Commit f5795554 authored by Debargha Mukherjee's avatar Debargha Mukherjee Committed by Gerrit Code Review
Browse files

Merge "Minor transform code cleanup" into nextgenv2

parents ff4e315d a1ddae59
......@@ -11,14 +11,14 @@
#include <math.h>
#include "./av1_rtcd.h"
#include "./aom_dsp_rtcd.h"
#include "./av1_rtcd.h"
#include "aom_dsp/inv_txfm.h"
#include "aom_ports/mem.h"
#include "av1/common/av1_inv_txfm2d_cfg.h"
#include "av1/common/blockd.h"
#include "av1/common/enums.h"
#include "av1/common/idct.h"
#include "av1/common/av1_inv_txfm2d_cfg.h"
#include "aom_dsp/inv_txfm.h"
#include "aom_ports/mem.h"
int get_tx_scale(const MACROBLOCKD *const xd, const TX_TYPE tx_type,
const TX_SIZE tx_size) {
......@@ -179,249 +179,6 @@ static void maybe_flip_strides(uint8_t **dst, int *dstride, tran_low_t **src,
}
#if CONFIG_AOM_HIGHBITDEPTH
void highbd_idst4_c(const tran_low_t *input, tran_low_t *output, int bd) {
tran_low_t step[4];
tran_high_t temp1, temp2;
(void)bd;
// stage 1
temp1 = (input[3] + input[1]) * cospi_16_64;
temp2 = (input[3] - input[1]) * cospi_16_64;
step[0] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
step[1] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = input[2] * cospi_24_64 - input[0] * cospi_8_64;
temp2 = input[2] * cospi_8_64 + input[0] * cospi_24_64;
step[2] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
step[3] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
// stage 2
output[0] = HIGHBD_WRAPLOW(step[0] + step[3], bd);
output[1] = HIGHBD_WRAPLOW(-step[1] - step[2], bd);
output[2] = HIGHBD_WRAPLOW(step[1] - step[2], bd);
output[3] = HIGHBD_WRAPLOW(step[3] - step[0], bd);
}
void highbd_idst8_c(const tran_low_t *input, tran_low_t *output, int bd) {
tran_low_t step1[8], step2[8];
tran_high_t temp1, temp2;
(void)bd;
// 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] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
step1[7] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = input[2] * cospi_12_64 - input[4] * cospi_20_64;
temp2 = input[2] * cospi_20_64 + input[4] * cospi_12_64;
step1[5] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
step1[6] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
// stage 2
temp1 = (step1[0] + step1[2]) * cospi_16_64;
temp2 = (step1[0] - step1[2]) * cospi_16_64;
step2[0] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
step2[1] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = step1[1] * cospi_24_64 - step1[3] * cospi_8_64;
temp2 = step1[1] * cospi_8_64 + step1[3] * cospi_24_64;
step2[2] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
step2[3] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
step2[4] = HIGHBD_WRAPLOW(step1[4] + step1[5], bd);
step2[5] = HIGHBD_WRAPLOW(step1[4] - step1[5], bd);
step2[6] = HIGHBD_WRAPLOW(-step1[6] + step1[7], bd);
step2[7] = HIGHBD_WRAPLOW(step1[6] + step1[7], bd);
// stage 3
step1[0] = HIGHBD_WRAPLOW(step2[0] + step2[3], bd);
step1[1] = HIGHBD_WRAPLOW(step2[1] + step2[2], bd);
step1[2] = HIGHBD_WRAPLOW(step2[1] - step2[2], bd);
step1[3] = HIGHBD_WRAPLOW(step2[0] - step2[3], bd);
step1[4] = step2[4];
temp1 = (step2[6] - step2[5]) * cospi_16_64;
temp2 = (step2[5] + step2[6]) * cospi_16_64;
