summaryrefslogtreecommitdiff
path: root/vp9
diff options
context:
space:
mode:
Diffstat (limited to 'vp9')
-rw-r--r--vp9/common/vp9_findnearmv.c4
-rw-r--r--vp9/encoder/vp9_encodeframe.c32
-rw-r--r--vp9/encoder/vp9_encodeintra.c105
-rw-r--r--vp9/encoder/vp9_encodemb.c55
-rw-r--r--vp9/encoder/vp9_mbgraph.c14
-rw-r--r--vp9/encoder/vp9_mbgraph.h2
-rw-r--r--vp9/encoder/vp9_mcomp.c109
-rw-r--r--vp9/encoder/vp9_mcomp.h32
-rw-r--r--vp9/encoder/vp9_onyx_if.c2
-rw-r--r--vp9/encoder/x86/vp9_dct_sse2_intrinsics.c (renamed from vp9/encoder/x86/vp9_dct_sse2.c)0
-rw-r--r--vp9/vp9cx.mk6
11 files changed, 154 insertions, 207 deletions
diff --git a/vp9/common/vp9_findnearmv.c b/vp9/common/vp9_findnearmv.c
index f713212a7..77b79395e 100644
--- a/vp9/common/vp9_findnearmv.c
+++ b/vp9/common/vp9_findnearmv.c
@@ -190,7 +190,11 @@ void vp9_find_best_ref_mvs(MACROBLOCKD *xd,
col_offset = this_mv.as_mv.col >> 3;
offset = ref_y_stride * row_offset + col_offset;
score = 0;
+#if !CONFIG_ABOVESPREFMV
if (xd->up_available) {
+#else
+ if (xd->up_available && xd->left_available) {
+#endif
vp9_sub_pixel_variance16x2(above_ref + offset, ref_y_stride,
SP(this_mv.as_mv.col),
SP(this_mv.as_mv.row),
diff --git a/vp9/encoder/vp9_encodeframe.c b/vp9/encoder/vp9_encodeframe.c
index c0fe5ac76..5271a597c 100644
--- a/vp9/encoder/vp9_encodeframe.c
+++ b/vp9/encoder/vp9_encodeframe.c
@@ -45,7 +45,7 @@
int enc_debug = 0;
#endif
-extern void select_interp_filter_type(VP9_COMP *cpi);
+void vp9_select_interp_filter_type(VP9_COMP *cpi);
static void encode_macroblock(VP9_COMP *cpi, TOKENEXTRA **t,
int output_enabled, int mb_row, int mb_col);
@@ -100,7 +100,7 @@ static unsigned int tt_activity_measure(VP9_COMP *cpi, MACROBLOCK *x) {
*/
act = vp9_variance16x16(x->src.y_buffer, x->src.y_stride, VP9_VAR_OFFS, 0,
&sse);
- act = act << 4;
+ act <<= 4;
/* If the region is flat, lower the activity some more. */
if (act < 8 << 12)
@@ -882,13 +882,10 @@ static void pick_sb64_modes(VP9_COMP *cpi,
/* Find best coding mode & reconstruct the MB so it is available
* as a predictor for MBs that follow in the SB */
if (cm->frame_type == KEY_FRAME) {
- vp9_rd_pick_intra_mode_sb64(cpi, x,
- totalrate,
- totaldist);
+ vp9_rd_pick_intra_mode_sb64(cpi, x, totalrate, totaldist);
/* Save the coding context */
- vpx_memcpy(&x->sb64_context.mic, xd->mode_info_context,
- sizeof(MODE_INFO));
+ vpx_memcpy(&x->sb64_context.mic, xd->mode_info_context, sizeof(MODE_INFO));
} else {
vp9_rd_pick_inter_mode_sb64(cpi, x, mb_row, mb_col, totalrate, totaldist);
}
@@ -1221,9 +1218,7 @@ static void init_encode_frame_mb_context(VP9_COMP *cpi) {
vpx_memset(cm->above_context, 0,
sizeof(ENTROPY_CONTEXT_PLANES) * cm->mb_cols);
- xd->fullpixel_mask = 0xffffffff;
- if (cm->full_pixel)
- xd->fullpixel_mask = 0xfffffff8;
+ xd->fullpixel_mask = cm->full_pixel ? 0xfffffff8 : 0xffffffff;
}
static void switch_lossless_mode(VP9_COMP *cpi, int lossless) {
@@ -1421,9 +1416,8 @@ static void set_txfm_flag(MODE_INFO *mi, int mis, int ymbs, int xmbs,
int x, y;
for (y = 0; y < ymbs; y++) {
- for (x = 0; x < xmbs; x++) {
+ for (x = 0; x < xmbs; x++)
mi[y * mis + x].mbmi.txfm_size = txfm_size;
- }
}
}
@@ -1678,7 +1672,7 @@ void vp9_encode_frame(VP9_COMP *cpi) {
// Update interpolation filter strategy for next frame.
