-#if BITS_IN_JSAMPLE == 8
-#define CONST_BITS 13
-#define PASS1_BITS 4 /* set this to 2 if 16x16 multiplies are faster */
-#else
-#define CONST_BITS 13
-#define PASS1_BITS 1 /* lose a little precision to avoid overflow */
-#endif
-
-/* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
- * causing a lot of useless floating-point operations at run time.
- * To get around this we use the following pre-calculated constants.
- * If you change CONST_BITS you may want to add appropriate values.
- * (With a reasonable C compiler, you can just rely on the FIX() macro...)
- */
-
-#if CONST_BITS == 13
-#define FIX_0_298631336 ((int32_t) 2446) /* FIX(0.298631336) */
-#define FIX_0_390180644 ((int32_t) 3196) /* FIX(0.390180644) */
-#define FIX_0_541196100 ((int32_t) 4433) /* FIX(0.541196100) */
-#define FIX_0_765366865 ((int32_t) 6270) /* FIX(0.765366865) */
-#define FIX_0_899976223 ((int32_t) 7373) /* FIX(0.899976223) */
-#define FIX_1_175875602 ((int32_t) 9633) /* FIX(1.175875602) */
-#define FIX_1_501321110 ((int32_t) 12299) /* FIX(1.501321110) */
-#define FIX_1_847759065 ((int32_t) 15137) /* FIX(1.847759065) */
-#define FIX_1_961570560 ((int32_t) 16069) /* FIX(1.961570560) */
-#define FIX_2_053119869 ((int32_t) 16819) /* FIX(2.053119869) */
-#define FIX_2_562915447 ((int32_t) 20995) /* FIX(2.562915447) */
-#define FIX_3_072711026 ((int32_t) 25172) /* FIX(3.072711026) */
-#else
-#define FIX_0_298631336 FIX(0.298631336)
-#define FIX_0_390180644 FIX(0.390180644)
-#define FIX_0_541196100 FIX(0.541196100)
-#define FIX_0_765366865 FIX(0.765366865)
-#define FIX_0_899976223 FIX(0.899976223)
-#define FIX_1_175875602 FIX(1.175875602)
-#define FIX_1_501321110 FIX(1.501321110)
-#define FIX_1_847759065 FIX(1.847759065)
-#define FIX_1_961570560 FIX(1.961570560)
-#define FIX_2_053119869 FIX(2.053119869)
-#define FIX_2_562915447 FIX(2.562915447)
-#define FIX_3_072711026 FIX(3.072711026)
-#endif
-
-
-/* Multiply an int32_t variable by an int32_t constant to yield an int32_t result.
- * For 8-bit samples with the recommended scaling, all the variable
- * and constant values involved are no more than 16 bits wide, so a
- * 16x16->32 bit multiply can be used instead of a full 32x32 multiply.
- * For 12-bit samples, a full 32-bit multiplication will be needed.
- */
-
-#if BITS_IN_JSAMPLE == 8 && CONST_BITS<=13 && PASS1_BITS<=2
-#define MULTIPLY(var,const) MULTIPLY16C16(var,const)
-#else
-#define MULTIPLY(var,const) ((var) * (const))
-#endif
-
-
-static av_always_inline void row_fdct(DCTELEM * data){
- int_fast32_t tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
- int_fast32_t tmp10, tmp11, tmp12, tmp13;
- int_fast32_t z1, z2, z3, z4, z5;
- DCTELEM *dataptr;
- int ctr;
-
- /* Pass 1: process rows. */
- /* Note results are scaled up by sqrt(8) compared to a true DCT; */
- /* furthermore, we scale the results by 2**PASS1_BITS. */
-
- dataptr = data;
- for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {
- tmp0 = dataptr[0] + dataptr[7];
- tmp7 = dataptr[0] - dataptr[7];
- tmp1 = dataptr[1] + dataptr[6];
- tmp6 = dataptr[1] - dataptr[6];
- tmp2 = dataptr[2] + dataptr[5];
- tmp5 = dataptr[2] - dataptr[5];
- tmp3 = dataptr[3] + dataptr[4];
- tmp4 = dataptr[3] - dataptr[4];
-
- /* Even part per LL&M figure 1 --- note that published figure is faulty;
- * rotator "sqrt(2)*c1" should be "sqrt(2)*c6".
