Refactor intra block prediction function

This commit simplifies the intra block boundary condition logic.
It removes the block index from the argument set.

Change-Id: If00142512eb88992613d6609356dfd73ba390138

7 files changed, 32 insertions, 41 deletions

diff --git a/vp9/common/vp9_reconintra.c b/vp9/common/vp9_reconintra.c
index 1e9acb8d5..f969ff1a8 100644
--- a/vp9/common/vp9_reconintra.c
+++ b/vp9/common/vp9_reconintra.c
@@ -1059,20 +1059,19 @@ static void build_intra_predictors(const MACROBLOCKD *xd, const uint8_t *ref,
   }
 }
 
-void vp9_predict_intra_block(const MACROBLOCKD *xd, int block_idx, int bwl_in,
+void vp9_predict_intra_block(const MACROBLOCKD *xd, int bwl_in,
                              TX_SIZE tx_size, PREDICTION_MODE mode,
                              const uint8_t *ref, int ref_stride,
                              uint8_t *dst, int dst_stride,
                              int aoff, int loff, int plane) {
-  const int bwl = bwl_in - tx_size;
-  const int wmask = (1 << bwl) - 1;
-  const int have_top = (block_idx >> bwl) || xd->up_available;
-  const int have_left = (block_idx & wmask) || xd->left_available;
-  const int have_right = ((block_idx & wmask) != wmask);
+  const int bw = (1 << bwl_in);
+  const int txw = (1 << tx_size);
+  const int have_top = loff || xd->up_available;
+  const int have_left = aoff || xd->left_available;
+  const int have_right = (aoff + txw) < bw;
   const int x = aoff * 4;
   const int y = loff * 4;
 
-  assert(bwl >= 0);
 #if CONFIG_VP9_HIGHBITDEPTH
   if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
     build_intra_predictors_high(xd, ref, ref_stride, dst, dst_stride, mode,
diff --git a/vp9/common/vp9_reconintra.h b/vp9/common/vp9_reconintra.h
index da5e435b1..de453808b 100644
--- a/vp9/common/vp9_reconintra.h
+++ b/vp9/common/vp9_reconintra.h
@@ -20,7 +20,7 @@ extern "C" {
 
 void vp9_init_intra_predictors(void);
 
-void vp9_predict_intra_block(const MACROBLOCKD *xd, int block_idx, int bwl_in,
+void vp9_predict_intra_block(const MACROBLOCKD *xd, int bwl_in,
                              TX_SIZE tx_size, PREDICTION_MODE mode,
                              const uint8_t *ref, int ref_stride,
                              uint8_t *dst, int dst_stride,
diff --git a/vp9/decoder/vp9_decodeframe.c b/vp9/decoder/vp9_decodeframe.c
index b3cf3fd4f..aeca7e857 100644
--- a/vp9/decoder/vp9_decodeframe.c
+++ b/vp9/decoder/vp9_decodeframe.c
@@ -391,8 +391,7 @@ static void predict_and_reconstruct_intra_block(int plane, int block,
   dec_txfrm_block_to_raster_xy(bwl, tx_size, block, &x, &y);
   dst = &pd->dst.buf[4 * y * pd->dst.stride + 4 * x];
 
-  vp9_predict_intra_block(xd, block >> (tx_size << 1),
-                          bwl, tx_size, mode,
+  vp9_predict_intra_block(xd, bwl, tx_size, mode,
                           dst, pd->dst.stride, dst, pd->dst.stride,
                           x, y, plane);
 
diff --git a/vp9/encoder/vp9_encodemb.c b/vp9/encoder/vp9_encodemb.c
index 659ce72b6..f74cdd8aa 100644
--- a/vp9/encoder/vp9_encodemb.c
+++ b/vp9/encoder/vp9_encodemb.c
@@ -795,7 +795,7 @@ void vp9_encode_block_intra(int plane, int block, BLOCK_SIZE plane_bsize,
       case TX_32X32:
         scan_order = &vp9_default_scan_orders[TX_32X32];
         mode = plane == 0 ? mbmi->mode : mbmi->uv_mode;
-        vp9_predict_intra_block(xd, block >> 6, bwl, TX_32X32, mode,
+        vp9_predict_intra_block(xd, bwl, TX_32X32, mode,
                                 x->skip_encode ? src : dst,
                                 x->skip_encode ? src_stride : dst_stride,
                                 dst, dst_stride, i, j, plane);
@@ -816,7 +816,7 @@ void vp9_encode_block_intra(int plane, int block, BLOCK_SIZE plane_bsize,
         tx_type = get_tx_type(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,
+        vp9_predict_intra_block(xd, bwl, TX_16X16, mode,
                                 x->skip_encode ? src : dst,
                                 x->skip_encode ? src_stride : dst_stride,
                                 dst, dst_stride, i, j, plane);
@@ -838,7 +838,7 @@ void vp9_encode_block_intra(int plane, int block, BLOCK_SIZE plane_bsize,
         tx_type = get_tx_type(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,
+        vp9_predict_intra_block(xd, bwl, TX_8X8, mode,
                                 x->skip_encode ? src : dst,
                                 x->skip_encode ? src_stride : dst_stride,
                                 dst, dst_stride, i, j, plane);
@@ -860,7 +860,7 @@ void vp9_encode_block_intra(int plane, int block, BLOCK_SIZE plane_bsize,
         tx_type = get_tx_type_4x4(pd->plane_type, xd, block);
         scan_order = &vp9_scan_orders[TX_4X4][tx_type];
         mode = plane == 0 ? get_y_mode(xd->mi[0], block) : mbmi->uv_mode;
-        vp9_predict_intra_block(xd, block, bwl, TX_4X4, mode,
+        vp9_predict_intra_block(xd, bwl, TX_4X4, mode,
                                 x->skip_encode ? src : dst,
                                 x->skip_encode ? src_stride : dst_stride,
                                 dst, dst_stride, i, j, plane);
@@ -903,7 +903,7 @@ void vp9_encode_block_intra(int plane, int block, BLOCK_SIZE plane_bsize,
     case TX_32X32:
       scan_order = &vp9_default_scan_orders[TX_32X32];
       mode = plane == 0 ? mbmi->mode : mbmi->uv_mode;
-      vp9_predict_intra_block(xd, block >> 6, bwl, TX_32X32, mode,
+      vp9_predict_intra_block(xd, bwl, TX_32X32, mode,
                               x->skip_encode ? src : dst,
                               x->skip_encode ? src_stride : dst_stride,
                               dst, dst_stride, i, j, plane);
@@ -923,7 +923,7 @@ void vp9_encode_block_intra(int plane, int block, BLOCK_SIZE plane_bsize,
       tx_type = get_tx_type(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,
+      vp9_predict_intra_block(xd, bwl, TX_16X16, mode,
                               x->skip_encode ? src : dst,
                               x->skip_encode ? src_stride : dst_stride,
                               dst, dst_stride, i, j, plane);
@@ -943,7 +943,7 @@ void vp9_encode_block_intra(int plane, int block, BLOCK_SIZE plane_bsize,
       tx_type = get_tx_type(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,
+      vp9_predict_intra_block(xd, bwl, TX_8X8, mode,
                               x->skip_encode ? src : dst,
                               x->skip_encode ? src_stride : dst_stride,
                               dst, dst_stride, i, j, plane);
@@ -963,7 +963,7 @@ void vp9_encode_block_intra(int plane, int block, BLOCK_SIZE plane_bsize,
       tx_type = get_tx_type_4x4(pd->plane_type, xd, block);
       scan_order = &vp9_scan_orders[TX_4X4][tx_type];
       mode = plane == 0 ? get_y_mode(xd->mi[0], block) : mbmi->uv_mode;
-      vp9_predict_intra_block(xd, block, bwl, TX_4X4, mode,
+      vp9_predict_intra_block(xd, bwl, TX_4X4, mode,
                               x->skip_encode ? src : dst,
                               x->skip_encode ? src_stride : dst_stride,
                               dst, dst_stride, i, j, plane);
diff --git a/vp9/encoder/vp9_mbgraph.c b/vp9/encoder/vp9_mbgraph.c
index d5eeb9cc5..acbd7dd43 100644
--- a/vp9/encoder/vp9_mbgraph.c
+++ b/vp9/encoder/vp9_mbgraph.c
@@ -145,7 +145,7 @@ static int find_best_16x16_intra(VP9_COMP *cpi, PREDICTION_MODE *pbest_mode) {
     unsigned int err;
 
