[PATCH v2] arm64: Implement optimised IP checksum helpers

[Robin Murphy]

On 31/05/16 15:39, Robin Murphy wrote:
> Hi James,
>
> On 31/05/16 15:24, James Morse wrote:
>> Hi Robin,
>>
>> On 31/05/16 12:22, Robin Murphy wrote:
>>> AArch64 is capable of 128-bit memory accesses without alignment
>>> restrictions, which makes it both possible and highly practical to slurp
>>> up a typical 20-byte IP header in just 2 loads. Implement our own
>>> version of ip_fast_checksum() to take advantage of that, resulting in
>>> considerably fewer instructions and memory accesses than the generic
>>> version. We can also get more optimal code generation for csum_fold() by
>>> defining it a slightly different way round from the generic version, so
>>> throw that into the mix too.
>>>
>>> Suggested-by: Luke Starrett <luke.starrett at broadcom.com>
>>> Acked-by: Luke Starrett <luke.starrett at broadcom.com>
>>> Signed-off-by: Robin Murphy <robin.murphy at arm.com>
>>> ---
>>>
>>> Minor changes: include types.h for correctness, add Luke's ack.
>>>
>>>   arch/arm64/include/asm/checksum.h | 51
>>> +++++++++++++++++++++++++++++++++++++++
>>
>> Maybe a nit, don't you need to remove the 'generic-y += checksum.h'
>> line from
>> arch/arm64/include/asm/Kbuild to avoid the generated version being
>> created too? [0]
>
> I guess you skimmed the very end of the patch ;)
>
> +#include <asm-generic/checksum.h>

Oh wait, I'm being a massive idiot - include/asm-generic is not the same 
thing as arch/arm/include/generated/asm, is it? I see what you mean now, 
and yes it probably does want fixing...

Sorry!
Robin.

>> The compiler on my box picks your header in preference to the
>> generated one, but
>> [1] suggests it isn't to be trusted!
>
> We only cherry-pick a couple of functions to optimise, so not building
> the generic file would necessitate copy-pasting the rest of it. I'd hope
> this pattern is sufficiently robust, since it's already in use on a
> number of other architectures! AFAICS [1] only applies when the generic
> file shouldn't be built at all, so cross fingers we're probably OK.
>
> Robin.
>
>> Thanks,
>>
>> James
>>
>> [0] d8ecc5cd8e22 ("kbuild: asm-generic support")
>> [1] https://lkml.org/lkml/2016/5/23/78
>>
>> _______________________________________________
>> linux-arm-kernel mailing list
>> linux-arm-kernel at lists.infradead.org
>> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
>>
>
>
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel