generic-library/vpx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Adds 1D transforms for ADST/FlipADST to make 16

Browse Source 
Makes a set of 16 transforms total, adding all 1D
combinations of ADST and FlipADST, and removng all DST
transforms.

lowres, midres both improve by about 0.1% and hdres by
-0.378% in BDRATE but with fewer transforms that are also
simpler.

Further experiments to continue later.

Change-Id: I7348a4c0e12078fdea5ae3a2d36a89a319ffcc6e
This commit is contained in:
Debargha Mukherjee
2016-03-14 22:30:09 -07:00
parent 4914ae4622
commit 1b17559327
13 changed files with 445 additions and 2044 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
test/vp10_fht16x16_test.cc
View File

@@ -103,20 +103,6 @@ INSTANTIATE_TEST_CASE_P(
make_tuple(&vp10_fht16x16_sse2, &vp10_iht16x16_256_add_sse2, 7,
VPX_BITS_8, 256),
make_tuple(&vp10_fht16x16_sse2, &vp10_iht16x16_256_add_sse2, 8,
VPX_BITS_8, 256),
make_tuple(&vp10_fht16x16_sse2, &vp10_iht16x16_256_add_sse2, 9,
VPX_BITS_8, 256),
make_tuple(&vp10_fht16x16_sse2, &vp10_iht16x16_256_add_sse2, 10,
VPX_BITS_8, 256),
make_tuple(&vp10_fht16x16_sse2, &vp10_iht16x16_256_add_sse2, 11,
VPX_BITS_8, 256),
make_tuple(&vp10_fht16x16_sse2, &vp10_iht16x16_256_add_sse2, 12,
VPX_BITS_8, 256),
make_tuple(&vp10_fht16x16_sse2, &vp10_iht16x16_256_add_sse2, 13,
VPX_BITS_8, 256),
make_tuple(&vp10_fht16x16_sse2, &vp10_iht16x16_256_add_sse2, 14,
VPX_BITS_8, 256),
make_tuple(&vp10_fht16x16_sse2, &vp10_iht16x16_256_add_sse2, 15,
VPX_BITS_8, 256)));
#endif // !CONFIG_EXT_TX
#endif // HAVE_SSE2

14
test/vp10_fht4x4_test.cc
View File

@@ -102,20 +102,6 @@ INSTANTIATE_TEST_CASE_P(
make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 7,
VPX_BITS_8, 16),
make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 8,
VPX_BITS_8, 16),
make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 9,
VPX_BITS_8, 16),
make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 10,
VPX_BITS_8, 16),
make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 11,
VPX_BITS_8, 16),
make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 12,
VPX_BITS_8, 16),
make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 13,
VPX_BITS_8, 16),
make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 14,
VPX_BITS_8, 16),
make_tuple(&vp10_fht4x4_sse2, &vp10_iht4x4_16_add_sse2, 15,
VPX_BITS_8, 16)));
#endif // !CONFIG_EXT_TX
#endif // HAVE_SSE2

14
test/vp10_fht8x8_test.cc
View File

@@ -102,20 +102,6 @@ INSTANTIATE_TEST_CASE_P(
make_tuple(&vp10_fht8x8_sse2, &vp10_iht8x8_64_add_sse2, 7,
VPX_BITS_8, 64),
make_tuple(&vp10_fht8x8_sse2, &vp10_iht8x8_64_add_sse2, 8,
VPX_BITS_8, 64),
make_tuple(&vp10_fht8x8_sse2, &vp10_iht8x8_64_add_sse2, 9,
VPX_BITS_8, 64),
make_tuple(&vp10_fht8x8_sse2, &vp10_iht8x8_64_add_sse2, 10,
VPX_BITS_8, 64),
make_tuple(&vp10_fht8x8_sse2, &vp10_iht8x8_64_add_sse2, 11,
VPX_BITS_8, 64),
make_tuple(&vp10_fht8x8_sse2, &vp10_iht8x8_64_add_sse2, 12,
VPX_BITS_8, 64),
make_tuple(&vp10_fht8x8_sse2, &vp10_iht8x8_64_add_sse2, 13,
VPX_BITS_8, 64),
make_tuple(&vp10_fht8x8_sse2, &vp10_iht8x8_64_add_sse2, 14,
VPX_BITS_8, 64),
make_tuple(&vp10_fht8x8_sse2, &vp10_iht8x8_64_add_sse2, 15,
VPX_BITS_8, 64)));
#endif // !CONFIG_EXT_TX
#endif // HAVE_SSE2

18
vp10/common/blockd.h
View File

@@ -383,10 +383,10 @@ static INLINE int supertx_enabled(const MB_MODE_INFO *mbmi) {
#define USE_MSKTX_FOR_32X32 0
static const int num_ext_tx_set_inter[EXT_TX_SETS_INTER] = {
1, 19, 12, 2
1, 16, 12, 2
};
static const int num_ext_tx_set_intra[EXT_TX_SETS_INTRA] = {
1, 17, 10
1, 12, 10
};
#if EXT_TX_SIZES == 4
@@ -437,17 +437,17 @@ static const int use_inter_ext_tx_for_txsize[EXT_TX_SETS_INTER][TX_SIZES] = {
// Transform types used in each intra set
static const int ext_tx_used_intra[EXT_TX_SETS_INTRA][TX_TYPES] = {
{ 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, },
{ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, },
{ 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, },
{1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
{1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0},
{1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0},
};
// Transform types used in each inter set
static const int ext_tx_used_inter[EXT_TX_SETS_INTER][TX_TYPES] = {
{ 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 },
{ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1},
{ 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1},
{ 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0},
{1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0},
{1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1},
{1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1},
{1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0},
};
static INLINE int get_ext_tx_types(TX_SIZE tx_size, BLOCK_SIZE bs,

436
vp10/common/entropymode.c
View File

@@ -836,47 +836,27 @@ static const vpx_prob default_switchable_interp_prob[SWITCHABLE_FILTER_CONTEXTS]
const vpx_tree_index vp10_ext_tx_inter_tree[EXT_TX_SETS_INTER]
[TREE_SIZE(TX_TYPES)] = {
{ // ToDo(yaowu): remove used entry 0.
-IDTX, 2,
-V_DCT, 4,
-H_DCT, 6,
-DCT_DCT, 8,
-DST_DST, 10,
12, 22,
14, 16,
-DST_DCT, -DCT_DST,
18, 20,
-ADST_DCT, -DCT_ADST,
-FLIPADST_DCT, -DCT_FLIPADST,
24, 30,
26, 28,
-DST_ADST, -ADST_DST,
-DST_FLIPADST, -FLIPADST_DST,
32, 34,
-ADST_ADST, -FLIPADST_FLIPADST,
-ADST_FLIPADST, -FLIPADST_ADST,
0
}, {
-IDTX, 2,
-V_DCT, 4,
-H_DCT, 6,
-DCT_DCT, 8,
-DST_DST, 10,
12, 22,
14, 16,
-DST_DCT, -DCT_DST,
18, 20,
4, 14,
6, 8,
-V_DCT, -H_DCT,
10, 12,
-V_ADST, -H_ADST,
-V_FLIPADST, -H_FLIPADST,
-DCT_DCT, 16,
18, 24,
20, 22,
-ADST_DCT, -DCT_ADST,
-FLIPADST_DCT, -DCT_FLIPADST,
24, 30,
26, 28,
-DST_ADST, -ADST_DST,
-DST_FLIPADST, -FLIPADST_DST,
32, 34,
-ADST_ADST, -FLIPADST_FLIPADST,
-ADST_FLIPADST, -FLIPADST_ADST,
-ADST_FLIPADST, -FLIPADST_ADST
}, {
-IDTX, 2,
-V_DCT, 4,
-H_DCT, 6,
4, 6,
-V_DCT, -H_DCT,
-DCT_DCT, 8,
10, 16,
12, 14,
@@ -893,39 +873,19 @@ const vpx_tree_index vp10_ext_tx_inter_tree[EXT_TX_SETS_INTER]
const vpx_tree_index vp10_ext_tx_intra_tree[EXT_TX_SETS_INTRA]
[TREE_SIZE(TX_TYPES)] = {
{ // ToDo(yaowu): remove unused entry 0.
-IDTX, 2,
-DCT_DCT, 4,
-DST_DST, 6,
8, 18,
10, 12,
-DST_DCT, -DCT_DST,
14, 16,
-ADST_DCT, -DCT_ADST,
-FLIPADST_DCT, -DCT_FLIPADST,
20, 26,
22, 24,
-DST_ADST, -ADST_DST,
-DST_FLIPADST, -FLIPADST_DST,
28, 30,
-ADST_ADST, -FLIPADST_FLIPADST,
-ADST_FLIPADST, -FLIPADST_ADST,
0
}, {
-IDTX, 2,
-DCT_DCT, 4,
-DST_DST, 6,
8, 18,
10, 12,
-DST_DCT, -DCT_DST,
14, 16,
6, 8,
-V_DCT, -H_DCT,
10, 16,
12, 14,
-ADST_DCT, -DCT_ADST,
-FLIPADST_DCT, -DCT_FLIPADST,
20, 26,
22, 24,
-DST_ADST, -ADST_DST,
-DST_FLIPADST, -FLIPADST_DST,
28, 30,
18, 20,
-ADST_ADST, -FLIPADST_FLIPADST,
-ADST_FLIPADST, -FLIPADST_ADST,
-ADST_FLIPADST, -FLIPADST_ADST
}, {
-IDTX, 2,
-DCT_DCT, 4,
@@ -942,33 +902,25 @@ const vpx_tree_index vp10_ext_tx_intra_tree[EXT_TX_SETS_INTRA]
static const vpx_prob
default_inter_ext_tx_prob[EXT_TX_SETS_INTER][EXT_TX_SIZES][TX_TYPES - 1] = {
{ // ToDo(yaowu): remove unused entry 0.
{ 12, 15, 15, 112, 16, 128, 128, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128 },
{ 12, 15, 15, 112, 16, 128, 128, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128 },
{ 12, 15, 15, 112, 16, 128, 128, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128 },
{ 0 },
{ 0 },
{ 0 },
#if EXT_TX_SIZES == 4
{ 12, 15, 15, 112, 16, 128, 128, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128 },
{ 0 },
#endif
}, {
{ 12, 15, 15, 112, 16, 128, 128, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128 },
{ 12, 15, 15, 112, 16, 128, 128, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128 },
{ 12, 15, 15, 112, 16, 128, 128, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128 },
{ 10, 24, 30, 128, 128, 128, 128, 112, 160, 128, 128, 128, 128, 128, 128},
{ 10, 24, 30, 128, 128, 128, 128, 112, 160, 128, 128, 128, 128, 128, 128},
{ 10, 24, 30, 128, 128, 128, 128, 112, 160, 128, 128, 128, 128, 128, 128},
#if EXT_TX_SIZES == 4
{ 12, 15, 15, 160, 16, 144, 160, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128 },
{ 10, 24, 30, 128, 128, 128, 128, 112, 160, 128, 128, 128, 128, 128, 128},
#endif
}, {
{ 12, 15, 15, 112, 128, 128, 128, 128, 128, 128, 128 },
{ 12, 15, 15, 112, 128, 128, 128, 128, 128, 128, 128 },
{ 12, 15, 15, 112, 128, 128, 128, 128, 128, 128, 128 },
{ 10, 30, 128, 112, 160, 128, 128, 128, 128, 128, 128 },
{ 10, 30, 128, 112, 160, 128, 128, 128, 128, 128, 128 },
{ 10, 30, 128, 112, 160, 128, 128, 128, 128, 128, 128 },
#if EXT_TX_SIZES == 4
{ 12, 15, 15, 160, 128, 128, 128, 128, 128, 128, 128 },
{ 10, 30, 128, 112, 160, 128, 128, 128, 128, 128, 128 },
#endif
}, {
{ 12, },
@@ -985,266 +937,110 @@ default_intra_ext_tx_prob[EXT_TX_SETS_INTRA][EXT_TX_SIZES]
[INTRA_MODES][TX_TYPES - 1] = {
{ // ToDo(yaowu): remove unused entry 0.
{
{ 8, 11, 24, 112, 87, 137, 127, 134,
128, 86, 128, 124, 125, 133, 176, 123, },
{ 10, 9, 39, 106, 73, 155, 163, 228,
35, 62, 129, 127, 133, 114, 213, 234, },
{ 10, 9, 14, 88, 91, 127, 151, 51,
210, 89, 126, 58, 52, 116, 217, 24, },
{ 9, 6, 29, 113, 98, 131, 149, 210,
119, 60, 124, 93, 90, 143, 170, 197, },
{ 8, 8, 38, 101, 111, 166, 167, 141,
130, 105, 128, 75, 75, 118, 197, 117, },
{ 7, 8, 39, 91, 101, 153, 166, 200,
99, 77, 123, 90, 83, 144, 224, 192, },
{ 7, 10, 26, 86, 119, 154, 130, 101,
152, 91, 129, 75, 79, 137, 219, 77, },
{ 10, 13, 20, 86, 102, 162, 112, 76,
171, 86, 134, 122, 106, 124, 196, 44, },
{ 8, 9, 33, 108, 100, 144, 148, 215,
77, 60, 125, 125, 128, 126, 198, 220, },
{ 3, 10, 29, 111, 69, 141, 204, 141,
139, 93, 120, 75, 77, 163, 242, 124, },
{ 0 }, { 0 }, { 0 }, { 0 }, { 0 }, { 0 }, { 0 }, { 0 }, { 0 }, { 0 },
}, {
{ 2, 53, 18, 147, 96, 98, 136, 133,
131, 120, 153, 163, 169, 137, 173, 124, },
{ 4, 18, 34, 133, 54, 130, 179, 228,
28, 72, 153, 164, 168, 118, 227, 239, },
{ 4, 18, 13, 125, 72, 110, 176, 36,
221, 104, 148, 75, 72, 117, 225, 19, },
{ 8, 33, 24, 162, 113, 99, 147, 226,
103, 85, 153, 143, 153, 124, 155, 210, },
{ 2, 15, 35, 107, 127, 158, 192, 128,
126, 116, 151, 95, 88, 182, 241, 119, },
{ 3, 15, 36, 112, 100, 146, 194, 189,
90, 98, 152, 99, 100, 165, 235, 175, },
{ 3, 16, 29, 109, 103, 140, 182, 76,
173, 104, 147, 82, 85, 159, 235, 70, },
{ 9, 24, 14, 120, 86, 156, 161, 34,
177, 121, 142, 128, 128, 126, 185, 37, },
{ 5, 24, 29, 152, 98, 99, 174, 228,
82, 76, 147, 149, 128, 132, 191, 225, },
{ 2, 15, 29, 111, 77, 126, 200, 135,
117, 93, 152, 96, 84, 191, 245, 135, },
{ 0 }, { 0 }, { 0 }, { 0 }, { 0 }, { 0 }, { 0 }, { 0 }, { 0 }, { 0 },
}, {
{ 2, 69, 13, 173, 111, 69, 137, 159,
159, 146, 151, 193, 203, 131, 180, 123, },
{ 1, 12, 33, 164, 32, 98, 204, 242,
23, 99, 149, 215, 232, 110, 239, 245, },
{ 1, 17, 9, 136, 82, 83, 171, 28,
231, 128, 135, 76, 64, 118, 235, 17, },
{ 4, 41, 17, 195, 131, 58, 161, 237,
141, 97, 153, 189, 191, 117, 182, 202, },
{ 2, 17, 36, 104, 149, 137, 217, 139,
191, 119, 125, 107, 115, 223, 249, 110, },
{ 2, 14, 24, 127, 91, 135, 219, 198,
113, 91, 164, 125, 173, 211, 250, 116, },
{ 3, 19, 24, 120, 102, 130, 209, 81,
187, 95, 143, 102, 50, 190, 244, 56, },
{ 4, 27, 10, 128, 91, 157, 181, 33,
181, 150, 141, 141, 166, 114, 215, 25, },
{ 2, 34, 27, 187, 102, 77, 210, 245,
113, 107, 136, 184, 188, 121, 210, 234, },
{ 1, 15, 22, 141, 59, 94, 208, 133,
154, 95, 152, 112, 105, 191, 242, 111, },
{ 0 }, { 0 }, { 0 }, { 0 }, { 0 }, { 0 }, { 0 }, { 0 }, { 0 }, { 0 },
#if EXT_TX_SIZES == 4
