RE: [lwip-devel] UDP - transmission inefficiences

[Bill Auerbach]

>> It does highlight the problem of trying to optimise lwIP in this way
>> though, as improvements for one platform can unintentionally hurt
>> another.
>>
>> Ideally inlining would be left up to the compiler as it should be in a
>> good position to know when it would help and when it would not, but
>not
>> all compilers are sophisticated enough to do this well.
>
>Maybe the name Bill gave its patch is a bit misleading: it does not
>inline a function call but calculates the checksum when IP header bytes
>are saved. No compiler can do this on its own :-)
>
>Of course this is architecture dependent: architectures with a big
>register file will be able to do the calculation in registers - they
>will gain from the patch. While architectures with a small register file
>will need many small memory accesses (even if it's only the cache which
>is accessed), which will slow them down.

This is true, but highly unlikely in this case for this change.

>Anyway, Bill's patch should be an option which is turned off by default.
>If so, I think it's worth including it.

I hate to agree but do with a report that it's slower for one application,
as dubious as I am about that finding.

Bill