Hyperthreading in P4
Many of your questions may have already been answered in earlier discussions or in the FAQ. The search results page will indicate current discussions as well as past list serves, articles, and papers.
Daniel Kidger Daniel.Kidger at quadrics.comMon Apr 29 10:54:35 PDT 2002
- Previous message: Processor contention(?) and network bandwidth on AMD
- Next message: Hyperthreading in P4
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Ole W. Saastad [mailto:ole at scali.no] wrote: >New Pentium 4 processors has hyper threading capabilities >and when setting this the linux sees 4 cpus on each dual node. >I have done some testing with OpenMP programs and found that >for OpenMP threaded programs there is no performance gain in using >the hypertheading. Using a number of threads that equal the number >of real processors seems to be optimal. Having a multi-threaded processor should help codes which are limited by memory *latency*. I doubt if memory-bandwidth limited codes would benefit much, since memory bandwidth is a limited resource which is already oversubscribed on many dual-P4 nodes. Also there are no more floating-point units that a standard P4, so CPU limited codes wont see any improvement either. Perhaps the interesting area though is where the CPU can issue instructions to the FPU *AND* to the integer execution units concurrently but for different threads. This would perhaps allow general Linux system services to not impact the performance on application codes? Yours, Daniel. -------------------------------------------------------------- Dr. Dan Kidger, Quadrics Ltd. daniel.kidger at quadrics.com One Bridewell St., Bristol, BS1 2AA, UK 0117 915 5505 ----------------------- www.quadrics.com --------------------
- Previous message: Processor contention(?) and network bandwidth on AMD
- Next message: Hyperthreading in P4
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Beowulf mailing list
