void vpx_idct4x4_1_add_neon(
int16_t *input,
uint8_t *dest,
int dest_stride) {
uint8x8_t d6u8;
uint32x2_t d2u32 = vdup_n_u32(0);
uint16x8_t q8u16;
int16x8_t q0s16;
uint8_t *d1, *d2;
int16_t i, a1, cospi_16_64 = 11585;
int16_t out = dct_const_round_shift(input[0] * cospi_16_64);
out = dct_const_round_shift(out * cospi_16_64);
a1 = ROUND_POWER_OF_TWO(out, 4);
q0s16 = vdupq_n_s16(a1);
// dc_only_idct_add
d1 = d2 = dest;
for (i = 0; i < 2; i++) {
d2u32 = vld1_lane_u32((const uint32_t *)d1, d2u32, 0);
d1 += dest_stride;
d2u32 = vld1_lane_u32((const uint32_t *)d1, d2u32, 1);
d1 += dest_stride;
q8u16 = vaddw_u8(vreinterpretq_u16_s16(q0s16),
vreinterpret_u8_u32(d2u32));
d6u8 = vqmovun_s16(vreinterpretq_s16_u16(q8u16));
vst1_lane_u32((uint32_t *)d2, vreinterpret_u32_u8(d6u8), 0);
d2 += dest_stride;
vst1_lane_u32((uint32_t *)d2, vreinterpret_u32_u8(d6u8), 1);
d2 += dest_stride;
}
return;
}
void vpx_highbd_idct32x32_1_add_neon(const tran_low_t *input, uint16_t *dest,
int stride, int bd) {
const tran_low_t out0 = HIGHBD_WRAPLOW(
dct_const_round_shift(input[0] * (tran_high_t)cospi_16_64), bd);
const tran_low_t out1 = HIGHBD_WRAPLOW(
dct_const_round_shift(out0 * (tran_high_t)cospi_16_64), bd);
const int16_t a1 = ROUND_POWER_OF_TWO(out1, 6);
const int16x8_t dc = vdupq_n_s16(a1);
int i;
if (a1 >= 0) {
const int16x8_t max = vdupq_n_s16((1 << bd) - 1);
for (i = 0; i < 8; ++i) {
highbd_idct32x32_1_add_pos_kernel(&dest, stride, dc, max);
highbd_idct32x32_1_add_pos_kernel(&dest, stride, dc, max);
highbd_idct32x32_1_add_pos_kernel(&dest, stride, dc, max);
highbd_idct32x32_1_add_pos_kernel(&dest, stride, dc, max);
}
} else {
for (i = 0; i < 8; ++i) {
highbd_idct32x32_1_add_neg_kernel(&dest, stride, dc);
highbd_idct32x32_1_add_neg_kernel(&dest, stride, dc);
highbd_idct32x32_1_add_neg_kernel(&dest, stride, dc);
highbd_idct32x32_1_add_neg_kernel(&dest, stride, dc);
}
}
}
void vp9_idct16x16_1_add_c(const int16_t *input, uint8_t *dest, int stride) {
int i, j;
int a1;
int16_t out = dct_const_round_shift(input[0] * cospi_16_64);
out = dct_const_round_shift(out * cospi_16_64);
a1 = ROUND_POWER_OF_TWO(out, 6);
for (j = 0; j < 16; ++j) {
for (i = 0; i < 16; ++i)
dest[i] = clip_pixel(dest[i] + a1);
dest += stride;
}
}
void vp9_short_idct8x8_1_add_c(int16_t *input, uint8_t *dest, int dest_stride) {
int i, j;
int a1;
int16_t out = dct_const_round_shift(input[0] * cospi_16_64);
out = dct_const_round_shift(out * cospi_16_64);
a1 = ROUND_POWER_OF_TWO(out, 5);
for (j = 0; j < 8; ++j) {
for (i = 0; i < 8; ++i)
dest[i] = clip_pixel(dest[i] + a1);
dest += dest_stride;
}
}
void vpx_highbd_idct4x4_1_add_neon(const tran_low_t *input, uint16_t *dest,
int stride, int bd) {
const int16x8_t max = vdupq_n_s16((1 << bd) - 1);
const tran_low_t out0 = HIGHBD_WRAPLOW(
dct_const_round_shift(input[0] * (tran_high_t)cospi_16_64), bd);
const tran_low_t out1 = HIGHBD_WRAPLOW(
dct_const_round_shift(out0 * (tran_high_t)cospi_16_64), bd);
const int16_t a1 = ROUND_POWER_OF_TWO(out1, 4);
const int16x8_t dc = vdupq_n_s16(a1);
highbd_idct4x4_1_add_kernel1(&dest, stride, dc, max);
highbd_idct4x4_1_add_kernel1(&dest, stride, dc, max);
}
void vpx_idct4x4_1_add_neon(const tran_low_t *input, uint8_t *dest,
int stride) {
const int16_t out0 = WRAPLOW(dct_const_round_shift(input[0] * cospi_16_64));
const int16_t out1 = WRAPLOW(dct_const_round_shift(out0 * cospi_16_64));
const int16_t a1 = ROUND_POWER_OF_TWO(out1, 4);
const int16x8_t dc = vdupq_n_s16(a1);
uint32x2_t d = vdup_n_u32(0);
assert(!((intptr_t)dest % sizeof(uint32_t)));
assert(!(stride % sizeof(uint32_t)));
idct4x4_1_add_kernel(&dest, stride, dc, &d);
idct4x4_1_add_kernel(&dest, stride, dc, &d);
}
void vp9_idct8x8_1_add_neon(
int16_t *input,
uint8_t *dest,
int dest_stride) {
uint8x8_t d2u8, d3u8, d30u8, d31u8;
uint64x1_t d2u64, d3u64, d4u64, d5u64;
uint16x8_t q0u16, q9u16, q10u16, q11u16, q12u16;
int16x8_t q0s16;
uint8_t *d1, *d2;
int16_t i, a1, cospi_16_64 = 11585;
int16_t out = dct_const_round_shift(input[0] * cospi_16_64);
out = dct_const_round_shift(out * cospi_16_64);
a1 = ROUND_POWER_OF_TWO(out, 5);
q0s16 = vdupq_n_s16(a1);
q0u16 = vreinterpretq_u16_s16(q0s16);
d1 = d2 = dest;
for (i = 0; i < 2; i++) {
d2u64 = vld1_u64((const uint64_t *)d1);
d1 += dest_stride;
d3u64 = vld1_u64((const uint64_t *)d1);
d1 += dest_stride;
d4u64 = vld1_u64((const uint64_t *)d1);
d1 += dest_stride;
