[PATCH] net/crc: make NEON CRC folding logic similar to x86 SSE

[Thomas Monjalon]

06/10/2026 17:18, Shreesh Adiga:
> This patch includes the following minor changes:
> 1) Use TBL based shift left for variable len shift instead of using
>    vshift_bytes_left which adds a switch case for 16 len cases.
> 2) Align the fold constants to use PMULL and PMULL2 instructions
>    instead of using cross lane operands which requires additional shift.
> 3) Remove unnecessary static arrays used in crc32_reduce_64_to_32 by
>    adapting x86 SSE logic.
> 4) Use x86 SSE logic for processing partial bytes sequence to avoid
>    using multiple variable len shifts which inserts switch case.
> 
> Signed-off-by: Shreesh Adiga <16567adigashreesh at gmail.com>

Applied, thanks.