ref: 97096f5fa355a27cd7262bf05f05156e40090c03
parent: 7906ed091ab5d16a7fbdc21109a034644aa6b9a3
parent: f72fdf1c7f443b8b62559279e55b439b0814e58a
author: Ronald S. Bultje <rbultje@google.com>
date: Mon Oct 22 08:54:39 EDT 2012
Merge changes I02e7f64a,Ide954b00,Idc8b5977 into experimental * changes: Fix another typo in 4x4-transform-for-i8x8-intra-pred coeff contexts. 8x8 transform support in splitmv. Use SPLITMV_PARTITIONING instead of a plain integer type.
--- a/vp8/common/blockd.h
+++ b/vp8/common/blockd.h
@@ -175,6 +175,14 @@
#define VP8_BINTRAMODES (B_HU_PRED + 1) /* 10 */
#define VP8_SUBMVREFS (1 + NEW4X4 - LEFT4X4)
+typedef enum {+ PARTITIONING_16X8 = 0,
+ PARTITIONING_8X16,
+ PARTITIONING_8X8,
+ PARTITIONING_4X4,
+ NB_PARTITIONINGS,
+} SPLITMV_PARTITIONING_TYPE;
+
/* For keyframes, intra block modes are predicted by the (already decoded)
modes for the Y blocks to the left and above us; for interframes, there
is a single probability table. */
@@ -216,7 +224,7 @@
int mv_ref_index[MAX_REF_FRAMES];
#endif
- unsigned char partitioning;
+ SPLITMV_PARTITIONING_TYPE partitioning;
unsigned char mb_skip_coeff; /* does this mb has coefficients at all, 1=no coefficients, 0=need decode tokens */
unsigned char need_to_clamp_mvs;
unsigned char need_to_clamp_secondmv;
--- a/vp8/common/entropymode.c
+++ b/vp8/common/entropymode.c
@@ -215,9 +215,9 @@
};
const vp8_tree_index vp8_mbsplit_tree[6] = {- -3, 2,
- -2, 4,
- -0, -1
+ -PARTITIONING_4X4, 2,
+ -PARTITIONING_8X8, 4,
+ -PARTITIONING_16X8, -PARTITIONING_8X16,
};
const vp8_tree_index vp8_mv_ref_tree[8] = {--- a/vp8/common/postproc.c
+++ b/vp8/common/postproc.c
@@ -783,7 +783,7 @@
if (mi->mbmi.mode == SPLITMV) { switch (mi->mbmi.partitioning) {- case 0 : { /* mv_top_bottom */+ case PARTITIONING_16X8 : { /* mv_top_bottom */union b_mode_info *bmi = &mi->bmi[0];
MV *mv = &bmi->mv.as_mv;
@@ -803,7 +803,7 @@
break;
}
- case 1 : { /* mv_left_right */+ case PARTITIONING_8X16 : { /* mv_left_right */union b_mode_info *bmi = &mi->bmi[0];
MV *mv = &bmi->mv.as_mv;
@@ -823,7 +823,7 @@
break;
}
- case 2 : { /* mv_quarters */+ case PARTITIONING_8X8 : { /* mv_quarters */union b_mode_info *bmi = &mi->bmi[0];
MV *mv = &bmi->mv.as_mv;
@@ -858,6 +858,7 @@
vp8_blit_line(x0 + 12, x1, y0 + 12, y1, y_buffer, y_stride);
break;
}
+ case PARTITIONING_4X4:
default : {union b_mode_info *bmi = mi->bmi;
int bx0, by0;
--- a/vp8/common/reconinter.c
+++ b/vp8/common/reconinter.c
@@ -965,7 +965,7 @@
MB_MODE_INFO * mbmi = &xd->mode_info_context->mbmi;
BLOCKD *blockd = xd->block;
- if (xd->mode_info_context->mbmi.partitioning < 3) {+ if (xd->mode_info_context->mbmi.partitioning != PARTITIONING_4X4) {blockd[ 0].bmi = xd->mode_info_context->bmi[ 0];
blockd[ 2].bmi = xd->mode_info_context->bmi[ 2];
blockd[ 8].bmi = xd->mode_info_context->bmi[ 8];
--- a/vp8/decoder/decodemv.c
+++ b/vp8/decoder/decodemv.c
@@ -1285,10 +1285,12 @@
#if CONFIG_TX_SELECT
if (cm->txfm_mode == TX_MODE_SELECT && mbmi->mb_skip_coeff == 0 &&
((mbmi->ref_frame == INTRA_FRAME && mbmi->mode <= I8X8_PRED) ||
- (mbmi->ref_frame != INTRA_FRAME && mbmi->mode != SPLITMV))) {+ (mbmi->ref_frame != INTRA_FRAME && !(mbmi->mode == SPLITMV &&
+ mbmi->partitioning == PARTITIONING_4X4)))) {// FIXME(rbultje) code ternary symbol once all experiments are merged
mbmi->txfm_size = vp8_read(bc, cm->prob_tx[0]);
- if (mbmi->txfm_size != TX_4X4 && mbmi->mode != I8X8_PRED)
+ if (mbmi->txfm_size != TX_4X4 && mbmi->mode != I8X8_PRED &&
+ mbmi->mode != SPLITMV)
mbmi->txfm_size += vp8_read(bc, cm->prob_tx[1]);
} else
#endif
@@ -1297,8 +1299,9 @@
(mbmi->ref_frame != INTRA_FRAME && mbmi->mode != SPLITMV))) {mbmi->txfm_size = TX_16X16;
} else if (cm->txfm_mode >= ALLOW_8X8 &&
- ((mbmi->ref_frame == INTRA_FRAME && mbmi->mode != B_PRED) ||
- (mbmi->ref_frame != INTRA_FRAME && mbmi->mode != SPLITMV))) {+ (!(mbmi->ref_frame == INTRA_FRAME && mbmi->mode == B_PRED) &&
+ !(mbmi->ref_frame != INTRA_FRAME && mbmi->mode == SPLITMV &&
+ mbmi->partitioning == PARTITIONING_4X4))) {mbmi->txfm_size = TX_8X8;
