Skip to content

GitLab

A
aom-rav1e
Commit f19f35f7 authored by David Barker's avatar David Barker Committed by Debargha Mukherjee
Browse files
Options

ext-inter: Further cleanup 

* Rename the 'masked_compound_*' functions to just 'masked_*'.
  The previous names were intended to be temporary, to distinguish
  the old and new masked motion search pipelines. But now that the
  old pipeline has been removed, we can reuse the old names.

* Simplify the new ext-inter compound motion search pipeline
  a bit.

* Harmonize names: Rename
  aom_highbd_masked_compound_sub_pixel_variance* to
  aom_highbd_8_masked_sub_pixel_variance*, to match the naming of
  the corresponding non-masked functions

Change-Id: I988768ffe2f42a942405b7d8e93a2757a012dca3
parent 73826033
Expand all Hide whitespace changes
Inline Side-by-side
Showing with 465 additions and 556 deletions
aom_dsp/aom_dsp_rtcd_defs.pl
View file @ f19f35f7
......@@ -740,13 +740,13 @@ if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") {
if (aom_config("CONFIG_EXT_INTER") eq "yes") {
foreach (@block_sizes) {
($w, $h) = @$_;
add_proto qw/unsigned int/, "aom_masked_compound_sad${w}x${h}", "const uint8_t *src, int src_stride, const uint8_t *ref, int ref_stride, const uint8_t *second_pred, const uint8_t *msk, int msk_stride, int invert_mask";
add_proto qw/unsigned int/, "aom_masked_sad${w}x${h}", "const uint8_t *src, int src_stride, const uint8_t *ref, int ref_stride, const uint8_t *second_pred, const uint8_t *msk, int msk_stride, int invert_mask";
}
if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") {
foreach (@block_sizes) {
($w, $h) = @$_;
add_proto qw/unsigned int/, "aom_highbd_masked_compound_sad${w}x${h}", "const uint8_t *src8, int src_stride, const uint8_t *ref8, int ref_stride, const uint8_t *second_pred8, const uint8_t *msk, int msk_stride, int invert_mask";
add_proto qw/unsigned int/, "aom_highbd_masked_sad${w}x${h}", "const uint8_t *src8, int src_stride, const uint8_t *ref8, int ref_stride, const uint8_t *second_pred8, const uint8_t *msk, int msk_stride, int invert_mask";
}
}
}
......@@ -1045,14 +1045,14 @@ if (aom_config("CONFIG_EXT_INTER") eq "yes") {
#
foreach (@block_sizes) {
($w, $h) = @$_;
add_proto qw/unsigned int/, "aom_masked_compound_sub_pixel_variance${w}x${h}", "const uint8_t *src, int src_stride, int xoffset, int yoffset, const uint8_t *ref, int ref_stride, const uint8_t *second_pred, const uint8_t *msk, int msk_stride, int invert_mask, unsigned int *sse";
add_proto qw/unsigned int/, "aom_masked_sub_pixel_variance${w}x${h}", "const uint8_t *src, int src_stride, int xoffset, int yoffset, const uint8_t *ref, int ref_stride, const uint8_t *second_pred, const uint8_t *msk, int msk_stride, int invert_mask, unsigned int *sse";
}
if (aom_config("CONFIG_HIGHBITDEPTH") eq "yes") {
foreach $bd ("_", "_10_", "_12_") {
foreach $bd ("_8_", "_10_", "_12_") {
foreach (@block_sizes) {
($w, $h) = @$_;
add_proto qw/unsigned int/, "aom_highbd${bd}masked_compound_sub_pixel_variance${w}x${h}", "const uint8_t *src, int src_stride, int xoffset, int yoffset, const uint8_t *ref, int ref_stride, const uint8_t *second_pred, const uint8_t *msk, int msk_stride, int invert_mask, unsigned int *sse";
add_proto qw/unsigned int/, "aom_highbd${bd}masked_sub_pixel_variance${w}x${h}", "const uint8_t *src, int src_stride, int xoffset, int yoffset, const uint8_t *ref, int ref_stride, const uint8_t *second_pred, const uint8_t *msk, int msk_stride, int invert_mask, unsigned int *sse";
}
}
}
......
