libvpx: Pull from upstream

source/libvpx/vp9/encoder/vp9_encodemb.c

 /*
  *  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
  *
  *  Use of this source code is governed by a BSD-style license
  *  that can be found in the LICENSE file in the root of the source
  *  tree. An additional intellectual property rights grant can be found
  *  in the file PATENTS.  All contributing project authors may
  *  be found in the AUTHORS file in the root of the source tree.
  */
 
 
 #include "./vp9_rtcd.h"
 #include "./vpx_config.h"
 
 #include "vpx_mem/vpx_mem.h"
 
 #include "vp9/common/vp9_idct.h"
 #include "vp9/common/vp9_reconinter.h"
 #include "vp9/common/vp9_reconintra.h"
 #include "vp9/common/vp9_systemdependent.h"
 
-#include "vp9/encoder/vp9_dct.h"
 #include "vp9/encoder/vp9_encodemb.h"
 #include "vp9/encoder/vp9_quantize.h"
 #include "vp9/encoder/vp9_rdopt.h"
 #include "vp9/encoder/vp9_tokenize.h"
 
-void vp9_setup_interp_filters(MACROBLOCKD *xd, INTERP_FILTER filter,
-                              VP9_COMMON *cm) {
-  xd->interp_kernel = vp9_get_interp_kernel(filter == SWITCHABLE ? EIGHTTAP
-                                                                 : filter);
-  assert(((intptr_t)xd->interp_kernel & 0xff) == 0);
-}
+struct optimize_ctx {
+  ENTROPY_CONTEXT ta[MAX_MB_PLANE][16];
+  ENTROPY_CONTEXT tl[MAX_MB_PLANE][16];
+};
+
+struct encode_b_args {
+  MACROBLOCK *x;
+  struct optimize_ctx *ctx;
+  unsigned char *skip;
+};
 
 void vp9_subtract_block_c(int rows, int cols,
                           int16_t *diff_ptr, ptrdiff_t diff_stride,
                           const uint8_t *src_ptr, ptrdiff_t src_stride,
                           const uint8_t *pred_ptr, ptrdiff_t pred_stride) {
   int r, c;
 
   for (r = 0; r < rows; r++) {
     for (c = 0; c < cols; c++)
       diff_ptr[c] = src_ptr[c] - pred_ptr[c];
@@ skipping 294 matching lines @@
   const BLOCK_SIZE plane_bsize = get_plane_block_size(bsize, pd);
   const int num_4x4_w = num_4x4_blocks_wide_lookup[plane_bsize];
   const int num_4x4_h = num_4x4_blocks_high_lookup[plane_bsize];
   const MB_MODE_INFO *mbmi = &xd->mi_8x8[0]->mbmi;
   const TX_SIZE tx_size = plane ? get_uv_tx_size(mbmi) : mbmi->tx_size;
 
   vp9_get_entropy_contexts(tx_size, args->ctx->ta[plane], args->ctx->tl[plane],
                            pd->above_context, pd->left_context,
                            num_4x4_w, num_4x4_h);
 }
-
-void vp9_xform_quant(int plane, int block, BLOCK_SIZE plane_bsize,
-                     TX_SIZE tx_size, void *arg) {
-  struct encode_b_args* const args = arg;
-  MACROBLOCK* const x = args->x;
-  MACROBLOCKD* const xd = &x->e_mbd;
+void vp9_xform_quant(MACROBLOCK *x, int plane, int block,
+                     BLOCK_SIZE plane_bsize, TX_SIZE tx_size) {
+  MACROBLOCKD *const xd = &x->e_mbd;
   struct macroblock_plane *const p = &x->plane[plane];
   struct macroblockd_plane *const pd = &xd->plane[plane];
   int16_t *coeff = BLOCK_OFFSET(p->coeff, block);
   int16_t *qcoeff = BLOCK_OFFSET(p->qcoeff, block);
   int16_t *dqcoeff = BLOCK_OFFSET(pd->dqcoeff, block);
   const scan_order *scan_order;
   uint16_t *eob = &p->eobs[block];
   const int diff_stride = 4 * num_4x4_blocks_wide_lookup[plane_bsize];
   int i, j;
   int16_t *src_diff;
@@ skipping 58 matching lines @@
   // TODO(jingning): per transformed block zero forcing only enabled for
   // luma component. will integrate chroma components as well.
   if (x->zcoeff_blk[tx_size][block] && plane == 0) {
     p->eobs[block] = 0;
     ctx->ta[plane][i] = 0;
     ctx->tl[plane][j] = 0;
     return;
   }
 