d5u64 = vld1_u64((const uint64_t *)d1);
d1 += dest_stride;
q9u16 = vaddw_u8(q0u16, vreinterpret_u8_u64(d2u64));
q10u16 = vaddw_u8(q0u16, vreinterpret_u8_u64(d3u64));
q11u16 = vaddw_u8(q0u16, vreinterpret_u8_u64(d4u64));
q12u16 = vaddw_u8(q0u16, vreinterpret_u8_u64(d5u64));
d2u8 = vqmovun_s16(vreinterpretq_s16_u16(q9u16));
d3u8 = vqmovun_s16(vreinterpretq_s16_u16(q10u16));
d30u8 = vqmovun_s16(vreinterpretq_s16_u16(q11u16));
d31u8 = vqmovun_s16(vreinterpretq_s16_u16(q12u16));
vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d2u8));
d2 += dest_stride;
vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d3u8));
d2 += dest_stride;
vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d30u8));
d2 += dest_stride;
vst1_u64((uint64_t *)d2, vreinterpret_u64_u8(d31u8));
d2 += dest_stride;
}
return;
}
void vp9_short_idct4x4_1_add_c(int16_t *input, uint8_t *dest, int dest_stride) {
int i;
int a1;
int16_t out = dct_const_round_shift(input[0] * cospi_16_64);
out = dct_const_round_shift(out * cospi_16_64);
a1 = ROUND_POWER_OF_TWO(out, 4);
for (i = 0; i < 4; i++) {
dest[0] = clip_pixel(dest[0] + a1);
dest[1] = clip_pixel(dest[1] + a1);
dest[2] = clip_pixel(dest[2] + a1);
dest[3] = clip_pixel(dest[3] + a1);
dest += dest_stride;
}
}
void aom_idct32x32_1_add_neon(int16_t *input, uint8_t *dest, int dest_stride) {
uint8x16_t q0u8, q8u8, q9u8, q10u8, q11u8, q12u8, q13u8, q14u8, q15u8;
int i, j, dest_stride8;
uint8_t *d;
int16_t a1, cospi_16_64 = 11585;
int16_t out = dct_const_round_shift(input[0] * cospi_16_64);
out = dct_const_round_shift(out * cospi_16_64);
a1 = ROUND_POWER_OF_TWO(out, 6);
dest_stride8 = dest_stride * 8;
if (a1 >= 0) { // diff_positive_32_32
a1 = a1 < 0 ? 0 : a1 > 255 ? 255 : a1;
q0u8 = vdupq_n_u8(a1);
for (i = 0; i < 2; i++, dest += 16) { // diff_positive_32_32_loop
d = dest;
for (j = 0; j < 4; j++) {
LD_16x8(d, dest_stride, &q8u8, &q9u8, &q10u8, &q11u8, &q12u8, &q13u8,
&q14u8, &q15u8);
ADD_DIFF_16x8(q0u8, &q8u8, &q9u8, &q10u8, &q11u8, &q12u8, &q13u8,
&q14u8, &q15u8);
ST_16x8(d, dest_stride, &q8u8, &q9u8, &q10u8, &q11u8, &q12u8, &q13u8,
&q14u8, &q15u8);
d += dest_stride8;
}
}
} else { // diff_negative_32_32
a1 = -a1;
a1 = a1 < 0 ? 0 : a1 > 255 ? 255 : a1;
q0u8 = vdupq_n_u8(a1);
for (i = 0; i < 2; i++, dest += 16) { // diff_negative_32_32_loop
d = dest;
for (j = 0; j < 4; j++) {
LD_16x8(d, dest_stride, &q8u8, &q9u8, &q10u8, &q11u8, &q12u8, &q13u8,
&q14u8, &q15u8);
SUB_DIFF_16x8(q0u8, &q8u8, &q9u8, &q10u8, &q11u8, &q12u8, &q13u8,
&q14u8, &q15u8);
ST_16x8(d, dest_stride, &q8u8, &q9u8, &q10u8, &q11u8, &q12u8, &q13u8,
&q14u8, &q15u8);
d += dest_stride8;
}
}
}
return;
}
void vp9_idct4_1d_c(int16_t *input, int16_t *output) {
int16_t step[4];
int temp1, temp2;
// stage 1
temp1 = (input[0] + input[2]) * cospi_16_64;
temp2 = (input[0] - input[2]) * cospi_16_64;
step[0] = dct_const_round_shift(temp1);
step[1] = dct_const_round_shift(temp2);
temp1 = input[1] * cospi_24_64 - input[3] * cospi_8_64;
temp2 = input[1] * cospi_8_64 + input[3] * cospi_24_64;
step[2] = dct_const_round_shift(temp1);
step[3] = dct_const_round_shift(temp2);
// stage 2
output[0] = step[0] + step[3];
output[1] = step[1] + step[2];
output[2] = step[1] - step[2];
output[3] = step[0] - step[3];
}
void vpx_idct32x32_1_add_neon(const tran_low_t *input, uint8_t *dest,
int stride) {
int i;
const int16_t out0 =
WRAPLOW(dct_const_round_shift((int16_t)input[0] * cospi_16_64));
const int16_t out1 = WRAPLOW(dct_const_round_shift(out0 * cospi_16_64));
const int16_t a1 = ROUND_POWER_OF_TWO(out1, 6);
if (a1 >= 0) {
const uint8x16_t dc = create_dcq(a1);
for (i = 0; i < 32; i++) {
idct32x32_1_add_pos_kernel(&dest, stride, dc);
}
} else {
const uint8x16_t dc = create_dcq(-a1);
for (i = 0; i < 32; i++) {
idct32x32_1_add_neg_kernel(&dest, stride, dc);
}
}
}
static void iadst4_1d(int16_t *input, int16_t *output) {
int s0, s1, s2, s3, s4, s5, s6, s7;
int x0 = input[0];
int x1 = input[1];
int x2 = input[2];
int x3 = input[3];
if (!(x0 | x1 | x2 | x3)) {
output[0] = output[1] = output[2] = output[3] = 0;
return;
}
s0 = sinpi_1_9 * x0;
s1 = sinpi_2_9 * x0;
s2 = sinpi_3_9 * x1;
s3 = sinpi_4_9 * x2;
s4 = sinpi_1_9 * x2;
s5 = sinpi_2_9 * x3;
s6 = sinpi_4_9 * x3;
s7 = x0 - x2 + x3;
x0 = s0 + s3 + s5;
x1 = s1 - s4 - s6;
x2 = sinpi_3_9 * s7;
x3 = s2;
s0 = x0 + x3;
s1 = x1 + x3;
s2 = x2;
s3 = x0 + x1 - x3;
// 1-D transform scaling factor is sqrt(2).