if ((cpi->common.frame_type != KEY_FRAME) && (cpi->sf.search_best_filter))
- select_interp_filter_type(cpi);
+ vp9_select_interp_filter_type(cpi);
} else {
encode_frame_internal(cpi);
}
@@ -1690,27 +1684,23 @@ void vp9_setup_block_ptrs(MACROBLOCK *x) {
int i;
for (r = 0; r < 4; r++) {
- for (c = 0; c < 4; c++) {
+ for (c = 0; c < 4; c++)
x->block[r * 4 + c].src_diff = x->src_diff + r * 4 * 16 + c * 4;
- }
}
for (r = 0; r < 2; r++) {
- for (c = 0; c < 2; c++) {
+ for (c = 0; c < 2; c++)
x->block[16 + r * 2 + c].src_diff = x->src_diff + 256 + r * 4 * 8 + c * 4;
- }
}
for (r = 0; r < 2; r++) {
- for (c = 0; c < 2; c++) {
+ for (c = 0; c < 2; c++)
x->block[20 + r * 2 + c].src_diff = x->src_diff + 320 + r * 4 * 8 + c * 4;
- }
}
- for (i = 0; i < 24; i++) {
+ for (i = 0; i < 24; i++)
x->block[i].coeff = x->coeff + i * 16;
- }
}
void vp9_build_block_offsets(MACROBLOCK *x) {
diff --git a/vp9/encoder/vp9_encodeintra.c b/vp9/encoder/vp9_encodeintra.c
index 076e1ec0d..be9c224b3 100644
--- a/vp9/encoder/vp9_encodeintra.c
+++ b/vp9/encoder/vp9_encodeintra.c
@@ -17,8 +17,6 @@
#include "vp9/encoder/vp9_encodeintra.h"
int vp9_encode_intra(VP9_COMP *cpi, MACROBLOCK *x, int use_16x16_pred) {
- int i;
- int intra_pred_var = 0;
MB_MODE_INFO * mbmi = &x->e_mbd.mode_info_context->mbmi;
(void) cpi;
@@ -29,15 +27,15 @@ int vp9_encode_intra(VP9_COMP *cpi, MACROBLOCK *x, int use_16x16_pred) {
vp9_encode_intra16x16mby(x);
} else {
+ int i;
+
for (i = 0; i < 16; i++) {
x->e_mbd.block[i].bmi.as_mode.first = B_DC_PRED;
vp9_encode_intra4x4block(x, i);
}
}
- intra_pred_var = vp9_get_mb_ss(x->src_diff);
-
- return intra_pred_var;
+ return vp9_get_mb_ss(x->src_diff);
}
void vp9_encode_intra4x4block(MACROBLOCK *x, int ib) {
@@ -71,7 +69,6 @@ void vp9_encode_intra4x4mby(MACROBLOCK *mb) {
for (i = 0; i < 16; i++)
vp9_encode_intra4x4block(mb, i);
- return;
}
void vp9_encode_intra16x16mby(MACROBLOCK *x) {
@@ -83,24 +80,28 @@ void vp9_encode_intra16x16mby(MACROBLOCK *x) {
vp9_subtract_mby(x->src_diff, *(b->base_src), xd->predictor, b->src_stride);
- if (tx_size == TX_16X16) {
- vp9_transform_mby_16x16(x);
- vp9_quantize_mby_16x16(x);
- if (x->optimize)
- vp9_optimize_mby_16x16(x);
- vp9_inverse_transform_mby_16x16(xd);
- } else if (tx_size == TX_8X8) {
- vp9_transform_mby_8x8(x);
- vp9_quantize_mby_8x8(x);
- if (x->optimize)
- vp9_optimize_mby_8x8(x);
- vp9_inverse_transform_mby_8x8(xd);
- } else {
- vp9_transform_mby_4x4(x);
- vp9_quantize_mby_4x4(x);
- if (x->optimize)
- vp9_optimize_mby_4x4(x);
- vp9_inverse_transform_mby_4x4(xd);
+ switch (tx_size) {
+ case TX_16X16:
+ vp9_transform_mby_16x16(x);
+ vp9_quantize_mby_16x16(x);
+ if (x->optimize)
+ vp9_optimize_mby_16x16(x);
+ vp9_inverse_transform_mby_16x16(xd);
+ break;
+ case TX_8X8:
+ vp9_transform_mby_8x8(x);
+ vp9_quantize_mby_8x8(x);
+ if (x->optimize)
+ vp9_optimize_mby_8x8(x);
