Re: [PATCH][3/7] perfctr-2.7.2 for 2.6.6-mm2: x86_64

[Andi Kleen]

> I want the kernel to avoid every delay that can be avoided.  Do not force

There are a lot of other more frutiful targets I would focus my 
attention then.  e.g. you could get rid of __SLOW_DOWN_IO on modern 
machines which hits all the time and is a *lot* slower than even hundreds 
of pipeline flushes. Or of all the mdelays in driver code like IDE.

> a pipeline flush for speculative rdtsc.  Besides those 20-30 cycles

But how useful is the "work" when it results in wrong answers? 

Speculative RDTSC is not a theoretical problem, it has actually
caused problems in practice in other codes by making CPU timestamps
non monotonic.

-Andi
-
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/