[Coco] Bechmarks

[Boisy G. Pitre]

On May 20, 2010, at 12:58 PM, Darren A wrote:

> On 5/20/10, Boisy G. Pitre wrote:
>> 
>> According to my calculations, the maximum throughput that your pak can
>> attain is 45.51 seconds per megabyte of data read.  It will be interesting
>> to see how close you can get to that number.
>> 
> 
> ---
> 
> 45.51 seconds per magabyte represents the maximum throughput for
> serial communication at 230,400 bps.  No driver for the MicroDrive
> module could achieve that due to the fact that the module will only
> transfer 512 bytes for each READ command.  You really need to include
> the command overhead when calculating the device's throughput
> capability.

Yes, there is some cost associated with sending commands to the device and awaiting the 512 bytes, then sending a similar command again and waiting for the next 512 bytes.  It does account for some percentage of time that each sector transaction takes.

> The datasheet also states that the device takes some time to respond
> to each command, but doesn't say how much.  I believe that will depend
> on the particular MicroSD card that is being used.

Agreed.  This will affect the timing as well.  Megaread probably won't be sensitive enough to pick up on the timing effects of those, however.

At this point, any benchmarking of the Drive Pak serves to test subsequent releases of the driver to see how performance can be improved.  Roger points out that the 254 second megaread doesn't accurately reflect the true speed of the device, and based on the theoretical throughput, I agree with him.

The height of the ceiling is known; now the goal is to get as close to it as possible.
--
Boisy G. Pitre
http://www.tee-boy.com/