+ vp9_inverse_transform_mby_8x8(xd);
+ break;
+ default:
+ vp9_transform_mby_4x4(x);
+ vp9_quantize_mby_4x4(x);
+ if (x->optimize)
+ vp9_optimize_mby_4x4(x);
+ vp9_inverse_transform_mby_4x4(xd);
+ break;
}
vp9_recon_mby(xd);
@@ -115,19 +116,22 @@ void vp9_encode_intra16x16mbuv(MACROBLOCK *x) {
vp9_subtract_mbuv(x->src_diff, x->src.u_buffer, x->src.v_buffer,
xd->predictor, x->src.uv_stride);
- if (tx_size == TX_4X4) {
- vp9_transform_mbuv_4x4(x);
- vp9_quantize_mbuv_4x4(x);
- if (x->optimize)
- vp9_optimize_mbuv_4x4(x);
- vp9_inverse_transform_mbuv_4x4(xd);
- } else /* 16x16 or 8x8 */ {
- vp9_transform_mbuv_8x8(x);
- vp9_quantize_mbuv_8x8(x);
- if (x->optimize)
- vp9_optimize_mbuv_8x8(x);
- vp9_inverse_transform_mbuv_8x8(xd);
- }
+ switch (tx_size) {
+ case TX_4X4:
+ vp9_transform_mbuv_4x4(x);
+ vp9_quantize_mbuv_4x4(x);
+ if (x->optimize)
+ vp9_optimize_mbuv_4x4(x);
+ vp9_inverse_transform_mbuv_4x4(xd);
+ break;
+ default: // 16x16 or 8x8
+ vp9_transform_mbuv_8x8(x);
+ vp9_quantize_mbuv_8x8(x);
+ if (x->optimize)
+ vp9_optimize_mbuv_8x8(x);
+ vp9_inverse_transform_mbuv_8x8(xd);
+ break;
+ }
vp9_recon_intra_mbuv(xd);
}
@@ -190,16 +194,13 @@ void vp9_encode_intra8x8(MACROBLOCK *x, int ib) {
}
void vp9_encode_intra8x8mby(MACROBLOCK *x) {
- int i, ib;
+ int i;
- for (i = 0; i < 4; i++) {
- ib = vp9_i8x8_block[i];
- vp9_encode_intra8x8(x, ib);
- }
+ for (i = 0; i < 4; i++)
+ vp9_encode_intra8x8(x, vp9_i8x8_block[i]);
}
-static void encode_intra_uv4x4(MACROBLOCK *x, int ib,
- int mode) {
+static void encode_intra_uv4x4(MACROBLOCK *x, int ib, int mode) {
BLOCKD *b = &x->e_mbd.block[ib];
BLOCK *be = &x->block[ib];
@@ -216,17 +217,13 @@ static void encode_intra_uv4x4(MACROBLOCK *x, int ib,
}
void vp9_encode_intra8x8mbuv(MACROBLOCK *x) {
- int i, ib, mode;
- BLOCKD *b;
+ int i;
for (i = 0; i < 4; i++) {
- ib = vp9_i8x8_block[i];
- b = &x->e_mbd.block[ib];
- mode = b->bmi.as_mode.first;
-
- /*u */
- encode_intra_uv4x4(x, i + 16, mode);
- /*v */
- encode_intra_uv4x4(x, i + 20, mode);
+ BLOCKD *b = &x->e_mbd.block[vp9_i8x8_block[i]];
+ int mode = b->bmi.as_mode.first;
+
+ encode_intra_uv4x4(x, i + 16, mode); // u
+ encode_intra_uv4x4(x, i + 20, mode); // v
}
}
diff --git a/vp9/encoder/vp9_encodemb.c b/vp9/encoder/vp9_encodemb.c
index edb301895..62f1a2a30 100644
--- a/vp9/encoder/vp9_encodemb.c
+++ b/vp9/encoder/vp9_encodemb.c
@@ -29,9 +29,8 @@ void vp9_subtract_b_c(BLOCK *be, BLOCKD *bd, int pitch) {
int r, c;
for (r = 0; r < 4; r++) {
- for (c = 0; c < 4; c++) {
+ for (c = 0; c < 4; c++)
diff_ptr[c] = src_ptr[c] - pred_ptr[c];
- }
diff_ptr += pitch;
pred_ptr += pitch;
@@ -47,9 +46,9 @@ void vp9_subtract_4b_c(BLOCK *be, BLOCKD *bd, int pitch) {
int r, c;
for (r = 0; r < 8; r++) {
- for (c = 0; c < 8; c++) {
+ for (c = 0; c < 8; c++)
diff_ptr[c] = src_ptr[c] - pred_ptr[c];
- }
+
diff_ptr += pitch;
pred_ptr += pitch;