     xd->mi[0]->mbmi.mode = mode;
-    vp9_predict_intra_block(xd, 0, 2, TX_16X16, mode,
+    vp9_predict_intra_block(xd, 2, TX_16X16, mode,
                             x->plane[0].src.buf, x->plane[0].src.stride,
                             xd->plane[0].dst.buf, xd->plane[0].dst.stride,
                             0, 0, 0);
diff --git a/vp9/encoder/vp9_pickmode.c b/vp9/encoder/vp9_pickmode.c
index e99cbc7a0..9bc5629c0 100644
--- a/vp9/encoder/vp9_pickmode.c
+++ b/vp9/encoder/vp9_pickmode.c
@@ -902,8 +902,7 @@ static void estimate_block_intra(int plane, int block, BLOCK_SIZE plane_bsize,
   p->src.buf = &src_buf_base[4 * (j * src_stride + i)];
   pd->dst.buf = &dst_buf_base[4 * (j * dst_stride + i)];
   // Use source buffer as an approximation for the fully reconstructed buffer.
-  vp9_predict_intra_block(xd, block >> (2 * tx_size),
-                          b_width_log2_lookup[plane_bsize],
+  vp9_predict_intra_block(xd, b_width_log2_lookup[plane_bsize],
                           tx_size, args->mode,
                           x->skip_encode ? p->src.buf : pd->dst.buf,
                           x->skip_encode ? src_stride : dst_stride,
diff --git a/vp9/encoder/vp9_rdopt.c b/vp9/encoder/vp9_rdopt.c
index 9985f8947..09cff00d9 100644
--- a/vp9/encoder/vp9_rdopt.c
+++ b/vp9/encoder/vp9_rdopt.c
@@ -760,7 +760,8 @@ static int conditional_skipintra(PREDICTION_MODE mode,
   return 0;
 }
 
-static int64_t rd_pick_intra4x4block(VP9_COMP *cpi, MACROBLOCK *x, int ib,
+static int64_t rd_pick_intra4x4block(VP9_COMP *cpi, MACROBLOCK *x,
+                                     int row, int col,
                                      PREDICTION_MODE *best_mode,
                                      const int *bmode_costs,
                                      ENTROPY_CONTEXT *a, ENTROPY_CONTEXT *l,
@@ -770,18 +771,14 @@ static int64_t rd_pick_intra4x4block(VP9_COMP *cpi, MACROBLOCK *x, int ib,
   PREDICTION_MODE mode;
   MACROBLOCKD *const xd = &x->e_mbd;
   int64_t best_rd = rd_thresh;
-
   struct macroblock_plane *p = &x->plane[0];
   struct macroblockd_plane *pd = &xd->plane[0];
   const int src_stride = p->src.stride;
   const int dst_stride = pd->dst.stride;
-  const uint8_t *src_init = &p->src.buf[vp9_raster_block_offset(BLOCK_8X8, ib,
-                                                                src_stride)];
-  uint8_t *dst_init = &pd->dst.buf[vp9_raster_block_offset(BLOCK_8X8, ib,
-                                                           dst_stride)];
+  const uint8_t *src_init = &p->src.buf[row * 4 * src_stride + col * 4];
+  uint8_t *dst_init = &pd->dst.buf[row * 4 * src_stride + col * 4];
   ENTROPY_CONTEXT ta[2], tempa[2];
   ENTROPY_CONTEXT tl[2], templ[2];
-
   const int num_4x4_blocks_wide = num_4x4_blocks_wide_lookup[bsize];
   const int num_4x4_blocks_high = num_4x4_blocks_high_lookup[bsize];
   int idx, idy;
@@ -790,8 +787,6 @@ static int64_t rd_pick_intra4x4block(VP9_COMP *cpi, MACROBLOCK *x, int ib,
   uint16_t best_dst16[8 * 8];
 #endif
 
-  assert(ib < 4);
-
   memcpy(ta, a, sizeof(ta));
   memcpy(tl, l, sizeof(tl));
   xd->mi[0]->mbmi.tx_size = TX_4X4;
@@ -819,7 +814,7 @@ static int64_t rd_pick_intra4x4block(VP9_COMP *cpi, MACROBLOCK *x, int ib,
 
       for (idy = 0; idy < num_4x4_blocks_high; ++idy) {
         for (idx = 0; idx < num_4x4_blocks_wide; ++idx) {
-          const int block = ib + idy * 2 + idx;
+          const int block = (row + idy) * 2 + (col + idx);
           const uint8_t *const src = &src_init[idx * 4 + idy * 4 * src_stride];
           uint8_t *const dst = &dst_init[idx * 4 + idy * 4 * dst_stride];
           int16_t *const src_diff = vp9_raster_block_offset_int16(BLOCK_8X8,
@@ -827,11 +822,11 @@ static int64_t rd_pick_intra4x4block(VP9_COMP *cpi, MACROBLOCK *x, int ib,
                                                                   p->src_diff);
           tran_low_t *const coeff = BLOCK_OFFSET(x->plane[0].coeff, block);
           xd->mi[0]->bmi[block].as_mode = mode;
-          vp9_predict_intra_block(xd, block, 1,
-                                  TX_4X4, mode,
+          vp9_predict_intra_block(xd, 1, TX_4X4, mode,
                                   x->skip_encode ? src : dst,
                                   x->skip_encode ? src_stride : dst_stride,
-                                  dst, dst_stride, idx, idy, 0);
+                                  dst, dst_stride,
+                                  col + idx, row + idy, 0);
           vpx_highbd_subtract_block(4, 4, src_diff, 8, src, src_stride,
                                     dst, dst_stride, xd->bd);
           if (xd->lossless) {
@@ -920,18 +915,17 @@ static int64_t rd_pick_intra4x4block(VP9_COMP *cpi, MACROBLOCK *x, int ib,
 
     for (idy = 0; idy < num_4x4_blocks_high; ++idy) {
       for (idx = 0; idx < num_4x4_blocks_wide; ++idx) {
-        const int block = ib + idy * 2 + idx;
+        const int block = (row + idy) * 2 + (col + idx);
         const uint8_t *const src = &src_init[idx * 4 + idy * 4 * src_stride];
         uint8_t *const dst = &dst_init[idx * 4 + idy * 4 * dst_stride];
         int16_t *const src_diff =
             vp9_raster_block_offset_int16(BLOCK_8X8, block, p->src_diff);
         tran_low_t *const coeff = BLOCK_OFFSET(x->plane[0].coeff, block);
         xd->mi[0]->bmi[block].as_mode = mode;
-        vp9_predict_intra_block(xd, block, 1,
-                                TX_4X4, mode,
+        vp9_predict_intra_block(xd, 1, TX_4X4, mode,
                                 x->skip_encode ? src : dst,
                                 x->skip_encode ? src_stride : dst_stride,
-                                dst, dst_stride, idx, idy, 0);
+                                dst, dst_stride, col + idx, row + idy, 0);
         vpx_subtract_block(4, 4, src_diff, 8, src, src_stride, dst, dst_stride);
 
         if (xd->lossless) {
@@ -1030,9 +1024,9 @@ static int64_t rd_pick_intra_sub_8x8_y_mode(VP9_COMP *cpi, MACROBLOCK *mb,
         bmode_costs  = cpi->y_mode_costs[A][L];
       }
 
-      this_rd = rd_pick_intra4x4block(cpi, mb, i, &best_mode, bmode_costs,
-                                      t_above + idx, t_left + idy, &r, &ry, &d,
-                                      bsize, best_rd - total_rd);
+      this_rd = rd_pick_intra4x4block(cpi, mb, idy, idx, &best_mode,
+                                      bmode_costs, t_above + idx, t_left + idy,
+                                      &r, &ry, &d, bsize, best_rd - total_rd);
       if (this_rd >= best_rd - total_rd)
         return INT64_MAX;