step1[5] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
step1[6] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
step1[7] = step2[7];
// stage 4
output[0] = HIGHBD_WRAPLOW(step1[0] + step1[7], bd);
output[1] = HIGHBD_WRAPLOW(-step1[1] - step1[6], bd);
output[2] = HIGHBD_WRAPLOW(step1[2] + step1[5], bd);
output[3] = HIGHBD_WRAPLOW(-step1[3] - step1[4], bd);
output[4] = HIGHBD_WRAPLOW(step1[3] - step1[4], bd);
output[5] = HIGHBD_WRAPLOW(-step1[2] + step1[5], bd);
output[6] = HIGHBD_WRAPLOW(step1[1] - step1[6], bd);
output[7] = HIGHBD_WRAPLOW(-step1[0] + step1[7], bd);
}
void highbd_idst16_c(const tran_low_t *input, tran_low_t *output, int bd) {
// av1_highbd_igentx16(input, output, bd, Tx16);
tran_low_t step1[16], step2[16];
tran_high_t temp1, temp2;
(void)bd;
// 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] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
step2[15] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = step1[9] * cospi_14_64 - step1[14] * cospi_18_64;
temp2 = step1[9] * cospi_18_64 + step1[14] * cospi_14_64;
step2[9] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
step2[14] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = step1[10] * cospi_22_64 - step1[13] * cospi_10_64;
temp2 = step1[10] * cospi_10_64 + step1[13] * cospi_22_64;
step2[10] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
step2[13] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = step1[11] * cospi_6_64 - step1[12] * cospi_26_64;
temp2 = step1[11] * cospi_26_64 + step1[12] * cospi_6_64;
step2[11] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
step2[12] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
// 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] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
step1[7] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = step2[5] * cospi_12_64 - step2[6] * cospi_20_64;
temp2 = step2[5] * cospi_20_64 + step2[6] * cospi_12_64;
step1[5] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
step1[6] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
step1[8] = HIGHBD_WRAPLOW(step2[8] + step2[9], bd);
step1[9] = HIGHBD_WRAPLOW(step2[8] - step2[9], bd);
step1[10] = HIGHBD_WRAPLOW(-step2[10] + step2[11], bd);
step1[11] = HIGHBD_WRAPLOW(step2[10] + step2[11], bd);
step1[12] = HIGHBD_WRAPLOW(step2[12] + step2[13], bd);
step1[13] = HIGHBD_WRAPLOW(step2[12] - step2[13], bd);
step1[14] = HIGHBD_WRAPLOW(-step2[14] + step2[15], bd);
step1[15] = HIGHBD_WRAPLOW(step2[14] + step2[15], bd);
// stage 4
temp1 = (step1[0] + step1[1]) * cospi_16_64;
temp2 = (step1[0] - step1[1]) * cospi_16_64;
step2[0] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
step2[1] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = step1[2] * cospi_24_64 - step1[3] * cospi_8_64;
temp2 = step1[2] * cospi_8_64 + step1[3] * cospi_24_64;
step2[2] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
step2[3] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
step2[4] = HIGHBD_WRAPLOW(step1[4] + step1[5], bd);
step2[5] = HIGHBD_WRAPLOW(step1[4] - step1[5], bd);
step2[6] = HIGHBD_WRAPLOW(-step1[6] + step1[7], bd);
step2[7] = HIGHBD_WRAPLOW(step1[6] + step1[7], bd);