src_ptr += src_stride;
@@ -65,9 +64,8 @@ void vp9_subtract_mbuv_s_c(int16_t *diff, const uint8_t *usrc,
int r, c;
for (r = 0; r < 8; r++) {
- for (c = 0; c < 8; c++) {
+ for (c = 0; c < 8; c++)
udiff[c] = usrc[c] - upred[c];
- }
udiff += 8;
upred += dst_stride;
@@ -98,9 +96,8 @@ void vp9_subtract_mby_s_c(int16_t *diff, const uint8_t *src, int src_stride,
int r, c;
for (r = 0; r < 16; r++) {
- for (c = 0; c < 16; c++) {
+ for (c = 0; c < 16; c++)
diff[c] = src[c] - pred[c];
- }
diff += 16;
pred += dst_stride;
@@ -113,9 +110,8 @@ void vp9_subtract_sby_s_c(int16_t *diff, const uint8_t *src, int src_stride,
int r, c;
for (r = 0; r < 32; r++) {
- for (c = 0; c < 32; c++) {
+ for (c = 0; c < 32; c++)
diff[c] = src[c] - pred[c];
- }
diff += 32;
pred += dst_stride;
@@ -132,9 +128,8 @@ void vp9_subtract_sbuv_s_c(int16_t *diff, const uint8_t *usrc,
int r, c;
for (r = 0; r < 16; r++) {
- for (c = 0; c < 16; c++) {
+ for (c = 0; c < 16; c++)
udiff[c] = usrc[c] - upred[c];
- }
udiff += 16;
upred += dst_stride;
@@ -142,9 +137,8 @@ void vp9_subtract_sbuv_s_c(int16_t *diff, const uint8_t *usrc,
}
for (r = 0; r < 16; r++) {
- for (c = 0; c < 16; c++) {
+ for (c = 0; c < 16; c++)
vdiff[c] = vsrc[c] - vpred[c];
- }
vdiff += 16;
vpred += dst_stride;
@@ -176,12 +170,10 @@ void vp9_transform_mby_4x4(MACROBLOCK *x) {
if (tx_type != DCT_DCT) {
vp9_short_fht4x4(b->src_diff, b->coeff, 16, tx_type);
} else if (!(i & 1) && get_tx_type_4x4(xd, &xd->block[i + 1]) == DCT_DCT) {
- x->fwd_txm8x4(&x->block[i].src_diff[0],
- &x->block[i].coeff[0], 32);
+ x->fwd_txm8x4(x->block[i].src_diff, x->block[i].coeff, 32);
i++;
} else {
- x->fwd_txm4x4(&x->block[i].src_diff[0],
- &x->block[i].coeff[0], 32);
+ x->fwd_txm4x4(x->block[i].src_diff, x->block[i].coeff, 32);
}
}
}
@@ -189,10 +181,8 @@ void vp9_transform_mby_4x4(MACROBLOCK *x) {
void vp9_transform_mbuv_4x4(MACROBLOCK *x) {
int i;
- for (i = 16; i < 24; i += 2) {
- x->fwd_txm8x4(&x->block[i].src_diff[0],
- &x->block[i].coeff[0], 16);
- }
+ for (i = 16; i < 24; i += 2)
+ x->fwd_txm8x4(x->block[i].src_diff, x->block[i].coeff, 16);
}
static void transform_mb_4x4(MACROBLOCK *x) {
@@ -211,8 +201,7 @@ void vp9_transform_mby_8x8(MACROBLOCK *x) {
if (tx_type != DCT_DCT) {
vp9_short_fht8x8(b->src_diff, b->coeff, 16, tx_type);
} else {
- x->fwd_txm8x8(&x->block[i].src_diff[0],
- &x->block[i].coeff[0], 32);
+ x->fwd_txm8x8(x->block[i].src_diff, x->block[i].coeff, 32);
}
}
for (i = 2; i < 11; i += 8) {
@@ -221,8 +210,7 @@ void vp9_transform_mby_8x8(MACROBLOCK *x) {
if (tx_type != DCT_DCT) {
vp9_short_fht8x8(b->src_diff, (b + 2)->coeff, 16, tx_type);
} else {
- x->fwd_txm8x8(&x->block[i].src_diff[0],
- &x->block[i + 2].coeff[0], 32);
+ x->fwd_txm8x8(x->block[i].src_diff, x->block[i + 2].coeff, 32);
}
}
}
@@ -230,10 +218,8 @@ void vp9_transform_mby_8x8(MACROBLOCK *x) {
void vp9_transform_mbuv_8x8(MACROBLOCK *x) {