// The overall dynamic range is 14b (input) + 14b (multiplication scaling)
// + 1b (addition) = 29b.
// Hence the output bit depth is 15b.
output[0] = dct_const_round_shift(s0);
output[1] = dct_const_round_shift(s1);
output[2] = dct_const_round_shift(s2);
output[3] = dct_const_round_shift(s3);
}
static void idct8_1d(int16_t *input, int16_t *output) {
int16_t step1[8], step2[8];
int temp1, temp2;
// stage 1
step1[0] = input[0];
step1[2] = input[4];
step1[1] = input[2];
step1[3] = input[6];
temp1 = input[1] * cospi_28_64 - input[7] * cospi_4_64;
temp2 = input[1] * cospi_4_64 + input[7] * cospi_28_64;
step1[4] = dct_const_round_shift(temp1);
step1[7] = dct_const_round_shift(temp2);
temp1 = input[5] * cospi_12_64 - input[3] * cospi_20_64;
temp2 = input[5] * cospi_20_64 + input[3] * cospi_12_64;
step1[5] = dct_const_round_shift(temp1);
step1[6] = dct_const_round_shift(temp2);
// stage 2 & stage 3 - even half
vp9_idct4_1d(step1, step1);
// stage 2 - odd half
step2[4] = step1[4] + step1[5];
step2[5] = step1[4] - step1[5];
step2[6] = -step1[6] + step1[7];
step2[7] = step1[6] + step1[7];
// stage 3 -odd half
step1[4] = step2[4];
temp1 = (step2[6] - step2[5]) * cospi_16_64;
temp2 = (step2[5] + step2[6]) * cospi_16_64;
step1[5] = dct_const_round_shift(temp1);
step1[6] = dct_const_round_shift(temp2);
step1[7] = step2[7];
// stage 4
output[0] = step1[0] + step1[7];
output[1] = step1[1] + step1[6];
output[2] = step1[2] + step1[5];
output[3] = step1[3] + step1[4];
output[4] = step1[3] - step1[4];
output[5] = step1[2] - step1[5];
output[6] = step1[1] - step1[6];
output[7] = step1[0] - step1[7];
}
static void idct32_1d(int16_t *input, int16_t *output) {
int16_t step1[32], step2[32];
int temp1, temp2;
// stage 1
step1[0] = input[0];
step1[1] = input[16];
step1[2] = input[8];
step1[3] = input[24];
step1[4] = input[4];
step1[5] = input[20];
step1[6] = input[12];
step1[7] = input[28];
step1[8] = input[2];
step1[9] = input[18];
step1[10] = input[10];
step1[11] = input[26];
step1[12] = input[6];
step1[13] = input[22];
step1[14] = input[14];
step1[15] = input[30];
temp1 = input[1] * cospi_31_64 - input[31] * cospi_1_64;
temp2 = input[1] * cospi_1_64 + input[31] * cospi_31_64;
step1[16] = dct_const_round_shift(temp1);
step1[31] = dct_const_round_shift(temp2);
temp1 = input[17] * cospi_15_64 - input[15] * cospi_17_64;
temp2 = input[17] * cospi_17_64 + input[15] * cospi_15_64;
step1[17] = dct_const_round_shift(temp1);
step1[30] = dct_const_round_shift(temp2);
temp1 = input[9] * cospi_23_64 - input[23] * cospi_9_64;
temp2 = input[9] * cospi_9_64 + input[23] * cospi_23_64;
step1[18] = dct_const_round_shift(temp1);
step1[29] = dct_const_round_shift(temp2);
temp1 = input[25] * cospi_7_64 - input[7] * cospi_25_64;
temp2 = input[25] * cospi_25_64 + input[7] * cospi_7_64;
step1[19] = dct_const_round_shift(temp1);
step1[28] = dct_const_round_shift(temp2);
temp1 = input[5] * cospi_27_64 - input[27] * cospi_5_64;
temp2 = input[5] * cospi_5_64 + input[27] * cospi_27_64;
step1[20] = dct_const_round_shift(temp1);
step1[27] = dct_const_round_shift(temp2);
temp1 = input[21] * cospi_11_64 - input[11] * cospi_21_64;
temp2 = input[21] * cospi_21_64 + input[11] * cospi_11_64;
step1[21] = dct_const_round_shift(temp1);
step1[26] = dct_const_round_shift(temp2);
temp1 = input[13] * cospi_19_64 - input[19] * cospi_13_64;
temp2 = input[13] * cospi_13_64 + input[19] * cospi_19_64;
step1[22] = dct_const_round_shift(temp1);
step1[25] = dct_const_round_shift(temp2);
temp1 = input[29] * cospi_3_64 - input[3] * cospi_29_64;
temp2 = input[29] * cospi_29_64 + input[3] * cospi_3_64;
step1[23] = dct_const_round_shift(temp1);
step1[24] = dct_const_round_shift(temp2);
// 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] = dct_const_round_shift(temp1);
step2[15] = dct_const_round_shift(temp2);
temp1 = step1[9] * cospi_14_64 - step1[14] * cospi_18_64;
temp2 = step1[9] * cospi_18_64 + step1[14] * cospi_14_64;
step2[9] = dct_const_round_shift(temp1);
step2[14] = dct_const_round_shift(temp2);
temp1 = step1[10] * cospi_22_64 - step1[13] * cospi_10_64;
temp2 = step1[10] * cospi_10_64 + step1[13] * cospi_22_64;