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] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
step2[14] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = -step1[10] * cospi_24_64 - step1[13] * cospi_8_64;
temp2 = -step1[10] * cospi_8_64 + step1[13] * cospi_24_64;
step2[10] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
step2[13] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
step2[11] = step1[11];
step2[12] = step1[12];
// stage 5
step1[0] = HIGHBD_WRAPLOW(step2[0] + step2[3], bd);
step1[1] = HIGHBD_WRAPLOW(step2[1] + step2[2], bd);
step1[2] = HIGHBD_WRAPLOW(step2[1] - step2[2], bd);
step1[3] = HIGHBD_WRAPLOW(step2[0] - step2[3], bd);
step1[4] = step2[4];
temp1 = (step2[6] - step2[5]) * cospi_16_64;
temp2 = (step2[5] + step2[6]) * cospi_16_64;
step1[5] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
step1[6] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
step1[7] = step2[7];
step1[8] = HIGHBD_WRAPLOW(step2[8] + step2[11], bd);
step1[9] = HIGHBD_WRAPLOW(step2[9] + step2[10], bd);
step1[10] = HIGHBD_WRAPLOW(step2[9] - step2[10], bd);
step1[11] = HIGHBD_WRAPLOW(step2[8] - step2[11], bd);
step1[12] = HIGHBD_WRAPLOW(-step2[12] + step2[15], bd);
step1[13] = HIGHBD_WRAPLOW(-step2[13] + step2[14], bd);
step1[14] = HIGHBD_WRAPLOW(step2[13] + step2[14], bd);
step1[15] = HIGHBD_WRAPLOW(step2[12] + step2[15], bd);
// stage 6
step2[0] = HIGHBD_WRAPLOW(step1[0] + step1[7], bd);
step2[1] = HIGHBD_WRAPLOW(step1[1] + step1[6], bd);
step2[2] = HIGHBD_WRAPLOW(step1[2] + step1[5], bd);
step2[3] = HIGHBD_WRAPLOW(step1[3] + step1[4], bd);
step2[4] = HIGHBD_WRAPLOW(step1[3] - step1[4], bd);
step2[5] = HIGHBD_WRAPLOW(step1[2] - step1[5], bd);
step2[6] = HIGHBD_WRAPLOW(step1[1] - step1[6], bd);
step2[7] = HIGHBD_WRAPLOW(step1[0] - step1[7], bd);
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] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
step2[13] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
temp1 = (-step1[11] + step1[12]) * cospi_16_64;
temp2 = (step1[11] + step1[12]) * cospi_16_64;
step2[11] = HIGHBD_WRAPLOW(dct_const_round_shift(temp1), bd);
step2[12] = HIGHBD_WRAPLOW(dct_const_round_shift(temp2), bd);
step2[14] = step1[14];
step2[15] = step1[15];
// stage 7
output[0] = HIGHBD_WRAPLOW(step2[0] + step2[15], bd);
output[1] = HIGHBD_WRAPLOW(-step2[1] - step2[14], bd);
output[2] = HIGHBD_WRAPLOW(step2[2] + step2[13], bd);
output[3] = HIGHBD_WRAPLOW(-step2[3] - step2[12], bd);
output[4] = HIGHBD_WRAPLOW(step2[4] + step2[11], bd);
output[5] = HIGHBD_WRAPLOW(-step2[5] - step2[10], bd);
output[6] = HIGHBD_WRAPLOW(step2[6] + step2[9], bd);
output[7] = HIGHBD_WRAPLOW(-step2[7] - step2[8], bd);
output[8] = HIGHBD_WRAPLOW(step2[7] - step2[8], bd);
output[9] = HIGHBD_WRAPLOW(-step2[6] + step2[9], bd);
output[10] = HIGHBD_WRAPLOW(step2[5] - step2[10], bd);
output[11] = HIGHBD_WRAPLOW(-step2[4] + step2[11], bd);
output[12] = HIGHBD_WRAPLOW(step2[3] - step2[12], bd);
output[13] = HIGHBD_WRAPLOW(-step2[2] + step2[13], bd);
output[14] = HIGHBD_WRAPLOW(step2[1] - step2[14], bd);
output[15] = HIGHBD_WRAPLOW(-step2[0] + step2[15], bd);
}
static void highbd_inv_idtx_add_c(const tran_low_t *input, uint8_t *dest8,
int stride, int bs, int tx_type, int bd) {
int r, c;
......