aom_dsp/sad.c
View file @ f19f35f7
......@@ -312,11 +312,10 @@ highbd_sadMxNx4D(4, 4)
#if CONFIG_AV1 && CONFIG_EXT_INTER
static INLINE
unsigned int masked_compound_sad(const uint8_t *src, int src_stride,
const uint8_t *a, int a_stride,
const uint8_t *b, int b_stride,
const uint8_t *m, int m_stride, int width,
int height) {
unsigned int masked_sad(const uint8_t *src, int src_stride,
const uint8_t *a, int a_stride, const uint8_t *b,
int b_stride, const uint8_t *m, int m_stride,
int width, int height) {
int y, x;
unsigned int sad = 0;
......@@ -336,17 +335,17 @@ highbd_sadMxNx4D(4, 4)
return sad;
}
#define MASKSADMxN(m, n) \
unsigned int aom_masked_compound_sad##m##x##n##_c( \
const uint8_t *src, int src_stride, const uint8_t *ref, int ref_stride, \
const uint8_t *second_pred, const uint8_t *msk, int msk_stride, \
int invert_mask) { \
if (!invert_mask) \
return masked_compound_sad(src, src_stride, ref, ref_stride, \
second_pred, m, msk, msk_stride, m, n); \
else \
return masked_compound_sad(src, src_stride, second_pred, m, ref, \
ref_stride, msk, msk_stride, m, n); \
#define MASKSADMxN(m, n) \
unsigned int aom_masked_sad##m##x##n##_c( \
const uint8_t *src, int src_stride, const uint8_t *ref, int ref_stride, \
const uint8_t *second_pred, const uint8_t *msk, int msk_stride, \
int invert_mask) { \
if (!invert_mask) \
return masked_sad(src, src_stride, ref, ref_stride, second_pred, m, msk, \
msk_stride, m, n); \
else \
return masked_sad(src, src_stride, second_pred, m, ref, ref_stride, msk, \
msk_stride, m, n); \
}
/* clang-format off */
......@@ -372,11 +371,11 @@ MASKSADMxN(4, 4)
#if CONFIG_HIGHBITDEPTH
static INLINE
unsigned int highbd_masked_compound_sad(const uint8_t *src8, int src_stride,
const uint8_t *a8, int a_stride,
const uint8_t *b8, int b_stride,
const uint8_t *m, int m_stride,
int width, int height) {
unsigned int highbd_masked_sad(const uint8_t *src8, int src_stride,
const uint8_t *a8, int a_stride,
const uint8_t *b8, int b_stride,
const uint8_t *m, int m_stride, int width,
int height) {
int y, x;
unsigned int sad = 0;
const uint16_t *src = CONVERT_TO_SHORTPTR(src8);
......@@ -399,19 +398,17 @@ MASKSADMxN(4, 4)
return sad;
}
#define HIGHBD_MASKSADMXN(m, n) \
unsigned int aom_highbd_masked_compound_sad##m##x##n##_c( \
const uint8_t *src8, int src_stride, const uint8_t *ref8, \
int ref_stride, const uint8_t *second_pred8, const uint8_t *msk, \
int msk_stride, int invert_mask) { \
if (!invert_mask) \
return highbd_masked_compound_sad(src8, src_stride, ref8, ref_stride, \
second_pred8, m, msk, msk_stride, m, \
n); \
else \
return highbd_masked_compound_sad(src8, src_stride, second_pred8, m, \
ref8, ref_stride, msk, msk_stride, m, \
n); \
#define HIGHBD_MASKSADMXN(m, n) \
unsigned int aom_highbd_masked_sad##m##x##n##_c( \
const uint8_t *src8, int src_stride, const uint8_t *ref8, \
int ref_stride, const uint8_t *second_pred8, const uint8_t *msk, \
int msk_stride, int invert_mask) { \
if (!invert_mask) \
return highbd_masked_sad(src8, src_stride, ref8, ref_stride, \
second_pred8, m, msk, msk_stride, m, n); \
else \
return highbd_masked_sad(src8, src_stride, second_pred8, m, ref8, \
ref_stride, msk, msk_stride, m, n); \
}
#if CONFIG_EXT_PARTITION
......
aom_dsp/variance.c
View file @ f19f35f7
......@@ -714,24 +714,24 @@ void aom_comp_mask_upsampled_pred_c(uint8_t *comp_pred, const uint8_t *pred,
}
}
#define MASK_SUBPIX_VAR(W, H) \
unsigned int aom_masked_compound_sub_pixel_variance##W##x##H##_c( \
const uint8_t *src, int src_stride, int xoffset, int yoffset, \
const uint8_t *ref, int ref_stride, const uint8_t *second_pred, \
const uint8_t *msk, int msk_stride, int invert_mask, \
unsigned int *sse) { \
uint16_t fdata3[(H + 1) * W]; \
uint8_t temp2[H * W]; \
DECLARE_ALIGNED(16, uint8_t, temp3[H * W]); \
\
var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, W, \
bilinear_filters_2t[xoffset]); \
var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \
bilinear_filters_2t[yoffset]); \
\
aom_comp_mask_pred(temp3, second_pred, W, H, temp2, W, msk, msk_stride, \
invert_mask); \
return aom_variance##W##x##H##_c(temp3, W, ref, ref_stride, sse); \
#define MASK_SUBPIX_VAR(W, H) \
unsigned int aom_masked_sub_pixel_variance##W##x##H##_c( \
const uint8_t *src, int src_stride, int xoffset, int yoffset, \
const uint8_t *ref, int ref_stride, const uint8_t *second_pred, \
const uint8_t *msk, int msk_stride, int invert_mask, \
unsigned int *sse) { \
uint16_t fdata3[(H + 1) * W]; \
uint8_t temp2[H * W]; \
DECLARE_ALIGNED(16, uint8_t, temp3[H * W]); \
\
var_filter_block2d_bil_first_pass(src, fdata3, src_stride, 1, H + 1, W, \
bilinear_filters_2t[xoffset]); \
var_filter_block2d_bil_second_pass(fdata3, temp2, W, W, H, W, \
bilinear_filters_2t[yoffset]); \
\
aom_comp_mask_pred_c(temp3, second_pred, W, H, temp2, W, msk, msk_stride, \
invert_mask); \
return aom_variance##W##x##H##_c(temp3, W, ref, ref_stride, sse); \
}
MASK_SUBPIX_VAR(4, 4)
......@@ -800,7 +800,7 @@ void aom_highbd_comp_mask_upsampled_pred_c(uint16_t *comp_pred,
}
#define HIGHBD_MASK_SUBPIX_VAR(W, H) \
unsigned int aom_highbd_masked_compound_sub_pixel_variance##W##x##H##_c( \
unsigned int aom_highbd_8_masked_sub_pixel_variance##W##x##H##_c( \
const uint8_t *src, int src_stride, int xoffset, int yoffset, \
const uint8_t *ref, int ref_stride, const uint8_t *second_pred, \
const uint8_t *msk, int msk_stride, int invert_mask, \
......@@ -822,7 +822,7 @@ void aom_highbd_comp_mask_upsampled_pred_c(uint16_t *comp_pred,
ref, ref_stride, sse); \
} \
\
unsigned int aom_highbd_10_masked_compound_sub_pixel_variance##W##x##H##_c( \
unsigned int aom_highbd_10_masked_sub_pixel_variance##W##x##H##_c( \
const uint8_t *src, int src_stride, int xoffset, int yoffset, \
const uint8_t *ref, int ref_stride, const uint8_t *second_pred, \
const uint8_t *msk, int msk_stride, int invert_mask, \
......@@ -844,7 +844,7 @@ void aom_highbd_comp_mask_upsampled_pred_c(uint16_t *comp_pred,
ref, ref_stride, sse); \
} \
\
unsigned int aom_highbd_12_masked_compound_sub_pixel_variance##W##x##H##_c( \
unsigned int aom_highbd_12_masked_sub_pixel_variance##W##x##H##_c( \
const uint8_t *src, int src_stride, int xoffset, int yoffset, \
const uint8_t *ref, int ref_stride, const uint8_t *second_pred, \
const uint8_t *msk, int msk_stride, int invert_mask, \
......
aom_dsp/variance.h
View file @ f19f35f7
......@@ -55,11 +55,12 @@ typedef unsigned int (*aom_subp_avg_variance_fn_t)(
int b_stride, unsigned int *sse, const uint8_t *second_pred);
#if CONFIG_AV1 && CONFIG_EXT_INTER
typedef unsigned int (*aom_masked_compound_sad_fn_t)(
const uint8_t *src, int src_stride, const uint8_t *ref, int ref_stride,
const uint8_t *second_pred, const uint8_t *msk, int msk_stride,
int invert_mask);
typedef unsigned int (*aom_masked_compound_subpixvariance_fn_t)(
typedef unsigned int (*aom_masked_sad_fn_t)(const uint8_t *src, int src_stride,
const uint8_t *ref, int ref_stride,
const uint8_t *second_pred,
const uint8_t *msk, int msk_stride,
int invert_mask);
typedef unsigned int (*aom_masked_subpixvariance_fn_t)(
const uint8_t *src, int src_stride, int xoffset, int yoffset,
const uint8_t *ref, int ref_stride, const uint8_t *second_pred,
const uint8_t *msk, int msk_stride, int invert_mask, unsigned int *sse);
......@@ -90,8 +91,8 @@ typedef struct aom_variance_vtable {
aom_sad_multi_fn_t sdx8f;
aom_sad_multi_d_fn_t sdx4df;
#if CONFIG_EXT_INTER
aom_masked_compound_sad_fn_t mcsdf;
aom_masked_compound_subpixvariance_fn_t mcsvf;
aom_masked_sad_fn_t msdf;
aom_masked_subpixvariance_fn_t msvf;
#endif // CONFIG_EXT_INTER
#if CONFIG_MOTION_VAR
aom_obmc_sad_fn_t osdf;
......
av1/encoder/encoder.c
View file @ f19f35f7
......@@ -1172,8 +1172,8 @@ MAKE_BFP_SAD4D_WRAPPER(aom_highbd_sad4x4x4d)
#if CONFIG_EXT_INTER
#define HIGHBD_MBFP(BT, MCSDF, MCSVF) \
cpi->fn_ptr[BT].mcsdf = MCSDF; \
cpi->fn_ptr[BT].mcsvf = MCSVF;
