Merge "No need to store the deringing filter direction variance in an array" into nextgenv2
diff --git a/av1/common/av1_fwd_txfm2d_cfg.h b/av1/common/av1_fwd_txfm2d_cfg.h
index 49d324d..5a7c218 100644
--- a/av1/common/av1_fwd_txfm2d_cfg.h
+++ b/av1/common/av1_fwd_txfm2d_cfg.h
@@ -109,7 +109,7 @@
}; // .txfm_type_row
// ---------------- config fwd_dct_dct_64 ----------------
-static const int8_t fwd_shift_dct_dct_64[3] = { 2, -2, -2 };
+static const int8_t fwd_shift_dct_dct_64[3] = { 0, -2, -2 };
static const int8_t fwd_stage_range_col_dct_dct_64[12] = {
13, 14, 15, 16, 17, 18, 19, 19, 19, 19, 19, 19
};
diff --git a/av1/common/av1_rtcd_defs.pl b/av1/common/av1_rtcd_defs.pl
index f3b1328..597d5b2 100644
--- a/av1/common/av1_rtcd_defs.pl
+++ b/av1/common/av1_rtcd_defs.pl
@@ -61,25 +61,23 @@
add_proto qw/void av1_iht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht4x4_16_add/;
- if (aom_config("CONFIG_EXT_TX") eq "yes") {
- add_proto qw/void av1_iht4x8_32_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ add_proto qw/void av1_iht4x8_32_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht4x8_32_add/;
- add_proto qw/void av1_iht8x4_32_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ add_proto qw/void av1_iht8x4_32_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht8x4_32_add/;
- add_proto qw/void av1_iht8x16_128_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ add_proto qw/void av1_iht8x16_128_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht8x16_128_add/;
- add_proto qw/void av1_iht16x8_128_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ add_proto qw/void av1_iht16x8_128_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht16x8_128_add/;
- add_proto qw/void av1_iht16x32_512_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ add_proto qw/void av1_iht16x32_512_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht16x32_512_add/;
- add_proto qw/void av1_iht32x16_512_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ add_proto qw/void av1_iht32x16_512_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht32x16_512_add/;
- }
add_proto qw/void av1_iht8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht8x8_64_add/;
@@ -90,25 +88,23 @@
add_proto qw/void av1_iht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht4x4_16_add sse2/;
- if (aom_config("CONFIG_EXT_TX") eq "yes") {
- add_proto qw/void av1_iht4x8_32_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ add_proto qw/void av1_iht4x8_32_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht4x8_32_add sse2/;
- add_proto qw/void av1_iht8x4_32_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ add_proto qw/void av1_iht8x4_32_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht8x4_32_add sse2/;
- add_proto qw/void av1_iht8x16_128_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ add_proto qw/void av1_iht8x16_128_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht8x16_128_add sse2/;
- add_proto qw/void av1_iht16x8_128_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ add_proto qw/void av1_iht16x8_128_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht16x8_128_add sse2/;
- add_proto qw/void av1_iht16x32_512_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ add_proto qw/void av1_iht16x32_512_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht16x32_512_add sse2/;
- add_proto qw/void av1_iht32x16_512_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ add_proto qw/void av1_iht32x16_512_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht32x16_512_add sse2/;
- }
add_proto qw/void av1_iht8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht8x8_64_add sse2/;
@@ -122,25 +118,23 @@
add_proto qw/void av1_iht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht4x4_16_add/;
- if (aom_config("CONFIG_EXT_TX") eq "yes") {
- add_proto qw/void av1_iht4x8_32_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ add_proto qw/void av1_iht4x8_32_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht4x8_32_add/;
- add_proto qw/void av1_iht8x4_32_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ add_proto qw/void av1_iht8x4_32_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht8x4_32_add/;
- add_proto qw/void av1_iht8x16_128_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ add_proto qw/void av1_iht8x16_128_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht8x16_128_add/;
- add_proto qw/void av1_iht16x8_128_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ add_proto qw/void av1_iht16x8_128_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht16x8_128_add/;
- add_proto qw/void av1_iht16x32_512_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ add_proto qw/void av1_iht16x32_512_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht16x32_512_add/;
- add_proto qw/void av1_iht32x16_512_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ add_proto qw/void av1_iht32x16_512_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht32x16_512_add/;
- }
add_proto qw/void av1_iht8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht8x8_64_add/;
@@ -151,25 +145,23 @@
add_proto qw/void av1_iht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht4x4_16_add sse2 neon dspr2/;
- if (aom_config("CONFIG_EXT_TX") eq "yes") {
- add_proto qw/void av1_iht4x8_32_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ add_proto qw/void av1_iht4x8_32_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht4x8_32_add sse2/;
- add_proto qw/void av1_iht8x4_32_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ add_proto qw/void av1_iht8x4_32_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht8x4_32_add sse2/;
- add_proto qw/void av1_iht8x16_128_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ add_proto qw/void av1_iht8x16_128_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht8x16_128_add sse2/;
- add_proto qw/void av1_iht16x8_128_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ add_proto qw/void av1_iht16x8_128_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht16x8_128_add sse2/;
- add_proto qw/void av1_iht16x32_512_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ add_proto qw/void av1_iht16x32_512_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht16x32_512_add sse2/;
- add_proto qw/void av1_iht32x16_512_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
+ add_proto qw/void av1_iht32x16_512_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht32x16_512_add sse2/;
- }
add_proto qw/void av1_iht8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type";
specialize qw/av1_iht8x8_64_add sse2 neon dspr2/;
@@ -283,25 +275,23 @@
add_proto qw/void av1_highbd_iht4x4_16_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type, int bd";
specialize qw/av1_highbd_iht4x4_16_add/;
- if (aom_config("CONFIG_EXT_TX") eq "yes") {
- add_proto qw/void av1_highbd_iht4x8_32_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type, int bd";
+ add_proto qw/void av1_highbd_iht4x8_32_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type, int bd";
specialize qw/av1_highbd_iht4x8_32_add/;
- add_proto qw/void av1_highbd_iht8x4_32_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type, int bd";
+ add_proto qw/void av1_highbd_iht8x4_32_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type, int bd";
specialize qw/av1_highbd_iht8x4_32_add/;
- add_proto qw/void av1_highbd_iht8x16_128_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type, int bd";
+ add_proto qw/void av1_highbd_iht8x16_128_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type, int bd";
specialize qw/av1_highbd_iht8x16_128_add/;
- add_proto qw/void av1_highbd_iht16x8_128_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type, int bd";
+ add_proto qw/void av1_highbd_iht16x8_128_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type, int bd";
specialize qw/av1_highbd_iht16x8_128_add/;
- add_proto qw/void av1_highbd_iht16x32_512_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type, int bd";
+ add_proto qw/void av1_highbd_iht16x32_512_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type, int bd";
specialize qw/av1_highbd_iht16x32_512_add/;
- add_proto qw/void av1_highbd_iht32x16_512_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type, int bd";
+ add_proto qw/void av1_highbd_iht32x16_512_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type, int bd";
specialize qw/av1_highbd_iht32x16_512_add/;
- }
add_proto qw/void av1_highbd_iht8x8_64_add/, "const tran_low_t *input, uint8_t *dest, int dest_stride, int tx_type, int bd";
specialize qw/av1_highbd_iht8x8_64_add/;
@@ -394,26 +384,29 @@
add_proto qw/void av1_fht32x32/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
specialize qw/av1_fht32x32 avx2/;
-if (aom_config("CONFIG_EXT_TX") eq "yes") {
- add_proto qw/void av1_fht4x8/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
- specialize qw/av1_fht4x8 sse2/;
-
- add_proto qw/void av1_fht8x4/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
- specialize qw/av1_fht8x4 sse2/;
-
- add_proto qw/void av1_fht8x16/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
- specialize qw/av1_fht8x16 sse2/;
-
- add_proto qw/void av1_fht16x8/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
- specialize qw/av1_fht16x8 sse2/;
-
- add_proto qw/void av1_fht16x32/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
- specialize qw/av1_fht16x32 sse2/;
-
- add_proto qw/void av1_fht32x16/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
- specialize qw/av1_fht32x16 sse2/;
+if (aom_config("CONFIG_TX64X64") eq "yes") {
+ add_proto qw/void av1_fht64x64/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
+ specialize qw/av1_fht64x64/;
}
+add_proto qw/void av1_fht4x8/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
+specialize qw/av1_fht4x8 sse2/;
+
+add_proto qw/void av1_fht8x4/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
+specialize qw/av1_fht8x4 sse2/;
+
+add_proto qw/void av1_fht8x16/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
+specialize qw/av1_fht8x16 sse2/;
+
+add_proto qw/void av1_fht16x8/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
+specialize qw/av1_fht16x8 sse2/;
+
+add_proto qw/void av1_fht16x32/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
+specialize qw/av1_fht16x32 sse2/;
+
+add_proto qw/void av1_fht32x16/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
+specialize qw/av1_fht32x16 sse2/;
+
if (aom_config("CONFIG_AOM_HIGHBITDEPTH") ne "yes") {
if (aom_config("CONFIG_EXT_TX") ne "yes") {
specialize qw/av1_fht4x4 msa/;
@@ -536,6 +529,11 @@
add_proto qw/void av1_highbd_fht32x32/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
specialize qw/av1_highbd_fht32x32/;
+ if (aom_config("CONFIG_TX64X64") eq "yes") {
+ add_proto qw/void av1_highbd_fht64x64/, "const int16_t *input, tran_low_t *output, int stride, int tx_type";
+ specialize qw/av1_highbd_fht64x64/;
+ }
+
add_proto qw/void av1_highbd_fwht4x4/, "const int16_t *input, tran_low_t *output, int stride";
specialize qw/av1_highbd_fwht4x4/;
diff --git a/av1/common/enums.h b/av1/common/enums.h
index 35b9e5f..2ec83ec 100644
--- a/av1/common/enums.h
+++ b/av1/common/enums.h
@@ -134,18 +134,16 @@
#if CONFIG_CB4X4
TX_2X2, // 2x2 transform
#endif
- TX_4X4, // 4x4 transform
- TX_8X8, // 8x8 transform
- TX_16X16, // 16x16 transform
- TX_32X32, // 32x32 transform
-#if CONFIG_EXT_TX
+ TX_4X4, // 4x4 transform
+ TX_8X8, // 8x8 transform
+ TX_16X16, // 16x16 transform
+ TX_32X32, // 32x32 transform
TX_4X8, // 4x8 transform
TX_8X4, // 8x4 transform
TX_8X16, // 8x16 transform
TX_16X8, // 16x8 transform
TX_16X32, // 16x32 transform
TX_32X16, // 32x16 transform
-#endif // CONFIG_EXT_TX
TX_SIZES_ALL, // Includes rectangular transforms
TX_SIZES = TX_32X32 + 1, // Does NOT include rectangular transforms
TX_INVALID = 255 // Invalid transform size
diff --git a/av1/common/idct.c b/av1/common/idct.c
index 96b34ee..223c577 100644
--- a/av1/common/idct.c
+++ b/av1/common/idct.c
@@ -58,6 +58,7 @@
int i;
for (i = 0; i < 32; ++i) output[i] = input[i] * 4;
}
+#endif // CONFIG_EXT_TX
// For use in lieu of ADST
static void ihalfright32_c(const tran_low_t *input, tran_low_t *output) {
@@ -74,7 +75,49 @@
// Note overall scaling factor is 4 times orthogonal
}
+#if CONFIG_TX64X64
+static void idct64_col_c(const tran_low_t *input, tran_low_t *output) {
+ int32_t in[64], out[64];
+ int i;
+ for (i = 0; i < 64; ++i) in[i] = (int32_t)input[i];
+ av1_idct64_new(in, out, inv_cos_bit_col_dct_dct_64,
+ inv_stage_range_col_dct_dct_64);
+ for (i = 0; i < 64; ++i) output[i] = (tran_low_t)out[i];
+}
+
+static void idct64_row_c(const tran_low_t *input, tran_low_t *output) {
+ int32_t in[64], out[64];
+ int i;
+ for (i = 0; i < 64; ++i) in[i] = (int32_t)input[i];
+ av1_idct64_new(in, out, inv_cos_bit_row_dct_dct_64,
+ inv_stage_range_row_dct_dct_64);
+ for (i = 0; i < 64; ++i) output[i] = (tran_low_t)out[i];
+}
+
+static void iidtx64_c(const tran_low_t *input, tran_low_t *output) {
+ int i;
+ for (i = 0; i < 64; ++i)
+ output[i] = (tran_low_t)dct_const_round_shift(input[i] * 4 * Sqrt2);
+}
+
+// For use in lieu of ADST
+static void ihalfright64_c(const tran_low_t *input, tran_low_t *output) {
+ int i;
+ tran_low_t inputhalf[32];
+ // Multiply input by sqrt(2)
+ for (i = 0; i < 32; ++i) {
+ inputhalf[i] = (tran_low_t)dct_const_round_shift(input[i] * Sqrt2);
+ }
+ for (i = 0; i < 32; ++i) {
+ output[i] = (tran_low_t)dct_const_round_shift(input[32 + i] * 4 * Sqrt2);
+ }
+ aom_idct32_c(inputhalf, output + 32);
+ // Note overall scaling factor is 4 * sqrt(2) times orthogonal
+}
+#endif // CONFIG_TX64X64
+
#if CONFIG_AOM_HIGHBITDEPTH
+#if CONFIG_EXT_TX
static void highbd_iidtx4_c(const tran_low_t *input, tran_low_t *output,
int bd) {
int i;
@@ -104,6 +147,7 @@
(void)bd;
for (i = 0; i < 32; ++i) output[i] = input[i] * 4;
}
+#endif // CONFIG_EXT_TX
static void highbd_ihalfright32_c(const tran_low_t *input, tran_low_t *output,
int bd) {
@@ -120,9 +164,62 @@
aom_highbd_idct16_c(inputhalf, output + 16, bd);
// Note overall scaling factor is 4 times orthogonal
}
+
+#if CONFIG_EXT_TX
+#if CONFIG_TX64X64
+static void highbd_iidtx64_c(const tran_low_t *input, tran_low_t *output,
+ int bd) {
+ int i;
+ for (i = 0; i < 64; ++i)
+ output[i] =
+ HIGHBD_WRAPLOW(highbd_dct_const_round_shift(input[i] * 4 * Sqrt2), bd);
+}
+
+// For use in lieu of ADST
+static void highbd_ihalfright64_c(const tran_low_t *input, tran_low_t *output,
+ int bd) {
+ int i;
+ tran_low_t inputhalf[32];
+ // Multiply input by sqrt(2)
+ for (i = 0; i < 32; ++i) {
+ inputhalf[i] =
+ HIGHBD_WRAPLOW(highbd_dct_const_round_shift(input[i] * Sqrt2), bd);
+ }
+ for (i = 0; i < 32; ++i) {
+ output[i] = HIGHBD_WRAPLOW(
+ highbd_dct_const_round_shift(input[32 + i] * 4 * Sqrt2), bd);
+ }
+ aom_highbd_idct32_c(inputhalf, output + 32, bd);
+ // Note overall scaling factor is 4 * sqrt(2) times orthogonal
+}
+
+static void highbd_idct64_col_c(const tran_low_t *input, tran_low_t *output,
+ int bd) {
+ int32_t in[64], out[64];
+ int i;
+ (void)bd;
+ for (i = 0; i < 64; ++i) in[i] = (int32_t)input[i];
+ av1_idct64_new(in, out, inv_cos_bit_col_dct_dct_64,
+ inv_stage_range_col_dct_dct_64);
+ for (i = 0; i < 64; ++i) output[i] = (tran_low_t)out[i];
+}
+
+static void highbd_idct64_row_c(const tran_low_t *input, tran_low_t *output,
+ int bd) {
+ int32_t in[64], out[64];
+ int i;
+ (void)bd;
+ for (i = 0; i < 64; ++i) in[i] = (int32_t)input[i];
+ av1_idct64_new(in, out, inv_cos_bit_row_dct_dct_64,
+ inv_stage_range_row_dct_dct_64);
+ for (i = 0; i < 64; ++i) output[i] = (tran_low_t)out[i];
+}
+#endif // CONFIG_TX64X64
+#endif // CONFIG_EXT_TX
#endif // CONFIG_AOM_HIGHBITDEPTH
// Inverse identity transform and add.
