]> granicus.if.org Git - libvpx/commit
reuse common loopfilter code
authorJohann <johannkoenig@google.com>
Mon, 18 Oct 2010 18:57:40 +0000 (14:57 -0400)
committerJohann <johannkoenig@google.com>
Mon, 25 Oct 2010 13:48:50 +0000 (09:48 -0400)
commit1376f061dae199af3378339367a1bc9f95fd4187
tree1485107af7910cf72e9e80cbdc8e1529fb0e5414
parentb71962fdc98c3fb9fdf95d74191452cd51cbc2b5
reuse common loopfilter code

there were four versions for the regular and
macroblock loopfilters:
horizontal [y|uv]
vertical [y|uv]

this moves all the common code into 2 functions:
vp8_loop_filter_neon
vp8_mbloop_filter_neon

this provides no gain in performance. there's a bit
of jitter, but it trends down ~0.25-0.5%. however,
this is a huge gain maintenance. also, there is the
potential to drop some stack usage in the macroblock
loopfilter.

Change-Id: I91506f07d2f449631ff67ad6f1b3f3be63b81a92
vp8/common/arm/neon/loopfilter_neon.asm [new file with mode: 0644]
vp8/common/arm/neon/loopfilterhorizontaledge_uv_neon.asm [deleted file]
vp8/common/arm/neon/loopfilterhorizontaledge_y_neon.asm [deleted file]
vp8/common/arm/neon/loopfilterverticaledge_uv_neon.asm [deleted file]
vp8/common/arm/neon/loopfilterverticaledge_y_neon.asm [deleted file]
vp8/common/arm/neon/mbloopfilter_neon.asm [new file with mode: 0644]
vp8/common/arm/neon/mbloopfilterhorizontaledge_uv_neon.asm [deleted file]
vp8/common/arm/neon/mbloopfilterhorizontaledge_y_neon.asm [deleted file]
vp8/common/arm/neon/mbloopfilterverticaledge_uv_neon.asm [deleted file]
vp8/common/arm/neon/mbloopfilterverticaledge_y_neon.asm [deleted file]
vp8/vp8_common.mk