Ok. Here are screen dumps of my JMP test of FOP.

Scenario:

Three PDF's are produced, a letter, application form and an
envelope. For a mail-out for a non-profit organization. (My other
choice was Access 2000).

There are 400, 800 and 400 page images produced from something
that used to run in Cocoon, now wrapped in a variation of
SimpleJaxp.java from the O'Reilly Java and XSLT book (Eric Burke ?)
to run on the client's computer.

Each PDF has the same structure: two XSLT's (i) simple set-up
merging content with database data (400 members) (ii) XSL-FO run
which feeds FOP through javax.xml.transform.SAXResult().

I won't share the data right now but the program normally runs for
about five minutes on an Athlon 800, RH 9 and Java 1.4.1. I thought
this was too slow even if my XSLT code is less than divinely
inspired.

I built and installed Java Memory Profiler from:
http://www.khelekore.org/jmp/performance.html and started a test.

This runs slowly but it appears to identify some code that bears
investigation. Here are some screen shots of my results:

The last file list_jmp3.jpg was taken just before the program
terminated.

Obviously, PropertyList->HashMap lookups are taking a LOT of time.

The PropertyList.java class is a lot more complex than a simple hash
lookup and I suspect it might be profitable to take a look. Right
out of the starting blocks, the constructor makes no reference
to Initial Capacity or Load Factor. I am confused by the code
and the seeming hurry to look everything up in the parent property
lists.

-- 
John Austin <[EMAIL PROTECTED]>

<<attachment: list_jmp1.jpg>>

<<attachment: list_jmp2.jpg>>

<<attachment: list_jmp3.jpg>>

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to