+#if CONFIG_EXT_TX
static void inv_idtx_add_c(const tran_low_t *input, uint8_t *dest, int stride,
int bs, int tx_type) {
int r, c;
@@ -136,6 +233,7 @@
}
}
}
+#endif // CONFIG_EXT_TX
#define FLIPUD_PTR(dest, stride, size) \
do { \
@@ -143,6 +241,7 @@
(stride) = -(stride); \
} while (0)
+#if CONFIG_EXT_TX
static void maybe_flip_strides(uint8_t **dst, int *dstride, tran_low_t **src,
int *sstride, int tx_type, int sizey,
int sizex) {
@@ -180,8 +279,10 @@
default: assert(0); break;
}
}
+#endif // CONFIG_EXT_TX
#if CONFIG_AOM_HIGHBITDEPTH
+#if CONFIG_EXT_TX
static void highbd_inv_idtx_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int bs, int tx_type, int bd) {
int r, c;
@@ -235,8 +336,8 @@
default: assert(0); break;
}
}
-#endif // CONFIG_AOM_HIGHBITDEPTH
#endif // CONFIG_EXT_TX
+#endif // CONFIG_AOM_HIGHBITDEPTH
void av1_iht4x4_16_add_c(const tran_low_t *input, uint8_t *dest, int stride,
int tx_type) {
@@ -301,7 +402,6 @@
}
}
-#if CONFIG_EXT_TX
void av1_iht4x8_32_add_c(const tran_low_t *input, uint8_t *dest, int stride,
int tx_type) {
static const transform_2d IHT_4x8[] = {
@@ -309,6 +409,7 @@
{ aom_iadst8_c, aom_idct4_c }, // ADST_DCT
{ aom_idct8_c, aom_iadst4_c }, // DCT_ADST
{ aom_iadst8_c, aom_iadst4_c }, // ADST_ADST
+#if CONFIG_EXT_TX
{ aom_iadst8_c, aom_idct4_c }, // FLIPADST_DCT
{ aom_idct8_c, aom_iadst4_c }, // DCT_FLIPADST
{ aom_iadst8_c, aom_iadst4_c }, // FLIPADST_FLIPADST
@@ -321,6 +422,7 @@
{ iidtx8_c, aom_iadst4_c }, // H_ADST
{ aom_iadst8_c, iidtx4_c }, // V_FLIPADST
{ iidtx8_c, aom_iadst4_c }, // H_FLIPADST
+#endif
};
const int n = 4;
@@ -343,7 +445,9 @@
IHT_4x8[tx_type].cols(out[i], out[i]);
}
+#if CONFIG_EXT_TX
maybe_flip_strides(&dest, &stride, &outp, &outstride, tx_type, n2, n);
+#endif
// Sum with the destination
for (i = 0; i < n2; ++i) {
@@ -362,6 +466,7 @@
{ aom_iadst4_c, aom_idct8_c }, // ADST_DCT
{ aom_idct4_c, aom_iadst8_c }, // DCT_ADST
{ aom_iadst4_c, aom_iadst8_c }, // ADST_ADST
+#if CONFIG_EXT_TX
{ aom_iadst4_c, aom_idct8_c }, // FLIPADST_DCT
{ aom_idct4_c, aom_iadst8_c }, // DCT_FLIPADST
{ aom_iadst4_c, aom_iadst8_c }, // FLIPADST_FLIPADST
@@ -374,6 +479,7 @@
{ iidtx4_c, aom_iadst8_c }, // H_ADST
{ aom_iadst4_c, iidtx8_c }, // V_FLIPADST
{ iidtx4_c, aom_iadst8_c }, // H_FLIPADST
+#endif
};
const int n = 4;
const int n2 = 8;
@@ -396,7 +502,9 @@
IHT_8x4[tx_type].cols(out[i], out[i]);
}
+#if CONFIG_EXT_TX
maybe_flip_strides(&dest, &stride, &outp, &outstride, tx_type, n, n2);
+#endif
// Sum with the destination
for (i = 0; i < n; ++i) {
@@ -415,6 +523,7 @@
{ aom_iadst16_c, aom_idct8_c }, // ADST_DCT
{ aom_idct16_c, aom_iadst8_c }, // DCT_ADST
{ aom_iadst16_c, aom_iadst8_c }, // ADST_ADST
+#if CONFIG_EXT_TX
{ aom_iadst16_c, aom_idct8_c }, // FLIPADST_DCT
{ aom_idct16_c, aom_iadst8_c }, // DCT_FLIPADST
{ aom_iadst16_c, aom_iadst8_c }, // FLIPADST_FLIPADST
@@ -427,6 +536,7 @@
{ iidtx16_c, aom_iadst8_c }, // H_ADST
{ aom_iadst16_c, iidtx8_c }, // V_FLIPADST
{ iidtx16_c, aom_iadst8_c }, // H_FLIPADST
+#endif
};
const int n = 8;
@@ -449,7 +559,9 @@
IHT_8x16[tx_type].cols(out[i], out[i]);
}
+#if CONFIG_EXT_TX
maybe_flip_strides(&dest, &stride, &outp, &outstride, tx_type, n2, n);
+#endif
// Sum with the destination
for (i = 0; i < n2; ++i) {
@@ -468,6 +580,7 @@
{ aom_iadst8_c, aom_idct16_c }, // ADST_DCT
{ aom_idct8_c, aom_iadst16_c }, // DCT_ADST
{ aom_iadst8_c, aom_iadst16_c }, // ADST_ADST
+#if CONFIG_EXT_TX
{ aom_iadst8_c, aom_idct16_c }, // FLIPADST_DCT
{ aom_idct8_c, aom_iadst16_c }, // DCT_FLIPADST
{ aom_iadst8_c, aom_iadst16_c }, // FLIPADST_FLIPADST
@@ -480,6 +593,7 @@
{ iidtx8_c, aom_iadst16_c }, // H_ADST
{ aom_iadst8_c, iidtx16_c }, // V_FLIPADST
{ iidtx8_c, aom_iadst16_c }, // H_FLIPADST
+#endif
};
const int n = 8;
const int n2 = 16;
@@ -502,7 +616,9 @@
IHT_16x8[tx_type].cols(out[i], out[i]);
}
+#if CONFIG_EXT_TX
maybe_flip_strides(&dest, &stride, &outp, &outstride, tx_type, n, n2);
+#endif
// Sum with the destination
for (i = 0; i < n; ++i) {
@@ -521,6 +637,7 @@
{ ihalfright32_c, aom_idct16_c }, // ADST_DCT
{ aom_idct32_c, aom_iadst16_c }, // DCT_ADST
{ ihalfright32_c, aom_iadst16_c }, // ADST_ADST
+#if CONFIG_EXT_TX
{ ihalfright32_c, aom_idct16_c }, // FLIPADST_DCT
{ aom_idct32_c, aom_iadst16_c }, // DCT_FLIPADST
{ ihalfright32_c, aom_iadst16_c }, // FLIPADST_FLIPADST
@@ -533,6 +650,7 @@
{ iidtx32_c, aom_iadst16_c }, // H_ADST
{ ihalfright32_c, iidtx16_c }, // V_FLIPADST
{ iidtx32_c, aom_iadst16_c }, // H_FLIPADST
+#endif
};
const int n = 16;
@@ -555,7 +673,9 @@
IHT_16x32[tx_type].cols(out[i], out[i]);
}
+#if CONFIG_EXT_TX
maybe_flip_strides(&dest, &stride, &outp, &outstride, tx_type, n2, n);
+#endif
// Sum with the destination
for (i = 0; i < n2; ++i) {
@@ -574,6 +694,7 @@
{ aom_iadst16_c, aom_idct32_c }, // ADST_DCT
{ aom_idct16_c, ihalfright32_c }, // DCT_ADST
{ aom_iadst16_c, ihalfright32_c }, // ADST_ADST
+#if CONFIG_EXT_TX
{ aom_iadst16_c, aom_idct32_c }, // FLIPADST_DCT
{ aom_idct16_c, ihalfright32_c }, // DCT_FLIPADST
{ aom_iadst16_c, ihalfright32_c }, // FLIPADST_FLIPADST
@@ -586,6 +707,7 @@
{ iidtx16_c, ihalfright32_c }, // H_ADST
{ aom_iadst16_c, iidtx32_c }, // V_FLIPADST
{ iidtx16_c, ihalfright32_c }, // H_FLIPADST
+#endif
};
const int n = 16;
const int n2 = 32;
@@ -608,7 +730,9 @@
IHT_32x16[tx_type].cols(out[i], out[i]);
}
+#if CONFIG_EXT_TX
maybe_flip_strides(&dest, &stride, &outp, &outstride, tx_type, n, n2);
+#endif
// Sum with the destination
for (i = 0; i < n; ++i) {
@@ -619,7 +743,6 @@
}
}
}
-#endif // CONFIG_EXT_TX
void av1_iht8x8_64_add_c(const tran_low_t *input, uint8_t *dest, int stride,
int tx_type) {
@@ -763,10 +886,10 @@
{ iidtx32_c, iidtx32_c }, // IDTX
{ aom_idct32_c, iidtx32_c }, // V_DCT
{ iidtx32_c, aom_idct32_c }, // H_DCT
- { ihalfright32_c, iidtx16_c }, // V_ADST
- { iidtx16_c, ihalfright32_c }, // H_ADST
- { ihalfright32_c, iidtx16_c }, // V_FLIPADST
- { iidtx16_c, ihalfright32_c }, // H_FLIPADST
+ { ihalfright32_c, iidtx32_c }, // V_ADST
+ { iidtx32_c, ihalfright32_c }, // H_ADST
+ { ihalfright32_c, iidtx32_c }, // V_FLIPADST
+ { iidtx32_c, ihalfright32_c }, // H_FLIPADST
};
int i, j;
@@ -806,6 +929,68 @@
}
}
}
+
+#if CONFIG_TX64X64
+void av1_iht64x64_4096_add_c(const tran_low_t *input, uint8_t *dest, int stride,
+ int tx_type) {
+ static const transform_2d IHT_64[] = {
+ { idct64_col_c, idct64_row_c }, // DCT_DCT
+ { ihalfright64_c, idct64_row_c }, // ADST_DCT
+ { idct64_col_c, ihalfright64_c }, // DCT_ADST
+ { ihalfright64_c, ihalfright64_c }, // ADST_ADST
+ { ihalfright64_c, idct64_row_c }, // FLIPADST_DCT
+ { idct64_col_c, ihalfright64_c }, // DCT_FLIPADST
+ { ihalfright64_c, ihalfright64_c }, // FLIPADST_FLIPADST
+ { ihalfright64_c, ihalfright64_c }, // ADST_FLIPADST
+ { ihalfright64_c, ihalfright64_c }, // FLIPADST_ADST
+ { iidtx64_c, iidtx64_c }, // IDTX
+ { idct64_col_c, iidtx64_c }, // V_DCT
+ { iidtx64_c, idct64_row_c }, // H_DCT
+ { ihalfright64_c, iidtx64_c }, // V_ADST
+ { iidtx64_c, ihalfright64_c }, // H_ADST
+ { ihalfright64_c, iidtx64_c }, // V_FLIPADST
+ { iidtx64_c, ihalfright64_c }, // H_FLIPADST
+ };
+
+ int i, j;
+ tran_low_t tmp;
+ tran_low_t out[64][64];
+ tran_low_t *outp = &out[0][0];
+ int outstride = 64;
+
+ // inverse transform row vectors
+ for (i = 0; i < 64; ++i) {
+ IHT_64[tx_type].rows(input, out[i]);
+ for (j = 0; j < 64; ++j) out[i][j] = ROUND_POWER_OF_TWO(out[i][j], 1);
+ input += 64;
+ }
+
+ // transpose
+ for (i = 1; i < 64; i++) {
+ for (j = 0; j < i; j++) {
+ tmp = out[i][j];
+ out[i][j] = out[j][i];
+ out[j][i] = tmp;
+ }
+ }
+
+ // inverse transform column vectors
+ for (i = 0; i < 64; ++i) {
+ IHT_64[tx_type].cols(out[i], out[i]);
+ }
+
+ maybe_flip_strides(&dest, &stride, &outp, &outstride, tx_type, 64, 64);
+
+ // Sum with the destination
+ for (i = 0; i < 64; ++i) {
+ for (j = 0; j < 64; ++j) {
+ int d = i * stride + j;
+ int s = j * outstride + i;
+ dest[d] = clip_pixel_add(dest[d], ROUND_POWER_OF_TWO(outp[s], 5));
+ }
+ }
+}
+#endif // CONFIG_TX64X64
#endif // CONFIG_EXT_TX
// idct
@@ -905,7 +1090,6 @@
}
}
-#if CONFIG_EXT_TX
void av1_inv_txfm_add_4x8(const tran_low_t *input, uint8_t *dest, int stride,
int eob, TX_TYPE tx_type) {
(void)eob;
@@ -941,7 +1125,6 @@
(void)eob;
av1_iht32x16_512_add(input, dest, stride, tx_type);
}
-#endif // CONFIG_EXT_TX
void av1_inv_txfm_add_8x8(const tran_low_t *input, uint8_t *dest, int stride,
int eob, TX_TYPE tx_type) {
@@ -1090,7 +1273,6 @@
}
}
-#if CONFIG_EXT_TX
void av1_highbd_iht4x8_32_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int tx_type, int bd) {
static const highbd_transform_2d HIGH_IHT_4x8[] = {
@@ -1098,6 +1280,7 @@
{ aom_highbd_iadst8_c, aom_highbd_idct4_c }, // ADST_DCT
{ aom_highbd_idct8_c, aom_highbd_iadst4_c }, // DCT_ADST
{ aom_highbd_iadst8_c, aom_highbd_iadst4_c }, // ADST_ADST
+#if CONFIG_EXT_TX
{ aom_highbd_iadst8_c, aom_highbd_idct4_c }, // FLIPADST_DCT
{ aom_highbd_idct8_c, aom_highbd_iadst4_c }, // DCT_FLIPADST
{ aom_highbd_iadst8_c, aom_highbd_iadst4_c }, // FLIPADST_FLIPADST
@@ -1110,6 +1293,7 @@
{ highbd_iidtx8_c, aom_highbd_iadst4_c }, // H_ADST
{ aom_highbd_iadst8_c, highbd_iidtx4_c }, // V_FLIPADST
{ highbd_iidtx8_c, aom_highbd_iadst4_c }, // H_FLIPADST
+#endif // CONFIG_EXT_TX
};
const int n = 4;
const int n2 = 8;
@@ -1136,7 +1320,9 @@
HIGH_IHT_4x8[tx_type].cols(out[i], out[i], bd);
}
+#if CONFIG_EXT_TX
maybe_flip_strides16(&dest, &stride, &outp, &outstride, tx_type, n2, n);
+#endif // CONFIG_EXT_TX
// Sum with the destination
for (i = 0; i < n2; ++i) {
@@ -1156,6 +1342,7 @@
{ aom_highbd_iadst4_c, aom_highbd_idct8_c }, // ADST_DCT
{ aom_highbd_idct4_c, aom_highbd_iadst8_c }, // DCT_ADST
{ aom_highbd_iadst4_c, aom_highbd_iadst8_c }, // ADST_ADST
+#if CONFIG_EXT_TX
{ aom_highbd_iadst4_c, aom_highbd_idct8_c }, // FLIPADST_DCT
{ aom_highbd_idct4_c, aom_highbd_iadst8_c }, // DCT_FLIPADST
{ aom_highbd_iadst4_c, aom_highbd_iadst8_c }, // FLIPADST_FLIPADST
@@ -1168,6 +1355,7 @@
{ highbd_iidtx4_c, aom_highbd_iadst8_c }, // H_ADST
{ aom_highbd_iadst4_c, highbd_iidtx8_c }, // V_FLIPADST
{ highbd_iidtx4_c, aom_highbd_iadst8_c }, // H_FLIPADST
+#endif // CONFIG_EXT_TX
};
const int n = 4;
const int n2 = 8;
@@ -1194,7 +1382,9 @@
HIGH_IHT_8x4[tx_type].cols(out[i], out[i], bd);
}
+#if CONFIG_EXT_TX
maybe_flip_strides16(&dest, &stride, &outp, &outstride, tx_type, n, n2);
+#endif // CONFIG_EXT_TX
// Sum with the destination
for (i = 0; i < n; ++i) {
@@ -1214,6 +1404,7 @@
{ aom_highbd_iadst16_c, aom_highbd_idct8_c }, // ADST_DCT
{ aom_highbd_idct16_c, aom_highbd_iadst8_c }, // DCT_ADST
{ aom_highbd_iadst16_c, aom_highbd_iadst8_c }, // ADST_ADST
+#if CONFIG_EXT_TX
{ aom_highbd_iadst16_c, aom_highbd_idct8_c }, // FLIPADST_DCT
{ aom_highbd_idct16_c, aom_highbd_iadst8_c }, // DCT_FLIPADST
{ aom_highbd_iadst16_c, aom_highbd_iadst8_c }, // FLIPADST_FLIPADST
@@ -1226,6 +1417,7 @@
{ highbd_iidtx16_c, aom_highbd_iadst8_c }, // H_ADST
{ aom_highbd_iadst16_c, highbd_iidtx8_c }, // V_FLIPADST
{ highbd_iidtx16_c, aom_highbd_iadst8_c }, // H_FLIPADST