cpi->fn_ptr[BT].msdf = MCSDF; \
cpi->fn_ptr[BT].msvf = MCSVF;
#define MAKE_MBFP_COMPOUND_SAD_WRAPPER(fnname) \
static unsigned int fnname##_bits8( \
......@@ -1201,23 +1201,23 @@ MAKE_BFP_SAD4D_WRAPPER(aom_highbd_sad4x4x4d)
}
#if CONFIG_EXT_PARTITION
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_compound_sad128x128)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_compound_sad128x64)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_compound_sad64x128)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_sad128x128)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_sad128x64)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_sad64x128)
#endif // CONFIG_EXT_PARTITION
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_compound_sad64x64)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_compound_sad64x32)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_compound_sad32x64)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_compound_sad32x32)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_compound_sad32x16)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_compound_sad16x32)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_compound_sad16x16)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_compound_sad16x8)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_compound_sad8x16)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_compound_sad8x8)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_compound_sad8x4)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_compound_sad4x8)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_compound_sad4x4)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_sad64x64)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_sad64x32)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_sad32x64)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_sad32x32)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_sad32x16)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_sad16x32)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_sad16x16)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_sad16x8)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_sad8x16)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_sad8x8)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_sad8x4)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_sad4x8)
MAKE_MBFP_COMPOUND_SAD_WRAPPER(aom_highbd_masked_sad4x4)
#endif // CONFIG_EXT_INTER
#if CONFIG_MOTION_VAR
......@@ -1382,39 +1382,39 @@ static void highbd_set_var_fns(AV1_COMP *const cpi) {
#if CONFIG_EXT_INTER
#if CONFIG_EXT_PARTITION
HIGHBD_MBFP(BLOCK_128X128, aom_highbd_masked_compound_sad128x128_bits8,
aom_highbd_masked_compound_sub_pixel_variance128x128)
HIGHBD_MBFP(BLOCK_128X64, aom_highbd_masked_compound_sad128x64_bits8,
aom_highbd_masked_compound_sub_pixel_variance128x64)
HIGHBD_MBFP(BLOCK_64X128, aom_highbd_masked_compound_sad64x128_bits8,
aom_highbd_masked_compound_sub_pixel_variance64x128)
HIGHBD_MBFP(BLOCK_128X128, aom_highbd_masked_sad128x128_bits8,
aom_highbd_8_masked_sub_pixel_variance128x128)
HIGHBD_MBFP(BLOCK_128X64, aom_highbd_masked_sad128x64_bits8,
aom_highbd_8_masked_sub_pixel_variance128x64)
HIGHBD_MBFP(BLOCK_64X128, aom_highbd_masked_sad64x128_bits8,
aom_highbd_8_masked_sub_pixel_variance64x128)
#endif // CONFIG_EXT_PARTITION
HIGHBD_MBFP(BLOCK_64X64, aom_highbd_masked_compound_sad64x64_bits8,
aom_highbd_masked_compound_sub_pixel_variance64x64)
HIGHBD_MBFP(BLOCK_64X32, aom_highbd_masked_compound_sad64x32_bits8,
aom_highbd_masked_compound_sub_pixel_variance64x32)
HIGHBD_MBFP(BLOCK_32X64, aom_highbd_masked_compound_sad32x64_bits8,
aom_highbd_masked_compound_sub_pixel_variance32x64)
HIGHBD_MBFP(BLOCK_32X32, aom_highbd_masked_compound_sad32x32_bits8,
aom_highbd_masked_compound_sub_pixel_variance32x32)
HIGHBD_MBFP(BLOCK_32X16, aom_highbd_masked_compound_sad32x16_bits8,
