I dont think the Nt box did spread the load well; 58% per cpu is a balanced 116% of one part...I bet if you looked at the kmode portion of the load (show kernel times in taskman), that 16% would be file IO, not java.
By default ant single threads, javac is also single threaded. nobody has added any explicit support for synchronization in their tasks so going MP would be dangerous. if you use <parallel> to run stuff in parallel, dont try and run multiple <javac> copies simultaneously, unless you set fork=true to run them in their own JVM ----- Original Message ----- From: "Kyle Adams" <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Sent: Wednesday, February 27, 2002 11:54 Subject: Re: AW: Win2K vs Sun More info: JDK 1.3 and Sun's javac were used on all machines. While jikes will speed up the build, we're currently using javac for any builds at the integration level or above (test, production environments). The chip on the Compaq machine is a Pentium III. Linux is not an option - the Unix group at our company has decided Linux is an unsupported OS. Question: If using the <parallel> task allows Ant to more effectively use MP, then why did the Win2K machine appear to do a much better job of distributing the load (without the <parallel> task)? As stated in my first post, both processors showed a 58% load. This is in contrast to the Sun E450 that had a 100% load on one processor while the other three had no load. Kyle -- To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]> For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
