hello Rickard,
I've implemented your suggestion in JBoss and found strange results (under
W2K, both 1.3 and 1.4 JVM with hotspot enabled).
The difference is frequently not significant at all: sometimes the
optimized version is faster, sometimes not (and by running tens of
thousands of serialisations, after a hotspot-warm-up)
In fact, when using the toByteArray'ed version, System.arraycopy is used
internally: this is the only additional step that differentiates both
versions. And if you take a look at it, as all elements are of the same type
(byte), the arraycopy is implemented as a C memmove. Maybe, internally,
the memmove is handled by the OS memory manager as a simple duplicated view
of the same memory page that is only *really* copied if/once a modification
occurs on this page (copy-on-write). In this case, as no modification is
performed, no copy could be done. Thus, the cost may not be that big (except
for the native call to System.arraycopy). But it may change depending on the
OS/JVM.
It is just an attempt to understand, not a strong in-depth analysis.
Any other opinion?
Cheers,
Sacha
-Message d'origine-
De : [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED]]De la part de
Rickard Öberg
Envoyé : mardi, 13 août 2002 15:01
À : [EMAIL PROTECTED]
Objet : [JBoss-dev] Optimizing MarshalledObject, or variants thereof
Hey
I just found out a way to optimize the use of MarshalledObject, if it's
used to send stuff over the network (JBoss has it's own version, but it
has the same basic code).
The MO serializes the data into a ByteArrayOutputStream, which then is
toByteArray'ed in order to get the data. This is bad, since the data is
then copied into a new array of the same length. This may not mean much,
but if you're doing it all the time it makes a difference.
There's a trick to get around this: simply store the BAOS in an instance
variable of the MO, and call out.write(BAOS.size());BAOS.writeTo(out);
during serialization. This will avoid the copy. Read the data during
deserialization, and the rest is the same.
This simple trick will hence reduce the number of byte arrays created by
50%. You'll probably want to introduce this fix into MarshalledValue,
and any other place where MO's are used.
/Rickard
--
Rickard Öberg
[EMAIL PROTECTED]
Senselogic
Got blog? I do. http://dreambean.com
---
This sf.net email is sponsored by: Dice - The leading online job board
for high-tech professionals. Search and apply for tech jobs today!
http://seeker.dice.com/seeker.epl?rel_code1
___
Jboss-development mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/jboss-development
---
This sf.net email is sponsored by: OSDN - Tired of that same old
cell phone? Get a new here for FREE!
https://www.inphonic.com/r.asp?r=sourceforge1refcode1=vs3390
___
Jboss-development mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/jboss-development