   if (!x->skip_recode)
-    vp9_xform_quant(plane, block, plane_bsize, tx_size, arg);
+    vp9_xform_quant(x, plane, block, plane_bsize, tx_size);
 
   if (x->optimize && (!x->skip_recode || !x->skip_optimize)) {
     vp9_optimize_b(plane, block, plane_bsize, tx_size, x, ctx);
   } else {
     ctx->ta[plane][i] = p->eobs[block] > 0;
     ctx->tl[plane][j] = p->eobs[block] > 0;
   }
 
   if (p->eobs[block])
-    *(args->skip_coeff) = 0;
+    *(args->skip) = 0;
 
   if (x->skip_encode || p->eobs[block] == 0)
     return;
 
   switch (tx_size) {
     case TX_32X32:
       vp9_idct32x32_add(dqcoeff, dst, pd->dst.stride, p->eobs[block]);
       break;
     case TX_16X16:
       vp9_idct16x16_add(dqcoeff, dst, pd->dst.stride, p->eobs[block]);
@@ skipping 17 matching lines @@
   MACROBLOCK *const x = args->x;
   MACROBLOCKD *const xd = &x->e_mbd;
   struct macroblock_plane *const p = &x->plane[plane];
   struct macroblockd_plane *const pd = &xd->plane[plane];
   int16_t *const dqcoeff = BLOCK_OFFSET(pd->dqcoeff, block);
   int i, j;
   uint8_t *dst;
   txfrm_block_to_raster_xy(plane_bsize, tx_size, block, &i, &j);
   dst = &pd->dst.buf[4 * j * pd->dst.stride + 4 * i];
 
-  vp9_xform_quant(plane, block, plane_bsize, tx_size, arg);
+  vp9_xform_quant(x, plane, block, plane_bsize, tx_size);
 
   if (p->eobs[block] == 0)
     return;
 
   xd->itxm_add(dqcoeff, dst, pd->dst.stride, p->eobs[block]);
 }
 
 void vp9_encode_sby(MACROBLOCK *x, BLOCK_SIZE bsize) {
   MACROBLOCKD *const xd = &x->e_mbd;
   struct optimize_ctx ctx;
   MB_MODE_INFO *mbmi = &xd->mi_8x8[0]->mbmi;
-  struct encode_b_args arg = {x, &ctx, &mbmi->skip_coeff};
+  struct encode_b_args arg = {x, &ctx, &mbmi->skip};
 
   vp9_subtract_sby(x, bsize);
   if (x->optimize)
     optimize_init_b(0, bsize, &arg);
 
-  foreach_transformed_block_in_plane(xd, bsize, 0, encode_block_pass1, &arg);
+  vp9_foreach_transformed_block_in_plane(xd, bsize, 0, encode_block_pass1,
+                                         &arg);
 }
 
 void vp9_encode_sb(MACROBLOCK *x, BLOCK_SIZE bsize) {
   MACROBLOCKD *const xd = &x->e_mbd;
   struct optimize_ctx ctx;
   MB_MODE_INFO *mbmi = &xd->mi_8x8[0]->mbmi;
-  struct encode_b_args arg = {x, &ctx, &mbmi->skip_coeff};
+  struct encode_b_args arg = {x, &ctx, &mbmi->skip};
 
   if (!x->skip_recode)
     vp9_subtract_sb(x, bsize);
 
   if (x->optimize && (!x->skip_recode || !x->skip_optimize)) {
     int i;
     for (i = 0; i < MAX_MB_PLANE; ++i)
       optimize_init_b(i, bsize, &arg);
   }
 
-  foreach_transformed_block(xd, bsize, encode_block, &arg);
+  vp9_foreach_transformed_block(xd, bsize, encode_block, &arg);
 }
 
