[x265] [PATCH 0/4] Optimise AArch64 quant/nquant primitives

[chen]

Hi Hari,




Thank for the patches, it looks good to me, I have not comment in these patches.




Regards,

Chen

At 2024-08-13 05:14:28, "Hari Limaye" <hari.limaye at arm.com> wrote:
>Hi,
>
>This series optimises the AArch64 Neon implementations of quant and
>nquant primitives.
>
>The series has no dependencies on any other on-list patch sets, except
>for patch:
>    [PATCH 1/4] AArch64: Remove SVE assembly implementation of quant
>which is based on refactoring changes to aarch64/asm-primitives.cpp made
>in the SAD patch series:
>    https://mailman.videolan.org/pipermail/x265-devel/2024-May/013707.html
>
>Relative performance observed compared to the existing Neon
>implementation:
>
>quant_neon:
>
>  Neoverse N1: 1.57x
>  Neoverse V1: 1.59x
>  Neoverse N2: 1.54x
>  Neoverse V2: 1.59x
>
>nquant_neon:
>
>  Neoverse N1: 1.79x
>  Neoverse V1: 1.77x
>  Neoverse N2: 1.70x
>  Neoverse V2: 1.73x
>
>Many thanks,
>
>Hari
>
>Hari Limaye (4):
>  AArch64: Remove SVE assembly implementation of quant
>  AArch64: Optimise quant_neon
>  Test: Update values used in check_nquant_primitive
>  AArch64: Optimise nquant_neon
>
> source/common/aarch64/asm-primitives.cpp |   3 -
> source/common/aarch64/fun-decls.h        |   2 -
> source/common/aarch64/pixel-util-sve.S   |  57 -------------
> source/common/aarch64/pixel-util.S       | 103 ++++++++++++-----------
> source/test/mbdstharness.cpp             |  10 ++-
> 5 files changed, 61 insertions(+), 114 deletions(-)
>
>-- 
>2.42.1
>
>_______________________________________________
>x265-devel mailing list
>x265-devel at videolan.org
>https://mailman.videolan.org/listinfo/x265-devel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/x265-devel/attachments/20240813/8c75ac1c/attachment.htm>