......@@ -12,13 +12,13 @@
#include <assert.h>
#include <math.h>
#include "./av1_rtcd.h"
#include "./aom_config.h"
#include "./aom_dsp_rtcd.h"
#include "av1/common/blockd.h"
#include "av1/common/idct.h"
#include "./av1_rtcd.h"
#include "aom_dsp/fwd_txfm.h"
#include "aom_ports/mem.h"
#include "av1/common/blockd.h"
#include "av1/common/idct.h"
static INLINE void range_check(const tran_low_t *input, const int size,
const int bit) {
......@@ -1132,7 +1132,12 @@ static void maybe_flip_input(const int16_t **src, int *src_stride, int l, int w,
}
#endif // CONFIG_EXT_TX
static const transform_2d FHT_4[] = {
void av1_fht4x4_c(const int16_t *input, tran_low_t *output, int stride,
int tx_type) {
if (tx_type == DCT_DCT) {
aom_fdct4x4_c(input, output, stride);
} else {
static const transform_2d FHT[] = {
{ fdct4, fdct4 }, // DCT_DCT
{ fadst4, fdct4 }, // ADST_DCT
{ fdct4, fadst4 }, // DCT_ADST
......@@ -1151,194 +1156,11 @@ static const transform_2d FHT_4[] = {
{ fadst4, fidtx4 }, // V_FLIPADST
{ fidtx4, fadst4 }, // H_FLIPADST
#endif // CONFIG_EXT_TX
};
static const transform_2d FHT_8[] = {
{ fdct8, fdct8 }, // DCT_DCT
{ fadst8, fdct8 }, // ADST_DCT
{ fdct8, fadst8 }, // DCT_ADST
{ fadst8, fadst8 }, // ADST_ADST
#if CONFIG_EXT_TX
{ 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
{ fadst16, fdct16 }, // ADST_DCT
{ fdct16, fadst16 }, // DCT_ADST
{ fadst16, fadst16 }, // ADST_ADST
#if CONFIG_EXT_TX
{ 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
{ 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
};
static const transform_2d FHT_4x8[] = {
{ fdct8, fdct4 }, // DCT_DCT
{ fadst8, fdct4 }, // ADST_DCT
{ fdct8, fadst4 }, // DCT_ADST
{ fadst8, fadst4 }, // ADST_ADST
{ fadst8, fdct4 }, // FLIPADST_DCT
{ fdct8, fadst4 }, // DCT_FLIPADST
{ fadst8, fadst4 }, // FLIPADST_FLIPADST
{ fadst8, fadst4 }, // ADST_FLIPADST
{ fadst8, fadst4 }, // FLIPADST_ADST
{ fidtx8, fidtx4 }, // IDTX
{ fdct8, fidtx4 }, // V_DCT
{ fidtx8, fdct4 }, // H_DCT
{ fadst8, fidtx4 }, // V_ADST
{ fidtx8, fadst4 }, // H_ADST
{ fadst8, fidtx4 }, // V_FLIPADST
{ fidtx8, fadst4 }, // H_FLIPADST
};
static const transform_2d FHT_8x4[] = {
{ fdct4, fdct8 }, // DCT_DCT
{ fadst4, fdct8 }, // ADST_DCT
{ fdct4, fadst8 }, // DCT_ADST
{ fadst4, fadst8 }, // ADST_ADST
{ fadst4, fdct8 }, // FLIPADST_DCT
{ fdct4, fadst8 }, // DCT_FLIPADST
{ fadst4, fadst8 }, // FLIPADST_FLIPADST
{ fadst4, fadst8 }, // ADST_FLIPADST
{ fadst4, fadst8 }, // FLIPADST_ADST
{ fidtx4, fidtx8 }, // IDTX
{ fdct4, fidtx8 }, // V_DCT
{ fidtx4, fdct8 }, // H_DCT
{ fadst4, fidtx8 }, // V_ADST
{ fidtx4, fadst8 }, // H_ADST
{ fadst4, fidtx8 }, // V_FLIPADST
{ fidtx4, fadst8 }, // H_FLIPADST
};
static const transform_2d FHT_8x16[] = {
{ fdct16, fdct8 }, // DCT_DCT
{ fadst16, fdct8 }, // ADST_DCT
{ fdct16, fadst8 }, // DCT_ADST
{ fadst16, fadst8 }, // ADST_ADST
{ fadst16, fdct8 }, // FLIPADST_DCT
{ fdct16, fadst8 }, // DCT_FLIPADST
{ fadst16, fadst8 }, // FLIPADST_FLIPADST
{ fadst16, fadst8 }, // ADST_FLIPADST
{ fadst16, fadst8 }, // FLIPADST_ADST
{ fidtx16, fidtx8 }, // IDTX
{ fdct16, fidtx8 }, // V_DCT
{ fidtx16, fdct8 }, // H_DCT
{ fadst16, fidtx8 }, // V_ADST
{ fidtx16, fadst8 }, // H_ADST
{ fadst16, fidtx8 }, // V_FLIPADST
{ fidtx16, fadst8 }, // H_FLIPADST
};
static const transform_2d FHT_16x8[] = {
{ fdct8, fdct16 }, // DCT_DCT
{ fadst8, fdct16 }, // ADST_DCT
{ fdct8, fadst16 }, // DCT_ADST
{ fadst8, fadst16 }, // ADST_ADST
{ fadst8, fdct16 }, // FLIPADST_DCT
{ fdct8, fadst16 }, // DCT_FLIPADST
{ fadst8, fadst16 }, // FLIPADST_FLIPADST
{ fadst8, fadst16 }, // ADST_FLIPADST
{ fadst8, fadst16 }, // FLIPADST_ADST
{ fidtx8, fidtx16 }, // IDTX
{ fdct8, fidtx16 }, // V_DCT
{ fidtx8, fdct16 }, // H_DCT
{ fadst8, fidtx16 }, // V_ADST
{ fidtx8, fadst16 }, // H_ADST
{ fadst8, fidtx16 }, // V_FLIPADST
{ fidtx8, fadst16 }, // H_FLIPADST
};
static const transform_2d FHT_16x32[] = {
{ fdct32, fdct16 }, // DCT_DCT
{ fhalfright32, fdct16 }, // ADST_DCT
{ fdct32, fadst16 }, // DCT_ADST
{ fhalfright32, fadst16 }, // ADST_ADST
{ fhalfright32, fdct16 }, // FLIPADST_DCT
{ fdct32, fadst16 }, // DCT_FLIPADST
{ fhalfright32, fadst16 }, // FLIPADST_FLIPADST
{ fhalfright32, fadst16 }, // ADST_FLIPADST
{ fhalfright32, fadst16 }, // FLIPADST_ADST
{ fidtx32, fidtx16 }, // IDTX
{ fdct32, fidtx16 }, // V_DCT
{ fidtx32, fdct16 }, // H_DCT
{ fhalfright32, fidtx16 }, // V_ADST
{ fidtx32, fadst16 }, // H_ADST
{ fhalfright32, fidtx16 }, // V_FLIPADST
{ fidtx32, fadst16 }, // H_FLIPADST
};
static const transform_2d FHT_32x16[] = {
{ fdct16, fdct32 }, // DCT_DCT
{ fadst16, fdct32 }, // ADST_DCT
{ fdct16, fhalfright32 }, // DCT_ADST
{ fadst16, fhalfright32 }, // ADST_ADST
{ fadst16, fdct32 }, // FLIPADST_DCT
{ fdct16, fhalfright32 }, // DCT_FLIPADST
{ fadst16, fhalfright32 }, // FLIPADST_FLIPADST
{ fadst16, fhalfright32 }, // ADST_FLIPADST
{ fadst16, fhalfright32 }, // FLIPADST_ADST
{ fidtx16, fidtx32 }, // IDTX
{ fdct16, fidtx32 }, // V_DCT
{ fidtx16, fdct32 }, // H_DCT
{ fadst16, fidtx32 }, // V_ADST
{ fidtx16, fhalfright32 }, // H_ADST
{ fadst16, fidtx32 }, // V_FLIPADST
{ fidtx16, fhalfright32 }, // H_FLIPADST
};
#endif // CONFIG_EXT_TX
void av1_fht4x4_c(const int16_t *input, tran_low_t *output, int stride,
int tx_type) {
if (tx_type == DCT_DCT) {
aom_fdct4x4_c(input, output, stride);
} else {
};
const transform_2d ht = FHT[tx_type];
tran_low_t out[4 * 4];
int i, j;
tran_low_t temp_in[4], temp_out[4];
const transform_2d ht = FHT_4[tx_type];
#if CONFIG_EXT_TX
int16_t flipped_input[4 * 4];
......@@ -1365,12 +1187,30 @@ void av1_fht4x4_c(const int16_t *input, tran_low_t *output, int stride,
#if CONFIG_EXT_TX
