Final 350 project
Dependencies: uzair Camera_LS_Y201 F7_Ethernet LCD_DISCO_F746NG NetworkAPI SDFileSystem mbed
includes/jfdctint.c@0:791a779d6220, 2017-07-31 (annotated)
- Committer:
- shoaib_ahmed
- Date:
- Mon Jul 31 09:16:35 2017 +0000
- Revision:
- 0:791a779d6220
final project;
Who changed what in which revision?
User | Revision | Line number | New contents of line |
---|---|---|---|
shoaib_ahmed | 0:791a779d6220 | 1 | /* |
shoaib_ahmed | 0:791a779d6220 | 2 | * jfdctint.c |
shoaib_ahmed | 0:791a779d6220 | 3 | * |
shoaib_ahmed | 0:791a779d6220 | 4 | * Copyright (C) 1991-1996, Thomas G. Lane. |
shoaib_ahmed | 0:791a779d6220 | 5 | * Modification developed 2003-2015 by Guido Vollbeding. |
shoaib_ahmed | 0:791a779d6220 | 6 | * This file is part of the Independent JPEG Group's software. |
shoaib_ahmed | 0:791a779d6220 | 7 | * For conditions of distribution and use, see the accompanying README file. |
shoaib_ahmed | 0:791a779d6220 | 8 | * |
shoaib_ahmed | 0:791a779d6220 | 9 | * This file contains a slow-but-accurate integer implementation of the |
shoaib_ahmed | 0:791a779d6220 | 10 | * forward DCT (Discrete Cosine Transform). |
shoaib_ahmed | 0:791a779d6220 | 11 | * |
shoaib_ahmed | 0:791a779d6220 | 12 | * A 2-D DCT can be done by 1-D DCT on each row followed by 1-D DCT |
shoaib_ahmed | 0:791a779d6220 | 13 | * on each column. Direct algorithms are also available, but they are |
shoaib_ahmed | 0:791a779d6220 | 14 | * much more complex and seem not to be any faster when reduced to code. |
shoaib_ahmed | 0:791a779d6220 | 15 | * |
shoaib_ahmed | 0:791a779d6220 | 16 | * This implementation is based on an algorithm described in |
shoaib_ahmed | 0:791a779d6220 | 17 | * C. Loeffler, A. Ligtenberg and G. Moschytz, "Practical Fast 1-D DCT |
shoaib_ahmed | 0:791a779d6220 | 18 | * Algorithms with 11 Multiplications", Proc. Int'l. Conf. on Acoustics, |
shoaib_ahmed | 0:791a779d6220 | 19 | * Speech, and Signal Processing 1989 (ICASSP '89), pp. 988-991. |
shoaib_ahmed | 0:791a779d6220 | 20 | * The primary algorithm described there uses 11 multiplies and 29 adds. |
shoaib_ahmed | 0:791a779d6220 | 21 | * We use their alternate method with 12 multiplies and 32 adds. |
shoaib_ahmed | 0:791a779d6220 | 22 | * The advantage of this method is that no data path contains more than one |
shoaib_ahmed | 0:791a779d6220 | 23 | * multiplication; this allows a very simple and accurate implementation in |
shoaib_ahmed | 0:791a779d6220 | 24 | * scaled fixed-point arithmetic, with a minimal number of shifts. |
shoaib_ahmed | 0:791a779d6220 | 25 | * |
shoaib_ahmed | 0:791a779d6220 | 26 | * We also provide FDCT routines with various input sample block sizes for |
shoaib_ahmed | 0:791a779d6220 | 27 | * direct resolution reduction or enlargement and for direct resolving the |
shoaib_ahmed | 0:791a779d6220 | 28 | * common 2x1 and 1x2 subsampling cases without additional resampling: NxN |
shoaib_ahmed | 0:791a779d6220 | 29 | * (N=1...16), 2NxN, and Nx2N (N=1...8) pixels for one 8x8 output DCT block. |
shoaib_ahmed | 0:791a779d6220 | 30 | * |
shoaib_ahmed | 0:791a779d6220 | 31 | * For N<8 we fill the remaining block coefficients with zero. |
shoaib_ahmed | 0:791a779d6220 | 32 | * For N>8 we apply a partial N-point FDCT on the input samples, computing |
shoaib_ahmed | 0:791a779d6220 | 33 | * just the lower 8 frequency coefficients and discarding the rest. |
shoaib_ahmed | 0:791a779d6220 | 34 | * |
shoaib_ahmed | 0:791a779d6220 | 35 | * We must scale the output coefficients of the N-point FDCT appropriately |
shoaib_ahmed | 0:791a779d6220 | 36 | * to the standard 8-point FDCT level by 8/N per 1-D pass. This scaling |
shoaib_ahmed | 0:791a779d6220 | 37 | * is folded into the constant multipliers (pass 2) and/or final/initial |
shoaib_ahmed | 0:791a779d6220 | 38 | * shifting. |
shoaib_ahmed | 0:791a779d6220 | 39 | * |
shoaib_ahmed | 0:791a779d6220 | 40 | * CAUTION: We rely on the FIX() macro except for the N=1,2,4,8 cases |
shoaib_ahmed | 0:791a779d6220 | 41 | * since there would be too many additional constants to pre-calculate. |
shoaib_ahmed | 0:791a779d6220 | 42 | */ |
shoaib_ahmed | 0:791a779d6220 | 43 | |
shoaib_ahmed | 0:791a779d6220 | 44 | #define JPEG_INTERNALS |
shoaib_ahmed | 0:791a779d6220 | 45 | #include "jinclude.h" |
shoaib_ahmed | 0:791a779d6220 | 46 | #include "jpeglib.h" |
shoaib_ahmed | 0:791a779d6220 | 47 | #include "jdct.h" /* Private declarations for DCT subsystem */ |
shoaib_ahmed | 0:791a779d6220 | 48 | |
shoaib_ahmed | 0:791a779d6220 | 49 | #ifdef DCT_ISLOW_SUPPORTED |
shoaib_ahmed | 0:791a779d6220 | 50 | |
shoaib_ahmed | 0:791a779d6220 | 51 | |
shoaib_ahmed | 0:791a779d6220 | 52 | /* |
shoaib_ahmed | 0:791a779d6220 | 53 | * This module is specialized to the case DCTSIZE = 8. |
shoaib_ahmed | 0:791a779d6220 | 54 | */ |
shoaib_ahmed | 0:791a779d6220 | 55 | |
shoaib_ahmed | 0:791a779d6220 | 56 | #if DCTSIZE != 8 |
shoaib_ahmed | 0:791a779d6220 | 57 | Sorry, this code only copes with 8x8 DCT blocks. /* deliberate syntax err */ |
shoaib_ahmed | 0:791a779d6220 | 58 | #endif |
shoaib_ahmed | 0:791a779d6220 | 59 | |
shoaib_ahmed | 0:791a779d6220 | 60 | |
shoaib_ahmed | 0:791a779d6220 | 61 | /* |
shoaib_ahmed | 0:791a779d6220 | 62 | * The poop on this scaling stuff is as follows: |
shoaib_ahmed | 0:791a779d6220 | 63 | * |
shoaib_ahmed | 0:791a779d6220 | 64 | * Each 1-D DCT step produces outputs which are a factor of sqrt(N) |
shoaib_ahmed | 0:791a779d6220 | 65 | * larger than the true DCT outputs. The final outputs are therefore |
shoaib_ahmed | 0:791a779d6220 | 66 | * a factor of N larger than desired; since N=8 this can be cured by |
shoaib_ahmed | 0:791a779d6220 | 67 | * a simple right shift at the end of the algorithm. The advantage of |
shoaib_ahmed | 0:791a779d6220 | 68 | * this arrangement is that we save two multiplications per 1-D DCT, |
shoaib_ahmed | 0:791a779d6220 | 69 | * because the y0 and y4 outputs need not be divided by sqrt(N). |
shoaib_ahmed | 0:791a779d6220 | 70 | * In the IJG code, this factor of 8 is removed by the quantization step |
shoaib_ahmed | 0:791a779d6220 | 71 | * (in jcdctmgr.c), NOT in this module. |
shoaib_ahmed | 0:791a779d6220 | 72 | * |
shoaib_ahmed | 0:791a779d6220 | 73 | * We have to do addition and subtraction of the integer inputs, which |
shoaib_ahmed | 0:791a779d6220 | 74 | * is no problem, and multiplication by fractional constants, which is |
shoaib_ahmed | 0:791a779d6220 | 75 | * a problem to do in integer arithmetic. We multiply all the constants |
shoaib_ahmed | 0:791a779d6220 | 76 | * by CONST_SCALE and convert them to integer constants (thus retaining |
shoaib_ahmed | 0:791a779d6220 | 77 | * CONST_BITS bits of precision in the constants). After doing a |
shoaib_ahmed | 0:791a779d6220 | 78 | * multiplication we have to divide the product by CONST_SCALE, with proper |
shoaib_ahmed | 0:791a779d6220 | 79 | * rounding, to produce the correct output. This division can be done |
shoaib_ahmed | 0:791a779d6220 | 80 | * cheaply as a right shift of CONST_BITS bits. We postpone shifting |
shoaib_ahmed | 0:791a779d6220 | 81 | * as long as possible so that partial sums can be added together with |
shoaib_ahmed | 0:791a779d6220 | 82 | * full fractional precision. |
shoaib_ahmed | 0:791a779d6220 | 83 | * |
shoaib_ahmed | 0:791a779d6220 | 84 | * The outputs of the first pass are scaled up by PASS1_BITS bits so that |
shoaib_ahmed | 0:791a779d6220 | 85 | * they are represented to better-than-integral precision. These outputs |
shoaib_ahmed | 0:791a779d6220 | 86 | * require BITS_IN_JSAMPLE + PASS1_BITS + 3 bits; this fits in a 16-bit word |
shoaib_ahmed | 0:791a779d6220 | 87 | * with the recommended scaling. (For 12-bit sample data, the intermediate |
shoaib_ahmed | 0:791a779d6220 | 88 | * array is INT32 anyway.) |
shoaib_ahmed | 0:791a779d6220 | 89 | * |
shoaib_ahmed | 0:791a779d6220 | 90 | * To avoid overflow of the 32-bit intermediate results in pass 2, we must |
shoaib_ahmed | 0:791a779d6220 | 91 | * have BITS_IN_JSAMPLE + CONST_BITS + PASS1_BITS <= 26. Error analysis |
shoaib_ahmed | 0:791a779d6220 | 92 | * shows that the values given below are the most effective. |
shoaib_ahmed | 0:791a779d6220 | 93 | */ |
shoaib_ahmed | 0:791a779d6220 | 94 | |
shoaib_ahmed | 0:791a779d6220 | 95 | #if BITS_IN_JSAMPLE == 8 |
shoaib_ahmed | 0:791a779d6220 | 96 | #define CONST_BITS 13 |
shoaib_ahmed | 0:791a779d6220 | 97 | #define PASS1_BITS 2 |
shoaib_ahmed | 0:791a779d6220 | 98 | #else |
shoaib_ahmed | 0:791a779d6220 | 99 | #define CONST_BITS 13 |
shoaib_ahmed | 0:791a779d6220 | 100 | #define PASS1_BITS 1 /* lose a little precision to avoid overflow */ |
shoaib_ahmed | 0:791a779d6220 | 101 | #endif |
shoaib_ahmed | 0:791a779d6220 | 102 | |
shoaib_ahmed | 0:791a779d6220 | 103 | /* Some C compilers fail to reduce "FIX(constant)" at compile time, thus |
shoaib_ahmed | 0:791a779d6220 | 104 | * causing a lot of useless floating-point operations at run time. |
shoaib_ahmed | 0:791a779d6220 | 105 | * To get around this we use the following pre-calculated constants. |
shoaib_ahmed | 0:791a779d6220 | 106 | * If you change CONST_BITS you may want to add appropriate values. |
shoaib_ahmed | 0:791a779d6220 | 107 | * (With a reasonable C compiler, you can just rely on the FIX() macro...) |
shoaib_ahmed | 0:791a779d6220 | 108 | */ |
shoaib_ahmed | 0:791a779d6220 | 109 | |
shoaib_ahmed | 0:791a779d6220 | 110 | #if CONST_BITS == 13 |
shoaib_ahmed | 0:791a779d6220 | 111 | #define FIX_0_298631336 ((INT32) 2446) /* FIX(0.298631336) */ |
shoaib_ahmed | 0:791a779d6220 | 112 | #define FIX_0_390180644 ((INT32) 3196) /* FIX(0.390180644) */ |
shoaib_ahmed | 0:791a779d6220 | 113 | #define FIX_0_541196100 ((INT32) 4433) /* FIX(0.541196100) */ |
shoaib_ahmed | 0:791a779d6220 | 114 | #define FIX_0_765366865 ((INT32) 6270) /* FIX(0.765366865) */ |
shoaib_ahmed | 0:791a779d6220 | 115 | #define FIX_0_899976223 ((INT32) 7373) /* FIX(0.899976223) */ |
shoaib_ahmed | 0:791a779d6220 | 116 | #define FIX_1_175875602 ((INT32) 9633) /* FIX(1.175875602) */ |
shoaib_ahmed | 0:791a779d6220 | 117 | #define FIX_1_501321110 ((INT32) 12299) /* FIX(1.501321110) */ |
shoaib_ahmed | 0:791a779d6220 | 118 | #define FIX_1_847759065 ((INT32) 15137) /* FIX(1.847759065) */ |
shoaib_ahmed | 0:791a779d6220 | 119 | #define FIX_1_961570560 ((INT32) 16069) /* FIX(1.961570560) */ |
shoaib_ahmed | 0:791a779d6220 | 120 | #define FIX_2_053119869 ((INT32) 16819) /* FIX(2.053119869) */ |
shoaib_ahmed | 0:791a779d6220 | 121 | #define FIX_2_562915447 ((INT32) 20995) /* FIX(2.562915447) */ |
shoaib_ahmed | 0:791a779d6220 | 122 | #define FIX_3_072711026 ((INT32) 25172) /* FIX(3.072711026) */ |
shoaib_ahmed | 0:791a779d6220 | 123 | #else |
shoaib_ahmed | 0:791a779d6220 | 124 | #define FIX_0_298631336 FIX(0.298631336) |
shoaib_ahmed | 0:791a779d6220 | 125 | #define FIX_0_390180644 FIX(0.390180644) |
shoaib_ahmed | 0:791a779d6220 | 126 | #define FIX_0_541196100 FIX(0.541196100) |
shoaib_ahmed | 0:791a779d6220 | 127 | #define FIX_0_765366865 FIX(0.765366865) |
shoaib_ahmed | 0:791a779d6220 | 128 | #define FIX_0_899976223 FIX(0.899976223) |
shoaib_ahmed | 0:791a779d6220 | 129 | #define FIX_1_175875602 FIX(1.175875602) |
shoaib_ahmed | 0:791a779d6220 | 130 | #define FIX_1_501321110 FIX(1.501321110) |
shoaib_ahmed | 0:791a779d6220 | 131 | #define FIX_1_847759065 FIX(1.847759065) |
shoaib_ahmed | 0:791a779d6220 | 132 | #define FIX_1_961570560 FIX(1.961570560) |
shoaib_ahmed | 0:791a779d6220 | 133 | #define FIX_2_053119869 FIX(2.053119869) |
shoaib_ahmed | 0:791a779d6220 | 134 | #define FIX_2_562915447 FIX(2.562915447) |
shoaib_ahmed | 0:791a779d6220 | 135 | #define FIX_3_072711026 FIX(3.072711026) |
shoaib_ahmed | 0:791a779d6220 | 136 | #endif |
shoaib_ahmed | 0:791a779d6220 | 137 | |
shoaib_ahmed | 0:791a779d6220 | 138 | |
shoaib_ahmed | 0:791a779d6220 | 139 | /* Multiply an INT32 variable by an INT32 constant to yield an INT32 result. |