aom_highbd_masked_compound_sub_pixel_variance32x16)
HIGHBD_MBFP(BLOCK_16X32, aom_highbd_masked_compound_sad16x32_bits8,
aom_highbd_masked_compound_sub_pixel_variance16x32)
HIGHBD_MBFP(BLOCK_16X16, aom_highbd_masked_compound_sad16x16_bits8,
aom_highbd_masked_compound_sub_pixel_variance16x16)
HIGHBD_MBFP(BLOCK_8X16, aom_highbd_masked_compound_sad8x16_bits8,
aom_highbd_masked_compound_sub_pixel_variance8x16)
HIGHBD_MBFP(BLOCK_16X8, aom_highbd_masked_compound_sad16x8_bits8,
aom_highbd_masked_compound_sub_pixel_variance16x8)
HIGHBD_MBFP(BLOCK_8X8, aom_highbd_masked_compound_sad8x8_bits8,
aom_highbd_masked_compound_sub_pixel_variance8x8)
HIGHBD_MBFP(BLOCK_4X8, aom_highbd_masked_compound_sad4x8_bits8,
aom_highbd_masked_compound_sub_pixel_variance4x8)
HIGHBD_MBFP(BLOCK_8X4, aom_highbd_masked_compound_sad8x4_bits8,
aom_highbd_masked_compound_sub_pixel_variance8x4)
HIGHBD_MBFP(BLOCK_4X4, aom_highbd_masked_compound_sad4x4_bits8,
aom_highbd_masked_compound_sub_pixel_variance4x4)
HIGHBD_MBFP(BLOCK_64X64, aom_highbd_masked_sad64x64_bits8,
aom_highbd_8_masked_sub_pixel_variance64x64)
HIGHBD_MBFP(BLOCK_64X32, aom_highbd_masked_sad64x32_bits8,
aom_highbd_8_masked_sub_pixel_variance64x32)
HIGHBD_MBFP(BLOCK_32X64, aom_highbd_masked_sad32x64_bits8,
aom_highbd_8_masked_sub_pixel_variance32x64)
HIGHBD_MBFP(BLOCK_32X32, aom_highbd_masked_sad32x32_bits8,
aom_highbd_8_masked_sub_pixel_variance32x32)
HIGHBD_MBFP(BLOCK_32X16, aom_highbd_masked_sad32x16_bits8,
aom_highbd_8_masked_sub_pixel_variance32x16)
HIGHBD_MBFP(BLOCK_16X32, aom_highbd_masked_sad16x32_bits8,
aom_highbd_8_masked_sub_pixel_variance16x32)
HIGHBD_MBFP(BLOCK_16X16, aom_highbd_masked_sad16x16_bits8,
aom_highbd_8_masked_sub_pixel_variance16x16)
HIGHBD_MBFP(BLOCK_8X16, aom_highbd_masked_sad8x16_bits8,
aom_highbd_8_masked_sub_pixel_variance8x16)
HIGHBD_MBFP(BLOCK_16X8, aom_highbd_masked_sad16x8_bits8,
aom_highbd_8_masked_sub_pixel_variance16x8)
HIGHBD_MBFP(BLOCK_8X8, aom_highbd_masked_sad8x8_bits8,
aom_highbd_8_masked_sub_pixel_variance8x8)
HIGHBD_MBFP(BLOCK_4X8, aom_highbd_masked_sad4x8_bits8,
aom_highbd_8_masked_sub_pixel_variance4x8)
HIGHBD_MBFP(BLOCK_8X4, aom_highbd_masked_sad8x4_bits8,
aom_highbd_8_masked_sub_pixel_variance8x4)
HIGHBD_MBFP(BLOCK_4X4, aom_highbd_masked_sad4x4_bits8,
aom_highbd_8_masked_sub_pixel_variance4x4)
#endif // CONFIG_EXT_INTER
#if CONFIG_MOTION_VAR
#if CONFIG_EXT_PARTITION
......@@ -1589,39 +1589,39 @@ static void highbd_set_var_fns(AV1_COMP *const cpi) {
#if CONFIG_EXT_INTER
#if CONFIG_EXT_PARTITION
HIGHBD_MBFP(BLOCK_128X128, aom_highbd_masked_compound_sad128x128_bits10,
aom_highbd_10_masked_compound_sub_pixel_variance128x128)
HIGHBD_MBFP(BLOCK_128X64, aom_highbd_masked_compound_sad128x64_bits10,
aom_highbd_10_masked_compound_sub_pixel_variance128x64)
HIGHBD_MBFP(BLOCK_64X128, aom_highbd_masked_compound_sad64x128_bits10,
aom_highbd_10_masked_compound_sub_pixel_variance64x128)
HIGHBD_MBFP(BLOCK_128X128, aom_highbd_masked_sad128x128_bits10,
aom_highbd_10_masked_sub_pixel_variance128x128)
HIGHBD_MBFP(BLOCK_128X64, aom_highbd_masked_sad128x64_bits10,
aom_highbd_10_masked_sub_pixel_variance128x64)
HIGHBD_MBFP(BLOCK_64X128, aom_highbd_masked_sad64x128_bits10,
aom_highbd_10_masked_sub_pixel_variance64x128)
#endif // CONFIG_EXT_PARTITION
HIGHBD_MBFP(BLOCK_64X64, aom_highbd_masked_compound_sad64x64_bits10,
aom_highbd_10_masked_compound_sub_pixel_variance64x64)
HIGHBD_MBFP(BLOCK_64X32, aom_highbd_masked_compound_sad64x32_bits10,
aom_highbd_10_masked_compound_sub_pixel_variance64x32)
HIGHBD_MBFP(BLOCK_32X64, aom_highbd_masked_compound_sad32x64_bits10,
aom_highbd_10_masked_compound_sub_pixel_variance32x64)
HIGHBD_MBFP(BLOCK_32X32, aom_highbd_masked_compound_sad32x32_bits10,
aom_highbd_10_masked_compound_sub_pixel_variance32x32)
HIGHBD_MBFP(BLOCK_32X16, aom_highbd_masked_compound_sad32x16_bits10,
aom_highbd_10_masked_compound_sub_pixel_variance32x16)
HIGHBD_MBFP(BLOCK_16X32, aom_highbd_masked_compound_sad16x32_bits10,
