> There is an interesting article about underperformance of pre-Pro x86
> processors on:
> http://www.intelligentfirm.com/membench/index.shtml
i suspect this is the same 'cache line access stall' thing observed &
posted by Robert Collins to comp.sys.intel a few months ago? If you look
at the Linux pentium FPU-memcpy patch, it's basically already optimized
for such an access type, although the correct order of memory accesses was
not found by analyzing the processor's (undocumented) behavior, it was
found via experimentation (according to the author). This patch has been
available for more than 2 years now AFAIR ...
-- mingo
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu