projects / platform / upstream / libvpx.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 5595e18) | patch
Optimize Neon high bitdepth subpel variance functions
authorSalome Thirot <salome.thirot@arm.com>
Tue, 7 Feb 2023 14:08:33 +0000 (14:08 +0000)
committerSalome Thirot <salome.thirot@arm.com>
Mon, 13 Feb 2023 20:19:30 +0000 (20:19 +0000)
commit42cb3dbf94706dd8477f9313ceba1fc3a9a14e92
tree50755a830b6f39317b0afa2b1cd436167f05ca83tree | snapshot
parent5595e18870b24fdfd8fdeb4cfa5ea51824486273commit | diff
Optimize Neon high bitdepth subpel variance functions

Use the same general code style as in the standard bitdepth Neon
implementation. Additionally, do not unnecessarily widen to 32-bit data
types when doing bilinear filtering - allowing us to process twice as
many elements per instruction.

Change-Id: I1e178991d2aa71f5f77a376e145d19257481e90f
vpx_dsp/arm/highbd_subpel_variance_neon.c [new file with mode: 0644] blob
vpx_dsp/arm/highbd_variance_neon.c diff | blob | history
vpx_dsp/arm/mem_neon.h diff | blob | history
vpx_dsp/vpx_dsp.mk diff | blob | history
Domain: Multimedia / Codec;