shoaib_ahmed | 0:791a779d6220 | 140 | * For 8-bit samples with the recommended scaling, all the variable |
shoaib_ahmed | 0:791a779d6220 | 141 | * and constant values involved are no more than 16 bits wide, so a |
shoaib_ahmed | 0:791a779d6220 | 142 | * 16x16->32 bit multiply can be used instead of a full 32x32 multiply. |
shoaib_ahmed | 0:791a779d6220 | 143 | * For 12-bit samples, a full 32-bit multiplication will be needed. |
shoaib_ahmed | 0:791a779d6220 | 144 | */ |
shoaib_ahmed | 0:791a779d6220 | 145 | |
shoaib_ahmed | 0:791a779d6220 | 146 | #if BITS_IN_JSAMPLE == 8 |
shoaib_ahmed | 0:791a779d6220 | 147 | #define MULTIPLY(var,const) MULTIPLY16C16(var,const) |
shoaib_ahmed | 0:791a779d6220 | 148 | #else |
shoaib_ahmed | 0:791a779d6220 | 149 | #define MULTIPLY(var,const) ((var) * (const)) |
shoaib_ahmed | 0:791a779d6220 | 150 | #endif |
shoaib_ahmed | 0:791a779d6220 | 151 | |
shoaib_ahmed | 0:791a779d6220 | 152 | |
shoaib_ahmed | 0:791a779d6220 | 153 | /* |
shoaib_ahmed | 0:791a779d6220 | 154 | * Perform the forward DCT on one block of samples. |
shoaib_ahmed | 0:791a779d6220 | 155 | */ |
shoaib_ahmed | 0:791a779d6220 | 156 | |
shoaib_ahmed | 0:791a779d6220 | 157 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 158 | jpeg_fdct_islow (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 159 | { |
shoaib_ahmed | 0:791a779d6220 | 160 | INT32 tmp0, tmp1, tmp2, tmp3; |
shoaib_ahmed | 0:791a779d6220 | 161 | INT32 tmp10, tmp11, tmp12, tmp13; |
shoaib_ahmed | 0:791a779d6220 | 162 | INT32 z1; |
shoaib_ahmed | 0:791a779d6220 | 163 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 164 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 165 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 166 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 167 | |
shoaib_ahmed | 0:791a779d6220 | 168 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 169 | * Note results are scaled up by sqrt(8) compared to a true DCT; |
shoaib_ahmed | 0:791a779d6220 | 170 | * furthermore, we scale the results by 2**PASS1_BITS. |
shoaib_ahmed | 0:791a779d6220 | 171 | * cK represents sqrt(2) * cos(K*pi/16). |
shoaib_ahmed | 0:791a779d6220 | 172 | */ |
shoaib_ahmed | 0:791a779d6220 | 173 | |
shoaib_ahmed | 0:791a779d6220 | 174 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 175 | for (ctr = 0; ctr < DCTSIZE; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 176 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 177 | |
shoaib_ahmed | 0:791a779d6220 | 178 | /* Even part per LL&M figure 1 --- note that published figure is faulty; |
shoaib_ahmed | 0:791a779d6220 | 179 | * rotator "c1" should be "c6". |
shoaib_ahmed | 0:791a779d6220 | 180 | */ |
shoaib_ahmed | 0:791a779d6220 | 181 | |
shoaib_ahmed | 0:791a779d6220 | 182 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[7]); |
shoaib_ahmed | 0:791a779d6220 | 183 | tmp1 = GETJSAMPLE(elemptr[1]) + GETJSAMPLE(elemptr[6]); |
shoaib_ahmed | 0:791a779d6220 | 184 | tmp2 = GETJSAMPLE(elemptr[2]) + GETJSAMPLE(elemptr[5]); |
shoaib_ahmed | 0:791a779d6220 | 185 | tmp3 = GETJSAMPLE(elemptr[3]) + GETJSAMPLE(elemptr[4]); |
shoaib_ahmed | 0:791a779d6220 | 186 | |
shoaib_ahmed | 0:791a779d6220 | 187 | tmp10 = tmp0 + tmp3; |
shoaib_ahmed | 0:791a779d6220 | 188 | tmp12 = tmp0 - tmp3; |
shoaib_ahmed | 0:791a779d6220 | 189 | tmp11 = tmp1 + tmp2; |
shoaib_ahmed | 0:791a779d6220 | 190 | tmp13 = tmp1 - tmp2; |
shoaib_ahmed | 0:791a779d6220 | 191 | |
shoaib_ahmed | 0:791a779d6220 | 192 | tmp0 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[7]); |
shoaib_ahmed | 0:791a779d6220 | 193 | tmp1 = GETJSAMPLE(elemptr[1]) - GETJSAMPLE(elemptr[6]); |
shoaib_ahmed | 0:791a779d6220 | 194 | tmp2 = GETJSAMPLE(elemptr[2]) - GETJSAMPLE(elemptr[5]); |
shoaib_ahmed | 0:791a779d6220 | 195 | tmp3 = GETJSAMPLE(elemptr[3]) - GETJSAMPLE(elemptr[4]); |
shoaib_ahmed | 0:791a779d6220 | 196 | |
shoaib_ahmed | 0:791a779d6220 | 197 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 198 | dataptr[0] = (DCTELEM) ((tmp10 + tmp11 - 8 * CENTERJSAMPLE) << PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 199 | dataptr[4] = (DCTELEM) ((tmp10 - tmp11) << PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 200 | |
shoaib_ahmed | 0:791a779d6220 | 201 | z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 202 | /* Add fudge factor here for final descale. */ |
shoaib_ahmed | 0:791a779d6220 | 203 | z1 += ONE << (CONST_BITS-PASS1_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 204 | |
shoaib_ahmed | 0:791a779d6220 | 205 | dataptr[2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 206 | RIGHT_SHIFT(z1 + MULTIPLY(tmp12, FIX_0_765366865), /* c2-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 207 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 208 | dataptr[6] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 209 | RIGHT_SHIFT(z1 - MULTIPLY(tmp13, FIX_1_847759065), /* c2+c6 */ |
shoaib_ahmed | 0:791a779d6220 | 210 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 211 | |
shoaib_ahmed | 0:791a779d6220 | 212 | /* Odd part per figure 8 --- note paper omits factor of sqrt(2). |
shoaib_ahmed | 0:791a779d6220 | 213 | * i0..i3 in the paper are tmp0..tmp3 here. |
shoaib_ahmed | 0:791a779d6220 | 214 | */ |
shoaib_ahmed | 0:791a779d6220 | 215 | |
shoaib_ahmed | 0:791a779d6220 | 216 | tmp12 = tmp0 + tmp2; |
shoaib_ahmed | 0:791a779d6220 | 217 | tmp13 = tmp1 + tmp3; |
shoaib_ahmed | 0:791a779d6220 | 218 | |
shoaib_ahmed | 0:791a779d6220 | 219 | z1 = MULTIPLY(tmp12 + tmp13, FIX_1_175875602); /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 220 | /* Add fudge factor here for final descale. */ |
shoaib_ahmed | 0:791a779d6220 | 221 | z1 += ONE << (CONST_BITS-PASS1_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 222 | |
shoaib_ahmed | 0:791a779d6220 | 223 | tmp12 = MULTIPLY(tmp12, - FIX_0_390180644); /* -c3+c5 */ |
shoaib_ahmed | 0:791a779d6220 | 224 | tmp13 = MULTIPLY(tmp13, - FIX_1_961570560); /* -c3-c5 */ |
shoaib_ahmed | 0:791a779d6220 | 225 | tmp12 += z1; |
shoaib_ahmed | 0:791a779d6220 | 226 | tmp13 += z1; |
shoaib_ahmed | 0:791a779d6220 | 227 | |
shoaib_ahmed | 0:791a779d6220 | 228 | z1 = MULTIPLY(tmp0 + tmp3, - FIX_0_899976223); /* -c3+c7 */ |
shoaib_ahmed | 0:791a779d6220 | 229 | tmp0 = MULTIPLY(tmp0, FIX_1_501321110); /* c1+c3-c5-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 230 | tmp3 = MULTIPLY(tmp3, FIX_0_298631336); /* -c1+c3+c5-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 231 | tmp0 += z1 + tmp12; |
shoaib_ahmed | 0:791a779d6220 | 232 | tmp3 += z1 + tmp13; |
shoaib_ahmed | 0:791a779d6220 | 233 | |
shoaib_ahmed | 0:791a779d6220 | 234 | z1 = MULTIPLY(tmp1 + tmp2, - FIX_2_562915447); /* -c1-c3 */ |
shoaib_ahmed | 0:791a779d6220 | 235 | tmp1 = MULTIPLY(tmp1, FIX_3_072711026); /* c1+c3+c5-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 236 | tmp2 = MULTIPLY(tmp2, FIX_2_053119869); /* c1+c3-c5+c7 */ |
shoaib_ahmed | 0:791a779d6220 | 237 | tmp1 += z1 + tmp13; |
shoaib_ahmed | 0:791a779d6220 | 238 | tmp2 += z1 + tmp12; |
shoaib_ahmed | 0:791a779d6220 | 239 | |
shoaib_ahmed | 0:791a779d6220 | 240 | dataptr[1] = (DCTELEM) RIGHT_SHIFT(tmp0, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 241 | dataptr[3] = (DCTELEM) RIGHT_SHIFT(tmp1, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 242 | dataptr[5] = (DCTELEM) RIGHT_SHIFT(tmp2, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 243 | dataptr[7] = (DCTELEM) RIGHT_SHIFT(tmp3, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 244 | |
shoaib_ahmed | 0:791a779d6220 | 245 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 246 | } |
shoaib_ahmed | 0:791a779d6220 | 247 | |
shoaib_ahmed | 0:791a779d6220 | 248 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 249 | * We remove the PASS1_BITS scaling, but leave the results scaled up |
shoaib_ahmed | 0:791a779d6220 | 250 | * by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 251 | * cK represents sqrt(2) * cos(K*pi/16). |
shoaib_ahmed | 0:791a779d6220 | 252 | */ |
shoaib_ahmed | 0:791a779d6220 | 253 | |
shoaib_ahmed | 0:791a779d6220 | 254 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 255 | for (ctr = DCTSIZE-1; ctr >= 0; ctr--) { |
shoaib_ahmed | 0:791a779d6220 | 256 | /* Even part per LL&M figure 1 --- note that published figure is faulty; |
shoaib_ahmed | 0:791a779d6220 | 257 | * rotator "c1" should be "c6". |
shoaib_ahmed | 0:791a779d6220 | 258 | */ |
shoaib_ahmed | 0:791a779d6220 | 259 | |
shoaib_ahmed | 0:791a779d6220 | 260 | tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 261 | tmp1 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 262 | tmp2 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 263 | tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 264 | |
shoaib_ahmed | 0:791a779d6220 | 265 | /* Add fudge factor here for final descale. */ |
shoaib_ahmed | 0:791a779d6220 | 266 | tmp10 = tmp0 + tmp3 + (ONE << (PASS1_BITS-1)); |
shoaib_ahmed | 0:791a779d6220 | 267 | tmp12 = tmp0 - tmp3; |
shoaib_ahmed | 0:791a779d6220 | 268 | tmp11 = tmp1 + tmp2; |
shoaib_ahmed | 0:791a779d6220 | 269 | tmp13 = tmp1 - tmp2; |
shoaib_ahmed | 0:791a779d6220 | 270 | |
shoaib_ahmed | 0:791a779d6220 | 271 | tmp0 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 272 | tmp1 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 273 | tmp2 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 274 | tmp3 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 275 | |
shoaib_ahmed | 0:791a779d6220 | 276 | dataptr[DCTSIZE*0] = (DCTELEM) RIGHT_SHIFT(tmp10 + tmp11, PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 277 | dataptr[DCTSIZE*4] = (DCTELEM) RIGHT_SHIFT(tmp10 - tmp11, PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 278 | |
shoaib_ahmed | 0:791a779d6220 | 279 | z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 280 | /* Add fudge factor here for final descale. */ |
shoaib_ahmed | 0:791a779d6220 | 281 | z1 += ONE << (CONST_BITS+PASS1_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 282 | |
shoaib_ahmed | 0:791a779d6220 | 283 | dataptr[DCTSIZE*2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 284 | RIGHT_SHIFT(z1 + MULTIPLY(tmp12, FIX_0_765366865), /* c2-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 285 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 286 | dataptr[DCTSIZE*6] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 287 | RIGHT_SHIFT(z1 - MULTIPLY(tmp13, FIX_1_847759065), /* c2+c6 */ |
shoaib_ahmed | 0:791a779d6220 | 288 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 289 | |
shoaib_ahmed | 0:791a779d6220 | 290 | /* Odd part per figure 8 --- note paper omits factor of sqrt(2). |
shoaib_ahmed | 0:791a779d6220 | 291 | * i0..i3 in the paper are tmp0..tmp3 here. |
shoaib_ahmed | 0:791a779d6220 | 292 | */ |
shoaib_ahmed | 0:791a779d6220 | 293 | |
shoaib_ahmed | 0:791a779d6220 | 294 | tmp12 = tmp0 + tmp2; |
shoaib_ahmed | 0:791a779d6220 | 295 | tmp13 = tmp1 + tmp3; |
shoaib_ahmed | 0:791a779d6220 | 296 | |
shoaib_ahmed | 0:791a779d6220 | 297 | z1 = MULTIPLY(tmp12 + tmp13, FIX_1_175875602); /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 298 | /* Add fudge factor here for final descale. */ |
shoaib_ahmed | 0:791a779d6220 | 299 | z1 += ONE << (CONST_BITS+PASS1_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 300 | |
shoaib_ahmed | 0:791a779d6220 | 301 | tmp12 = MULTIPLY(tmp12, - FIX_0_390180644); /* -c3+c5 */ |
shoaib_ahmed | 0:791a779d6220 | 302 | tmp13 = MULTIPLY(tmp13, - FIX_1_961570560); /* -c3-c5 */ |
shoaib_ahmed | 0:791a779d6220 | 303 | tmp12 += z1; |
shoaib_ahmed | 0:791a779d6220 | 304 | tmp13 += z1; |
shoaib_ahmed | 0:791a779d6220 | 305 | |
shoaib_ahmed | 0:791a779d6220 | 306 | z1 = MULTIPLY(tmp0 + tmp3, - FIX_0_899976223); /* -c3+c7 */ |