void av1_fht4x8_c(const int16_t *input, tran_low_t *output, int stride,
int tx_type) {
static const transform_2d FHT[] = {
{ fdct8, fdct4 }, // DCT_DCT
{ fadst8, fdct4 }, // ADST_DCT
{ fdct8, fadst4 }, // DCT_ADST
{ fadst8, fadst4 }, // ADST_ADST
{ fadst8, fdct4 }, // FLIPADST_DCT
{ fdct8, fadst4 }, // DCT_FLIPADST
{ fadst8, fadst4 }, // FLIPADST_FLIPADST
{ fadst8, fadst4 }, // ADST_FLIPADST
{ fadst8, fadst4 }, // FLIPADST_ADST
{ fidtx8, fidtx4 }, // IDTX
{ fdct8, fidtx4 }, // V_DCT
{ fidtx8, fdct4 }, // H_DCT
{ fadst8, fidtx4 }, // V_ADST
{ fidtx8, fadst4 }, // H_ADST
{ fadst8, fidtx4 }, // V_FLIPADST
{ fidtx8, fadst4 }, // H_FLIPADST
};
const transform_2d ht = FHT[tx_type];
const int n = 4;
const int n2 = 8;
tran_low_t out[8 * 4];
tran_low_t temp_in[8], temp_out[8];
int i, j;
const transform_2d ht = FHT_4x8[tx_type];
int16_t flipped_input[8 * 4];
maybe_flip_input(&input, &stride, n2, n, flipped_input, tx_type);
......@@ -1394,12 +1234,30 @@ void av1_fht4x8_c(const int16_t *input, tran_low_t *output, int stride,
void av1_fht8x4_c(const int16_t *input, tran_low_t *output, int stride,
int tx_type) {
static const transform_2d FHT[] = {
{ fdct4, fdct8 }, // DCT_DCT
{ fadst4, fdct8 }, // ADST_DCT
{ fdct4, fadst8 }, // DCT_ADST
{ fadst4, fadst8 }, // ADST_ADST
{ fadst4, fdct8 }, // FLIPADST_DCT
{ fdct4, fadst8 }, // DCT_FLIPADST
{ fadst4, fadst8 }, // FLIPADST_FLIPADST
{ fadst4, fadst8 }, // ADST_FLIPADST
{ fadst4, fadst8 }, // FLIPADST_ADST
{ fidtx4, fidtx8 }, // IDTX
{ fdct4, fidtx8 }, // V_DCT
{ fidtx4, fdct8 }, // H_DCT
{ fadst4, fidtx8 }, // V_ADST
{ fidtx4, fadst8 }, // H_ADST
{ fadst4, fidtx8 }, // V_FLIPADST
{ fidtx4, fadst8 }, // H_FLIPADST
};
const transform_2d ht = FHT[tx_type];
const int n = 4;
const int n2 = 8;
tran_low_t out[8 * 4];
tran_low_t temp_in[8], temp_out[8];
int i, j;
const transform_2d ht = FHT_8x4[tx_type];
int16_t flipped_input[8 * 4];
maybe_flip_input(&input, &stride, n, n2, flipped_input, tx_type);
......@@ -1423,12 +1281,30 @@ void av1_fht8x4_c(const int16_t *input, tran_low_t *output, int stride,
void av1_fht8x16_c(const int16_t *input, tran_low_t *output, int stride,
int tx_type) {
static const transform_2d FHT[] = {
{ fdct16, fdct8 }, // DCT_DCT
{ fadst16, fdct8 }, // ADST_DCT
{ fdct16, fadst8 }, // DCT_ADST
{ fadst16, fadst8 }, // ADST_ADST
{ fadst16, fdct8 }, // FLIPADST_DCT
{ fdct16, fadst8 }, // DCT_FLIPADST
{ fadst16, fadst8 }, // FLIPADST_FLIPADST
{ fadst16, fadst8 }, // ADST_FLIPADST
{ fadst16, fadst8 }, // FLIPADST_ADST
{ fidtx16, fidtx8 }, // IDTX
{ fdct16, fidtx8 }, // V_DCT
{ fidtx16, fdct8 }, // H_DCT
{ fadst16, fidtx8 }, // V_ADST
{ fidtx16, fadst8 }, // H_ADST
{ fadst16, fidtx8 }, // V_FLIPADST
{ fidtx16, fadst8 }, // H_FLIPADST
};
const transform_2d ht = FHT[tx_type];
const int n = 8;
const int n2 = 16;
tran_low_t out[16 * 8];