You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@tomcat.apache.org by bu...@apache.org on 2002/05/29 17:30:21 UTC
DO NOT REPLY [Bug 6858] -
Easy & Significant Performance Improvement
DO NOT REPLY TO THIS EMAIL, BUT PLEASE POST YOUR BUG
RELATED COMMENTS THROUGH THE WEB INTERFACE AVAILABLE AT
<http://nagoya.apache.org/bugzilla/show_bug.cgi?id=6858>.
ANY REPLY MADE TO THIS MESSAGE WILL NOT BE COLLECTED AND
INSERTED IN THE BUG DATABASE.
http://nagoya.apache.org/bugzilla/show_bug.cgi?id=6858
Easy & Significant Performance Improvement
------- Additional Comments From osvaldo@visionnaire.com.br 2002-05-29 15:30 -------
Just some additional data.
* Allocating big arrays w/o necessity may hurt some GC engines. The allocator
may have a policy of skipping the young generation for objects bigger than
some maximum limit, so these objects would be allocated directly in the
old-generation and totally screw with thee benefits of generational GC;
the result is that you need full-GC's instead of generational-GCs for
deleting these buffers, which could explain big and frequent GC pauses.
Other VMs may segregate objects by size, having many small heaps for different
ranges of size, so the freespace control is faster as each heap is an array
and a single bit can track each slot's position; IBM is one JVM that's known
to have this architecture -- and very large objects will spoil the trick again
because the segregated heaps exist only for a small number of size ranges.
* A default policy of 512 bytes looks good, this should be good enough for the
output of most tags. Some tags produce a lot of output and it would be
important to set a bigger buffer for making the I/O more efficient, but this
is less severe than the memory management problem. For pages we can set the
buffer size, unfortunately not for tags so this may be a good RFE for the
JSP spec? Meanwhile, the lesser evil is using a small buffer.
--
To unsubscribe, e-mail: <ma...@jakarta.apache.org>
For additional commands, e-mail: <ma...@jakarta.apache.org>