shoaib_ahmed | 0:791a779d6220 | 307 | tmp0 = MULTIPLY(tmp0, FIX_1_501321110); /* c1+c3-c5-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 308 | tmp3 = MULTIPLY(tmp3, FIX_0_298631336); /* -c1+c3+c5-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 309 | tmp0 += z1 + tmp12; |
shoaib_ahmed | 0:791a779d6220 | 310 | tmp3 += z1 + tmp13; |
shoaib_ahmed | 0:791a779d6220 | 311 | |
shoaib_ahmed | 0:791a779d6220 | 312 | z1 = MULTIPLY(tmp1 + tmp2, - FIX_2_562915447); /* -c1-c3 */ |
shoaib_ahmed | 0:791a779d6220 | 313 | tmp1 = MULTIPLY(tmp1, FIX_3_072711026); /* c1+c3+c5-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 314 | tmp2 = MULTIPLY(tmp2, FIX_2_053119869); /* c1+c3-c5+c7 */ |
shoaib_ahmed | 0:791a779d6220 | 315 | tmp1 += z1 + tmp13; |
shoaib_ahmed | 0:791a779d6220 | 316 | tmp2 += z1 + tmp12; |
shoaib_ahmed | 0:791a779d6220 | 317 | |
shoaib_ahmed | 0:791a779d6220 | 318 | dataptr[DCTSIZE*1] = (DCTELEM) RIGHT_SHIFT(tmp0, CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 319 | dataptr[DCTSIZE*3] = (DCTELEM) RIGHT_SHIFT(tmp1, CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 320 | dataptr[DCTSIZE*5] = (DCTELEM) RIGHT_SHIFT(tmp2, CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 321 | dataptr[DCTSIZE*7] = (DCTELEM) RIGHT_SHIFT(tmp3, CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 322 | |
shoaib_ahmed | 0:791a779d6220 | 323 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 324 | } |
shoaib_ahmed | 0:791a779d6220 | 325 | } |
shoaib_ahmed | 0:791a779d6220 | 326 | |
shoaib_ahmed | 0:791a779d6220 | 327 | #ifdef DCT_SCALING_SUPPORTED |
shoaib_ahmed | 0:791a779d6220 | 328 | |
shoaib_ahmed | 0:791a779d6220 | 329 | |
shoaib_ahmed | 0:791a779d6220 | 330 | /* |
shoaib_ahmed | 0:791a779d6220 | 331 | * Perform the forward DCT on a 7x7 sample block. |
shoaib_ahmed | 0:791a779d6220 | 332 | */ |
shoaib_ahmed | 0:791a779d6220 | 333 | |
shoaib_ahmed | 0:791a779d6220 | 334 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 335 | jpeg_fdct_7x7 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 336 | { |
shoaib_ahmed | 0:791a779d6220 | 337 | INT32 tmp0, tmp1, tmp2, tmp3; |
shoaib_ahmed | 0:791a779d6220 | 338 | INT32 tmp10, tmp11, tmp12; |
shoaib_ahmed | 0:791a779d6220 | 339 | INT32 z1, z2, z3; |
shoaib_ahmed | 0:791a779d6220 | 340 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 341 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 342 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 343 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 344 | |
shoaib_ahmed | 0:791a779d6220 | 345 | /* Pre-zero output coefficient block. */ |
shoaib_ahmed | 0:791a779d6220 | 346 | MEMZERO(data, SIZEOF(DCTELEM) * DCTSIZE2); |
shoaib_ahmed | 0:791a779d6220 | 347 | |
shoaib_ahmed | 0:791a779d6220 | 348 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 349 | * Note results are scaled up by sqrt(8) compared to a true DCT; |
shoaib_ahmed | 0:791a779d6220 | 350 | * furthermore, we scale the results by 2**PASS1_BITS. |
shoaib_ahmed | 0:791a779d6220 | 351 | * cK represents sqrt(2) * cos(K*pi/14). |
shoaib_ahmed | 0:791a779d6220 | 352 | */ |
shoaib_ahmed | 0:791a779d6220 | 353 | |
shoaib_ahmed | 0:791a779d6220 | 354 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 355 | for (ctr = 0; ctr < 7; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 356 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 357 | |
shoaib_ahmed | 0:791a779d6220 | 358 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 359 | |
shoaib_ahmed | 0:791a779d6220 | 360 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[6]); |
shoaib_ahmed | 0:791a779d6220 | 361 | tmp1 = GETJSAMPLE(elemptr[1]) + GETJSAMPLE(elemptr[5]); |
shoaib_ahmed | 0:791a779d6220 | 362 | tmp2 = GETJSAMPLE(elemptr[2]) + GETJSAMPLE(elemptr[4]); |
shoaib_ahmed | 0:791a779d6220 | 363 | tmp3 = GETJSAMPLE(elemptr[3]); |
shoaib_ahmed | 0:791a779d6220 | 364 | |
shoaib_ahmed | 0:791a779d6220 | 365 | tmp10 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[6]); |
shoaib_ahmed | 0:791a779d6220 | 366 | tmp11 = GETJSAMPLE(elemptr[1]) - GETJSAMPLE(elemptr[5]); |
shoaib_ahmed | 0:791a779d6220 | 367 | tmp12 = GETJSAMPLE(elemptr[2]) - GETJSAMPLE(elemptr[4]); |
shoaib_ahmed | 0:791a779d6220 | 368 | |
shoaib_ahmed | 0:791a779d6220 | 369 | z1 = tmp0 + tmp2; |
shoaib_ahmed | 0:791a779d6220 | 370 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 371 | dataptr[0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 372 | ((z1 + tmp1 + tmp3 - 7 * CENTERJSAMPLE) << PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 373 | tmp3 += tmp3; |
shoaib_ahmed | 0:791a779d6220 | 374 | z1 -= tmp3; |
shoaib_ahmed | 0:791a779d6220 | 375 | z1 -= tmp3; |
shoaib_ahmed | 0:791a779d6220 | 376 | z1 = MULTIPLY(z1, FIX(0.353553391)); /* (c2+c6-c4)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 377 | z2 = MULTIPLY(tmp0 - tmp2, FIX(0.920609002)); /* (c2+c4-c6)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 378 | z3 = MULTIPLY(tmp1 - tmp2, FIX(0.314692123)); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 379 | dataptr[2] = (DCTELEM) DESCALE(z1 + z2 + z3, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 380 | z1 -= z2; |
shoaib_ahmed | 0:791a779d6220 | 381 | z2 = MULTIPLY(tmp0 - tmp1, FIX(0.881747734)); /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 382 | dataptr[4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 383 | DESCALE(z2 + z3 - MULTIPLY(tmp1 - tmp3, FIX(0.707106781)), /* c2+c6-c4 */ |
shoaib_ahmed | 0:791a779d6220 | 384 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 385 | dataptr[6] = (DCTELEM) DESCALE(z1 + z2, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 386 | |
shoaib_ahmed | 0:791a779d6220 | 387 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 388 | |
shoaib_ahmed | 0:791a779d6220 | 389 | tmp1 = MULTIPLY(tmp10 + tmp11, FIX(0.935414347)); /* (c3+c1-c5)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 390 | tmp2 = MULTIPLY(tmp10 - tmp11, FIX(0.170262339)); /* (c3+c5-c1)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 391 | tmp0 = tmp1 - tmp2; |
shoaib_ahmed | 0:791a779d6220 | 392 | tmp1 += tmp2; |
shoaib_ahmed | 0:791a779d6220 | 393 | tmp2 = MULTIPLY(tmp11 + tmp12, - FIX(1.378756276)); /* -c1 */ |
shoaib_ahmed | 0:791a779d6220 | 394 | tmp1 += tmp2; |
shoaib_ahmed | 0:791a779d6220 | 395 | tmp3 = MULTIPLY(tmp10 + tmp12, FIX(0.613604268)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 396 | tmp0 += tmp3; |
shoaib_ahmed | 0:791a779d6220 | 397 | tmp2 += tmp3 + MULTIPLY(tmp12, FIX(1.870828693)); /* c3+c1-c5 */ |
shoaib_ahmed | 0:791a779d6220 | 398 | |
shoaib_ahmed | 0:791a779d6220 | 399 | dataptr[1] = (DCTELEM) DESCALE(tmp0, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 400 | dataptr[3] = (DCTELEM) DESCALE(tmp1, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 401 | dataptr[5] = (DCTELEM) DESCALE(tmp2, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 402 | |
shoaib_ahmed | 0:791a779d6220 | 403 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 404 | } |
shoaib_ahmed | 0:791a779d6220 | 405 | |
shoaib_ahmed | 0:791a779d6220 | 406 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 407 | * We remove the PASS1_BITS scaling, but leave the results scaled up |
shoaib_ahmed | 0:791a779d6220 | 408 | * by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 409 | * We must also scale the output by (8/7)**2 = 64/49, which we fold |
shoaib_ahmed | 0:791a779d6220 | 410 | * into the constant multipliers: |
shoaib_ahmed | 0:791a779d6220 | 411 | * cK now represents sqrt(2) * cos(K*pi/14) * 64/49. |
shoaib_ahmed | 0:791a779d6220 | 412 | */ |
shoaib_ahmed | 0:791a779d6220 | 413 | |
shoaib_ahmed | 0:791a779d6220 | 414 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 415 | for (ctr = 0; ctr < 7; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 416 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 417 | |
shoaib_ahmed | 0:791a779d6220 | 418 | tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 419 | tmp1 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 420 | tmp2 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 421 | tmp3 = dataptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 422 | |
shoaib_ahmed | 0:791a779d6220 | 423 | tmp10 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 424 | tmp11 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 425 | tmp12 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 426 | |
shoaib_ahmed | 0:791a779d6220 | 427 | z1 = tmp0 + tmp2; |
shoaib_ahmed | 0:791a779d6220 | 428 | dataptr[DCTSIZE*0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 429 | DESCALE(MULTIPLY(z1 + tmp1 + tmp3, FIX(1.306122449)), /* 64/49 */ |
shoaib_ahmed | 0:791a779d6220 | 430 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 431 | tmp3 += tmp3; |
shoaib_ahmed | 0:791a779d6220 | 432 | z1 -= tmp3; |
shoaib_ahmed | 0:791a779d6220 | 433 | z1 -= tmp3; |
shoaib_ahmed | 0:791a779d6220 | 434 | z1 = MULTIPLY(z1, FIX(0.461784020)); /* (c2+c6-c4)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 435 | z2 = MULTIPLY(tmp0 - tmp2, FIX(1.202428084)); /* (c2+c4-c6)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 436 | z3 = MULTIPLY(tmp1 - tmp2, FIX(0.411026446)); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 437 | dataptr[DCTSIZE*2] = (DCTELEM) DESCALE(z1 + z2 + z3, CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 438 | z1 -= z2; |
shoaib_ahmed | 0:791a779d6220 | 439 | z2 = MULTIPLY(tmp0 - tmp1, FIX(1.151670509)); /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 440 | dataptr[DCTSIZE*4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 441 | DESCALE(z2 + z3 - MULTIPLY(tmp1 - tmp3, FIX(0.923568041)), /* c2+c6-c4 */ |
shoaib_ahmed | 0:791a779d6220 | 442 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 443 | dataptr[DCTSIZE*6] = (DCTELEM) DESCALE(z1 + z2, CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 444 | |
shoaib_ahmed | 0:791a779d6220 | 445 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 446 | |
shoaib_ahmed | 0:791a779d6220 | 447 | tmp1 = MULTIPLY(tmp10 + tmp11, FIX(1.221765677)); /* (c3+c1-c5)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 448 | tmp2 = MULTIPLY(tmp10 - tmp11, FIX(0.222383464)); /* (c3+c5-c1)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 449 | tmp0 = tmp1 - tmp2; |
shoaib_ahmed | 0:791a779d6220 | 450 | tmp1 += tmp2; |
shoaib_ahmed | 0:791a779d6220 | 451 | tmp2 = MULTIPLY(tmp11 + tmp12, - FIX(1.800824523)); /* -c1 */ |
shoaib_ahmed | 0:791a779d6220 | 452 | tmp1 += tmp2; |
shoaib_ahmed | 0:791a779d6220 | 453 | tmp3 = MULTIPLY(tmp10 + tmp12, FIX(0.801442310)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 454 | tmp0 += tmp3; |
shoaib_ahmed | 0:791a779d6220 | 455 | tmp2 += tmp3 + MULTIPLY(tmp12, FIX(2.443531355)); /* c3+c1-c5 */ |
shoaib_ahmed | 0:791a779d6220 | 456 | |
shoaib_ahmed | 0:791a779d6220 | 457 | dataptr[DCTSIZE*1] = (DCTELEM) DESCALE(tmp0, CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 458 | dataptr[DCTSIZE*3] = (DCTELEM) DESCALE(tmp1, CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 459 | dataptr[DCTSIZE*5] = (DCTELEM) DESCALE(tmp2, CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 460 | |
shoaib_ahmed | 0:791a779d6220 | 461 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 462 | } |
shoaib_ahmed | 0:791a779d6220 | 463 | } |
shoaib_ahmed | 0:791a779d6220 | 464 | |
shoaib_ahmed | 0:791a779d6220 | 465 | |
shoaib_ahmed | 0:791a779d6220 | 466 | /* |
shoaib_ahmed | 0:791a779d6220 | 467 | * Perform the forward DCT on a 6x6 sample block. |
shoaib_ahmed | 0:791a779d6220 | 468 | */ |
shoaib_ahmed | 0:791a779d6220 | 469 | |
shoaib_ahmed | 0:791a779d6220 | 470 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 471 | jpeg_fdct_6x6 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 472 | { |
shoaib_ahmed | 0:791a779d6220 | 473 | INT32 tmp0, tmp1, tmp2; |
shoaib_ahmed | 0:791a779d6220 | 474 | INT32 tmp10, tmp11, tmp12; |