- */
-
- tmp10 = tmp0 + tmp3;
- tmp13 = tmp0 - tmp3;
- tmp11 = tmp1 + tmp2;
- tmp12 = tmp1 - tmp2;
-
- dataptr[0] = (DCTELEM) ((tmp10 + tmp11) << PASS1_BITS);
- dataptr[4] = (DCTELEM) ((tmp10 - tmp11) << PASS1_BITS);
-
- z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100);
- dataptr[2] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, FIX_0_765366865),
- CONST_BITS-PASS1_BITS);
- dataptr[6] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, - FIX_1_847759065),
- CONST_BITS-PASS1_BITS);
-
- /* Odd part per figure 8 --- note paper omits factor of sqrt(2).
- * cK represents cos(K*pi/16).
- * i0..i3 in the paper are tmp4..tmp7 here.
- */
-
- z1 = tmp4 + tmp7;
- z2 = tmp5 + tmp6;
- z3 = tmp4 + tmp6;
- z4 = tmp5 + tmp7;
- z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
-
- tmp4 = MULTIPLY(tmp4, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
- tmp5 = MULTIPLY(tmp5, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
- tmp6 = MULTIPLY(tmp6, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
- tmp7 = MULTIPLY(tmp7, FIX_1_501321110); /* sqrt(2) * ( c1+c3-c5-c7) */
- z1 = MULTIPLY(z1, - FIX_0_899976223); /* sqrt(2) * (c7-c3) */
- z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
- z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
- z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */
-
- z3 += z5;
- z4 += z5;
-
- dataptr[7] = (DCTELEM) DESCALE(tmp4 + z1 + z3, CONST_BITS-PASS1_BITS);
- dataptr[5] = (DCTELEM) DESCALE(tmp5 + z2 + z4, CONST_BITS-PASS1_BITS);
- dataptr[3] = (DCTELEM) DESCALE(tmp6 + z2 + z3, CONST_BITS-PASS1_BITS);
- dataptr[1] = (DCTELEM) DESCALE(tmp7 + z1 + z4, CONST_BITS-PASS1_BITS);
-
- dataptr += DCTSIZE; /* advance pointer to next row */
- }
-}
-
-/*
- * Perform the forward DCT on one block of samples.
- */
-
-GLOBAL(void)
-ff_jpeg_fdct_islow (DCTELEM * data)
-{
- int_fast32_t tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
- int_fast32_t tmp10, tmp11, tmp12, tmp13;
- int_fast32_t z1, z2, z3, z4, z5;
- DCTELEM *dataptr;
- int ctr;
-
- row_fdct(data);
-
- /* Pass 2: process columns.
- * We remove the PASS1_BITS scaling, but leave the results scaled up
- * by an overall factor of 8.
- */
-
- dataptr = data;
- for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {
- tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*7];
- tmp7 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*7];
- tmp1 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*6];
- tmp6 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*6];
- tmp2 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*5];
- tmp5 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*5];
- tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*4];
- tmp4 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*4];
-
- /* Even part per LL&M figure 1 --- note that published figure is faulty;
- * rotator "sqrt(2)*c1" should be "sqrt(2)*c6".
- */
-
- tmp10 = tmp0 + tmp3;
- tmp13 = tmp0 - tmp3;
- tmp11 = tmp1 + tmp2;
- tmp12 = tmp1 - tmp2;
-
- dataptr[DCTSIZE*0] = (DCTELEM) DESCALE(tmp10 + tmp11, PASS1_BITS);
- dataptr[DCTSIZE*4] = (DCTELEM) DESCALE(tmp10 - tmp11, PASS1_BITS);
-
- z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100);
- dataptr[DCTSIZE*2] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, FIX_0_765366865),
- CONST_BITS+PASS1_BITS);
- dataptr[DCTSIZE*6] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, - FIX_1_847759065),
- CONST_BITS+PASS1_BITS);
-
- /* Odd part per figure 8 --- note paper omits factor of sqrt(2).
- * cK represents cos(K*pi/16).
