The performance tests I made were extremely rough. The main purpose was to
find out an efficient pooling implementation. Actually, the benchmark code
contained a major fault in direct instantation. References to instances were
not stored making it possible for the JVM to reuse the same memory block. So
the time for direct instantiation measures in practice pure looping 100 000
times. Time required for "real" instantiation seems to vary a lot depending
on the state of the memory and typically exceeding the time required for
taking a pooled object into use. The main goal of pooling is to get rid of
these variations and guarantee a predictable access time for critical
objects.
The tests were run under JBuilder 4.0/Windows NT using Sun's JVM 1.3.0-C in
mixed mode. I repeat that the absolute values are useless and even the
relative ones approximate.
The updates are not yet (2/16/2001) in CVS but should be within a few days.
-- Ilkka
--
Nokia Networks
http://www.nokia.com/networks
mail: [EMAIL PROTECTED]
Subject: Re: Performance of PoolService
From: "Santiago Gala" <[EMAIL PROTECTED]>
Date: Fri, 16 Feb 2001 13:30:41 +0100
[EMAIL PROTECTED] wrote:
> Below are some performance comparisons of the new pool service. The =
tests
> have been executed for 100 000 empty objects in a 300 MHz P2, so the
A very interesting work. Thanks for your tests. I wanted to ask a few =
questions:
Can you tell which Virtual Machine have you used?
Also, did you allowed for "heating", if the machine is Hotspot Server =
specially?
Third, is the optimized code of the last benchmark already in CVS?
------------------------------------------------------------
To subscribe: [EMAIL PROTECTED]
To unsubscribe: [EMAIL PROTECTED]
Search: <http://www.mail-archive.com/turbine%40list.working-dogs.com/>
Problems?: [EMAIL PROTECTED]