step2[10] = dct_const_round_shift(temp1);
step2[13] = dct_const_round_shift(temp2);
temp1 = step1[11] * cospi_6_64 - step1[12] * cospi_26_64;
temp2 = step1[11] * cospi_26_64 + step1[12] * cospi_6_64;
step2[11] = dct_const_round_shift(temp1);
step2[12] = dct_const_round_shift(temp2);
step2[16] = step1[16] + step1[17];
step2[17] = step1[16] - step1[17];
step2[18] = -step1[18] + step1[19];
step2[19] = step1[18] + step1[19];
step2[20] = step1[20] + step1[21];
step2[21] = step1[20] - step1[21];
step2[22] = -step1[22] + step1[23];
step2[23] = step1[22] + step1[23];
step2[24] = step1[24] + step1[25];
step2[25] = step1[24] - step1[25];
step2[26] = -step1[26] + step1[27];
step2[27] = step1[26] + step1[27];
step2[28] = step1[28] + step1[29];
step2[29] = step1[28] - step1[29];
step2[30] = -step1[30] + step1[31];
step2[31] = step1[30] + step1[31];
// 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] = dct_const_round_shift(temp1);
step1[7] = dct_const_round_shift(temp2);
temp1 = step2[5] * cospi_12_64 - step2[6] * cospi_20_64;
temp2 = step2[5] * cospi_20_64 + step2[6] * cospi_12_64;
step1[5] = dct_const_round_shift(temp1);
step1[6] = dct_const_round_shift(temp2);
step1[8] = step2[8] + step2[9];
step1[9] = step2[8] - step2[9];
step1[10] = -step2[10] + step2[11];
step1[11] = step2[10] + step2[11];
step1[12] = step2[12] + step2[13];
step1[13] = step2[12] - step2[13];
step1[14] = -step2[14] + step2[15];
step1[15] = step2[14] + step2[15];
step1[16] = step2[16];
step1[31] = step2[31];
temp1 = -step2[17] * cospi_4_64 + step2[30] * cospi_28_64;
temp2 = step2[17] * cospi_28_64 + step2[30] * cospi_4_64;
step1[17] = dct_const_round_shift(temp1);
step1[30] = dct_const_round_shift(temp2);
temp1 = -step2[18] * cospi_28_64 - step2[29] * cospi_4_64;
temp2 = -step2[18] * cospi_4_64 + step2[29] * cospi_28_64;
step1[18] = dct_const_round_shift(temp1);
step1[29] = dct_const_round_shift(temp2);
step1[19] = step2[19];
step1[20] = step2[20];
temp1 = -step2[21] * cospi_20_64 + step2[26] * cospi_12_64;
temp2 = step2[21] * cospi_12_64 + step2[26] * cospi_20_64;
step1[21] = dct_const_round_shift(temp1);
step1[26] = dct_const_round_shift(temp2);
temp1 = -step2[22] * cospi_12_64 - step2[25] * cospi_20_64;
temp2 = -step2[22] * cospi_20_64 + step2[25] * cospi_12_64;
step1[22] = dct_const_round_shift(temp1);
step1[25] = dct_const_round_shift(temp2);
step1[23] = step2[23];
step1[24] = step2[24];
step1[27] = step2[27];
step1[28] = step2[28];
// stage 4
temp1 = (step1[0] + step1[1]) * cospi_16_64;
temp2 = (step1[0] - step1[1]) * cospi_16_64;
step2[0] = dct_const_round_shift(temp1);
step2[1] = dct_const_round_shift(temp2);
temp1 = step1[2] * cospi_24_64 - step1[3] * cospi_8_64;
temp2 = step1[2] * cospi_8_64 + step1[3] * cospi_24_64;
step2[2] = dct_const_round_shift(temp1);
step2[3] = dct_const_round_shift(temp2);
step2[4] = step1[4] + step1[5];
step2[5] = step1[4] - step1[5];
step2[6] = -step1[6] + step1[7];
step2[7] = step1[6] + step1[7];
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] = dct_const_round_shift(temp1);
step2[14] = dct_const_round_shift(temp2);
temp1 = -step1[10] * cospi_24_64 - step1[13] * cospi_8_64;
temp2 = -step1[10] * cospi_8_64 + step1[13] * cospi_24_64;
step2[10] = dct_const_round_shift(temp1);
step2[13] = dct_const_round_shift(temp2);
step2[11] = step1[11];
step2[12] = step1[12];
step2[16] = step1[16] + step1[19];
step2[17] = step1[17] + step1[18];
step2[18] = step1[17] - step1[18];
step2[19] = step1[16] - step1[19];
step2[20] = -step1[20] + step1[23];
step2[21] = -step1[21] + step1[22];
step2[22] = step1[21] + step1[22];
step2[23] = step1[20] + step1[23];
step2[24] = step1[24] + step1[27];
step2[25] = step1[25] + step1[26];
step2[26] = step1[25] - step1[26];
step2[27] = step1[24] - step1[27];
step2[28] = -step1[28] + step1[31];
step2[29] = -step1[29] + step1[30];
step2[30] = step1[29] + step1[30];
step2[31] = step1[28] + step1[31];
// stage 5
step1[0] = step2[0] + step2[3];
step1[1] = step2[1] + step2[2];
step1[2] = step2[1] - step2[2];
step1[3] = step2[0] - step2[3];
step1[4] = step2[4];
