[En-Nut-Discussion] IP Checksum calculation broken. Reenabled old algorithm for the meantime.

[Ulrich Prinz]

Am 18.06.2012 17:37, schrieb Harald Kipp:
> Hi Nathan,
>
> On 18.06.2012 15:45, Nathan Moore wrote:
>
>> Years ago I was trying to squeeze performance out of Nut/OS on an
>> Atmega and fiddled with a gcc inline assembly checksum calculation
>> for it.  I'm sure that all targets could benefit from their own
>> checksumming routines for both performance and correctness (endian
>> and alignment issues).
>
> I fully agree and it's on the list. However, I wanted to make a
> generic routine available first to simplify porting to new
> platforms.
>

I think that is the most important point:

If a platform supports HW-CRC we should use it. If not, we should used
architecture optimized calculations. I.e. you don't use byte-copy
routines on a 32-bit system cause it is generic.

Best regards
Ulrich