} else {mbmi->txfm_size = TX_4X4;
--- a/vp8/decoder/decodframe.c
+++ b/vp8/decoder/decodframe.c
@@ -384,10 +384,16 @@
}
}
} else if (mode == SPLITMV) {- DEQUANT_INVOKE(&pbi->dequant, idct_add_y_block)
- (xd->qcoeff, xd->block[0].dequant,
- xd->predictor, xd->dst.y_buffer,
- xd->dst.y_stride, xd->eobs);
+ if (tx_size == TX_8X8) {+ vp8_dequant_idct_add_y_block_8x8_c(xd->qcoeff, xd->block[0].dequant,
+ xd->predictor, xd->dst.y_buffer,
+ xd->dst.y_stride, xd->eobs, xd);
+ } else {+ DEQUANT_INVOKE(&pbi->dequant,
+ idct_add_y_block)(xd->qcoeff, xd->block[0].dequant,
+ xd->predictor, xd->dst.y_buffer,
+ xd->dst.y_stride, xd->eobs);
+ }
} else {BLOCKD *b = &xd->block[24];
@@ -489,8 +495,10 @@
if (!xd->mode_info_context->mbmi.encoded_as_sb) {#endif
if ((tx_size == TX_8X8 &&
- xd->mode_info_context->mbmi.mode != I8X8_PRED)
- || tx_size == TX_16X16)
+ xd->mode_info_context->mbmi.mode != I8X8_PRED &&
+ xd->mode_info_context->mbmi.mode != SPLITMV)
+ || tx_size == TX_16X16
+ )
DEQUANT_INVOKE(&pbi->dequant, idct_add_uv_block_8x8) //
(xd->qcoeff + 16 * 16, xd->block[16].dequant,
xd->predictor + 16 * 16, xd->dst.u_buffer, xd->dst.v_buffer,
--- a/vp8/decoder/detokenize.c
+++ b/vp8/decoder/detokenize.c
@@ -493,7 +493,8 @@
INT16 *qcoeff_ptr = &xd->qcoeff[0];
TX_TYPE tx_type = DCT_DCT;
- int bufthred = (xd->mode_info_context->mbmi.mode == I8X8_PRED) ? 16 : 24;
+ int bufthred = (xd->mode_info_context->mbmi.mode == I8X8_PRED ||
+ xd->mode_info_context->mbmi.mode == SPLITMV) ? 16 : 24;
if (xd->mode_info_context->mbmi.mode != B_PRED &&
xd->mode_info_context->mbmi.mode != SPLITMV &&
xd->mode_info_context->mbmi.mode != I8X8_PRED) {--- a/vp8/encoder/bitstream.c
+++ b/vp8/encoder/bitstream.c
@@ -1219,7 +1219,8 @@
#if CONFIG_TX_SELECT
if (((rf == INTRA_FRAME && mode <= I8X8_PRED) ||
- (rf != INTRA_FRAME && mode != SPLITMV)) &&
+ (rf != INTRA_FRAME && !(mode == SPLITMV &&
+ mi->partitioning == PARTITIONING_4X4))) &&
pc->txfm_mode == TX_MODE_SELECT &&
!((pc->mb_no_coeff_skip && mi->mb_skip_coeff) ||
(segfeature_active(xd, segment_id, SEG_LVL_EOB) &&
@@ -1227,7 +1228,7 @@
TX_SIZE sz = mi->txfm_size;
// FIXME(rbultje) code ternary symbol once all experiments are merged
vp8_write(bc, sz != TX_4X4, pc->prob_tx[0]);
- if (sz != TX_4X4 && mode != I8X8_PRED)
+ if (sz != TX_4X4 && mode != I8X8_PRED && mode != SPLITMV)
vp8_write(bc, sz != TX_8X8, pc->prob_tx[1]);
}
#endif
--- a/vp8/encoder/encodeframe.c
+++ b/vp8/encoder/encodeframe.c
@@ -2161,7 +2161,9 @@
if (mbmi->mode != B_PRED && mbmi->mode != I8X8_PRED &&
mbmi->mode != SPLITMV) {cpi->txfm_count[mbmi->txfm_size]++;
- } else if (mbmi->mode == I8X8_PRED) {+ } else if (mbmi->mode == I8X8_PRED ||
+ (mbmi->mode == SPLITMV &&
+ mbmi->partitioning != PARTITIONING_4X4)) {cpi->txfm_count_8x8p[mbmi->txfm_size]++;
}
} else
@@ -2169,8 +2171,10 @@
if (mbmi->mode != B_PRED && mbmi->mode != I8X8_PRED &&
mbmi->mode != SPLITMV && cpi->common.txfm_mode >= ALLOW_16X16) {mbmi->txfm_size = TX_16X16;
- } else if (mbmi->mode != B_PRED && mbmi->mode != SPLITMV &&
- cpi->common.txfm_mode >= ALLOW_8X8) {+ } else if (mbmi->mode != B_PRED &&
+ !(mbmi->mode == SPLITMV &&
+ mbmi->partitioning == PARTITIONING_4X4) &&
+ cpi->common.txfm_mode >= ALLOW_8X8) {mbmi->txfm_size = TX_8X8;
} else {mbmi->txfm_size = TX_4X4;
--- a/vp8/encoder/encodemb.c
+++ b/vp8/encoder/encodemb.c
@@ -636,6 +636,7 @@
ENTROPY_CONTEXT_PLANES t_above, t_left;
ENTROPY_CONTEXT *ta;
ENTROPY_CONTEXT *tl;
+ int has_2nd_order = x->e_mbd.mode_info_context->mbmi.mode != SPLITMV;
if (!x->e_mbd.above_context || !x->e_mbd.left_context)
return;
@@ -645,7 +646,7 @@
ta = (ENTROPY_CONTEXT *)&t_above;
tl = (ENTROPY_CONTEXT *)&t_left;
- type = PLANE_TYPE_Y_NO_DC;
+ type = has_2nd_order ? PLANE_TYPE_Y_NO_DC : PLANE_TYPE_Y_WITH_DC;
for (b = 0; b < 16; b += 4) {optimize_b(x, b, type,
ta + vp8_block2above_8x8[b], tl + vp8_block2left_8x8[b],
@@ -655,8 +656,11 @@
}
// 8x8 always have 2nd roder haar block
- check_reset_8x8_2nd_coeffs(&x->e_mbd,
- ta + vp8_block2above_8x8[24], tl + vp8_block2left_8x8[24]);
