Here is an IBM document that answers many of the questions raised in this thread.
Match 31-bit WebSphere Application Server performance with new features in 64-bit Java on System z http://www-03.ibm.com/servers/enable/site/education/wp/1d71a/1d71a.pdf Doug On Mon, 4 Jan 2010 12:00:25 -0500, Scott Rowe <[email protected]> wrote: >IIRC, this is the type of thing they are doing, and as Jim mentioned: it improves performance because more pointers fit in a cache line, >and therefor use less cache. With today's processor designs, using cache effectively can have huge performance benefits. > > >>>> Tom Marchant <[email protected]> 1/4/2010 10:45 AM >>> >>On Sat, 2 Jan 2010 15:37:15 -0500, Jim Mulder <[email protected]> wrote: >>> >>> I was told that compressed pointers for storage below 32GB fit >>>into a smaller space, so more compressed pointers fit in a cache >>>line, leading to more effective cache utilization. Performance is >>>all about the caches these days. I am not a Java person. I don't >>>know what a compressed pointer is. >>> > >Just a wild guess. If all pointers are to storage on a doubleword boundary, >the address can be shifted right three bits. Then you can point to any >doubleword below 32 GB using an unsigned 32-bit address. How that might >help performance is a mystery to me. > >-- >Tom Marchant > ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