int i;
- for (i = 16; i < 24; i += 4) {
- x->fwd_txm8x8(&x->block[i].src_diff[0],
- &x->block[i].coeff[0], 16);
- }
+ for (i = 16; i < 24; i += 4)
+ x->fwd_txm8x8(x->block[i].src_diff, x->block[i].coeff, 16);
}
void vp9_transform_mb_8x8(MACROBLOCK *x) {
@@ -249,8 +235,7 @@ void vp9_transform_mby_16x16(MACROBLOCK *x) {
if (tx_type != DCT_DCT) {
vp9_short_fht16x16(b->src_diff, b->coeff, 16, tx_type);
} else {
- x->fwd_txm16x16(&x->block[0].src_diff[0],
- &x->block[0].coeff[0], 32);
+ x->fwd_txm16x16(x->block[0].src_diff, x->block[0].coeff, 32);
}
}
@@ -267,10 +252,8 @@ void vp9_transform_sby_32x32(MACROBLOCK *x) {
void vp9_transform_sbuv_16x16(MACROBLOCK *x) {
SUPERBLOCK * const x_sb = &x->sb_coeff_data;
vp9_clear_system_state();
- x->fwd_txm16x16(x_sb->src_diff + 1024,
- x_sb->coeff + 1024, 32);
- x->fwd_txm16x16(x_sb->src_diff + 1280,
- x_sb->coeff + 1280, 32);
+ x->fwd_txm16x16(x_sb->src_diff + 1024, x_sb->coeff + 1024, 32);
+ x->fwd_txm16x16(x_sb->src_diff + 1280, x_sb->coeff + 1280, 32);
}
#define RDTRUNC(RM,DM,R,D) ( (128+(R)*(RM)) & 0xFF )
diff --git a/vp9/encoder/vp9_mbgraph.c b/vp9/encoder/vp9_mbgraph.c
index d6644c2aa..121de653f 100644
--- a/vp9/encoder/vp9_mbgraph.c
+++ b/vp9/encoder/vp9_mbgraph.c
@@ -29,7 +29,7 @@ static unsigned int do_16x16_motion_iteration(VP9_COMP *cpi,
BLOCKD *d = &xd->block[0];
vp9_variance_fn_ptr_t v_fn_ptr = cpi->fn_ptr[BLOCK_16X16];
unsigned int best_err;
- int step_param;
+
int tmp_col_min = x->mv_col_min;
int tmp_col_max = x->mv_col_max;
@@ -38,11 +38,8 @@ static unsigned int do_16x16_motion_iteration(VP9_COMP *cpi,
int_mv ref_full;
// Further step/diamond searches as necessary
- if (cpi->Speed < 8) {
- step_param = cpi->sf.first_step + ((cpi->Speed > 5) ? 1 : 0);
- } else {
- step_param = cpi->sf.first_step + 2;
- }
+ int step_param = cpi->sf.first_step +
+ (cpi->Speed < 8 ? (cpi->Speed > 5 ? 1 : 0) : 2);
vp9_clamp_mv_min_max(x, ref_mv);
@@ -438,10 +435,7 @@ static void separate_arf_mbs(VP9_COMP *cpi) {
vpx_free(arf_not_zz);
}
-void vp9_update_mbgraph_stats
-(
- VP9_COMP *cpi
-) {
+void vp9_update_mbgraph_stats(VP9_COMP *cpi) {
VP9_COMMON *const cm = &cpi->common;
int i, n_frames = vp9_lookahead_depth(cpi->lookahead);
YV12_BUFFER_CONFIG *golden_ref =
diff --git a/vp9/encoder/vp9_mbgraph.h b/vp9/encoder/vp9_mbgraph.h
index db23eca33..c5bca4d01 100644
--- a/vp9/encoder/vp9_mbgraph.h
+++ b/vp9/encoder/vp9_mbgraph.h
@@ -11,6 +11,6 @@
#ifndef VP9_ENCODER_VP9_MBGRAPH_H_
#define VP9_ENCODER_VP9_MBGRAPH_H_
-extern void vp9_update_mbgraph_stats(VP9_COMP *cpi);
+void vp9_update_mbgraph_stats(VP9_COMP *cpi);
#endif // VP9_ENCODER_VP9_MBGRAPH_H_
diff --git a/vp9/encoder/vp9_mcomp.c b/vp9/encoder/vp9_mcomp.c
index 5287f9785..300d9f85c 100644
--- a/vp9/encoder/vp9_mcomp.c
+++ b/vp9/encoder/vp9_mcomp.c
@@ -8,14 +8,14 @@
* be found in the AUTHORS file in the root of the source tree.