+ if (has_2nd_order) {+ check_reset_8x8_2nd_coeffs(&x->e_mbd,
+ ta + vp8_block2above_8x8[24],
+ tl + vp8_block2left_8x8[24]);
+ }
}
void vp8_optimize_mbuv_8x8(MACROBLOCK *x, const VP8_ENCODER_RTCD *rtcd) {@@ -896,11 +900,25 @@
optimize_mb_16x16(x, rtcd);
vp8_inverse_transform_mb_16x16(IF_RTCD(&rtcd->common->idct), xd);
} else if (tx_size == TX_8X8) {- vp8_transform_mb_8x8(x);
- vp8_quantize_mb_8x8(x);
- if (x->optimize)
- optimize_mb_8x8(x, rtcd);
- vp8_inverse_transform_mb_8x8(IF_RTCD(&rtcd->common->idct), xd);
+ if (xd->mode_info_context->mbmi.mode == SPLITMV) {+ assert(xd->mode_info_context->mbmi.partitioning != PARTITIONING_4X4);
+ vp8_transform_mby_8x8(x);
+ vp8_transform_mbuv_4x4(x);
+ vp8_quantize_mby_8x8(x);
+ vp8_quantize_mbuv_4x4(x);
+ if (x->optimize) {+ vp8_optimize_mby_8x8(x, rtcd);
+ vp8_optimize_mbuv_4x4(x, rtcd);
+ }
+ vp8_inverse_transform_mby_8x8(IF_RTCD(&rtcd->common->idct), xd);
+ vp8_inverse_transform_mbuv_4x4(IF_RTCD(&rtcd->common->idct), xd);
+ } else {+ vp8_transform_mb_8x8(x);
+ vp8_quantize_mb_8x8(x);
+ if (x->optimize)
+ optimize_mb_8x8(x, rtcd);
+ vp8_inverse_transform_mb_8x8(IF_RTCD(&rtcd->common->idct), xd);
+ }
} else {transform_mb_4x4(x);
vp8_quantize_mb_4x4(x);
--- a/vp8/encoder/onyx_int.h
+++ b/vp8/encoder/onyx_int.h
@@ -366,10 +366,10 @@
} VP8_ENCODER_RTCD;
enum {- BLOCK_16X8,
- BLOCK_8X16,
- BLOCK_8X8,
- BLOCK_4X4,
+ BLOCK_16X8 = PARTITIONING_16X8,
+ BLOCK_8X16 = PARTITIONING_8X16,
+ BLOCK_8X8 = PARTITIONING_8X8,
+ BLOCK_4X4 = PARTITIONING_4X4,
BLOCK_16X16,
BLOCK_MAX_SEGMENTS,
BLOCK_32X32 = BLOCK_MAX_SEGMENTS,
--- a/vp8/encoder/rdopt.c
+++ b/vp8/encoder/rdopt.c
@@ -1583,8 +1583,8 @@
} else {*(a + vp8_block2above[ib]) = besta0;
*(a + vp8_block2above[ib + 1]) = besta1;
- *(l + vp8_block2above[ib]) = bestl0;
- *(l + vp8_block2above[ib + 4]) = bestl1;
+ *(l + vp8_block2left[ib]) = bestl0;
+ *(l + vp8_block2left[ib + 4]) = bestl1;
}
return best_rd;
@@ -2178,6 +2178,7 @@
int which_label,
int *labelyrate,
int *distortion,
+ int64_t *otherrd,
ENTROPY_CONTEXT *ta,
ENTROPY_CONTEXT *tl,
const VP8_ENCODER_RTCD *rtcd) {@@ -2184,7 +2185,16 @@
int i, j;
MACROBLOCKD *xd = &x->e_mbd;
const int iblock[4] = { 0, 1, 4, 5 };+ int othercost = 0, otherdist = 0;
+ ENTROPY_CONTEXT_PLANES tac, tlc;
+ ENTROPY_CONTEXT *tacp = (ENTROPY_CONTEXT *) &tac,
+ *tlcp = (ENTROPY_CONTEXT *) &tlc;
+ if (otherrd) {+ memcpy(&tac, ta, sizeof(ENTROPY_CONTEXT_PLANES));
+ memcpy(&tlc, tl, sizeof(ENTROPY_CONTEXT_PLANES));
+ }
+
*distortion = 0;
*labelyrate = 0;
for (i = 0; i < 4; i++) {@@ -2191,8 +2201,9 @@
int ib = vp8_i8x8_block[i];
if (labels[ib] == which_label) {- BLOCKD *bd = &xd->block[ib];
- BLOCK *be = &x->block[ib];
+ int idx = (ib & 8) + ((ib & 2) << 1);
+ BLOCKD *bd = &xd->block[ib], *bd2 = &xd->block[idx];
+ BLOCK *be = &x->block[ib], *be2 = &x->block[idx];
int thisdistortion;
vp8_build_inter_predictors4b(xd, bd, 16);
@@ -2200,24 +2211,66 @@
vp8_build_2nd_inter_predictors4b(xd, bd, 16);
vp8_subtract_4b_c(be, bd, 16);
- for (j = 0; j < 4; j += 2) {- bd = &xd->block[ib + iblock[j]];
- be = &x->block[ib + iblock[j]];
- x->vp8_short_fdct8x4(be->src_diff, be->coeff, 32);
- x->quantize_b_4x4_pair(be, be + 1, bd, bd + 1);
- thisdistortion = vp8_block_error_c(be->coeff, bd->dqcoeff, 32);
+ if (xd->mode_info_context->mbmi.txfm_size == TX_4X4) {+ if (otherrd) {+ x->vp8_short_fdct8x8(be->src_diff, be2->coeff, 32);
+ x->quantize_b_8x8(be2, bd2);
+ thisdistortion = vp8_block_error_c(be2->coeff, bd2->dqcoeff, 64);
+ otherdist += thisdistortion;
+ othercost += cost_coeffs(x, bd2, PLANE_TYPE_Y_WITH_DC,
+ tacp + vp8_block2above_8x8[idx],
+ tlcp + vp8_block2left_8x8[idx], TX_8X8);
+ }
+ for (j = 0; j < 4; j += 2) {+ bd = &xd->block[ib + iblock[j]];
+ be = &x->block[ib + iblock[j]];
+ x->vp8_short_fdct8x4(be->src_diff, be->coeff, 32);
+ x->quantize_b_4x4_pair(be, be + 1, bd, bd + 1);
+ thisdistortion = vp8_block_error_c(be->coeff, bd->dqcoeff, 32);
+ *distortion += thisdistortion;
+ *labelyrate += cost_coeffs(x, bd, PLANE_TYPE_Y_WITH_DC,
