irqbalance question on openSuSE 10.1
Hi, From the Description of irqbalance, it is supposed to balance IRQ activity between two or more processors. On /etc/init.d/irq_balancer LINE 41 there is the following line if [ $PHYS -gt 1 -o \( $PROC -gt 1 -a $PHYS -eq 0 \) ] ; then witch explicitly states that the $PHYS variable has to be greater then one. This means that there has to be two physical processors, and the daemon will not load if the computer has either a Hyper-Threading or Dual-Core CPU. I'm sure that SuSE folks dis this on purpose but, is it possible to someone to tell me why irqbalance will not work on this CPU's, specially on a Dual-Core System ? I've done a goggling on the subject and find nothing substantial on this subject. My system is a Dual-Core Intel(R) Pentium(R) D CPU 3.20GHz Thanks for your time -- Rui Santos http://www.ruisantos.com/ They that would give up essential liberty for temporary safety deserve neither one of them.
Rui Santos wrote:
On /etc/init.d/irq_balancer LINE 41 there is the following line
if [ $PHYS -gt 1 -o \( $PROC -gt 1 -a $PHYS -eq 0 \) ] ; then
witch explicitly states that the $PHYS variable has to be greater then one.
Actually, it says "PHYS>0" _OR_ " "PHYS==0 and PROC>1". The latter is the case on my 4-way box, so irqbalance is active.
This means that there has to be two physical processors, and the daemon will not load if the computer has either a Hyper-Threading or Dual-Core CPU.
I don't know what /proc/cpuinfo looks like for a dual-core, but a dual-core machine certainly has two physical processors, even they're not on different carriers. Hyperhreading is different - only one processor that pretends to do more things at the same time :-) Post your /proc/cpuinfo. /Per Jessen, Zürich
Per, On Wednesday 31 May 2006 07:16, Per Jessen wrote:
...
Hyperhreading is different - only one processor that pretends to do more things at the same time :-)
The difference is more a matter of degree. In an HT CPU there is less duplicated internal machinery. A dual-core CPU has two full CPUs that operate largely independently. In the current dual-core chips, I believe, there is a single, shared level-2 cache. The HT CPUs don't double all of the CPU's hardware, but they still hold two full sets of the CPU's internal state and a good-sized chunk of the processing machinery, making possible both (some) concurrent processing and faster context switching between the processes or threads that occupy those register sets than on a conventional single CPU. It seems HT CPUs were just a transitional state between older single CPU chips and modern multi-core chips.
...
/Per Jessen, Zürich
Randall Schulz
Randall R Schulz wrote:
On Wednesday 31 May 2006 07:16, Per Jessen wrote:
Hyperhreading is different - only one processor that pretends to do more things at the same time :-)
The difference is more a matter of degree. In an HT CPU there is less duplicated internal machinery. A dual-core CPU has two full CPUs that operate largely independently. In the current dual-core chips, I believe, there is a single, shared level-2 cache.
I've been thinking of getting a dual-core Pentium D for a new workstation, and the specs I've seen all say 2 x 1Mb L2 cache. I believ the two CPUs operate fully independently.
It seems HT CPUs were just a transitional state between older single CPU chips and modern multi-core chips.
Hmm, that's stretching it a bit I think. The P4 Hyperthreading is AFAIK an improved version of the Xeon <something>threading. The HT bit still resides on the same die, whereas the dual-core has two dies. I tend to think of a dual-core as simply two processors, nothing more, nothing less. But a lot cheaper to buy and to run! /Per Jessen, Zürich
Per Jessen wrote:
It seems HT CPUs were just a transitional state between older single CPU chips and modern multi-core chips.
Hmm, that's stretching it a bit I think.
Actually, Intels marketing says exactly what you're saying ... but at least it confirms what I said about the 2 x 1Mb L2-cache :-) /Per Jessen, Zürich
Per Jessen wrote:
I tend to think of a dual-core as simply two processors, nothing more, nothing less. But a lot cheaper to buy and to run!
Interesting article: http://www.pugetsystems.com/articles.php?id=23 /Per Jessen, Zürich
participants (3)
-
Per Jessen -
Randall R Schulz -
Rui Santos