+#endif // CONFIG_EXT_TX
};
const int n = 8;
const int n2 = 16;
@@ -1251,7 +1443,9 @@
HIGH_IHT_8x16[tx_type].cols(out[i], out[i], bd);
}
+#if CONFIG_EXT_TX
maybe_flip_strides16(&dest, &stride, &outp, &outstride, tx_type, n2, n);
+#endif // CONFIG_EXT_TX
// Sum with the destination
for (i = 0; i < n2; ++i) {
@@ -1271,6 +1465,7 @@
{ aom_highbd_iadst8_c, aom_highbd_idct16_c }, // ADST_DCT
{ aom_highbd_idct8_c, aom_highbd_iadst16_c }, // DCT_ADST
{ aom_highbd_iadst8_c, aom_highbd_iadst16_c }, // ADST_ADST
+#if CONFIG_EXT_TX
{ aom_highbd_iadst8_c, aom_highbd_idct16_c }, // FLIPADST_DCT
{ aom_highbd_idct8_c, aom_highbd_iadst16_c }, // DCT_FLIPADST
{ aom_highbd_iadst8_c, aom_highbd_iadst16_c }, // FLIPADST_FLIPADST
@@ -1283,6 +1478,7 @@
{ highbd_iidtx8_c, aom_highbd_iadst16_c }, // H_ADST
{ aom_highbd_iadst8_c, highbd_iidtx16_c }, // V_FLIPADST
{ highbd_iidtx8_c, aom_highbd_iadst16_c }, // H_FLIPADST
+#endif // CONFIG_EXT_TX
};
const int n = 8;
const int n2 = 16;
@@ -1308,7 +1504,9 @@
HIGH_IHT_16x8[tx_type].cols(out[i], out[i], bd);
}
+#if CONFIG_EXT_TX
maybe_flip_strides16(&dest, &stride, &outp, &outstride, tx_type, n, n2);
+#endif // CONFIG_EXT_TX
// Sum with the destination
for (i = 0; i < n; ++i) {
@@ -1328,6 +1526,7 @@
{ highbd_ihalfright32_c, aom_highbd_idct16_c }, // ADST_DCT
{ aom_highbd_idct32_c, aom_highbd_iadst16_c }, // DCT_ADST
{ highbd_ihalfright32_c, aom_highbd_iadst16_c }, // ADST_ADST
+#if CONFIG_EXT_TX
{ highbd_ihalfright32_c, aom_highbd_idct16_c }, // FLIPADST_DCT
{ aom_highbd_idct32_c, aom_highbd_iadst16_c }, // DCT_FLIPADST
{ highbd_ihalfright32_c, aom_highbd_iadst16_c }, // FLIPADST_FLIPADST
@@ -1340,6 +1539,7 @@
{ highbd_iidtx32_c, aom_highbd_iadst16_c }, // H_ADST
{ highbd_ihalfright32_c, highbd_iidtx16_c }, // V_FLIPADST
{ highbd_iidtx32_c, aom_highbd_iadst16_c }, // H_FLIPADST
+#endif // CONFIG_EXT_TX
};
const int n = 16;
const int n2 = 32;
@@ -1365,7 +1565,9 @@
HIGH_IHT_16x32[tx_type].cols(out[i], out[i], bd);
}
+#if CONFIG_EXT_TX
maybe_flip_strides16(&dest, &stride, &outp, &outstride, tx_type, n2, n);
+#endif // CONFIG_EXT_TX
// Sum with the destination
for (i = 0; i < n2; ++i) {
@@ -1385,6 +1587,7 @@
{ aom_highbd_iadst16_c, aom_highbd_idct32_c }, // ADST_DCT
{ aom_highbd_idct16_c, highbd_ihalfright32_c }, // DCT_ADST
{ aom_highbd_iadst16_c, highbd_ihalfright32_c }, // ADST_ADST
+#if CONFIG_EXT_TX
{ aom_highbd_iadst16_c, aom_highbd_idct32_c }, // FLIPADST_DCT
{ aom_highbd_idct16_c, highbd_ihalfright32_c }, // DCT_FLIPADST
{ aom_highbd_iadst16_c, highbd_ihalfright32_c }, // FLIPADST_FLIPADST
@@ -1397,6 +1600,7 @@
{ highbd_iidtx16_c, highbd_ihalfright32_c }, // H_ADST
{ aom_highbd_iadst16_c, highbd_iidtx32_c }, // V_FLIPADST
{ highbd_iidtx16_c, highbd_ihalfright32_c }, // H_FLIPADST
+#endif // CONFIG_EXT_TX
};
const int n = 16;
const int n2 = 32;
@@ -1422,7 +1626,9 @@
HIGH_IHT_32x16[tx_type].cols(out[i], out[i], bd);
}
+#if CONFIG_EXT_TX
maybe_flip_strides16(&dest, &stride, &outp, &outstride, tx_type, n, n2);
+#endif // CONFIG_EXT_TX
// Sum with the destination
for (i = 0; i < n; ++i) {
@@ -1434,7 +1640,6 @@
}
}
}
-#endif // CONFIG_EXT_TX
void av1_highbd_iht8x8_64_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int tx_type, int bd) {
@@ -1630,6 +1835,71 @@
}
}
}
+
+#if CONFIG_TX64X64
+void av1_highbd_iht64x64_4096_add_c(const tran_low_t *input, uint8_t *dest8,
+ int stride, int tx_type, int bd) {
+ static const highbd_transform_2d HIGH_IHT_64[] = {
+ { highbd_idct64_col_c, highbd_idct64_row_c }, // DCT_DCT
+ { highbd_ihalfright64_c, highbd_idct64_row_c }, // ADST_DCT
+ { highbd_idct64_col_c, highbd_ihalfright64_c }, // DCT_ADST
+ { highbd_ihalfright64_c, highbd_ihalfright64_c }, // ADST_ADST
+ { highbd_ihalfright64_c, highbd_idct64_row_c }, // FLIPADST_DCT
+ { highbd_idct64_col_c, highbd_ihalfright64_c }, // DCT_FLIPADST
+ { highbd_ihalfright64_c, highbd_ihalfright64_c }, // FLIPADST_FLIPADST
+ { highbd_ihalfright64_c, highbd_ihalfright64_c }, // ADST_FLIPADST
+ { highbd_ihalfright64_c, highbd_ihalfright64_c }, // FLIPADST_ADST
+ { highbd_iidtx64_c, highbd_iidtx64_c }, // IDTX
+ { highbd_idct64_col_c, highbd_iidtx64_c }, // V_DCT
+ { highbd_iidtx64_c, highbd_idct64_row_c }, // H_DCT
+ { highbd_ihalfright64_c, highbd_iidtx64_c }, // V_ADST
+ { highbd_iidtx64_c, highbd_ihalfright64_c }, // H_ADST
+ { highbd_ihalfright64_c, highbd_iidtx64_c }, // V_FLIPADST
+ { highbd_iidtx64_c, highbd_ihalfright64_c }, // H_FLIPADST
+ };
+
+ uint16_t *dest = CONVERT_TO_SHORTPTR(dest8);
+
+ int i, j;
+ tran_low_t tmp;
+ tran_low_t out[64][64];
+ tran_low_t *outp = &out[0][0];
+ int outstride = 64;
+
+ // inverse transform row vectors
+ for (i = 0; i < 64; ++i) {
+ HIGH_IHT_64[tx_type].rows(input, out[i], bd);
+ for (j = 0; j < 64; ++j) out[i][j] = ROUND_POWER_OF_TWO(out[i][j], 1);
+ input += 64;
+ }
+
+ // transpose
+ for (i = 1; i < 64; i++) {
+ for (j = 0; j < i; j++) {
+ tmp = out[i][j];
+ out[i][j] = out[j][i];
+ out[j][i] = tmp;
+ }
+ }
+
+ // inverse transform column vectors
+ for (i = 0; i < 64; ++i) {
+ HIGH_IHT_64[tx_type].cols(out[i], out[i], bd);
+ }
+
+ maybe_flip_strides16(&dest, &stride, &outp, &outstride, tx_type, 64, 64);
+
+ // Sum with the destination
+ for (i = 0; i < 64; ++i) {
+ for (j = 0; j < 64; ++j) {
+ int d = i * stride + j;
+ int s = j * outstride + i;
+ dest[d] =
+ highbd_clip_pixel_add(dest[d], ROUND_POWER_OF_TWO(outp[s], 5), bd);
+ }
+ }
+}
+#endif // CONFIG_TX64X64
#endif // CONFIG_EXT_TX
// idct
@@ -1737,7 +2007,6 @@
}
}
-#if CONFIG_EXT_TX
void av1_highbd_inv_txfm_add_4x8(const tran_low_t *input, uint8_t *dest,
int stride, int eob, int bd, TX_TYPE tx_type) {
(void)eob;
@@ -1777,7 +2046,6 @@
(void)eob;
av1_highbd_iht32x16_512_add_c(input, dest, stride, tx_type, bd);
}
-#endif // CONFIG_EXT_TX
void av1_highbd_inv_txfm_add_8x8(const tran_low_t *input, uint8_t *dest,
int stride, int eob, int bd, TX_TYPE tx_type) {
@@ -1904,7 +2172,6 @@
av1_inv_txfm_add_16x16(input, dest, stride, eob, tx_type);
break;
case TX_8X8: av1_inv_txfm_add_8x8(input, dest, stride, eob, tx_type); break;
-#if CONFIG_EXT_TX
case TX_4X8: av1_inv_txfm_add_4x8(input, dest, stride, eob, tx_type); break;
case TX_8X4: av1_inv_txfm_add_8x4(input, dest, stride, eob, tx_type); break;
case TX_8X16:
@@ -1919,7 +2186,6 @@
case TX_32X16:
av1_inv_txfm_add_32x16(input, dest, stride, eob, tx_type);
break;
-#endif // CONFIG_EXT_TX
case TX_4X4:
// this is like av1_short_idct4x4 but has a special case around eob<=1
// which is significant (not just an optimization) for the lossless
@@ -1949,7 +2215,6 @@
case TX_8X8:
av1_highbd_inv_txfm_add_8x8(input, dest, stride, eob, bd, tx_type);
break;
-#if CONFIG_EXT_TX
case TX_4X8:
av1_highbd_inv_txfm_add_4x8(input, dest, stride, eob, bd, tx_type);
break;
@@ -1968,7 +2233,6 @@
case TX_32X16:
av1_highbd_inv_txfm_add_32x16(input, dest, stride, eob, bd, tx_type);
break;
-#endif // CONFIG_EXT_TX
case TX_4X4:
// this is like av1_short_idct4x4 but has a special case around eob<=1
// which is significant (not just an optimization) for the lossless
diff --git a/av1/common/idct.h b/av1/common/idct.h
index 1acc825..db9a6e2 100644
--- a/av1/common/idct.h
+++ b/av1/common/idct.h
@@ -67,12 +67,10 @@
void av1_inv_txfm_add_4x4(const tran_low_t *input, uint8_t *dest, int stride,
int eob, TX_TYPE tx_type, int lossless);
-#if CONFIG_EXT_TX
void av1_inv_txfm_add_8x4(const tran_low_t *input, uint8_t *dest, int stride,
int eob, TX_TYPE tx_type);
void av1_inv_txfm_add_4x8(const tran_low_t *input, uint8_t *dest, int stride,
int eob, TX_TYPE tx_type);
-#endif // CONFIG_EXT_TX
void av1_inv_txfm_add_8x8(const tran_low_t *input, uint8_t *dest, int stride,
int eob, TX_TYPE tx_type);
void av1_inv_txfm_add_16x16(const tran_low_t *input, uint8_t *dest, int stride,
@@ -95,12 +93,10 @@
void av1_highbd_inv_txfm_add_4x4(const tran_low_t *input, uint8_t *dest,
int stride, int eob, int bd, TX_TYPE tx_type,
int lossless);
-#if CONFIG_EXT_TX
void av1_highbd_inv_txfm_add_4x8(const tran_low_t *input, uint8_t *dest,
int stride, int eob, int bd, TX_TYPE tx_type);
void av1_highbd_inv_txfm_add_8x4(const tran_low_t *input, uint8_t *dest,
int stride, int eob, int bd, TX_TYPE tx_type);
-#endif // CONFIG_EXT_TX
void av1_highbd_inv_txfm_add_8x8(const tran_low_t *input, uint8_t *dest,
int stride, int eob, int bd, TX_TYPE tx_type);
void av1_highbd_inv_txfm_add_16x16(const tran_low_t *input, uint8_t *dest,
diff --git a/av1/common/scan.c b/av1/common/scan.c
index cab3729..b2386b9 100644
--- a/av1/common/scan.c
+++ b/av1/common/scan.c
@@ -36,12 +36,12 @@
0, 1, 4, 2, 5, 3, 6, 8, 9, 7, 12, 10, 13, 11, 14, 15,
};
-#if CONFIG_EXT_TX
DECLARE_ALIGNED(16, static const int16_t, default_scan_4x8[32]) = {
0, 1, 4, 5, 2, 8, 6, 9, 10, 3, 12, 7, 13, 11, 14, 16,
17, 15, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31,
};
+#if CONFIG_EXT_TX
DECLARE_ALIGNED(16, static const int16_t, mcol_scan_4x8[32]) = {
0, 4, 8, 12, 16, 20, 24, 28, 1, 5, 9, 13, 17, 21, 25, 29,
2, 6, 10, 14, 18, 22, 26, 30, 3, 7, 11, 15, 19, 23, 27, 31,
@@ -51,12 +51,14 @@
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31,
};
+#endif
DECLARE_ALIGNED(16, static const int16_t, default_scan_8x4[32]) = {
0, 1, 8, 9, 2, 16, 10, 17, 18, 3, 24, 11, 25, 19, 26, 4,
12, 27, 20, 5, 28, 13, 21, 29, 6, 14, 22, 30, 7, 15, 23, 31,
};
+#if CONFIG_EXT_TX
DECLARE_ALIGNED(16, static const int16_t, mcol_scan_8x4[32]) = {
0, 8, 16, 24, 1, 9, 17, 25, 2, 10, 18, 26, 3, 11, 19, 27,
4, 12, 20, 28, 5, 13, 21, 29, 6, 14, 22, 30, 7, 15, 23, 31,
@@ -66,7 +68,7 @@
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31,
};
-#endif // CONFIG_EXT_TX
+#endif
DECLARE_ALIGNED(16, static const int16_t, default_scan_8x8[64]) = {
0, 8, 1, 16, 9, 2, 17, 24, 10, 3, 18, 25, 32, 11, 4, 26,
@@ -105,7 +107,6 @@
58, 45, 38, 52, 31, 59, 53, 46, 60, 39, 61, 47, 54, 55, 62, 63,
};
-#if CONFIG_EXT_TX
DECLARE_ALIGNED(16, static const int16_t, default_scan_8x16[128]) = {
0, 1, 8, 2, 9, 16, 3, 10, 17, 24, 4, 11, 18, 25, 32,
5, 12, 19, 26, 33, 40, 6, 13, 20, 27, 34, 41, 48, 7, 14,
@@ -129,6 +130,7 @@
122, 63, 78, 93, 108, 123, 79, 94, 109, 124, 95, 110, 125, 111, 126, 127,
};
+#if CONFIG_EXT_TX
DECLARE_ALIGNED(16, static const int16_t, mcol_scan_8x16[128]) = {
0, 8, 16, 24, 32, 40, 48, 56, 64, 72, 80, 88, 96, 104, 112, 120,
1, 9, 17, 25, 33, 41, 49, 57, 65, 73, 81, 89, 97, 105, 113, 121,
@@ -174,6 +176,7 @@
105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119,
120, 121, 122, 123, 124, 125, 126, 127,
};
+#endif
DECLARE_ALIGNED(16, static const int16_t, default_scan_16x32[512]) = {
0, 1, 16, 2, 17, 32, 3, 18, 33, 48, 4, 19, 34, 49, 64,
@@ -251,6 +254,7 @@
510, 511,
};
+#if CONFIG_EXT_TX
DECLARE_ALIGNED(16, static const int16_t, mcol_scan_16x32[512]) = {
0, 16, 32, 48, 64, 80, 96, 112, 128, 144, 160, 176, 192, 208, 224,
240, 256, 272, 288, 304, 320, 336, 352, 368, 384, 400, 416, 432, 448, 464,
@@ -1034,7 +1038,6 @@
8, 3, 6, 8, 9, 6, 9, 9, 12, 7, 10, 10, 13, 11, 14, 0, 0,
};
-#if CONFIG_EXT_TX
DECLARE_ALIGNED(16, static const int16_t,