-void vp9_encode_block_intra(int plane, int block, BLOCK_SIZE plane_bsize,
-                            TX_SIZE tx_size, void *arg) {
+static void encode_block_intra(int plane, int block, BLOCK_SIZE plane_bsize,
+                               TX_SIZE tx_size, void *arg) {
   struct encode_b_args* const args = arg;
   MACROBLOCK *const x = args->x;
   MACROBLOCKD *const xd = &x->e_mbd;
   MB_MODE_INFO *mbmi = &xd->mi_8x8[0]->mbmi;
   struct macroblock_plane *const p = &x->plane[plane];
   struct macroblockd_plane *const pd = &xd->plane[plane];
   int16_t *coeff = BLOCK_OFFSET(p->coeff, block);
   int16_t *qcoeff = BLOCK_OFFSET(p->qcoeff, block);
   int16_t *dqcoeff = BLOCK_OFFSET(pd->dqcoeff, block);
   const scan_order *scan_order;
@@ skipping 40 matching lines @@
       tx_type = get_tx_type_16x16(pd->plane_type, xd);
       scan_order = &vp9_scan_orders[TX_16X16][tx_type];
       mode = plane == 0 ? mbmi->mode : mbmi->uv_mode;
       vp9_predict_intra_block(xd, block >> 4, bwl, TX_16X16, mode,
                               x->skip_encode ? src : dst,
                               x->skip_encode ? p->src.stride : pd->dst.stride,
                               dst, pd->dst.stride, i, j, plane);
       if (!x->skip_recode) {
         vp9_subtract_block(16, 16, src_diff, diff_stride,
                            src, p->src.stride, dst, pd->dst.stride);
-        vp9_fht16x16(tx_type, src_diff, coeff, diff_stride);
+        vp9_fht16x16(src_diff, coeff, diff_stride, tx_type);
         vp9_quantize_b(coeff, 256, x->skip_block, p->zbin, p->round,
                        p->quant, p->quant_shift, qcoeff, dqcoeff,
                        pd->dequant, p->zbin_extra, eob, scan_order->scan,
                        scan_order->iscan);
       }
       if (!x->skip_encode && *eob)
         vp9_iht16x16_add(tx_type, dqcoeff, dst, pd->dst.stride, *eob);
       break;
     case TX_8X8:
       tx_type = get_tx_type_8x8(pd->plane_type, xd);
       scan_order = &vp9_scan_orders[TX_8X8][tx_type];
       mode = plane == 0 ? mbmi->mode : mbmi->uv_mode;
       vp9_predict_intra_block(xd, block >> 2, bwl, TX_8X8, mode,
                               x->skip_encode ? src : dst,
                               x->skip_encode ? p->src.stride : pd->dst.stride,
                               dst, pd->dst.stride, i, j, plane);
       if (!x->skip_recode) {
         vp9_subtract_block(8, 8, src_diff, diff_stride,
                            src, p->src.stride, dst, pd->dst.stride);
-        vp9_fht8x8(tx_type, src_diff, coeff, diff_stride);
+        vp9_fht8x8(src_diff, coeff, diff_stride, tx_type);
         vp9_quantize_b(coeff, 64, x->skip_block, p->zbin, p->round, p->quant,
                        p->quant_shift, qcoeff, dqcoeff,
                        pd->dequant, p->zbin_extra, eob, scan_order->scan,
                        scan_order->iscan);
       }
       if (!x->skip_encode && *eob)
         vp9_iht8x8_add(tx_type, dqcoeff, dst, pd->dst.stride, *eob);
       break;
     case TX_4X4:
       tx_type = get_tx_type_4x4(pd->plane_type, xd, block);
       scan_order = &vp9_scan_orders[TX_4X4][tx_type];
       if (mbmi->sb_type < BLOCK_8X8 && plane == 0)
         mode = xd->mi_8x8[0]->bmi[block].as_mode;
       else
         mode = plane == 0 ? mbmi->mode : mbmi->uv_mode;
 