+ ta + vp8_block2above[ib + iblock[j]],
+ tl + vp8_block2left[ib + iblock[j]],
+ TX_4X4);
+ *labelyrate += cost_coeffs(x, bd + 1, PLANE_TYPE_Y_WITH_DC,
+ ta + vp8_block2above[ib + iblock[j] + 1],
+ tl + vp8_block2left[ib + iblock[j]],
+ TX_4X4);
+ }
+ } else /* 8x8 */ {+ if (otherrd) {+ for (j = 0; j < 4; j += 2) {+ BLOCKD *bd3 = &xd->block[ib + iblock[j]];
+ BLOCK *be3 = &x->block[ib + iblock[j]];
+ x->vp8_short_fdct8x4(be3->src_diff, be3->coeff, 32);
+ x->quantize_b_4x4_pair(be3, be3 + 1, bd3, bd3 + 1);
+ thisdistortion = vp8_block_error_c(be3->coeff, bd3->dqcoeff, 32);
+ otherdist += thisdistortion;
+ othercost += cost_coeffs(x, bd3, PLANE_TYPE_Y_WITH_DC,
+ tacp + vp8_block2above[ib + iblock[j]],
+ tlcp + vp8_block2left[ib + iblock[j]],
+ TX_4X4);
+ othercost += cost_coeffs(x, bd3 + 1, PLANE_TYPE_Y_WITH_DC,
+ tacp + vp8_block2above[ib + iblock[j] + 1],
+ tlcp + vp8_block2left[ib + iblock[j]],
+ TX_4X4);
+ }
+ }
+ x->vp8_short_fdct8x8(be->src_diff, be2->coeff, 32);
+ x->quantize_b_8x8(be2, bd2);
+ thisdistortion = vp8_block_error_c(be2->coeff, bd2->dqcoeff, 64);
*distortion += thisdistortion;
- *labelyrate += cost_coeffs(x, bd, PLANE_TYPE_Y_WITH_DC,
- ta + vp8_block2above[ib + iblock[j]],
- tl + vp8_block2left[ib + iblock[j]], TX_4X4);
- *labelyrate += cost_coeffs(x, bd + 1, PLANE_TYPE_Y_WITH_DC,
- ta + vp8_block2above[ib + iblock[j] + 1],
- tl + vp8_block2left[ib + iblock[j]],
- TX_4X4);
+ *labelyrate += cost_coeffs(x, bd2, PLANE_TYPE_Y_WITH_DC,
+ ta + vp8_block2above_8x8[idx],
+ tl + vp8_block2left_8x8[idx], TX_8X8);
}
}
}
*distortion >>= 2;
+ if (otherrd) {+ othercost >>= 2;
+ *otherrd = RDCOST(x->rdmult, x->rddiv, othercost, otherdist);
+ }
return RDCOST(x->rdmult, x->rddiv, *labelyrate, *distortion);
}
@@ -2229,7 +2282,8 @@
int_mv mvp;
int64_t segment_rd;
- int segment_num;
+ SPLITMV_PARTITIONING_TYPE segment_num;
+ TX_SIZE txfm_size;
int r;
int d;
int segment_yrate;
@@ -2255,9 +2309,14 @@
return r;
}
-static void rd_check_segment(VP8_COMP *cpi, MACROBLOCK *x,
- BEST_SEG_INFO *bsi, unsigned int segmentation,
- int_mv seg_mvs[16 /* n_blocks */][MAX_REF_FRAMES - 1]) {+static void rd_check_segment_txsize(VP8_COMP *cpi, MACROBLOCK *x,
+ BEST_SEG_INFO *bsi,
+ SPLITMV_PARTITIONING_TYPE segmentation,
+ TX_SIZE tx_size, int64_t *otherrds,
+ int64_t *rds, int *completed,
+ /* 16 = n_blocks */
+ int_mv seg_mvs[16 /* n_blocks */]
+ [MAX_REF_FRAMES - 1]) {int i, j;
int const *labels;
int br = 0, bd = 0;
@@ -2265,12 +2324,12 @@
MB_MODE_INFO * mbmi = &x->e_mbd.mode_info_context->mbmi;
int label_count;
- int64_t this_segment_rd = 0;
+ int64_t this_segment_rd = 0, other_segment_rd;
int label_mv_thresh;
int rate = 0;
int sbr = 0, sbd = 0;
int segmentyrate = 0;
- uint8_t best_eobs[16];
+ uint8_t best_eobs[16] = { 0 };vp8_variance_fn_ptr_t *v_fn_ptr;
@@ -2298,20 +2357,23 @@
label_mv_thresh = 1 * bsi->mvthresh / label_count;
// Segmentation method overheads
- rate = vp8_cost_token(vp8_mbsplit_tree, vp8_mbsplit_probs, vp8_mbsplit_encodings + segmentation);
+ rate = vp8_cost_token(vp8_mbsplit_tree, vp8_mbsplit_probs,
+ vp8_mbsplit_encodings + segmentation);
rate += vp8_cost_mv_ref(cpi, SPLITMV, bsi->mdcounts);
this_segment_rd += RDCOST(x->rdmult, x->rddiv, rate, 0);
br += rate;
+ other_segment_rd = this_segment_rd;
- for (i = 0; i < label_count; i++) {+ mbmi->txfm_size = tx_size;
+ for (i = 0; i < label_count && this_segment_rd < bsi->segment_rd; i++) {int_mv mode_mv[B_MODE_COUNT], second_mode_mv[B_MODE_COUNT];
- int64_t best_label_rd = INT64_MAX;
+ int64_t best_label_rd = INT64_MAX, best_other_rd = INT64_MAX;
B_PREDICTION_MODE mode_selected = ZERO4X4;
int bestlabelyrate = 0;
// search for the best motion vector on this segment
for (this_mode = LEFT4X4; this_mode <= NEW4X4; this_mode ++) {- int64_t this_rd;
+ int64_t this_rd, other_rd;
int distortion;
int labelyrate;
ENTROPY_CONTEXT_PLANES t_above_s, t_left_s;
@@ -2333,14 +2395,16 @@
BLOCK *c;
BLOCKD *e;
- // Is the best so far sufficiently good that we cant justify doing and new motion search.