default_scan_4x8_neighbors[33 * MAX_NEIGHBORS]) = {
0, 0, 0, 0, 0, 0, 1, 4, 1, 1, 4, 4, 2, 5, 5, 8, 6,
@@ -1043,6 +1046,7 @@
24, 22, 25, 23, 26, 24, 24, 25, 28, 26, 29, 27, 30, 0, 0
};
+#if CONFIG_EXT_TX
DECLARE_ALIGNED(16, static const int16_t,
mcol_scan_4x8_neighbors[33 * MAX_NEIGHBORS]) = {
0, 0, 0, 0, 4, 4, 8, 8, 12, 12, 16, 16, 20, 20, 24, 24, 0,
@@ -1058,6 +1062,7 @@
13, 16, 14, 17, 15, 18, 16, 16, 17, 20, 18, 21, 19, 22, 20, 20, 21,
24, 22, 25, 23, 26, 24, 24, 25, 28, 26, 29, 27, 30, 0, 0
};
+#endif
DECLARE_ALIGNED(16, static const int16_t,
default_scan_8x4_neighbors[33 * MAX_NEIGHBORS]) = {
@@ -1067,6 +1072,7 @@
13, 14, 21, 22, 29, 6, 6, 7, 14, 15, 22, 23, 30, 0, 0
};
+#if CONFIG_EXT_TX
DECLARE_ALIGNED(16, static const int16_t,
mcol_scan_8x4_neighbors[33 * MAX_NEIGHBORS]) = {
0, 0, 0, 0, 8, 8, 16, 16, 0, 0, 1, 8, 9, 16, 17, 24, 1,
@@ -1141,7 +1147,6 @@
31, 38, 53, 60, 46, 53, 39, 46, 54, 61, 47, 54, 55, 62, 0, 0,
};
-#if CONFIG_EXT_TX
DECLARE_ALIGNED(16, static const int16_t,
default_scan_8x16_neighbors[129 * MAX_NEIGHBORS]) = {
0, 0, 0, 0, 0, 0, 1, 1, 1, 8, 8, 8, 2, 2, 2,
@@ -1186,6 +1191,7 @@
126, 0, 0
};
+#if CONFIG_EXT_TX
DECLARE_ALIGNED(16, static const int16_t,
mcol_scan_8x16_neighbors[129 * MAX_NEIGHBORS]) = {
0, 0, 0, 0, 8, 8, 16, 16, 24, 24, 32, 32, 40, 40, 48, 48,
@@ -1271,6 +1277,7 @@
104, 119, 105, 120, 106, 121, 107, 122, 108, 123, 109, 124, 110, 125, 111,
126, 0, 0
};
+#endif
DECLARE_ALIGNED(16, static const int16_t,
default_scan_16x32_neighbors[513 * MAX_NEIGHBORS]) = {
@@ -1418,6 +1425,7 @@
478, 509, 479, 510, 0, 0
};
+#if CONFIG_EXT_TX
DECLARE_ALIGNED(16, static const int16_t,
mcol_scan_16x32_neighbors[513 * MAX_NEIGHBORS]) = {
0, 0, 0, 0, 16, 16, 32, 32, 48, 48, 64, 64, 80, 80, 96,
@@ -2841,12 +2849,12 @@
0, 1, 3, 5, 2, 4, 6, 9, 7, 8, 11, 13, 10, 12, 14, 15,
};
-#if CONFIG_EXT_TX
DECLARE_ALIGNED(16, static const int16_t, av1_default_iscan_4x8[32]) = {
0, 1, 4, 9, 2, 3, 6, 11, 5, 7, 8, 13, 10, 12, 14, 17,
15, 16, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31,
};
+#if CONFIG_EXT_TX
DECLARE_ALIGNED(16, static const int16_t, av1_mcol_iscan_4x8[32]) = {
0, 8, 16, 24, 1, 9, 17, 25, 2, 10, 18, 26, 3, 11, 19, 27,
4, 12, 20, 28, 5, 13, 21, 29, 6, 14, 22, 30, 7, 15, 23, 31,
@@ -2856,12 +2864,14 @@
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31,
};
+#endif
DECLARE_ALIGNED(16, static const int16_t, av1_default_iscan_8x4[32]) = {
0, 1, 4, 9, 15, 19, 24, 28, 2, 3, 6, 11, 16, 21, 25, 29,
5, 7, 8, 13, 18, 22, 26, 30, 10, 12, 14, 17, 20, 23, 27, 31,
};
+#if CONFIG_EXT_TX
DECLARE_ALIGNED(16, static const int16_t, av1_mcol_iscan_8x4[32]) = {
0, 4, 8, 12, 16, 20, 24, 28, 1, 5, 9, 13, 17, 21, 25, 29,
2, 6, 10, 14, 18, 22, 26, 30, 3, 7, 11, 15, 19, 23, 27, 31,
@@ -2910,7 +2920,6 @@
25, 32, 39, 45, 50, 55, 59, 62, 33, 40, 46, 51, 54, 58, 61, 63,
};
-#if CONFIG_EXT_TX
DECLARE_ALIGNED(16, static const int16_t, av1_default_iscan_8x16[128]) = {
0, 1, 3, 6, 10, 15, 21, 28, 2, 4, 7, 11, 16, 22, 29, 36,
5, 8, 12, 17, 23, 30, 37, 44, 9, 13, 18, 24, 31, 38, 45, 52,
@@ -2933,6 +2942,7 @@
35, 43, 51, 59, 67, 75, 83, 91, 99, 106, 112, 117, 121, 124, 126, 127,
};
+#if CONFIG_EXT_TX
DECLARE_ALIGNED(16, static const int16_t, av1_mcol_iscan_8x16[128]) = {
0, 16, 32, 48, 64, 80, 96, 112, 1, 17, 33, 49, 65, 81, 97, 113,
2, 18, 34, 50, 66, 82, 98, 114, 3, 19, 35, 51, 67, 83, 99, 115,
@@ -2978,6 +2988,7 @@
105, 106, 107, 108, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119,
120, 121, 122, 123, 124, 125, 126, 127,
};
+#endif
DECLARE_ALIGNED(16, static const int16_t, av1_default_iscan_16x32[512]) = {
0, 1, 3, 6, 10, 15, 21, 28, 36, 45, 55, 66, 78, 91, 105,
@@ -3055,6 +3066,7 @@
510, 511,
};
+#if CONFIG_EXT_TX
DECLARE_ALIGNED(16, static const int16_t, av1_mcol_iscan_16x32[512]) = {
0, 32, 64, 96, 128, 160, 192, 224, 256, 288, 320, 352, 384, 416, 448, 480,
1, 33, 65, 97, 129, 161, 193, 225, 257, 289, 321, 353, 385, 417, 449, 481,
@@ -3810,7 +3822,6 @@
{ default_scan_32x32, av1_default_iscan_32x32, default_scan_32x32_neighbors },
};
-#if CONFIG_EXT_TX
const SCAN_ORDER av1_intra_scan_orders[TX_SIZES][TX_TYPES] = {
#if CONFIG_CB4X4
{
@@ -3819,6 +3830,7 @@
{ row_scan_4x4, av1_row_iscan_4x4, row_scan_4x4_neighbors },
{ col_scan_4x4, av1_col_iscan_4x4, col_scan_4x4_neighbors },
{ default_scan_4x4, av1_default_iscan_4x4, default_scan_4x4_neighbors },
+#if CONFIG_EXT_TX
{ default_scan_4x4, av1_default_iscan_4x4, default_scan_4x4_neighbors },
{ default_scan_4x4, av1_default_iscan_4x4, default_scan_4x4_neighbors },
{ default_scan_4x4, av1_default_iscan_4x4, default_scan_4x4_neighbors },
@@ -3831,6 +3843,7 @@
{ col_scan_4x4, av1_col_iscan_4x4, col_scan_4x4_neighbors },
{ row_scan_4x4, av1_row_iscan_4x4, row_scan_4x4_neighbors },
{ col_scan_4x4, av1_col_iscan_4x4, col_scan_4x4_neighbors },
+#endif // CONFIG_EXT_TX
},
#endif
{
@@ -3839,6 +3852,7 @@
{ row_scan_4x4, av1_row_iscan_4x4, row_scan_4x4_neighbors },
{ col_scan_4x4, av1_col_iscan_4x4, col_scan_4x4_neighbors },
{ default_scan_4x4, av1_default_iscan_4x4, default_scan_4x4_neighbors },
+#if CONFIG_EXT_TX
{ default_scan_4x4, av1_default_iscan_4x4, default_scan_4x4_neighbors },
{ default_scan_4x4, av1_default_iscan_4x4, default_scan_4x4_neighbors },
{ default_scan_4x4, av1_default_iscan_4x4, default_scan_4x4_neighbors },
@@ -3851,6 +3865,7 @@
{ col_scan_4x4, av1_col_iscan_4x4, col_scan_4x4_neighbors },
{ row_scan_4x4, av1_row_iscan_4x4, row_scan_4x4_neighbors },
{ col_scan_4x4, av1_col_iscan_4x4, col_scan_4x4_neighbors },
+#endif // CONFIG_EXT_TX
},
{
// TX_8X8
@@ -3858,6 +3873,7 @@
{ row_scan_8x8, av1_row_iscan_8x8, row_scan_8x8_neighbors },
{ col_scan_8x8, av1_col_iscan_8x8, col_scan_8x8_neighbors },
{ default_scan_8x8, av1_default_iscan_8x8, default_scan_8x8_neighbors },
+#if CONFIG_EXT_TX
{ default_scan_8x8, av1_default_iscan_8x8, default_scan_8x8_neighbors },
{ default_scan_8x8, av1_default_iscan_8x8, default_scan_8x8_neighbors },
{ default_scan_8x8, av1_default_iscan_8x8, default_scan_8x8_neighbors },
@@ -3870,6 +3886,7 @@
{ col_scan_8x8, av1_col_iscan_8x8, col_scan_8x8_neighbors },
{ row_scan_8x8, av1_row_iscan_8x8, row_scan_8x8_neighbors },
{ col_scan_8x8, av1_col_iscan_8x8, col_scan_8x8_neighbors },
+#endif // CONFIG_EXT_TX
},
{
// TX_16X16
@@ -3879,6 +3896,7 @@
{ col_scan_16x16, av1_col_iscan_16x16, col_scan_16x16_neighbors },
{ default_scan_16x16, av1_default_iscan_16x16,
default_scan_16x16_neighbors },
+#if CONFIG_EXT_TX
{ default_scan_16x16, av1_default_iscan_16x16,
default_scan_16x16_neighbors },
{ default_scan_16x16, av1_default_iscan_16x16,
@@ -3896,11 +3914,13 @@
{ col_scan_16x16, av1_col_iscan_16x16, col_scan_16x16_neighbors },
{ row_scan_16x16, av1_row_iscan_16x16, row_scan_16x16_neighbors },
{ col_scan_16x16, av1_col_iscan_16x16, col_scan_16x16_neighbors },
+#endif // CONFIG_EXT_TX
},
{
// TX_32X32
{ default_scan_32x32, av1_default_iscan_32x32,
default_scan_32x32_neighbors },
+#if CONFIG_EXT_TX
{ h2_scan_32x32, av1_h2_iscan_32x32, h2_scan_32x32_neighbors },
{ v2_scan_32x32, av1_v2_iscan_32x32, v2_scan_32x32_neighbors },
{ qtr_scan_32x32, av1_qtr_iscan_32x32, qtr_scan_32x32_neighbors },
@@ -3916,6 +3936,7 @@
{ mcol_scan_32x32, av1_mcol_iscan_32x32, mcol_scan_32x32_neighbors },
{ mrow_scan_32x32, av1_mrow_iscan_32x32, mrow_scan_32x32_neighbors },
{ mcol_scan_32x32, av1_mcol_iscan_32x32, mcol_scan_32x32_neighbors },
+#endif // CONFIG_EXT_TX
}
};
@@ -3927,6 +3948,7 @@
{ default_scan_4x4, av1_default_iscan_4x4, default_scan_4x4_neighbors },
{ default_scan_4x4, av1_default_iscan_4x4, default_scan_4x4_neighbors },
{ default_scan_4x4, av1_default_iscan_4x4, default_scan_4x4_neighbors },
+#if CONFIG_EXT_TX
{ default_scan_4x4, av1_default_iscan_4x4, default_scan_4x4_neighbors },
{ default_scan_4x4, av1_default_iscan_4x4, default_scan_4x4_neighbors },
{ default_scan_4x4, av1_default_iscan_4x4, default_scan_4x4_neighbors },
@@ -3939,6 +3961,7 @@
{ mcol_scan_4x4, av1_mcol_iscan_4x4, mcol_scan_4x4_neighbors },
{ mrow_scan_4x4, av1_mrow_iscan_4x4, mrow_scan_4x4_neighbors },
{ mcol_scan_4x4, av1_mcol_iscan_4x4, mcol_scan_4x4_neighbors },
+#endif // CONFIG_EXT_TX
},
#endif
{
@@ -3947,6 +3970,7 @@
{ default_scan_4x4, av1_default_iscan_4x4, default_scan_4x4_neighbors },
{ default_scan_4x4, av1_default_iscan_4x4, default_scan_4x4_neighbors },
{ default_scan_4x4, av1_default_iscan_4x4, default_scan_4x4_neighbors },
+#if CONFIG_EXT_TX
{ default_scan_4x4, av1_default_iscan_4x4, default_scan_4x4_neighbors },
{ default_scan_4x4, av1_default_iscan_4x4, default_scan_4x4_neighbors },
{ default_scan_4x4, av1_default_iscan_4x4, default_scan_4x4_neighbors },
@@ -3959,6 +3983,7 @@
{ mcol_scan_4x4, av1_mcol_iscan_4x4, mcol_scan_4x4_neighbors },
{ mrow_scan_4x4, av1_mrow_iscan_4x4, mrow_scan_4x4_neighbors },
{ mcol_scan_4x4, av1_mcol_iscan_4x4, mcol_scan_4x4_neighbors },
+#endif // CONFIG_EXT_TX
},
{
// TX_8X8
@@ -3966,6 +3991,7 @@
{ default_scan_8x8, av1_default_iscan_8x8, default_scan_8x8_neighbors },
{ default_scan_8x8, av1_default_iscan_8x8, default_scan_8x8_neighbors },
{ default_scan_8x8, av1_default_iscan_8x8, default_scan_8x8_neighbors },
+#if CONFIG_EXT_TX
{ default_scan_8x8, av1_default_iscan_8x8, default_scan_8x8_neighbors },
{ default_scan_8x8, av1_default_iscan_8x8, default_scan_8x8_neighbors },
{ default_scan_8x8, av1_default_iscan_8x8, default_scan_8x8_neighbors },
@@ -3978,6 +4004,7 @@
{ mcol_scan_8x8, av1_mcol_iscan_8x8, mcol_scan_8x8_neighbors },
{ mrow_scan_8x8, av1_mrow_iscan_8x8, mrow_scan_8x8_neighbors },
{ mcol_scan_8x8, av1_mcol_iscan_8x8, mcol_scan_8x8_neighbors },
+#endif // CONFIG_EXT_TX
},
{
// TX_16X16
@@ -3989,6 +4016,7 @@
default_scan_16x16_neighbors },
{ default_scan_16x16, av1_default_iscan_16x16,
default_scan_16x16_neighbors },
+#if CONFIG_EXT_TX
{ default_scan_16x16, av1_default_iscan_16x16,
default_scan_16x16_neighbors },
{ default_scan_16x16, av1_default_iscan_16x16,
@@ -4006,11 +4034,13 @@
{ mcol_scan_16x16, av1_mcol_iscan_16x16, mcol_scan_16x16_neighbors },
{ mrow_scan_16x16, av1_mrow_iscan_16x16, mrow_scan_16x16_neighbors },
{ mcol_scan_16x16, av1_mcol_iscan_16x16, mcol_scan_16x16_neighbors },
+#endif // CONFIG_EXT_TX
},
{
// TX_32X32
{ default_scan_32x32, av1_default_iscan_32x32,
default_scan_32x32_neighbors },
+#if CONFIG_EXT_TX
{ h2_scan_32x32, av1_h2_iscan_32x32, h2_scan_32x32_neighbors },
{ v2_scan_32x32, av1_v2_iscan_32x32, v2_scan_32x32_neighbors },
{ qtr_scan_32x32, av1_qtr_iscan_32x32, qtr_scan_32x32_neighbors },
@@ -4026,6 +4056,7 @@
{ mcol_scan_32x32, av1_mcol_iscan_32x32, mcol_scan_32x32_neighbors },
{ mrow_scan_32x32, av1_mrow_iscan_32x32, mrow_scan_32x32_neighbors },
{ mcol_scan_32x32, av1_mcol_iscan_32x32, mcol_scan_32x32_neighbors },
+#endif // CONFIG_EXT_TX
},
{
// TX_4X8
@@ -4033,6 +4064,7 @@
{ default_scan_4x8, av1_default_iscan_4x8, default_scan_4x8_neighbors },
{ default_scan_4x8, av1_default_iscan_4x8, default_scan_4x8_neighbors },
{ default_scan_4x8, av1_default_iscan_4x8, default_scan_4x8_neighbors },