       vp9_predict_intra_block(xd, block, bwl, TX_4X4, mode,
                               x->skip_encode ? src : dst,
                               x->skip_encode ? p->src.stride : pd->dst.stride,
                               dst, pd->dst.stride, i, j, plane);
 
       if (!x->skip_recode) {
         vp9_subtract_block(4, 4, src_diff, diff_stride,
                            src, p->src.stride, dst, pd->dst.stride);
         if (tx_type != DCT_DCT)
-          vp9_short_fht4x4(src_diff, coeff, diff_stride, tx_type);
+          vp9_fht4x4(src_diff, coeff, diff_stride, tx_type);
         else
           x->fwd_txm4x4(src_diff, coeff, diff_stride);
         vp9_quantize_b(coeff, 16, x->skip_block, p->zbin, p->round, p->quant,
                        p->quant_shift, qcoeff, dqcoeff,
                        pd->dequant, p->zbin_extra, eob, scan_order->scan,
                        scan_order->iscan);
       }
 
       if (!x->skip_encode && *eob) {
         if (tx_type == DCT_DCT)
           // this is like vp9_short_idct4x4 but has a special case around eob<=1
           // which is significant (not just an optimization) for the lossless
           // case.
           xd->itxm_add(dqcoeff, dst, pd->dst.stride, *eob);
         else
           vp9_iht4x4_16_add(dqcoeff, dst, pd->dst.stride, tx_type);
       }
       break;
     default:
       assert(0);
   }
   if (*eob)
-    *(args->skip_coeff) = 0;
+    *(args->skip) = 0;
 }
 
-void vp9_encode_intra_block_y(MACROBLOCK *x, BLOCK_SIZE bsize) {
-  MACROBLOCKD* const xd = &x->e_mbd;
-  struct optimize_ctx ctx;
-  MB_MODE_INFO *mbmi = &xd->mi_8x8[0]->mbmi;
-  struct encode_b_args arg = {x, &ctx, &mbmi->skip_coeff};
+void vp9_encode_block_intra(MACROBLOCK *x, int plane, int block,
+                            BLOCK_SIZE plane_bsize, TX_SIZE tx_size,
+                            unsigned char *skip) {
+  struct encode_b_args arg = {x, NULL, skip};
+  encode_block_intra(plane, block, plane_bsize, tx_size, &arg);
+}
 
-  foreach_transformed_block_in_plane(xd, bsize, 0, vp9_encode_block_intra,
-                                     &arg);
-}
-void vp9_encode_intra_block_uv(MACROBLOCK *x, BLOCK_SIZE bsize) {
-  MACROBLOCKD* const xd = &x->e_mbd;
-  struct optimize_ctx ctx;
-  MB_MODE_INFO *mbmi = &xd->mi_8x8[0]->mbmi;
-  struct encode_b_args arg = {x, &ctx, &mbmi->skip_coeff};
-  foreach_transformed_block_uv(xd, bsize, vp9_encode_block_intra, &arg);
+
+void vp9_encode_intra_block_plane(MACROBLOCK *x, BLOCK_SIZE bsize, int plane) {
+  const MACROBLOCKD *const xd = &x->e_mbd;
+  struct encode_b_args arg = {x, NULL, &xd->mi_8x8[0]->mbmi.skip};
+
+  vp9_foreach_transformed_block_in_plane(xd, bsize, plane, encode_block_intra,
+                                         &arg);
 }
 
 int vp9_encode_intra(MACROBLOCK *x, int use_16x16_pred) {
   MB_MODE_INFO * mbmi = &x->e_mbd.mi_8x8[0]->mbmi;
   x->skip_encode = 0;
   mbmi->mode = DC_PRED;
   mbmi->ref_frame[0] = INTRA_FRAME;
   mbmi->tx_size = use_16x16_pred ? (mbmi->sb_type >= BLOCK_16X16 ? TX_16X16
                                                                  : TX_8X8)
                                    : TX_4X4;
-  vp9_encode_intra_block_y(x, mbmi->sb_type);
+  vp9_encode_intra_block_plane(x, mbmi->sb_type, 0);
   return vp9_get_mb_ss(x->plane[0].src_diff);
 }