- * i0..i3 in the paper are tmp4..tmp7 here.
- */
-
- z1 = tmp4 + tmp7;
- z2 = tmp5 + tmp6;
- z3 = tmp4 + tmp6;
- z4 = tmp5 + tmp7;
- z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
-
- tmp4 = MULTIPLY(tmp4, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
- tmp5 = MULTIPLY(tmp5, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
- tmp6 = MULTIPLY(tmp6, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
- tmp7 = MULTIPLY(tmp7, FIX_1_501321110); /* sqrt(2) * ( c1+c3-c5-c7) */
- z1 = MULTIPLY(z1, - FIX_0_899976223); /* sqrt(2) * (c7-c3) */
- z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
- z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
- z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */
-
- z3 += z5;
- z4 += z5;
-
- dataptr[DCTSIZE*7] = (DCTELEM) DESCALE(tmp4 + z1 + z3,
- CONST_BITS+PASS1_BITS);
- dataptr[DCTSIZE*5] = (DCTELEM) DESCALE(tmp5 + z2 + z4,
- CONST_BITS+PASS1_BITS);
- dataptr[DCTSIZE*3] = (DCTELEM) DESCALE(tmp6 + z2 + z3,
- CONST_BITS+PASS1_BITS);
- dataptr[DCTSIZE*1] = (DCTELEM) DESCALE(tmp7 + z1 + z4,
- CONST_BITS+PASS1_BITS);
-
- dataptr++; /* advance pointer to next column */
- }
-}
-
-/*
- * The secret of DCT2-4-8 is really simple -- you do the usual 1-DCT
- * on the rows and then, instead of doing even and odd, part on the colums
- * you do even part two times.
- */
-GLOBAL(void)
-ff_fdct248_islow (DCTELEM * data)
-{
- int_fast32_t tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
- int_fast32_t tmp10, tmp11, tmp12, tmp13;
- int_fast32_t z1;
- DCTELEM *dataptr;
- int ctr;
-
- row_fdct(data);
-
- /* Pass 2: process columns.
- * We remove the PASS1_BITS scaling, but leave the results scaled up
- * by an overall factor of 8.
- */
-
- dataptr = data;
- for (ctr = DCTSIZE-1; ctr >= 0; ctr--) {
- tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*1];
- tmp1 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*3];
- tmp2 = dataptr[DCTSIZE*4] + dataptr[DCTSIZE*5];
- tmp3 = dataptr[DCTSIZE*6] + dataptr[DCTSIZE*7];
- tmp4 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*1];
- tmp5 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*3];
- tmp6 = dataptr[DCTSIZE*4] - dataptr[DCTSIZE*5];
- tmp7 = dataptr[DCTSIZE*6] - dataptr[DCTSIZE*7];
-
- tmp10 = tmp0 + tmp3;
- tmp11 = tmp1 + tmp2;
- tmp12 = tmp1 - tmp2;
- tmp13 = tmp0 - tmp3;
-
- dataptr[DCTSIZE*0] = (DCTELEM) DESCALE(tmp10 + tmp11, PASS1_BITS);
- dataptr[DCTSIZE*4] = (DCTELEM) DESCALE(tmp10 - tmp11, PASS1_BITS);
-
- z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100);
- dataptr[DCTSIZE*2] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, FIX_0_765366865),
- CONST_BITS+PASS1_BITS);
- dataptr[DCTSIZE*6] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, - FIX_1_847759065),
- CONST_BITS+PASS1_BITS);
-
- tmp10 = tmp4 + tmp7;
- tmp11 = tmp5 + tmp6;
- tmp12 = tmp5 - tmp6;
- tmp13 = tmp4 - tmp7;
-
- dataptr[DCTSIZE*1] = (DCTELEM) DESCALE(tmp10 + tmp11, PASS1_BITS);
- dataptr[DCTSIZE*5] = (DCTELEM) DESCALE(tmp10 - tmp11, PASS1_BITS);
-
- z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100);
- dataptr[DCTSIZE*3] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, FIX_0_765366865),
- CONST_BITS+PASS1_BITS);
- dataptr[DCTSIZE*7] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, - FIX_1_847759065),
- CONST_BITS+PASS1_BITS);