temp1 = (step2[6] - step2[5]) * cospi_16_64;
temp2 = (step2[5] + step2[6]) * cospi_16_64;
step1[5] = dct_const_round_shift(temp1);
step1[6] = dct_const_round_shift(temp2);
step1[7] = step2[7];
step1[8] = step2[8] + step2[11];
step1[9] = step2[9] + step2[10];
step1[10] = step2[9] - step2[10];
step1[11] = step2[8] - step2[11];
step1[12] = -step2[12] + step2[15];
step1[13] = -step2[13] + step2[14];
step1[14] = step2[13] + step2[14];
step1[15] = step2[12] + step2[15];
step1[16] = step2[16];
step1[17] = step2[17];
temp1 = -step2[18] * cospi_8_64 + step2[29] * cospi_24_64;
temp2 = step2[18] * cospi_24_64 + step2[29] * cospi_8_64;
step1[18] = dct_const_round_shift(temp1);
step1[29] = dct_const_round_shift(temp2);
temp1 = -step2[19] * cospi_8_64 + step2[28] * cospi_24_64;
temp2 = step2[19] * cospi_24_64 + step2[28] * cospi_8_64;
step1[19] = dct_const_round_shift(temp1);
step1[28] = dct_const_round_shift(temp2);
temp1 = -step2[20] * cospi_24_64 - step2[27] * cospi_8_64;
temp2 = -step2[20] * cospi_8_64 + step2[27] * cospi_24_64;
step1[20] = dct_const_round_shift(temp1);
step1[27] = dct_const_round_shift(temp2);
temp1 = -step2[21] * cospi_24_64 - step2[26] * cospi_8_64;
temp2 = -step2[21] * cospi_8_64 + step2[26] * cospi_24_64;
step1[21] = dct_const_round_shift(temp1);
step1[26] = dct_const_round_shift(temp2);
step1[22] = step2[22];
step1[23] = step2[23];
step1[24] = step2[24];
step1[25] = step2[25];
step1[30] = step2[30];
step1[31] = step2[31];
// stage 6
step2[0] = step1[0] + step1[7];
step2[1] = step1[1] + step1[6];
step2[2] = step1[2] + step1[5];
step2[3] = step1[3] + step1[4];
step2[4] = step1[3] - step1[4];
step2[5] = step1[2] - step1[5];
step2[6] = step1[1] - step1[6];
step2[7] = step1[0] - step1[7];
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] = dct_const_round_shift(temp1);
step2[13] = dct_const_round_shift(temp2);
temp1 = (-step1[11] + step1[12]) * cospi_16_64;
temp2 = (step1[11] + step1[12]) * cospi_16_64;
step2[11] = dct_const_round_shift(temp1);
step2[12] = dct_const_round_shift(temp2);
step2[14] = step1[14];
step2[15] = step1[15];
step2[16] = step1[16] + step1[23];
step2[17] = step1[17] + step1[22];
step2[18] = step1[18] + step1[21];
step2[19] = step1[19] + step1[20];
step2[20] = step1[19] - step1[20];
step2[21] = step1[18] - step1[21];
step2[22] = step1[17] - step1[22];
step2[23] = step1[16] - step1[23];
step2[24] = -step1[24] + step1[31];
step2[25] = -step1[25] + step1[30];
step2[26] = -step1[26] + step1[29];
step2[27] = -step1[27] + step1[28];
step2[28] = step1[27] + step1[28];
step2[29] = step1[26] + step1[29];
step2[30] = step1[25] + step1[30];
step2[31] = step1[24] + step1[31];
// stage 7
step1[0] = step2[0] + step2[15];
step1[1] = step2[1] + step2[14];
step1[2] = step2[2] + step2[13];
step1[3] = step2[3] + step2[12];
step1[4] = step2[4] + step2[11];
step1[5] = step2[5] + step2[10];
step1[6] = step2[6] + step2[9];
step1[7] = step2[7] + step2[8];
step1[8] = step2[7] - step2[8];
step1[9] = step2[6] - step2[9];
step1[10] = step2[5] - step2[10];
step1[11] = step2[4] - step2[11];
step1[12] = step2[3] - step2[12];
step1[13] = step2[2] - step2[13];
step1[14] = step2[1] - step2[14];
step1[15] = step2[0] - step2[15];
step1[16] = step2[16];
step1[17] = step2[17];
step1[18] = step2[18];
step1[19] = step2[19];
temp1 = (-step2[20] + step2[27]) * cospi_16_64;
temp2 = (step2[20] + step2[27]) * cospi_16_64;
step1[20] = dct_const_round_shift(temp1);
step1[27] = dct_const_round_shift(temp2);
temp1 = (-step2[21] + step2[26]) * cospi_16_64;
temp2 = (step2[21] + step2[26]) * cospi_16_64;
step1[21] = dct_const_round_shift(temp1);
step1[26] = dct_const_round_shift(temp2);
temp1 = (-step2[22] + step2[25]) * cospi_16_64;
temp2 = (step2[22] + step2[25]) * cospi_16_64;
step1[22] = dct_const_round_shift(temp1);
step1[25] = dct_const_round_shift(temp2);
temp1 = (-step2[23] + step2[24]) * cospi_16_64;
temp2 = (step2[23] + step2[24]) * cospi_16_64;
step1[23] = dct_const_round_shift(temp1);
step1[24] = dct_const_round_shift(temp2);
step1[28] = step2[28];
step1[29] = step2[29];
step1[30] = step2[30];
step1[31] = step2[31];
// final stage
output[0] = step1[0] + step1[31];