+ /* Is the best so far sufficiently good that we cant justify doing
+ * and new motion search. */
if (best_label_rd < label_mv_thresh)
break;
if (cpi->compressor_speed) {- if (segmentation == BLOCK_8X16 || segmentation == BLOCK_16X8) {+ if (segmentation == PARTITIONING_8X16 ||
+ segmentation == PARTITIONING_16X8) {bsi->mvp.as_int = bsi->sv_mvp[i].as_int;
- if (i == 1 && segmentation == BLOCK_16X8)
+ if (i == 1 && segmentation == PARTITIONING_16X8)
bsi->mvp.as_int = bsi->sv_mvp[2].as_int;
step_param = bsi->sv_istep[i];
@@ -2347,7 +2411,7 @@
}
// use previous block's result as next block's MV predictor.
- if (segmentation == BLOCK_4X4 && i > 0) {+ if (segmentation == PARTITIONING_4X4 && i > 0) {bsi->mvp.as_int = x->e_mbd.block[i - 1].bmi.as_mv.first.as_int;
if (i == 4 || i == 8 || i == 12)
bsi->mvp.as_int = x->e_mbd.block[i - 4].bmi.as_mv.first.as_int;
@@ -2379,7 +2443,8 @@
// Should we do a full search (best quality only)
if ((cpi->compressor_speed == 0) && (bestsme >> sseshift) > 4000) {/* Check if mvp_full is within the range. */
- vp8_clamp_mv(&mvp_full, x->mv_col_min, x->mv_col_max, x->mv_row_min, x->mv_row_max);
+ vp8_clamp_mv(&mvp_full, x->mv_col_min, x->mv_col_max,
+ x->mv_row_min, x->mv_row_max);
thissme = cpi->full_search_sad(x, c, e, &mvp_full,
sadpb, 16, v_fn_ptr,
@@ -2389,7 +2454,8 @@
bestsme = thissme;
mode_mv[NEW4X4].as_int = e->bmi.as_mv.first.as_int;
} else {- // The full search result is actually worse so re-instate the previous best vector
+ /* The full search result is actually worse so re-instate the
+ * previous best vector */
e->bmi.as_mv.first.as_int = mode_mv[NEW4X4].as_int;
}
}
@@ -2399,8 +2465,8 @@
int distortion;
unsigned int sse;
cpi->find_fractional_mv_step(x, c, e, &mode_mv[NEW4X4],
- bsi->ref_mv, x->errorperbit, v_fn_ptr, XMVCOST,
- &distortion, &sse);
+ bsi->ref_mv, x->errorperbit, v_fn_ptr,
+ XMVCOST, &distortion, &sse);
// safe motion search result for use in compound prediction
seg_mvs[i][mbmi->ref_frame - 1].as_int = mode_mv[NEW4X4].as_int;
@@ -2407,7 +2473,8 @@
}
} /* NEW4X4 */
else if (mbmi->second_ref_frame && this_mode == NEW4X4) {- // motion search not completed? Then skip newmv for this block with comppred
+ /* motion search not completed? Then skip newmv for this block with
+ * comppred */
if (seg_mvs[i][mbmi->second_ref_frame - 1].as_int == INVALID_MV ||
seg_mvs[i][mbmi->ref_frame - 1].as_int == INVALID_MV) {continue;
@@ -2429,14 +2496,15 @@
mv_check_bounds(x, &second_mode_mv[this_mode]))
continue;
- if (segmentation == BLOCK_4X4) {+ if (segmentation == PARTITIONING_4X4) {this_rd = encode_inter_mb_segment(x, labels, i, &labelyrate,
&distortion,
ta_s, tl_s, IF_RTCD(&cpi->rtcd));
+ other_rd = this_rd;
} else {this_rd = encode_inter_mb_segment_8x8(x, labels, i, &labelyrate,
- &distortion, ta_s, tl_s,
- IF_RTCD(&cpi->rtcd));
+ &distortion, &other_rd,
+ ta_s, tl_s, IF_RTCD(&cpi->rtcd));
}
this_rd += RDCOST(x->rdmult, x->rddiv, rate, 0);
rate += labelyrate;
@@ -2447,10 +2515,21 @@
bestlabelyrate = labelyrate;
mode_selected = this_mode;
best_label_rd = this_rd;
- for (j = 0; j < 16; j++)
- if (labels[j] == i)
- best_eobs[j] = x->e_mbd.block[j].eob;
+ if (x->e_mbd.mode_info_context->mbmi.txfm_size == TX_4X4) {+ for (j = 0; j < 16; j++)
+ if (labels[j] == i)
+ best_eobs[j] = x->e_mbd.block[j].eob;
+ } else {+ for (j = 0; j < 4; j++) {+ int ib = vp8_i8x8_block[j], idx = j * 4;
+ if (labels[ib] == i)
+ best_eobs[idx] = x->e_mbd.block[idx].eob;
+ }
+ }
+ if (other_rd < best_other_rd)
+ best_other_rd = other_rd;
+
vpx_memcpy(ta_b, ta_s, sizeof(ENTROPY_CONTEXT_PLANES));
vpx_memcpy(tl_b, tl_s, sizeof(ENTROPY_CONTEXT_PLANES));
@@ -2461,18 +2540,18 @@
vpx_memcpy(tl, tl_b, sizeof(ENTROPY_CONTEXT_PLANES));
labels2mode(x, labels, i, mode_selected, &mode_mv[mode_selected],
- &second_mode_mv[mode_selected], seg_mvs[i], bsi->ref_mv, bsi->second_ref_mv, XMVCOST);
+ &second_mode_mv[mode_selected], seg_mvs[i],
+ bsi->ref_mv, bsi->second_ref_mv, XMVCOST);
br += sbr;
bd += sbd;
segmentyrate += bestlabelyrate;
this_segment_rd += best_label_rd;
-
- if (this_segment_rd >= bsi->segment_rd) {- break;
- }
-
-
+ other_segment_rd += best_other_rd;
+ if (rds)
+ rds[i] = this_segment_rd;
+ if (otherrds)
+ rds[i] = other_segment_rd;
} /* for each label */
if (this_segment_rd < bsi->segment_rd) {@@ -2481,6 +2560,7 @@
bsi->segment_yrate = segmentyrate;
bsi->segment_rd = this_segment_rd;
bsi->segment_num = segmentation;
+ bsi->txfm_size = mbmi->txfm_size;
// store everything needed to come back to this!!