*/
+#include <stdio.h>
+#include <limits.h>
+#include <math.h>
#include "vp9/encoder/vp9_onyx_int.h"
#include "vp9/encoder/vp9_mcomp.h"
#include "vpx_mem/vpx_mem.h"
#include "./vpx_config.h"
-#include <stdio.h>
-#include <limits.h>
-#include <math.h>
#include "vp9/common/vp9_findnearmv.h"
#include "vp9/common/vp9_common.h"
@@ -39,21 +39,20 @@ void vp9_clamp_mv_min_max(MACROBLOCK *x, int_mv *ref_mv) {
}
int vp9_mv_bit_cost(int_mv *mv, int_mv *ref, int *mvjcost, int *mvcost[2],
- int Weight, int ishp) {
+ int weight, int ishp) {
MV v;
- v.row = (mv->as_mv.row - ref->as_mv.row);
- v.col = (mv->as_mv.col - ref->as_mv.col);
+ v.row = mv->as_mv.row - ref->as_mv.row;
+ v.col = mv->as_mv.col - ref->as_mv.col;
return ((mvjcost[vp9_get_mv_joint(v)] +
- mvcost[0][v.row] + mvcost[1][v.col]) *
- Weight) >> 7;
+ mvcost[0][v.row] + mvcost[1][v.col]) * weight) >> 7;
}
static int mv_err_cost(int_mv *mv, int_mv *ref, int *mvjcost, int *mvcost[2],
int error_per_bit, int ishp) {
if (mvcost) {
MV v;
- v.row = (mv->as_mv.row - ref->as_mv.row);
- v.col = (mv->as_mv.col - ref->as_mv.col);
+ v.row = mv->as_mv.row - ref->as_mv.row;
+ v.col = mv->as_mv.col - ref->as_mv.col;
return ((mvjcost[vp9_get_mv_joint(v)] +
mvcost[0][v.row] + mvcost[1][v.col]) *
error_per_bit + 128) >> 8;
@@ -63,11 +62,10 @@ static int mv_err_cost(int_mv *mv, int_mv *ref, int *mvjcost, int *mvcost[2],
static int mvsad_err_cost(int_mv *mv, int_mv *ref, int *mvjsadcost,
int *mvsadcost[2], int error_per_bit) {
-
if (mvsadcost) {
MV v;
- v.row = (mv->as_mv.row - ref->as_mv.row);
- v.col = (mv->as_mv.col - ref->as_mv.col);
+ v.row = mv->as_mv.row - ref->as_mv.row;
+ v.col = mv->as_mv.col - ref->as_mv.col;
return ((mvjsadcost[vp9_get_mv_joint(v)] +
mvsadcost[0][v.row] + mvsadcost[1][v.col]) *
error_per_bit + 128) >> 8;
@@ -76,45 +74,39 @@ static int mvsad_err_cost(int_mv *mv, int_mv *ref, int *mvjsadcost,
}
void vp9_init_dsmotion_compensation(MACROBLOCK *x, int stride) {
- int Len;
+ int len;
int search_site_count = 0;
-
// Generate offsets for 4 search sites per step.
- Len = MAX_FIRST_STEP;
x->ss[search_site_count].mv.col = 0;
x->ss[search_site_count].mv.row = 0;
x->ss[search_site_count].offset = 0;
search_site_count++;
- while (Len > 0) {
-
+ for (len = MAX_FIRST_STEP; len > 0; len /= 2) {
// Compute offsets for search sites.
x->ss[search_site_count].mv.col = 0;
- x->ss[search_site_count].mv.row = -Len;
- x->ss[search_site_count].offset = -Len * stride;
+ x->ss[search_site_count].mv.row = -len;
+ x->ss[search_site_count].offset = -len * stride;
search_site_count++;
// Compute offsets for search sites.
x->ss[search_site_count].mv.col = 0;
- x->ss[search_site_count].mv.row = Len;
- x->ss[search_site_count].offset = Len * stride;
+ x->ss[search_site_count].mv.row = len;
+ x->ss[search_site_count].offset = len * stride;
search_site_count++;
// Compute offsets for search sites.
- x->ss[search_site_count].mv.col = -Len;
+ x->ss[search_site_count].mv.col = -len;
x->ss[search_site_count].mv.row = 0;
- x->ss[search_site_count].offset = -Len;
+ x->ss[search_site_count].offset = -len;
search_site_count++;
// Compute offsets for search sites.
- x->ss[search_site_count].mv.col = Len;
+ x->ss[search_site_count].mv.col = len;
x->ss[search_site_count].mv.row = 0;
- x->ss[search_site_count].offset = Len;
+ x->ss[search_site_count].offset = len;
search_site_count++;
-
- // Contract.
- Len /= 2;
}
x->ss_count = search_site_count;
@@ -122,68 +114,63 @@ void vp9_init_dsmotion_compensation(MACROBLOCK *x, int stride) {
}
void vp9_init3smotion_compensation(MACROBLOCK *x, int stride) {
- int Len;
+ int len;
int search_site_count = 0;
// Generate offsets for 8 search sites per step.