output[1] = step1[1] + step1[30];
output[2] = step1[2] + step1[29];
output[3] = step1[3] + step1[28];
output[4] = step1[4] + step1[27];
output[5] = step1[5] + step1[26];
output[6] = step1[6] + step1[25];
output[7] = step1[7] + step1[24];
output[8] = step1[8] + step1[23];
output[9] = step1[9] + step1[22];
output[10] = step1[10] + step1[21];
output[11] = step1[11] + step1[20];
output[12] = step1[12] + step1[19];
output[13] = step1[13] + step1[18];
output[14] = step1[14] + step1[17];
output[15] = step1[15] + step1[16];
output[16] = step1[15] - step1[16];
output[17] = step1[14] - step1[17];
output[18] = step1[13] - step1[18];
output[19] = step1[12] - step1[19];
output[20] = step1[11] - step1[20];
output[21] = step1[10] - step1[21];
output[22] = step1[9] - step1[22];
output[23] = step1[8] - step1[23];
output[24] = step1[7] - step1[24];
output[25] = step1[6] - step1[25];
output[26] = step1[5] - step1[26];
output[27] = step1[4] - step1[27];
output[28] = step1[3] - step1[28];
output[29] = step1[2] - step1[29];
output[30] = step1[1] - step1[30];
output[31] = step1[0] - step1[31];
}
void iadst16_1d(int16_t *input, int16_t *output) {
int s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, s10, s11, s12, s13, s14, s15;
int x0 = input[15];
int x1 = input[0];
int x2 = input[13];
int x3 = input[2];
int x4 = input[11];
int x5 = input[4];
int x6 = input[9];
int x7 = input[6];
int x8 = input[7];
int x9 = input[8];
int x10 = input[5];
int x11 = input[10];
int x12 = input[3];
int x13 = input[12];
int x14 = input[1];
int x15 = input[14];
if (!(x0 | x1 | x2 | x3 | x4 | x5 | x6 | x7 | x8
| x9 | x10 | x11 | x12 | x13 | x14 | x15)) {
output[0] = output[1] = output[2] = output[3] = output[4]
= output[5] = output[6] = output[7] = output[8]
= output[9] = output[10] = output[11] = output[12]
= output[13] = output[14] = output[15] = 0;
return;
}
// stage 1
s0 = x0 * cospi_1_64 + x1 * cospi_31_64;
s1 = x0 * cospi_31_64 - x1 * cospi_1_64;
s2 = x2 * cospi_5_64 + x3 * cospi_27_64;
s3 = x2 * cospi_27_64 - x3 * cospi_5_64;
s4 = x4 * cospi_9_64 + x5 * cospi_23_64;
s5 = x4 * cospi_23_64 - x5 * cospi_9_64;
s6 = x6 * cospi_13_64 + x7 * cospi_19_64;
s7 = x6 * cospi_19_64 - x7 * cospi_13_64;
s8 = x8 * cospi_17_64 + x9 * cospi_15_64;
s9 = x8 * cospi_15_64 - x9 * cospi_17_64;
s10 = x10 * cospi_21_64 + x11 * cospi_11_64;
s11 = x10 * cospi_11_64 - x11 * cospi_21_64;
s12 = x12 * cospi_25_64 + x13 * cospi_7_64;
s13 = x12 * cospi_7_64 - x13 * cospi_25_64;
s14 = x14 * cospi_29_64 + x15 * cospi_3_64;
s15 = x14 * cospi_3_64 - x15 * cospi_29_64;
x0 = dct_const_round_shift(s0 + s8);
x1 = dct_const_round_shift(s1 + s9);
x2 = dct_const_round_shift(s2 + s10);
x3 = dct_const_round_shift(s3 + s11);
x4 = dct_const_round_shift(s4 + s12);
x5 = dct_const_round_shift(s5 + s13);
x6 = dct_const_round_shift(s6 + s14);
x7 = dct_const_round_shift(s7 + s15);
x8 = dct_const_round_shift(s0 - s8);
x9 = dct_const_round_shift(s1 - s9);
x10 = dct_const_round_shift(s2 - s10);
x11 = dct_const_round_shift(s3 - s11);
x12 = dct_const_round_shift(s4 - s12);
x13 = dct_const_round_shift(s5 - s13);
x14 = dct_const_round_shift(s6 - s14);
x15 = dct_const_round_shift(s7 - s15);
// stage 2
s0 = x0;
s1 = x1;
s2 = x2;
s3 = x3;
s4 = x4;
s5 = x5;
s6 = x6;
s7 = x7;
s8 = x8 * cospi_4_64 + x9 * cospi_28_64;
s9 = x8 * cospi_28_64 - x9 * cospi_4_64;
s10 = x10 * cospi_20_64 + x11 * cospi_12_64;
s11 = x10 * cospi_12_64 - x11 * cospi_20_64;
s12 = - x12 * cospi_28_64 + x13 * cospi_4_64;
s13 = x12 * cospi_4_64 + x13 * cospi_28_64;
s14 = - x14 * cospi_12_64 + x15 * cospi_20_64;
s15 = x14 * cospi_20_64 + x15 * cospi_12_64;
x0 = s0 + s4;
x1 = s1 + s5;
x2 = s2 + s6;
x3 = s3 + s7;
x4 = s0 - s4;
x5 = s1 - s5;
x6 = s2 - s6;
x7 = s3 - s7;
x8 = dct_const_round_shift(s8 + s12);
x9 = dct_const_round_shift(s9 + s13);
x10 = dct_const_round_shift(s10 + s14);
x11 = dct_const_round_shift(s11 + s15);
x12 = dct_const_round_shift(s8 - s12);
x13 = dct_const_round_shift(s9 - s13);
x14 = dct_const_round_shift(s10 - s14);
x15 = dct_const_round_shift(s11 - s15);
// stage 3
s0 = x0;
s1 = x1;
s2 = x2;
s3 = x3;
s4 = x4 * cospi_8_64 + x5 * cospi_24_64;
