mlib_status
__mlib_VideoIDCT8x8_S16_S16_Q1(
mlib_s16 *block,
const mlib_s16 *coeffs)
{
const mlib_s16 *coeffPtr = coeffs;
mlib_s16 *blockPtr = block;
mlib_s64 workspace[64];
mlib_s64 *workPtr = workspace;
mlib_s64 x0, x1, x2, x3, x4, x5, x6, x7, x8;
mlib_s32 str = 8;
mlib_s32 i;
IDCT1(coeffPtr, workPtr, 16384);
workPtr = workspace;
IDCT2(workPtr, blockPtr, 2048);
return (MLIB_SUCCESS);
}
mlib_status
__mlib_VideoIDCT8x8_U8_S16_Q1(
mlib_u8 *block,
const mlib_s16 *coeffs,
mlib_s32 stride)
{
const mlib_s16 *coeffPtr = coeffs;
mlib_u8 *blockPtr = block;
mlib_s64 workspace[64];
mlib_s64 *workPtr = workspace;
mlib_s64 x0, x1, x2, x3, x4, x5, x6, x7, x8;
mlib_s32 i;
mlib_s32 str = 8;
mlib_s64 *inPtr;
IDCT1(coeffPtr, workPtr, (-128 * 8));
inPtr = workspace;
for (i = 0; i < 8; i++) {
/* first stage */
x4 = RCOS_1_16 * inPtr[8 * 1];
x5 = RCOS_7_16 * inPtr[8 * 1];
x6 = RCOS_3_16 * inPtr[8 * 3];
x7 = RCOS_5_16 * inPtr[8 * 3];
/* second stage */
x8 = RTWOSQRT2 * (inPtr[8 * 0]);
x0 = RTWOSQRT2 * (inPtr[8 * 0]);
x2 = RCOS_6_16 * inPtr[8 * 2];
x3 = RCOS_2_16 * inPtr[8 * 2];
x1 = x4 + x6;
x4 = ROUND(COS_4_16 * (x4 - x6));
x6 = x5 - x7;
x5 = ROUND(COS_4_16 * (x5 + x7));
/* third stage */
x7 = x8 + x3;
x8 -= x3;
x3 = x0 + x2;
x0 -= x2;
x2 = (x4 + x5);
x4 -= x5;
/* fourth stage */
SATURATE(x7 + x1, blockPtr[0]);
SATURATE(x3 + x2, blockPtr[1]);
SATURATE(x0 + x4, blockPtr[2]);
SATURATE(x8 + x6, blockPtr[3]);
SATURATE(x8 - x6, blockPtr[4]);
SATURATE(x0 - x4, blockPtr[5]);
SATURATE(x3 - x2, blockPtr[6]);
SATURATE(x7 - x1, blockPtr[7]);
inPtr++;
blockPtr += stride;
}
return (MLIB_SUCCESS);
}
void IDCT(BLOCK *block,int k)
{
int tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
int z5, z10, z11, z12, z13;
BLOCK *ptr;
int i;
/* Pass 1: process columns from input, store into work array. */
switch(k){
case 1:IDCT1(block); return;
}
ptr = block;
for (i = 0; i< DCTSIZE; i++,ptr++) {
/* Due to quantization, we will usually find that many of the input
* coefficients are zero, especially the AC terms. We can exploit this
* by short-circuiting the IDCT calculation for any column in which all
* the AC terms are zero. In that case each output is equal to the
* DC coefficient (with scale factor as needed).
* With typical images and quantization tables, half or more of the
* column DCT calculations can be simplified this way.
