Pádraig Brady wrote:
> I'm a little wary of this being too aggressive,
> but it does give a noticeable (13%) boost on my new laptop.
> Here are the numbers from dd bs=$blksize if=/dev/zero of=/dev/null
>
> blksize system-1 system-2
> ----------------------------
> 1024 734 MB/s 1.7 GB/s
> 2048 1.3 GB/s 3.0 GB/s
> 4096 2.4 GB/s 5.1 GB/s
> 8192 3.5 GB/s 7.3 GB/s
> 16384 3.9 GB/s 9.4 GB/s
> 32768 5.2 GB/s 9.9 GB/s
> 65536 5.3 GB/s 11.2 GB/s
> 131072 5.5 GB/s 11.8 GB/s
> 262144 5.7 GB/s 11.6 GB/s
> 524288 5.7 GB/s 11.4 GB/s
> 1048576 5.8 GB/s 11.4 GB/s
I think it's time to apply your patch:
http://lists.gnu.org/archive/html/coreutils/2011-07/msg00059.html
Here are numbers from a 3.2GHz i7-970 with 1333MHz DDR3:
1024=2.6 GB/s
2048=4.4 GB/s
4096=6.5 GB/s
8192=8.5 GB/s
16384=10.1 GB/s
32768=11.1 GB/s
65536=12.0 GB/s
131072=12.3 GB/s
262144=12.5 GB/s
524288=12.5 GB/s
1048576=12.6 GB/s
One minor suggestion: change s/1/2/ in "-sINT 1" here.
+ timeout --foreground -sINT 1 \
+ dd bs=$bs if=/dev/zero of=/dev/null 2>&1 |
+ sed -n 's/.* \([0-9.]* [GM]B\/s\)/\1/p'
In this comment:
for i in $(seq 0 10); do
bs=$((1024*2**$i))
printf "%7s=" $bs
timeout --foreground -sINT 2 \
dd bs=$bs if=/dev/zero of=/dev/null 2>&1 |
sed -n 's/.* \([0-9.]* [GM]B\/s\)/\1/p'
done
That makes the numbers more reproducible.
Otherwise, for large block sizes, I'm guessing that the
interrupted transfer can make too large a difference.
