![]() |
|
FreeBSD Installation and Upgrading Installing and upgrading FreeBSD. |
![]() |
|
Thread Tools | Display Modes |
|
|||
![]()
Any suggested CPU optimizations for FreeBSD 7 i386 system with dual Xeon E5320? Its a Clovertown core with 2x4MB of L2 Cache?
I have none right now. Using Google I found this: /etc/make.conf CPUTYPE?=nocona |
|
|||
![]() Quote:
# Old CPUs nocpu I486_CPU nocpu I586_CPU # Change the scheduler nooptions SCHED_4BSD options SCHED_ULE # I do not have AGP nodevice agp |
|
|||
![]()
Wait for 7.1 RELEASE or track 7 STABLE.
Building custom kernel for 7.0 RELEASE is of very little gain. |
|
|||
![]()
That is the idea.
|
|
||||
![]()
No need to curse. That's the recommended method for those that want to build kernels based on GENERIC, in such a way that any updates to GENERIC are also updates to their kernel.
Just include GENERIC in the config file. Any devices/options you don't want, you disable via nodevice/nooption. That way, when devices/options are added to GENERIC they are automatically added to your custom kernel as well. See the SMP config file for the include syntax. |
|
||||
![]()
Eh, stupid question...
'include GENERIC' as in the sense of actually puting that in the file, or # cat GENERIC CUSTOM > THEKERNCONF
__________________
My Journal Thou shalt check the array bounds of all strings (indeed, all arrays), for surely where thou typest ``foo'' someone someday shall type ``supercalifragilisticexpialidocious''. |
|
||||
![]()
The generic kernel is an SMP kernel now, but taking a quick grep of PAE, I see what you mean -> sorry, and thanks!
__________________
My Journal Thou shalt check the array bounds of all strings (indeed, all arrays), for surely where thou typest ``foo'' someone someday shall type ``supercalifragilisticexpialidocious''. |
|
|||
![]() Quote:
Only interesting info I found is about npx(4) and i586, and that is applicable for older machines. I use core2 CPU without i586, and performance is not decreased comparing with GENERIC. |
|
||||
![]()
There's a thread about this from 2005 that shows memory accesses to be faster on P3 CPUs when I586_CPU and I686_CPU are enabled in the kernel config, compared to just I686_CPU.
Doing a google search for "matt dillon I586_CPU" brings up more hits. It seems I585_CPU enables a few features that I686_CPU doesn't, even on CPUs newer than the Pentium. |
|
|||
![]()
That is obsolete information (there is no hard evidence), but I will explore this.
|
|
||||
![]()
Please post back or sort a thread if you turn anything (recent) up richard, thanks.
__________________
My Journal Thou shalt check the array bounds of all strings (indeed, all arrays), for surely where thou typest ``foo'' someone someday shall type ``supercalifragilisticexpialidocious''. |
|
||||
![]()
It might be "obsolete" in that it is old, but there hasn't been any mention of enabling bcopy and similar operations in the FreeBSD kernel when using I686_CPU. Granted, this may not make a difference on Athlon64/Opteron/Core/Core2 CPUs. I haven't heard/read anything about this subject since that thread, though. Guess someone with access to a few different CPUs will have to benchmark kernels with combinations of I586_CPU and I686_CPU.
|
|
|||
![]()
Here is short overview where I586_CPU is used in kernel:
http://fxr.watson.org/fxr/search?string=586_CPU I can't see anything that 586 have and that 686 lack. |
|
||||
![]()
This page shows all the different copy functions configured when I586_CPU is enabled. There are no entries for I686_CPU in there, which tells me that all the optimisations are enabled for I585_CPU and not for I686_CPU. Haven't looked through any of the other pages in that list just yet.
|
![]() |
Thread Tools | |
Display Modes | |
|
|
![]() |
||||
Thread | Thread Starter | Forum | Replies | Last Post |
optimizations in kernel and apps | chill | FreeBSD General | 5 | 1st October 2008 05:49 AM |
quick question about optimizations in /etc/make.conf | thevirtuesofxen | FreeBSD Installation and Upgrading | 7 | 15th July 2008 10:29 AM |