static void filter(MpegAudioContext *s, int ch, short *samples, int incr)
{
short *p, *q;
int sum, offset, i, j;
int tmp[64];
int tmp1[32];
int *out;
// print_pow1(samples, 1152);
offset = s->samples_offset[ch];
out = &s->sb_samples[ch][0][0][0];
for(j=0; j<36; j++) {
/* 32 samples at once */
for(i=0; i<32; i++) {
s->samples_buf[ch][offset + (31 - i)] = be2me_16(samples[0]);
samples += incr;
}
/* filter */
p = s->samples_buf[ch] + offset;
q = filter_bank;
/* maxsum = 23169 */
for(i=0; i<64; i++) {
sum = p[0*64] * q[0*64];
sum += p[1*64] * q[1*64];
sum += p[2*64] * q[2*64];
sum += p[3*64] * q[3*64];
sum += p[4*64] * q[4*64];
sum += p[5*64] * q[5*64];
sum += p[6*64] * q[6*64];
sum += p[7*64] * q[7*64];
tmp[i] = sum;
p++;
q++;
}
tmp1[0] = tmp[16] >> WSHIFT;
for( i=1; i<=16; i++ ) tmp1[i] = (tmp[i+16]+tmp[16-i]) >> WSHIFT;
for( i=17; i<=31; i++ ) tmp1[i] = (tmp[i+16]-tmp[80-i]) >> WSHIFT;
idct32(out, tmp1);
/* advance of 32 samples */
offset -= 32;
out += 32;
/* handle the wrap around */
if (offset < 0) {
memmove(s->samples_buf[ch] + SAMPLES_BUF_SIZE - (512 - 32),
s->samples_buf[ch], (512 - 32) * 2);
offset = SAMPLES_BUF_SIZE - 512;
}
}
s->samples_offset[ch] = offset;
// print_pow(s->sb_samples, 1152);
}
void vp9_idct32x32_34_add_c(const int16_t *input, uint8_t *dest, int stride) {
int16_t out[32 * 32] = {0};
int16_t *outptr = out;
int i, j;
int16_t temp_in[32], temp_out[32];
// Rows
// only upper-left 8x8 has non-zero coeff
for (i = 0; i < 8; ++i) {
idct32(input, outptr);
input += 32;
outptr += 32;
}
// Columns
for (i = 0; i < 32; ++i) {
for (j = 0; j < 32; ++j)
temp_in[j] = out[j * 32 + i];
idct32(temp_in, temp_out);
for (j = 0; j < 32; ++j)
dest[j * stride + i] = clip_pixel(ROUND_POWER_OF_TWO(temp_out[j], 6)
+ dest[j * stride + i]);
}
}
void vp9_idct32x32_1024_add_c(const int16_t *input, uint8_t *dest, int stride) {
int16_t out[32 * 32];
int16_t *outptr = out;
int i, j;
int16_t temp_in[32], temp_out[32];
// Rows
for (i = 0; i < 32; ++i) {
int16_t zero_coeff[16];
for (j = 0; j < 16; ++j)
zero_coeff[j] = input[2 * j] | input[2 * j + 1];
for (j = 0; j < 8; ++j)
zero_coeff[j] = zero_coeff[2 * j] | zero_coeff[2 * j + 1];
for (j = 0; j < 4; ++j)
zero_coeff[j] = zero_coeff[2 * j] | zero_coeff[2 * j + 1];
for (j = 0; j < 2; ++j)
zero_coeff[j] = zero_coeff[2 * j] | zero_coeff[2 * j + 1];
if (zero_coeff[0] | zero_coeff[1])
idct32(input, outptr);
else
vpx_memset(outptr, 0, sizeof(int16_t) * 32);
input += 32;
outptr += 32;
}
// Columns
for (i = 0; i < 32; ++i) {
for (j = 0; j < 32; ++j)
temp_in[j] = out[j * 32 + i];
idct32(temp_in, temp_out);
for (j = 0; j < 32; ++j)
dest[j * stride + i] = clip_pixel(ROUND_POWER_OF_TWO(temp_out[j], 6)
+ dest[j * stride + i]);
}
}