+#if CONFIG_EXT_TX
{ default_scan_4x8, av1_default_iscan_4x8, default_scan_4x8_neighbors },
{ default_scan_4x8, av1_default_iscan_4x8, default_scan_4x8_neighbors },
{ default_scan_4x8, av1_default_iscan_4x8, default_scan_4x8_neighbors },
@@ -4045,6 +4077,7 @@
{ mcol_scan_4x8, av1_mcol_iscan_4x8, mcol_scan_4x8_neighbors },
{ mrow_scan_4x8, av1_mrow_iscan_4x8, mrow_scan_4x8_neighbors },
{ mcol_scan_4x8, av1_mcol_iscan_4x8, mcol_scan_4x8_neighbors },
+#endif // CONFIG_EXT_TX
},
{
// TX_8X4
@@ -4052,6 +4085,7 @@
{ default_scan_8x4, av1_default_iscan_8x4, default_scan_8x4_neighbors },
{ default_scan_8x4, av1_default_iscan_8x4, default_scan_8x4_neighbors },
{ default_scan_8x4, av1_default_iscan_8x4, default_scan_8x4_neighbors },
+#if CONFIG_EXT_TX
{ default_scan_8x4, av1_default_iscan_8x4, default_scan_8x4_neighbors },
{ default_scan_8x4, av1_default_iscan_8x4, default_scan_8x4_neighbors },
{ default_scan_8x4, av1_default_iscan_8x4, default_scan_8x4_neighbors },
@@ -4064,6 +4098,7 @@
{ mcol_scan_8x4, av1_mcol_iscan_8x4, mcol_scan_8x4_neighbors },
{ mrow_scan_8x4, av1_mrow_iscan_8x4, mrow_scan_8x4_neighbors },
{ mcol_scan_8x4, av1_mcol_iscan_8x4, mcol_scan_8x4_neighbors },
+#endif // CONFIG_EXT_TX
},
{
// TX_8X16
@@ -4075,6 +4110,7 @@
default_scan_8x16_neighbors },
{ default_scan_8x16, av1_default_iscan_8x16,
default_scan_8x16_neighbors },
+#if CONFIG_EXT_TX
{ default_scan_8x16, av1_default_iscan_8x16,
default_scan_8x16_neighbors },
{ default_scan_8x16, av1_default_iscan_8x16,
@@ -4092,6 +4128,7 @@
{ mcol_scan_8x16, av1_mcol_iscan_8x16, mcol_scan_8x16_neighbors },
{ mrow_scan_8x16, av1_mrow_iscan_8x16, mrow_scan_8x16_neighbors },
{ mcol_scan_8x16, av1_mcol_iscan_8x16, mcol_scan_8x16_neighbors },
+#endif // CONFIG_EXT_TX
},
{
// TX_16X8
@@ -4103,6 +4140,7 @@
default_scan_16x8_neighbors },
{ default_scan_16x8, av1_default_iscan_16x8,
default_scan_16x8_neighbors },
+#if CONFIG_EXT_TX
{ default_scan_16x8, av1_default_iscan_16x8,
default_scan_16x8_neighbors },
{ default_scan_16x8, av1_default_iscan_16x8,
@@ -4120,6 +4158,7 @@
{ mcol_scan_16x8, av1_mcol_iscan_16x8, mcol_scan_16x8_neighbors },
{ mrow_scan_16x8, av1_mrow_iscan_16x8, mrow_scan_16x8_neighbors },
{ mcol_scan_16x8, av1_mcol_iscan_16x8, mcol_scan_16x8_neighbors },
+#endif // CONFIG_EXT_TX
},
{
// TX_16X32
@@ -4131,6 +4170,7 @@
default_scan_16x32_neighbors },
{ default_scan_16x32, av1_default_iscan_16x32,
default_scan_16x32_neighbors },
+#if CONFIG_EXT_TX
{ default_scan_16x32, av1_default_iscan_16x32,
default_scan_16x32_neighbors },
{ default_scan_16x32, av1_default_iscan_16x32,
@@ -4148,6 +4188,7 @@
{ mcol_scan_16x32, av1_mcol_iscan_16x32, mcol_scan_16x32_neighbors },
{ mrow_scan_16x32, av1_mrow_iscan_16x32, mrow_scan_16x32_neighbors },
{ mcol_scan_16x32, av1_mcol_iscan_16x32, mcol_scan_16x32_neighbors },
+#endif // CONFIG_EXT_TX
},
{
// TX_32X16
@@ -4159,6 +4200,7 @@
default_scan_32x16_neighbors },
{ default_scan_32x16, av1_default_iscan_32x16,
default_scan_32x16_neighbors },
+#if CONFIG_EXT_TX
{ default_scan_32x16, av1_default_iscan_32x16,
default_scan_32x16_neighbors },
{ default_scan_32x16, av1_default_iscan_32x16,
@@ -4176,49 +4218,9 @@
{ mcol_scan_32x16, av1_mcol_iscan_32x16, mcol_scan_32x16_neighbors },
{ mrow_scan_32x16, av1_mrow_iscan_32x16, mrow_scan_32x16_neighbors },
{ mcol_scan_32x16, av1_mcol_iscan_32x16, mcol_scan_32x16_neighbors },
- }
-};
-
-#else // CONFIG_EXT_TX
-
-const SCAN_ORDER av1_intra_scan_orders[TX_SIZES][TX_TYPES] = {
-#if CONFIG_CB4X4
- { // TX_2X2
- { default_scan_4x4, av1_default_iscan_4x4, default_scan_4x4_neighbors },
- { row_scan_4x4, av1_row_iscan_4x4, row_scan_4x4_neighbors },
- { col_scan_4x4, av1_col_iscan_4x4, col_scan_4x4_neighbors },
- { default_scan_4x4, av1_default_iscan_4x4, default_scan_4x4_neighbors } },
-#endif
- { // TX_4X4
- { default_scan_4x4, av1_default_iscan_4x4, default_scan_4x4_neighbors },
- { row_scan_4x4, av1_row_iscan_4x4, row_scan_4x4_neighbors },
- { col_scan_4x4, av1_col_iscan_4x4, col_scan_4x4_neighbors },
- { default_scan_4x4, av1_default_iscan_4x4, default_scan_4x4_neighbors } },
- { // TX_8X8
- { default_scan_8x8, av1_default_iscan_8x8, default_scan_8x8_neighbors },
- { row_scan_8x8, av1_row_iscan_8x8, row_scan_8x8_neighbors },
- { col_scan_8x8, av1_col_iscan_8x8, col_scan_8x8_neighbors },
- { default_scan_8x8, av1_default_iscan_8x8, default_scan_8x8_neighbors } },
- { // TX_16X16
- { default_scan_16x16, av1_default_iscan_16x16,
- default_scan_16x16_neighbors },
- { row_scan_16x16, av1_row_iscan_16x16, row_scan_16x16_neighbors },
- { col_scan_16x16, av1_col_iscan_16x16, col_scan_16x16_neighbors },
- { default_scan_16x16, av1_default_iscan_16x16,
- default_scan_16x16_neighbors } },
- {
- // TX_32X32
- { default_scan_32x32, av1_default_iscan_32x32,
- default_scan_32x32_neighbors },
- { default_scan_32x32, av1_default_iscan_32x32,
- default_scan_32x32_neighbors },
- { default_scan_32x32, av1_default_iscan_32x32,
- default_scan_32x32_neighbors },
- { default_scan_32x32, av1_default_iscan_32x32,
- default_scan_32x32_neighbors },
- }
-};
#endif // CONFIG_EXT_TX
+ }
+};
#if CONFIG_ADAPT_SCAN
// TX_32X32 will has 1024 coefficients whose indexes can be represented in 10
diff --git a/av1/common/scan.h b/av1/common/scan.h
index af39993..2078e99 100644
--- a/av1/common/scan.h
+++ b/av1/common/scan.h
@@ -27,6 +27,7 @@
extern const SCAN_ORDER av1_default_scan_orders[TX_SIZES];
extern const SCAN_ORDER av1_intra_scan_orders[TX_SIZES][TX_TYPES];
+extern const SCAN_ORDER av1_inter_scan_orders[TX_SIZES_ALL][TX_TYPES];
#if CONFIG_ADAPT_SCAN
void av1_update_scan_prob(AV1_COMMON *cm, TX_SIZE tx_size, TX_TYPE tx_type,
@@ -87,7 +88,7 @@
return &cm->fc->sc[tx_size][tx_type];
#else // CONFIG_ADAPT_SCAN
(void)cm;
-#if CONFIG_EXT_TX
+#if CONFIG_EXT_TX || CONFIG_VAR_TX
return is_inter ? &av1_inter_scan_orders[tx_size][tx_type]
: &av1_intra_scan_orders[tx_size][tx_type];
#else
diff --git a/av1/common/warped_motion.c b/av1/common/warped_motion.c
index e5ed39d..f73e777 100644
--- a/av1/common/warped_motion.c
+++ b/av1/common/warped_motion.c
@@ -35,18 +35,18 @@
const int x = *(points++), y = *(points++);
if (subsampling_x)
*(proj++) = ROUND_POWER_OF_TWO_SIGNED(
- ((x << (WARPEDMODEL_PREC_BITS + 1)) + mat[1]),
+ ((x * (1 << (WARPEDMODEL_PREC_BITS + 1))) + mat[1]),
WARPEDDIFF_PREC_BITS + 1);
else
*(proj++) = ROUND_POWER_OF_TWO_SIGNED(
- ((x << WARPEDMODEL_PREC_BITS) + mat[1]), WARPEDDIFF_PREC_BITS);
+ ((x * (1 << WARPEDMODEL_PREC_BITS)) + mat[1]), WARPEDDIFF_PREC_BITS);
if (subsampling_y)
*(proj++) = ROUND_POWER_OF_TWO_SIGNED(
- ((y << (WARPEDMODEL_PREC_BITS + 1)) + mat[0]),
+ ((y * (1 << (WARPEDMODEL_PREC_BITS + 1))) + mat[0]),
WARPEDDIFF_PREC_BITS + 1);
else
*(proj++) = ROUND_POWER_OF_TWO_SIGNED(
- ((y << WARPEDMODEL_PREC_BITS)) + mat[0], WARPEDDIFF_PREC_BITS);
+ ((y * (1 << WARPEDMODEL_PREC_BITS))) + mat[0], WARPEDDIFF_PREC_BITS);
points += stride_points - 2;
proj += stride_proj - 2;
}
@@ -119,12 +119,12 @@
y = (subsampling_y ? 4 * y + 1 : 2 * y);
Z = (mat[7] * x + mat[6] * y + (1 << (WARPEDMODEL_ROW3HOMO_PREC_BITS + 1)));
- xp = (mat[1] * x + mat[0] * y + 2 * mat[3])
- << (WARPEDPIXEL_PREC_BITS + WARPEDMODEL_ROW3HOMO_PREC_BITS -
- WARPEDMODEL_PREC_BITS);
- yp = (mat[2] * x + mat[5] * y + 2 * mat[4])
- << (WARPEDPIXEL_PREC_BITS + WARPEDMODEL_ROW3HOMO_PREC_BITS -
- WARPEDMODEL_PREC_BITS);
+ xp = (mat[1] * x + mat[0] * y + 2 * mat[3]) *
+ (1 << (WARPEDPIXEL_PREC_BITS + WARPEDMODEL_ROW3HOMO_PREC_BITS -
+ WARPEDMODEL_PREC_BITS));
+ yp = (mat[2] * x + mat[5] * y + 2 * mat[4]) *
+ (1 << (WARPEDPIXEL_PREC_BITS + WARPEDMODEL_ROW3HOMO_PREC_BITS -
+ WARPEDMODEL_PREC_BITS));
xp = xp > 0 ? (xp + Z / 2) / Z : (xp - Z / 2) / Z;
yp = yp > 0 ? (yp + Z / 2) / Z : (yp - Z / 2) / Z;
@@ -220,9 +220,9 @@
const int64_t v3 = x * (p[1] - p[-1]);
const int64_t v4 = 2 * p[0];
return (int32_t)ROUND_POWER_OF_TWO_SIGNED(
- (v4 << (3 * WARPEDPIXEL_PREC_BITS)) +
- (v3 << (2 * WARPEDPIXEL_PREC_BITS)) +
- (v2 << WARPEDPIXEL_PREC_BITS) + v1,
+ (v4 * (1 << (3 * WARPEDPIXEL_PREC_BITS))) +
+ (v3 * (1 << (2 * WARPEDPIXEL_PREC_BITS))) +
+ (v2 * (1 << WARPEDPIXEL_PREC_BITS)) + v1,
3 * WARPEDPIXEL_PREC_BITS + 1 - WARPEDPIXEL_FILTER_BITS);
}
}
@@ -246,10 +246,10 @@
i + k + 1 - WARPEDPIXEL_FILTER_TAPS / 2,
j + 1 - WARPEDPIXEL_FILTER_TAPS / 2);
arr[k] = do_ntap_filter(arr_temp + WARPEDPIXEL_FILTER_TAPS / 2 - 1,
- y - (j << WARPEDPIXEL_PREC_BITS));
+ y - (j * (1 << WARPEDPIXEL_PREC_BITS)));
}
val = do_ntap_filter(arr + WARPEDPIXEL_FILTER_TAPS / 2 - 1,
- x - (i << WARPEDPIXEL_PREC_BITS));
+ x - (i * (1 << WARPEDPIXEL_PREC_BITS)));
val = ROUND_POWER_OF_TWO_SIGNED(val, WARPEDPIXEL_FILTER_BITS * 2);
return (uint8_t)clip_pixel(val);
}
@@ -262,9 +262,10 @@
for (k = 0; k < 4; ++k) {
int32_t arr_temp[4];
get_subcolumn(4, ref, arr_temp, stride, i + k - 1, j - 1);
- arr[k] = do_cubic_filter(arr_temp + 1, y - (j << WARPEDPIXEL_PREC_BITS));
+ arr[k] =
+ do_cubic_filter(arr_temp + 1, y - (j * (1 << WARPEDPIXEL_PREC_BITS)));
}
- val = do_cubic_filter(arr + 1, x - (i << WARPEDPIXEL_PREC_BITS));
+ val = do_cubic_filter(arr + 1, x - (i * (1 << WARPEDPIXEL_PREC_BITS)));
val = ROUND_POWER_OF_TWO_SIGNED(val, WARPEDPIXEL_FILTER_BITS * 2);
return (uint8_t)clip_pixel(val);
}
@@ -272,8 +273,8 @@
static uint8_t bi_linear_filter(uint8_t *ref, int x, int y, int stride) {
const int ix = x >> WARPEDPIXEL_PREC_BITS;
const int iy = y >> WARPEDPIXEL_PREC_BITS;
- const int sx = x - (ix << WARPEDPIXEL_PREC_BITS);
- const int sy = y - (iy << WARPEDPIXEL_PREC_BITS);
+ const int sx = x - (ix * (1 << WARPEDPIXEL_PREC_BITS));
+ const int sy = y - (iy * (1 << WARPEDPIXEL_PREC_BITS));
int32_t val;
val = ROUND_POWER_OF_TWO_SIGNED(
ref[iy * stride + ix] * (WARPEDPIXEL_PREC_SHIFTS - sy) *
@@ -289,8 +290,8 @@
int height, int stride) {
int ix = x >> WARPEDPIXEL_PREC_BITS;
int iy = y >> WARPEDPIXEL_PREC_BITS;
- int sx = x - (ix << WARPEDPIXEL_PREC_BITS);
- int sy = y - (iy << WARPEDPIXEL_PREC_BITS);
+ int sx = x - (ix * (1 << WARPEDPIXEL_PREC_BITS));
+ int sy = y - (iy * (1 << WARPEDPIXEL_PREC_BITS));
int32_t v;
if (ix < 0 && iy < 0)
@@ -357,10 +358,10 @@
i + k + 1 - WARPEDPIXEL_FILTER_TAPS / 2,
j + 1 - WARPEDPIXEL_FILTER_TAPS / 2);
arr[k] = do_ntap_filter(arr_temp + WARPEDPIXEL_FILTER_TAPS / 2 - 1,
- y - (j << WARPEDPIXEL_PREC_BITS));
+ y - (j * (1 << WARPEDPIXEL_PREC_BITS)));
}
val = do_ntap_filter(arr + WARPEDPIXEL_FILTER_TAPS / 2 - 1,
- x - (i << WARPEDPIXEL_PREC_BITS));
+ x - (i * (1 << WARPEDPIXEL_PREC_BITS)));
val = ROUND_POWER_OF_TWO_SIGNED(val, WARPEDPIXEL_FILTER_BITS * 2);
return (uint16_t)clip_pixel_highbd(val, bd);
}
@@ -374,9 +375,10 @@
for (k = 0; k < 4; ++k) {
int32_t arr_temp[4];
highbd_get_subcolumn(4, ref, arr_temp, stride, i + k - 1, j - 1);
- arr[k] = do_cubic_filter(arr_temp + 1, y - (j << WARPEDPIXEL_PREC_BITS));
+ arr[k] =
+ do_cubic_filter(arr_temp + 1, y - (j * (1 << WARPEDPIXEL_PREC_BITS)));
}
- val = do_cubic_filter(arr + 1, x - (i << WARPEDPIXEL_PREC_BITS));