s5 = x4 * cospi_24_64 - x5 * cospi_8_64;
s6 = - x6 * cospi_24_64 + x7 * cospi_8_64;
s7 = x6 * cospi_8_64 + x7 * cospi_24_64;
s8 = x8;
s9 = x9;
s10 = x10;
s11 = x11;
s12 = x12 * cospi_8_64 + x13 * cospi_24_64;
s13 = x12 * cospi_24_64 - x13 * cospi_8_64;
s14 = - x14 * cospi_24_64 + x15 * cospi_8_64;
s15 = x14 * cospi_8_64 + x15 * cospi_24_64;
x0 = s0 + s2;
x1 = s1 + s3;
x2 = s0 - s2;
x3 = s1 - s3;
x4 = dct_const_round_shift(s4 + s6);
x5 = dct_const_round_shift(s5 + s7);
x6 = dct_const_round_shift(s4 - s6);
x7 = dct_const_round_shift(s5 - s7);
x8 = s8 + s10;
x9 = s9 + s11;
x10 = s8 - s10;
x11 = s9 - s11;
x12 = dct_const_round_shift(s12 + s14);
x13 = dct_const_round_shift(s13 + s15);
x14 = dct_const_round_shift(s12 - s14);
x15 = dct_const_round_shift(s13 - s15);
// stage 4
s2 = (- cospi_16_64) * (x2 + x3);
s3 = cospi_16_64 * (x2 - x3);
s6 = cospi_16_64 * (x6 + x7);
s7 = cospi_16_64 * (- x6 + x7);
s10 = cospi_16_64 * (x10 + x11);
s11 = cospi_16_64 * (- x10 + x11);
s14 = (- cospi_16_64) * (x14 + x15);
s15 = cospi_16_64 * (x14 - x15);
x2 = dct_const_round_shift(s2);
x3 = dct_const_round_shift(s3);
x6 = dct_const_round_shift(s6);
x7 = dct_const_round_shift(s7);
x10 = dct_const_round_shift(s10);
x11 = dct_const_round_shift(s11);
x14 = dct_const_round_shift(s14);
x15 = dct_const_round_shift(s15);
output[0] = x0;
output[1] = -x8;
output[2] = x12;
output[3] = -x4;
output[4] = x6;
output[5] = x14;
output[6] = x10;
output[7] = x2;
output[8] = x3;
output[9] = x11;
output[10] = x15;
output[11] = x7;
output[12] = x5;
output[13] = -x13;
output[14] = x9;
output[15] = -x1;
}
static void idct16_1d(int16_t *input, int16_t *output) {
int16_t step1[16], step2[16];
int temp1, temp2;
// stage 1
step1[0] = input[0/2];
step1[1] = input[16/2];
step1[2] = input[8/2];
step1[3] = input[24/2];
step1[4] = input[4/2];
step1[5] = input[20/2];
step1[6] = input[12/2];
step1[7] = input[28/2];
step1[8] = input[2/2];
step1[9] = input[18/2];
step1[10] = input[10/2];
step1[11] = input[26/2];
step1[12] = input[6/2];
step1[13] = input[22/2];
step1[14] = input[14/2];
step1[15] = input[30/2];
// 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] = dct_const_round_shift(temp1);
step2[15] = dct_const_round_shift(temp2);
temp1 = step1[9] * cospi_14_64 - step1[14] * cospi_18_64;
temp2 = step1[9] * cospi_18_64 + step1[14] * cospi_14_64;
step2[9] = dct_const_round_shift(temp1);
step2[14] = dct_const_round_shift(temp2);
temp1 = step1[10] * cospi_22_64 - step1[13] * cospi_10_64;
temp2 = step1[10] * cospi_10_64 + step1[13] * cospi_22_64;
step2[10] = dct_const_round_shift(temp1);
step2[13] = dct_const_round_shift(temp2);
temp1 = step1[11] * cospi_6_64 - step1[12] * cospi_26_64;
temp2 = step1[11] * cospi_26_64 + step1[12] * cospi_6_64;
step2[11] = dct_const_round_shift(temp1);
step2[12] = dct_const_round_shift(temp2);
// 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] = dct_const_round_shift(temp1);
step1[7] = dct_const_round_shift(temp2);
temp1 = step2[5] * cospi_12_64 - step2[6] * cospi_20_64;
temp2 = step2[5] * cospi_20_64 + step2[6] * cospi_12_64;
step1[5] = dct_const_round_shift(temp1);
step1[6] = dct_const_round_shift(temp2);
step1[8] = step2[8] + step2[9];
step1[9] = step2[8] - step2[9];
step1[10] = -step2[10] + step2[11];
step1[11] = step2[10] + step2[11];
step1[12] = step2[12] + step2[13];
step1[13] = step2[12] - step2[13];
step1[14] = -step2[14] + step2[15];
step1[15] = step2[14] + step2[15];
// stage 4
temp1 = (step1[0] + step1[1]) * cospi_16_64;
temp2 = (step1[0] - step1[1]) * cospi_16_64;
step2[0] = dct_const_round_shift(temp1);
step2[1] = dct_const_round_shift(temp2);
temp1 = step1[2] * cospi_24_64 - step1[3] * cospi_8_64;
temp2 = step1[2] * cospi_8_64 + step1[3] * cospi_24_64;
step2[2] = dct_const_round_shift(temp1);
step2[3] = dct_const_round_shift(temp2);
step2[4] = step1[4] + step1[5];
step2[5] = step1[4] - step1[5];
step2[6] = -step1[6] + step1[7];
step2[7] = step1[6] + step1[7];
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] = dct_const_round_shift(temp1);
step2[14] = dct_const_round_shift(temp2);
