[lwip-devel] bug entry #27352 code changes

[Leon Woestenberg]

Hello,

I'm doubtful whether the code changes introduced by bug entry #27352
are necessary.

There are no benchmarks, or even assembly dumps of the generated code,
a lot of assumptions untested and the resulting code change are
reasonably intrusive.

Short intro to that bug report:
----
Sun 30 Aug 2009 11:11:40 PM CEST, original submission:
This was brought up on lwip-users by Bill Auerbach.
It would both speed up and decrease memory footprint by not using a
packed struct where we don't need it:

typedef u32_t ip_addr;

instead of

PACK_STRUCT_BEGIN
struct ip_addr {
PACK_STRUCT_FIELD(u32_t addr);
} PACK_STRUCT_STRUCT;
PACK_STRUCT_END

Sun 30 Aug 2009 11:14:40 PM CEST, comment #1:

Forgot to add: the downsize is of course that every bit of code using
lwIP has to be changed. However, I'm not at all OK with using packing
where we absolutely don't need it!

Since this is rather a big API change, I'd love to hear comments from
active developers, please.
        Simon Goldschmidt <goldsimon>
Project MemberIn charge of this item.
----


Is it really true that the benefit of some 32-bit direct comparisons
(instead of two words, our four bytes) outperforms the extra copy of
the padded to the unpadded structures on NIOS-2?

How about other architectures?

In general I have seen a lot of code changes and bug reports that
related to performance, without any proof of improvement.

Regards,
-- 
Leon