+ val = do_cubic_filter(arr + 1, x - (i * (1 << WARPEDPIXEL_PREC_BITS)));
val = ROUND_POWER_OF_TWO_SIGNED(val, WARPEDPIXEL_FILTER_BITS * 2);
return (uint16_t)clip_pixel_highbd(val, bd);
}
@@ -385,8 +387,8 @@
int bd) {
const int ix = x >> WARPEDPIXEL_PREC_BITS;
const int iy = y >> WARPEDPIXEL_PREC_BITS;
- const int sx = x - (ix << WARPEDPIXEL_PREC_BITS);
- const int sy = y - (iy << WARPEDPIXEL_PREC_BITS);
+ const int sx = x - (ix * (1 << WARPEDPIXEL_PREC_BITS));
+ const int sy = y - (iy * (1 << WARPEDPIXEL_PREC_BITS));
int32_t val;
val = ROUND_POWER_OF_TWO_SIGNED(
ref[iy * stride + ix] * (WARPEDPIXEL_PREC_SHIFTS - sy) *
@@ -402,8 +404,8 @@
int height, int stride, int bd) {
int ix = x >> WARPEDPIXEL_PREC_BITS;
int iy = y >> WARPEDPIXEL_PREC_BITS;
- int sx = x - (ix << WARPEDPIXEL_PREC_BITS);
- int sy = y - (iy << WARPEDPIXEL_PREC_BITS);
+ int sx = x - (ix * (1 << WARPEDPIXEL_PREC_BITS));
+ int sy = y - (iy * (1 << WARPEDPIXEL_PREC_BITS));
int32_t v;
if (ix < 0 && iy < 0)
diff --git a/av1/common/x86/idct_intrin_sse2.c b/av1/common/x86/idct_intrin_sse2.c
index 5d28a28..a6b6e1e 100644
--- a/av1/common/x86/idct_intrin_sse2.c
+++ b/av1/common/x86/idct_intrin_sse2.c
@@ -571,6 +571,7 @@
in[6] = mm_reverse_epi16(in[6]);
in[7] = mm_reverse_epi16(in[7]);
}
+#endif // CONFIG_EXT_TX
static INLINE void scale_sqrt2_8x4(__m128i *in) {
// Implements 'ROUND_POWER_OF_TWO(input * Sqrt2, DCT_CONST_BITS)'
@@ -693,8 +694,10 @@
switch (tx_type) {
case DCT_DCT:
case ADST_DCT:
+#if CONFIG_EXT_TX
case FLIPADST_DCT:
case H_DCT:
+#endif
aom_idct8_sse2(in);
array_transpose_8x8(in, in);
aom_idct8_sse2(in + 8);
@@ -702,17 +705,20 @@
break;
case DCT_ADST:
case ADST_ADST:
+#if CONFIG_EXT_TX
case DCT_FLIPADST:
case FLIPADST_FLIPADST:
case ADST_FLIPADST:
case FLIPADST_ADST:
case H_ADST:
case H_FLIPADST:
+#endif
aom_iadst8_sse2(in);
array_transpose_8x8(in, in);
aom_iadst8_sse2(in + 8);
array_transpose_8x8(in + 8, in + 8);
break;
+#if CONFIG_EXT_TX
case V_FLIPADST:
case V_ADST:
case V_DCT:
@@ -720,6 +726,7 @@
iidtx8_sse2(in);
iidtx8_sse2(in + 8);
break;
+#endif
default: assert(0); break;
}
scale_sqrt2_8x8(in);
@@ -729,33 +736,50 @@
switch (tx_type) {
case DCT_DCT:
case DCT_ADST:
+#if CONFIG_EXT_TX
case DCT_FLIPADST:
- case V_DCT: idct16_8col(in); break;
+ case V_DCT:
+#endif
+ idct16_8col(in);
+ break;
case ADST_DCT:
case ADST_ADST:
+#if CONFIG_EXT_TX
case FLIPADST_ADST:
case ADST_FLIPADST:
case FLIPADST_FLIPADST:
case FLIPADST_DCT:
case V_ADST:
- case V_FLIPADST: iadst16_8col(in); break;
+ case V_FLIPADST:
+#endif
+ iadst16_8col(in);
+ break;
+#if CONFIG_EXT_TX
case H_DCT:
case H_ADST:
case H_FLIPADST:
case IDTX: iidtx16_8col(in); break;
+#endif
default: assert(0); break;
}
switch (tx_type) {
case DCT_DCT:
case ADST_DCT:
+#if CONFIG_EXT_TX
case H_DCT:
+#endif
case DCT_ADST:
case ADST_ADST:
+#if CONFIG_EXT_TX
case H_ADST:
case V_ADST:
case V_DCT:
- case IDTX: write_buffer_8x16(dest, in, stride); break;
+ case IDTX:
+#endif
+ write_buffer_8x16(dest, in, stride);
+ break;
+#if CONFIG_EXT_TX
case FLIPADST_DCT:
case FLIPADST_ADST:
case V_FLIPADST: write_buffer_8x16(dest + stride * 15, in, -stride); break;
@@ -771,6 +795,7 @@
flip_buffer_lr_8x8(in + 8);
write_buffer_8x16(dest + stride * 15, in, -stride);
break;
+#endif
default: assert(0); break;
}
}
@@ -837,20 +862,30 @@
switch (tx_type) {
case DCT_DCT:
case ADST_DCT:
+#if CONFIG_EXT_TX
case FLIPADST_DCT:
- case H_DCT: idct16_8col(in); break;
+ case H_DCT:
+#endif
+ idct16_8col(in);
+ break;
case DCT_ADST:
case ADST_ADST:
+#if CONFIG_EXT_TX
case DCT_FLIPADST:
case FLIPADST_FLIPADST:
case ADST_FLIPADST:
case FLIPADST_ADST:
case H_ADST:
- case H_FLIPADST: iadst16_8col(in); break;
+ case H_FLIPADST:
+#endif
+ iadst16_8col(in);
+ break;
+#if CONFIG_EXT_TX
case V_FLIPADST:
case V_ADST:
case V_DCT:
case IDTX: iidtx16_8col(in); break;
+#endif
default: assert(0); break;
}
@@ -862,22 +897,27 @@
switch (tx_type) {
case DCT_DCT:
case DCT_ADST:
+#if CONFIG_EXT_TX
case DCT_FLIPADST:
case V_DCT:
+#endif
aom_idct8_sse2(in);
aom_idct8_sse2(in + 8);
break;
case ADST_DCT:
case ADST_ADST:
+#if CONFIG_EXT_TX
case FLIPADST_ADST:
case ADST_FLIPADST:
case FLIPADST_FLIPADST:
case FLIPADST_DCT:
case V_ADST:
case V_FLIPADST:
+#endif
aom_iadst8_sse2(in);
aom_iadst8_sse2(in + 8);
break;
+#if CONFIG_EXT_TX
case H_DCT:
case H_ADST:
case H_FLIPADST:
@@ -887,22 +927,26 @@
iidtx8_sse2(in);
iidtx8_sse2(in + 8);
break;
+#endif
default: assert(0); break;
}
switch (tx_type) {
case DCT_DCT:
case ADST_DCT:
- case H_DCT:
case DCT_ADST:
case ADST_ADST:
+#if CONFIG_EXT_TX
+ case H_DCT:
case H_ADST:
case V_ADST:
case V_DCT:
case IDTX:
+#endif
write_buffer_8x8_round6(dest, in, stride);
write_buffer_8x8_round6(dest + 8, in + 8, stride);
break;
+#if CONFIG_EXT_TX
case FLIPADST_DCT:
case FLIPADST_ADST:
case V_FLIPADST:
@@ -923,6 +967,7 @@
write_buffer_8x8_round6(dest + stride * 7, in + 8, -stride);
write_buffer_8x8_round6(dest + stride * 7 + 8, in, -stride);
break;
+#endif
default: assert(0); break;
}
}
@@ -961,10 +1006,15 @@
switch (tx_type) {
case DCT_DCT:
case ADST_DCT:
+#if CONFIG_EXT_TX
case FLIPADST_DCT:
- case H_DCT: aom_idct8_sse2(in); break;
+ case H_DCT:
+#endif
+ aom_idct8_sse2(in);
+ break;
case DCT_ADST:
case ADST_ADST:
+#if CONFIG_EXT_TX
case DCT_FLIPADST:
case FLIPADST_FLIPADST:
case ADST_FLIPADST:
@@ -974,9 +1024,8 @@
case V_FLIPADST:
case V_ADST:
case V_DCT:
- case IDTX:
- iidtx8_sse2(in);
- array_transpose_8x8(in, in);
+ case IDTX: iidtx8_sse2(in); array_transpose_8x8(in, in);
+#endif
break;
default: assert(0); break;
}
@@ -995,22 +1044,27 @@
switch (tx_type) {
case DCT_DCT:
case DCT_ADST:
+#if CONFIG_EXT_TX
case DCT_FLIPADST:
case V_DCT:
+#endif
aom_idct4_sse2(in + 4);
aom_idct4_sse2(in + 6);
break;
case ADST_DCT:
case ADST_ADST:
+#if CONFIG_EXT_TX
case FLIPADST_ADST:
case ADST_FLIPADST:
case FLIPADST_FLIPADST:
case FLIPADST_DCT:
case V_ADST:
case V_FLIPADST:
+#endif
aom_iadst4_sse2(in + 4);
aom_iadst4_sse2(in + 6);
break;
+#if CONFIG_EXT_TX
case H_DCT:
case H_ADST:
case H_FLIPADST:
@@ -1020,6 +1074,7 @@
iidtx4_sse2(in + 6);
array_transpose_4x4(in + 6);
break;
+#endif
default: assert(0); break;
}
@@ -1032,9 +1087,10 @@
switch (tx_type) {
case DCT_DCT:
case ADST_DCT:
- case H_DCT:
case DCT_ADST:
case ADST_ADST:
+#if CONFIG_EXT_TX
+ case H_DCT:
case H_ADST:
case V_ADST:
case V_DCT:
@@ -1056,6 +1112,7 @@
in[2] = mm_reverse_epi16(in[2]);
in[3] = mm_reverse_epi16(in[3]);
FLIPUD_PTR(dest, stride, 4);
+#endif
break;
default: assert(0); break;
}
@@ -1139,22 +1196,27 @@
switch (tx_type) {
case DCT_DCT:
case ADST_DCT:
+#if CONFIG_EXT_TX
case FLIPADST_DCT:
case H_DCT:
+#endif
aom_idct4_sse2(in + 4);
aom_idct4_sse2(in + 6);
break;
case DCT_ADST:
case ADST_ADST:
+#if CONFIG_EXT_TX
case DCT_FLIPADST:
case FLIPADST_FLIPADST:
case ADST_FLIPADST:
case FLIPADST_ADST:
case H_ADST:
case H_FLIPADST:
+#endif
aom_iadst4_sse2(in + 4);
aom_iadst4_sse2(in + 6);
break;
+#if CONFIG_EXT_TX
case V_FLIPADST:
case V_ADST:
case V_DCT:
@@ -1164,6 +1226,7 @@
iidtx4_sse2(in + 6);
array_transpose_4x4(in + 6);
break;
+#endif
default: assert(0); break;
}
@@ -1177,16 +1240,25 @@
switch (tx_type) {
case DCT_DCT:
case DCT_ADST:
+#if CONFIG_EXT_TX
case DCT_FLIPADST:
- case V_DCT: aom_idct8_sse2(in); break;
+ case V_DCT:
+#endif
+ aom_idct8_sse2(in);
+ break;
case ADST_DCT:
case ADST_ADST:
+#if CONFIG_EXT_TX
case FLIPADST_ADST:
case ADST_FLIPADST:
case FLIPADST_FLIPADST:
case FLIPADST_DCT:
case V_ADST:
- case V_FLIPADST: aom_iadst8_sse2(in); break;
+ case V_FLIPADST:
+#endif
+ aom_iadst8_sse2(in);
+ break;
+#if CONFIG_EXT_TX
case H_DCT:
case H_ADST:
case H_FLIPADST:
@@ -1194,19 +1266,24 @@
iidtx8_sse2(in);
array_transpose_8x8(in, in);
break;
+#endif
default: assert(0); break;
}
switch (tx_type) {
case DCT_DCT:
case ADST_DCT:
- case H_DCT:
case DCT_ADST:
case ADST_ADST:
+#if CONFIG_EXT_TX
+ case H_DCT:
case H_ADST:
case V_ADST:
case V_DCT:
- case IDTX: break;
+ case IDTX:
+#endif
+ break;
+#if CONFIG_EXT_TX
case FLIPADST_DCT:
case FLIPADST_ADST:
case V_FLIPADST: FLIPUD_PTR(dest, stride, 8); break;
@@ -1233,6 +1310,7 @@
in[7] = _mm_shufflelo_epi16(in[7], 0x1b);
FLIPUD_PTR(dest, stride, 8);
break;
+#endif
default: assert(0); break;
}
in[0] = _mm_unpacklo_epi64(in[0], in[1]);
@@ -1283,6 +1361,7 @@
aom_idct16_sse2(bl, br); // Includes a transposition
}
+#if CONFIG_EXT_TX
static INLINE void iidtx32_16col(__m128i *tl, __m128i *tr, __m128i *bl,
__m128i *br) {
int i;
@@ -1295,6 +1374,7 @@
br[i] = _mm_slli_epi16(br[i], 2);
}
}
+#endif // CONFIG_EXT_TX
static INLINE void write_buffer_16x32_round6(uint8_t *dest, __m128i *intl,
__m128i *intr, __m128i *inbl,
@@ -1335,22 +1415,27 @@
switch (tx_type) {
case DCT_DCT:
case ADST_DCT:
+#if CONFIG_EXT_TX
case FLIPADST_DCT:
case H_DCT:
+#endif
aom_idct16_sse2(intl, intr);
aom_idct16_sse2(inbl, inbr);
break;
case DCT_ADST:
case ADST_ADST:
+#if CONFIG_EXT_TX
case DCT_FLIPADST:
case FLIPADST_FLIPADST:
case ADST_FLIPADST:
case FLIPADST_ADST:
case H_ADST:
case H_FLIPADST:
+#endif
aom_iadst16_sse2(intl, intr);
aom_iadst16_sse2(inbl, inbr);
break;
+#if CONFIG_EXT_TX
case V_FLIPADST:
case V_ADST:
case V_DCT:
@@ -1358,6 +1443,7 @@
iidtx16_sse2(intl, intr);
iidtx16_sse2(inbl, inbr);
break;
+#endif
default: assert(0); break;
}
@@ -1370,33 +1456,47 @@
switch (tx_type) {
case DCT_DCT:
case DCT_ADST:
+#if CONFIG_EXT_TX
case DCT_FLIPADST:
- case V_DCT: idct32_16col(intl, intr, inbl, inbr); break;
+ case V_DCT:
+#endif
+ idct32_16col(intl, intr, inbl, inbr);
+ break;
case ADST_DCT:
case ADST_ADST:
+#if CONFIG_EXT_TX
case FLIPADST_ADST:
case ADST_FLIPADST:
case FLIPADST_FLIPADST:
case FLIPADST_DCT:
case V_ADST:
- case V_FLIPADST: ihalfright32_16col(intl, intr, inbl, inbr); break;
+ case V_FLIPADST:
+#endif
+ ihalfright32_16col(intl, intr, inbl, inbr);
+ break;
+#if CONFIG_EXT_TX
case H_DCT:
case H_ADST:
case H_FLIPADST:
case IDTX: iidtx32_16col(intl, intr, inbl, inbr); break;
+#endif
default: assert(0); break;
}
switch (tx_type) {
case DCT_DCT:
case ADST_DCT:
- case H_DCT:
case DCT_ADST:
case ADST_ADST:
+#if CONFIG_EXT_TX
+ case H_DCT:
case H_ADST:
case V_ADST:
case V_DCT:
- case IDTX: break;
+ case IDTX:
+#endif
+ break;
+#if CONFIG_EXT_TX
case FLIPADST_DCT:
case FLIPADST_ADST:
case V_FLIPADST: FLIPUD_PTR(dest, stride, 32); break;
@@ -1423,6 +1523,7 @@
}
FLIPUD_PTR(dest, stride, 32);
break;
+#endif
default: assert(0); break;
}
write_buffer_16x32_round6(dest, intl, intr, inbl, inbr, stride);
@@ -1467,20 +1568,30 @@
switch (tx_type) {
case DCT_DCT:
case ADST_DCT:
+#if CONFIG_EXT_TX
case FLIPADST_DCT:
- case H_DCT: idct32_16col(in0, in1, in2, in3); break;
+ case H_DCT:
+#endif
+ idct32_16col(in0, in1, in2, in3);
+ break;
case DCT_ADST:
case ADST_ADST:
+#if CONFIG_EXT_TX
case DCT_FLIPADST:
case FLIPADST_FLIPADST:
case ADST_FLIPADST:
case FLIPADST_ADST:
case H_ADST:
- case H_FLIPADST: ihalfright32_16col(in0, in1, in2, in3); break;
+ case H_FLIPADST:
+#endif
+ ihalfright32_16col(in0, in1, in2, in3);
+ break;
+#if CONFIG_EXT_TX
case V_FLIPADST:
case V_ADST:
case V_DCT:
case IDTX: iidtx32_16col(in0, in1, in2, in3); break;
+#endif
default: assert(0); break;
}
@@ -1493,22 +1604,27 @@
switch (tx_type) {
case DCT_DCT:
case DCT_ADST:
+#if CONFIG_EXT_TX
case DCT_FLIPADST:
case V_DCT:
+#endif
aom_idct16_sse2(in0, in1);
aom_idct16_sse2(in2, in3);
break;
case ADST_DCT:
case ADST_ADST:
+#if CONFIG_EXT_TX
case FLIPADST_ADST:
case ADST_FLIPADST:
case FLIPADST_FLIPADST:
case FLIPADST_DCT:
case V_ADST:
case V_FLIPADST:
+#endif
aom_iadst16_sse2(in0, in1);
aom_iadst16_sse2(in2, in3);
break;
+#if CONFIG_EXT_TX
case H_DCT:
case H_ADST:
case H_FLIPADST:
@@ -1516,19 +1632,24 @@
iidtx16_sse2(in0, in1);
iidtx16_sse2(in2, in3);
break;
+#endif
default: assert(0); break;
}
switch (tx_type) {
case DCT_DCT:
case ADST_DCT:
- case H_DCT:
case DCT_ADST:
case ADST_ADST:
+#if CONFIG_EXT_TX
+ case H_DCT:
case H_ADST:
case V_ADST:
case V_DCT:
- case IDTX: break;
+ case IDTX:
+#endif
+ break;
+#if CONFIG_EXT_TX
case FLIPADST_DCT:
case FLIPADST_ADST:
case V_FLIPADST: FLIPUD_PTR(dest, stride, 16); break;
@@ -1555,8 +1676,8 @@
}
FLIPUD_PTR(dest, stride, 16);
break;
+#endif
default: assert(0); break;
}
write_buffer_32x16_round6(dest, in0, in1, in2, in3, stride);
}
-#endif // CONFIG_EXT_TX
diff --git a/av1/encoder/dct.c b/av1/encoder/dct.c
index 221e3cd..c002dab 100644
--- a/av1/encoder/dct.c
+++ b/av1/encoder/dct.c
@@ -18,6 +18,8 @@
#include "aom_dsp/fwd_txfm.h"
#include "aom_ports/mem.h"
#include "av1/common/blockd.h"
+#include "av1/common/av1_fwd_txfm1d.h"
+#include "av1/common/av1_fwd_txfm2d_cfg.h"
#include "av1/common/idct.h"
static INLINE void range_check(const tran_low_t *input, const int size,
@@ -997,6 +999,21 @@
output[15] = (tran_low_t)-x1;
}
+// For use in lieu of ADST
+static void fhalfright32(const tran_low_t *input, tran_low_t *output) {
+ int i;
+ tran_low_t inputhalf[16];
+ for (i = 0; i < 16; ++i) {
+ output[16 + i] = input[i] * 4;
+ }
+ // Multiply input by sqrt(2)
+ for (i = 0; i < 16; ++i) {
+ inputhalf[i] = (tran_low_t)fdct_round_shift(input[i + 16] * Sqrt2);
+ }
+ fdct16(inputhalf, output);
+ // Note overall scaling factor is 4 times orthogonal
+}
+
#if CONFIG_EXT_TX
static void fidtx4(const tran_low_t *input, tran_low_t *output) {
int i;
@@ -1020,21 +1037,6 @@
for (i = 0; i < 32; ++i) output[i] = input[i] * 4;
}
-// For use in lieu of ADST
-static void fhalfright32(const tran_low_t *input, tran_low_t *output) {
- int i;
- tran_low_t inputhalf[16];
- for (i = 0; i < 16; ++i) {
- output[16 + i] = input[i] * 4;
- }
- // Multiply input by sqrt(2)
- for (i = 0; i < 16; ++i) {
- inputhalf[i] = (tran_low_t)fdct_round_shift(input[i + 16] * Sqrt2);
- }
- fdct16(inputhalf, output);
- // Note overall scaling factor is 4 times orthogonal
-}
-
static void copy_block(const int16_t *src, int src_stride, int l, int w,
int16_t *dest, int dest_stride) {
int i;
@@ -1182,7 +1184,6 @@
}
}
-#if CONFIG_EXT_TX
void av1_fht4x8_c(const int16_t *input, tran_low_t *output, int stride,
int tx_type) {
static const transform_2d FHT[] = {
@@ -1190,6 +1191,7 @@
{ fadst8, fdct4 }, // ADST_DCT
{ fdct8, fadst4 }, // DCT_ADST
{ fadst8, fadst4 }, // ADST_ADST
+#if CONFIG_EXT_TX
{ fadst8, fdct4 }, // FLIPADST_DCT
{ fdct8, fadst4 }, // DCT_FLIPADST
{ fadst8, fadst4 }, // FLIPADST_FLIPADST
@@ -1202,6 +1204,7 @@
{ fidtx8, fadst4 }, // H_ADST
{ fadst8, fidtx4 }, // V_FLIPADST
{ fidtx8, fadst4 }, // H_FLIPADST
+#endif
};
const transform_2d ht = FHT[tx_type];
const int n = 4;
@@ -1209,8 +1212,10 @@
tran_low_t out[8 * 4];
tran_low_t temp_in[8], temp_out[8];
int i, j;
+#if CONFIG_EXT_TX
int16_t flipped_input[8 * 4];
maybe_flip_input(&input, &stride, n2, n, flipped_input, tx_type);
+#endif
// Columns
for (i = 0; i < n; ++i) {
@@ -1237,6 +1242,7 @@
{ fadst4, fdct8 }, // ADST_DCT
{ fdct4, fadst8 }, // DCT_ADST
{ fadst4, fadst8 }, // ADST_ADST
+#if CONFIG_EXT_TX
{ fadst4, fdct8 }, // FLIPADST_DCT
{ fdct4, fadst8 }, // DCT_FLIPADST
{ fadst4, fadst8 }, // FLIPADST_FLIPADST
@@ -1249,6 +1255,7 @@
{ fidtx4, fadst8 }, // H_ADST
{ fadst4, fidtx8 }, // V_FLIPADST
{ fidtx4, fadst8 }, // H_FLIPADST
+#endif
};
const transform_2d ht = FHT[tx_type];
const int n = 4;
@@ -1256,8 +1263,10 @@
tran_low_t out[8 * 4];
tran_low_t temp_in[8], temp_out[8];
int i, j;
+#if CONFIG_EXT_TX
int16_t flipped_input[8 * 4];
maybe_flip_input(&input, &stride, n, n2, flipped_input, tx_type);
+#endif
// Columns
for (i = 0; i < n2; ++i) {
@@ -1284,6 +1293,7 @@
{ fadst16, fdct8 }, // ADST_DCT
{ fdct16, fadst8 }, // DCT_ADST
{ fadst16, fadst8 }, // ADST_ADST
+#if CONFIG_EXT_TX
{ fadst16, fdct8 }, // FLIPADST_DCT
{ fdct16, fadst8 }, // DCT_FLIPADST
{ fadst16, fadst8 }, // FLIPADST_FLIPADST
@@ -1296,6 +1306,7 @@
{ fidtx16, fadst8 }, // H_ADST
{ fadst16, fidtx8 }, // V_FLIPADST
{ fidtx16, fadst8 }, // H_FLIPADST
+#endif
};
const transform_2d ht = FHT[tx_type];
const int n = 8;
@@ -1303,8 +1314,10 @@
tran_low_t out[16 * 8];
tran_low_t temp_in[16], temp_out[16];
int i, j;
+#if CONFIG_EXT_TX
int16_t flipped_input[16 * 8];
maybe_flip_input(&input, &stride, n2, n, flipped_input, tx_type);
+#endif
// Columns
for (i = 0; i < n; ++i) {
@@ -1332,6 +1345,7 @@
{ fadst8, fdct16 }, // ADST_DCT
{ fdct8, fadst16 }, // DCT_ADST
{ fadst8, fadst16 }, // ADST_ADST
+#if CONFIG_EXT_TX
{ fadst8, fdct16 }, // FLIPADST_DCT
{ fdct8, fadst16 }, // DCT_FLIPADST
{ fadst8, fadst16 }, // FLIPADST_FLIPADST
@@ -1344,6 +1358,7 @@
{ fidtx8, fadst16 }, // H_ADST
{ fadst8, fidtx16 }, // V_FLIPADST
{ fidtx8, fadst16 }, // H_FLIPADST
+#endif
};
const transform_2d ht = FHT[tx_type];
const int n = 8;
@@ -1351,8 +1366,10 @@
tran_low_t out[16 * 8];
tran_low_t temp_in[16], temp_out[16];
int i, j;
+#if CONFIG_EXT_TX
int16_t flipped_input[16 * 8];
maybe_flip_input(&input, &stride, n, n2, flipped_input, tx_type);
+#endif
// Columns
for (i = 0; i < n2; ++i) {
@@ -1380,6 +1397,7 @@
{ fhalfright32, fdct16 }, // ADST_DCT
{ fdct32, fadst16 }, // DCT_ADST
{ fhalfright32, fadst16 }, // ADST_ADST
+#if CONFIG_EXT_TX
{ fhalfright32, fdct16 }, // FLIPADST_DCT
{ fdct32, fadst16 }, // DCT_FLIPADST
{ fhalfright32, fadst16 }, // FLIPADST_FLIPADST
@@ -1392,6 +1410,7 @@
{ fidtx32, fadst16 }, // H_ADST
{ fhalfright32, fidtx16 }, // V_FLIPADST
{ fidtx32, fadst16 }, // H_FLIPADST
+#endif
};
const transform_2d ht = FHT[tx_type];
const int n = 16;
@@ -1399,8 +1418,10 @@
tran_low_t out[32 * 16];
tran_low_t temp_in[32], temp_out[32];
int i, j;
+#if CONFIG_EXT_TX
int16_t flipped_input[32 * 16];
maybe_flip_input(&input, &stride, n2, n, flipped_input, tx_type);
+#endif
// Columns
for (i = 0; i < n; ++i) {
@@ -1428,6 +1449,7 @@
{ fadst16, fdct32 }, // ADST_DCT
{ fdct16, fhalfright32 }, // DCT_ADST
{ fadst16, fhalfright32 }, // ADST_ADST
+#if CONFIG_EXT_TX
{ fadst16, fdct32 }, // FLIPADST_DCT
{ fdct16, fhalfright32 }, // DCT_FLIPADST
{ fadst16, fhalfright32 }, // FLIPADST_FLIPADST
@@ -1440,6 +1462,7 @@
{ fidtx16, fhalfright32 }, // H_ADST
{ fadst16, fidtx32 }, // V_FLIPADST
{ fidtx16, fhalfright32 }, // H_FLIPADST
+#endif
};
const transform_2d ht = FHT[tx_type];
const int n = 16;
@@ -1447,8 +1470,10 @@
tran_low_t out[32 * 16];
tran_low_t temp_in[32], temp_out[32];
int i, j;
+#if CONFIG_EXT_TX
int16_t flipped_input[32 * 16];
maybe_flip_input(&input, &stride, n, n2, flipped_input, tx_type);
+#endif
// Columns
for (i = 0; i < n2; ++i) {
@@ -1469,8 +1494,6 @@
// Note: overall scale factor of transform is 4 times unitary
}
-#endif // CONFIG_EXT_TX
-
void av1_fdct8x8_quant_c(const int16_t *input, int stride,
tran_low_t *coeff_ptr, intptr_t n_coeffs,
int skip_block, const int16_t *zbin_ptr,
@@ -1759,7 +1782,6 @@
av1_fht4x4_c(input, output, stride, tx_type);
}
-#if CONFIG_EXT_TX
void av1_highbd_fht4x8_c(const int16_t *input, tran_low_t *output, int stride,
int tx_type) {
av1_fht4x8_c(input, output, stride, tx_type);
@@ -1789,7 +1811,6 @@
int tx_type) {
av1_fht32x16_c(input, output, stride, tx_type);
}
-#endif // CONFIG_EXT_TX
void av1_highbd_fht8x8_c(const int16_t *input, tran_low_t *output, int stride,
int tx_type) {
@@ -1874,12 +1895,103 @@
}
}
+#if CONFIG_TX64X64
+#if CONFIG_EXT_TX
+static void fidtx64(const tran_low_t *input, tran_low_t *output) {
+ int i;
+ for (i = 0; i < 64; ++i)
+ output[i] = (tran_low_t)fdct_round_shift(input[i] * 4 * Sqrt2);
+}
+
+// For use in lieu of ADST
+static void fhalfright64(const tran_low_t *input, tran_low_t *output) {
+ int i;
+ tran_low_t inputhalf[32];
+ for (i = 0; i < 32; ++i) {
+ output[32 + i] = (tran_low_t)fdct_round_shift(input[i] * 4 * Sqrt2);
+ }
+ // Multiply input by sqrt(2)
+ for (i = 0; i < 32; ++i) {
+ inputhalf[i] = (tran_low_t)fdct_round_shift(input[i + 32] * Sqrt2);
+ }
+ fdct32(inputhalf, output);
+ // Note overall scaling factor is 2 times unitary
+}
+#endif // CONFIG_EXT_TX
+
+static void fdct64_col(const tran_low_t *input, tran_low_t *output) {
+ int32_t in[64], out[64];
+ int i;
+ for (i = 0; i < 64; ++i) in[i] = (int32_t)input[i];
+ av1_fdct64_new(in, out, fwd_cos_bit_col_dct_dct_64,
+ fwd_stage_range_col_dct_dct_64);
+ for (i = 0; i < 64; ++i) output[i] = (tran_low_t)out[i];
+}
+
+static void fdct64_row(const tran_low_t *input, tran_low_t *output) {
+ int32_t in[64], out[64];
+ int i;
+ for (i = 0; i < 64; ++i) in[i] = (int32_t)input[i];
+ av1_fdct64_new(in, out, fwd_cos_bit_row_dct_dct_64,
+ fwd_stage_range_row_dct_dct_64);
+ for (i = 0; i < 64; ++i) output[i] = (tran_low_t)out[i];
+}
+
+void av1_fht64x64_c(const int16_t *input, tran_low_t *output, int stride,
+ int tx_type) {
+ static const transform_2d FHT[] = {
+ { fdct64_col, fdct64_row }, // DCT_DCT
+#if CONFIG_EXT_TX
+ { fhalfright64, fdct64_row }, // ADST_DCT
+ { fdct64_col, fhalfright64 }, // DCT_ADST
+ { fhalfright64, fhalfright64 }, // ADST_ADST
+ { fhalfright64, fdct64_row }, // FLIPADST_DCT
+ { fdct64_col, fhalfright64 }, // DCT_FLIPADST
+ { fhalfright64, fhalfright64 }, // FLIPADST_FLIPADST
+ { fhalfright64, fhalfright64 }, // ADST_FLIPADST
+ { fhalfright64, fhalfright64 }, // FLIPADST_ADST
+ { fidtx64, fidtx64 }, // IDTX
+ { fdct64_col, fidtx64 }, // V_DCT
+ { fidtx64, fdct64_row }, // H_DCT
+ { fhalfright64, fidtx64 }, // V_ADST
+ { fidtx64, fhalfright64 }, // H_ADST
+ { fhalfright64, fidtx64 }, // V_FLIPADST
+ { fidtx64, fhalfright64 }, // H_FLIPADST
+#endif
+ };
+ const transform_2d ht = FHT[tx_type];
+ tran_low_t out[4096];
+ int i, j;
+ tran_low_t temp_in[64], temp_out[64];
+#if CONFIG_EXT_TX
+ int16_t flipped_input[64 * 64];
+ maybe_flip_input(&input, &stride, 64, 64, flipped_input, tx_type);
+#endif
+ // Columns
+ for (i = 0; i < 64; ++i) {
+ for (j = 0; j < 64; ++j) temp_in[j] = input[j * stride + i];
+ ht.cols(temp_in, temp_out);
+ for (j = 0; j < 64; ++j)
+ out[j * 64 + i] = (temp_out[j] + 1 + (temp_out[j] > 0)) >> 2;
+ }
+
+ // Rows
+ for (i = 0; i < 64; ++i) {
+ for (j = 0; j < 64; ++j) temp_in[j] = out[j + i * 64];
+ ht.rows(temp_in, temp_out);
+ for (j = 0; j < 64; ++j)
+ output[j + i * 64] =
+ (tran_low_t)((temp_out[j] + 1 + (temp_out[j] < 0)) >> 2);
+ }
+}
+#endif // CONFIG_TX64X64
+
#if CONFIG_EXT_TX
// Forward identity transform.