aom_highbd_10_masked_compound_sub_pixel_variance16x32)
HIGHBD_MBFP(BLOCK_16X16, aom_highbd_masked_compound_sad16x16_bits10,
aom_highbd_10_masked_compound_sub_pixel_variance16x16)
HIGHBD_MBFP(BLOCK_8X16, aom_highbd_masked_compound_sad8x16_bits10,
aom_highbd_10_masked_compound_sub_pixel_variance8x16)
HIGHBD_MBFP(BLOCK_16X8, aom_highbd_masked_compound_sad16x8_bits10,
aom_highbd_10_masked_compound_sub_pixel_variance16x8)
HIGHBD_MBFP(BLOCK_8X8, aom_highbd_masked_compound_sad8x8_bits10,
aom_highbd_10_masked_compound_sub_pixel_variance8x8)
HIGHBD_MBFP(BLOCK_4X8, aom_highbd_masked_compound_sad4x8_bits10,
aom_highbd_10_masked_compound_sub_pixel_variance4x8)
HIGHBD_MBFP(BLOCK_8X4, aom_highbd_masked_compound_sad8x4_bits10,
aom_highbd_10_masked_compound_sub_pixel_variance8x4)
HIGHBD_MBFP(BLOCK_4X4, aom_highbd_masked_compound_sad4x4_bits10,
aom_highbd_10_masked_compound_sub_pixel_variance4x4)
HIGHBD_MBFP(BLOCK_64X64, aom_highbd_masked_sad64x64_bits10,
aom_highbd_10_masked_sub_pixel_variance64x64)
HIGHBD_MBFP(BLOCK_64X32, aom_highbd_masked_sad64x32_bits10,
aom_highbd_10_masked_sub_pixel_variance64x32)
HIGHBD_MBFP(BLOCK_32X64, aom_highbd_masked_sad32x64_bits10,
aom_highbd_10_masked_sub_pixel_variance32x64)
HIGHBD_MBFP(BLOCK_32X32, aom_highbd_masked_sad32x32_bits10,
aom_highbd_10_masked_sub_pixel_variance32x32)
HIGHBD_MBFP(BLOCK_32X16, aom_highbd_masked_sad32x16_bits10,
aom_highbd_10_masked_sub_pixel_variance32x16)
HIGHBD_MBFP(BLOCK_16X32, aom_highbd_masked_sad16x32_bits10,
aom_highbd_10_masked_sub_pixel_variance16x32)
HIGHBD_MBFP(BLOCK_16X16, aom_highbd_masked_sad16x16_bits10,
aom_highbd_10_masked_sub_pixel_variance16x16)
HIGHBD_MBFP(BLOCK_8X16, aom_highbd_masked_sad8x16_bits10,
aom_highbd_10_masked_sub_pixel_variance8x16)
HIGHBD_MBFP(BLOCK_16X8, aom_highbd_masked_sad16x8_bits10,
aom_highbd_10_masked_sub_pixel_variance16x8)
HIGHBD_MBFP(BLOCK_8X8, aom_highbd_masked_sad8x8_bits10,
aom_highbd_10_masked_sub_pixel_variance8x8)
HIGHBD_MBFP(BLOCK_4X8, aom_highbd_masked_sad4x8_bits10,
aom_highbd_10_masked_sub_pixel_variance4x8)
HIGHBD_MBFP(BLOCK_8X4, aom_highbd_masked_sad8x4_bits10,
aom_highbd_10_masked_sub_pixel_variance8x4)
HIGHBD_MBFP(BLOCK_4X4, aom_highbd_masked_sad4x4_bits10,
aom_highbd_10_masked_sub_pixel_variance4x4)
#endif // CONFIG_EXT_INTER
#if CONFIG_MOTION_VAR
#if CONFIG_EXT_PARTITION
......@@ -1796,39 +1796,39 @@ static void highbd_set_var_fns(AV1_COMP *const cpi) {
#if CONFIG_EXT_INTER
#if CONFIG_EXT_PARTITION
HIGHBD_MBFP(BLOCK_128X128, aom_highbd_masked_compound_sad128x128_bits12,
aom_highbd_12_masked_compound_sub_pixel_variance128x128)
HIGHBD_MBFP(BLOCK_128X64, aom_highbd_masked_compound_sad128x64_bits12,
aom_highbd_12_masked_compound_sub_pixel_variance128x64)
HIGHBD_MBFP(BLOCK_64X128, aom_highbd_masked_compound_sad64x128_bits12,
aom_highbd_12_masked_compound_sub_pixel_variance64x128)
HIGHBD_MBFP(BLOCK_128X128, aom_highbd_masked_sad128x128_bits12,
aom_highbd_12_masked_sub_pixel_variance128x128)
HIGHBD_MBFP(BLOCK_128X64, aom_highbd_masked_sad128x64_bits12,
aom_highbd_12_masked_sub_pixel_variance128x64)
HIGHBD_MBFP(BLOCK_64X128, aom_highbd_masked_sad64x128_bits12,
aom_highbd_12_masked_sub_pixel_variance64x128)
#endif // CONFIG_EXT_PARTITION
HIGHBD_MBFP(BLOCK_64X64, aom_highbd_masked_compound_sad64x64_bits12,
aom_highbd_12_masked_compound_sub_pixel_variance64x64)
HIGHBD_MBFP(BLOCK_64X32, aom_highbd_masked_compound_sad64x32_bits12,
aom_highbd_12_masked_compound_sub_pixel_variance64x32)
HIGHBD_MBFP(BLOCK_32X64, aom_highbd_masked_compound_sad32x64_bits12,
aom_highbd_12_masked_compound_sub_pixel_variance32x64)
HIGHBD_MBFP(BLOCK_32X32, aom_highbd_masked_compound_sad32x32_bits12,
aom_highbd_12_masked_compound_sub_pixel_variance32x32)
HIGHBD_MBFP(BLOCK_32X16, aom_highbd_masked_compound_sad32x16_bits12,
aom_highbd_12_masked_compound_sub_pixel_variance32x16)
HIGHBD_MBFP(BLOCK_16X32, aom_highbd_masked_compound_sad16x32_bits12,