shoaib_ahmed | 0:791a779d6220 | 475 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 476 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 477 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 478 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 479 | |
shoaib_ahmed | 0:791a779d6220 | 480 | /* Pre-zero output coefficient block. */ |
shoaib_ahmed | 0:791a779d6220 | 481 | MEMZERO(data, SIZEOF(DCTELEM) * DCTSIZE2); |
shoaib_ahmed | 0:791a779d6220 | 482 | |
shoaib_ahmed | 0:791a779d6220 | 483 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 484 | * Note results are scaled up by sqrt(8) compared to a true DCT; |
shoaib_ahmed | 0:791a779d6220 | 485 | * furthermore, we scale the results by 2**PASS1_BITS. |
shoaib_ahmed | 0:791a779d6220 | 486 | * cK represents sqrt(2) * cos(K*pi/12). |
shoaib_ahmed | 0:791a779d6220 | 487 | */ |
shoaib_ahmed | 0:791a779d6220 | 488 | |
shoaib_ahmed | 0:791a779d6220 | 489 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 490 | for (ctr = 0; ctr < 6; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 491 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 492 | |
shoaib_ahmed | 0:791a779d6220 | 493 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 494 | |
shoaib_ahmed | 0:791a779d6220 | 495 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[5]); |
shoaib_ahmed | 0:791a779d6220 | 496 | tmp11 = GETJSAMPLE(elemptr[1]) + GETJSAMPLE(elemptr[4]); |
shoaib_ahmed | 0:791a779d6220 | 497 | tmp2 = GETJSAMPLE(elemptr[2]) + GETJSAMPLE(elemptr[3]); |
shoaib_ahmed | 0:791a779d6220 | 498 | |
shoaib_ahmed | 0:791a779d6220 | 499 | tmp10 = tmp0 + tmp2; |
shoaib_ahmed | 0:791a779d6220 | 500 | tmp12 = tmp0 - tmp2; |
shoaib_ahmed | 0:791a779d6220 | 501 | |
shoaib_ahmed | 0:791a779d6220 | 502 | tmp0 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[5]); |
shoaib_ahmed | 0:791a779d6220 | 503 | tmp1 = GETJSAMPLE(elemptr[1]) - GETJSAMPLE(elemptr[4]); |
shoaib_ahmed | 0:791a779d6220 | 504 | tmp2 = GETJSAMPLE(elemptr[2]) - GETJSAMPLE(elemptr[3]); |
shoaib_ahmed | 0:791a779d6220 | 505 | |
shoaib_ahmed | 0:791a779d6220 | 506 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 507 | dataptr[0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 508 | ((tmp10 + tmp11 - 6 * CENTERJSAMPLE) << PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 509 | dataptr[2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 510 | DESCALE(MULTIPLY(tmp12, FIX(1.224744871)), /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 511 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 512 | dataptr[4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 513 | DESCALE(MULTIPLY(tmp10 - tmp11 - tmp11, FIX(0.707106781)), /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 514 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 515 | |
shoaib_ahmed | 0:791a779d6220 | 516 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 517 | |
shoaib_ahmed | 0:791a779d6220 | 518 | tmp10 = DESCALE(MULTIPLY(tmp0 + tmp2, FIX(0.366025404)), /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 519 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 520 | |
shoaib_ahmed | 0:791a779d6220 | 521 | dataptr[1] = (DCTELEM) (tmp10 + ((tmp0 + tmp1) << PASS1_BITS)); |
shoaib_ahmed | 0:791a779d6220 | 522 | dataptr[3] = (DCTELEM) ((tmp0 - tmp1 - tmp2) << PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 523 | dataptr[5] = (DCTELEM) (tmp10 + ((tmp2 - tmp1) << PASS1_BITS)); |
shoaib_ahmed | 0:791a779d6220 | 524 | |
shoaib_ahmed | 0:791a779d6220 | 525 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 526 | } |
shoaib_ahmed | 0:791a779d6220 | 527 | |
shoaib_ahmed | 0:791a779d6220 | 528 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 529 | * We remove the PASS1_BITS scaling, but leave the results scaled up |
shoaib_ahmed | 0:791a779d6220 | 530 | * by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 531 | * We must also scale the output by (8/6)**2 = 16/9, which we fold |
shoaib_ahmed | 0:791a779d6220 | 532 | * into the constant multipliers: |
shoaib_ahmed | 0:791a779d6220 | 533 | * cK now represents sqrt(2) * cos(K*pi/12) * 16/9. |
shoaib_ahmed | 0:791a779d6220 | 534 | */ |
shoaib_ahmed | 0:791a779d6220 | 535 | |
shoaib_ahmed | 0:791a779d6220 | 536 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 537 | for (ctr = 0; ctr < 6; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 538 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 539 | |
shoaib_ahmed | 0:791a779d6220 | 540 | tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 541 | tmp11 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 542 | tmp2 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 543 | |
shoaib_ahmed | 0:791a779d6220 | 544 | tmp10 = tmp0 + tmp2; |
shoaib_ahmed | 0:791a779d6220 | 545 | tmp12 = tmp0 - tmp2; |
shoaib_ahmed | 0:791a779d6220 | 546 | |
shoaib_ahmed | 0:791a779d6220 | 547 | tmp0 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 548 | tmp1 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 549 | tmp2 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 550 | |
shoaib_ahmed | 0:791a779d6220 | 551 | dataptr[DCTSIZE*0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 552 | DESCALE(MULTIPLY(tmp10 + tmp11, FIX(1.777777778)), /* 16/9 */ |
shoaib_ahmed | 0:791a779d6220 | 553 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 554 | dataptr[DCTSIZE*2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 555 | DESCALE(MULTIPLY(tmp12, FIX(2.177324216)), /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 556 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 557 | dataptr[DCTSIZE*4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 558 | DESCALE(MULTIPLY(tmp10 - tmp11 - tmp11, FIX(1.257078722)), /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 559 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 560 | |
shoaib_ahmed | 0:791a779d6220 | 561 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 562 | |
shoaib_ahmed | 0:791a779d6220 | 563 | tmp10 = MULTIPLY(tmp0 + tmp2, FIX(0.650711829)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 564 | |
shoaib_ahmed | 0:791a779d6220 | 565 | dataptr[DCTSIZE*1] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 566 | DESCALE(tmp10 + MULTIPLY(tmp0 + tmp1, FIX(1.777777778)), /* 16/9 */ |
shoaib_ahmed | 0:791a779d6220 | 567 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 568 | dataptr[DCTSIZE*3] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 569 | DESCALE(MULTIPLY(tmp0 - tmp1 - tmp2, FIX(1.777777778)), /* 16/9 */ |
shoaib_ahmed | 0:791a779d6220 | 570 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 571 | dataptr[DCTSIZE*5] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 572 | DESCALE(tmp10 + MULTIPLY(tmp2 - tmp1, FIX(1.777777778)), /* 16/9 */ |
shoaib_ahmed | 0:791a779d6220 | 573 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 574 | |
shoaib_ahmed | 0:791a779d6220 | 575 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 576 | } |
shoaib_ahmed | 0:791a779d6220 | 577 | } |
shoaib_ahmed | 0:791a779d6220 | 578 | |
shoaib_ahmed | 0:791a779d6220 | 579 | |
shoaib_ahmed | 0:791a779d6220 | 580 | /* |
shoaib_ahmed | 0:791a779d6220 | 581 | * Perform the forward DCT on a 5x5 sample block. |
shoaib_ahmed | 0:791a779d6220 | 582 | */ |
shoaib_ahmed | 0:791a779d6220 | 583 | |
shoaib_ahmed | 0:791a779d6220 | 584 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 585 | jpeg_fdct_5x5 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 586 | { |
shoaib_ahmed | 0:791a779d6220 | 587 | INT32 tmp0, tmp1, tmp2; |
shoaib_ahmed | 0:791a779d6220 | 588 | INT32 tmp10, tmp11; |
shoaib_ahmed | 0:791a779d6220 | 589 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 590 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 591 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 592 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 593 | |
shoaib_ahmed | 0:791a779d6220 | 594 | /* Pre-zero output coefficient block. */ |
shoaib_ahmed | 0:791a779d6220 | 595 | MEMZERO(data, SIZEOF(DCTELEM) * DCTSIZE2); |
shoaib_ahmed | 0:791a779d6220 | 596 | |
shoaib_ahmed | 0:791a779d6220 | 597 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 598 | * Note results are scaled up by sqrt(8) compared to a true DCT; |
shoaib_ahmed | 0:791a779d6220 | 599 | * furthermore, we scale the results by 2**PASS1_BITS. |
shoaib_ahmed | 0:791a779d6220 | 600 | * We scale the results further by 2 as part of output adaption |
shoaib_ahmed | 0:791a779d6220 | 601 | * scaling for different DCT size. |
shoaib_ahmed | 0:791a779d6220 | 602 | * cK represents sqrt(2) * cos(K*pi/10). |
shoaib_ahmed | 0:791a779d6220 | 603 | */ |
shoaib_ahmed | 0:791a779d6220 | 604 | |
shoaib_ahmed | 0:791a779d6220 | 605 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 606 | for (ctr = 0; ctr < 5; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 607 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 608 | |
shoaib_ahmed | 0:791a779d6220 | 609 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 610 | |
shoaib_ahmed | 0:791a779d6220 | 611 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[4]); |
shoaib_ahmed | 0:791a779d6220 | 612 | tmp1 = GETJSAMPLE(elemptr[1]) + GETJSAMPLE(elemptr[3]); |
shoaib_ahmed | 0:791a779d6220 | 613 | tmp2 = GETJSAMPLE(elemptr[2]); |
shoaib_ahmed | 0:791a779d6220 | 614 | |
shoaib_ahmed | 0:791a779d6220 | 615 | tmp10 = tmp0 + tmp1; |
shoaib_ahmed | 0:791a779d6220 | 616 | tmp11 = tmp0 - tmp1; |
shoaib_ahmed | 0:791a779d6220 | 617 | |
shoaib_ahmed | 0:791a779d6220 | 618 | tmp0 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[4]); |
shoaib_ahmed | 0:791a779d6220 | 619 | tmp1 = GETJSAMPLE(elemptr[1]) - GETJSAMPLE(elemptr[3]); |
shoaib_ahmed | 0:791a779d6220 | 620 | |
shoaib_ahmed | 0:791a779d6220 | 621 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 622 | dataptr[0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 623 | ((tmp10 + tmp2 - 5 * CENTERJSAMPLE) << (PASS1_BITS+1)); |
shoaib_ahmed | 0:791a779d6220 | 624 | tmp11 = MULTIPLY(tmp11, FIX(0.790569415)); /* (c2+c4)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 625 | tmp10 -= tmp2 << 2; |
shoaib_ahmed | 0:791a779d6220 | 626 | tmp10 = MULTIPLY(tmp10, FIX(0.353553391)); /* (c2-c4)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 627 | dataptr[2] = (DCTELEM) DESCALE(tmp11 + tmp10, CONST_BITS-PASS1_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 628 | dataptr[4] = (DCTELEM) DESCALE(tmp11 - tmp10, CONST_BITS-PASS1_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 629 | |
shoaib_ahmed | 0:791a779d6220 | 630 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 631 | |
shoaib_ahmed | 0:791a779d6220 | 632 | tmp10 = MULTIPLY(tmp0 + tmp1, FIX(0.831253876)); /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 633 | |
shoaib_ahmed | 0:791a779d6220 | 634 | dataptr[1] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 635 | DESCALE(tmp10 + MULTIPLY(tmp0, FIX(0.513743148)), /* c1-c3 */ |
shoaib_ahmed | 0:791a779d6220 | 636 | CONST_BITS-PASS1_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 637 | dataptr[3] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 638 | DESCALE(tmp10 - MULTIPLY(tmp1, FIX(2.176250899)), /* c1+c3 */ |
shoaib_ahmed | 0:791a779d6220 | 639 | CONST_BITS-PASS1_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 640 | |
shoaib_ahmed | 0:791a779d6220 | 641 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 642 | } |
shoaib_ahmed | 0:791a779d6220 | 643 | |
shoaib_ahmed | 0:791a779d6220 | 644 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 645 | * We remove the PASS1_BITS scaling, but leave the results scaled up |
shoaib_ahmed | 0:791a779d6220 | 646 | * by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 647 | * We must also scale the output by (8/5)**2 = 64/25, which we partially |
shoaib_ahmed | 0:791a779d6220 | 648 | * fold into the constant multipliers (other part was done in pass 1): |
shoaib_ahmed | 0:791a779d6220 | 649 | * cK now represents sqrt(2) * cos(K*pi/10) * 32/25. |