for (i = 0; i < 16; i++) {@@ -2493,8 +2573,114 @@
bsi->eobs[i] = best_eobs[i];
}
}
+
+ if (completed) {+ *completed = i;
+ }
}
+static void rd_check_segment(VP8_COMP *cpi, MACROBLOCK *x,
+ BEST_SEG_INFO *bsi,
+ unsigned int segmentation,
+ /* 16 = n_blocks */
+ int_mv seg_mvs[16][MAX_REF_FRAMES - 1],
+ int64_t txfm_cache[NB_TXFM_MODES]) {+#if CONFIG_TX_SELECT
+ int i, n, c = vp8_mbsplit_count[segmentation];
+
+ if (segmentation == PARTITIONING_4X4) {+ int64_t rd[16];
+
+ rd_check_segment_txsize(cpi, x, bsi, segmentation, TX_4X4, NULL,
+ rd, &n, seg_mvs);
+ if (n == c) {+ for (i = 0; i < NB_TXFM_MODES; i++) {+ if (rd[c - 1] < txfm_cache[i])
+ txfm_cache[i] = rd[c - 1];
+ }
+ }
+ } else {+ int64_t diff, base_rd;
+ int cost4x4 = vp8_cost_bit(cpi->common.prob_tx[0], 0);
+ int cost8x8 = vp8_cost_bit(cpi->common.prob_tx[0], 1);
+
+ if (cpi->common.txfm_mode == TX_MODE_SELECT) {+ int64_t rd4x4[4], rd8x8[4];
+ int n4x4, n8x8, nmin;
+ BEST_SEG_INFO bsi4x4, bsi8x8;
+
+ /* factor in cost of cost4x4/8x8 in decision */
+ vpx_memcpy(&bsi4x4, bsi, sizeof(*bsi));
+ vpx_memcpy(&bsi8x8, bsi, sizeof(*bsi));
+ rd_check_segment_txsize(cpi, x, &bsi4x4, segmentation,
+ TX_4X4, NULL, rd4x4, &n4x4, seg_mvs);
+ rd_check_segment_txsize(cpi, x, &bsi8x8, segmentation,
+ TX_8X8, NULL, rd8x8, &n8x8, seg_mvs);
+ if (bsi4x4.segment_num == segmentation) {+ bsi4x4.segment_rd += RDCOST(x->rdmult, x->rddiv, cost4x4, 0);
+ if (bsi4x4.segment_rd < bsi->segment_rd)
+ vpx_memcpy(bsi, &bsi4x4, sizeof(*bsi));
+ }
+ if (bsi8x8.segment_num == segmentation) {+ bsi8x8.segment_rd += RDCOST(x->rdmult, x->rddiv, cost8x8, 0);
+ if (bsi8x8.segment_rd < bsi->segment_rd)
+ vpx_memcpy(bsi, &bsi8x8, sizeof(*bsi));
+ }
+ n = n4x4 > n8x8 ? n4x4 : n8x8;
+ if (n == c) {+ nmin = n4x4 < n8x8 ? n4x4 : n8x8;
+ diff = rd8x8[nmin - 1] - rd4x4[nmin - 1];
+ if (n == n4x4) {+ base_rd = rd4x4[c - 1];
+ } else {+ base_rd = rd8x8[c - 1] - diff;
+ }
+ }
+ } else {+ int64_t rd[4], otherrd[4];
+
+ if (cpi->common.txfm_mode == ONLY_4X4) {+ rd_check_segment_txsize(cpi, x, bsi, segmentation, TX_4X4, otherrd,
+ rd, &n, seg_mvs);
+ if (n == c) {+ base_rd = rd[c - 1];
+ diff = otherrd[c - 1] - rd[c - 1];
+ }
+ } else /* use 8x8 transform */ {+ rd_check_segment_txsize(cpi, x, bsi, segmentation, TX_8X8, otherrd,
+ rd, &n, seg_mvs);
+ if (n == c) {+ diff = rd[c - 1] - otherrd[c - 1];
+ base_rd = otherrd[c - 1];
+ }
+ }
+ }
+
+ if (n == c) {+ if (base_rd < txfm_cache[ONLY_4X4]) {+ txfm_cache[ONLY_4X4] = base_rd;
+ }
+ if (base_rd + diff < txfm_cache[1]) {+ txfm_cache[ALLOW_8X8] = txfm_cache[ALLOW_16X16] = base_rd + diff;
+ }
+ if (diff < 0) {+ base_rd += diff + RDCOST(x->rdmult, x->rddiv, cost8x8, 0);
+ } else {+ base_rd += RDCOST(x->rdmult, x->rddiv, cost4x4, 0);
+ }
+ if (base_rd < txfm_cache[TX_MODE_SELECT]) {+ txfm_cache[TX_MODE_SELECT] = base_rd;
+ }
+ }
+ }
+#else
+ rd_check_segment_txsize(cpi, x, bsi, segmentation,
+ (segmentation == PARTITIONING_4X4 ||
+ cpi->common.txfm_mode == ONLY_4X4) ? TX_4X4 : TX_8X8,
+ NULL, NULL, NULL, seg_mvs);
+#endif
+}
+
static __inline
void vp8_cal_step_param(int sr, int *sp) {int step = 0;
@@ -2508,17 +2694,26 @@
*sp = MAX_MVSEARCH_STEPS - 1 - step;
}
-static int vp8_rd_pick_best_mbsegmentation(VP8_COMP *cpi, MACROBLOCK *x,
- int_mv *best_ref_mv, int_mv *second_best_ref_mv, int64_t best_rd,
- int *mdcounts, int *returntotrate,
- int *returnyrate, int *returndistortion,
- int *skippable, int mvthresh,
- int_mv seg_mvs[BLOCK_MAX_SEGMENTS - 1][16 /* n_blocks */][MAX_REF_FRAMES - 1]) {+static int rd_pick_best_mbsegmentation(VP8_COMP *cpi, MACROBLOCK *x,
+ int_mv *best_ref_mv,
+ int_mv *second_best_ref_mv,
+ int64_t best_rd,
+ int *mdcounts,
+ int *returntotrate,
+ int *returnyrate,
+ int *returndistortion,
+ int *skippable, int mvthresh,
+ int_mv seg_mvs[NB_PARTITIONINGS]
+ [16 /* n_blocks */]
+ [MAX_REF_FRAMES - 1],
+ int64_t txfm_cache[NB_TXFM_MODES]) {int i;
BEST_SEG_INFO bsi;
MB_MODE_INFO * mbmi = &x->e_mbd.mode_info_context->mbmi;
vpx_memset(&bsi, 0, sizeof(bsi));
+ for (i = 0; i < NB_TXFM_MODES; i++)
+ txfm_cache[i] = INT64_MAX;
bsi.segment_rd = best_rd;
bsi.ref_mv = best_ref_mv;
@@ -2526,6 +2721,7 @@
bsi.mvp.as_int = best_ref_mv->as_int;