}, {
{ 2, 69, 13, 173, 111, 69, 137, 159,
159, 146, 151, 193, 203, 131, 180, 123, },
{ 1, 12, 33, 164, 32, 98, 204, 242,
23, 99, 149, 215, 232, 110, 239, 245, },
{ 1, 17, 9, 136, 82, 83, 171, 28,
231, 128, 135, 76, 64, 118, 235, 17, },
{ 4, 41, 17, 195, 131, 58, 161, 237,
141, 97, 153, 189, 191, 117, 182, 202, },
{ 2, 17, 36, 104, 149, 137, 217, 139,
191, 119, 125, 107, 115, 223, 249, 110, },
{ 2, 14, 24, 127, 91, 135, 219, 198,
113, 91, 164, 125, 173, 211, 250, 116, },
{ 3, 19, 24, 120, 102, 130, 209, 81,
187, 95, 143, 102, 50, 190, 244, 56, },
{ 4, 27, 10, 128, 91, 157, 181, 33,
181, 150, 141, 141, 166, 114, 215, 25, },
{ 2, 34, 27, 187, 102, 77, 210, 245,
113, 107, 136, 184, 188, 121, 210, 234, },
{ 1, 15, 22, 141, 59, 94, 208, 133,
154, 95, 152, 112, 105, 191, 242, 111, },
{ 0 }, { 0 }, { 0 }, { 0 }, { 0 }, { 0 }, { 0 }, { 0 }, { 0 }, { 0 },
#endif
},
}, {
{
{ 8, 11, 24, 112, 87, 137, 127, 134,
128, 86, 128, 124, 125, 133, 176, 123, },
{ 10, 9, 39, 106, 73, 155, 163, 228,
35, 62, 129, 127, 133, 114, 213, 234, },
{ 10, 9, 14, 88, 91, 127, 151, 51,
210, 89, 126, 58, 52, 116, 217, 24, },
{ 9, 6, 29, 113, 98, 131, 149, 210,
119, 60, 124, 93, 90, 143, 170, 197, },
{ 8, 8, 38, 101, 111, 166, 167, 141,
130, 105, 128, 75, 75, 118, 197, 117, },
{ 7, 8, 39, 91, 101, 153, 166, 200,
99, 77, 123, 90, 83, 144, 224, 192, },
{ 7, 10, 26, 86, 119, 154, 130, 101,
152, 91, 129, 75, 79, 137, 219, 77, },
{ 10, 13, 20, 86, 102, 162, 112, 76,
171, 86, 134, 122, 106, 124, 196, 44, },
{ 8, 9, 33, 108, 100, 144, 148, 215,
77, 60, 125, 125, 128, 126, 198, 220, },
{ 3, 10, 29, 111, 69, 141, 204, 141,
139, 93, 120, 75, 77, 163, 242, 124, },
{ 8, 176, 32, 128, 128, 128, 128, 128, 128, 128, 128, },
{ 10, 28, 32, 128, 176, 192, 208, 128, 128, 128, 128, },
{ 10, 28, 32, 128, 176, 192, 48, 128, 128, 128, 128, },
{ 9, 160, 32, 128, 128, 128, 128, 128, 128, 128, 128, },
{ 8, 28, 32, 128, 96, 128, 128, 128, 160, 192, 128, },
{ 7, 28, 32, 128, 160, 176, 192, 128, 128, 128, 128, },
{ 7, 20, 32, 128, 160, 176, 64, 128, 128, 128, 128, },
{ 10, 23, 32, 128, 160, 176, 64, 128, 128, 128, 128, },
{ 8, 29, 32, 128, 160, 176, 192, 128, 128, 128, 128, },
{ 3, 20, 32, 128, 96, 128, 128, 128, 160, 192, 128, },
}, {
{ 2, 53, 18, 147, 96, 98, 136, 133,
131, 120, 153, 163, 169, 137, 173, 124, },
{ 4, 18, 34, 133, 54, 130, 179, 228,
28, 72, 153, 164, 168, 118, 227, 239, },
{ 4, 18, 13, 125, 72, 110, 176, 36,
221, 104, 148, 75, 72, 117, 225, 19, },
{ 8, 33, 24, 162, 113, 99, 147, 226,
103, 85, 153, 143, 153, 124, 155, 210, },
{ 2, 15, 35, 107, 127, 158, 192, 128,
126, 116, 151, 95, 88, 182, 241, 119, },
{ 3, 15, 36, 112, 100, 146, 194, 189,
90, 98, 152, 99, 100, 165, 235, 175, },
{ 3, 16, 29, 109, 103, 140, 182, 76,
173, 104, 147, 82, 85, 159, 235, 70, },
{ 9, 24, 14, 120, 86, 156, 161, 34,
177, 121, 142, 128, 128, 126, 185, 37, },
{ 5, 24, 29, 152, 98, 99, 174, 228,
82, 76, 147, 149, 128, 132, 191, 225, },
{ 2, 15, 29, 111, 77, 126, 200, 135,
117, 93, 152, 96, 84, 191, 245, 135, },
{ 2, 176, 32, 128, 128, 128, 128, 128, 128, 128, 128, },
{ 4, 28, 32, 128, 176, 192, 208, 128, 128, 128, 128, },
{ 4, 28, 32, 128, 176, 192, 48, 128, 128, 128, 128, },
{ 8, 160, 32, 128, 128, 128, 128, 128, 128, 128, 128, },
{ 2, 28, 32, 128, 96, 128, 128, 128, 160, 192, 128, },
{ 3, 28, 32, 128, 160, 176, 192, 128, 128, 128, 128, },
{ 3, 26, 32, 128, 160, 176, 64, 128, 128, 128, 128, },
{ 9, 24, 32, 128, 160, 176, 64, 128, 128, 128, 128, },
{ 5, 24, 32, 128, 160, 176, 192, 128, 128, 128, 128, },
{ 2, 25, 32, 128, 96, 128, 128, 128, 160, 192, 128, },
}, {
{ 2, 69, 13, 173, 111, 69, 137, 159,
159, 146, 151, 193, 203, 131, 180, 123, },
{ 1, 12, 33, 164, 32, 98, 204, 242,
23, 99, 149, 215, 232, 110, 239, 245, },
{ 1, 17, 9, 136, 82, 83, 171, 28,
231, 128, 135, 76, 64, 118, 235, 17, },
{ 4, 41, 17, 195, 131, 58, 161, 237,
141, 97, 153, 189, 191, 117, 182, 202, },
{ 2, 17, 36, 104, 149, 137, 217, 139,
191, 119, 125, 107, 115, 223, 249, 110, },
{ 2, 14, 24, 127, 91, 135, 219, 198,
113, 91, 164, 125, 173, 211, 250, 116, },
{ 3, 19, 24, 120, 102, 130, 209, 81,
187, 95, 143, 102, 50, 190, 244, 56, },
{ 4, 27, 10, 128, 91, 157, 181, 33,
181, 150, 141, 141, 166, 114, 215, 25, },
{ 2, 34, 27, 187, 102, 77, 210, 245,
113, 107, 136, 184, 188, 121, 210, 234, },
{ 1, 15, 22, 141, 59, 94, 208, 133,
154, 95, 152, 112, 105, 191, 242, 111, },
{ 2, 176, 32, 128, 128, 128, 128, 128, 128, 128, 128, },
{ 1, 28, 32, 128, 176, 192, 208, 128, 128, 128, 128, },
{ 1, 28, 32, 128, 176, 192, 48, 128, 128, 128, 128, },
{ 4, 160, 32, 128, 128, 128, 128, 128, 128, 128, 128, },
{ 2, 28, 32, 128, 96, 128, 128, 128, 160, 192, 128, },
{ 2, 28, 32, 128, 160, 176, 192, 128, 128, 128, 128, },
{ 3, 29, 32, 128, 160, 176, 64, 128, 128, 128, 128, },
{ 4, 27, 32, 128, 160, 176, 64, 128, 128, 128, 128, },
{ 2, 34, 32, 128, 160, 176, 192, 128, 128, 128, 128, },
{ 1, 25, 32, 128, 96, 128, 128, 128, 160, 192, 128, },
#if EXT_TX_SIZES == 4
}, {
{ 2, 69, 13, 173, 111, 69, 137, 159,
159, 146, 151, 193, 203, 131, 180, 123, },
{ 1, 12, 33, 164, 32, 98, 204, 242,
23, 99, 149, 215, 232, 110, 239, 245, },
{ 1, 17, 9, 136, 82, 83, 171, 28,
231, 128, 135, 76, 64, 118, 235, 17, },
{ 4, 41, 17, 195, 131, 58, 161, 237,
141, 97, 153, 189, 191, 117, 182, 202, },
{ 2, 17, 36, 104, 149, 137, 217, 139,
191, 119, 125, 107, 115, 223, 249, 110, },
{ 2, 14, 24, 127, 91, 135, 219, 198,
113, 91, 164, 125, 173, 211, 250, 116, },
{ 3, 19, 24, 120, 102, 130, 209, 81,
187, 95, 143, 102, 50, 190, 244, 56, },
{ 4, 27, 10, 128, 91, 157, 181, 33,
181, 150, 141, 141, 166, 114, 215, 25, },
{ 2, 34, 27, 187, 102, 77, 210, 245,
113, 107, 136, 184, 188, 121, 210, 234, },
{ 1, 15, 22, 141, 59, 94, 208, 133,
154, 95, 152, 112, 105, 191, 242, 111, },
{ 2, 176, 32, 128, 128, 128, 128, 128, 128, 128, 128, },
{ 1, 12, 32, 128, 160, 176, 192, 128, 128, 128, 128, },
{ 1, 17, 32, 128, 160, 176, 64, 128, 128, 128, 128, },
{ 4, 41, 32, 128, 128, 128, 128, 128, 128, 128, 128, },
{ 2, 17, 32, 128, 96, 128, 128, 128, 160, 192, 128, },
{ 2, 14, 32, 128, 160, 176, 192, 128, 128, 128, 128, },
{ 3, 19, 32, 128, 160, 176, 64, 128, 128, 128, 128, },
{ 4, 27, 32, 128, 160, 176, 64, 128, 128, 128, 128, },
{ 2, 34, 32, 128, 160, 176, 192, 128, 128, 128, 128, },
{ 1, 15, 32, 128, 96, 128, 128, 128, 160, 192, 128, },
#endif
},
}, {
{
{ 8, 176, 128, 128, 128, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 10, 28, 176, 192, 208, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 10, 28, 176, 192, 48, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 9, 160, 128, 128, 128, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 8, 28, 96, 128, 128, 128, 160, 192,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 7, 28, 160, 176, 192, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 7, 20, 160, 176, 64, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 10, 23, 160, 176, 64, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 8, 29, 160, 176, 192, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 3, 20, 96, 128, 128, 128, 160, 192,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 8, 176, 128, 128, 128, 128, 128, 128, 128, },
{ 10, 28, 176, 192, 208, 128, 128, 128, 128, },
{ 10, 28, 176, 192, 48, 128, 128, 128, 128, },
{ 9, 160, 128, 128, 128, 128, 128, 128, 128, },
{ 8, 28, 96, 128, 128, 128, 160, 192, 128, },
{ 7, 28, 160, 176, 192, 128, 128, 128, 128, },
{ 7, 20, 160, 176, 64, 128, 128, 128, 128, },
{ 10, 23, 160, 176, 64, 128, 128, 128, 128, },
{ 8, 29, 160, 176, 192, 128, 128, 128, 128, },
{ 3, 20, 96, 128, 128, 128, 160, 192, 128, },
}, {
{ 2, 176, 128, 128, 128, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 4, 28, 176, 192, 208, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 4, 28, 176, 192, 48, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 8, 160, 128, 128, 128, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 2, 28, 96, 128, 128, 128, 160, 192,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 3, 28, 160, 176, 192, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 3, 26, 160, 176, 64, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 9, 24, 160, 176, 64, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 5, 24, 160, 176, 192, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 2, 25, 96, 128, 128, 128, 160, 192,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 2, 176, 128, 128, 128, 128, 128, 128, 128, },
{ 4, 28, 176, 192, 208, 128, 128, 128, 128, },
{ 4, 28, 176, 192, 48, 128, 128, 128, 128, },
{ 8, 160, 128, 128, 128, 128, 128, 128, 128, },
{ 2, 28, 96, 128, 128, 128, 160, 192, 128, },
{ 3, 28, 160, 176, 192, 128, 128, 128, 128, },
{ 3, 26, 160, 176, 64, 128, 128, 128, 128, },
{ 9, 24, 160, 176, 64, 128, 128, 128, 128, },
{ 5, 24, 160, 176, 192, 128, 128, 128, 128, },
{ 2, 25, 96, 128, 128, 128, 160, 192, 128, },
}, {
{ 2, 176, 128, 128, 128, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 1, 28, 176, 192, 208, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 1, 28, 176, 192, 48, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 4, 160, 128, 128, 128, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 2, 28, 96, 128, 128, 128, 160, 192,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 2, 28, 160, 176, 192, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 3, 29, 160, 176, 64, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 4, 27, 160, 176, 64, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 2, 34, 160, 176, 192, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 1, 25, 96, 128, 128, 128, 160, 192,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 2, 176, 128, 128, 128, 128, 128, 128, 128, },
{ 1, 28, 176, 192, 208, 128, 128, 128, 128, },
{ 1, 28, 176, 192, 48, 128, 128, 128, 128, },
{ 4, 160, 128, 128, 128, 128, 128, 128, 128, },
{ 2, 28, 96, 128, 128, 128, 160, 192, 128, },
{ 2, 28, 160, 176, 192, 128, 128, 128, 128, },
{ 3, 29, 160, 176, 64, 128, 128, 128, 128, },
{ 4, 27, 160, 176, 64, 128, 128, 128, 128, },
{ 2, 34, 160, 176, 192, 128, 128, 128, 128, },
{ 1, 25, 96, 128, 128, 128, 160, 192, 128, },
#if EXT_TX_SIZES == 4
}, {
{ 2, 176, 128, 128, 128, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 1, 12, 160, 176, 192, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 1, 17, 160, 176, 64, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 4, 41, 128, 128, 128, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 2, 17, 96, 128, 128, 128, 160, 192,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 2, 14, 160, 176, 192, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 3, 19, 160, 176, 64, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 4, 27, 160, 176, 64, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 2, 34, 160, 176, 192, 128, 128, 128,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 1, 15, 96, 128, 128, 128, 160, 192,
128, 128, 128, 128, 128, 128, 128, 128, },
{ 2, 176, 128, 128, 128, 128, 128, 128, 128, },
{ 1, 12, 160, 176, 192, 128, 128, 128, 128, },
{ 1, 17, 160, 176, 64, 128, 128, 128, 128, },
{ 4, 41, 128, 128, 128, 128, 128, 128, 128, },
{ 2, 17, 96, 128, 128, 128, 160, 192, 128, },
{ 2, 14, 160, 176, 192, 128, 128, 128, 128, },
{ 3, 19, 160, 176, 64, 128, 128, 128, 128, },
{ 4, 27, 160, 176, 64, 128, 128, 128, 128, },
{ 2, 34, 160, 176, 192, 128, 128, 128, 128, },
{ 1, 15, 96, 128, 128, 128, 160, 192, 128, },
#endif
},
},

