Re: Do I want to debootstrap Debian armhf on my iMX53 board?
On Fri, Aug 19, 2011 at 09:41:50PM +0100, David Given wrote:
> On 19/08/11 21:13, Lennart Sorensen wrote:
> > /dev/sda:
> > Timing cached reads: 512 MB in 2.00 seconds = 255.59 MB/sec
> > Timing buffered disk reads: 134 MB in 3.01 seconds = 44.49 MB/sec
> > /dev/mmcblk0:
> > Timing cached reads: 490 MB in 2.00 seconds = 244.49 MB/sec
> > Timing buffered disk reads: 36 MB in 3.09 seconds = 11.63 MB/sec
> 44 MB/s is surprisingly slow for SATA; that's (anecdotally) about what
> the TrimSlice gets through its emasculated SATA-via-USB interface.
It's a 5400rpm laptop drive. Perfectly normal. It happened to be lying
around unused. I didn't expect it to be any faster. The fact it doesn't
use a ton of CPU cycles to do 44MB/s unlike the microSD makes a much
bigger difference. I will probably try with a real full size SATA
> For comparison, my SheevaPlug gets:
> Timing cached reads: 568 MB in 2.00 seconds = 283.81 MB/sec
> Timing buffered disk reads: 78 MB in 3.06 seconds = 25.50 MB/sec
> Timing cached reads: 564 MB in 2.00 seconds = 281.88 MB/sec
> Timing buffered disk reads: 56 MB in 3.02 seconds = 18.57 MB/sec
> /dev/sda is a 64MB Kingston SSD (feeling, I expect, extremely frustrated
> at being hooked up via USB). Interesting that the cache read figure is
> faster on the SheevaPlug, isn't it?
Yeah USB hurts. Wastes tons of CPU cycles.
I am not sure what the ram specs are on the sheeva versus the imx53.
> > MEMORY INDEX : 5.943
> > INTEGER INDEX : 5.664
> > FLOATING-POINT INDEX: 1.669
> > I guess that's quite a bit faster than your sheevaplug then. Especially
> > on floating point. No idea if the default gcc settings the makefile
> > used are even ideal.
> Memory's about twice as fast, integer about 20%. Not as much difference
> as I would have thought... I compiled mine with -O3.
They are both single core and about the same clock speed as far as
> Anyway, thanks very much --- interesting to see the figures.