> To help me decide whether to include /dev/random in my kernel I wanted to
> measure how it affected disk access speed. To do it I ran hdparm -t on my
> IDE drive with a freshly booted computer that was otherwise idle.
> The speed increase obtained by removing /dev/random from the kernel was
> too low to distinguish from measurement error but when I performed the
> same experiment running 1.2.13 I got a 5% speed improvement.
thats probably due to the buffer cache code change. (using a unified
buffer cache for most data).
the second value in "hdparm -t" shows the "raw buffering throughput". This
is now a bit slower. This change shows up in the "summary value" too.
-- mingo