Mailing list::: Increasing kern.hz rate

I'm installing KDE4 from ports on a VM (VirtualBox) FreeBSD 9.1 and decided to increase kern.hz to 1000.

Motherboard: Intel Desktop Board D525MW (all-in-one) CPU Atom D525
HD: Seagate Momentus Thin 320 GB SATA 3
RAM: 2 x Kingstone 2 GB DDR3 1333 MHz SO-DIMM

All seems to work fine but it's not much faster than before with kern.hz=100.
 
Performance is different on PowerPC/POWER architectures than on the standard i386/AMD64 that is used. The kernel Hertz rate makes a large difference. Both OpenBSD and Debian would choke when compiling programs; yet, that did not happen with FreeBSD. (This is in reference to PPC.)
 
kern.hz=100 over here, too. And works like a charm ;)

On my laptop I once had audio latencies too which I "resolved" by reverting it to kern.hz=1000.
 
xibo said:
kern.hz=100 over here, too. And works like a charm ;)

On my laptop I once had audio latencies too which I "resolved" by reverting it to kern.hz=1000.
Click to expand...

The above is an example of why kern.hz=X cannot and does not work the same because of the hardware and hardware usage. Being that the machines will be in use for audio, a higher hertz rate will not hurt.

Not enough RISC usage in FreeBSD;I am SOL.
 
This year FreeBSD 10-HEAD got a new reincarnation of the kernel event timers backend callout(9), no more limited by or even related to Hz rate. It is used to implement most of the user-level primitives, such as usleep(), poll(), select(), etc. Not all primitives are covered yet, but work should progress further. For modern x86 hardware it is realistic now to get event delays as low as 1.5 microseconds, even in user-level.
 
Sweet deal. The CISC runs with 32 bit environment on a 64 bit processor- I needed to copy the exact same environment due to the other laptop being i386. Is this- kernel event timers- in RECENT?

I should be with a RISC setting soon- PowerPC 32&64- and has it been tested on those architectures?
 
You must log in or register to reply here.
Back
Top