- Len = MAX_FIRST_STEP;
x->ss[search_site_count].mv.col = 0;
x->ss[search_site_count].mv.row = 0;
x->ss[search_site_count].offset = 0;
search_site_count++;
- while (Len > 0) {
-
+ for (len = MAX_FIRST_STEP; len > 0; len /= 2) {
// Compute offsets for search sites.
x->ss[search_site_count].mv.col = 0;
- x->ss[search_site_count].mv.row = -Len;
- x->ss[search_site_count].offset = -Len * stride;
+ x->ss[search_site_count].mv.row = -len;
+ x->ss[search_site_count].offset = -len * stride;
search_site_count++;
// Compute offsets for search sites.
x->ss[search_site_count].mv.col = 0;
- x->ss[search_site_count].mv.row = Len;
- x->ss[search_site_count].offset = Len * stride;
+ x->ss[search_site_count].mv.row = len;
+ x->ss[search_site_count].offset = len * stride;
search_site_count++;
// Compute offsets for search sites.
- x->ss[search_site_count].mv.col = -Len;
+ x->ss[search_site_count].mv.col = -len;
x->ss[search_site_count].mv.row = 0;
- x->ss[search_site_count].offset = -Len;
+ x->ss[search_site_count].offset = -len;
search_site_count++;
// Compute offsets for search sites.
- x->ss[search_site_count].mv.col = Len;
+ x->ss[search_site_count].mv.col = len;
x->ss[search_site_count].mv.row = 0;
- x->ss[search_site_count].offset = Len;
+ x->ss[search_site_count].offset = len;
search_site_count++;
// Compute offsets for search sites.
- x->ss[search_site_count].mv.col = -Len;
- x->ss[search_site_count].mv.row = -Len;
- x->ss[search_site_count].offset = -Len * stride - Len;
+ x->ss[search_site_count].mv.col = -len;
+ x->ss[search_site_count].mv.row = -len;
+ x->ss[search_site_count].offset = -len * stride - len;
search_site_count++;
// Compute offsets for search sites.
- x->ss[search_site_count].mv.col = Len;
- x->ss[search_site_count].mv.row = -Len;
- x->ss[search_site_count].offset = -Len * stride + Len;
+ x->ss[search_site_count].mv.col = len;
+ x->ss[search_site_count].mv.row = -len;
+ x->ss[search_site_count].offset = -len * stride + len;
search_site_count++;
// Compute offsets for search sites.
- x->ss[search_site_count].mv.col = -Len;
- x->ss[search_site_count].mv.row = Len;
- x->ss[search_site_count].offset = Len * stride - Len;
+ x->ss[search_site_count].mv.col = -len;
+ x->ss[search_site_count].mv.row = len;
+ x->ss[search_site_count].offset = len * stride - len;
search_site_count++;
// Compute offsets for search sites.
- x->ss[search_site_count].mv.col = Len;
- x->ss[search_site_count].mv.row = Len;
- x->ss[search_site_count].offset = Len * stride + Len;
+ x->ss[search_site_count].mv.col = len;
+ x->ss[search_site_count].mv.row = len;
+ x->ss[search_site_count].offset = len * stride + len;
search_site_count++;
-
- // Contract.
- Len /= 2;
}
x->ss_count = search_site_count;
@@ -2018,12 +2005,10 @@ int vp9_refining_search_sadx4(MACROBLOCK *x, BLOCK *b, BLOCKD *d,
for (i = 0; i < search_range; i++) {
int best_site = -1;
- int all_in = 1;
-
- all_in &= ((ref_mv->as_mv.row - 1) > x->mv_row_min);
- all_in &= ((ref_mv->as_mv.row + 1) < x->mv_row_max);
- all_in &= ((ref_mv->as_mv.col - 1) > x->mv_col_min);
- all_in &= ((ref_mv->as_mv.col + 1) < x->mv_col_max);
+ int all_in = ((ref_mv->as_mv.row - 1) > x->mv_row_min) &
+ ((ref_mv->as_mv.row + 1) < x->mv_row_max) &
+ ((ref_mv->as_mv.col - 1) > x->mv_col_min) &
+ ((ref_mv->as_mv.col + 1) < x->mv_col_max);
if (all_in) {
unsigned int sad_array[4];
diff --git a/vp9/encoder/vp9_mcomp.h b/vp9/encoder/vp9_mcomp.h
