fast_idct_8 (short *in, int stride)
{
INT32 tmp10, tmp11, tmp12, tmp13;
INT32 tmp20, tmp21, tmp22, tmp23;
INT32 tmp30, tmp31;
INT32 tmp40, tmp41, tmp42, tmp43;
INT32 tmp50, tmp51, tmp52, tmp53;
INT32 in0, in1, in2, in3, in4, in5, in6, in7;
int i, j;
in0 = in[ 0];
in1 = in[stride ];
in2 = in[stride*2];
in3 = in[stride*3];
in4 = in[stride*4];
in5 = in[stride*5];
in6 = in[stride*6];
in7 = in[stride*7];
tmp10 = (in0 + in4) * COS_1_4;
tmp11 = (in0 - in4) * COS_1_4;
tmp12 = in2 * SIN_1_8 - in6 * COS_1_8;
tmp13 = in6 * SIN_1_8 + in2 * COS_1_8;
tmp20 = tmp10 + tmp13;
tmp21 = tmp11 + tmp12;
tmp22 = tmp11 - tmp12;
tmp23 = tmp10 - tmp13;
tmp30 = UNFIXO((in3 + in5) * COS_1_4);
tmp31 = UNFIXO((in3 - in5) * COS_1_4);
tmp40 = OVERSH(in1) + tmp30;
tmp41 = OVERSH(in7) + tmp31;
tmp42 = OVERSH(in1) - tmp30;
tmp43 = OVERSH(in7) - tmp31;
tmp50 = tmp40 * OCOS_1_16 + tmp41 * OSIN_1_16;
tmp51 = tmp40 * OSIN_1_16 - tmp41 * OCOS_1_16;
tmp52 = tmp42 * OCOS_5_16 + tmp43 * OSIN_5_16;
tmp53 = tmp42 * OSIN_5_16 - tmp43 * OCOS_5_16;
in[ 0] = UNFIXH(tmp20 + tmp50);
in[stride ] = UNFIXH(tmp21 + tmp53);
in[stride*2] = UNFIXH(tmp22 + tmp52);
in[stride*3] = UNFIXH(tmp23 + tmp51);
in[stride*4] = UNFIXH(tmp23 - tmp51);
in[stride*5] = UNFIXH(tmp22 - tmp52);
in[stride*6] = UNFIXH(tmp21 - tmp53);
in[stride*7] = UNFIXH(tmp20 - tmp50);
}
/*
__inline__ voidmp_fwd_dct_fast(data2d, dest2d)
Block data2d, dest2d;
*/
__inline__ void mp_fwd_dct_fast(Block data2d, Block dest2d)
{
int32 *data = (int32 *) data2d; /* this algorithm wants
* a 1-d array */
int32 *dest = (int32 *) dest2d;
int rowctr, columncounter;
register int32 *inptr, *outptr;
int32 workspace[DCTSIZE_SQ];
int32 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
int32 tmp10, tmp11, tmp12, tmp13;
int32 tmp14, tmp15, tmp16, tmp17;
int32 tmp25, tmp26;
SHIFT_TEMPS;
/*
* Each iteration of the inner loop performs one 8-point 1-D DCT. It
* reads from a *row* of the input matrix and stores into a *column*
* of the output matrix. In the first pass, we read from the data[]
* array and store into the local workspace[]. In the second pass,
* we read from the workspace[] array and store into data[], thus
* performing the equivalent of a columnar DCT pass with no variable
* array indexing.