*/
if ((ptr[DCTSIZE*1] | ptr[DCTSIZE*2] | ptr[DCTSIZE*3] |
ptr[DCTSIZE*4] | ptr[DCTSIZE*5] | ptr[DCTSIZE*6] |
ptr[DCTSIZE*7]) == 0) {
/* AC terms all zero */
ptr[DCTSIZE*0] =
ptr[DCTSIZE*1] =
ptr[DCTSIZE*2] =
ptr[DCTSIZE*3] =
ptr[DCTSIZE*4] =
ptr[DCTSIZE*5] =
ptr[DCTSIZE*6] =
ptr[DCTSIZE*7] =
ptr[DCTSIZE*0];
continue;
}
/* Even part */
z10 = ptr[DCTSIZE*0] + ptr[DCTSIZE*4]; /* phase 3 */
z11 = ptr[DCTSIZE*0] - ptr[DCTSIZE*4];
z13 = ptr[DCTSIZE*2] + ptr[DCTSIZE*6]; /* phases 5-3 */
z12 = MULTIPLY(ptr[DCTSIZE*2] - ptr[DCTSIZE*6], FIX_1_414213562) - z13; /* 2*c4 */
tmp0 = z10 + z13; /* phase 2 */
tmp3 = z10 - z13;
tmp1 = z11 + z12;
tmp2 = z11 - z12;
/* Odd part */
z13 = ptr[DCTSIZE*3] + ptr[DCTSIZE*5]; /* phase 6 */
z10 = ptr[DCTSIZE*3] - ptr[DCTSIZE*5];
z11 = ptr[DCTSIZE*1] + ptr[DCTSIZE*7];
z12 = ptr[DCTSIZE*1] - ptr[DCTSIZE*7];
z5 = MULTIPLY(z12 - z10, FIX_1_847759065);
tmp7 = z11 + z13; /* phase 5 */
tmp6 = MULTIPLY(z10, FIX_2_613125930) + z5 - tmp7; /* phase 2 */
tmp5 = MULTIPLY(z11 - z13, FIX_1_414213562) - tmp6;
tmp4 = MULTIPLY(z12, FIX_1_082392200) - z5 + tmp5;
ptr[DCTSIZE*0] = (tmp0 + tmp7);
ptr[DCTSIZE*7] = (tmp0 - tmp7);
ptr[DCTSIZE*1] = (tmp1 + tmp6);
ptr[DCTSIZE*6] = (tmp1 - tmp6);
ptr[DCTSIZE*2] = (tmp2 + tmp5);
ptr[DCTSIZE*5] = (tmp2 - tmp5);
ptr[DCTSIZE*4] = (tmp3 + tmp4);
ptr[DCTSIZE*3] = (tmp3 - tmp4);
}
/* Pass 2: process rows from work array, store into output array. */
/* Note that we must descale the results by a factor of 8 == 2**3, */
/* and also undo the PASS1_BITS scaling. */
ptr = block;
for (i = 0; i < DCTSIZE; i++ ,ptr+=DCTSIZE) {
/* Rows of zeroes can be exploited in the same way as we did with columns.
* However, the column calculation has created many nonzero AC terms, so
* the simplification applies less often (typically 5% to 10% of the time).
* On machines with very fast multiplication, it's possible that the
* test takes more time than it's worth. In that case this section
* may be commented out.
*/
#ifndef NO_ZERO_ROW_TEST
if ((ptr[1] | ptr[2] | ptr[3] | ptr[4] | ptr[5] | ptr[6] |
ptr[7]) == 0) {
/* AC terms all zero */
ptr[0] =
ptr[1] =
ptr[2] =
ptr[3] =
ptr[4] =
ptr[5] =
ptr[6] =
ptr[7] =
RANGE(DESCALE(ptr[0], PASS1_BITS+3));;
continue;
}
#endif
/* Even part */
z10 = ptr[0] + ptr[4];
z11 = ptr[0] - ptr[4];
z13 = ptr[2] + ptr[6];
z12 = MULTIPLY(ptr[2] - ptr[6], FIX_1_414213562) - z13;
tmp0 = z10 + z13;
tmp3 = z10 - z13;
tmp1 = z11 + z12;
tmp2 = z11 - z12;
/* Odd part */
z13 = ptr[3] + ptr[5];
z10 = ptr[3] - ptr[5];
z11 = ptr[1] + ptr[7];
z12 = ptr[1] - ptr[7];
z5 = MULTIPLY(z12 - z10, FIX_1_847759065);
tmp7 = z11 + z13; /* phase 5 */
tmp6 = MULTIPLY(z10, FIX_2_613125930) + z5 - tmp7; /* phase 2 */
tmp5 = MULTIPLY(z11 - z13, FIX_1_414213562) - tmp6;
tmp4 = MULTIPLY(z12, FIX_1_082392200) - z5 + tmp5;
/* Final output stage: scale down by a factor of 8 and range-limit */
ptr[0] = RANGE(DESCALE(tmp0 + tmp7, PASS1_BITS+3));;
ptr[7] = RANGE(DESCALE(tmp0 - tmp7, PASS1_BITS+3));;
ptr[1] = RANGE(DESCALE(tmp1 + tmp6, PASS1_BITS+3));;
ptr[6] = RANGE(DESCALE(tmp1 - tmp6, PASS1_BITS+3));;
ptr[2] = RANGE(DESCALE(tmp2 + tmp5, PASS1_BITS+3));;
ptr[5] = RANGE(DESCALE(tmp2 - tmp5, PASS1_BITS+3));;
ptr[4] = RANGE(DESCALE(tmp3 + tmp4, PASS1_BITS+3));;
ptr[3] = RANGE(DESCALE(tmp3 - tmp4, PASS1_BITS+3));;
}
}