdown Java-Linux Team
http://www.blackdown.org/java-linux.html
JVM'01: http://www.usenix.org/events/jvm01/
----------------------------------------------------------------------
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