18
vp10/common/enums.h
View File

@@ -111,21 +111,17 @@ typedef enum {
FLIPADST_FLIPADST = 6,
ADST_FLIPADST = 7,
FLIPADST_ADST = 8,
DST_DCT = 9,
DCT_DST = 10,
DST_ADST = 11,
ADST_DST = 12,
DST_FLIPADST = 13,
FLIPADST_DST = 14,
DST_DST = 15,
IDTX = 16,
V_DCT = 17,
H_DCT = 18,
IDTX = 9,
V_DCT = 10,
H_DCT = 11,
V_ADST = 12,
H_ADST = 13,
V_FLIPADST = 14,
H_FLIPADST = 15,
#endif // CONFIG_EXT_TX
TX_TYPES,
} TX_TYPE;
#if CONFIG_EXT_TX
#define EXT_TX_SIZES 4 // number of sizes that use extended transforms
#define EXT_TX_SETS_INTER 4 // Sets of transform selections for INTER

643
vp10/common/idct.c
View File

@@ -18,247 +18,6 @@
#include "vpx_dsp/inv_txfm.h"
#include "vpx_ports/mem.h"
#if CONFIG_EXT_TX
void idst4_c(const tran_low_t *input, tran_low_t *output) {
tran_low_t step[4];
tran_high_t temp1, temp2;
// stage 1
temp1 = (input[3] + input[1]) * cospi_16_64;
temp2 = (input[3] - input[1]) * cospi_16_64;
step[0] = WRAPLOW(dct_const_round_shift(temp1), 8);
step[1] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = input[2] * cospi_24_64 - input[0] * cospi_8_64;
temp2 = input[2] * cospi_8_64 + input[0] * cospi_24_64;
step[2] = WRAPLOW(dct_const_round_shift(temp1), 8);
step[3] = WRAPLOW(dct_const_round_shift(temp2), 8);
// stage 2
output[0] = WRAPLOW(step[0] + step[3], 8);
output[1] = WRAPLOW(-step[1] - step[2], 8);
output[2] = WRAPLOW(step[1] - step[2], 8);
output[3] = WRAPLOW(step[3] - step[0], 8);
}
void idst8_c(const tran_low_t *input, tran_low_t *output) {
// vp9_igentx8(input, output, Tx8);
tran_low_t step1[8], step2[8];
tran_high_t temp1, temp2;
// stage 1
step1[0] = input[7];
step1[2] = input[3];
step1[1] = input[5];
step1[3] = input[1];
temp1 = input[6] * cospi_28_64 - input[0] * cospi_4_64;
temp2 = input[6] * cospi_4_64 + input[0] * cospi_28_64;
step1[4] = WRAPLOW(dct_const_round_shift(temp1), 8);
step1[7] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = input[2] * cospi_12_64 - input[4] * cospi_20_64;
temp2 = input[2] * cospi_20_64 + input[4] * cospi_12_64;
step1[5] = WRAPLOW(dct_const_round_shift(temp1), 8);
step1[6] = WRAPLOW(dct_const_round_shift(temp2), 8);
// stage 2
temp1 = (step1[0] + step1[2]) * cospi_16_64;
temp2 = (step1[0] - step1[2]) * cospi_16_64;
step2[0] = WRAPLOW(dct_const_round_shift(temp1), 8);
step2[1] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = step1[1] * cospi_24_64 - step1[3] * cospi_8_64;
temp2 = step1[1] * cospi_8_64 + step1[3] * cospi_24_64;
step2[2] = WRAPLOW(dct_const_round_shift(temp1), 8);
step2[3] = WRAPLOW(dct_const_round_shift(temp2), 8);
step2[4] = WRAPLOW(step1[4] + step1[5], 8);
step2[5] = WRAPLOW(step1[4] - step1[5], 8);
step2[6] = WRAPLOW(-step1[6] + step1[7], 8);
step2[7] = WRAPLOW(step1[6] + step1[7], 8);
// stage 3
step1[0] = WRAPLOW(step2[0] + step2[3], 8);
step1[1] = WRAPLOW(step2[1] + step2[2], 8);
step1[2] = WRAPLOW(step2[1] - step2[2], 8);
step1[3] = WRAPLOW(step2[0] - step2[3], 8);
step1[4] = step2[4];
temp1 = (step2[6] - step2[5]) * cospi_16_64;
temp2 = (step2[5] + step2[6]) * cospi_16_64;
step1[5] = WRAPLOW(dct_const_round_shift(temp1), 8);
step1[6] = WRAPLOW(dct_const_round_shift(temp2), 8);
step1[7] = step2[7];
// stage 4
output[0] = WRAPLOW(step1[0] + step1[7], 8);
output[1] = WRAPLOW(-step1[1] - step1[6], 8);
output[2] = WRAPLOW(step1[2] + step1[5], 8);
output[3] = WRAPLOW(-step1[3] - step1[4], 8);
output[4] = WRAPLOW(step1[3] - step1[4], 8);
output[5] = WRAPLOW(-step1[2] + step1[5], 8);
output[6] = WRAPLOW(step1[1] - step1[6], 8);
output[7] = WRAPLOW(-step1[0] + step1[7], 8);
}
void idst16_c(const tran_low_t *input, tran_low_t *output) {
tran_low_t step1[16], step2[16];
tran_high_t temp1, temp2;
// stage 1
step1[0] = input[15];
step1[1] = input[7];
step1[2] = input[11];
step1[3] = input[3];
step1[4] = input[13];
step1[5] = input[5];
step1[6] = input[9];
step1[7] = input[1];
step1[8] = input[14];
step1[9] = input[6];
step1[10] = input[10];
step1[11] = input[2];
step1[12] = input[12];
step1[13] = input[4];
step1[14] = input[8];
step1[15] = input[0];
// stage 2
step2[0] = step1[0];
step2[1] = step1[1];
step2[2] = step1[2];
step2[3] = step1[3];
step2[4] = step1[4];
step2[5] = step1[5];
step2[6] = step1[6];
step2[7] = step1[7];
temp1 = step1[8] * cospi_30_64 - step1[15] * cospi_2_64;
temp2 = step1[8] * cospi_2_64 + step1[15] * cospi_30_64;
step2[8] = WRAPLOW(dct_const_round_shift(temp1), 8);
step2[15] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = step1[9] * cospi_14_64 - step1[14] * cospi_18_64;
temp2 = step1[9] * cospi_18_64 + step1[14] * cospi_14_64;
step2[9] = WRAPLOW(dct_const_round_shift(temp1), 8);
step2[14] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = step1[10] * cospi_22_64 - step1[13] * cospi_10_64;
temp2 = step1[10] * cospi_10_64 + step1[13] * cospi_22_64;
step2[10] = WRAPLOW(dct_const_round_shift(temp1), 8);
step2[13] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = step1[11] * cospi_6_64 - step1[12] * cospi_26_64;
temp2 = step1[11] * cospi_26_64 + step1[12] * cospi_6_64;
step2[11] = WRAPLOW(dct_const_round_shift(temp1), 8);
step2[12] = WRAPLOW(dct_const_round_shift(temp2), 8);
// stage 3
step1[0] = step2[0];
step1[1] = step2[1];
step1[2] = step2[2];
step1[3] = step2[3];
temp1 = step2[4] * cospi_28_64 - step2[7] * cospi_4_64;
temp2 = step2[4] * cospi_4_64 + step2[7] * cospi_28_64;
step1[4] = WRAPLOW(dct_const_round_shift(temp1), 8);
step1[7] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = step2[5] * cospi_12_64 - step2[6] * cospi_20_64;
temp2 = step2[5] * cospi_20_64 + step2[6] * cospi_12_64;
step1[5] = WRAPLOW(dct_const_round_shift(temp1), 8);
step1[6] = WRAPLOW(dct_const_round_shift(temp2), 8);
step1[8] = WRAPLOW(step2[8] + step2[9], 8);
step1[9] = WRAPLOW(step2[8] - step2[9], 8);
step1[10] = WRAPLOW(-step2[10] + step2[11], 8);
step1[11] = WRAPLOW(step2[10] + step2[11], 8);
step1[12] = WRAPLOW(step2[12] + step2[13], 8);
step1[13] = WRAPLOW(step2[12] - step2[13], 8);
step1[14] = WRAPLOW(-step2[14] + step2[15], 8);
step1[15] = WRAPLOW(step2[14] + step2[15], 8);
// stage 4
temp1 = (step1[0] + step1[1]) * cospi_16_64;
temp2 = (step1[0] - step1[1]) * cospi_16_64;
step2[0] = WRAPLOW(dct_const_round_shift(temp1), 8);
step2[1] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = step1[2] * cospi_24_64 - step1[3] * cospi_8_64;
temp2 = step1[2] * cospi_8_64 + step1[3] * cospi_24_64;
step2[2] = WRAPLOW(dct_const_round_shift(temp1), 8);
step2[3] = WRAPLOW(dct_const_round_shift(temp2), 8);
step2[4] = WRAPLOW(step1[4] + step1[5], 8);
step2[5] = WRAPLOW(step1[4] - step1[5], 8);
step2[6] = WRAPLOW(-step1[6] + step1[7], 8);
step2[7] = WRAPLOW(step1[6] + step1[7], 8);
step2[8] = step1[8];
step2[15] = step1[15];
temp1 = -step1[9] * cospi_8_64 + step1[14] * cospi_24_64;
temp2 = step1[9] * cospi_24_64 + step1[14] * cospi_8_64;
step2[9] = WRAPLOW(dct_const_round_shift(temp1), 8);
step2[14] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = -step1[10] * cospi_24_64 - step1[13] * cospi_8_64;
temp2 = -step1[10] * cospi_8_64 + step1[13] * cospi_24_64;
step2[10] = WRAPLOW(dct_const_round_shift(temp1), 8);
step2[13] = WRAPLOW(dct_const_round_shift(temp2), 8);
step2[11] = step1[11];
step2[12] = step1[12];
// stage 5
step1[0] = WRAPLOW(step2[0] + step2[3], 8);
step1[1] = WRAPLOW(step2[1] + step2[2], 8);
step1[2] = WRAPLOW(step2[1] - step2[2], 8);
step1[3] = WRAPLOW(step2[0] - step2[3], 8);
step1[4] = step2[4];
temp1 = (step2[6] - step2[5]) * cospi_16_64;
temp2 = (step2[5] + step2[6]) * cospi_16_64;
step1[5] = WRAPLOW(dct_const_round_shift(temp1), 8);
step1[6] = WRAPLOW(dct_const_round_shift(temp2), 8);
step1[7] = step2[7];
step1[8] = WRAPLOW(step2[8] + step2[11], 8);
step1[9] = WRAPLOW(step2[9] + step2[10], 8);
step1[10] = WRAPLOW(step2[9] - step2[10], 8);
step1[11] = WRAPLOW(step2[8] - step2[11], 8);
step1[12] = WRAPLOW(-step2[12] + step2[15], 8);
step1[13] = WRAPLOW(-step2[13] + step2[14], 8);
step1[14] = WRAPLOW(step2[13] + step2[14], 8);
step1[15] = WRAPLOW(step2[12] + step2[15], 8);
// stage 6
step2[0] = WRAPLOW(step1[0] + step1[7], 8);
step2[1] = WRAPLOW(step1[1] + step1[6], 8);
step2[2] = WRAPLOW(step1[2] + step1[5], 8);
step2[3] = WRAPLOW(step1[3] + step1[4], 8);
step2[4] = WRAPLOW(step1[3] - step1[4], 8);
step2[5] = WRAPLOW(step1[2] - step1[5], 8);
step2[6] = WRAPLOW(step1[1] - step1[6], 8);
step2[7] = WRAPLOW(step1[0] - step1[7], 8);
step2[8] = step1[8];
step2[9] = step1[9];
temp1 = (-step1[10] + step1[13]) * cospi_16_64;
temp2 = (step1[10] + step1[13]) * cospi_16_64;
step2[10] = WRAPLOW(dct_const_round_shift(temp1), 8);
step2[13] = WRAPLOW(dct_const_round_shift(temp2), 8);
temp1 = (-step1[11] + step1[12]) * cospi_16_64;
temp2 = (step1[11] + step1[12]) * cospi_16_64;
step2[11] = WRAPLOW(dct_const_round_shift(temp1), 8);
step2[12] = WRAPLOW(dct_const_round_shift(temp2), 8);
step2[14] = step1[14];
step2[15] = step1[15];
// stage 7
output[0] = WRAPLOW(step2[0] + step2[15], 8);
output[1] = WRAPLOW(-step2[1] - step2[14], 8);
output[2] = WRAPLOW(step2[2] + step2[13], 8);
output[3] = WRAPLOW(-step2[3] - step2[12], 8);
output[4] = WRAPLOW(step2[4] + step2[11], 8);
output[5] = WRAPLOW(-step2[5] - step2[10], 8);
output[6] = WRAPLOW(step2[6] + step2[9], 8);
output[7] = WRAPLOW(-step2[7] - step2[8], 8);
output[8] = WRAPLOW(step2[7] - step2[8], 8);
output[9] = WRAPLOW(-step2[6] + step2[9], 8);
output[10] = WRAPLOW(step2[5] - step2[10], 8);
output[11] = WRAPLOW(-step2[4] + step2[11], 8);
output[12] = WRAPLOW(step2[3] - step2[12], 8);
output[13] = WRAPLOW(-step2[2] + step2[13], 8);
output[14] = WRAPLOW(step2[1] - step2[14], 8);
output[15] = WRAPLOW(-step2[0] + step2[15], 8);
}
#if CONFIG_EXT_TX
static void iidtx4_c(const tran_low_t *input, tran_low_t *output) {
int i;
@@ -285,21 +44,6 @@ static void iidtx32_c(const tran_low_t *input, tran_low_t *output) {
}
// For use in lieu of DST
static void ihalfcenter32_c(const tran_low_t *input, tran_low_t *output) {
int i;
tran_low_t inputhalf[16];
for (i = 0; i < 8; ++i) {
output[i] = input[16 + i] * 4;
output[24 + i] = input[24 + i] * 4;
}
// Multiply input by sqrt(2)
for (i = 0; i < 16; ++i) {
inputhalf[i] = (tran_low_t)dct_const_round_shift(input[i] * Sqrt2);
}
idct16_c(inputhalf, output + 8);
// Note overall scaling factor is 4 times orthogonal
}
static void ihalfright32_c(const tran_low_t *input, tran_low_t *output) {
int i;
tran_low_t inputhalf[16];
@@ -379,7 +123,6 @@ static void highbd_ihalfright32_c(const tran_low_t *input, tran_low_t *output,
// Note overall scaling factor is 4 times orthogonal
}
#endif // CONFIG_VP9_HIGHBITDEPTH
#endif // CONFIG_EXT_TX
// Inverse identity transform and add.
static void inv_idtx_add_c(const tran_low_t *input, uint8_t *dest, int stride,
@@ -412,24 +155,21 @@ static void maybe_flip_strides(uint8_t **dst, int *dstride,
case ADST_DCT:
case DCT_ADST:
case ADST_ADST:
case DST_DST:
case DCT_DST:
case DST_DCT:
case DST_ADST:
case ADST_DST:
case IDTX:
case V_DCT:
case H_DCT:
case V_ADST:
case H_ADST:
break;
case FLIPADST_DCT:
case FLIPADST_ADST:
case FLIPADST_DST:
case V_FLIPADST:
// flip UD
FLIPUD_PTR(*dst, *dstride, size);
break;
case DCT_FLIPADST:
case ADST_FLIPADST:
case DST_FLIPADST:
case H_FLIPADST:
// flip LR
FLIPUD_PTR(*src, *sstride, size);
break;
@@ -716,24 +456,21 @@ static void maybe_flip_strides16(uint16_t **dst, int *dstride,
case ADST_DCT:
case DCT_ADST:
case ADST_ADST:
case DST_DST:
case DCT_DST:
case DST_DCT:
case DST_ADST:
case ADST_DST:
case IDTX:
case V_DCT:
case H_DCT:
case V_ADST:
case H_ADST:
break;
case FLIPADST_DCT:
case FLIPADST_ADST:
case FLIPADST_DST:
case V_FLIPADST:
// flip UD
FLIPUD_PTR(*dst, *dstride, size);
break;
case DCT_FLIPADST:
case ADST_FLIPADST:
case DST_FLIPADST:
case H_FLIPADST:
// flip LR
FLIPUD_PTR(*src, *sstride, size);
break;
@@ -754,26 +491,23 @@ static void maybe_flip_strides16(uint16_t **dst, int *dstride,
void vp10_iht4x4_16_add_c(const tran_low_t *input, uint8_t *dest, int stride,
int tx_type) {
static const transform_2d IHT_4[] = {
{ idct4_c, idct4_c }, // DCT_DCT = 0,
{ iadst4_c, idct4_c }, // ADST_DCT = 1,
{ idct4_c, iadst4_c }, // DCT_ADST = 2,
{ iadst4_c, iadst4_c }, // ADST_ADST = 3,
{ idct4_c, idct4_c }, // DCT_DCT
{ iadst4_c, idct4_c }, // ADST_DCT
{ idct4_c, iadst4_c }, // DCT_ADST
{ iadst4_c, iadst4_c }, // ADST_ADST
#if CONFIG_EXT_TX
{ iadst4_c, idct4_c }, // FLIPADST_DCT = 4,
{ idct4_c, iadst4_c }, // DCT_FLIPADST = 5,
{ iadst4_c, iadst4_c }, // FLIPADST_FLIPADST = 6,
{ iadst4_c, iadst4_c }, // ADST_FLIPADST = 7,
{ iadst4_c, iadst4_c }, // FLIPADST_ADST = 8,
{ idst4_c, idct4_c }, // DST_DCT = 9,
{ idct4_c, idst4_c }, // DCT_DST = 10,
{ idst4_c, iadst4_c }, // DST_ADST = 11,
{ iadst4_c, idst4_c }, // ADST_DST = 12,
{ idst4_c, iadst4_c }, // DST_FLIPADST = 13,
{ iadst4_c, idst4_c }, // FLIPADST_DST = 14,
{ idst4_c, idst4_c }, // DST_DST = 15
{ iidtx4_c, iidtx4_c }, // IDTX = 16
{ idct4_c, iidtx4_c }, // V_DCT = 17
{ iidtx4_c, idct4_c }, // H_DCT = 18
{ iadst4_c, idct4_c }, // FLIPADST_DCT
{ idct4_c, iadst4_c }, // DCT_FLIPADST
{ iadst4_c, iadst4_c }, // FLIPADST_FLIPADST
{ iadst4_c, iadst4_c }, // ADST_FLIPADST
{ iadst4_c, iadst4_c }, // FLIPADST_ADST
{ iidtx4_c, iidtx4_c }, // IDTX
{ idct4_c, iidtx4_c }, // V_DCT
{ iidtx4_c, idct4_c }, // H_DCT
{ iadst4_c, iidtx4_c }, // V_ADST
{ iidtx4_c, iadst4_c }, // H_ADST
{ iadst4_c, iidtx4_c }, // V_FLIPADST
{ iidtx4_c, iadst4_c }, // H_FLIPADST
#endif // CONFIG_EXT_TX
};
@@ -820,26 +554,23 @@ void vp10_iht4x4_16_add_c(const tran_low_t *input, uint8_t *dest, int stride,
void vp10_iht8x8_64_add_c(const tran_low_t *input, uint8_t *dest, int stride,
int tx_type) {
static const transform_2d IHT_8[] = {
{ idct8_c, idct8_c }, // DCT_DCT = 0,
{ iadst8_c, idct8_c }, // ADST_DCT = 1,
{ idct8_c, iadst8_c }, // DCT_ADST = 2,
{ iadst8_c, iadst8_c }, // ADST_ADST = 3,
{ idct8_c, idct8_c }, // DCT_DCT
{ iadst8_c, idct8_c }, // ADST_DCT
{ idct8_c, iadst8_c }, // DCT_ADST
{ iadst8_c, iadst8_c }, // ADST_ADST
#if CONFIG_EXT_TX
{ iadst8_c, idct8_c }, // FLIPADST_DCT = 4,
{ idct8_c, iadst8_c }, // DCT_FLIPADST = 5,
{ iadst8_c, iadst8_c }, // FLIPADST_FLIPADST = 6,
{ iadst8_c, iadst8_c }, // ADST_FLIPADST = 7,
{ iadst8_c, iadst8_c }, // FLIPADST_ADST = 8,
{ idst8_c, idct8_c }, // DST_DCT = 9,
{ idct8_c, idst8_c }, // DCT_DST = 10,
{ idst8_c, iadst8_c }, // DST_ADST = 11,
{ iadst8_c, idst8_c }, // ADST_DST = 12,
{ idst8_c, iadst8_c }, // DST_FLIPADST = 13,
{ iadst8_c, idst8_c }, // FLIPADST_DST = 14,
{ idst8_c, idst8_c }, // DST_DST = 15
{ iidtx8_c, iidtx8_c }, // IDTX = 16
{ idct8_c, iidtx8_c }, // V_DCT = 17
{ iidtx8_c, idct8_c }, // H_DCT = 18
{ iadst8_c, idct8_c }, // FLIPADST_DCT
{ idct8_c, iadst8_c }, // DCT_FLIPADST
{ iadst8_c, iadst8_c }, // FLIPADST_FLIPADST
{ iadst8_c, iadst8_c }, // ADST_FLIPADST
{ iadst8_c, iadst8_c }, // FLIPADST_ADST
{ iidtx8_c, iidtx8_c }, // IDTX
{ idct8_c, iidtx8_c }, // V_DCT
{ iidtx8_c, idct8_c }, // H_DCT
{ iadst8_c, iidtx8_c }, // V_ADST
{ iidtx8_c, iadst8_c }, // H_ADST
{ iadst8_c, iidtx8_c }, // V_FLIPADST
{ iidtx8_c, iadst8_c }, // H_FLIPADST
#endif // CONFIG_EXT_TX
};
@@ -886,26 +617,23 @@ void vp10_iht8x8_64_add_c(const tran_low_t *input, uint8_t *dest, int stride,
void vp10_iht16x16_256_add_c(const tran_low_t *input, uint8_t *dest, int stride,
int tx_type) {
static const transform_2d IHT_16[] = {
{ idct16_c, idct16_c }, // DCT_DCT = 0,
{ iadst16_c, idct16_c }, // ADST_DCT = 1,
{ idct16_c, iadst16_c }, // DCT_ADST = 2,
{ iadst16_c, iadst16_c }, // ADST_ADST = 3,
{ idct16_c, idct16_c }, // DCT_DCT