*/
inptr = data; /* initialize pointers for first pass */
outptr = workspace;
/* PASS ONE */
for (rowctr = DCTSIZE - 1; rowctr >= 0; rowctr--) {
/*
* many tmps have nonoverlapping lifetime -- flashy
* register colourers should be able to do this lot
* very well
*/
/* SHIFT_TEMPS */
/* temp0 through tmp7: -512 to +512 */
/* if I-block, then -256 to +256 */
tmp0 = inptr[7] + inptr[0];
tmp1 = inptr[6] + inptr[1];
tmp2 = inptr[5] + inptr[2];
tmp3 = inptr[4] + inptr[3];
tmp4 = inptr[3] - inptr[4];
tmp5 = inptr[2] - inptr[5];
tmp6 = inptr[1] - inptr[6];
tmp7 = inptr[0] - inptr[7];
/* tmp10 through tmp13: -1024 to +1024 */
/* if I-block, then -512 to +512 */
tmp10 = tmp3 + tmp0;
tmp11 = tmp2 + tmp1;
tmp12 = tmp1 - tmp2;
tmp13 = tmp0 - tmp3;
outptr[0] = (int32) UNFIXH((tmp10 + tmp11) * SIN_1_4);
outptr[DCTSIZE * 4] = (int32) UNFIXH((tmp10 - tmp11) * COS_1_4);
outptr[DCTSIZE * 2] = (int32) UNFIXH(tmp13 * COS_1_8 + tmp12 * SIN_1_8);
outptr[DCTSIZE * 6] = (int32) UNFIXH(tmp13 * SIN_1_8 - tmp12 * COS_1_8);
tmp16 = UNFIXO((tmp6 + tmp5) * SIN_1_4);
tmp15 = UNFIXO((tmp6 - tmp5) * COS_1_4);
OVERSHIFT(tmp4);
OVERSHIFT(tmp7);
/*
* tmp4, tmp7, tmp15, tmp16 are overscaled by
* OVERSCALE
*/
tmp14 = tmp4 + tmp15;
tmp25 = tmp4 - tmp15;
tmp26 = tmp7 - tmp16;
tmp17 = tmp7 + tmp16;
outptr[DCTSIZE] = (int32) UNFIXH(tmp17 * OCOS_1_16 + tmp14 * OSIN_1_16);
outptr[DCTSIZE * 7] = (int32) UNFIXH(tmp17 * OCOS_7_16 - tmp14 * OSIN_7_16);
outptr[DCTSIZE * 5] = (int32) UNFIXH(tmp26 * OCOS_5_16 + tmp25 * OSIN_5_16);
outptr[DCTSIZE * 3] = (int32) UNFIXH(tmp26 * OCOS_3_16 - tmp25 * OSIN_3_16);
inptr += DCTSIZE; /* advance inptr to next row */
outptr++; /* advance outptr to next column */
}
/* end of pass; in case it was pass 1, set up for pass 2 */
inptr = workspace;
outptr = dest;
columncounter = 0;
/* PASS TWO */
for (rowctr = DCTSIZE - 1; rowctr >= 0; rowctr--) {
/*
* many tmps have nonoverlapping lifetime -- flashy
* register colourers should be able to do this lot
* very well
*/
/* SHIFT_TEMPS */
/* temp0 through tmp7: -512 to +512 */
/* if I-block, then -256 to +256 */
tmp0 = inptr[7] + inptr[0];
tmp1 = inptr[6] + inptr[1];
tmp2 = inptr[5] + inptr[2];
tmp3 = inptr[4] + inptr[3];
tmp4 = inptr[3] - inptr[4];
tmp5 = inptr[2] - inptr[5];
tmp6 = inptr[1] - inptr[6];
tmp7 = inptr[0] - inptr[7];
/* tmp10 through tmp13: -1024 to +1024 */
/* if I-block, then -512 to +512 */
tmp10 = tmp3 + tmp0;
tmp11 = tmp2 + tmp1;
tmp12 = tmp1 - tmp2;
tmp13 = tmp0 - tmp3;
outptr[ zigzag[0][columncounter] ] = (int32) UNFIXH((tmp10 + tmp11) * SIN_1_4);
outptr[ zigzag[4][columncounter] ] = (int32) UNFIXH((tmp10 - tmp11) * COS_1_4);
outptr[ zigzag[2][columncounter] ] = (int32) UNFIXH(tmp13 * COS_1_8 + tmp12 * SIN_1_8);
outptr[ zigzag[6][columncounter] ] = (int32) UNFIXH(tmp13 * SIN_1_8 - tmp12 * COS_1_8);
tmp16 = UNFIXO((tmp6 + tmp5) * SIN_1_4);
tmp15 = UNFIXO((tmp6 - tmp5) * COS_1_4);
OVERSHIFT(tmp4);
OVERSHIFT(tmp7);
/*
* tmp4, tmp7, tmp15, tmp16 are overscaled by
* OVERSCALE
*/
tmp14 = tmp4 + tmp15;
tmp25 = tmp4 - tmp15;
tmp26 = tmp7 - tmp16;
tmp17 = tmp7 + tmp16;
outptr[ zigzag[1][columncounter] ] = (int32) UNFIXH(tmp17 * OCOS_1_16 + tmp14 * OSIN_1_16);
outptr[ zigzag[7][columncounter] ] = (int32) UNFIXH(tmp17 * OCOS_7_16 - tmp14 * OSIN_7_16);
outptr[ zigzag[5][columncounter] ] = (int32) UNFIXH(tmp26 * OCOS_5_16 + tmp25 * OSIN_5_16);
outptr[ zigzag[3][columncounter] ] = (int32) UNFIXH(tmp26 * OCOS_3_16 - tmp25 * OSIN_3_16);
inptr += DCTSIZE; /* advance inptr to next row */
/* outptr++;*/ /* advance outptr to next column */
columncounter++;
}
/* END OF PASS TWO */
}
