Re: x86, ARM, PARISC, PPC, MIPS and Sparc folks please run this

[Jamie Lokier]

Kars de Jong wrote:
> And no, this board has no way of getting a better time resolution than
> the 100 Hz tick timer either ;-)

The coherency test is fine.  That's just logic.

The clock granularity got me wondering whether the timing measurement
is meaningful on these machines.  It's possible for the shared test to
take 2000 microseconds and the unshared test to take 10 microseconds,
and they can still show up as 10ms if they both cross a clock tick
boundary.

The minimum of 128 tests of each type is likely to report 0 until
timing_loops is larger enough to make all 128 consistently almost
10ms, according to the timing when each test starts.  Then as we only
care if there is an approximately 2:1 ratio or more, it is fine.

That depends on the timing of each test not being synchronised with
the clock ticks, or when they are, that not affecting the result.

I'm not sure, but I have a feeling that the random shuffle makes it ok.

Hmm.

-- Jamie
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/