Please tell me more about this.
>SMP=1 is upto 10% slower for single CPU
>on real tasks and 40% slower on certain FPU/sleep/FPU/sleep patterns
Is this situation permanent or just a matter of some tuning
needing to be done (e.g, the granularity of some locks needing to be
adjusted)? I would hope, for example, that spin locks to do not
actually "spin" if there is only one CPU.
Also, there is a lot of code in the kernel, especially in
the filesystems, that was written in a roundabout way to work
with an environment where there was no locking but in which
a kernel process had the CPU until it would do something that
blocked. So, this code does things like starting to allocate a
block in a file and then rechecking to make sure that that block
has not been allocated since the CPU blocked. This makes me wonder
if some speedups would be possible when SMP=1.
So, I wonder if the overhead of SMP=1 is really that great
once SMP=1 is as well tuned as non-SMP.
Adam J. Richter __ ______________ 4880 Stevens Creek Blvd, Suite 205
adam@yggdrasil.com \ / San Jose, California 95129-1034
+1 408 261-6630 | g g d r a s i l United States of America
fax +1 408 261-6631 "Free Software For The Rest Of Us."
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu