Support adaptive scan order in cb4x4 mode

This commit adds 2x2 transform block scan order to make the
adaptive scan order support cb4x4 mode.

BUG=aomedia:135

Change-Id: Ic8c3ae9ed65d577df629524b617b386b5e799d4c

av1/common/entropy.c

@@ -3588,7 +3588,7 @@ void av1_adapt_coef_probs(AV1_COMMON *cm) {
     adapt_coef_probs(cm, tx_size, count_sat, update_factor);
 
 #if CONFIG_ADAPT_SCAN
-  for (tx_size = TX_4X4; tx_size < TX_SIZES; ++tx_size)
+  for (tx_size = 0; tx_size < TX_SIZES; ++tx_size)
     for (tx_type = DCT_DCT; tx_type < TX_TYPES; ++tx_type) {
       av1_update_scan_prob(cm, tx_size, tx_type, ADAPT_SCAN_UPDATE_RATE_16);
       av1_update_scan_order_facade(cm, tx_size, tx_type);

av1/common/entropymode.h

@@ -67,24 +67,35 @@ typedef struct frame_contexts {
 #endif  // CONFIG_EC_MULTISYMBOL
   aom_prob switchable_interp_prob[SWITCHABLE_FILTER_CONTEXTS]
                                  [SWITCHABLE_FILTERS - 1];
-
 #if CONFIG_ADAPT_SCAN
-  // TODO(angiebird): try aom_prob
+// TODO(angiebird): try aom_prob
+#if CONFIG_CB4X4
+  uint32_t non_zero_prob_2x2[TX_TYPES][4];
+#endif
   uint32_t non_zero_prob_4X4[TX_TYPES][16];
   uint32_t non_zero_prob_8X8[TX_TYPES][64];
   uint32_t non_zero_prob_16X16[TX_TYPES][256];
   uint32_t non_zero_prob_32X32[TX_TYPES][1024];
 
+#if CONFIG_CB4X4
+  DECLARE_ALIGNED(16, int16_t, scan_2x2[TX_TYPES][4]);
+#endif
   DECLARE_ALIGNED(16, int16_t, scan_4X4[TX_TYPES][16]);
   DECLARE_ALIGNED(16, int16_t, scan_8X8[TX_TYPES][64]);
   DECLARE_ALIGNED(16, int16_t, scan_16X16[TX_TYPES][256]);
   DECLARE_ALIGNED(16, int16_t, scan_32X32[TX_TYPES][1024]);
 
+#if CONFIG_CB4X4
+  DECLARE_ALIGNED(16, int16_t, iscan_2x2[TX_TYPES][4]);
+#endif
   DECLARE_ALIGNED(16, int16_t, iscan_4X4[TX_TYPES][16]);
   DECLARE_ALIGNED(16, int16_t, iscan_8X8[TX_TYPES][64]);
   DECLARE_ALIGNED(16, int16_t, iscan_16X16[TX_TYPES][256]);
   DECLARE_ALIGNED(16, int16_t, iscan_32X32[TX_TYPES][1024]);
 
+#if CONFIG_CB4X4
+  int16_t nb_2x2[TX_TYPES][(4 + 1) * 2];
+#endif
   int16_t nb_4X4[TX_TYPES][(16 + 1) * 2];
   int16_t nb_8X8[TX_TYPES][(64 + 1) * 2];
   int16_t nb_16X16[TX_TYPES][(256 + 1) * 2];
@@ -199,12 +210,15 @@ typedef struct FRAME_COUNTS {
   unsigned int switchable_interp[SWITCHABLE_FILTER_CONTEXTS]
                                 [SWITCHABLE_FILTERS];
 #if CONFIG_ADAPT_SCAN
+#if CONFIG_CB4X4
+  unsigned int non_zero_count_2x2[TX_TYPES][4];
+#endif  // CONFIG_CB4X4
   unsigned int non_zero_count_4X4[TX_TYPES][16];
   unsigned int non_zero_count_8X8[TX_TYPES][64];
   unsigned int non_zero_count_16X16[TX_TYPES][256];
   unsigned int non_zero_count_32X32[TX_TYPES][1024];
   unsigned int txb_count[TX_SIZES][TX_TYPES];
-#endif
+#endif  // CONFIG_ADAPT_SCAN
 