{ iadst16_c, idct16_c }, // ADST_DCT
{ idct16_c, iadst16_c }, // DCT_ADST
{ iadst16_c, iadst16_c }, // ADST_ADST
#if CONFIG_EXT_TX
{ iadst16_c, idct16_c }, // FLIPADST_DCT = 4,
{ idct16_c, iadst16_c }, // DCT_FLIPADST = 5,
{ iadst16_c, iadst16_c }, // FLIPADST_FLIPADST = 6,
{ iadst16_c, iadst16_c }, // ADST_FLIPADST = 7,
{ iadst16_c, iadst16_c }, // FLIPADST_ADST = 8,
{ idst16_c, idct16_c }, // DST_DCT = 9,
{ idct16_c, idst16_c }, // DCT_DST = 10,
{ idst16_c, iadst16_c }, // DST_ADST = 11,
{ iadst16_c, idst16_c }, // ADST_DST = 12,
{ idst16_c, iadst16_c }, // DST_FLIPADST = 13,
{ iadst16_c, idst16_c }, // FLIPADST_DST = 14,
{ idst16_c, idst16_c }, // DST_DST = 15
{ iidtx16_c, iidtx16_c }, // IDTX = 16
{ idct16_c, iidtx16_c }, // V_DCT = 17
{ iidtx16_c, idct16_c }, // H_DCT = 18
{ iadst16_c, idct16_c }, // FLIPADST_DCT
{ idct16_c, iadst16_c }, // DCT_FLIPADST
{ iadst16_c, iadst16_c }, // FLIPADST_FLIPADST
{ iadst16_c, iadst16_c }, // ADST_FLIPADST
{ iadst16_c, iadst16_c }, // FLIPADST_ADST
{ iidtx16_c, iidtx16_c }, // IDTX
{ idct16_c, iidtx16_c }, // V_DCT
{ iidtx16_c, idct16_c }, // H_DCT
{ iadst16_c, iidtx16_c }, // V_ADST
{ iidtx16_c, iadst16_c }, // H_ADST
{ iadst16_c, iidtx16_c }, // V_FLIPADST
{ iidtx16_c, iadst16_c }, // H_FLIPADST
#endif // CONFIG_EXT_TX
};
@@ -953,25 +681,22 @@ void vp10_iht16x16_256_add_c(const tran_low_t *input, uint8_t *dest, int stride,
void vp10_iht32x32_1024_add_c(const tran_low_t *input, uint8_t *dest,
int stride, int tx_type) {
static const transform_2d IHT_32[] = {
{ idct32_c, idct32_c }, // DCT_DCT = 0,
{ ihalfright32_c, idct32_c }, // ADST_DCT = 1,
{ idct32_c, ihalfright32_c }, // DCT_ADST = 2,
{ ihalfright32_c, ihalfright32_c }, // ADST_ADST = 3,
{ ihalfright32_c, idct32_c }, // FLIPADST_DCT = 4,
{ idct32_c, ihalfright32_c }, // DCT_FLIPADST = 5,
{ ihalfright32_c, ihalfright32_c }, // FLIPADST_FLIPADST = 6,
{ ihalfright32_c, ihalfright32_c }, // ADST_FLIPADST = 7,
{ ihalfright32_c, ihalfright32_c }, // FLIPADST_ADST = 8,
{ ihalfcenter32_c, idct32_c }, // DST_DCT = 9,
{ idct32_c, ihalfcenter32_c }, // DCT_DST = 10,
{ ihalfcenter32_c, ihalfright32_c }, // DST_ADST = 11,
{ ihalfright32_c, ihalfcenter32_c }, // ADST_DST = 12,
{ ihalfcenter32_c, ihalfright32_c }, // DST_FLIPADST = 13,
{ ihalfright32_c, ihalfcenter32_c }, // FLIPADST_DST = 14,
{ ihalfcenter32_c, ihalfcenter32_c }, // DST_DST = 15
{ iidtx32_c, iidtx32_c }, // IDTX = 16
{ idct32_c, iidtx32_c }, // V_DCT = 17
{ iidtx32_c, idct32_c }, // H_DCT = 18
{ idct32_c, idct32_c }, // DCT_DCT
{ ihalfright32_c, idct32_c }, // ADST_DCT
{ idct32_c, ihalfright32_c }, // DCT_ADST
{ ihalfright32_c, ihalfright32_c }, // ADST_ADST
{ ihalfright32_c, idct32_c }, // FLIPADST_DCT
{ idct32_c, ihalfright32_c }, // DCT_FLIPADST
{ ihalfright32_c, ihalfright32_c }, // FLIPADST_FLIPADST
{ ihalfright32_c, ihalfright32_c }, // ADST_FLIPADST
{ ihalfright32_c, ihalfright32_c }, // FLIPADST_ADST
{ iidtx32_c, iidtx32_c }, // IDTX
{ idct32_c, iidtx32_c }, // V_DCT
{ iidtx32_c, 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
};
int i, j;
@@ -1098,15 +823,12 @@ void vp10_inv_txfm_add_4x4(const tran_low_t *input, uint8_t *dest,
case FLIPADST_ADST:
vp10_iht4x4_16_add(input, dest, stride, tx_type);
break;
case DST_DST:
case DST_DCT:
case DCT_DST:
case DST_ADST:
case ADST_DST:
case FLIPADST_DST:
case DST_FLIPADST:
case H_DCT:
case V_DCT:
case H_DCT:
case V_ADST:
case H_ADST:
case V_FLIPADST:
case H_FLIPADST:
// Use C version since DST only exists in C code
vp10_iht4x4_16_add_c(input, dest, stride, tx_type);
break;
@@ -1139,15 +861,12 @@ void vp10_inv_txfm_add_8x8(const tran_low_t *input, uint8_t *dest,
case FLIPADST_ADST:
vp10_iht8x8_64_add(input, dest, stride, tx_type);
break;
case DST_DST:
case DST_DCT:
case DCT_DST:
case DST_ADST:
case ADST_DST:
case FLIPADST_DST:
case DST_FLIPADST:
case H_DCT:
case V_DCT:
case H_DCT:
case V_ADST:
case H_ADST:
case V_FLIPADST:
case H_FLIPADST:
// Use C version since DST only exists in C code
vp10_iht8x8_64_add_c(input, dest, stride, tx_type);
break;
@@ -1180,15 +899,12 @@ void vp10_inv_txfm_add_16x16(const tran_low_t *input, uint8_t *dest,
case FLIPADST_ADST:
vp10_iht16x16_256_add(input, dest, stride, tx_type);
break;
case DST_DST:
case DST_DCT:
case DCT_DST:
case DST_ADST:
case ADST_DST:
case FLIPADST_DST:
case DST_FLIPADST:
case H_DCT:
case V_DCT:
case H_DCT:
case V_ADST:
case H_ADST:
case V_FLIPADST:
case H_FLIPADST:
// Use C version since DST only exists in C code
vp10_iht16x16_256_add_c(input, dest, stride, tx_type);
break;
@@ -1217,15 +933,12 @@ void vp10_inv_txfm_add_32x32(const tran_low_t *input, uint8_t *dest,
case FLIPADST_FLIPADST:
case ADST_FLIPADST:
case FLIPADST_ADST:
case DST_DST:
case DST_DCT:
case DCT_DST:
case DST_ADST:
case ADST_DST:
case FLIPADST_DST:
case DST_FLIPADST:
case H_DCT:
case V_DCT:
case H_DCT:
case V_ADST:
case H_ADST:
case V_FLIPADST:
case H_FLIPADST:
vp10_iht32x32_1024_add_c(input, dest, stride, tx_type);
break;
case IDTX:
@@ -1242,26 +955,23 @@ void vp10_inv_txfm_add_32x32(const tran_low_t *input, uint8_t *dest,
void vp10_highbd_iht4x4_16_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int tx_type, int bd) {
static const highbd_transform_2d HIGH_IHT_4[] = {
{ vpx_highbd_idct4_c, vpx_highbd_idct4_c }, // DCT_DCT = 0,
{ vpx_highbd_iadst4_c, vpx_highbd_idct4_c }, // ADST_DCT = 1,
{ vpx_highbd_idct4_c, vpx_highbd_iadst4_c }, // DCT_ADST = 2,
{ vpx_highbd_iadst4_c, vpx_highbd_iadst4_c }, // ADST_ADST = 3,
{ vpx_highbd_idct4_c, vpx_highbd_idct4_c }, // DCT_DCT
{ vpx_highbd_iadst4_c, vpx_highbd_idct4_c }, // ADST_DCT
{ vpx_highbd_idct4_c, vpx_highbd_iadst4_c }, // DCT_ADST
{ vpx_highbd_iadst4_c, vpx_highbd_iadst4_c }, // ADST_ADST
#if CONFIG_EXT_TX
{ vpx_highbd_iadst4_c, vpx_highbd_idct4_c }, // FLIPADST_DCT = 4,
{ vpx_highbd_idct4_c, vpx_highbd_iadst4_c }, // DCT_FLIPADST = 5,
{ vpx_highbd_iadst4_c, vpx_highbd_iadst4_c }, // FLIPADST_FLIPADST = 6,
{ vpx_highbd_iadst4_c, vpx_highbd_iadst4_c }, // ADST_FLIPADST = 7,
{ vpx_highbd_iadst4_c, vpx_highbd_iadst4_c }, // FLIPADST_ADST = 8,
{ highbd_idst4_c, vpx_highbd_idct4_c }, // DST_DCT = 9,
{ vpx_highbd_idct4_c, highbd_idst4_c }, // DCT_DST = 10,
{ highbd_idst4_c, vpx_highbd_iadst4_c }, // DST_ADST = 11,
{ vpx_highbd_iadst4_c, highbd_idst4_c }, // ADST_DST = 12,
{ highbd_idst4_c, vpx_highbd_iadst4_c }, // DST_FLIPADST = 13,
{ vpx_highbd_iadst4_c, highbd_idst4_c }, // FLIPADST_DST = 14,
{ highbd_idst4_c, highbd_idst4_c }, // DST_DST = 15
{ highbd_iidtx4_c, highbd_iidtx4_c }, // IDTX = 16
{ vpx_highbd_idct4_c, highbd_iidtx4_c }, // V_DCT = 17
{ highbd_iidtx4_c, vpx_highbd_idct4_c }, // H_DCT = 18
{ vpx_highbd_iadst4_c, vpx_highbd_idct4_c }, // FLIPADST_DCT
{ vpx_highbd_idct4_c, vpx_highbd_iadst4_c }, // DCT_FLIPADST
{ vpx_highbd_iadst4_c, vpx_highbd_iadst4_c }, // FLIPADST_FLIPADST
{ vpx_highbd_iadst4_c, vpx_highbd_iadst4_c }, // ADST_FLIPADST
{ vpx_highbd_iadst4_c, vpx_highbd_iadst4_c }, // FLIPADST_ADST
{ highbd_iidtx4_c, highbd_iidtx4_c }, // IDTX
{ vpx_highbd_idct4_c, highbd_iidtx4_c }, // V_DCT
{ highbd_iidtx4_c, vpx_highbd_idct4_c }, // H_DCT
{ vpx_highbd_iadst4_c, highbd_iidtx4_c }, // V_ADST
{ highbd_iidtx4_c, vpx_highbd_iadst4_c }, // H_ADST
{ vpx_highbd_iadst4_c, highbd_iidtx4_c }, // V_FLIPADST
{ highbd_iidtx4_c, vpx_highbd_iadst4_c }, // H_FLIPADST
#endif // CONFIG_EXT_TX
};
@@ -1311,26 +1021,23 @@ void vp10_highbd_iht4x4_16_add_c(const tran_low_t *input, uint8_t *dest8,
void vp10_highbd_iht8x8_64_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int tx_type, int bd) {
static const highbd_transform_2d HIGH_IHT_8[] = {
{ vpx_highbd_idct8_c, vpx_highbd_idct8_c }, // DCT_DCT = 0,
{ vpx_highbd_iadst8_c, vpx_highbd_idct8_c }, // ADST_DCT = 1,
{ vpx_highbd_idct8_c, vpx_highbd_iadst8_c }, // DCT_ADST = 2,
{ vpx_highbd_iadst8_c, vpx_highbd_iadst8_c }, // ADST_ADST = 3,
{ vpx_highbd_idct8_c, vpx_highbd_idct8_c }, // DCT_DCT
{ vpx_highbd_iadst8_c, vpx_highbd_idct8_c }, // ADST_DCT
{ vpx_highbd_idct8_c, vpx_highbd_iadst8_c }, // DCT_ADST
{ vpx_highbd_iadst8_c, vpx_highbd_iadst8_c }, // ADST_ADST
#if CONFIG_EXT_TX
{ vpx_highbd_iadst8_c, vpx_highbd_idct8_c }, // FLIPADST_DCT = 4,
{ vpx_highbd_idct8_c, vpx_highbd_iadst8_c }, // DCT_FLIPADST = 5,
{ vpx_highbd_iadst8_c, vpx_highbd_iadst8_c }, // FLIPADST_FLIPADST = 6,
{ vpx_highbd_iadst8_c, vpx_highbd_iadst8_c }, // ADST_FLIPADST = 7,
{ vpx_highbd_iadst8_c, vpx_highbd_iadst8_c }, // FLIPADST_ADST = 8,
{ highbd_idst8_c, vpx_highbd_idct8_c }, // DST_DCT = 9,
{ vpx_highbd_idct8_c, highbd_idst8_c }, // DCT_DST = 10,
{ highbd_idst8_c, vpx_highbd_iadst8_c }, // DST_ADST = 11,
{ vpx_highbd_iadst8_c, highbd_idst8_c }, // ADST_DST = 12,
{ highbd_idst8_c, vpx_highbd_iadst8_c }, // DST_FLIPADST = 13,
{ vpx_highbd_iadst8_c, highbd_idst8_c }, // FLIPADST_DST = 14,
{ highbd_idst8_c, highbd_idst8_c }, // DST_DST = 15
{ highbd_iidtx8_c, highbd_iidtx8_c }, // IDTX = 16
{ vpx_highbd_idct8_c, highbd_iidtx8_c }, // V_DCT = 17
{ highbd_iidtx8_c, vpx_highbd_idct8_c }, // H_DCT = 18
{ vpx_highbd_iadst8_c, vpx_highbd_idct8_c }, // FLIPADST_DCT
{ vpx_highbd_idct8_c, vpx_highbd_iadst8_c }, // DCT_FLIPADST
{ vpx_highbd_iadst8_c, vpx_highbd_iadst8_c }, // FLIPADST_FLIPADST
{ vpx_highbd_iadst8_c, vpx_highbd_iadst8_c }, // ADST_FLIPADST
{ vpx_highbd_iadst8_c, vpx_highbd_iadst8_c }, // FLIPADST_ADST
{ highbd_iidtx8_c, highbd_iidtx8_c }, // IDTX
{ vpx_highbd_idct8_c, highbd_iidtx8_c }, // V_DCT
{ highbd_iidtx8_c, vpx_highbd_idct8_c }, // H_DCT
{ vpx_highbd_iadst8_c, highbd_iidtx8_c }, // V_ADST
{ highbd_iidtx8_c, vpx_highbd_iadst8_c }, // H_ADST
{ vpx_highbd_iadst8_c, highbd_iidtx8_c }, // V_FLIPADST
{ highbd_iidtx8_c, vpx_highbd_iadst8_c }, // H_FLIPADST
#endif // CONFIG_EXT_TX
};
@@ -1380,26 +1087,23 @@ void vp10_highbd_iht8x8_64_add_c(const tran_low_t *input, uint8_t *dest8,
void vp10_highbd_iht16x16_256_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int tx_type, int bd) {
static const highbd_transform_2d HIGH_IHT_16[] = {
{ vpx_highbd_idct16_c, vpx_highbd_idct16_c }, // DCT_DCT = 0,
{ vpx_highbd_iadst16_c, vpx_highbd_idct16_c }, // ADST_DCT = 1,
{ vpx_highbd_idct16_c, vpx_highbd_iadst16_c }, // DCT_ADST = 2,
{ vpx_highbd_iadst16_c, vpx_highbd_iadst16_c }, // ADST_ADST = 3,
{ vpx_highbd_idct16_c, vpx_highbd_idct16_c }, // DCT_DCT
{ vpx_highbd_iadst16_c, vpx_highbd_idct16_c }, // ADST_DCT
{ vpx_highbd_idct16_c, vpx_highbd_iadst16_c }, // DCT_ADST
{ vpx_highbd_iadst16_c, vpx_highbd_iadst16_c }, // ADST_ADST
#if CONFIG_EXT_TX
{ vpx_highbd_iadst16_c, vpx_highbd_idct16_c }, // FLIPADST_DCT = 4,
{ vpx_highbd_idct16_c, vpx_highbd_iadst16_c }, // DCT_FLIPADST = 5,
{ vpx_highbd_iadst16_c, vpx_highbd_iadst16_c }, // FLIPADST_FLIPADST = 6,
{ vpx_highbd_iadst16_c, vpx_highbd_iadst16_c }, // ADST_FLIPADST = 7,
{ vpx_highbd_iadst16_c, vpx_highbd_iadst16_c }, // FLIPADST_ADST = 8,
{ highbd_idst16_c, vpx_highbd_idct16_c }, // DST_DCT = 9,
{ vpx_highbd_idct16_c, highbd_idst16_c }, // DCT_DST = 10,
{ highbd_idst16_c, vpx_highbd_iadst16_c }, // DST_ADST = 11,
{ vpx_highbd_iadst16_c, highbd_idst16_c }, // ADST_DST = 12,
{ highbd_idst16_c, vpx_highbd_iadst16_c }, // DST_FLIPADST = 13,
{ vpx_highbd_iadst16_c, highbd_idst16_c }, // FLIPADST_DST = 14,
{ highbd_idst16_c, highbd_idst16_c }, // DST_DST = 15
{ highbd_iidtx16_c, highbd_iidtx16_c }, // IDTX = 16
{ vpx_highbd_idct16_c, highbd_iidtx16_c }, // V_DCT = 17
{ highbd_iidtx16_c, vpx_highbd_idct16_c }, // H_DCT = 18
{ vpx_highbd_iadst16_c, vpx_highbd_idct16_c }, // FLIPADST_DCT
{ vpx_highbd_idct16_c, vpx_highbd_iadst16_c }, // DCT_FLIPADST
{ vpx_highbd_iadst16_c, vpx_highbd_iadst16_c }, // FLIPADST_FLIPADST
{ vpx_highbd_iadst16_c, vpx_highbd_iadst16_c }, // ADST_FLIPADST
{ vpx_highbd_iadst16_c, vpx_highbd_iadst16_c }, // FLIPADST_ADST
{ highbd_iidtx16_c, highbd_iidtx16_c }, // IDTX
{ vpx_highbd_idct16_c, highbd_iidtx16_c }, // V_DCT
{ highbd_iidtx16_c, vpx_highbd_idct16_c }, // H_DCT
{ vpx_highbd_iadst16_c, highbd_iidtx16_c }, // V_ADST
{ highbd_iidtx16_c, vpx_highbd_iadst16_c }, // H_ADST
{ vpx_highbd_iadst16_c, highbd_iidtx16_c }, // V_FLIPADST
{ highbd_iidtx16_c, vpx_highbd_iadst16_c }, // H_FLIPADST
#endif // CONFIG_EXT_TX
};
@@ -1459,16 +1163,13 @@ void vp10_highbd_iht32x32_1024_add_c(const tran_low_t *input, uint8_t *dest8,
{ highbd_ihalfright32_c, highbd_ihalfright32_c }, // FLIPADST_FLIPADST
{ highbd_ihalfright32_c, highbd_ihalfright32_c }, // ADST_FLIPADST
{ highbd_ihalfright32_c, highbd_ihalfright32_c }, // FLIPADST_ADST
{ highbd_ihalfcenter32_c, vpx_highbd_idct32_c }, // DST_DCT
{ vpx_highbd_idct32_c, highbd_ihalfcenter32_c }, // DCT_DST
{ highbd_ihalfcenter32_c, highbd_ihalfright32_c }, // DST_ADST
{ highbd_ihalfright32_c, highbd_ihalfcenter32_c }, // ADST_DST
{ highbd_ihalfcenter32_c, highbd_ihalfright32_c }, // DST_FLIPADST
{ highbd_ihalfright32_c, highbd_ihalfcenter32_c }, // FLIPADST_DST
{ highbd_ihalfcenter32_c, highbd_ihalfcenter32_c }, // DST_DST
{ highbd_iidtx32_c, highbd_iidtx32_c }, // IDTX
{ highbd_iidtx32_c, highbd_iidtx32_c }, // IDTX
{ vpx_highbd_idct32_c, highbd_iidtx32_c }, // V_DCT
{ highbd_iidtx32_c, vpx_highbd_idct32_c }, // H_DCT
{ highbd_iidtx32_c, vpx_highbd_idct32_c }, // H_DCT
{ highbd_ihalfright32_c, highbd_iidtx32_c }, // V_ADST
{ highbd_iidtx32_c, highbd_ihalfright32_c }, // H_ADST
{ highbd_ihalfright32_c, highbd_iidtx32_c }, // V_FLIPADST
{ highbd_iidtx32_c, highbd_ihalfright32_c }, // H_FLIPADST
};
uint16_t *dest = CONVERT_TO_SHORTPTR(dest8);
@@ -1602,15 +1303,12 @@ void vp10_highbd_inv_txfm_add_4x4(const tran_low_t *input, uint8_t *dest,
case FLIPADST_ADST:
vp10_highbd_iht4x4_16_add(input, dest, stride, tx_type, bd);
break;
case DST_DST:
case DST_DCT:
case DCT_DST:
case DST_ADST:
case ADST_DST:
case FLIPADST_DST:
case DST_FLIPADST:
case H_DCT:
case V_DCT:
case H_DCT:
case V_ADST:
case H_ADST:
case V_FLIPADST:
case H_FLIPADST:
// Use C version since DST only exists in C code
vp10_highbd_iht4x4_16_add_c(input, dest, stride, tx_type, bd);
break;
@@ -1644,15 +1342,12 @@ void vp10_highbd_inv_txfm_add_8x8(const tran_low_t *input, uint8_t *dest,
case FLIPADST_ADST:
vp10_highbd_iht8x8_64_add(input, dest, stride, tx_type, bd);
break;
case DST_DST:
case DST_DCT:
case DCT_DST:
case DST_ADST:
case ADST_DST:
case FLIPADST_DST:
case DST_FLIPADST:
case H_DCT:
case V_DCT:
case H_DCT:
case V_ADST:
case H_ADST:
case V_FLIPADST:
case H_FLIPADST:
// Use C version since DST only exists in C code
vp10_highbd_iht8x8_64_add_c(input, dest, stride, tx_type, bd);
break;
@@ -1686,15 +1381,12 @@ void vp10_highbd_inv_txfm_add_16x16(const tran_low_t *input, uint8_t *dest,
case FLIPADST_ADST:
vp10_highbd_iht16x16_256_add(input, dest, stride, tx_type, bd);
break;
case DST_DST:
case DST_DCT:
case DCT_DST:
case DST_ADST:
case ADST_DST:
case FLIPADST_DST:
case DST_FLIPADST:
case H_DCT:
case V_DCT:
case H_DCT:
case V_ADST:
case H_ADST:
case V_FLIPADST:
case H_FLIPADST:
// Use C version since DST only exists in C code
vp10_highbd_iht16x16_256_add_c(input, dest, stride, tx_type, bd);
break;
@@ -1724,15 +1416,12 @@ void vp10_highbd_inv_txfm_add_32x32(const tran_low_t *input, uint8_t *dest,
case FLIPADST_FLIPADST:
case ADST_FLIPADST:
case FLIPADST_ADST:
case DST_DST:
case DST_DCT:
case DCT_DST:
case DST_ADST:
case ADST_DST:
case FLIPADST_DST:
case DST_FLIPADST:
case H_DCT:
case V_DCT:
case H_DCT:
case V_ADST:
case H_ADST:
case V_FLIPADST:
case H_FLIPADST:
vp10_highbd_iht32x32_1024_add_c(input, dest, stride, tx_type, bd);
break;
case IDTX:

130
vp10/common/scan.c
View File

@@ -2882,13 +2882,10 @@ const scan_order vp10_intra_scan_orders[TX_SIZES][TX_TYPES] = {
{default_scan_4x4, vp10_default_iscan_4x4, default_scan_4x4_neighbors},
{default_scan_4x4, vp10_default_iscan_4x4, default_scan_4x4_neighbors},
{default_scan_4x4, vp10_default_iscan_4x4, default_scan_4x4_neighbors},
{default_scan_4x4, vp10_default_iscan_4x4, default_scan_4x4_neighbors},
{default_scan_4x4, vp10_default_iscan_4x4, default_scan_4x4_neighbors},
{default_scan_4x4, vp10_default_iscan_4x4, default_scan_4x4_neighbors},
{default_scan_4x4, vp10_default_iscan_4x4, default_scan_4x4_neighbors},
{default_scan_4x4, vp10_default_iscan_4x4, default_scan_4x4_neighbors},
{default_scan_4x4, vp10_default_iscan_4x4, default_scan_4x4_neighbors},
{default_scan_4x4, vp10_default_iscan_4x4, default_scan_4x4_neighbors},
{row_scan_4x4, vp10_row_iscan_4x4, row_scan_4x4_neighbors},
{col_scan_4x4, vp10_col_iscan_4x4, col_scan_4x4_neighbors},
{row_scan_4x4, vp10_row_iscan_4x4, row_scan_4x4_neighbors},
{col_scan_4x4, vp10_col_iscan_4x4, col_scan_4x4_neighbors},
{row_scan_4x4, vp10_row_iscan_4x4, row_scan_4x4_neighbors},
{col_scan_4x4, vp10_col_iscan_4x4, col_scan_4x4_neighbors},
}, { // TX_8X8
@@ -2902,13 +2899,10 @@ const scan_order vp10_intra_scan_orders[TX_SIZES][TX_TYPES] = {
{default_scan_8x8, vp10_default_iscan_8x8, default_scan_8x8_neighbors},
{default_scan_8x8, vp10_default_iscan_8x8, default_scan_8x8_neighbors},
{default_scan_8x8, vp10_default_iscan_8x8, default_scan_8x8_neighbors},
{default_scan_8x8, vp10_default_iscan_8x8, default_scan_8x8_neighbors},
{default_scan_8x8, vp10_default_iscan_8x8, default_scan_8x8_neighbors},
{default_scan_8x8, vp10_default_iscan_8x8, default_scan_8x8_neighbors},
{default_scan_8x8, vp10_default_iscan_8x8, default_scan_8x8_neighbors},
{default_scan_8x8, vp10_default_iscan_8x8, default_scan_8x8_neighbors},
{default_scan_8x8, vp10_default_iscan_8x8, default_scan_8x8_neighbors},
{default_scan_8x8, vp10_default_iscan_8x8, default_scan_8x8_neighbors},
{row_scan_8x8, vp10_row_iscan_8x8, row_scan_8x8_neighbors},
{col_scan_8x8, vp10_col_iscan_8x8, col_scan_8x8_neighbors},
{row_scan_8x8, vp10_row_iscan_8x8, row_scan_8x8_neighbors},
{col_scan_8x8, vp10_col_iscan_8x8, col_scan_8x8_neighbors},
{row_scan_8x8, vp10_row_iscan_8x8, row_scan_8x8_neighbors},
{col_scan_8x8, vp10_col_iscan_8x8, col_scan_8x8_neighbors},
}, { // TX_16X16
@@ -2930,22 +2924,12 @@ const scan_order vp10_intra_scan_orders[TX_SIZES][TX_TYPES] = {
default_scan_16x16_neighbors},
{default_scan_16x16, vp10_default_iscan_16x16,
default_scan_16x16_neighbors},
{default_scan_16x16, vp10_default_iscan_16x16,
default_scan_16x16_neighbors},
{default_scan_16x16, vp10_default_iscan_16x16,
default_scan_16x16_neighbors},
{default_scan_16x16, vp10_default_iscan_16x16,
default_scan_16x16_neighbors},
{default_scan_16x16, vp10_default_iscan_16x16,
default_scan_16x16_neighbors},
{default_scan_16x16, vp10_default_iscan_16x16,
default_scan_16x16_neighbors},
{default_scan_16x16, vp10_default_iscan_16x16,
default_scan_16x16_neighbors},
{default_scan_16x16, vp10_default_iscan_16x16,
default_scan_16x16_neighbors},
{row_scan_16x16, vp10_row_iscan_16x16, row_scan_16x16_neighbors},
{col_scan_16x16, vp10_col_iscan_16x16, col_scan_16x16_neighbors},
{row_scan_16x16, vp10_row_iscan_16x16, row_scan_16x16_neighbors},
{col_scan_16x16, vp10_col_iscan_16x16, col_scan_16x16_neighbors},
{row_scan_16x16, vp10_row_iscan_16x16, row_scan_16x16_neighbors},
{col_scan_16x16, vp10_col_iscan_16x16, col_scan_16x16_neighbors},
{row_scan_16x16, vp10_row_iscan_16x16, row_scan_16x16_neighbors},
{col_scan_16x16, vp10_col_iscan_16x16, col_scan_16x16_neighbors},
}, { // TX_32X32
{default_scan_32x32, vp10_default_iscan_32x32,
default_scan_32x32_neighbors},
@@ -2965,26 +2949,14 @@ const scan_order vp10_intra_scan_orders[TX_SIZES][TX_TYPES] = {
qtr_scan_32x32_neighbors},
{qtr_scan_32x32, vp10_qtr_iscan_32x32,
qtr_scan_32x32_neighbors},
{h2_scan_32x32, vp10_h2_iscan_32x32,
h2_scan_32x32_neighbors},
{v2_scan_32x32, vp10_v2_iscan_32x32,
v2_scan_32x32_neighbors},
{qtr_scan_32x32, vp10_qtr_iscan_32x32,
qtr_scan_32x32_neighbors},
{qtr_scan_32x32, vp10_qtr_iscan_32x32,
qtr_scan_32x32_neighbors},
{qtr_scan_32x32, vp10_qtr_iscan_32x32,
qtr_scan_32x32_neighbors},
{qtr_scan_32x32, vp10_qtr_iscan_32x32,
qtr_scan_32x32_neighbors},
{qtr_scan_32x32, vp10_qtr_iscan_32x32,
qtr_scan_32x32_neighbors},
{default_scan_32x32, vp10_default_iscan_32x32,
default_scan_32x32_neighbors},
{h2_scan_32x32, vp10_h2_iscan_32x32,
h2_scan_32x32_neighbors},
{v2_scan_32x32, vp10_v2_iscan_32x32,
v2_scan_32x32_neighbors},
{mrow_scan_32x32, vp10_mrow_iscan_32x32, mrow_scan_32x32_neighbors},
{mcol_scan_32x32, vp10_mcol_iscan_32x32, mcol_scan_32x32_neighbors},
{mrow_scan_32x32, vp10_mrow_iscan_32x32, mrow_scan_32x32_neighbors},
{mcol_scan_32x32, vp10_mcol_iscan_32x32, mcol_scan_32x32_neighbors},
{mrow_scan_32x32, vp10_mrow_iscan_32x32, mrow_scan_32x32_neighbors},
{mcol_scan_32x32, vp10_mcol_iscan_32x32, mcol_scan_32x32_neighbors},
}
};
@@ -3000,13 +2972,10 @@ const scan_order vp10_inter_scan_orders[TX_SIZES][TX_TYPES] = {
{default_scan_4x4, vp10_default_iscan_4x4, default_scan_4x4_neighbors},
{default_scan_4x4, vp10_default_iscan_4x4, default_scan_4x4_neighbors},
{default_scan_4x4, vp10_default_iscan_4x4, default_scan_4x4_neighbors},
{default_scan_4x4, vp10_default_iscan_4x4, default_scan_4x4_neighbors},
{default_scan_4x4, vp10_default_iscan_4x4, default_scan_4x4_neighbors},
{default_scan_4x4, vp10_default_iscan_4x4, default_scan_4x4_neighbors},
{default_scan_4x4, vp10_default_iscan_4x4, default_scan_4x4_neighbors},
{default_scan_4x4, vp10_default_iscan_4x4, default_scan_4x4_neighbors},
{default_scan_4x4, vp10_default_iscan_4x4, default_scan_4x4_neighbors},
{default_scan_4x4, vp10_default_iscan_4x4, default_scan_4x4_neighbors},
{mrow_scan_4x4, vp10_mrow_iscan_4x4, mrow_scan_4x4_neighbors},
{mcol_scan_4x4, vp10_mcol_iscan_4x4, mcol_scan_4x4_neighbors},
{mrow_scan_4x4, vp10_mrow_iscan_4x4, mrow_scan_4x4_neighbors},
{mcol_scan_4x4, vp10_mcol_iscan_4x4, mcol_scan_4x4_neighbors},
{mrow_scan_4x4, vp10_mrow_iscan_4x4, mrow_scan_4x4_neighbors},
{mcol_scan_4x4, vp10_mcol_iscan_4x4, mcol_scan_4x4_neighbors},
}, { // TX_8X8
@@ -3020,13 +2989,10 @@ const scan_order vp10_inter_scan_orders[TX_SIZES][TX_TYPES] = {
{default_scan_8x8, vp10_default_iscan_8x8, default_scan_8x8_neighbors},
{default_scan_8x8, vp10_default_iscan_8x8, default_scan_8x8_neighbors},
{default_scan_8x8, vp10_default_iscan_8x8, default_scan_8x8_neighbors},
{default_scan_8x8, vp10_default_iscan_8x8, default_scan_8x8_neighbors},
{default_scan_8x8, vp10_default_iscan_8x8, default_scan_8x8_neighbors},
{default_scan_8x8, vp10_default_iscan_8x8, default_scan_8x8_neighbors},
{default_scan_8x8, vp10_default_iscan_8x8, default_scan_8x8_neighbors},
{default_scan_8x8, vp10_default_iscan_8x8, default_scan_8x8_neighbors},
{default_scan_8x8, vp10_default_iscan_8x8, default_scan_8x8_neighbors},
{default_scan_8x8, vp10_default_iscan_8x8, default_scan_8x8_neighbors},
{mrow_scan_8x8, vp10_mrow_iscan_8x8, mrow_scan_8x8_neighbors},
{mcol_scan_8x8, vp10_mcol_iscan_8x8, mcol_scan_8x8_neighbors},
{mrow_scan_8x8, vp10_mrow_iscan_8x8, mrow_scan_8x8_neighbors},
{mcol_scan_8x8, vp10_mcol_iscan_8x8, mcol_scan_8x8_neighbors},
{mrow_scan_8x8, vp10_mrow_iscan_8x8, mrow_scan_8x8_neighbors},
{mcol_scan_8x8, vp10_mcol_iscan_8x8, mcol_scan_8x8_neighbors},
}, { // TX_16X16
@@ -3050,22 +3016,12 @@ const scan_order vp10_inter_scan_orders[TX_SIZES][TX_TYPES] = {
default_scan_16x16_neighbors},
{default_scan_16x16, vp10_default_iscan_16x16,
default_scan_16x16_neighbors},
{default_scan_16x16, vp10_default_iscan_16x16,
default_scan_16x16_neighbors},
{default_scan_16x16, vp10_default_iscan_16x16,
default_scan_16x16_neighbors},
{default_scan_16x16, vp10_default_iscan_16x16,
default_scan_16x16_neighbors},
{default_scan_16x16, vp10_default_iscan_16x16,
default_scan_16x16_neighbors},
{default_scan_16x16, vp10_default_iscan_16x16,
default_scan_16x16_neighbors},
{default_scan_16x16, vp10_default_iscan_16x16,
default_scan_16x16_neighbors},
{default_scan_16x16, vp10_default_iscan_16x16,
default_scan_16x16_neighbors},
{mrow_scan_16x16, vp10_mrow_iscan_16x16, mrow_scan_16x16_neighbors},
{mcol_scan_16x16, vp10_mcol_iscan_16x16, mcol_scan_16x16_neighbors},
{mrow_scan_16x16, vp10_mrow_iscan_16x16, mrow_scan_16x16_neighbors},
{mcol_scan_16x16, vp10_mcol_iscan_16x16, mcol_scan_16x16_neighbors},
{mrow_scan_16x16, vp10_mrow_iscan_16x16, mrow_scan_16x16_neighbors},
{mcol_scan_16x16, vp10_mcol_iscan_16x16, mcol_scan_16x16_neighbors},
{mrow_scan_16x16, vp10_mrow_iscan_16x16, mrow_scan_16x16_neighbors},
{mcol_scan_16x16, vp10_mcol_iscan_16x16, mcol_scan_16x16_neighbors},
}, { // TX_32X32
{default_scan_32x32, vp10_default_iscan_32x32,
default_scan_32x32_neighbors},
@@ -3085,24 +3041,14 @@ const scan_order vp10_inter_scan_orders[TX_SIZES][TX_TYPES] = {
qtr_scan_32x32_neighbors},
{qtr_scan_32x32, vp10_qtr_iscan_32x32,
qtr_scan_32x32_neighbors},
{h2_scan_32x32, vp10_h2_iscan_32x32,
h2_scan_32x32_neighbors},
{v2_scan_32x32, vp10_v2_iscan_32x32,
v2_scan_32x32_neighbors},
{qtr_scan_32x32, vp10_qtr_iscan_32x32,
qtr_scan_32x32_neighbors},
{qtr_scan_32x32, vp10_qtr_iscan_32x32,
qtr_scan_32x32_neighbors},
{qtr_scan_32x32, vp10_qtr_iscan_32x32,
qtr_scan_32x32_neighbors},
{qtr_scan_32x32, vp10_qtr_iscan_32x32,
qtr_scan_32x32_neighbors},
{qtr_scan_32x32, vp10_qtr_iscan_32x32,
qtr_scan_32x32_neighbors},
{default_scan_32x32, vp10_default_iscan_32x32,
default_scan_32x32_neighbors},
{mrow_scan_32x32, vp10_mrow_iscan_32x32, mrow_scan_32x32_neighbors},
{mcol_scan_32x32, vp10_mcol_iscan_32x32, mcol_scan_32x32_neighbors},
{mrow_scan_32x32, vp10_mrow_iscan_32x32, mrow_scan_32x32_neighbors},
{mcol_scan_32x32, vp10_mcol_iscan_32x32, mcol_scan_32x32_neighbors},
{mrow_scan_32x32, vp10_mrow_iscan_32x32, mrow_scan_32x32_neighbors},
{mcol_scan_32x32, vp10_mcol_iscan_32x32, mcol_scan_32x32_neighbors},
}
};

