Merge "Support rectangular transform block units in the codebase" into nextgenv2
diff --git a/av1/encoder/dct.c b/av1/encoder/dct.c
index dd4031f..c002dab 100644
--- a/av1/encoder/dct.c
+++ b/av1/encoder/dct.c
@@ -999,6 +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;
@@ -1022,21 +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;
@@ -1184,7 +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[] = {
@@ -1192,6 +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
@@ -1204,6 +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;
@@ -1211,8 +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) {
@@ -1239,6 +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
@@ -1251,6 +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;
@@ -1258,8 +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) {
@@ -1286,6 +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
@@ -1298,6 +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;
@@ -1305,8 +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) {
@@ -1334,6 +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
@@ -1346,6 +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;
@@ -1353,8 +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) {
@@ -1382,6 +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
@@ -1394,6 +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;
@@ -1401,8 +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) {
@@ -1430,6 +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
@@ -1442,6 +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;
@@ -1449,8 +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) {
@@ -1471,8 +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,
@@ -1761,7 +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);
@@ -1791,7 +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) {
