diff options
| author | Linfeng Zhang <linfengz@google.com> | 2017-10-11 11:59:04 -0700 |
|---|---|---|
| committer | Linfeng Zhang <linfengz@google.com> | 2017-10-16 15:42:42 -0700 |
| commit | 580d32240f9a2f7783ced9b0e409de62ba07f4e1 (patch) | |
| tree | 291af741b11ba297faeaba257751922f49422314 /vp9/encoder/vp9_frame_scale.c | |
| parent | 1fa3ec3023ef1b841d993bc21cba32b658a790ea (diff) | |
| download | libvpx-580d32240f9a2f7783ced9b0e409de62ba07f4e1.tar libvpx-580d32240f9a2f7783ced9b0e409de62ba07f4e1.tar.gz libvpx-580d32240f9a2f7783ced9b0e409de62ba07f4e1.tar.bz2 libvpx-580d32240f9a2f7783ced9b0e409de62ba07f4e1.zip | |
Add 4 to 3 scaling SSSE3 optimization
Note this change will trigger the different C version on SSSE3 and
generate different scaled output.
Its speed is 2x compared with the version calling vpx_scaled_2d_ssse3().
Change-Id: I17fff122cd0a5ac8aa451d84daa606582da8e194
Diffstat (limited to 'vp9/encoder/vp9_frame_scale.c')
| -rw-r--r-- | vp9/encoder/vp9_frame_scale.c | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/vp9/encoder/vp9_frame_scale.c b/vp9/encoder/vp9_frame_scale.c index 832df18c8..a410d0407 100644 --- a/vp9/encoder/vp9_frame_scale.c +++ b/vp9/encoder/vp9_frame_scale.c @@ -28,7 +28,7 @@ void vp9_scale_and_extend_frame_c(const YV12_BUFFER_CONFIG *src, const InterpKernel *const kernel = vp9_filter_kernels[filter_type]; int x, y, i; -#if HAVE_NEON +#if HAVE_SSSE3 || HAVE_NEON // TODO(linfengz): The 4:3 specialized C code is disabled by default since // it's much slower than the general version which calls vpx_scaled_2d() even // if vpx_scaled_2d() is not optimized. It will only be enabled as a reference |