void av1_fwd_idtx_c(const int16_t *src_diff, tran_low_t *coeff, int stride,
int bs, int tx_type) {
int r, c;
- const int shift = bs < 32 ? 3 : 2;
+ const int shift = bs < 32 ? 3 : (bs < 64 ? 2 : 1);
if (tx_type == IDTX) {
for (r = 0; r < bs; ++r) {
for (c = 0; c < bs; ++c) coeff[c] = src_diff[c] * (1 << shift);
@@ -1894,5 +2006,12 @@
int tx_type) {
av1_fht32x32_c(input, output, stride, tx_type);
}
+
+#if CONFIG_TX64X64
+void av1_highbd_fht64x64_c(const int16_t *input, tran_low_t *output, int stride,
+ int tx_type) {
+ av1_fht64x64_c(input, output, stride, tx_type);
+}
+#endif // CONFIG_TX64X64
#endif // CONFIG_AOM_HIGHBITDEPTH
#endif // CONFIG_EXT_TX
diff --git a/av1/encoder/hybrid_fwd_txfm.c b/av1/encoder/hybrid_fwd_txfm.c
index 6d5eccd..ff03516 100644
--- a/av1/encoder/hybrid_fwd_txfm.c
+++ b/av1/encoder/hybrid_fwd_txfm.c
@@ -55,7 +55,6 @@
}
}
-#if CONFIG_EXT_TX
static void fwd_txfm_4x8(const int16_t *src_diff, tran_low_t *coeff,
int diff_stride, TX_TYPE tx_type,
FWD_TXFM_OPT fwd_txfm_opt) {
@@ -97,7 +96,6 @@
(void)fwd_txfm_opt;
av1_fht32x16(src_diff, coeff, diff_stride, tx_type);
}
-#endif // CONFIG_EXT_TX
static void fwd_txfm_8x8(const int16_t *src_diff, tran_low_t *coeff,
int diff_stride, TX_TYPE tx_type,
@@ -233,7 +231,6 @@
}
}
-#if CONFIG_EXT_TX
static void highbd_fwd_txfm_4x8(const int16_t *src_diff, tran_low_t *coeff,
int diff_stride, TX_TYPE tx_type,
FWD_TXFM_OPT fwd_txfm_opt, const int bd) {
@@ -281,7 +278,6 @@
(void)bd;
av1_highbd_fht32x16(src_diff, coeff, diff_stride, tx_type);
}
-#endif // CONFIG_EXT_TX
static void highbd_fwd_txfm_8x8(const int16_t *src_diff, tran_low_t *coeff,
int diff_stride, TX_TYPE tx_type,
@@ -403,7 +399,6 @@
case TX_8X8:
fwd_txfm_8x8(src_diff, coeff, diff_stride, tx_type, fwd_txfm_opt);
break;
-#if CONFIG_EXT_TX
case TX_4X8:
fwd_txfm_4x8(src_diff, coeff, diff_stride, tx_type, fwd_txfm_opt);
break;
@@ -422,7 +417,6 @@
case TX_32X16:
fwd_txfm_32x16(src_diff, coeff, diff_stride, tx_type, fwd_txfm_opt);
break;
-#endif // CONFIG_EXT_TX
case TX_4X4:
fwd_txfm_4x4(src_diff, coeff, diff_stride, tx_type, lossless);
break;
@@ -452,7 +446,6 @@
highbd_fwd_txfm_8x8(src_diff, coeff, diff_stride, tx_type, fwd_txfm_opt,
bd);
break;
-#if CONFIG_EXT_TX
case TX_4X8:
highbd_fwd_txfm_4x8(src_diff, coeff, diff_stride, tx_type, fwd_txfm_opt,
bd);
@@ -477,7 +470,6 @@
highbd_fwd_txfm_32x16(src_diff, coeff, diff_stride, tx_type, fwd_txfm_opt,
bd);
break;
-#endif // CONFIG_EXT_TX
case TX_4X4:
highbd_fwd_txfm_4x4(src_diff, coeff, diff_stride, tx_type, lossless, bd);
break;
diff --git a/av1/encoder/rdopt.c b/av1/encoder/rdopt.c
index b37714c..8682a3e 100644
--- a/av1/encoder/rdopt.c
+++ b/av1/encoder/rdopt.c
@@ -6745,24 +6745,24 @@
av1_set_mvcost(x, mbmi->ref_frame[0], 0, mbmi->ref_mv_idx);
#endif // CONFIG_REF_MV
rate_mv = av1_mv_bit_cost(&frame_mv[refs[0]].as_mv,
- &x->mbmi_ext->ref_mvs[refs[0]][0].as_mv,
+ &mbmi_ext->ref_mvs[refs[0]][0].as_mv,
x->nmvjointcost, x->mvcost, MV_COST_WEIGHT);
#if CONFIG_REF_MV
av1_set_mvcost(x, mbmi->ref_frame[1], 1, mbmi->ref_mv_idx);
#endif // CONFIG_REF_MV
rate_mv += av1_mv_bit_cost(
- &frame_mv[refs[1]].as_mv, &x->mbmi_ext->ref_mvs[refs[1]][0].as_mv,
+ &frame_mv[refs[1]].as_mv, &mbmi_ext->ref_mvs[refs[1]][0].as_mv,
x->nmvjointcost, x->mvcost, MV_COST_WEIGHT);
}
} else if (this_mode == NEAREST_NEWMV || this_mode == NEAR_NEWMV) {
frame_mv[refs[1]].as_int = single_newmv[refs[1]].as_int;
rate_mv = av1_mv_bit_cost(&frame_mv[refs[1]].as_mv,
- &x->mbmi_ext->ref_mvs[refs[1]][0].as_mv,
+ &mbmi_ext->ref_mvs[refs[1]][0].as_mv,
x->nmvjointcost, x->mvcost, MV_COST_WEIGHT);
} else {
frame_mv[refs[0]].as_int = single_newmv[refs[0]].as_int;
rate_mv = av1_mv_bit_cost(&frame_mv[refs[0]].as_mv,
- &x->mbmi_ext->ref_mvs[refs[0]][0].as_mv,
+ &mbmi_ext->ref_mvs[refs[0]][0].as_mv,
x->nmvjointcost, x->mvcost, MV_COST_WEIGHT);
}
#else
@@ -6778,13 +6778,13 @@
av1_set_mvcost(x, mbmi->ref_frame[0], 0, mbmi->ref_mv_idx);
#endif // CONFIG_REF_MV
rate_mv = av1_mv_bit_cost(&frame_mv[refs[0]].as_mv,
- &x->mbmi_ext->ref_mvs[refs[0]][0].as_mv,
+ &mbmi_ext->ref_mvs[refs[0]][0].as_mv,
x->nmvjointcost, x->mvcost, MV_COST_WEIGHT);
#if CONFIG_REF_MV
av1_set_mvcost(x, mbmi->ref_frame[1], 1, mbmi->ref_mv_idx);
#endif // CONFIG_REF_MV
rate_mv += av1_mv_bit_cost(&frame_mv[refs[1]].as_mv,
- &x->mbmi_ext->ref_mvs[refs[1]][0].as_mv,
+ &mbmi_ext->ref_mvs[refs[1]][0].as_mv,
x->nmvjointcost, x->mvcost, MV_COST_WEIGHT);
}
#endif // CONFIG_EXT_INTER
diff --git a/av1/encoder/x86/dct_intrin_sse2.c b/av1/encoder/x86/dct_intrin_sse2.c
index 22cc342..3596292 100644
--- a/av1/encoder/x86/dct_intrin_sse2.c
+++ b/av1/encoder/x86/dct_intrin_sse2.c
@@ -2592,7 +2592,6 @@
}
}
-#if CONFIG_EXT_TX
static INLINE void scale_sqrt2_8x4(__m128i *in) {
// Implements fdct_round_shift(input * Sqrt2), which is equivalent to
// ROUND_POWER_OF_TWO(input * Sqrt2, DCT_CONST_BITS),
@@ -2767,9 +2766,9 @@
int tx_type) {
__m128i in[8];
+ load_buffer_4x8(input, in, stride, 0, 0);
switch (tx_type) {
case DCT_DCT:
- load_buffer_4x8(input, in, stride, 0, 0);
fdct8_sse2(in);
// Repack data into two 4x4 blocks so we can reuse the 4x4 transforms
// The other cases (and the 8x4 transforms) all behave similarly
@@ -2781,7 +2780,6 @@
fdct4_sse2(in + 4);
break;
case ADST_DCT:
- load_buffer_4x8(input, in, stride, 0, 0);
fadst8_sse2(in);
in[4] = _mm_shuffle_epi32(in[0], 0xe);
in[5] = _mm_shuffle_epi32(in[1], 0xe);
@@ -2791,7 +2789,6 @@
fdct4_sse2(in + 4);
break;
case DCT_ADST:
- load_buffer_4x8(input, in, stride, 0, 0);
fdct8_sse2(in);
in[4] = _mm_shuffle_epi32(in[0], 0xe);
in[5] = _mm_shuffle_epi32(in[1], 0xe);
@@ -2801,7 +2798,6 @@
fadst4_sse2(in + 4);
break;
case ADST_ADST:
- load_buffer_4x8(input, in, stride, 0, 0);
fadst8_sse2(in);
in[4] = _mm_shuffle_epi32(in[0], 0xe);
in[5] = _mm_shuffle_epi32(in[1], 0xe);
@@ -2810,6 +2806,7 @@
fadst4_sse2(in);
fadst4_sse2(in + 4);
break;
+#if CONFIG_EXT_TX
case FLIPADST_DCT:
load_buffer_4x8(input, in, stride, 1, 0);
fadst8_sse2(in);
@@ -2930,6 +2927,7 @@
fadst4_sse2(in);
fadst4_sse2(in + 4);
break;
+#endif
default: assert(0); break;
}
write_buffer_4x8(output, in);
@@ -3023,6 +3021,7 @@
fadst4_sse2(in + 4);
fadst8_sse2(in);
break;
+#if CONFIG_EXT_TX
case FLIPADST_DCT:
load_buffer_8x4(input, in, stride, 1, 0);
fadst4_sse2(in);
@@ -3095,6 +3094,7 @@
fidtx4_sse2(in + 4);
fadst8_sse2(in);
break;
+#endif
default: assert(0); break;
}
write_buffer_8x4(output, in);
@@ -3158,6 +3158,7 @@
fadst8_sse2(t);
fadst8_sse2(b);
break;
+#if CONFIG_EXT_TX
case FLIPADST_DCT:
load_buffer_8x16(input, in, stride, 1, 0);
fadst16_8col(in);
@@ -3254,6 +3255,7 @@
fadst8_sse2(t);
fadst8_sse2(b);
break;
+#endif
default: assert(0); break;
}
right_shift_8x8(t, 2);
@@ -3314,6 +3316,7 @@
fadst8_sse2(r);
fadst16_8col(in);
break;
+#if CONFIG_EXT_TX
case FLIPADST_DCT:
load_buffer_16x8(input, in, stride, 1, 0);
fadst8_sse2(l);
@@ -3386,6 +3389,7 @@
fidtx8_sse2(r);
fadst16_8col(in);
break;
+#endif
default: assert(0); break;
}
array_transpose_8x8(l, l);
@@ -3436,6 +3440,7 @@
fdct16_sse2(tl, tr);
}
+#if CONFIG_EXT_TX
static INLINE void fidtx32_16col(__m128i *tl, __m128i *tr, __m128i *bl,
__m128i *br) {
int i;
@@ -3448,6 +3453,7 @@
array_transpose_16x16(tl, tr);
array_transpose_16x16(bl, br);
}
+#endif
static INLINE void load_buffer_16x32(const int16_t *input, __m128i *intl,
__m128i *intr, __m128i *inbl,
@@ -3507,7 +3513,7 @@
}
}
-// Note on data layout, for both this and the 32x16 tranforms:
+// Note on data layout, for both this and the 32x16 transforms:
// So that we can reuse the 16-element transforms easily,
// we want to split the input into 8x16 blocks.
// For 16x32, this means the input is a 2x2 grid of such blocks.
@@ -3541,6 +3547,7 @@
fadst16_sse2(intl, intr);
fadst16_sse2(inbl, inbr);
break;
+#if CONFIG_EXT_TX
case FLIPADST_DCT:
load_buffer_16x32(input, intl, intr, inbl, inbr, stride, 1, 0);
fhalfright32_16col(intl, intr, inbl, inbr);
@@ -3613,6 +3620,7 @@
fadst16_sse2(intl, intr);
fadst16_sse2(inbl, inbr);
break;
+#endif
default: assert(0); break;
}
write_buffer_16x32(output, intl, intr, inbl, inbr);
@@ -3671,31 +3679,29 @@
int tx_type) {
__m128i in0[16], in1[16], in2[16], in3[16];
+ load_buffer_32x16(input, in0, in1, in2, in3, stride, 0, 0);
switch (tx_type) {
case DCT_DCT:
- load_buffer_32x16(input, in0, in1, in2, in3, stride, 0, 0);
fdct16_sse2(in0, in1);
fdct16_sse2(in2, in3);
fdct32_16col(in0, in1, in2, in3);
break;
case ADST_DCT:
- load_buffer_32x16(input, in0, in1, in2, in3, stride, 0, 0);
fadst16_sse2(in0, in1);
fadst16_sse2(in2, in3);
fdct32_16col(in0, in1, in2, in3);
break;
case DCT_ADST:
- load_buffer_32x16(input, in0, in1, in2, in3, stride, 0, 0);
fdct16_sse2(in0, in1);
fdct16_sse2(in2, in3);
fhalfright32_16col(in0, in1, in2, in3);
break;
case ADST_ADST:
- load_buffer_32x16(input, in0, in1, in2, in3, stride, 0, 0);
fadst16_sse2(in0, in1);
fadst16_sse2(in2, in3);
fhalfright32_16col(in0, in1, in2, in3);
break;
+#if CONFIG_EXT_TX
case FLIPADST_DCT:
load_buffer_32x16(input, in0, in1, in2, in3, stride, 1, 0);
fadst16_sse2(in0, in1);
@@ -3768,8 +3774,8 @@
fidtx16_sse2(in2, in3);
fhalfright32_16col(in0, in1, in2, in3);
break;
+#endif
default: assert(0); break;
}
write_buffer_32x16(output, in0, in1, in2, in3);
}
-#endif // CONFIG_EXT_TX