bsi.mvthresh = mvthresh;
bsi.mdcounts = mdcounts;
+ bsi.txfm_size = TX_4X4;
for (i = 0; i < 16; i++)
bsi.modes[i] = ZERO4X4;
@@ -2533,16 +2729,20 @@
if (cpi->compressor_speed == 0) {/* for now, we will keep the original segmentation order
when in best quality mode */
- rd_check_segment(cpi, x, &bsi, BLOCK_16X8, seg_mvs[BLOCK_16X8]);
- rd_check_segment(cpi, x, &bsi, BLOCK_8X16, seg_mvs[BLOCK_8X16]);
- rd_check_segment(cpi, x, &bsi, BLOCK_8X8, seg_mvs[BLOCK_8X8]);
- rd_check_segment(cpi, x, &bsi, BLOCK_4X4, seg_mvs[BLOCK_4X4]);
+ rd_check_segment(cpi, x, &bsi, PARTITIONING_16X8,
+ seg_mvs[PARTITIONING_16X8], txfm_cache);
+ rd_check_segment(cpi, x, &bsi, PARTITIONING_8X16,
+ seg_mvs[PARTITIONING_8X16], txfm_cache);
+ rd_check_segment(cpi, x, &bsi, PARTITIONING_8X8,
+ seg_mvs[PARTITIONING_8X8], txfm_cache);
+ rd_check_segment(cpi, x, &bsi, PARTITIONING_4X4,
+ seg_mvs[PARTITIONING_4X4], txfm_cache);
} else {int sr;
- rd_check_segment(cpi, x, &bsi, BLOCK_8X8, seg_mvs[BLOCK_8X8]);
+ rd_check_segment(cpi, x, &bsi, PARTITIONING_8X8,
+ seg_mvs[PARTITIONING_8X8], txfm_cache);
-
if (bsi.segment_rd < best_rd) {int tmp_col_min = x->mv_col_min;
int tmp_col_max = x->mv_col_max;
@@ -2557,34 +2757,40 @@
bsi.sv_mvp[2].as_int = bsi.mvs[8].as_int;
bsi.sv_mvp[3].as_int = bsi.mvs[10].as_int;
- /* Use 8x8 result as 16x8/8x16's predictor MV. Adjust search range according to the closeness of 2 MV. */
+ /* Use 8x8 result as 16x8/8x16's predictor MV. Adjust search range
+ * according to the closeness of 2 MV. */
/* block 8X16 */
- {- sr = MAXF((abs(bsi.sv_mvp[0].as_mv.row - bsi.sv_mvp[2].as_mv.row)) >> 3, (abs(bsi.sv_mvp[0].as_mv.col - bsi.sv_mvp[2].as_mv.col)) >> 3);
- vp8_cal_step_param(sr, &bsi.sv_istep[0]);
+ sr = MAXF((abs(bsi.sv_mvp[0].as_mv.row - bsi.sv_mvp[2].as_mv.row)) >> 3,
+ (abs(bsi.sv_mvp[0].as_mv.col - bsi.sv_mvp[2].as_mv.col)) >> 3);
+ vp8_cal_step_param(sr, &bsi.sv_istep[0]);
- sr = MAXF((abs(bsi.sv_mvp[1].as_mv.row - bsi.sv_mvp[3].as_mv.row)) >> 3, (abs(bsi.sv_mvp[1].as_mv.col - bsi.sv_mvp[3].as_mv.col)) >> 3);
- vp8_cal_step_param(sr, &bsi.sv_istep[1]);
+ sr = MAXF((abs(bsi.sv_mvp[1].as_mv.row - bsi.sv_mvp[3].as_mv.row)) >> 3,
+ (abs(bsi.sv_mvp[1].as_mv.col - bsi.sv_mvp[3].as_mv.col)) >> 3);
+ vp8_cal_step_param(sr, &bsi.sv_istep[1]);
- rd_check_segment(cpi, x, &bsi, BLOCK_8X16, seg_mvs[BLOCK_8X16]);
- }
+ rd_check_segment(cpi, x, &bsi, PARTITIONING_8X16,
+ seg_mvs[PARTITIONING_8X16], txfm_cache);
/* block 16X8 */
- {- sr = MAXF((abs(bsi.sv_mvp[0].as_mv.row - bsi.sv_mvp[1].as_mv.row)) >> 3, (abs(bsi.sv_mvp[0].as_mv.col - bsi.sv_mvp[1].as_mv.col)) >> 3);
- vp8_cal_step_param(sr, &bsi.sv_istep[0]);
+ sr = MAXF((abs(bsi.sv_mvp[0].as_mv.row - bsi.sv_mvp[1].as_mv.row)) >> 3,
+ (abs(bsi.sv_mvp[0].as_mv.col - bsi.sv_mvp[1].as_mv.col)) >> 3);
+ vp8_cal_step_param(sr, &bsi.sv_istep[0]);
- sr = MAXF((abs(bsi.sv_mvp[2].as_mv.row - bsi.sv_mvp[3].as_mv.row)) >> 3, (abs(bsi.sv_mvp[2].as_mv.col - bsi.sv_mvp[3].as_mv.col)) >> 3);
- vp8_cal_step_param(sr, &bsi.sv_istep[1]);
+ sr = MAXF((abs(bsi.sv_mvp[2].as_mv.row - bsi.sv_mvp[3].as_mv.row)) >> 3,
+ (abs(bsi.sv_mvp[2].as_mv.col - bsi.sv_mvp[3].as_mv.col)) >> 3);
+ vp8_cal_step_param(sr, &bsi.sv_istep[1]);
- rd_check_segment(cpi, x, &bsi, BLOCK_16X8, seg_mvs[BLOCK_16X8]);
- }
+ rd_check_segment(cpi, x, &bsi, PARTITIONING_16X8,
+ seg_mvs[PARTITIONING_16X8], txfm_cache);
/* If 8x8 is better than 16x8/8x16, then do 4x4 search */
/* Not skip 4x4 if speed=0 (good quality) */
- if (cpi->sf.no_skip_block4x4_search || bsi.segment_num == BLOCK_8X8) { /* || (sv_segment_rd8x8-bsi.segment_rd) < sv_segment_rd8x8>>5) */+ if (cpi->sf.no_skip_block4x4_search ||
+ bsi.segment_num == PARTITIONING_8X8) {+ /* || (sv_segment_rd8x8-bsi.segment_rd) < sv_segment_rd8x8>>5) */
bsi.mvp.as_int = bsi.sv_mvp[0].as_int;
- rd_check_segment(cpi, x, &bsi, BLOCK_4X4, seg_mvs[BLOCK_4X4]);
+ rd_check_segment(cpi, x, &bsi, PARTITIONING_4X4,
+ seg_mvs[PARTITIONING_4X4], txfm_cache);
}
/* restore UMV window */
@@ -2608,9 +2814,12 @@
*returntotrate = bsi.r;
*returndistortion = bsi.d;
*returnyrate = bsi.segment_yrate;
- *skippable = mby_is_skippable_4x4(&x->e_mbd, 0);
+ *skippable = bsi.txfm_size == TX_4X4 ?