382
vp10/encoder/dct.c
View File

@@ -36,219 +36,6 @@ static INLINE void range_check(const tran_low_t *input, const int size,
#endif
}
#if CONFIG_EXT_TX
void fdst4(const tran_low_t *input, tran_low_t *output) {
tran_high_t step[4];
tran_high_t temp1, temp2;
step[0] = input[0] - input[3];
step[1] = -input[1] + input[2];
step[2] = -input[1] - input[2];
step[3] = input[0] + input[3];
temp1 = (step[0] + step[1]) * cospi_16_64;
temp2 = (step[0] - step[1]) * cospi_16_64;
output[3] = fdct_round_shift(temp1);
output[1] = fdct_round_shift(temp2);
temp1 = step[2] * cospi_24_64 + step[3] * cospi_8_64;
temp2 = -step[2] * cospi_8_64 + step[3] * cospi_24_64;
output[2] = fdct_round_shift(temp1);
output[0] = fdct_round_shift(temp2);
}
void fdst8(const tran_low_t *input, tran_low_t *output) {
tran_high_t s0, s1, s2, s3, s4, s5, s6, s7; // canbe16
tran_high_t t0, t1, t2, t3; // needs32
tran_high_t x0, x1, x2, x3; // canbe16
// stage 1
s0 = input[0] - input[7];
s1 = -input[1] + input[6];
s2 = input[2] - input[5];
s3 = -input[3] + input[4];
s4 = -input[3] - input[4];
s5 = input[2] + input[5];
s6 = -input[1] - input[6];
s7 = input[0] + input[7];
x0 = s0 + s3;
x1 = s1 + s2;
x2 = s1 - s2;
x3 = s0 - s3;
t0 = (x0 + x1) * cospi_16_64;
t1 = (x0 - x1) * cospi_16_64;
t2 = x2 * cospi_24_64 + x3 * cospi_8_64;
t3 = -x2 * cospi_8_64 + x3 * cospi_24_64;
output[7] = fdct_round_shift(t0);
output[5] = fdct_round_shift(t2);
output[3] = fdct_round_shift(t1);
output[1] = fdct_round_shift(t3);
// Stage 2
t0 = (s6 - s5) * cospi_16_64;
t1 = (s6 + s5) * cospi_16_64;
t2 = fdct_round_shift(t0);
t3 = fdct_round_shift(t1);
// Stage 3
x0 = s4 + t2;
x1 = s4 - t2;
x2 = s7 - t3;
x3 = s7 + t3;
// Stage 4
t0 = x0 * cospi_28_64 + x3 * cospi_4_64;
t1 = x1 * cospi_12_64 + x2 * cospi_20_64;
t2 = x2 * cospi_12_64 + x1 * -cospi_20_64;
t3 = x3 * cospi_28_64 + x0 * -cospi_4_64;
output[6] = fdct_round_shift(t0);
output[4] = fdct_round_shift(t2);
output[2] = fdct_round_shift(t1);
output[0] = fdct_round_shift(t3);
}
void fdst16(const tran_low_t *input, tran_low_t *output) {
tran_high_t step1[8]; // canbe16
tran_high_t step2[8]; // canbe16
tran_high_t step3[8]; // canbe16
tran_high_t in[8]; // canbe16
tran_high_t temp1, temp2; // needs32
// step 1
in[0] = input[0] - input[15];
in[1] = -input[1] + input[14];
in[2] = input[2] - input[13];
in[3] = -input[3] + input[12];
in[4] = input[4] - input[11];
in[5] = -input[5] + input[10];
in[6] = input[6] - input[ 9];
in[7] = -input[7] + input[ 8];
step1[0] = -input[7] - input[ 8];
step1[1] = input[6] + input[ 9];
step1[2] = -input[5] - input[10];
step1[3] = input[4] + input[11];
step1[4] = -input[3] - input[12];
step1[5] = input[2] + input[13];
step1[6] = -input[1] - input[14];
step1[7] = input[0] + input[15];
// fdct8(step, step);
{
tran_high_t s0, s1, s2, s3, s4, s5, s6, s7; // canbe16
tran_high_t t0, t1, t2, t3; // needs32
tran_high_t x0, x1, x2, x3; // canbe16
// stage 1
s0 = in[0] + in[7];
s1 = in[1] + in[6];
s2 = in[2] + in[5];
s3 = in[3] + in[4];
s4 = in[3] - in[4];
s5 = in[2] - in[5];
s6 = in[1] - in[6];
s7 = in[0] - in[7];
// fdct4(step, step);
x0 = s0 + s3;
x1 = s1 + s2;
x2 = s1 - s2;
x3 = s0 - s3;
t0 = (x0 + x1) * cospi_16_64;
t1 = (x0 - x1) * cospi_16_64;
t2 = x3 * cospi_8_64 + x2 * cospi_24_64;
t3 = x3 * cospi_24_64 - x2 * cospi_8_64;
output[15] = fdct_round_shift(t0);
output[11] = fdct_round_shift(t2);
output[7] = fdct_round_shift(t1);
output[3] = fdct_round_shift(t3);
// Stage 2
t0 = (s6 - s5) * cospi_16_64;
t1 = (s6 + s5) * cospi_16_64;
t2 = fdct_round_shift(t0);
t3 = fdct_round_shift(t1);
// Stage 3
x0 = s4 + t2;
x1 = s4 - t2;
x2 = s7 - t3;
x3 = s7 + t3;
// Stage 4
t0 = x0 * cospi_28_64 + x3 * cospi_4_64;
t1 = x1 * cospi_12_64 + x2 * cospi_20_64;
t2 = x2 * cospi_12_64 + x1 * -cospi_20_64;
t3 = x3 * cospi_28_64 + x0 * -cospi_4_64;
output[13] = fdct_round_shift(t0);
output[9] = fdct_round_shift(t2);
output[5] = fdct_round_shift(t1);
output[1] = fdct_round_shift(t3);
}
// step 2
temp1 = (step1[5] - step1[2]) * cospi_16_64;
temp2 = (step1[4] - step1[3]) * cospi_16_64;
step2[2] = fdct_round_shift(temp1);
step2[3] = fdct_round_shift(temp2);
temp1 = (step1[4] + step1[3]) * cospi_16_64;
temp2 = (step1[5] + step1[2]) * cospi_16_64;
step2[4] = fdct_round_shift(temp1);
step2[5] = fdct_round_shift(temp2);
// step 3
step3[0] = step1[0] + step2[3];
step3[1] = step1[1] + step2[2];
step3[2] = step1[1] - step2[2];
step3[3] = step1[0] - step2[3];
step3[4] = step1[7] - step2[4];
step3[5] = step1[6] - step2[5];
step3[6] = step1[6] + step2[5];
step3[7] = step1[7] + step2[4];
// step 4
temp1 = step3[1] * -cospi_8_64 + step3[6] * cospi_24_64;
temp2 = step3[2] * cospi_24_64 + step3[5] * cospi_8_64;
step2[1] = fdct_round_shift(temp1);
step2[2] = fdct_round_shift(temp2);
temp1 = step3[2] * cospi_8_64 - step3[5] * cospi_24_64;
temp2 = step3[1] * cospi_24_64 + step3[6] * cospi_8_64;
step2[5] = fdct_round_shift(temp1);
step2[6] = fdct_round_shift(temp2);
// step 5
step1[0] = step3[0] + step2[1];
step1[1] = step3[0] - step2[1];
step1[2] = step3[3] + step2[2];
step1[3] = step3[3] - step2[2];
step1[4] = step3[4] - step2[5];
step1[5] = step3[4] + step2[5];
step1[6] = step3[7] - step2[6];
step1[7] = step3[7] + step2[6];
// step 6
temp1 = step1[0] * cospi_30_64 + step1[7] * cospi_2_64;
temp2 = step1[1] * cospi_14_64 + step1[6] * cospi_18_64;
output[14] = fdct_round_shift(temp1);
output[6] = fdct_round_shift(temp2);
temp1 = step1[2] * cospi_22_64 + step1[5] * cospi_10_64;
temp2 = step1[3] * cospi_6_64 + step1[4] * cospi_26_64;
output[10] = fdct_round_shift(temp1);
output[2] = fdct_round_shift(temp2);
temp1 = step1[3] * -cospi_26_64 + step1[4] * cospi_6_64;
temp2 = step1[2] * -cospi_10_64 + step1[5] * cospi_22_64;
output[12] = fdct_round_shift(temp1);
output[4] = fdct_round_shift(temp2);
temp1 = step1[1] * -cospi_18_64 + step1[6] * cospi_14_64;
temp2 = step1[0] * -cospi_2_64 + step1[7] * cospi_30_64;
output[8] = fdct_round_shift(temp1);
output[0] = fdct_round_shift(temp2);
}
#endif // CONFIG_EXT_TX
static void fdct4(const tran_low_t *input, tran_low_t *output) {
tran_high_t temp;
tran_low_t step[4];
@@ -1236,22 +1023,6 @@ static void fidtx32(const tran_low_t *input, tran_low_t *output) {
output[i] = input[i] * 4;
}
// For use in lieu of DST
static void fhalfcenter32(const tran_low_t *input, tran_low_t *output) {
int i;
tran_low_t inputhalf[16];
for (i = 0; i < 8; ++i) {
output[16 + i] = input[i] * 4;
output[24 + i] = input[24 + i] * 4;
}
// Multiply input by sqrt(2)
for (i = 0; i < 16; ++i) {
inputhalf[i] = (tran_low_t)fdct_round_shift(input[i + 8] * Sqrt2);
}
fdct16(inputhalf, output);
// Note overall scaling factor is 4 times orthogonal
}
// For use in lieu of ADST
static void fhalfright32(const tran_low_t *input, tran_low_t *output) {
int i;
@@ -1334,25 +1105,22 @@ static void maybe_flip_input(const int16_t **src, int *src_stride, int l,
case ADST_DCT:
case DCT_ADST:
case ADST_ADST:
case DST_DST:
case DCT_DST:
case DST_DCT:
case DST_ADST:
case ADST_DST:
case IDTX:
case H_DCT:
case V_DCT:
case H_DCT:
case V_ADST:
case H_ADST:
break;
case FLIPADST_DCT:
case FLIPADST_ADST:
case FLIPADST_DST:
case V_FLIPADST:
copy_flipud(*src, *src_stride, l, buff, l);
*src = buff;
*src_stride = l;
break;
case DCT_FLIPADST:
case ADST_FLIPADST:
case DST_FLIPADST:
case H_FLIPADST:
copy_fliplr(*src, *src_stride, l, buff, l);
*src = buff;
*src_stride = l;
@@ -1370,98 +1138,86 @@ static void maybe_flip_input(const int16_t **src, int *src_stride, int l,
#endif // CONFIG_EXT_TX
static const transform_2d FHT_4[] = {
{ fdct4, fdct4 }, // DCT_DCT = 0,
{ fadst4, fdct4 }, // ADST_DCT = 1,
{ fdct4, fadst4 }, // DCT_ADST = 2,
{ fadst4, fadst4 }, // ADST_ADST = 3,
{ fdct4, fdct4 }, // DCT_DCT
{ fadst4, fdct4 }, // ADST_DCT
{ fdct4, fadst4 }, // DCT_ADST
{ fadst4, fadst4 }, // ADST_ADST
#if CONFIG_EXT_TX
{ fadst4, fdct4 }, // FLIPADST_DCT = 4,
{ fdct4, fadst4 }, // DCT_FLIPADST = 5,
{ fadst4, fadst4 }, // FLIPADST_FLIPADST = 6,
{ fadst4, fadst4 }, // ADST_FLIPADST = 7,
{ fadst4, fadst4 }, // FLIPADST_ADST = 8,
{ fdst4, fdct4 }, // DST_DCT = 9,
{ fdct4, fdst4 }, // DCT_DST = 10,
{ fdst4, fadst4 }, // DST_ADST = 11,
{ fadst4, fdst4 }, // ADST_DST = 12,
{ fdst4, fadst4 }, // DST_FLIPADST = 13,
{ fadst4, fdst4 }, // FLIPADST_DST = 14,
{ fdst4, fdst4 }, // DST_DST = 15
{ fidtx4, fidtx4 }, // IDTX = 16
{ fdct4, fidtx4 }, // V_DCT = 17
{ fidtx4, fdct4 }, // H_DCT = 18
{ fadst4, fdct4 }, // FLIPADST_DCT
{ fdct4, fadst4 }, // DCT_FLIPADST
{ fadst4, fadst4 }, // FLIPADST_FLIPADST
{ fadst4, fadst4 }, // ADST_FLIPADST
{ fadst4, fadst4 }, // FLIPADST_ADST
{ fidtx4, fidtx4 }, // IDTX
{ fdct4, fidtx4 }, // V_DCT
{ fidtx4, fdct4 }, // H_DCT
{ fadst4, fidtx4 }, // V_ADST
{ fidtx4, fadst4 }, // H_ADST
{ fadst4, fidtx4 }, // V_FLIPADST
{ fidtx4, fadst4 }, // H_FLIPADST
#endif // CONFIG_EXT_TX
};
static const transform_2d FHT_8[] = {
{ fdct8, fdct8 }, // DCT_DCT = 0,
{ fadst8, fdct8 }, // ADST_DCT = 1,
{ fdct8, fadst8 }, // DCT_ADST = 2,
{ fadst8, fadst8 }, // ADST_ADST = 3,
{ fdct8, fdct8 }, // DCT_DCT
{ fadst8, fdct8 }, // ADST_DCT
{ fdct8, fadst8 }, // DCT_ADST
{ fadst8, fadst8 }, // ADST_ADST
#if CONFIG_EXT_TX
{ fadst8, fdct8 }, // FLIPADST_DCT = 4,
{ fdct8, fadst8 }, // DCT_FLIPADST = 5,
{ fadst8, fadst8 }, // FLIPADST_FLIPADST = 6,
{ fadst8, fadst8 }, // ADST_FLIPADST = 7,
{ fadst8, fadst8 }, // FLIPADST_ADST = 8,
{ fdst8, fdct8 }, // DST_DCT = 9,
{ fdct8, fdst8 }, // DCT_DST = 10,
{ fdst8, fadst8 }, // DST_ADST = 11,
{ fadst8, fdst8 }, // ADST_DST = 12,
{ fdst8, fadst8 }, // DST_FLIPADST = 13,
{ fadst8, fdst8 }, // FLIPADST_DST = 14,
{ fdst8, fdst8 }, // DST_DST = 15
{ fidtx8, fidtx8 }, // IDTX = 16
{ fdct8, fidtx8 }, // V_DCT = 17
{ fidtx8, fdct8 }, // H_DCT = 18
{ fadst8, fdct8 }, // FLIPADST_DCT
{ fdct8, fadst8 }, // DCT_FLIPADST
{ fadst8, fadst8 }, // FLIPADST_FLIPADST
{ fadst8, fadst8 }, // ADST_FLIPADST
{ fadst8, fadst8 }, // FLIPADST_ADST
{ fidtx8, fidtx8 }, // IDTX
{ fdct8, fidtx8 }, // V_DCT
{ fidtx8, fdct8 }, // H_DCT
{ fadst8, fidtx8 }, // V_ADST
{ fidtx8, fadst8 }, // H_ADST
{ fadst8, fidtx8 }, // V_FLIPADST
{ fidtx8, fadst8 }, // H_FLIPADST
#endif // CONFIG_EXT_TX
};
static const transform_2d FHT_16[] = {
{ fdct16, fdct16 }, // DCT_DCT = 0,
{ fadst16, fdct16 }, // ADST_DCT = 1,
{ fdct16, fadst16 }, // DCT_ADST = 2,
{ fadst16, fadst16 }, // ADST_ADST = 3,
{ fdct16, fdct16 }, // DCT_DCT
{ fadst16, fdct16 }, // ADST_DCT
{ fdct16, fadst16 }, // DCT_ADST
{ fadst16, fadst16 }, // ADST_ADST
#if CONFIG_EXT_TX
{ fadst16, fdct16 }, // FLIPADST_DCT = 4,
{ fdct16, fadst16 }, // DCT_FLIPADST = 5,
{ fadst16, fadst16 }, // FLIPADST_FLIPADST = 6,
{ fadst16, fadst16 }, // ADST_FLIPADST = 7,
{ fadst16, fadst16 }, // FLIPADST_ADST = 8,
{ fdst16, fdct16 }, // DST_DCT = 9,
{ fdct16, fdst16 }, // DCT_DST = 10,
{ fdst16, fadst16 }, // DST_ADST = 11,
{ fadst16, fdst16 }, // ADST_DST = 12,
{ fdst16, fadst16 }, // DST_FLIPADST = 13,
{ fadst16, fdst16 }, // FLIPADST_DST = 14,
{ fdst16, fdst16 }, // DST_DST = 15
{ fidtx16, fidtx16 }, // IDTX = 16
{ fdct16, fidtx16 }, // V_DCT = 17
{ fidtx16, fdct16 }, // H_DCT = 18
{ fadst16, fdct16 }, // FLIPADST_DCT
{ fdct16, fadst16 }, // DCT_FLIPADST
{ fadst16, fadst16 }, // FLIPADST_FLIPADST
{ fadst16, fadst16 }, // ADST_FLIPADST
{ fadst16, fadst16 }, // FLIPADST_ADST
{ fidtx16, fidtx16 }, // IDTX
{ fdct16, fidtx16 }, // V_DCT
{ fidtx16, fdct16 }, // H_DCT
{ fadst16, fidtx16 }, // V_ADST
{ fidtx16, fadst16 }, // H_ADST
{ fadst16, fidtx16 }, // V_FLIPADST
{ fidtx16, fadst16 }, // H_FLIPADST
#endif // CONFIG_EXT_TX
};
#if CONFIG_EXT_TX
static const transform_2d FHT_32[] = {
{ fdct32, fdct32 }, // DCT_DCT = 0,
{ fhalfright32, fdct32 }, // ADST_DCT = 1,
{ fdct32, fhalfright32 }, // DCT_ADST = 2,
{ fhalfright32, fhalfright32 }, // ADST_ADST = 3,
{ fhalfright32, fdct32 }, // FLIPADST_DCT = 4,
{ fdct32, fhalfright32 }, // DCT_FLIPADST = 5,
{ fhalfright32, fhalfright32 }, // FLIPADST_FLIPADST = 6,
{ fhalfright32, fhalfright32 }, // ADST_FLIPADST = 7,
{ fhalfright32, fhalfright32 }, // FLIPADST_ADST = 8,
{ fhalfcenter32, fdct32 }, // DST_DCT = 9,
{ fdct32, fhalfcenter32 }, // DCT_DST = 10,
{ fhalfcenter32, fhalfright32 }, // DST_ADST = 11,
{ fhalfright32, fhalfcenter32 }, // ADST_DST = 12,
{ fhalfcenter32, fhalfright32 }, // DST_FLIPADST = 13,
{ fhalfright32, fhalfcenter32 }, // FLIPADST_DST = 14,
{ fhalfcenter32, fhalfcenter32 }, // DST_DST = 15
{ fidtx32, fidtx32 }, // IDTX = 16
{ fdct32, fidtx32 }, // V_DCT = 17
{ fidtx32, fdct32 }, // H_DCT = 18
{ fdct32, fdct32 }, // DCT_DCT
{ fhalfright32, fdct32 }, // ADST_DCT
{ fdct32, fhalfright32 }, // DCT_ADST
{ fhalfright32, fhalfright32 }, // ADST_ADST
{ fhalfright32, fdct32 }, // FLIPADST_DCT
{ fdct32, fhalfright32 }, // DCT_FLIPADST
{ fhalfright32, fhalfright32 }, // FLIPADST_FLIPADST
{ fhalfright32, fhalfright32 }, // ADST_FLIPADST
{ fhalfright32, fhalfright32 }, // FLIPADST_ADST
{ fidtx32, fidtx32 }, // IDTX
{ fdct32, fidtx32 }, // V_DCT
{ fidtx32, fdct32 }, // H_DCT
{ fhalfright32, fidtx32 }, // V_ADST
{ fidtx32, fhalfright32 }, // H_ADST
{ fhalfright32, fidtx32 }, // V_FLIPADST
{ fidtx32, fhalfright32 }, // H_FLIPADST
};
#endif // CONFIG_EXT_TX

5
vp10/encoder/encodeframe.c
View File

@@ -5314,11 +5314,6 @@ static void rd_supertx_sb(VP10_COMP *cpi, ThreadData *td,
#if CONFIG_EXT_TX
if (!ext_tx_used_inter[ext_tx_set][tx_type])
continue;
if (ext_tx_set == 1 &&
tx_type >= DST_ADST && tx_type < IDTX && *best_tx == DCT_DCT) {
tx_type = IDTX - 1;
continue;
}
#else
if (tx_size >= TX_32X32 && tx_type != DCT_DCT)
continue;

104
vp10/encoder/hybrid_fwd_txfm.c
View File

@@ -54,17 +54,14 @@ void vp10_fwd_txfm_4x4(const int16_t *src_diff, tran_low_t *coeff,
case FLIPADST_FLIPADST:
case ADST_FLIPADST:
case FLIPADST_ADST:
case DST_DST:
case DCT_DST:
case DST_DCT:
case DST_ADST:
case ADST_DST:
case DST_FLIPADST:
case FLIPADST_DST:
vp10_fht4x4(src_diff, coeff, diff_stride, tx_type);
break;
case H_DCT:
case V_DCT:
case H_DCT:
case V_ADST:
case H_ADST:
case V_FLIPADST:
case H_FLIPADST:
vp10_fht4x4_c(src_diff, coeff, diff_stride, tx_type);
break;
case IDTX:
@@ -96,17 +93,14 @@ static void fwd_txfm_8x8(const int16_t *src_diff, tran_low_t *coeff,
case FLIPADST_FLIPADST:
case ADST_FLIPADST:
case FLIPADST_ADST:
case DST_DST:
case DCT_DST:
case DST_DCT:
case DST_ADST:
case ADST_DST:
case DST_FLIPADST:
case FLIPADST_DST:
vp10_fht8x8(src_diff, coeff, diff_stride, tx_type);
break;
case H_DCT:
case V_DCT:
case H_DCT:
case V_ADST:
case H_ADST:
case V_FLIPADST:
case H_FLIPADST:
vp10_fht8x8_c(src_diff, coeff, diff_stride, tx_type);
break;
case IDTX:
@@ -138,17 +132,14 @@ static void fwd_txfm_16x16(const int16_t *src_diff, tran_low_t *coeff,
case FLIPADST_FLIPADST:
case ADST_FLIPADST:
case FLIPADST_ADST:
case DST_DST:
case DCT_DST:
case DST_DCT:
case DST_ADST:
case ADST_DST:
case DST_FLIPADST:
case FLIPADST_DST:
vp10_fht16x16(src_diff, coeff, diff_stride, tx_type);
break;
case H_DCT:
case V_DCT:
case H_DCT:
case V_ADST:
case H_ADST:
case V_FLIPADST:
case H_FLIPADST:
vp10_fht16x16_c(src_diff, coeff, diff_stride, tx_type);
break;
case IDTX:
@@ -180,17 +171,14 @@ static void fwd_txfm_32x32(int rd_transform, const int16_t *src_diff,
case FLIPADST_FLIPADST:
case ADST_FLIPADST:
case FLIPADST_ADST:
case DST_DST:
case DCT_DST:
case DST_DCT:
case DST_ADST:
case ADST_DST:
case DST_FLIPADST:
case FLIPADST_DST:
vp10_fht32x32_c(src_diff, coeff, diff_stride, tx_type);
break;
case H_DCT:
case V_DCT:
case H_DCT:
case V_ADST:
case H_ADST:
case V_FLIPADST:
case H_FLIPADST:
vp10_fht32x32_c(src_diff, coeff, diff_stride, tx_type);
break;
case IDTX:
@@ -227,15 +215,12 @@ void vp10_highbd_fwd_txfm_4x4(const int16_t *src_diff, tran_low_t *coeff,
case FLIPADST_ADST:
vp10_highbd_fht4x4(src_diff, coeff, diff_stride, tx_type);
break;
case DST_DST:
case DCT_DST:
case DST_DCT:
case DST_ADST:
case ADST_DST:
case DST_FLIPADST:
case FLIPADST_DST:
case H_DCT:
case V_DCT:
case H_DCT:
case V_ADST:
case H_ADST:
case V_FLIPADST:
case H_FLIPADST:
vp10_highbd_fht4x4_c(src_diff, coeff, diff_stride, tx_type);
break;
case IDTX:
@@ -270,15 +255,12 @@ static void highbd_fwd_txfm_8x8(const int16_t *src_diff, tran_low_t *coeff,
case FLIPADST_ADST:
vp10_highbd_fht8x8(src_diff, coeff, diff_stride, tx_type);
break;
case DST_DST:
case DCT_DST:
case DST_DCT:
case DST_ADST:
case ADST_DST:
case DST_FLIPADST:
case FLIPADST_DST:
case H_DCT:
case V_DCT:
case H_DCT:
case V_ADST:
case H_ADST:
case V_FLIPADST:
case H_FLIPADST:
// Use C version since DST exists only in C
vp10_highbd_fht8x8_c(src_diff, coeff, diff_stride, tx_type);
break;
@@ -314,15 +296,12 @@ static void highbd_fwd_txfm_16x16(const int16_t *src_diff, tran_low_t *coeff,
case FLIPADST_ADST:
vp10_highbd_fht16x16(src_diff, coeff, diff_stride, tx_type);
break;
case DST_DST:
case DCT_DST:
case DST_DCT:
case DST_ADST:
case ADST_DST:
case DST_FLIPADST:
case FLIPADST_DST:
case H_DCT:
case V_DCT:
case H_DCT:
case V_ADST:
case H_ADST:
case V_FLIPADST:
case H_FLIPADST:
// Use C version since DST exists only in C
vp10_highbd_fht16x16_c(src_diff, coeff, diff_stride, tx_type);
break;
@@ -355,15 +334,12 @@ static void highbd_fwd_txfm_32x32(int rd_transform, const int16_t *src_diff,
case FLIPADST_FLIPADST:
case ADST_FLIPADST:
case FLIPADST_ADST:
case DST_DST:
case DCT_DST:
case DST_DCT:
case DST_ADST:
case ADST_DST:
case DST_FLIPADST:
case FLIPADST_DST:
case H_DCT:
case V_DCT:
case H_DCT:
case V_ADST:
case H_ADST:
case V_FLIPADST:
case H_FLIPADST:
vp10_highbd_fht32x32_c(src_diff, coeff, diff_stride, tx_type);
break;
case IDTX:

33
vp10/encoder/rdopt.c
View File

@@ -1349,11 +1349,6 @@ static void choose_largest_tx_size(VP10_COMP *cpi, MACROBLOCK *x,
if (cpi->sf.tx_type_search > 0) {
if (!do_tx_type_search(tx_type, prune))
continue;
} else if (ext_tx_set == 1 &&
tx_type >= DST_ADST && tx_type < IDTX &&
best_tx_type == DCT_DCT) {
tx_type = IDTX - 1;
continue;
}
} else {
if (!ALLOW_INTRA_EXT_TX && bs >= BLOCK_8X8) {
@@ -1362,12 +1357,6 @@ static void choose_largest_tx_size(VP10_COMP *cpi, MACROBLOCK *x,
}
if (!ext_tx_used_intra[ext_tx_set][tx_type])
continue;
if (ext_tx_set == 1 &&
tx_type >= DST_ADST && tx_type < IDTX &&
best_tx_type == DCT_DCT) {
tx_type = IDTX - 1;
continue;
}
}
mbmi->tx_type = tx_type;
@@ -1533,11 +1522,6 @@ static void choose_tx_size_from_rd(VP10_COMP *cpi, MACROBLOCK *x,
if (cpi->sf.tx_type_search > 0) {
if (!do_tx_type_search(tx_type, prune))
continue;
} else if (ext_tx_set == 1 &&
tx_type >= DST_ADST && tx_type < IDTX &&
best_tx_type == DCT_DCT) {
tx_type = IDTX - 1;
continue;
}
} else {
if (!ALLOW_INTRA_EXT_TX && bs >= BLOCK_8X8) {
@@ -1546,12 +1530,6 @@ static void choose_tx_size_from_rd(VP10_COMP *cpi, MACROBLOCK *x,
}
if (!ext_tx_used_intra[ext_tx_set][tx_type])
continue;
if (ext_tx_set == 1 &&
tx_type >= DST_ADST && tx_type < IDTX &&
best_tx_type == DCT_DCT) {
tx_type = IDTX - 1;
break;
}
}
mbmi->tx_type = tx_type;
txfm_rd_in_plane(x,
@@ -3169,11 +3147,6 @@ static void select_tx_type_yrd(const VP10_COMP *cpi, MACROBLOCK *x,
if (cpi->sf.tx_type_search > 0) {
if (!do_tx_type_search(tx_type, prune))
continue;
} else if (ext_tx_set == 1 &&
tx_type >= DST_ADST && tx_type < IDTX &&
best_tx_type == DCT_DCT) {
tx_type = IDTX - 1;
continue;
}
} else {
if (!ALLOW_INTRA_EXT_TX && bsize >= BLOCK_8X8) {
@@ -3182,12 +3155,6 @@ static void select_tx_type_yrd(const VP10_COMP *cpi, MACROBLOCK *x,
}
if (!ext_tx_used_intra[ext_tx_set][tx_type])
continue;
if (ext_tx_set == 1 &&
tx_type >= DST_ADST && tx_type < IDTX &&
best_tx_type == DCT_DCT) {
tx_type = IDTX - 1;
break;
}
}
mbmi->tx_type = tx_type;