shoaib_ahmed | 0:791a779d6220 | 650 | */ |
shoaib_ahmed | 0:791a779d6220 | 651 | |
shoaib_ahmed | 0:791a779d6220 | 652 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 653 | for (ctr = 0; ctr < 5; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 654 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 655 | |
shoaib_ahmed | 0:791a779d6220 | 656 | tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 657 | tmp1 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 658 | tmp2 = dataptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 659 | |
shoaib_ahmed | 0:791a779d6220 | 660 | tmp10 = tmp0 + tmp1; |
shoaib_ahmed | 0:791a779d6220 | 661 | tmp11 = tmp0 - tmp1; |
shoaib_ahmed | 0:791a779d6220 | 662 | |
shoaib_ahmed | 0:791a779d6220 | 663 | tmp0 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 664 | tmp1 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 665 | |
shoaib_ahmed | 0:791a779d6220 | 666 | dataptr[DCTSIZE*0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 667 | DESCALE(MULTIPLY(tmp10 + tmp2, FIX(1.28)), /* 32/25 */ |
shoaib_ahmed | 0:791a779d6220 | 668 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 669 | tmp11 = MULTIPLY(tmp11, FIX(1.011928851)); /* (c2+c4)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 670 | tmp10 -= tmp2 << 2; |
shoaib_ahmed | 0:791a779d6220 | 671 | tmp10 = MULTIPLY(tmp10, FIX(0.452548340)); /* (c2-c4)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 672 | dataptr[DCTSIZE*2] = (DCTELEM) DESCALE(tmp11 + tmp10, CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 673 | dataptr[DCTSIZE*4] = (DCTELEM) DESCALE(tmp11 - tmp10, CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 674 | |
shoaib_ahmed | 0:791a779d6220 | 675 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 676 | |
shoaib_ahmed | 0:791a779d6220 | 677 | tmp10 = MULTIPLY(tmp0 + tmp1, FIX(1.064004961)); /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 678 | |
shoaib_ahmed | 0:791a779d6220 | 679 | dataptr[DCTSIZE*1] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 680 | DESCALE(tmp10 + MULTIPLY(tmp0, FIX(0.657591230)), /* c1-c3 */ |
shoaib_ahmed | 0:791a779d6220 | 681 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 682 | dataptr[DCTSIZE*3] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 683 | DESCALE(tmp10 - MULTIPLY(tmp1, FIX(2.785601151)), /* c1+c3 */ |
shoaib_ahmed | 0:791a779d6220 | 684 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 685 | |
shoaib_ahmed | 0:791a779d6220 | 686 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 687 | } |
shoaib_ahmed | 0:791a779d6220 | 688 | } |
shoaib_ahmed | 0:791a779d6220 | 689 | |
shoaib_ahmed | 0:791a779d6220 | 690 | |
shoaib_ahmed | 0:791a779d6220 | 691 | /* |
shoaib_ahmed | 0:791a779d6220 | 692 | * Perform the forward DCT on a 4x4 sample block. |
shoaib_ahmed | 0:791a779d6220 | 693 | */ |
shoaib_ahmed | 0:791a779d6220 | 694 | |
shoaib_ahmed | 0:791a779d6220 | 695 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 696 | jpeg_fdct_4x4 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 697 | { |
shoaib_ahmed | 0:791a779d6220 | 698 | INT32 tmp0, tmp1; |
shoaib_ahmed | 0:791a779d6220 | 699 | INT32 tmp10, tmp11; |
shoaib_ahmed | 0:791a779d6220 | 700 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 701 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 702 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 703 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 704 | |
shoaib_ahmed | 0:791a779d6220 | 705 | /* Pre-zero output coefficient block. */ |
shoaib_ahmed | 0:791a779d6220 | 706 | MEMZERO(data, SIZEOF(DCTELEM) * DCTSIZE2); |
shoaib_ahmed | 0:791a779d6220 | 707 | |
shoaib_ahmed | 0:791a779d6220 | 708 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 709 | * Note results are scaled up by sqrt(8) compared to a true DCT; |
shoaib_ahmed | 0:791a779d6220 | 710 | * furthermore, we scale the results by 2**PASS1_BITS. |
shoaib_ahmed | 0:791a779d6220 | 711 | * We must also scale the output by (8/4)**2 = 2**2, which we add here. |
shoaib_ahmed | 0:791a779d6220 | 712 | * cK represents sqrt(2) * cos(K*pi/16) [refers to 8-point FDCT]. |
shoaib_ahmed | 0:791a779d6220 | 713 | */ |
shoaib_ahmed | 0:791a779d6220 | 714 | |
shoaib_ahmed | 0:791a779d6220 | 715 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 716 | for (ctr = 0; ctr < 4; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 717 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 718 | |
shoaib_ahmed | 0:791a779d6220 | 719 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 720 | |
shoaib_ahmed | 0:791a779d6220 | 721 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[3]); |
shoaib_ahmed | 0:791a779d6220 | 722 | tmp1 = GETJSAMPLE(elemptr[1]) + GETJSAMPLE(elemptr[2]); |
shoaib_ahmed | 0:791a779d6220 | 723 | |
shoaib_ahmed | 0:791a779d6220 | 724 | tmp10 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[3]); |
shoaib_ahmed | 0:791a779d6220 | 725 | tmp11 = GETJSAMPLE(elemptr[1]) - GETJSAMPLE(elemptr[2]); |
shoaib_ahmed | 0:791a779d6220 | 726 | |
shoaib_ahmed | 0:791a779d6220 | 727 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 728 | dataptr[0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 729 | ((tmp0 + tmp1 - 4 * CENTERJSAMPLE) << (PASS1_BITS+2)); |
shoaib_ahmed | 0:791a779d6220 | 730 | dataptr[2] = (DCTELEM) ((tmp0 - tmp1) << (PASS1_BITS+2)); |
shoaib_ahmed | 0:791a779d6220 | 731 | |
shoaib_ahmed | 0:791a779d6220 | 732 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 733 | |
shoaib_ahmed | 0:791a779d6220 | 734 | tmp0 = MULTIPLY(tmp10 + tmp11, FIX_0_541196100); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 735 | /* Add fudge factor here for final descale. */ |
shoaib_ahmed | 0:791a779d6220 | 736 | tmp0 += ONE << (CONST_BITS-PASS1_BITS-3); |
shoaib_ahmed | 0:791a779d6220 | 737 | |
shoaib_ahmed | 0:791a779d6220 | 738 | dataptr[1] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 739 | RIGHT_SHIFT(tmp0 + MULTIPLY(tmp10, FIX_0_765366865), /* c2-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 740 | CONST_BITS-PASS1_BITS-2); |
shoaib_ahmed | 0:791a779d6220 | 741 | dataptr[3] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 742 | RIGHT_SHIFT(tmp0 - MULTIPLY(tmp11, FIX_1_847759065), /* c2+c6 */ |
shoaib_ahmed | 0:791a779d6220 | 743 | CONST_BITS-PASS1_BITS-2); |
shoaib_ahmed | 0:791a779d6220 | 744 | |
shoaib_ahmed | 0:791a779d6220 | 745 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 746 | } |
shoaib_ahmed | 0:791a779d6220 | 747 | |
shoaib_ahmed | 0:791a779d6220 | 748 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 749 | * We remove the PASS1_BITS scaling, but leave the results scaled up |
shoaib_ahmed | 0:791a779d6220 | 750 | * by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 751 | * cK represents sqrt(2) * cos(K*pi/16) [refers to 8-point FDCT]. |
shoaib_ahmed | 0:791a779d6220 | 752 | */ |
shoaib_ahmed | 0:791a779d6220 | 753 | |
shoaib_ahmed | 0:791a779d6220 | 754 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 755 | for (ctr = 0; ctr < 4; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 756 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 757 | |
shoaib_ahmed | 0:791a779d6220 | 758 | /* Add fudge factor here for final descale. */ |
shoaib_ahmed | 0:791a779d6220 | 759 | tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*3] + (ONE << (PASS1_BITS-1)); |
shoaib_ahmed | 0:791a779d6220 | 760 | tmp1 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 761 | |
shoaib_ahmed | 0:791a779d6220 | 762 | tmp10 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 763 | tmp11 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 764 | |
shoaib_ahmed | 0:791a779d6220 | 765 | dataptr[DCTSIZE*0] = (DCTELEM) RIGHT_SHIFT(tmp0 + tmp1, PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 766 | dataptr[DCTSIZE*2] = (DCTELEM) RIGHT_SHIFT(tmp0 - tmp1, PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 767 | |
shoaib_ahmed | 0:791a779d6220 | 768 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 769 | |
shoaib_ahmed | 0:791a779d6220 | 770 | tmp0 = MULTIPLY(tmp10 + tmp11, FIX_0_541196100); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 771 | /* Add fudge factor here for final descale. */ |
shoaib_ahmed | 0:791a779d6220 | 772 | tmp0 += ONE << (CONST_BITS+PASS1_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 773 | |
shoaib_ahmed | 0:791a779d6220 | 774 | dataptr[DCTSIZE*1] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 775 | RIGHT_SHIFT(tmp0 + MULTIPLY(tmp10, FIX_0_765366865), /* c2-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 776 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 777 | dataptr[DCTSIZE*3] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 778 | RIGHT_SHIFT(tmp0 - MULTIPLY(tmp11, FIX_1_847759065), /* c2+c6 */ |
shoaib_ahmed | 0:791a779d6220 | 779 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 780 | |
shoaib_ahmed | 0:791a779d6220 | 781 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 782 | } |
shoaib_ahmed | 0:791a779d6220 | 783 | } |
shoaib_ahmed | 0:791a779d6220 | 784 | |
shoaib_ahmed | 0:791a779d6220 | 785 | |
shoaib_ahmed | 0:791a779d6220 | 786 | /* |
shoaib_ahmed | 0:791a779d6220 | 787 | * Perform the forward DCT on a 3x3 sample block. |
shoaib_ahmed | 0:791a779d6220 | 788 | */ |
shoaib_ahmed | 0:791a779d6220 | 789 | |
shoaib_ahmed | 0:791a779d6220 | 790 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 791 | jpeg_fdct_3x3 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 792 | { |
shoaib_ahmed | 0:791a779d6220 | 793 | INT32 tmp0, tmp1, tmp2; |
shoaib_ahmed | 0:791a779d6220 | 794 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 795 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 796 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 797 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 798 | |
shoaib_ahmed | 0:791a779d6220 | 799 | /* Pre-zero output coefficient block. */ |
shoaib_ahmed | 0:791a779d6220 | 800 | MEMZERO(data, SIZEOF(DCTELEM) * DCTSIZE2); |
shoaib_ahmed | 0:791a779d6220 | 801 | |
shoaib_ahmed | 0:791a779d6220 | 802 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 803 | * Note results are scaled up by sqrt(8) compared to a true DCT; |
shoaib_ahmed | 0:791a779d6220 | 804 | * furthermore, we scale the results by 2**PASS1_BITS. |
shoaib_ahmed | 0:791a779d6220 | 805 | * We scale the results further by 2**2 as part of output adaption |
shoaib_ahmed | 0:791a779d6220 | 806 | * scaling for different DCT size. |
shoaib_ahmed | 0:791a779d6220 | 807 | * cK represents sqrt(2) * cos(K*pi/6). |
shoaib_ahmed | 0:791a779d6220 | 808 | */ |
shoaib_ahmed | 0:791a779d6220 | 809 | |
shoaib_ahmed | 0:791a779d6220 | 810 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 811 | for (ctr = 0; ctr < 3; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 812 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 813 | |
shoaib_ahmed | 0:791a779d6220 | 814 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 815 | |
shoaib_ahmed | 0:791a779d6220 | 816 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[2]); |
shoaib_ahmed | 0:791a779d6220 | 817 | tmp1 = GETJSAMPLE(elemptr[1]); |
shoaib_ahmed | 0:791a779d6220 | 818 | |
shoaib_ahmed | 0:791a779d6220 | 819 | tmp2 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[2]); |
shoaib_ahmed | 0:791a779d6220 | 820 | |
shoaib_ahmed | 0:791a779d6220 | 821 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 822 | dataptr[0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 823 | ((tmp0 + tmp1 - 3 * CENTERJSAMPLE) << (PASS1_BITS+2)); |
shoaib_ahmed | 0:791a779d6220 | 824 | dataptr[2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 825 | DESCALE(MULTIPLY(tmp0 - tmp1 - tmp1, FIX(0.707106781)), /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 826 | CONST_BITS-PASS1_BITS-2); |
shoaib_ahmed | 0:791a779d6220 | 827 | |
shoaib_ahmed | 0:791a779d6220 | 828 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 829 | |
