diff options
-rw-r--r-- | vp9/encoder/vp9_mcomp.c | 4 | ||||
-rw-r--r-- | vpx_dsp/sad.c | 3 | ||||
-rw-r--r-- | vpx_dsp/variance.c | 124 | ||||
-rw-r--r-- | vpx_dsp/vpx_dsp_rtcd_defs.pl | 2 |
4 files changed, 66 insertions, 67 deletions
diff --git a/vp9/encoder/vp9_mcomp.c b/vp9/encoder/vp9_mcomp.c index 37406c232..1cb978667 100644 --- a/vp9/encoder/vp9_mcomp.c +++ b/vp9/encoder/vp9_mcomp.c @@ -329,8 +329,8 @@ static unsigned int setup_center_error( if (second_pred != NULL) { if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) { DECLARE_ALIGNED(16, uint16_t, comp_pred16[64 * 64]); - vpx_highbd_comp_avg_pred(comp_pred16, second_pred, w, h, y + offset, - y_stride); + vpx_highbd_comp_avg_pred(comp_pred16, CONVERT_TO_SHORTPTR(second_pred), w, + h, CONVERT_TO_SHORTPTR(y + offset), y_stride); besterr = vfp->vf(CONVERT_TO_BYTEPTR(comp_pred16), w, src, src_stride, sse1); } else { diff --git a/vpx_dsp/sad.c b/vpx_dsp/sad.c index 18b6dc6e0..d4a532968 100644 --- a/vpx_dsp/sad.c +++ b/vpx_dsp/sad.c @@ -173,7 +173,8 @@ static INLINE unsigned int highbd_sadb(const uint8_t *a8, int a_stride, const uint8_t *src, int src_stride, const uint8_t *ref, int ref_stride, \ const uint8_t *second_pred) { \ DECLARE_ALIGNED(16, uint16_t, comp_pred[m * n]); \ - vpx_highbd_comp_avg_pred_c(comp_pred, second_pred, m, n, ref, ref_stride); \ + vpx_highbd_comp_avg_pred_c(comp_pred, CONVERT_TO_SHORTPTR(second_pred), m, \ + n, CONVERT_TO_SHORTPTR(ref), ref_stride); \ return highbd_sadb(src, src_stride, comp_pred, m, m, n); \ } diff --git a/vpx_dsp/variance.c b/vpx_dsp/variance.c index 93bd8f30d..a7eef0d0b 100644 --- a/vpx_dsp/variance.c +++ b/vpx_dsp/variance.c @@ -449,65 +449,65 @@ static void highbd_var_filter_block2d_bil_second_pass( dst, dst_stride, sse); \ } -#define HIGHBD_SUBPIX_AVG_VAR(W, H) \ - uint32_t vpx_highbd_8_sub_pixel_avg_variance##W##x##H##_c( \ - const uint8_t *src, int src_stride, int xoffset, int yoffset, \ - const uint8_t *dst, int dst_stride, uint32_t *sse, \ - const uint8_t *second_pred) { \ - uint16_t fdata3[(H + 1) * W]; \ - uint16_t temp2[H * W]; \ - DECLARE_ALIGNED(16, uint16_t, temp3[H * W]); \ - \ - highbd_var_filter_block2d_bil_first_pass( \ - src, fdata3, src_stride, 1, H + 1, W, bilinear_filters[xoffset]); \ - highbd_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \ - bilinear_filters[yoffset]); \ - \ - vpx_highbd_comp_avg_pred_c(temp3, second_pred, W, H, \ - CONVERT_TO_BYTEPTR(temp2), W); \ - \ - return vpx_highbd_8_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp3), W, \ - dst, dst_stride, sse); \ - } \ - \ - uint32_t vpx_highbd_10_sub_pixel_avg_variance##W##x##H##_c( \ - const uint8_t *src, int src_stride, int xoffset, int yoffset, \ - const uint8_t *dst, int dst_stride, uint32_t *sse, \ - const uint8_t *second_pred) { \ - uint16_t fdata3[(H + 1) * W]; \ - uint16_t temp2[H * W]; \ - DECLARE_ALIGNED(16, uint16_t, temp3[H * W]); \ - \ - highbd_var_filter_block2d_bil_first_pass( \ - src, fdata3, src_stride, 1, H + 1, W, bilinear_filters[xoffset]); \ - highbd_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \ - bilinear_filters[yoffset]); \ - \ - vpx_highbd_comp_avg_pred_c(temp3, second_pred, W, H, \ - CONVERT_TO_BYTEPTR(temp2), W); \ - \ - return vpx_highbd_10_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp3), W, \ - dst, dst_stride, sse); \ - } \ - \ - uint32_t vpx_highbd_12_sub_pixel_avg_variance##W##x##H##_c( \ - const uint8_t *src, int src_stride, int xoffset, int yoffset, \ - const uint8_t *dst, int dst_stride, uint32_t *sse, \ - const uint8_t *second_pred) { \ - uint16_t fdata3[(H + 1) * W]; \ - uint16_t temp2[H * W]; \ - DECLARE_ALIGNED(16, uint16_t, temp3[H * W]); \ - \ - highbd_var_filter_block2d_bil_first_pass( \ - src, fdata3, src_stride, 1, H + 1, W, bilinear_filters[xoffset]); \ - highbd_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \ - bilinear_filters[yoffset]); \ - \ - vpx_highbd_comp_avg_pred_c(temp3, second_pred, W, H, \ - CONVERT_TO_BYTEPTR(temp2), W); \ - \ - return vpx_highbd_12_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp3), W, \ - dst, dst_stride, sse); \ +#define HIGHBD_SUBPIX_AVG_VAR(W, H) \ + uint32_t vpx_highbd_8_sub_pixel_avg_variance##W##x##H##_c( \ + const uint8_t *src, int src_stride, int xoffset, int yoffset, \ + const uint8_t *dst, int dst_stride, uint32_t *sse, \ + const uint8_t *second_pred) { \ + uint16_t fdata3[(H + 1) * W]; \ + uint16_t temp2[H * W]; \ + DECLARE_ALIGNED(16, uint16_t, temp3[H * W]); \ + \ + highbd_var_filter_block2d_bil_first_pass( \ + src, fdata3, src_stride, 1, H + 1, W, bilinear_filters[xoffset]); \ + highbd_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \ + bilinear_filters[yoffset]); \ + \ + vpx_highbd_comp_avg_pred_c(temp3, CONVERT_TO_SHORTPTR(second_pred), W, H, \ + temp2, W); \ + \ + return vpx_highbd_8_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp3), W, \ + dst, dst_stride, sse); \ + } \ + \ + uint32_t vpx_highbd_10_sub_pixel_avg_variance##W##x##H##_c( \ + const uint8_t *src, int src_stride, int xoffset, int yoffset, \ + const uint8_t *dst, int dst_stride, uint32_t *sse, \ + const uint8_t *second_pred) { \ + uint16_t fdata3[(H + 1) * W]; \ + uint16_t temp2[H * W]; \ + DECLARE_ALIGNED(16, uint16_t, temp3[H * W]); \ + \ + highbd_var_filter_block2d_bil_first_pass( \ + src, fdata3, src_stride, 1, H + 1, W, bilinear_filters[xoffset]); \ + highbd_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \ + bilinear_filters[yoffset]); \ + \ + vpx_highbd_comp_avg_pred_c(temp3, CONVERT_TO_SHORTPTR(second_pred), W, H, \ + temp2, W); \ + \ + return vpx_highbd_10_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp3), W, \ + dst, dst_stride, sse); \ + } \ + \ + uint32_t vpx_highbd_12_sub_pixel_avg_variance##W##x##H##_c( \ + const uint8_t *src, int src_stride, int xoffset, int yoffset, \ + const uint8_t *dst, int dst_stride, uint32_t *sse, \ + const uint8_t *second_pred) { \ + uint16_t fdata3[(H + 1) * W]; \ + uint16_t temp2[H * W]; \ + DECLARE_ALIGNED(16, uint16_t, temp3[H * W]); \ + \ + highbd_var_filter_block2d_bil_first_pass( \ + src, fdata3, src_stride, 1, H + 1, W, bilinear_filters[xoffset]); \ + highbd_var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \ + bilinear_filters[yoffset]); \ + \ + vpx_highbd_comp_avg_pred_c(temp3, CONVERT_TO_SHORTPTR(second_pred), W, H, \ + temp2, W); \ + \ + return vpx_highbd_12_variance##W##x##H##_c(CONVERT_TO_BYTEPTR(temp3), W, \ + dst, dst_stride, sse); \ } /* All three forms of the variance are available in the same sizes. */ @@ -538,12 +538,10 @@ HIGHBD_MSE(16, 8) HIGHBD_MSE(8, 16) HIGHBD_MSE(8, 8) -void vpx_highbd_comp_avg_pred(uint16_t *comp_pred, const uint8_t *pred8, - int width, int height, const uint8_t *ref8, +void vpx_highbd_comp_avg_pred(uint16_t *comp_pred, const uint16_t *pred, + int width, int height, const uint16_t *ref, int ref_stride) { int i, j; - uint16_t *pred = CONVERT_TO_SHORTPTR(pred8); - uint16_t *ref = CONVERT_TO_SHORTPTR(ref8); for (i = 0; i < height; ++i) { for (j = 0; j < width; ++j) { const int tmp = pred[j] + ref[j]; diff --git a/vpx_dsp/vpx_dsp_rtcd_defs.pl b/vpx_dsp/vpx_dsp_rtcd_defs.pl index b943e157b..a51761cd3 100644 --- a/vpx_dsp/vpx_dsp_rtcd_defs.pl +++ b/vpx_dsp/vpx_dsp_rtcd_defs.pl @@ -1367,7 +1367,7 @@ if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") { add_proto qw/unsigned int vpx_highbd_12_mse8x8/, "const uint8_t *src_ptr, int source_stride, const uint8_t *ref_ptr, int recon_stride, unsigned int *sse"; specialize qw/vpx_highbd_12_mse8x8 sse2/; - add_proto qw/void vpx_highbd_comp_avg_pred/, "uint16_t *comp_pred, const uint8_t *pred8, int width, int height, const uint8_t *ref8, int ref_stride"; + add_proto qw/void vpx_highbd_comp_avg_pred/, "uint16_t *comp_pred, const uint16_t *pred, int width, int height, const uint16_t *ref, int ref_stride"; # # Subpixel Variance |