678
vp10/encoder/x86/dct_sse2.c
View File

@@ -172,42 +172,6 @@ static void fadst4_sse2(__m128i *in) {
transpose_4x4(in);
}
#if CONFIG_EXT_TX
static void fdst4_sse2(__m128i *in) {
const __m128i k__cospi_p16_p16 = _mm_set1_epi16((int16_t) cospi_16_64);
const __m128i k__cospi_p16_m16 = pair_set_epi16(cospi_16_64, -cospi_16_64);
const __m128i k__cospi_p08_m24 = pair_set_epi16(cospi_8_64, -cospi_24_64);
const __m128i k__cospi_p24_p08 = pair_set_epi16(cospi_24_64, cospi_8_64);
const __m128i k__DCT_CONST_ROUNDING = _mm_set1_epi32(DCT_CONST_ROUNDING);
__m128i u[4], v[4];
u[0] = _mm_unpacklo_epi16(in[0], in[1]);
u[1] = _mm_unpacklo_epi16(in[3], in[2]);
v[0] = _mm_add_epi16(u[0], u[1]);
v[1] = _mm_sub_epi16(u[0], u[1]);
u[0] = _mm_madd_epi16(v[0], k__cospi_p24_p08);
u[1] = _mm_madd_epi16(v[1], k__cospi_p16_p16);
u[2] = _mm_madd_epi16(v[0], k__cospi_p08_m24);
u[3] = _mm_madd_epi16(v[1], k__cospi_p16_m16);
v[0] = _mm_add_epi32(u[0], k__DCT_CONST_ROUNDING);
v[1] = _mm_add_epi32(u[1], k__DCT_CONST_ROUNDING);
v[2] = _mm_add_epi32(u[2], k__DCT_CONST_ROUNDING);
v[3] = _mm_add_epi32(u[3], k__DCT_CONST_ROUNDING);
u[0] = _mm_srai_epi32(v[0], DCT_CONST_BITS);
u[1] = _mm_srai_epi32(v[1], DCT_CONST_BITS);
u[2] = _mm_srai_epi32(v[2], DCT_CONST_BITS);
u[3] = _mm_srai_epi32(v[3], DCT_CONST_BITS);
in[0] = _mm_packs_epi32(u[0], u[2]);
in[1] = _mm_packs_epi32(u[1], u[3]);
transpose_4x4(in);
}
#endif // CONFIG_EXT_TX
void vp10_fht4x4_sse2(const int16_t *input, tran_low_t *output,
int stride, int tx_type) {
__m128i in[4];
@@ -265,48 +229,6 @@ void vp10_fht4x4_sse2(const int16_t *input, tran_low_t *output,
fadst4_sse2(in);
write_buffer_4x4(output, in);
break;
case DST_DST:
load_buffer_4x4(input, in, stride, 0, 0);
fdst4_sse2(in);
fdst4_sse2(in);
write_buffer_4x4(output, in);
break;
case DCT_DST:
load_buffer_4x4(input, in, stride, 0, 0);
fdct4_sse2(in);
fdst4_sse2(in);
write_buffer_4x4(output, in);
break;
case DST_DCT:
load_buffer_4x4(input, in, stride, 0, 0);
fdst4_sse2(in);
fdct4_sse2(in);
write_buffer_4x4(output, in);
break;
case DST_ADST:
load_buffer_4x4(input, in, stride, 0, 0);
fdst4_sse2(in);
fadst4_sse2(in);
write_buffer_4x4(output, in);
break;
case ADST_DST:
load_buffer_4x4(input, in, stride, 0, 0);
fadst4_sse2(in);
fdst4_sse2(in);
write_buffer_4x4(output, in);
break;
case DST_FLIPADST:
load_buffer_4x4(input, in, stride, 0, 1);
fdst4_sse2(in);
fadst4_sse2(in);
write_buffer_4x4(output, in);
break;
case FLIPADST_DST:
load_buffer_4x4(input, in, stride, 1, 0);
fadst4_sse2(in);
fdst4_sse2(in);
write_buffer_4x4(output, in);
break;
#endif // CONFIG_EXT_TX
default:
assert(0);
@@ -1288,155 +1210,6 @@ static void fadst8_sse2(__m128i *in) {
array_transpose_8x8(in, in);
}
#if CONFIG_EXT_TX
static void fdst8_sse2(__m128i *in) {
// Constants
const __m128i k__cospi_p16_p16 = _mm_set1_epi16((int16_t) cospi_16_64);
const __m128i k__cospi_p16_m16 = pair_set_epi16(cospi_16_64, -cospi_16_64);
const __m128i k__cospi_m16_m16 = _mm_set1_epi16((int16_t) -cospi_16_64);
const __m128i k__cospi_m16_p16 = pair_set_epi16(-cospi_16_64, cospi_16_64);
const __m128i k__cospi_m24_p08 = pair_set_epi16(-cospi_24_64, cospi_8_64);
const __m128i k__cospi_p08_p24 = pair_set_epi16(cospi_8_64, cospi_24_64);
const __m128i k__cospi_p28_p04 = pair_set_epi16(cospi_28_64, cospi_4_64);
const __m128i k__cospi_m04_p28 = pair_set_epi16(-cospi_4_64, cospi_28_64);
const __m128i k__cospi_p20_p12 = pair_set_epi16(cospi_20_64, cospi_12_64);
const __m128i k__cospi_m12_p20 = pair_set_epi16(-cospi_12_64, cospi_20_64);
const __m128i k__DCT_CONST_ROUNDING = _mm_set1_epi32(DCT_CONST_ROUNDING);
__m128i s0, s1, s2, s3, s4, s5, s6, s7;
__m128i x0, x1, x2, x3, x4, x5, x6, x7;
__m128i t0, t1, t2, t3, t4, t5, t6, t7;
s0 = _mm_sub_epi16(in[0], in[7]);
s1 = _mm_sub_epi16(in[1], in[6]); // -s1
s2 = _mm_sub_epi16(in[2], in[5]);
s3 = _mm_sub_epi16(in[3], in[4]); // -s3
s4 = _mm_add_epi16(in[3], in[4]); // -s4
s5 = _mm_add_epi16(in[2], in[5]);
s6 = _mm_add_epi16(in[1], in[6]); // -s6
s7 = _mm_add_epi16(in[0], in[7]);
x0 = _mm_sub_epi16(s0, s3);
x1 = _mm_sub_epi16(s1, s2); // -x1
x2 = _mm_add_epi16(s1, s2); // -x2
x3 = _mm_add_epi16(s0, s3);
// Interleave
t0 = _mm_unpacklo_epi16(x0, x1);
t1 = _mm_unpackhi_epi16(x0, x1);
t2 = _mm_unpacklo_epi16(x2, x3);
t3 = _mm_unpackhi_epi16(x2, x3);
// Perform butterfly multiplication/addition
x0 = _mm_madd_epi16(t0, k__cospi_p16_m16);
x1 = _mm_madd_epi16(t1, k__cospi_p16_m16);
x2 = _mm_madd_epi16(t0, k__cospi_p16_p16);
x3 = _mm_madd_epi16(t1, k__cospi_p16_p16);
x4 = _mm_madd_epi16(t2, k__cospi_m24_p08);
x5 = _mm_madd_epi16(t3, k__cospi_m24_p08);
x6 = _mm_madd_epi16(t2, k__cospi_p08_p24);
x7 = _mm_madd_epi16(t3, k__cospi_p08_p24);
// Rounding
t0 = _mm_add_epi32(x0, k__DCT_CONST_ROUNDING);
t1 = _mm_add_epi32(x1, k__DCT_CONST_ROUNDING);
t2 = _mm_add_epi32(x2, k__DCT_CONST_ROUNDING);
t3 = _mm_add_epi32(x3, k__DCT_CONST_ROUNDING);
t4 = _mm_add_epi32(x4, k__DCT_CONST_ROUNDING);
t5 = _mm_add_epi32(x5, k__DCT_CONST_ROUNDING);
t6 = _mm_add_epi32(x6, k__DCT_CONST_ROUNDING);
t7 = _mm_add_epi32(x7, k__DCT_CONST_ROUNDING);
// Shift
x0 = _mm_srai_epi32(t0, DCT_CONST_BITS);
x1 = _mm_srai_epi32(t1, DCT_CONST_BITS);
x2 = _mm_srai_epi32(t2, DCT_CONST_BITS);
x3 = _mm_srai_epi32(t3, DCT_CONST_BITS);
x4 = _mm_srai_epi32(t4, DCT_CONST_BITS);
x5 = _mm_srai_epi32(t5, DCT_CONST_BITS);
x6 = _mm_srai_epi32(t6, DCT_CONST_BITS);
x7 = _mm_srai_epi32(t7, DCT_CONST_BITS);
// Pack 32b integer to 16b with signed saturation
in[7] = _mm_packs_epi32(x0, x1);
in[5] = _mm_packs_epi32(x4, x5);
in[3] = _mm_packs_epi32(x2, x3);
in[1] = _mm_packs_epi32(x6, x7);
// Interleave
s0 = _mm_unpacklo_epi16(s6, s5);
s1 = _mm_unpackhi_epi16(s6, s5);
// Perform butterfly multiplication/addition
x0 = _mm_madd_epi16(s0, k__cospi_m16_m16);
x1 = _mm_madd_epi16(s1, k__cospi_m16_m16);
x2 = _mm_madd_epi16(s0, k__cospi_m16_p16);
x3 = _mm_madd_epi16(s1, k__cospi_m16_p16);
// Rounding
t0 = _mm_add_epi32(x0, k__DCT_CONST_ROUNDING);
t1 = _mm_add_epi32(x1, k__DCT_CONST_ROUNDING);
t2 = _mm_add_epi32(x2, k__DCT_CONST_ROUNDING);
t3 = _mm_add_epi32(x3, k__DCT_CONST_ROUNDING);
// Shift
x0 = _mm_srai_epi32(t0, DCT_CONST_BITS);
x1 = _mm_srai_epi32(t1, DCT_CONST_BITS);
x2 = _mm_srai_epi32(t2, DCT_CONST_BITS);
x3 = _mm_srai_epi32(t3, DCT_CONST_BITS);
// Pack 32b integer to 16b with signed saturation
t2 = _mm_packs_epi32(x0, x1);
t3 = _mm_packs_epi32(x2, x3);
x0 = _mm_sub_epi16(t2, s4);
x1 = _mm_add_epi16(t2, s4); // -x1
x2 = _mm_sub_epi16(s7, t3);
x3 = _mm_add_epi16(s7, t3);
s0 = _mm_unpacklo_epi16(x0, x3);
s1 = _mm_unpackhi_epi16(x0, x3);
s2 = _mm_unpacklo_epi16(x1, x2);
s3 = _mm_unpackhi_epi16(x1, x2);
t0 = _mm_madd_epi16(s0, k__cospi_p28_p04);
t1 = _mm_madd_epi16(s1, k__cospi_p28_p04);
t2 = _mm_madd_epi16(s2, k__cospi_m12_p20);
t3 = _mm_madd_epi16(s3, k__cospi_m12_p20);
t4 = _mm_madd_epi16(s2, k__cospi_p20_p12);
t5 = _mm_madd_epi16(s3, k__cospi_p20_p12);
t6 = _mm_madd_epi16(s0, k__cospi_m04_p28);
t7 = _mm_madd_epi16(s1, k__cospi_m04_p28);
// Rounding
x0 = _mm_add_epi32(t0, k__DCT_CONST_ROUNDING);
x1 = _mm_add_epi32(t1, k__DCT_CONST_ROUNDING);
x2 = _mm_add_epi32(t2, k__DCT_CONST_ROUNDING);
x3 = _mm_add_epi32(t3, k__DCT_CONST_ROUNDING);
x4 = _mm_add_epi32(t4, k__DCT_CONST_ROUNDING);
x5 = _mm_add_epi32(t5, k__DCT_CONST_ROUNDING);
x6 = _mm_add_epi32(t6, k__DCT_CONST_ROUNDING);
x7 = _mm_add_epi32(t7, k__DCT_CONST_ROUNDING);
// Shift
s0 = _mm_srai_epi32(x0, DCT_CONST_BITS);
s1 = _mm_srai_epi32(x1, DCT_CONST_BITS);
s2 = _mm_srai_epi32(x2, DCT_CONST_BITS);
s3 = _mm_srai_epi32(x3, DCT_CONST_BITS);
s4 = _mm_srai_epi32(x4, DCT_CONST_BITS);
s5 = _mm_srai_epi32(x5, DCT_CONST_BITS);
s6 = _mm_srai_epi32(x6, DCT_CONST_BITS);
s7 = _mm_srai_epi32(x7, DCT_CONST_BITS);
in[6] = _mm_packs_epi32(s0, s1);
in[4] = _mm_packs_epi32(s4, s5);
in[2] = _mm_packs_epi32(s2, s3);
in[0] = _mm_packs_epi32(s6, s7);
// coeffs: [x3 x2 x1 x0, x7 x6 x5 x4]
// Transpose
array_transpose_8x8(in, in);
}
#endif // CONFIG_EXT_TX
void vp10_fht8x8_sse2(const int16_t *input, tran_low_t *output,
int stride, int tx_type) {
__m128i in[8];
@@ -1502,55 +1275,6 @@ void vp10_fht8x8_sse2(const int16_t *input, tran_low_t *output,
right_shift_8x8(in, 1);
write_buffer_8x8(output, in, 8);
break;
case DST_DST:
load_buffer_8x8(input, in, stride, 0, 0);
fdst8_sse2(in);
fdst8_sse2(in);
right_shift_8x8(in, 1);
write_buffer_8x8(output, in, 8);
break;
case DCT_DST:
load_buffer_8x8(input, in, stride, 0, 0);
fdct8_sse2(in);
fdst8_sse2(in);
right_shift_8x8(in, 1);
write_buffer_8x8(output, in, 8);
break;
case DST_DCT:
load_buffer_8x8(input, in, stride, 0, 0);
fdst8_sse2(in);
fdct8_sse2(in);
right_shift_8x8(in, 1);
write_buffer_8x8(output, in, 8);
break;
case DST_ADST:
load_buffer_8x8(input, in, stride, 0, 0);
fdst8_sse2(in);
fadst8_sse2(in);
right_shift_8x8(in, 1);
write_buffer_8x8(output, in, 8);
break;
case ADST_DST:
load_buffer_8x8(input, in, stride, 0, 0);
fadst8_sse2(in);
fdst8_sse2(in);
right_shift_8x8(in, 1);
write_buffer_8x8(output, in, 8);
break;
case DST_FLIPADST:
load_buffer_8x8(input, in, stride, 0, 1);
fdst8_sse2(in);
fadst8_sse2(in);
right_shift_8x8(in, 1);
write_buffer_8x8(output, in, 8);
break;
case FLIPADST_DST:
load_buffer_8x8(input, in, stride, 1, 0);
fadst8_sse2(in);
fdst8_sse2(in);
right_shift_8x8(in, 1);
write_buffer_8x8(output, in, 8);
break;
#endif // CONFIG_EXT_TX
default:
assert(0);
@@ -2420,351 +2144,6 @@ static void fadst16_8col(__m128i *in) {
in[15] = _mm_sub_epi16(kZero, s[1]);
}
#if CONFIG_EXT_TX
static void fdst16_8col(__m128i *in) {
const __m128i k__cospi_p16_m16 = pair_set_epi16(cospi_16_64, -cospi_16_64);
const __m128i k__cospi_p16_p16 = _mm_set1_epi16((int16_t) cospi_16_64);
const __m128i k__cospi_m24_p08 = pair_set_epi16(-cospi_24_64, cospi_8_64);
const __m128i k__cospi_p08_p24 = pair_set_epi16(cospi_8_64, cospi_24_64);
const __m128i k__cospi_m16_m16 = _mm_set1_epi16((int16_t) -cospi_16_64);
const __m128i k__cospi_m16_p16 = pair_set_epi16(-cospi_16_64, cospi_16_64);
const __m128i k__cospi_m28_p04 = pair_set_epi16(-cospi_28_64, cospi_4_64);
const __m128i k__cospi_m12_p20 = pair_set_epi16(-cospi_12_64, cospi_20_64);
const __m128i k__cospi_p20_p12 = pair_set_epi16(cospi_20_64, cospi_12_64);
const __m128i k__cospi_p04_p28 = pair_set_epi16(cospi_4_64, cospi_28_64);
const __m128i k__cospi_m08_m24 = pair_set_epi16(-cospi_8_64, -cospi_24_64);
const __m128i k__cospi_p24_m08 = pair_set_epi16(cospi_24_64, -cospi_8_64);
const __m128i k__cospi_m30_p02 = pair_set_epi16(-cospi_30_64, cospi_2_64);
const __m128i k__cospi_m14_p18 = pair_set_epi16(-cospi_14_64, cospi_18_64);
const __m128i k__cospi_m22_p10 = pair_set_epi16(-cospi_22_64, cospi_10_64);
const __m128i k__cospi_m06_p26 = pair_set_epi16(-cospi_6_64, cospi_26_64);
const __m128i k__cospi_p26_p06 = pair_set_epi16(cospi_26_64, cospi_6_64);
const __m128i k__cospi_p10_p22 = pair_set_epi16(cospi_10_64, cospi_22_64);
const __m128i k__cospi_p18_p14 = pair_set_epi16(cospi_18_64, cospi_14_64);
const __m128i k__cospi_p02_p30 = pair_set_epi16(cospi_2_64, cospi_30_64);
const __m128i k__DCT_CONST_ROUNDING = _mm_set1_epi32(DCT_CONST_ROUNDING);
__m128i u0, u1, u2, u3, u4, u5, u6, u7;
__m128i v0, v1, v2, v3, v4, v5, v6, v7;
__m128i s0, s1, s2, s3, s4, s5, s6, s7;
__m128i x0, x1, x2, x3, t0, t1, t2, t3;
__m128i y0, y1, y2, y3, y4, y5, y6, y7;
__m128i w0, w1, w2, w3, w4, w5, w6, w7;
// (1)
u0 = _mm_sub_epi16(in[0], in[15]);
v7 = _mm_add_epi16(in[0], in[15]);
u1 = _mm_sub_epi16(in[1], in[14]); // -u1
v6 = _mm_add_epi16(in[1], in[14]); // -v6
u2 = _mm_sub_epi16(in[2], in[13]);
v5 = _mm_add_epi16(in[2], in[13]);
u3 = _mm_sub_epi16(in[3], in[12]); // -u3
v4 = _mm_add_epi16(in[3], in[12]); // -v4
u4 = _mm_sub_epi16(in[4], in[11]);
v3 = _mm_add_epi16(in[4], in[11]);
u5 = _mm_sub_epi16(in[5], in[10]); // -u5
v2 = _mm_add_epi16(in[5], in[10]); // -v2
u6 = _mm_sub_epi16(in[6], in[9]);