temp1 = -step1[10] * cospi_24_64 - step1[13] * cospi_8_64;
temp2 = -step1[10] * cospi_8_64 + step1[13] * cospi_24_64;
step2[10] = dct_const_round_shift(temp1);
step2[13] = dct_const_round_shift(temp2);
step2[11] = step1[11];
step2[12] = step1[12];
// stage 5
step1[0] = step2[0] + step2[3];
step1[1] = step2[1] + step2[2];
step1[2] = step2[1] - step2[2];
step1[3] = step2[0] - step2[3];
step1[4] = step2[4];
temp1 = (step2[6] - step2[5]) * cospi_16_64;
temp2 = (step2[5] + step2[6]) * cospi_16_64;
step1[5] = dct_const_round_shift(temp1);
step1[6] = dct_const_round_shift(temp2);
step1[7] = step2[7];
step1[8] = step2[8] + step2[11];
step1[9] = step2[9] + step2[10];
step1[10] = step2[9] - step2[10];
step1[11] = step2[8] - step2[11];
step1[12] = -step2[12] + step2[15];
step1[13] = -step2[13] + step2[14];
step1[14] = step2[13] + step2[14];
step1[15] = step2[12] + step2[15];
// stage 6
step2[0] = step1[0] + step1[7];
step2[1] = step1[1] + step1[6];
step2[2] = step1[2] + step1[5];
step2[3] = step1[3] + step1[4];
step2[4] = step1[3] - step1[4];
step2[5] = step1[2] - step1[5];
step2[6] = step1[1] - step1[6];
step2[7] = step1[0] - step1[7];
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] = dct_const_round_shift(temp1);
step2[13] = dct_const_round_shift(temp2);
temp1 = (-step1[11] + step1[12]) * cospi_16_64;
temp2 = (step1[11] + step1[12]) * cospi_16_64;
step2[11] = dct_const_round_shift(temp1);
step2[12] = dct_const_round_shift(temp2);
step2[14] = step1[14];
step2[15] = step1[15];
// stage 7
output[0] = step2[0] + step2[15];
output[1] = step2[1] + step2[14];
output[2] = step2[2] + step2[13];
output[3] = step2[3] + step2[12];
output[4] = step2[4] + step2[11];
output[5] = step2[5] + step2[10];
output[6] = step2[6] + step2[9];
output[7] = step2[7] + step2[8];
output[8] = step2[7] - step2[8];
output[9] = step2[6] - step2[9];
output[10] = step2[5] - step2[10];
output[11] = step2[4] - step2[11];
output[12] = step2[3] - step2[12];
output[13] = step2[2] - step2[13];
output[14] = step2[1] - step2[14];
output[15] = step2[0] - step2[15];
}
static void iadst8_1d(int16_t *input, int16_t *output) {
int s0, s1, s2, s3, s4, s5, s6, s7;
int x0 = input[7];
int x1 = input[0];
int x2 = input[5];
int x3 = input[2];
int x4 = input[3];
int x5 = input[4];
int x6 = input[1];
int x7 = input[6];
if (!(x0 | x1 | x2 | x3 | x4 | x5 | x6 | x7)) {
output[0] = output[1] = output[2] = output[3] = output[4]
= output[5] = output[6] = output[7] = 0;
return;
}
// stage 1
s0 = cospi_2_64 * x0 + cospi_30_64 * x1;
s1 = cospi_30_64 * x0 - cospi_2_64 * x1;
s2 = cospi_10_64 * x2 + cospi_22_64 * x3;
s3 = cospi_22_64 * x2 - cospi_10_64 * x3;
s4 = cospi_18_64 * x4 + cospi_14_64 * x5;
s5 = cospi_14_64 * x4 - cospi_18_64 * x5;
s6 = cospi_26_64 * x6 + cospi_6_64 * x7;
s7 = cospi_6_64 * x6 - cospi_26_64 * x7;
x0 = dct_const_round_shift(s0 + s4);
x1 = dct_const_round_shift(s1 + s5);
x2 = dct_const_round_shift(s2 + s6);
x3 = dct_const_round_shift(s3 + s7);
x4 = dct_const_round_shift(s0 - s4);
x5 = dct_const_round_shift(s1 - s5);
x6 = dct_const_round_shift(s2 - s6);
x7 = dct_const_round_shift(s3 - s7);
// stage 2
s0 = x0;
s1 = x1;
s2 = x2;
s3 = x3;
s4 = cospi_8_64 * x4 + cospi_24_64 * x5;
s5 = cospi_24_64 * x4 - cospi_8_64 * x5;
s6 = -cospi_24_64 * x6 + cospi_8_64 * x7;
s7 = cospi_8_64 * x6 + cospi_24_64 * x7;
x0 = s0 + s2;
x1 = s1 + s3;
x2 = s0 - s2;
x3 = s1 - s3;
x4 = dct_const_round_shift(s4 + s6);
x5 = dct_const_round_shift(s5 + s7);
x6 = dct_const_round_shift(s4 - s6);
x7 = dct_const_round_shift(s5 - s7);
// stage 3
s2 = cospi_16_64 * (x2 + x3);
s3 = cospi_16_64 * (x2 - x3);
s6 = cospi_16_64 * (x6 + x7);
s7 = cospi_16_64 * (x6 - x7);
x2 = dct_const_round_shift(s2);
x3 = dct_const_round_shift(s3);
x6 = dct_const_round_shift(s6);
x7 = dct_const_round_shift(s7);
output[0] = x0;
output[1] = -x4;
output[2] = x6;
output[3] = -x2;
output[4] = x3;
output[5] = -x7;
output[6] = x5;
output[7] = -x1;
}
void vp9_short_idct1_32x32_c(int16_t *input, int16_t *output) {
int16_t out = dct_const_round_shift(input[0] * cospi_16_64);
out = dct_const_round_shift(out * cospi_16_64);
output[0] = ROUND_POWER_OF_TWO(out, 6);
}