+ mby_is_skippable_4x4(&x->e_mbd, 0) :
+ mby_is_skippable_8x8(&x->e_mbd, 0);
/* save partitions */
+ mbmi->txfm_size = bsi.txfm_size;
mbmi->partitioning = bsi.segment_num;
x->partition_info->count = vp8_mbsplit_count[bsi.segment_num];
@@ -3299,7 +3508,7 @@
unsigned char *y_buffer[4], *u_buffer[4], *v_buffer[4];
unsigned int ref_costs[MAX_REF_FRAMES];
- int_mv seg_mvs[BLOCK_MAX_SEGMENTS - 1][16 /* n_blocks */][MAX_REF_FRAMES - 1];
+ int_mv seg_mvs[NB_PARTITIONINGS][16 /* n_blocks */][MAX_REF_FRAMES - 1];
vpx_memset(mode8x8, 0, sizeof(mode8x8));
vpx_memset(&frame_mv, 0, sizeof(frame_mv));
@@ -3314,7 +3523,7 @@
for (i = 0; i < NB_TXFM_MODES; i++)
best_txfm_rd[i] = INT64_MAX;
- for (i = 0; i < BLOCK_MAX_SEGMENTS - 1; i++) {+ for (i = 0; i < NB_PARTITIONINGS; i++) {int j, k;
for (j = 0; j < 16; j++)
@@ -3680,12 +3889,12 @@
(mbmi->ref_frame == GOLDEN_FRAME) ?
cpi->rd_threshes[THR_NEWG] : this_rd_thresh;
- mbmi->txfm_size = TX_4X4; // FIXME use 8x8 in case of 8x8/8x16/16x8
- tmp_rd = vp8_rd_pick_best_mbsegmentation(cpi, x, &best_ref_mv,
- second_ref, best_yrd, mdcounts,
- &rate, &rate_y, &distortion,
- &skippable,
- this_rd_thresh, seg_mvs);
+ tmp_rd = rd_pick_best_mbsegmentation(cpi, x, &best_ref_mv,
+ second_ref, best_yrd, mdcounts,
+ &rate, &rate_y, &distortion,
+ &skippable,
+ this_rd_thresh, seg_mvs,
+ txfm_cache);
rate2 += rate;
distortion2 += distortion;
@@ -4081,7 +4290,7 @@
if (!mode_excluded && this_rd != INT64_MAX) { for (i = 0; i < NB_TXFM_MODES; i++) {int64_t adj_rd;
- if (this_mode != B_PRED && this_mode != SPLITMV) {+ if (this_mode != B_PRED) {adj_rd = this_rd + txfm_cache[i] - txfm_cache[cm->txfm_mode];
} else {adj_rd = this_rd;
--- a/vp8/encoder/tokenize.c
+++ b/vp8/encoder/tokenize.c
@@ -589,7 +589,8 @@
xd->mode_info_context->mbmi.mb_skip_coeff = mb_is_skippable_16x16(xd);
break;
case TX_8X8:
- if (xd->mode_info_context->mbmi.mode == I8X8_PRED)
+ if (xd->mode_info_context->mbmi.mode == I8X8_PRED ||
+ xd->mode_info_context->mbmi.mode == SPLITMV)
xd->mode_info_context->mbmi.mb_skip_coeff = mb_is_skippable_8x8_4x4uv(xd, 0);
else
xd->mode_info_context->mbmi.mb_skip_coeff = mb_is_skippable_8x8(xd, has_y2_block);
@@ -665,7 +666,8 @@
*(A + vp8_block2above_8x8[b] + 1) = *(A + vp8_block2above_8x8[b]);
*(L + vp8_block2left_8x8[b] + 1) = *(L + vp8_block2left_8x8[b]);
}
- if (xd->mode_info_context->mbmi.mode == I8X8_PRED) {+ if (xd->mode_info_context->mbmi.mode == I8X8_PRED ||
+ xd->mode_info_context->mbmi.mode == SPLITMV) {tokenize1st_order_chroma_4x4(xd, t, cpi, dry_run);
} else { for (b = 16; b < 24; b += 4) {@@ -1260,7 +1262,8 @@
if (tx_size == TX_16X16) {vp8_stuff_mb_16x16(cpi, xd, t, dry_run);
} else if (tx_size == TX_8X8) {- if (xd->mode_info_context->mbmi.mode == I8X8_PRED) {+ if (xd->mode_info_context->mbmi.mode == I8X8_PRED ||
+ xd->mode_info_context->mbmi.mode == SPLITMV) {vp8_stuff_mb_8x8_4x4uv(cpi, xd, t, dry_run);
} else {vp8_stuff_mb_8x8(cpi, xd, t, dry_run);
--
⑨