diff options
Diffstat (limited to 'third_party/aom/aom_dsp/arm')
| -rw-r--r-- | third_party/aom/aom_dsp/arm/blend_a64_mask_neon.c | 6 | ||||
| -rw-r--r-- | third_party/aom/aom_dsp/arm/subpel_variance_neon.c | 22 |
2 files changed, 13 insertions, 15 deletions
diff --git a/third_party/aom/aom_dsp/arm/blend_a64_mask_neon.c b/third_party/aom/aom_dsp/arm/blend_a64_mask_neon.c index 82c0b0e28..e7f08a5fd 100644 --- a/third_party/aom/aom_dsp/arm/blend_a64_mask_neon.c +++ b/third_party/aom/aom_dsp/arm/blend_a64_mask_neon.c @@ -86,7 +86,8 @@ static INLINE void blend_4x4(uint8_t *dst, uint32_t dst_stride, const int16x8_t vec_round_bits) { int16x8_t src0_0, src0_1; int16x8_t src1_0, src1_1; - uint64x2_t tu0, tu1, tu2, tu3; + uint64x2_t tu0 = vdupq_n_u64(0), tu1 = vdupq_n_u64(0), tu2 = vdupq_n_u64(0), + tu3 = vdupq_n_u64(0); int16x8_t mask0_1, mask2_3; int16x8_t res0, res1; @@ -154,7 +155,8 @@ void aom_lowbd_blend_a64_d16_mask_neon( assert(IS_POWER_OF_TWO(w)); uint8x8_t s0, s1, s2, s3; - uint32x2_t tu0, tu1, tu2, tu3; + uint32x2_t tu0 = vdup_n_u32(0), tu1 = vdup_n_u32(0), tu2 = vdup_n_u32(0), + tu3 = vdup_n_u32(0); uint8x16_t t0, t1, t2, t3, t4, t5, t6, t7; int16x8_t mask0, mask1, mask2, mask3; int16x8_t mask4, mask5, mask6, mask7; diff --git a/third_party/aom/aom_dsp/arm/subpel_variance_neon.c b/third_party/aom/aom_dsp/arm/subpel_variance_neon.c index 44d821821..cf618eee7 100644 --- a/third_party/aom/aom_dsp/arm/subpel_variance_neon.c +++ b/third_party/aom/aom_dsp/arm/subpel_variance_neon.c @@ -17,13 +17,9 @@ #include "aom_ports/mem.h" #include "aom/aom_integer.h" +#include "aom_dsp/aom_filter.h" #include "aom_dsp/variance.h" -static const uint8_t bilinear_filters[8][2] = { - { 128, 0 }, { 112, 16 }, { 96, 32 }, { 80, 48 }, - { 64, 64 }, { 48, 80 }, { 32, 96 }, { 16, 112 }, -}; - static void var_filter_block2d_bil_w8(const uint8_t *src_ptr, uint8_t *output_ptr, unsigned int src_pixels_per_line, @@ -83,9 +79,9 @@ unsigned int aom_sub_pixel_variance8x8_neon(const uint8_t *src, int src_stride, DECLARE_ALIGNED(16, uint8_t, fdata3[9 * 8]); var_filter_block2d_bil_w8(src, fdata3, src_stride, 1, 9, 8, - bilinear_filters[xoffset]); + bilinear_filters_2t[xoffset]); var_filter_block2d_bil_w8(fdata3, temp2, 8, 8, 8, 8, - bilinear_filters[yoffset]); + bilinear_filters_2t[yoffset]); return aom_variance8x8_neon(temp2, 8, dst, dst_stride, sse); } @@ -98,9 +94,9 @@ unsigned int aom_sub_pixel_variance16x16_neon(const uint8_t *src, DECLARE_ALIGNED(16, uint8_t, fdata3[17 * 16]); var_filter_block2d_bil_w16(src, fdata3, src_stride, 1, 17, 16, - bilinear_filters[xoffset]); + bilinear_filters_2t[xoffset]); var_filter_block2d_bil_w16(fdata3, temp2, 16, 16, 16, 16, - bilinear_filters[yoffset]); + bilinear_filters_2t[yoffset]); return aom_variance16x16_neon(temp2, 16, dst, dst_stride, sse); } @@ -113,9 +109,9 @@ unsigned int aom_sub_pixel_variance32x32_neon(const uint8_t *src, DECLARE_ALIGNED(16, uint8_t, fdata3[33 * 32]); var_filter_block2d_bil_w16(src, fdata3, src_stride, 1, 33, 32, - bilinear_filters[xoffset]); + bilinear_filters_2t[xoffset]); var_filter_block2d_bil_w16(fdata3, temp2, 32, 32, 32, 32, - bilinear_filters[yoffset]); + bilinear_filters_2t[yoffset]); return aom_variance32x32_neon(temp2, 32, dst, dst_stride, sse); } @@ -128,8 +124,8 @@ unsigned int aom_sub_pixel_variance64x64_neon(const uint8_t *src, DECLARE_ALIGNED(16, uint8_t, fdata3[65 * 64]); var_filter_block2d_bil_w16(src, fdata3, src_stride, 1, 65, 64, - bilinear_filters[xoffset]); + bilinear_filters_2t[xoffset]); var_filter_block2d_bil_w16(fdata3, temp2, 64, 64, 64, 64, - bilinear_filters[yoffset]); + bilinear_filters_2t[yoffset]); return aom_variance64x64_neon(temp2, 64, dst, dst_stride, sse); } |