 #if CONFIG_REF_MV
   unsigned int newmv_mode[NEWMV_MODE_CONTEXTS][2];

av1/common/scan.c

@@ -6493,6 +6493,9 @@ const SCAN_ORDER av1_inter_scan_orders[TX_SIZES_ALL][TX_TYPES] = {
 static uint32_t *get_non_zero_prob(FRAME_CONTEXT *fc, TX_SIZE tx_size,
                                    TX_TYPE tx_type) {
   switch (tx_size) {
+#if CONFIG_CB4X4
+    case TX_2X2: return fc->non_zero_prob_2x2[tx_type];
+#endif
     case TX_4X4: return fc->non_zero_prob_4X4[tx_type];
     case TX_8X8: return fc->non_zero_prob_8X8[tx_type];
     case TX_16X16: return fc->non_zero_prob_16X16[tx_type];
@@ -6504,6 +6507,9 @@ static uint32_t *get_non_zero_prob(FRAME_CONTEXT *fc, TX_SIZE tx_size,
 static int16_t *get_adapt_scan(FRAME_CONTEXT *fc, TX_SIZE tx_size,
                                TX_TYPE tx_type) {
   switch (tx_size) {
+#if CONFIG_CB4X4
+    case TX_2X2: return fc->scan_2x2[tx_type];
+#endif
     case TX_4X4: return fc->scan_4X4[tx_type];
     case TX_8X8: return fc->scan_8X8[tx_type];
     case TX_16X16: return fc->scan_16X16[tx_type];
@@ -6515,6 +6521,9 @@ static int16_t *get_adapt_scan(FRAME_CONTEXT *fc, TX_SIZE tx_size,
 static int16_t *get_adapt_iscan(FRAME_CONTEXT *fc, TX_SIZE tx_size,
                                 TX_TYPE tx_type) {
   switch (tx_size) {
+#if CONFIG_CB4X4
+    case TX_2X2: return fc->iscan_2x2[tx_type];
+#endif
     case TX_4X4: return fc->iscan_4X4[tx_type];
     case TX_8X8: return fc->iscan_8X8[tx_type];
     case TX_16X16: return fc->iscan_16X16[tx_type];
@@ -6526,6 +6535,9 @@ static int16_t *get_adapt_iscan(FRAME_CONTEXT *fc, TX_SIZE tx_size,
 static int16_t *get_adapt_nb(FRAME_CONTEXT *fc, TX_SIZE tx_size,
                              TX_TYPE tx_type) {
   switch (tx_size) {
+#if CONFIG_CB4X4
+    case TX_2X2: return fc->nb_2x2[tx_type];
+#endif
     case TX_4X4: return fc->nb_4X4[tx_type];
     case TX_8X8: return fc->nb_8X8[tx_type];
     case TX_16X16: return fc->nb_16X16[tx_type];
@@ -6537,6 +6549,9 @@ static int16_t *get_adapt_nb(FRAME_CONTEXT *fc, TX_SIZE tx_size,
 static uint32_t *get_non_zero_counts(FRAME_COUNTS *counts, TX_SIZE tx_size,
                                      TX_TYPE tx_type) {
   switch (tx_size) {
+#if CONFIG_CB4X4
+    case TX_2X2: return counts->non_zero_count_2x2[tx_type];
+#endif
     case TX_4X4: return counts->non_zero_count_4X4[tx_type];
     case TX_8X8: return counts->non_zero_count_8X8[tx_type];
     case TX_16X16: return counts->non_zero_count_16X16[tx_type];
@@ -6716,7 +6731,7 @@ void av1_update_scan_order_facade(AV1_COMMON *cm, TX_SIZE tx_size,
 void av1_init_scan_order(AV1_COMMON *cm) {
   TX_SIZE tx_size;
   TX_TYPE tx_type;
-  for (tx_size = TX_4X4; tx_size < TX_SIZES; ++tx_size) {
+  for (tx_size = 0; tx_size < TX_SIZES; ++tx_size) {
     for (tx_type = DCT_DCT; tx_type < TX_TYPES; ++tx_type) {
       uint32_t *non_zero_prob = get_non_zero_prob(cm->fc, tx_size, tx_type);
       const int tx2d_size = tx_size_2d[tx_size];