On Aug 18, 23:34, Brian Buhrow wrote:
}
} Hello. Are there pros to having the kernel do the CPU
} switching? You don't seem to list } any, but there must be
} trade-offs. What are they?
It just depends on where you want the complexity. If userland
does it, then you use cpusets and schedular affinity to jump to a
particular cpu. This is how "cpuctl identify <n>" does it. For
my testing, I have done it this way. It makes the kernel side
really simple. If done in the kernel, then a xcall will be needed
which complicates the kernel side considerably, but makes the
userland side simple as you just specify which cpu you want.
}-- End of excerpt from Brian Buhrow