aom_highbd_12_masked_compound_sub_pixel_variance16x32)
HIGHBD_MBFP(BLOCK_16X16, aom_highbd_masked_compound_sad16x16_bits12,
aom_highbd_12_masked_compound_sub_pixel_variance16x16)
HIGHBD_MBFP(BLOCK_8X16, aom_highbd_masked_compound_sad8x16_bits12,
aom_highbd_12_masked_compound_sub_pixel_variance8x16)
HIGHBD_MBFP(BLOCK_16X8, aom_highbd_masked_compound_sad16x8_bits12,
aom_highbd_12_masked_compound_sub_pixel_variance16x8)
HIGHBD_MBFP(BLOCK_8X8, aom_highbd_masked_compound_sad8x8_bits12,
aom_highbd_12_masked_compound_sub_pixel_variance8x8)
HIGHBD_MBFP(BLOCK_4X8, aom_highbd_masked_compound_sad4x8_bits12,
aom_highbd_12_masked_compound_sub_pixel_variance4x8)
HIGHBD_MBFP(BLOCK_8X4, aom_highbd_masked_compound_sad8x4_bits12,
aom_highbd_12_masked_compound_sub_pixel_variance8x4)
HIGHBD_MBFP(BLOCK_4X4, aom_highbd_masked_compound_sad4x4_bits12,
aom_highbd_12_masked_compound_sub_pixel_variance4x4)
HIGHBD_MBFP(BLOCK_64X64, aom_highbd_masked_sad64x64_bits12,
aom_highbd_12_masked_sub_pixel_variance64x64)
HIGHBD_MBFP(BLOCK_64X32, aom_highbd_masked_sad64x32_bits12,
aom_highbd_12_masked_sub_pixel_variance64x32)
HIGHBD_MBFP(BLOCK_32X64, aom_highbd_masked_sad32x64_bits12,
aom_highbd_12_masked_sub_pixel_variance32x64)
HIGHBD_MBFP(BLOCK_32X32, aom_highbd_masked_sad32x32_bits12,
aom_highbd_12_masked_sub_pixel_variance32x32)
HIGHBD_MBFP(BLOCK_32X16, aom_highbd_masked_sad32x16_bits12,
aom_highbd_12_masked_sub_pixel_variance32x16)
HIGHBD_MBFP(BLOCK_16X32, aom_highbd_masked_sad16x32_bits12,
aom_highbd_12_masked_sub_pixel_variance16x32)
HIGHBD_MBFP(BLOCK_16X16, aom_highbd_masked_sad16x16_bits12,
aom_highbd_12_masked_sub_pixel_variance16x16)
HIGHBD_MBFP(BLOCK_8X16, aom_highbd_masked_sad8x16_bits12,
aom_highbd_12_masked_sub_pixel_variance8x16)
HIGHBD_MBFP(BLOCK_16X8, aom_highbd_masked_sad16x8_bits12,
aom_highbd_12_masked_sub_pixel_variance16x8)
HIGHBD_MBFP(BLOCK_8X8, aom_highbd_masked_sad8x8_bits12,
aom_highbd_12_masked_sub_pixel_variance8x8)
HIGHBD_MBFP(BLOCK_4X8, aom_highbd_masked_sad4x8_bits12,
aom_highbd_12_masked_sub_pixel_variance4x8)
HIGHBD_MBFP(BLOCK_8X4, aom_highbd_masked_sad8x4_bits12,
aom_highbd_12_masked_sub_pixel_variance8x4)
HIGHBD_MBFP(BLOCK_4X4, aom_highbd_masked_sad4x4_bits12,
aom_highbd_12_masked_sub_pixel_variance4x4)
#endif // CONFIG_EXT_INTER
#if CONFIG_MOTION_VAR
......@@ -2411,44 +2411,29 @@ AV1_COMP *av1_create_compressor(AV1EncoderConfig *oxcf,
#endif // CONFIG_MOTION_VAR
#if CONFIG_EXT_INTER
#define MBFP(BT, MCSDF, MCSVF) \
cpi->fn_ptr[BT].mcsdf = MCSDF; \
cpi->fn_ptr[BT].mcsvf = MCSVF;
#define MBFP(BT, MCSDF, MCSVF) \
cpi->fn_ptr[BT].msdf = MCSDF; \
cpi->fn_ptr[BT].msvf = MCSVF;
#if CONFIG_EXT_PARTITION
MBFP(BLOCK_128X128, aom_masked_compound_sad128x128,
aom_masked_compound_sub_pixel_variance128x128)
MBFP(BLOCK_128X64, aom_masked_compound_sad128x64,
aom_masked_compound_sub_pixel_variance128x64)
MBFP(BLOCK_64X128, aom_masked_compound_sad64x128,
aom_masked_compound_sub_pixel_variance64x128)
MBFP(BLOCK_128X128, aom_masked_sad128x128,
aom_masked_sub_pixel_variance128x128)
MBFP(BLOCK_128X64, aom_masked_sad128x64, aom_masked_sub_pixel_variance128x64)
MBFP(BLOCK_64X128, aom_masked_sad64x128, aom_masked_sub_pixel_variance64x128)
#endif // CONFIG_EXT_PARTITION
MBFP(BLOCK_64X64, aom_masked_compound_sad64x64,
aom_masked_compound_sub_pixel_variance64x64)
MBFP(BLOCK_64X32, aom_masked_compound_sad64x32,
aom_masked_compound_sub_pixel_variance64x32)
MBFP(BLOCK_32X64, aom_masked_compound_sad32x64,
aom_masked_compound_sub_pixel_variance32x64)
MBFP(BLOCK_32X32, aom_masked_compound_sad32x32,
aom_masked_compound_sub_pixel_variance32x32)
MBFP(BLOCK_32X16, aom_masked_compound_sad32x16,
aom_masked_compound_sub_pixel_variance32x16)
MBFP(BLOCK_16X32, aom_masked_compound_sad16x32,
aom_masked_compound_sub_pixel_variance16x32)
MBFP(BLOCK_16X16, aom_masked_compound_sad16x16,
aom_masked_compound_sub_pixel_variance16x16)
MBFP(BLOCK_16X8, aom_masked_compound_sad16x8,
aom_masked_compound_sub_pixel_variance16x8)