v1 = _mm_add_epi16(in[6], in[9]);
u7 = _mm_sub_epi16(in[7], in[8]); // -u7
v0 = _mm_add_epi16(in[7], in[8]); // -v0
s0 = _mm_sub_epi16(u0, u7);
s1 = _mm_sub_epi16(u1, u6); // -s1
s2 = _mm_sub_epi16(u2, u5);
s3 = _mm_sub_epi16(u3, u4); // -s3
s4 = _mm_add_epi16(u3, u4); // -s4
s5 = _mm_add_epi16(u2, u5);
s6 = _mm_add_epi16(u1, u6); // -s6
s7 = _mm_add_epi16(u0, u7);
x0 = _mm_sub_epi16(s0, s3);
x1 = _mm_sub_epi16(s1, s2); // -x1
x2 = _mm_add_epi16(s1, s2); // -x2
x3 = _mm_add_epi16(s0, s3);
y0 = _mm_unpacklo_epi16(x0, x1);
y1 = _mm_unpackhi_epi16(x0, x1);
y2 = _mm_unpacklo_epi16(x2, x3);
y3 = _mm_unpackhi_epi16(x2, x3);
t0 = _mm_madd_epi16(y0, k__cospi_p16_m16);
t1 = _mm_madd_epi16(y1, k__cospi_p16_m16);
t2 = _mm_madd_epi16(y0, k__cospi_p16_p16);
t3 = _mm_madd_epi16(y1, k__cospi_p16_p16);
x0 = _mm_madd_epi16(y2, k__cospi_m24_p08);
x1 = _mm_madd_epi16(y3, k__cospi_m24_p08);
x2 = _mm_madd_epi16(y2, k__cospi_p08_p24);
x3 = _mm_madd_epi16(y3, k__cospi_p08_p24);
y0 = _mm_add_epi32(t0, k__DCT_CONST_ROUNDING);
y1 = _mm_add_epi32(t1, k__DCT_CONST_ROUNDING);
y2 = _mm_add_epi32(t2, k__DCT_CONST_ROUNDING);
y3 = _mm_add_epi32(t3, k__DCT_CONST_ROUNDING);
y4 = _mm_add_epi32(x0, k__DCT_CONST_ROUNDING);
y5 = _mm_add_epi32(x1, k__DCT_CONST_ROUNDING);
y6 = _mm_add_epi32(x2, k__DCT_CONST_ROUNDING);
y7 = _mm_add_epi32(x3, k__DCT_CONST_ROUNDING);
t0 = _mm_srai_epi32(y0, DCT_CONST_BITS);
t1 = _mm_srai_epi32(y1, DCT_CONST_BITS);
t2 = _mm_srai_epi32(y2, DCT_CONST_BITS);
t3 = _mm_srai_epi32(y3, DCT_CONST_BITS);
x0 = _mm_srai_epi32(y4, DCT_CONST_BITS);
x1 = _mm_srai_epi32(y5, DCT_CONST_BITS);
x2 = _mm_srai_epi32(y6, DCT_CONST_BITS);
x3 = _mm_srai_epi32(y7, DCT_CONST_BITS);
in[15] = _mm_packs_epi32(t0, t1);
in[11] = _mm_packs_epi32(x0, x1);
in[7] = _mm_packs_epi32(t2, t3);
in[3] = _mm_packs_epi32(x2, x3);
// (2)
t0 = _mm_unpacklo_epi16(s6, s5);
t1 = _mm_unpackhi_epi16(s6, s5);
y0 = _mm_madd_epi16(t0, k__cospi_m16_m16);
y1 = _mm_madd_epi16(t1, k__cospi_m16_m16);
y2 = _mm_madd_epi16(t0, k__cospi_m16_p16);
y3 = _mm_madd_epi16(t1, k__cospi_m16_p16);
x0 = _mm_add_epi32(y0, k__DCT_CONST_ROUNDING);
x1 = _mm_add_epi32(y1, k__DCT_CONST_ROUNDING);
x2 = _mm_add_epi32(y2, k__DCT_CONST_ROUNDING);
x3 = _mm_add_epi32(y3, k__DCT_CONST_ROUNDING);
y4 = _mm_srai_epi32(x0, DCT_CONST_BITS);
y5 = _mm_srai_epi32(x1, DCT_CONST_BITS);
y6 = _mm_srai_epi32(x2, DCT_CONST_BITS);
y7 = _mm_srai_epi32(x3, DCT_CONST_BITS);
t2 = _mm_packs_epi32(y4, y5);
t3 = _mm_packs_epi32(y6, y7);
x0 = _mm_sub_epi16(s4, t2); // -x0
x1 = _mm_add_epi16(s4, t2); // -x1
x2 = _mm_sub_epi16(s7, t3);
x3 = _mm_add_epi16(s7, t3);
y0 = _mm_unpacklo_epi16(x0, x3);
y1 = _mm_unpackhi_epi16(x0, x3);
y2 = _mm_unpacklo_epi16(x1, x2);
y3 = _mm_unpackhi_epi16(x1, x2);
w0 = _mm_madd_epi16(y0, k__cospi_m28_p04);
w1 = _mm_madd_epi16(y1, k__cospi_m28_p04);
w2 = _mm_madd_epi16(y2, k__cospi_m12_p20);
w3 = _mm_madd_epi16(y3, k__cospi_m12_p20);
w4 = _mm_madd_epi16(y2, k__cospi_p20_p12);
w5 = _mm_madd_epi16(y3, k__cospi_p20_p12);
w6 = _mm_madd_epi16(y0, k__cospi_p04_p28);
w7 = _mm_madd_epi16(y1, k__cospi_p04_p28);
u0 = _mm_add_epi32(w0, k__DCT_CONST_ROUNDING);
u1 = _mm_add_epi32(w1, k__DCT_CONST_ROUNDING);
u2 = _mm_add_epi32(w2, k__DCT_CONST_ROUNDING);
u3 = _mm_add_epi32(w3, k__DCT_CONST_ROUNDING);
u4 = _mm_add_epi32(w4, k__DCT_CONST_ROUNDING);
u5 = _mm_add_epi32(w5, k__DCT_CONST_ROUNDING);
u6 = _mm_add_epi32(w6, k__DCT_CONST_ROUNDING);
u7 = _mm_add_epi32(w7, k__DCT_CONST_ROUNDING);
y0 = _mm_srai_epi32(u0, DCT_CONST_BITS);
y1 = _mm_srai_epi32(u1, DCT_CONST_BITS);
y2 = _mm_srai_epi32(u2, DCT_CONST_BITS);
y3 = _mm_srai_epi32(u3, DCT_CONST_BITS);
y4 = _mm_srai_epi32(u4, DCT_CONST_BITS);
y5 = _mm_srai_epi32(u5, DCT_CONST_BITS);
y6 = _mm_srai_epi32(u6, DCT_CONST_BITS);
y7 = _mm_srai_epi32(u7, DCT_CONST_BITS);
in[13] = _mm_packs_epi32(y0, y1);
in[9] = _mm_packs_epi32(y4, y5);
in[5] = _mm_packs_epi32(y2, y3);
in[1] = _mm_packs_epi32(y6, y7);
// (3)
y0 = _mm_unpacklo_epi16(v5, v2);
y1 = _mm_unpackhi_epi16(v5, v2);
y2 = _mm_unpacklo_epi16(v4, v3);
y3 = _mm_unpackhi_epi16(v4, v3);
u0 = _mm_madd_epi16(y0, k__cospi_p16_p16);
u1 = _mm_madd_epi16(y1, k__cospi_p16_p16);
u2 = _mm_madd_epi16(y2, k__cospi_m16_m16);
u3 = _mm_madd_epi16(y3, k__cospi_m16_m16);
u4 = _mm_madd_epi16(y2, k__cospi_m16_p16);
u5 = _mm_madd_epi16(y3, k__cospi_m16_p16);
u6 = _mm_madd_epi16(y0, k__cospi_p16_m16);
u7 = _mm_madd_epi16(y1, k__cospi_p16_m16);
w0 = _mm_add_epi32(u0, k__DCT_CONST_ROUNDING);
w1 = _mm_add_epi32(u1, k__DCT_CONST_ROUNDING);
w2 = _mm_add_epi32(u2, k__DCT_CONST_ROUNDING);
w3 = _mm_add_epi32(u3, k__DCT_CONST_ROUNDING);
w4 = _mm_add_epi32(u4, k__DCT_CONST_ROUNDING);
w5 = _mm_add_epi32(u5, k__DCT_CONST_ROUNDING);
w6 = _mm_add_epi32(u6, k__DCT_CONST_ROUNDING);
w7 = _mm_add_epi32(u7, k__DCT_CONST_ROUNDING);
s0 = _mm_srai_epi32(w0, DCT_CONST_BITS);
s1 = _mm_srai_epi32(w1, DCT_CONST_BITS);
s2 = _mm_srai_epi32(w2, DCT_CONST_BITS);
s3 = _mm_srai_epi32(w3, DCT_CONST_BITS);
s4 = _mm_srai_epi32(w4, DCT_CONST_BITS);
s5 = _mm_srai_epi32(w5, DCT_CONST_BITS);
s6 = _mm_srai_epi32(w6, DCT_CONST_BITS);
s7 = _mm_srai_epi32(w7, DCT_CONST_BITS);
y2 = _mm_packs_epi32(s0, s1);
y3 = _mm_packs_epi32(s2, s3);
y4 = _mm_packs_epi32(s4, s5);
y5 = _mm_packs_epi32(s6, s7);
// step 3
w0 = _mm_sub_epi16(v0, y3); // -w0
w1 = _mm_add_epi16(v1, y2);
w2 = _mm_sub_epi16(v1, y2);
w3 = _mm_add_epi16(v0, y3); // -w3
w4 = _mm_sub_epi16(v7, y4);
w5 = _mm_add_epi16(v6, y5); // -w5
w6 = _mm_sub_epi16(v6, y5); // -w6
w7 = _mm_add_epi16(v7, y4);
// step 4
x0 = _mm_unpacklo_epi16(w1, w6);
x1 = _mm_unpackhi_epi16(w1, w6);
x2 = _mm_unpacklo_epi16(w2, w5);
x3 = _mm_unpackhi_epi16(w2, w5);
u0 = _mm_madd_epi16(x0, k__cospi_m08_m24);
u1 = _mm_madd_epi16(x1, k__cospi_m08_m24);
u2 = _mm_madd_epi16(x2, k__cospi_p24_m08);
u3 = _mm_madd_epi16(x3, k__cospi_p24_m08);
u4 = _mm_madd_epi16(x2, k__cospi_p08_p24);
u5 = _mm_madd_epi16(x3, k__cospi_p08_p24);
u6 = _mm_madd_epi16(x0, k__cospi_p24_m08);
u7 = _mm_madd_epi16(x1, k__cospi_p24_m08);
s0 = _mm_add_epi32(u0, k__DCT_CONST_ROUNDING);
s1 = _mm_add_epi32(u1, k__DCT_CONST_ROUNDING);
s2 = _mm_add_epi32(u2, k__DCT_CONST_ROUNDING);
s3 = _mm_add_epi32(u3, k__DCT_CONST_ROUNDING);
s4 = _mm_add_epi32(u4, k__DCT_CONST_ROUNDING);
s5 = _mm_add_epi32(u5, k__DCT_CONST_ROUNDING);
s6 = _mm_add_epi32(u6, k__DCT_CONST_ROUNDING);
s7 = _mm_add_epi32(u7, k__DCT_CONST_ROUNDING);
u0 = _mm_srai_epi32(s0, DCT_CONST_BITS);
u1 = _mm_srai_epi32(s1, DCT_CONST_BITS);
u2 = _mm_srai_epi32(s2, DCT_CONST_BITS);
u3 = _mm_srai_epi32(s3, DCT_CONST_BITS);
u4 = _mm_srai_epi32(s4, DCT_CONST_BITS);
u5 = _mm_srai_epi32(s5, DCT_CONST_BITS);
u6 = _mm_srai_epi32(s6, DCT_CONST_BITS);
u7 = _mm_srai_epi32(s7, DCT_CONST_BITS);
y1 = _mm_packs_epi32(u0, u1);
y2 = _mm_packs_epi32(u2, u3);
y5 = _mm_packs_epi32(u4, u5);
y6 = _mm_packs_epi32(u6, u7);
// step 5
v0 = _mm_sub_epi16(w0, y1); // -v0
v1 = _mm_add_epi16(w0, y1); // -v1
v2 = _mm_sub_epi16(w3, y2); // -v2
v3 = _mm_add_epi16(w3, y2); // -v3
v4 = _mm_sub_epi16(w4, y5);
v5 = _mm_add_epi16(w4, y5);
v6 = _mm_sub_epi16(w7, y6);
v7 = _mm_add_epi16(w7, y6);
u0 = _mm_unpacklo_epi16(v0, v7);
u1 = _mm_unpackhi_epi16(v0, v7);
u2 = _mm_unpacklo_epi16(v1, v6);
u3 = _mm_unpackhi_epi16(v1, v6);
u4 = _mm_unpacklo_epi16(v2, v5);
u5 = _mm_unpackhi_epi16(v2, v5);
u6 = _mm_unpacklo_epi16(v3, v4);
u7 = _mm_unpackhi_epi16(v3, v4);
s0 = _mm_madd_epi16(u0, k__cospi_m30_p02); // x0
s1 = _mm_madd_epi16(u1, k__cospi_m30_p02);
s2 = _mm_madd_epi16(u2, k__cospi_m14_p18); // x1
s3 = _mm_madd_epi16(u3, k__cospi_m14_p18);
s4 = _mm_madd_epi16(u4, k__cospi_m22_p10); // x2
s5 = _mm_madd_epi16(u5, k__cospi_m22_p10);
s6 = _mm_madd_epi16(u6, k__cospi_m06_p26); // x3
s7 = _mm_madd_epi16(u7, k__cospi_m06_p26);
w0 = _mm_madd_epi16(u6, k__cospi_p26_p06); // x4
w1 = _mm_madd_epi16(u7, k__cospi_p26_p06);
w2 = _mm_madd_epi16(u4, k__cospi_p10_p22); // x5
w3 = _mm_madd_epi16(u5, k__cospi_p10_p22);
w4 = _mm_madd_epi16(u2, k__cospi_p18_p14); // x6
w5 = _mm_madd_epi16(u3, k__cospi_p18_p14);
w6 = _mm_madd_epi16(u0, k__cospi_p02_p30); // x7
w7 = _mm_madd_epi16(u1, k__cospi_p02_p30);
v0 = _mm_add_epi32(s0, k__DCT_CONST_ROUNDING);
v1 = _mm_add_epi32(s1, k__DCT_CONST_ROUNDING);
v2 = _mm_add_epi32(s2, k__DCT_CONST_ROUNDING);
v3 = _mm_add_epi32(s3, k__DCT_CONST_ROUNDING);
v4 = _mm_add_epi32(s4, k__DCT_CONST_ROUNDING);
v5 = _mm_add_epi32(s5, k__DCT_CONST_ROUNDING);
v6 = _mm_add_epi32(s6, k__DCT_CONST_ROUNDING);
v7 = _mm_add_epi32(s7, k__DCT_CONST_ROUNDING);
y0 = _mm_add_epi32(w0, k__DCT_CONST_ROUNDING);
y1 = _mm_add_epi32(w1, k__DCT_CONST_ROUNDING);
y2 = _mm_add_epi32(w2, k__DCT_CONST_ROUNDING);
y3 = _mm_add_epi32(w3, k__DCT_CONST_ROUNDING);
y4 = _mm_add_epi32(w4, k__DCT_CONST_ROUNDING);
y5 = _mm_add_epi32(w5, k__DCT_CONST_ROUNDING);
y6 = _mm_add_epi32(w6, k__DCT_CONST_ROUNDING);
y7 = _mm_add_epi32(w7, k__DCT_CONST_ROUNDING);
u0 = _mm_srai_epi32(v0, DCT_CONST_BITS);
u1 = _mm_srai_epi32(v1, DCT_CONST_BITS);
u2 = _mm_srai_epi32(v2, DCT_CONST_BITS);
u3 = _mm_srai_epi32(v3, DCT_CONST_BITS);
u4 = _mm_srai_epi32(v4, DCT_CONST_BITS);
u5 = _mm_srai_epi32(v5, DCT_CONST_BITS);
u6 = _mm_srai_epi32(v6, DCT_CONST_BITS);
u7 = _mm_srai_epi32(v7, DCT_CONST_BITS);
s0 = _mm_srai_epi32(y0, DCT_CONST_BITS);
s1 = _mm_srai_epi32(y1, DCT_CONST_BITS);
s2 = _mm_srai_epi32(y2, DCT_CONST_BITS);
s3 = _mm_srai_epi32(y3, DCT_CONST_BITS);
s4 = _mm_srai_epi32(y4, DCT_CONST_BITS);
s5 = _mm_srai_epi32(y5, DCT_CONST_BITS);
s6 = _mm_srai_epi32(y6, DCT_CONST_BITS);
s7 = _mm_srai_epi32(y7, DCT_CONST_BITS);
in[14] = _mm_packs_epi32(u0, u1);
in[6] = _mm_packs_epi32(u2, u3);
in[10] = _mm_packs_epi32(u4, u5);
in[2] = _mm_packs_epi32(u6, u7);
in[12] = _mm_packs_epi32(s0, s1);
in[4] = _mm_packs_epi32(s2, s3);
in[8] = _mm_packs_epi32(s4, s5);
in[0] = _mm_packs_epi32(s6, s7);
}
#endif // CONFIG_EXT_TX
static void fdct16_sse2(__m128i *in0, __m128i *in1) {
fdct16_8col(in0);
fdct16_8col(in1);
@@ -2777,14 +2156,6 @@ static void fadst16_sse2(__m128i *in0, __m128i *in1) {
array_transpose_16x16(in0, in1);
}
#if CONFIG_EXT_TX
static void fdst16_sse2(__m128i *in0, __m128i *in1) {
fdst16_8col(in0);
fdst16_8col(in1);
array_transpose_16x16(in0, in1);
}
#endif // CONFIG_EXT_TX
void vp10_fht16x16_sse2(const int16_t *input, tran_low_t *output,
int stride, int tx_type) {
__m128i in0[16], in1[16];
@@ -2850,55 +2221,6 @@ void vp10_fht16x16_sse2(const int16_t *input, tran_low_t *output,
fadst16_sse2(in0, in1);
write_buffer_16x16(output, in0, in1, 16);
break;
case DST_DST:
load_buffer_16x16(input, in0, in1, stride, 0, 0);
fdst16_sse2(in0, in1);
right_shift_16x16(in0, in1);
fdst16_sse2(in0, in1);
write_buffer_16x16(output, in0, in1, 16);
break;
case DCT_DST:
load_buffer_16x16(input, in0, in1, stride, 0, 0);
fdct16_sse2(in0, in1);
right_shift_16x16(in0, in1);
fdst16_sse2(in0, in1);
write_buffer_16x16(output, in0, in1, 16);
break;
case DST_DCT:
load_buffer_16x16(input, in0, in1, stride, 0, 0);
fdst16_sse2(in0, in1);
right_shift_16x16(in0, in1);
fdct16_sse2(in0, in1);
write_buffer_16x16(output, in0, in1, 16);
break;
case DST_ADST:
load_buffer_16x16(input, in0, in1, stride, 0, 0);
fdst16_sse2(in0, in1);
right_shift_16x16(in0, in1);
fadst16_sse2(in0, in1);
write_buffer_16x16(output, in0, in1, 16);
break;
case ADST_DST:
load_buffer_16x16(input, in0, in1, stride, 0, 0);
fadst16_sse2(in0, in1);
right_shift_16x16(in0, in1);
fdst16_sse2(in0, in1);
write_buffer_16x16(output, in0, in1, 16);
break;
case DST_FLIPADST:
load_buffer_16x16(input, in0, in1, stride, 0, 1);
fdst16_sse2(in0, in1);
right_shift_16x16(in0, in1);
fadst16_sse2(in0, in1);
write_buffer_16x16(output, in0, in1, 16);
break;
case FLIPADST_DST:
load_buffer_16x16(input, in0, in1, stride, 1, 0);
fadst16_sse2(in0, in1);
right_shift_16x16(in0, in1);
fdst16_sse2(in0, in1);
write_buffer_16x16(output, in0, in1, 16);
break;
#endif // CONFIG_EXT_TX
default:
assert(0);