>> The below one should be the patological case to show the performance hit
>> caused by the additional pte mapping and invlpg done in the page-fault
>> handler:
>
>Surely the pathalogical case is the above code but using CLONE_VM and one
>thread per CPU across an SMP box ?
You'll get an higer rate of page-faults/sec (fun ;), but not an higer kmap
latency. kmap is completly a per-CPU thing so it will SMP scale without
any problem (no SMP locking at all in it).
Doubling the page-fault rate theorically will halve the time for a bench
but then we'll add in the game the latency due the mmap semaphore... and
probably the kmap latency will be more biased without adding other
software latencies in the testcase.
Andrea
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/