Falk Hueffner wrote:
> It depends on their amount. I'd estimate an unaligned trap takes about
> 100 cycles to handle. So if for example on my SX, I get 1000 unaligned
> traps per second, I loose about 0.1% of the performance. You can check
> the amount of unaligned traps by watching /proc/cpuinfo.
I have often seen these 'unaligned traps' lurking in my logs so I am curious
about being able too see how many occur in cpuinfo and learning something
new about my Alpha.
Looking at my cpuinfo I see....
kernel unaligned acc : 0 (pc=0,va=0)
user unaligned acc : 5407 (pc=120009bbc,va=1201c1abc)
platform string : Digital AlphaPC 164LX 533 MHz
Is "acc" short for access? I gather the unaligned trap is where the kernal
detects and traps a memory access. Whats the number of unaligned traps, the
zero number in kernal unaligned acc or the 5407? I presume that my cpuinfo is
telling me that whatever causes these unaligned traps is something that
runs from user space?
Mike
--
--------------------------------------------------------------------
Michael Lake
Active caver, Linux enthusiast and interested in anything technical.
Safety Convenor, Australian Speleological Federation
Owner, Speleonics (Australia)
--------------------------------------------------------------------