MBFP(BLOCK_8X16, aom_masked_compound_sad8x16,
aom_masked_compound_sub_pixel_variance8x16)
MBFP(BLOCK_8X8, aom_masked_compound_sad8x8,
aom_masked_compound_sub_pixel_variance8x8)
MBFP(BLOCK_4X8, aom_masked_compound_sad4x8,
aom_masked_compound_sub_pixel_variance4x8)
MBFP(BLOCK_8X4, aom_masked_compound_sad8x4,
aom_masked_compound_sub_pixel_variance8x4)
MBFP(BLOCK_4X4, aom_masked_compound_sad4x4,
aom_masked_compound_sub_pixel_variance4x4)
MBFP(BLOCK_64X64, aom_masked_sad64x64, aom_masked_sub_pixel_variance64x64)
MBFP(BLOCK_64X32, aom_masked_sad64x32, aom_masked_sub_pixel_variance64x32)
MBFP(BLOCK_32X64, aom_masked_sad32x64, aom_masked_sub_pixel_variance32x64)
MBFP(BLOCK_32X32, aom_masked_sad32x32, aom_masked_sub_pixel_variance32x32)
MBFP(BLOCK_32X16, aom_masked_sad32x16, aom_masked_sub_pixel_variance32x16)
MBFP(BLOCK_16X32, aom_masked_sad16x32, aom_masked_sub_pixel_variance16x32)
MBFP(BLOCK_16X16, aom_masked_sad16x16, aom_masked_sub_pixel_variance16x16)
MBFP(BLOCK_16X8, aom_masked_sad16x8, aom_masked_sub_pixel_variance16x8)
MBFP(BLOCK_8X16, aom_masked_sad8x16, aom_masked_sub_pixel_variance8x16)
MBFP(BLOCK_8X8, aom_masked_sad8x8, aom_masked_sub_pixel_variance8x8)
MBFP(BLOCK_4X8, aom_masked_sad4x8, aom_masked_sub_pixel_variance4x8)
MBFP(BLOCK_8X4, aom_masked_sad8x4, aom_masked_sub_pixel_variance8x4)
MBFP(BLOCK_4X4, aom_masked_sad4x4, aom_masked_sub_pixel_variance4x4)
#endif // CONFIG_EXT_INTER
#if CONFIG_HIGHBITDEPTH
......
av1/encoder/mcomp.c
View file @ f19f35f7
......@@ -177,30 +177,30 @@ static INLINE const uint8_t *pre(const uint8_t *buf, int stride, int r, int c) {
/* checks if (r, c) has better score than previous best */
#if CONFIG_EXT_INTER
#define CHECK_BETTER(v, r, c) \
if (c >= minc && c <= maxc && r >= minr && r <= maxr) { \
MV this_mv = { r, c }; \
v = mv_err_cost(&this_mv, ref_mv, mvjcost, mvcost, error_per_bit); \
if (second_pred == NULL) \
thismse = vfp->svf(pre(y, y_stride, r, c), y_stride, sp(c), sp(r), \
src_address, src_stride, &sse); \
else if (mask) \
thismse = vfp->mcsvf(pre(y, y_stride, r, c), y_stride, sp(c), sp(r), \
src_address, src_stride, second_pred, mask, \
mask_stride, invert_mask, &sse); \
else \
thismse = vfp->svaf(pre(y, y_stride, r, c), y_stride, sp(c), sp(r), \
src_address, src_stride, &sse, second_pred); \
v += thismse; \
if (v < besterr) { \
besterr = v; \
br = r; \
bc = c; \
*distortion = thismse; \
*sse1 = sse; \
} \
} else { \
v = INT_MAX; \
#define CHECK_BETTER(v, r, c) \
if (c >= minc && c <= maxc && r >= minr && r <= maxr) { \
MV this_mv = { r, c }; \
v = mv_err_cost(&this_mv, ref_mv, mvjcost, mvcost, error_per_bit); \
if (second_pred == NULL) \
thismse = vfp->svf(pre(y, y_stride, r, c), y_stride, sp(c), sp(r), \
src_address, src_stride, &sse); \
else if (mask) \
thismse = vfp->msvf(pre(y, y_stride, r, c), y_stride, sp(c), sp(r), \
src_address, src_stride, second_pred, mask, \
mask_stride, invert_mask, &sse); \
else \
thismse = vfp->svaf(pre(y, y_stride, r, c), y_stride, sp(c), sp(r), \
src_address, src_stride, &sse, second_pred); \
v += thismse; \
if (v < besterr) { \
besterr = v; \
br = r; \
bc = c; \
*distortion = thismse; \
*sse1 = sse; \
} \
} else { \
v = INT_MAX; \
}
#else
#define CHECK_BETTER(v, r, c) \
......@@ -224,7 +224,7 @@ static INLINE const uint8_t *pre(const uint8_t *buf, int stride, int r, int c) {
} else { \
v = INT_MAX; \
}
#endif
#endif // CONFIG_EXT_INTER
#define CHECK_BETTER0(v, r, c) CHECK_BETTER(v, r, c)
......@@ -272,7 +272,7 @@ static INLINE const uint8_t *upre(const uint8_t *buf, int stride, int r,
} else { \
v = INT_MAX; \
}
#endif
#endif // CONFIG_EXT_INTER
#define FIRST_LEVEL_CHECKS \
{ \
......@@ -861,9 +861,9 @@ int av1_find_best_sub_pixel_tree(
src_address, src_stride, &sse);
#if CONFIG_EXT_INTER
else if (mask)
thismse = vfp->mcsvf(pre_address, y_stride, sp(tc), sp(tr),
src_address, src_stride, second_pred, mask,
mask_stride, invert_mask, &sse);