index 06acdbe58..2479d7235 100644
--- a/vp9/encoder/vp9_mcomp.h
+++ b/vp9/encoder/vp9_mcomp.h
@@ -24,11 +24,12 @@ void print_mode_context(VP9_COMMON *pc);
#define MAX_FULL_PEL_VAL ((1 << (MAX_MVSEARCH_STEPS)) - 1) // Max full pel mv specified in 1 pel units
#define MAX_FIRST_STEP (1 << (MAX_MVSEARCH_STEPS-1)) // Maximum size of the first step in full pel units
-extern void vp9_clamp_mv_min_max(MACROBLOCK *x, int_mv *ref_mv);
-extern int vp9_mv_bit_cost(int_mv *mv, int_mv *ref, int *mvjcost,
- int *mvcost[2], int Weight, int ishp);
-extern void vp9_init_dsmotion_compensation(MACROBLOCK *x, int stride);
-extern void vp9_init3smotion_compensation(MACROBLOCK *x, int stride);
+void vp9_clamp_mv_min_max(MACROBLOCK *x, int_mv *ref_mv);
+int vp9_mv_bit_cost(int_mv *mv, int_mv *ref, int *mvjcost,
+ int *mvcost[2], int weight, int ishp);
+void vp9_init_dsmotion_compensation(MACROBLOCK *x, int stride);
+void vp9_init3smotion_compensation(MACROBLOCK *x, int stride);
+
// Runs sequence of diamond searches in smaller steps for RD
struct VP9_COMP;
int vp9_full_pixel_diamond(struct VP9_COMP *cpi, MACROBLOCK *x, BLOCK *b,
@@ -37,20 +38,13 @@ int vp9_full_pixel_diamond(struct VP9_COMP *cpi, MACROBLOCK *x, BLOCK *b,
vp9_variance_fn_ptr_t *fn_ptr,
int_mv *ref_mv, int_mv *dst_mv);
-extern int vp9_hex_search
-(
- MACROBLOCK *x,
- BLOCK *b,
- BLOCKD *d,
- int_mv *ref_mv,
- int_mv *best_mv,
- int search_param,
- int error_per_bit,
- const vp9_variance_fn_ptr_t *vf,
- int *mvjsadcost, int *mvsadcost[2],
- int *mvjcost, int *mvcost[2],
- int_mv *center_mv
-);
+int vp9_hex_search(MACROBLOCK *x, BLOCK *b, BLOCKD *d,
+ int_mv *ref_mv, int_mv *best_mv,
+ int search_param, int error_per_bit,
+ const vp9_variance_fn_ptr_t *vf,
+ int *mvjsadcost, int *mvsadcost[2],
+ int *mvjcost, int *mvcost[2],
+ int_mv *center_mv);
typedef int (fractional_mv_step_fp) (MACROBLOCK *x, BLOCK *b, BLOCKD *d, int_mv
*bestmv, int_mv *ref_mv, int error_per_bit, const vp9_variance_fn_ptr_t *vfp,
diff --git a/vp9/encoder/vp9_onyx_if.c b/vp9/encoder/vp9_onyx_if.c
index b9d3049f2..95270ecd5 100644
--- a/vp9/encoder/vp9_onyx_if.c
+++ b/vp9/encoder/vp9_onyx_if.c
@@ -2538,7 +2538,7 @@ static void loopfilter_frame(VP9_COMP *cpi, VP9_COMMON *cm) {
}
-void select_interp_filter_type(VP9_COMP *cpi) {
+void vp9_select_interp_filter_type(VP9_COMP *cpi) {
int i;
int high_filter_index = 0;
unsigned int thresh;
diff --git a/vp9/encoder/x86/vp9_dct_sse2.c b/vp9/encoder/x86/vp9_dct_sse2_intrinsics.c
index ff884d999..ff884d999 100644
--- a/vp9/encoder/x86/vp9_dct_sse2.c
+++ b/vp9/encoder/x86/vp9_dct_sse2_intrinsics.c
diff --git a/vp9/vp9cx.mk b/vp9/vp9cx.mk
index 22fa565e5..43dba1373 100644
--- a/vp9/vp9cx.mk
+++ b/vp9/vp9cx.mk
@@ -110,10 +110,10 @@ VP9_CX_SRCS-$(ARCH_X86)$(ARCH_X86_64) += encoder/x86/vp9_quantize_mmx.asm
VP9_CX_SRCS-$(ARCH_X86)$(ARCH_X86_64) += encoder/x86/vp9_encodeopt.asm
VP9_CX_SRCS-$(ARCH_X86_64) += encoder/x86/vp9_ssim_opt.asm
-VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_dct_sse2.c
+VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_dct_sse2_intrinsics.c
ifeq ($(HAVE_SSE2),yes)
-vp9/encoder/x86/vp9_dct_sse2.c.d: CFLAGS += -msse2
-vp9/encoder/x86/vp9_dct_sse2.c.o: CFLAGS += -msse2
+vp9/encoder/x86/vp9_dct_sse2_intrinsics.c.d: CFLAGS += -msse2
+vp9/encoder/x86/vp9_dct_sse2_intrinsics.c.o: CFLAGS += -msse2
endif
generated by cgit v1.2.3 (git 2.25.1) at 2024-10-13 15:27:08 +0000