|
From: Leif M. <le...@ta...> - 2004-05-26 21:22:10
|
Jennifer Kolar wrote: > I ran my java application through optimize-it and looked at the % of > time the CPU spent in the Wrapper Manager - which is a static object > in my VM since I reference it to call restarts and also to enable and > disable monitoring pings. I also ran the WrapperSimpleApp w/ it > calling my app under optimize-it. > > I found that under 3.0.5 the % of time the CPU spent in wrapper code > was about 15% > under 3.1.0 (with ticktimer on) the % of time the CPU spent in wrapper > code was about 38% There is a thread in the WrapperManager that is cycling at a 100ms interval where each cycle is a "tick". That is a little tighter than the old thread which was simply pinging the JNI code for system signals. What is the rest of the application doing at this time. If it is idle then I am not surprised at the percentage jump. But the new thread should still be quite light weight. When running under reasonably high load. The % of CPU used by the thread in 3.0.5 and 3.1.0 should both be insignificant. Are you seeing results different from what I am expecting? > I then used purify to run the wrapper.exe to look for leaks on that > side of things, and I didn't find any particular problems.. Good to hear. Thank you. > Thanks for the memory footprint information and the expected CPU load > info. Cheers, Leif |