shoaib_ahmed | 0:791a779d6220 | 830 | dataptr[1] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 831 | DESCALE(MULTIPLY(tmp2, FIX(1.224744871)), /* c1 */ |
shoaib_ahmed | 0:791a779d6220 | 832 | CONST_BITS-PASS1_BITS-2); |
shoaib_ahmed | 0:791a779d6220 | 833 | |
shoaib_ahmed | 0:791a779d6220 | 834 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 835 | } |
shoaib_ahmed | 0:791a779d6220 | 836 | |
shoaib_ahmed | 0:791a779d6220 | 837 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 838 | * We remove the PASS1_BITS scaling, but leave the results scaled up |
shoaib_ahmed | 0:791a779d6220 | 839 | * by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 840 | * We must also scale the output by (8/3)**2 = 64/9, which we partially |
shoaib_ahmed | 0:791a779d6220 | 841 | * fold into the constant multipliers (other part was done in pass 1): |
shoaib_ahmed | 0:791a779d6220 | 842 | * cK now represents sqrt(2) * cos(K*pi/6) * 16/9. |
shoaib_ahmed | 0:791a779d6220 | 843 | */ |
shoaib_ahmed | 0:791a779d6220 | 844 | |
shoaib_ahmed | 0:791a779d6220 | 845 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 846 | for (ctr = 0; ctr < 3; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 847 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 848 | |
shoaib_ahmed | 0:791a779d6220 | 849 | tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 850 | tmp1 = dataptr[DCTSIZE*1]; |
shoaib_ahmed | 0:791a779d6220 | 851 | |
shoaib_ahmed | 0:791a779d6220 | 852 | tmp2 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 853 | |
shoaib_ahmed | 0:791a779d6220 | 854 | dataptr[DCTSIZE*0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 855 | DESCALE(MULTIPLY(tmp0 + tmp1, FIX(1.777777778)), /* 16/9 */ |
shoaib_ahmed | 0:791a779d6220 | 856 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 857 | dataptr[DCTSIZE*2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 858 | DESCALE(MULTIPLY(tmp0 - tmp1 - tmp1, FIX(1.257078722)), /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 859 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 860 | |
shoaib_ahmed | 0:791a779d6220 | 861 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 862 | |
shoaib_ahmed | 0:791a779d6220 | 863 | dataptr[DCTSIZE*1] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 864 | DESCALE(MULTIPLY(tmp2, FIX(2.177324216)), /* c1 */ |
shoaib_ahmed | 0:791a779d6220 | 865 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 866 | |
shoaib_ahmed | 0:791a779d6220 | 867 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 868 | } |
shoaib_ahmed | 0:791a779d6220 | 869 | } |
shoaib_ahmed | 0:791a779d6220 | 870 | |
shoaib_ahmed | 0:791a779d6220 | 871 | |
shoaib_ahmed | 0:791a779d6220 | 872 | /* |
shoaib_ahmed | 0:791a779d6220 | 873 | * Perform the forward DCT on a 2x2 sample block. |
shoaib_ahmed | 0:791a779d6220 | 874 | */ |
shoaib_ahmed | 0:791a779d6220 | 875 | |
shoaib_ahmed | 0:791a779d6220 | 876 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 877 | jpeg_fdct_2x2 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 878 | { |
shoaib_ahmed | 0:791a779d6220 | 879 | DCTELEM tmp0, tmp1, tmp2, tmp3; |
shoaib_ahmed | 0:791a779d6220 | 880 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 881 | |
shoaib_ahmed | 0:791a779d6220 | 882 | /* Pre-zero output coefficient block. */ |
shoaib_ahmed | 0:791a779d6220 | 883 | MEMZERO(data, SIZEOF(DCTELEM) * DCTSIZE2); |
shoaib_ahmed | 0:791a779d6220 | 884 | |
shoaib_ahmed | 0:791a779d6220 | 885 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 886 | * Note results are scaled up by sqrt(8) compared to a true DCT. |
shoaib_ahmed | 0:791a779d6220 | 887 | */ |
shoaib_ahmed | 0:791a779d6220 | 888 | |
shoaib_ahmed | 0:791a779d6220 | 889 | /* Row 0 */ |
shoaib_ahmed | 0:791a779d6220 | 890 | elemptr = sample_data[0] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 891 | |
shoaib_ahmed | 0:791a779d6220 | 892 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[1]); |
shoaib_ahmed | 0:791a779d6220 | 893 | tmp1 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[1]); |
shoaib_ahmed | 0:791a779d6220 | 894 | |
shoaib_ahmed | 0:791a779d6220 | 895 | /* Row 1 */ |
shoaib_ahmed | 0:791a779d6220 | 896 | elemptr = sample_data[1] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 897 | |
shoaib_ahmed | 0:791a779d6220 | 898 | tmp2 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[1]); |
shoaib_ahmed | 0:791a779d6220 | 899 | tmp3 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[1]); |
shoaib_ahmed | 0:791a779d6220 | 900 | |
shoaib_ahmed | 0:791a779d6220 | 901 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 902 | * We leave the results scaled up by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 903 | * We must also scale the output by (8/2)**2 = 2**4. |
shoaib_ahmed | 0:791a779d6220 | 904 | */ |
shoaib_ahmed | 0:791a779d6220 | 905 | |
shoaib_ahmed | 0:791a779d6220 | 906 | /* Column 0 */ |
shoaib_ahmed | 0:791a779d6220 | 907 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 908 | data[DCTSIZE*0] = (tmp0 + tmp2 - 4 * CENTERJSAMPLE) << 4; |
shoaib_ahmed | 0:791a779d6220 | 909 | data[DCTSIZE*1] = (tmp0 - tmp2) << 4; |
shoaib_ahmed | 0:791a779d6220 | 910 | |
shoaib_ahmed | 0:791a779d6220 | 911 | /* Column 1 */ |
shoaib_ahmed | 0:791a779d6220 | 912 | data[DCTSIZE*0+1] = (tmp1 + tmp3) << 4; |
shoaib_ahmed | 0:791a779d6220 | 913 | data[DCTSIZE*1+1] = (tmp1 - tmp3) << 4; |
shoaib_ahmed | 0:791a779d6220 | 914 | } |
shoaib_ahmed | 0:791a779d6220 | 915 | |
shoaib_ahmed | 0:791a779d6220 | 916 | |
shoaib_ahmed | 0:791a779d6220 | 917 | /* |
shoaib_ahmed | 0:791a779d6220 | 918 | * Perform the forward DCT on a 1x1 sample block. |
shoaib_ahmed | 0:791a779d6220 | 919 | */ |
shoaib_ahmed | 0:791a779d6220 | 920 | |
shoaib_ahmed | 0:791a779d6220 | 921 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 922 | jpeg_fdct_1x1 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 923 | { |
shoaib_ahmed | 0:791a779d6220 | 924 | DCTELEM dcval; |
shoaib_ahmed | 0:791a779d6220 | 925 | |
shoaib_ahmed | 0:791a779d6220 | 926 | /* Pre-zero output coefficient block. */ |
shoaib_ahmed | 0:791a779d6220 | 927 | MEMZERO(data, SIZEOF(DCTELEM) * DCTSIZE2); |
shoaib_ahmed | 0:791a779d6220 | 928 | |
shoaib_ahmed | 0:791a779d6220 | 929 | dcval = GETJSAMPLE(sample_data[0][start_col]); |
shoaib_ahmed | 0:791a779d6220 | 930 | |
shoaib_ahmed | 0:791a779d6220 | 931 | /* We leave the result scaled up by an overall factor of 8. */ |
shoaib_ahmed | 0:791a779d6220 | 932 | /* We must also scale the output by (8/1)**2 = 2**6. */ |
shoaib_ahmed | 0:791a779d6220 | 933 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 934 | data[0] = (dcval - CENTERJSAMPLE) << 6; |
shoaib_ahmed | 0:791a779d6220 | 935 | } |
shoaib_ahmed | 0:791a779d6220 | 936 | |
shoaib_ahmed | 0:791a779d6220 | 937 | |
shoaib_ahmed | 0:791a779d6220 | 938 | /* |
shoaib_ahmed | 0:791a779d6220 | 939 | * Perform the forward DCT on a 9x9 sample block. |
shoaib_ahmed | 0:791a779d6220 | 940 | */ |
shoaib_ahmed | 0:791a779d6220 | 941 | |
shoaib_ahmed | 0:791a779d6220 | 942 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 943 | jpeg_fdct_9x9 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 944 | { |
shoaib_ahmed | 0:791a779d6220 | 945 | INT32 tmp0, tmp1, tmp2, tmp3, tmp4; |
shoaib_ahmed | 0:791a779d6220 | 946 | INT32 tmp10, tmp11, tmp12, tmp13; |
shoaib_ahmed | 0:791a779d6220 | 947 | INT32 z1, z2; |
shoaib_ahmed | 0:791a779d6220 | 948 | DCTELEM workspace[8]; |
shoaib_ahmed | 0:791a779d6220 | 949 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 950 | DCTELEM *wsptr; |
shoaib_ahmed | 0:791a779d6220 | 951 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 952 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 953 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 954 | |
shoaib_ahmed | 0:791a779d6220 | 955 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 956 | * Note results are scaled up by sqrt(8) compared to a true DCT; |
shoaib_ahmed | 0:791a779d6220 | 957 | * we scale the results further by 2 as part of output adaption |
shoaib_ahmed | 0:791a779d6220 | 958 | * scaling for different DCT size. |
shoaib_ahmed | 0:791a779d6220 | 959 | * cK represents sqrt(2) * cos(K*pi/18). |
shoaib_ahmed | 0:791a779d6220 | 960 | */ |
shoaib_ahmed | 0:791a779d6220 | 961 | |
shoaib_ahmed | 0:791a779d6220 | 962 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 963 | ctr = 0; |
shoaib_ahmed | 0:791a779d6220 | 964 | for (;;) { |
shoaib_ahmed | 0:791a779d6220 | 965 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 966 | |
shoaib_ahmed | 0:791a779d6220 | 967 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 968 | |
shoaib_ahmed | 0:791a779d6220 | 969 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[8]); |
shoaib_ahmed | 0:791a779d6220 | 970 | tmp1 = GETJSAMPLE(elemptr[1]) + GETJSAMPLE(elemptr[7]); |
shoaib_ahmed | 0:791a779d6220 | 971 | tmp2 = GETJSAMPLE(elemptr[2]) + GETJSAMPLE(elemptr[6]); |
shoaib_ahmed | 0:791a779d6220 | 972 | tmp3 = GETJSAMPLE(elemptr[3]) + GETJSAMPLE(elemptr[5]); |
shoaib_ahmed | 0:791a779d6220 | 973 | tmp4 = GETJSAMPLE(elemptr[4]); |
shoaib_ahmed | 0:791a779d6220 | 974 | |
shoaib_ahmed | 0:791a779d6220 | 975 | tmp10 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[8]); |
shoaib_ahmed | 0:791a779d6220 | 976 | tmp11 = GETJSAMPLE(elemptr[1]) - GETJSAMPLE(elemptr[7]); |
shoaib_ahmed | 0:791a779d6220 | 977 | tmp12 = GETJSAMPLE(elemptr[2]) - GETJSAMPLE(elemptr[6]); |
shoaib_ahmed | 0:791a779d6220 | 978 | tmp13 = GETJSAMPLE(elemptr[3]) - GETJSAMPLE(elemptr[5]); |
shoaib_ahmed | 0:791a779d6220 | 979 | |
shoaib_ahmed | 0:791a779d6220 | 980 | z1 = tmp0 + tmp2 + tmp3; |
shoaib_ahmed | 0:791a779d6220 | 981 | z2 = tmp1 + tmp4; |
shoaib_ahmed | 0:791a779d6220 | 982 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 983 | dataptr[0] = (DCTELEM) ((z1 + z2 - 9 * CENTERJSAMPLE) << 1); |
shoaib_ahmed | 0:791a779d6220 | 984 | dataptr[6] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 985 | DESCALE(MULTIPLY(z1 - z2 - z2, FIX(0.707106781)), /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 986 | CONST_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 987 | z1 = MULTIPLY(tmp0 - tmp2, FIX(1.328926049)); /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 988 | z2 = MULTIPLY(tmp1 - tmp4 - tmp4, FIX(0.707106781)); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 989 | dataptr[2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 990 | DESCALE(MULTIPLY(tmp2 - tmp3, FIX(1.083350441)) /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 991 | + z1 + z2, CONST_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 992 | dataptr[4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 993 | DESCALE(MULTIPLY(tmp3 - tmp0, FIX(0.245575608)) /* c8 */ |
shoaib_ahmed | 0:791a779d6220 | 994 | + z1 - z2, CONST_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 995 | |
shoaib_ahmed | 0:791a779d6220 | 996 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 997 | |
shoaib_ahmed | 0:791a779d6220 | 998 | dataptr[3] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 999 | DESCALE(MULTIPLY(tmp10 - tmp12 - tmp13, FIX(1.224744871)), /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 1000 | CONST_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 1001 | |
shoaib_ahmed | 0:791a779d6220 | 1002 | tmp11 = MULTIPLY(tmp11, FIX(1.224744871)); /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 1003 | tmp0 = MULTIPLY(tmp10 + tmp12, FIX(0.909038955)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 1004 | tmp1 = MULTIPLY(tmp10 + tmp13, FIX(0.483689525)); /* c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1005 | |
shoaib_ahmed | 0:791a779d6220 | 1006 | dataptr[1] = (DCTELEM) DESCALE(tmp11 + tmp0 + tmp1, CONST_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 1007 | |
