diff options
| author | Martin Storsjö <martin@martin.st> | 2019-02-01 10:04:56 +0200 |
|---|---|---|
| committer | Martin Storsjö <martin@martin.st> | 2019-02-19 11:46:04 +0200 |
| commit | 52c9b0a6c0d02cff6caebcf6989e565e05b55200 (patch) | |
| tree | 64ea5d55b27e4483e3c45b21c83b66a7977b6617 /libavcodec/aarch64/vp8dsp_init_aarch64.c | |
| parent | c513fcd7d235aa4cef45a6c3125bd4dcc03bf276 (diff) | |
| download | ffmpeg-52c9b0a6c0d02cff6caebcf6989e565e05b55200.tar.gz | |
aarch64: vp8: Port vp8_luma_dc_wht and vp8_idct_dc_add4uv from arm version
Cortex A53 A72 A73
vp8_luma_dc_wht_c: 115.7 75.7 90.7
vp8_luma_dc_wht_neon: 60.7 41.2 45.7
vp8_idct_dc_add4uv_c: 376.1 262.9 282.5
vp8_idct_dc_add4uv_neon: 52.0 29.0 37.0
Signed-off-by: Martin Storsjö <martin@martin.st>
Diffstat (limited to 'libavcodec/aarch64/vp8dsp_init_aarch64.c')
| -rw-r--r-- | libavcodec/aarch64/vp8dsp_init_aarch64.c | 3 |
1 files changed, 3 insertions, 0 deletions
diff --git a/libavcodec/aarch64/vp8dsp_init_aarch64.c b/libavcodec/aarch64/vp8dsp_init_aarch64.c index 20bb607dfd..e296ccd893 100644 --- a/libavcodec/aarch64/vp8dsp_init_aarch64.c +++ b/libavcodec/aarch64/vp8dsp_init_aarch64.c @@ -28,6 +28,7 @@ void ff_vp8_luma_dc_wht_neon(int16_t block[4][4][16], int16_t dc[16]); void ff_vp8_idct_add_neon(uint8_t *dst, int16_t block[16], ptrdiff_t stride); void ff_vp8_idct_dc_add_neon(uint8_t *dst, int16_t block[16], ptrdiff_t stride); void ff_vp8_idct_dc_add4y_neon(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride); +void ff_vp8_idct_dc_add4uv_neon(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride); VP8_LF(neon); @@ -55,10 +56,12 @@ av_cold void ff_vp8dsp_init_aarch64(VP8DSPContext *dsp) { if (!have_neon(av_get_cpu_flags())) return; + dsp->vp8_luma_dc_wht = ff_vp8_luma_dc_wht_neon; dsp->vp8_idct_add = ff_vp8_idct_add_neon; dsp->vp8_idct_dc_add = ff_vp8_idct_dc_add_neon; dsp->vp8_idct_dc_add4y = ff_vp8_idct_dc_add4y_neon; + dsp->vp8_idct_dc_add4uv = ff_vp8_idct_dc_add4uv_neon; dsp->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16_neon; dsp->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16_neon; |