shoaib_ahmed | 0:791a779d6220 | 1008 | tmp2 = MULTIPLY(tmp12 - tmp13, FIX(1.392728481)); /* c1 */ |
shoaib_ahmed | 0:791a779d6220 | 1009 | |
shoaib_ahmed | 0:791a779d6220 | 1010 | dataptr[5] = (DCTELEM) DESCALE(tmp0 - tmp11 - tmp2, CONST_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 1011 | dataptr[7] = (DCTELEM) DESCALE(tmp1 - tmp11 + tmp2, CONST_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 1012 | |
shoaib_ahmed | 0:791a779d6220 | 1013 | ctr++; |
shoaib_ahmed | 0:791a779d6220 | 1014 | |
shoaib_ahmed | 0:791a779d6220 | 1015 | if (ctr != DCTSIZE) { |
shoaib_ahmed | 0:791a779d6220 | 1016 | if (ctr == 9) |
shoaib_ahmed | 0:791a779d6220 | 1017 | break; /* Done. */ |
shoaib_ahmed | 0:791a779d6220 | 1018 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 1019 | } else |
shoaib_ahmed | 0:791a779d6220 | 1020 | dataptr = workspace; /* switch pointer to extended workspace */ |
shoaib_ahmed | 0:791a779d6220 | 1021 | } |
shoaib_ahmed | 0:791a779d6220 | 1022 | |
shoaib_ahmed | 0:791a779d6220 | 1023 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 1024 | * We leave the results scaled up by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 1025 | * We must also scale the output by (8/9)**2 = 64/81, which we partially |
shoaib_ahmed | 0:791a779d6220 | 1026 | * fold into the constant multipliers and final/initial shifting: |
shoaib_ahmed | 0:791a779d6220 | 1027 | * cK now represents sqrt(2) * cos(K*pi/18) * 128/81. |
shoaib_ahmed | 0:791a779d6220 | 1028 | */ |
shoaib_ahmed | 0:791a779d6220 | 1029 | |
shoaib_ahmed | 0:791a779d6220 | 1030 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 1031 | wsptr = workspace; |
shoaib_ahmed | 0:791a779d6220 | 1032 | for (ctr = DCTSIZE-1; ctr >= 0; ctr--) { |
shoaib_ahmed | 0:791a779d6220 | 1033 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 1034 | |
shoaib_ahmed | 0:791a779d6220 | 1035 | tmp0 = dataptr[DCTSIZE*0] + wsptr[DCTSIZE*0]; |
shoaib_ahmed | 0:791a779d6220 | 1036 | tmp1 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 1037 | tmp2 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 1038 | tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 1039 | tmp4 = dataptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 1040 | |
shoaib_ahmed | 0:791a779d6220 | 1041 | tmp10 = dataptr[DCTSIZE*0] - wsptr[DCTSIZE*0]; |
shoaib_ahmed | 0:791a779d6220 | 1042 | tmp11 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 1043 | tmp12 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 1044 | tmp13 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 1045 | |
shoaib_ahmed | 0:791a779d6220 | 1046 | z1 = tmp0 + tmp2 + tmp3; |
shoaib_ahmed | 0:791a779d6220 | 1047 | z2 = tmp1 + tmp4; |
shoaib_ahmed | 0:791a779d6220 | 1048 | dataptr[DCTSIZE*0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1049 | DESCALE(MULTIPLY(z1 + z2, FIX(1.580246914)), /* 128/81 */ |
shoaib_ahmed | 0:791a779d6220 | 1050 | CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 1051 | dataptr[DCTSIZE*6] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1052 | DESCALE(MULTIPLY(z1 - z2 - z2, FIX(1.117403309)), /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 1053 | CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 1054 | z1 = MULTIPLY(tmp0 - tmp2, FIX(2.100031287)); /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 1055 | z2 = MULTIPLY(tmp1 - tmp4 - tmp4, FIX(1.117403309)); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 1056 | dataptr[DCTSIZE*2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1057 | DESCALE(MULTIPLY(tmp2 - tmp3, FIX(1.711961190)) /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 1058 | + z1 + z2, CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 1059 | dataptr[DCTSIZE*4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1060 | DESCALE(MULTIPLY(tmp3 - tmp0, FIX(0.388070096)) /* c8 */ |
shoaib_ahmed | 0:791a779d6220 | 1061 | + z1 - z2, CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 1062 | |
shoaib_ahmed | 0:791a779d6220 | 1063 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 1064 | |
shoaib_ahmed | 0:791a779d6220 | 1065 | dataptr[DCTSIZE*3] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1066 | DESCALE(MULTIPLY(tmp10 - tmp12 - tmp13, FIX(1.935399303)), /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 1067 | CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 1068 | |
shoaib_ahmed | 0:791a779d6220 | 1069 | tmp11 = MULTIPLY(tmp11, FIX(1.935399303)); /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 1070 | tmp0 = MULTIPLY(tmp10 + tmp12, FIX(1.436506004)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 1071 | tmp1 = MULTIPLY(tmp10 + tmp13, FIX(0.764348879)); /* c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1072 | |
shoaib_ahmed | 0:791a779d6220 | 1073 | dataptr[DCTSIZE*1] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1074 | DESCALE(tmp11 + tmp0 + tmp1, CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 1075 | |
shoaib_ahmed | 0:791a779d6220 | 1076 | tmp2 = MULTIPLY(tmp12 - tmp13, FIX(2.200854883)); /* c1 */ |
shoaib_ahmed | 0:791a779d6220 | 1077 | |
shoaib_ahmed | 0:791a779d6220 | 1078 | dataptr[DCTSIZE*5] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1079 | DESCALE(tmp0 - tmp11 - tmp2, CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 1080 | dataptr[DCTSIZE*7] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1081 | DESCALE(tmp1 - tmp11 + tmp2, CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 1082 | |
shoaib_ahmed | 0:791a779d6220 | 1083 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 1084 | wsptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 1085 | } |
shoaib_ahmed | 0:791a779d6220 | 1086 | } |
shoaib_ahmed | 0:791a779d6220 | 1087 | |
shoaib_ahmed | 0:791a779d6220 | 1088 | |
shoaib_ahmed | 0:791a779d6220 | 1089 | /* |
shoaib_ahmed | 0:791a779d6220 | 1090 | * Perform the forward DCT on a 10x10 sample block. |
shoaib_ahmed | 0:791a779d6220 | 1091 | */ |
shoaib_ahmed | 0:791a779d6220 | 1092 | |
shoaib_ahmed | 0:791a779d6220 | 1093 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 1094 | jpeg_fdct_10x10 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 1095 | { |
shoaib_ahmed | 0:791a779d6220 | 1096 | INT32 tmp0, tmp1, tmp2, tmp3, tmp4; |
shoaib_ahmed | 0:791a779d6220 | 1097 | INT32 tmp10, tmp11, tmp12, tmp13, tmp14; |
shoaib_ahmed | 0:791a779d6220 | 1098 | DCTELEM workspace[8*2]; |
shoaib_ahmed | 0:791a779d6220 | 1099 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 1100 | DCTELEM *wsptr; |
shoaib_ahmed | 0:791a779d6220 | 1101 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 1102 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 1103 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 1104 | |
shoaib_ahmed | 0:791a779d6220 | 1105 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 1106 | * Note results are scaled up by sqrt(8) compared to a true DCT; |
shoaib_ahmed | 0:791a779d6220 | 1107 | * we scale the results further by 2 as part of output adaption |
shoaib_ahmed | 0:791a779d6220 | 1108 | * scaling for different DCT size. |
shoaib_ahmed | 0:791a779d6220 | 1109 | * cK represents sqrt(2) * cos(K*pi/20). |
shoaib_ahmed | 0:791a779d6220 | 1110 | */ |
shoaib_ahmed | 0:791a779d6220 | 1111 | |
shoaib_ahmed | 0:791a779d6220 | 1112 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 1113 | ctr = 0; |
shoaib_ahmed | 0:791a779d6220 | 1114 | for (;;) { |
shoaib_ahmed | 0:791a779d6220 | 1115 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 1116 | |
shoaib_ahmed | 0:791a779d6220 | 1117 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 1118 | |
shoaib_ahmed | 0:791a779d6220 | 1119 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[9]); |
shoaib_ahmed | 0:791a779d6220 | 1120 | tmp1 = GETJSAMPLE(elemptr[1]) + GETJSAMPLE(elemptr[8]); |
shoaib_ahmed | 0:791a779d6220 | 1121 | tmp12 = GETJSAMPLE(elemptr[2]) + GETJSAMPLE(elemptr[7]); |
shoaib_ahmed | 0:791a779d6220 | 1122 | tmp3 = GETJSAMPLE(elemptr[3]) + GETJSAMPLE(elemptr[6]); |
shoaib_ahmed | 0:791a779d6220 | 1123 | tmp4 = GETJSAMPLE(elemptr[4]) + GETJSAMPLE(elemptr[5]); |
shoaib_ahmed | 0:791a779d6220 | 1124 | |
shoaib_ahmed | 0:791a779d6220 | 1125 | tmp10 = tmp0 + tmp4; |
shoaib_ahmed | 0:791a779d6220 | 1126 | tmp13 = tmp0 - tmp4; |
shoaib_ahmed | 0:791a779d6220 | 1127 | tmp11 = tmp1 + tmp3; |
shoaib_ahmed | 0:791a779d6220 | 1128 | tmp14 = tmp1 - tmp3; |
shoaib_ahmed | 0:791a779d6220 | 1129 | |
shoaib_ahmed | 0:791a779d6220 | 1130 | tmp0 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[9]); |
shoaib_ahmed | 0:791a779d6220 | 1131 | tmp1 = GETJSAMPLE(elemptr[1]) - GETJSAMPLE(elemptr[8]); |
shoaib_ahmed | 0:791a779d6220 | 1132 | tmp2 = GETJSAMPLE(elemptr[2]) - GETJSAMPLE(elemptr[7]); |
shoaib_ahmed | 0:791a779d6220 | 1133 | tmp3 = GETJSAMPLE(elemptr[3]) - GETJSAMPLE(elemptr[6]); |
shoaib_ahmed | 0:791a779d6220 | 1134 | tmp4 = GETJSAMPLE(elemptr[4]) - GETJSAMPLE(elemptr[5]); |
shoaib_ahmed | 0:791a779d6220 | 1135 | |
shoaib_ahmed | 0:791a779d6220 | 1136 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 1137 | dataptr[0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1138 | ((tmp10 + tmp11 + tmp12 - 10 * CENTERJSAMPLE) << 1); |
shoaib_ahmed | 0:791a779d6220 | 1139 | tmp12 += tmp12; |
shoaib_ahmed | 0:791a779d6220 | 1140 | dataptr[4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1141 | DESCALE(MULTIPLY(tmp10 - tmp12, FIX(1.144122806)) - /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 1142 | MULTIPLY(tmp11 - tmp12, FIX(0.437016024)), /* c8 */ |
shoaib_ahmed | 0:791a779d6220 | 1143 | CONST_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 1144 | tmp10 = MULTIPLY(tmp13 + tmp14, FIX(0.831253876)); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 1145 | dataptr[2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1146 | DESCALE(tmp10 + MULTIPLY(tmp13, FIX(0.513743148)), /* c2-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 1147 | CONST_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 1148 | dataptr[6] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1149 | DESCALE(tmp10 - MULTIPLY(tmp14, FIX(2.176250899)), /* c2+c6 */ |
shoaib_ahmed | 0:791a779d6220 | 1150 | CONST_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 1151 | |
shoaib_ahmed | 0:791a779d6220 | 1152 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 1153 | |
shoaib_ahmed | 0:791a779d6220 | 1154 | tmp10 = tmp0 + tmp4; |
shoaib_ahmed | 0:791a779d6220 | 1155 | tmp11 = tmp1 - tmp3; |
shoaib_ahmed | 0:791a779d6220 | 1156 | dataptr[5] = (DCTELEM) ((tmp10 - tmp11 - tmp2) << 1); |
shoaib_ahmed | 0:791a779d6220 | 1157 | tmp2 <<= CONST_BITS; |
shoaib_ahmed | 0:791a779d6220 | 1158 | dataptr[1] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1159 | DESCALE(MULTIPLY(tmp0, FIX(1.396802247)) + /* c1 */ |
shoaib_ahmed | 0:791a779d6220 | 1160 | MULTIPLY(tmp1, FIX(1.260073511)) + tmp2 + /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 1161 | MULTIPLY(tmp3, FIX(0.642039522)) + /* c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1162 | MULTIPLY(tmp4, FIX(0.221231742)), /* c9 */ |
shoaib_ahmed | 0:791a779d6220 | 1163 | CONST_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 1164 | tmp12 = MULTIPLY(tmp0 - tmp4, FIX(0.951056516)) - /* (c3+c7)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 1165 | MULTIPLY(tmp1 + tmp3, FIX(0.587785252)); /* (c1-c9)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 1166 | tmp13 = MULTIPLY(tmp10 + tmp11, FIX(0.309016994)) + /* (c3-c7)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 1167 | (tmp11 << (CONST_BITS - 1)) - tmp2; |
shoaib_ahmed | 0:791a779d6220 | 1168 | dataptr[3] = (DCTELEM) DESCALE(tmp12 + tmp13, CONST_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 1169 | dataptr[7] = (DCTELEM) DESCALE(tmp12 - tmp13, CONST_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 1170 | |
shoaib_ahmed | 0:791a779d6220 | 1171 | ctr++; |
shoaib_ahmed | 0:791a779d6220 | 1172 | |
shoaib_ahmed | 0:791a779d6220 | 1173 | if (ctr != DCTSIZE) { |
shoaib_ahmed | 0:791a779d6220 | 1174 | if (ctr == 10) |
shoaib_ahmed | 0:791a779d6220 | 1175 | break; /* Done. */ |
shoaib_ahmed | 0:791a779d6220 | 1176 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 1177 | } else |
shoaib_ahmed | 0:791a779d6220 | 1178 | dataptr = workspace; /* switch pointer to extended workspace */ |
shoaib_ahmed | 0:791a779d6220 | 1179 | } |
shoaib_ahmed | 0:791a779d6220 | 1180 | |
shoaib_ahmed | 0:791a779d6220 | 1181 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 1182 | * We leave the results scaled up by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 1183 | * We must also scale the output by (8/10)**2 = 16/25, which we partially |
shoaib_ahmed | 0:791a779d6220 | 1184 | * fold into the constant multipliers and final/initial shifting: |
shoaib_ahmed | 0:791a779d6220 | 1185 | * cK now represents sqrt(2) * cos(K*pi/20) * 32/25. |
shoaib_ahmed | 0:791a779d6220 | 1186 | */ |
shoaib_ahmed | 0:791a779d6220 | 1187 | |
shoaib_ahmed | 0:791a779d6220 | 1188 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 1189 | wsptr = workspace; |
shoaib_ahmed | 0:791a779d6220 | 1190 | for (ctr = DCTSIZE-1; ctr >= 0; ctr--) { |
shoaib_ahmed | 0:791a779d6220 | 1191 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 1192 | |
shoaib_ahmed | 0:791a779d6220 | 1193 | tmp0 = dataptr[DCTSIZE*0] + wsptr[DCTSIZE*1]; |
shoaib_ahmed | 0:791a779d6220 | 1194 | tmp1 = dataptr[DCTSIZE*1] + wsptr[DCTSIZE*0]; |
shoaib_ahmed | 0:791a779d6220 | 1195 | tmp12 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 1196 | tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 1197 | tmp4 = dataptr[DCTSIZE*4] + dataptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 1198 | |
shoaib_ahmed | 0:791a779d6220 | 1199 | tmp10 = tmp0 + tmp4; |
shoaib_ahmed | 0:791a779d6220 | 1200 | tmp13 = tmp0 - tmp4; |
shoaib_ahmed | 0:791a779d6220 | 1201 | tmp11 = tmp1 + tmp3; |
shoaib_ahmed | 0:791a779d6220 | 1202 | tmp14 = tmp1 - tmp3; |
shoaib_ahmed | 0:791a779d6220 | 1203 | |
shoaib_ahmed | 0:791a779d6220 | 1204 | tmp0 = dataptr[DCTSIZE*0] - wsptr[DCTSIZE*1]; |
shoaib_ahmed | 0:791a779d6220 | 1205 | tmp1 = dataptr[DCTSIZE*1] - wsptr[DCTSIZE*0]; |
shoaib_ahmed | 0:791a779d6220 | 1206 | tmp2 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 1207 | tmp3 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 1208 | tmp4 = dataptr[DCTSIZE*4] - dataptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 1209 | |
shoaib_ahmed | 0:791a779d6220 | 1210 | dataptr[DCTSIZE*0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1211 | DESCALE(MULTIPLY(tmp10 + tmp11 + tmp12, FIX(1.28)), /* 32/25 */ |
shoaib_ahmed | 0:791a779d6220 | 1212 | CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 1213 | tmp12 += tmp12; |
shoaib_ahmed | 0:791a779d6220 | 1214 | dataptr[DCTSIZE*4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1215 | DESCALE(MULTIPLY(tmp10 - tmp12, FIX(1.464477191)) - /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 1216 | MULTIPLY(tmp11 - tmp12, FIX(0.559380511)), /* c8 */ |
shoaib_ahmed | 0:791a779d6220 | 1217 | CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 1218 | tmp10 = MULTIPLY(tmp13 + tmp14, FIX(1.064004961)); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 1219 | dataptr[DCTSIZE*2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1220 | DESCALE(tmp10 + MULTIPLY(tmp13, FIX(0.657591230)), /* c2-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 1221 | CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 1222 | dataptr[DCTSIZE*6] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1223 | DESCALE(tmp10 - MULTIPLY(tmp14, FIX(2.785601151)), /* c2+c6 */ |
shoaib_ahmed | 0:791a779d6220 | 1224 | CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 1225 | |
shoaib_ahmed | 0:791a779d6220 | 1226 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 1227 | |
shoaib_ahmed | 0:791a779d6220 | 1228 | tmp10 = tmp0 + tmp4; |
shoaib_ahmed | 0:791a779d6220 | 1229 | tmp11 = tmp1 - tmp3; |
shoaib_ahmed | 0:791a779d6220 | 1230 | dataptr[DCTSIZE*5] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1231 | DESCALE(MULTIPLY(tmp10 - tmp11 - tmp2, FIX(1.28)), /* 32/25 */ |
shoaib_ahmed | 0:791a779d6220 | 1232 | CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 1233 | tmp2 = MULTIPLY(tmp2, FIX(1.28)); /* 32/25 */ |
shoaib_ahmed | 0:791a779d6220 | 1234 | dataptr[DCTSIZE*1] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1235 | DESCALE(MULTIPLY(tmp0, FIX(1.787906876)) + /* c1 */ |
shoaib_ahmed | 0:791a779d6220 | 1236 | MULTIPLY(tmp1, FIX(1.612894094)) + tmp2 + /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 1237 | MULTIPLY(tmp3, FIX(0.821810588)) + /* c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1238 | MULTIPLY(tmp4, FIX(0.283176630)), /* c9 */ |
shoaib_ahmed | 0:791a779d6220 | 1239 | CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 1240 | tmp12 = MULTIPLY(tmp0 - tmp4, FIX(1.217352341)) - /* (c3+c7)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 1241 | MULTIPLY(tmp1 + tmp3, FIX(0.752365123)); /* (c1-c9)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 1242 | tmp13 = MULTIPLY(tmp10 + tmp11, FIX(0.395541753)) + /* (c3-c7)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 1243 | MULTIPLY(tmp11, FIX(0.64)) - tmp2; /* 16/25 */ |
shoaib_ahmed | 0:791a779d6220 | 1244 | dataptr[DCTSIZE*3] = (DCTELEM) DESCALE(tmp12 + tmp13, CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 1245 | dataptr[DCTSIZE*7] = (DCTELEM) DESCALE(tmp12 - tmp13, CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 1246 | |
shoaib_ahmed | 0:791a779d6220 | 1247 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 1248 | wsptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 1249 | } |
shoaib_ahmed | 0:791a779d6220 | 1250 | } |
shoaib_ahmed | 0:791a779d6220 | 1251 | |
shoaib_ahmed | 0:791a779d6220 | 1252 | |
shoaib_ahmed | 0:791a779d6220 | 1253 | /* |
shoaib_ahmed | 0:791a779d6220 | 1254 | * Perform the forward DCT on an 11x11 sample block. |
shoaib_ahmed | 0:791a779d6220 | 1255 | */ |
shoaib_ahmed | 0:791a779d6220 | 1256 | |
shoaib_ahmed | 0:791a779d6220 | 1257 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 1258 | jpeg_fdct_11x11 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 1259 | { |
shoaib_ahmed | 0:791a779d6220 | 1260 | INT32 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5; |
shoaib_ahmed | 0:791a779d6220 | 1261 | INT32 tmp10, tmp11, tmp12, tmp13, tmp14; |
shoaib_ahmed | 0:791a779d6220 | 1262 | INT32 z1, z2, z3; |
shoaib_ahmed | 0:791a779d6220 | 1263 | DCTELEM workspace[8*3]; |
shoaib_ahmed | 0:791a779d6220 | 1264 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 1265 | DCTELEM *wsptr; |
shoaib_ahmed | 0:791a779d6220 | 1266 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 1267 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 1268 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 1269 | |
shoaib_ahmed | 0:791a779d6220 | 1270 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 1271 | * Note results are scaled up by sqrt(8) compared to a true DCT; |
shoaib_ahmed | 0:791a779d6220 | 1272 | * we scale the results further by 2 as part of output adaption |
shoaib_ahmed | 0:791a779d6220 | 1273 | * scaling for different DCT size. |
shoaib_ahmed | 0:791a779d6220 | 1274 | * cK represents sqrt(2) * cos(K*pi/22). |
shoaib_ahmed | 0:791a779d6220 | 1275 | */ |
shoaib_ahmed | 0:791a779d6220 | 1276 | |
shoaib_ahmed | 0:791a779d6220 | 1277 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 1278 | ctr = 0; |
shoaib_ahmed | 0:791a779d6220 | 1279 | for (;;) { |
shoaib_ahmed | 0:791a779d6220 | 1280 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 1281 | |
shoaib_ahmed | 0:791a779d6220 | 1282 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 1283 | |
shoaib_ahmed | 0:791a779d6220 | 1284 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[10]); |
shoaib_ahmed | 0:791a779d6220 | 1285 | tmp1 = GETJSAMPLE(elemptr[1]) + GETJSAMPLE(elemptr[9]); |
shoaib_ahmed | 0:791a779d6220 | 1286 | tmp2 = GETJSAMPLE(elemptr[2]) + GETJSAMPLE(elemptr[8]); |
shoaib_ahmed | 0:791a779d6220 | 1287 | tmp3 = GETJSAMPLE(elemptr[3]) + GETJSAMPLE(elemptr[7]); |
shoaib_ahmed | 0:791a779d6220 | 1288 | tmp4 = GETJSAMPLE(elemptr[4]) + GETJSAMPLE(elemptr[6]); |
shoaib_ahmed | 0:791a779d6220 | 1289 | tmp5 = GETJSAMPLE(elemptr[5]); |
shoaib_ahmed | 0:791a779d6220 | 1290 | |
shoaib_ahmed | 0:791a779d6220 | 1291 | tmp10 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[10]); |
shoaib_ahmed | 0:791a779d6220 | 1292 | tmp11 = GETJSAMPLE(elemptr[1]) - GETJSAMPLE(elemptr[9]); |
shoaib_ahmed | 0:791a779d6220 | 1293 | tmp12 = GETJSAMPLE(elemptr[2]) - GETJSAMPLE(elemptr[8]); |
shoaib_ahmed | 0:791a779d6220 | 1294 | tmp13 = GETJSAMPLE(elemptr[3]) - GETJSAMPLE(elemptr[7]); |
shoaib_ahmed | 0:791a779d6220 | 1295 | tmp14 = GETJSAMPLE(elemptr[4]) - GETJSAMPLE(elemptr[6]); |
shoaib_ahmed | 0:791a779d6220 | 1296 | |
shoaib_ahmed | 0:791a779d6220 | 1297 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 1298 | dataptr[0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1299 | ((tmp0 + tmp1 + tmp2 + tmp3 + tmp4 + tmp5 - 11 * CENTERJSAMPLE) << 1); |
shoaib_ahmed | 0:791a779d6220 | 1300 | tmp5 += tmp5; |
shoaib_ahmed | 0:791a779d6220 | 1301 | tmp0 -= tmp5; |
shoaib_ahmed | 0:791a779d6220 | 1302 | tmp1 -= tmp5; |
shoaib_ahmed | 0:791a779d6220 | 1303 | tmp2 -= tmp5; |
shoaib_ahmed | 0:791a779d6220 | 1304 | tmp3 -= tmp5; |
shoaib_ahmed | 0:791a779d6220 | 1305 | tmp4 -= tmp5; |
shoaib_ahmed | 0:791a779d6220 | 1306 | z1 = MULTIPLY(tmp0 + tmp3, FIX(1.356927976)) + /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 1307 | MULTIPLY(tmp2 + tmp4, FIX(0.201263574)); /* c10 */ |
shoaib_ahmed | 0:791a779d6220 | 1308 | z2 = MULTIPLY(tmp1 - tmp3, FIX(0.926112931)); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 1309 | z3 = MULTIPLY(tmp0 - tmp1, FIX(1.189712156)); /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 1310 | dataptr[2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1311 | DESCALE(z1 + z2 - MULTIPLY(tmp3, FIX(1.018300590)) /* c2+c8-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 1312 | - MULTIPLY(tmp4, FIX(1.390975730)), /* c4+c10 */ |
shoaib_ahmed | 0:791a779d6220 | 1313 | CONST_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 1314 | dataptr[4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1315 | DESCALE(z2 + z3 + MULTIPLY(tmp1, FIX(0.062335650)) /* c4-c6-c10 */ |
shoaib_ahmed | 0:791a779d6220 | 1316 | - MULTIPLY(tmp2, FIX(1.356927976)) /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 1317 | + MULTIPLY(tmp4, FIX(0.587485545)), /* c8 */ |
shoaib_ahmed | 0:791a779d6220 | 1318 | CONST_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 1319 | dataptr[6] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1320 | DESCALE(z1 + z3 - MULTIPLY(tmp0, FIX(1.620527200)) /* c2+c4-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 1321 | - MULTIPLY(tmp2, FIX(0.788749120)), /* c8+c10 */ |
shoaib_ahmed | 0:791a779d6220 | 1322 | CONST_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 1323 | |
shoaib_ahmed | 0:791a779d6220 | 1324 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 1325 | |
shoaib_ahmed | 0:791a779d6220 | 1326 | tmp1 = MULTIPLY(tmp10 + tmp11, FIX(1.286413905)); /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 1327 | tmp2 = MULTIPLY(tmp10 + tmp12, FIX(1.068791298)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 1328 | tmp3 = MULTIPLY(tmp10 + tmp13, FIX(0.764581576)); /* c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1329 | tmp0 = tmp1 + tmp2 + tmp3 - MULTIPLY(tmp10, FIX(1.719967871)) /* c7+c5+c3-c1 */ |
shoaib_ahmed | 0:791a779d6220 | 1330 | + MULTIPLY(tmp14, FIX(0.398430003)); /* c9 */ |
shoaib_ahmed | 0:791a779d6220 | 1331 | tmp4 = MULTIPLY(tmp11 + tmp12, - FIX(0.764581576)); /* -c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1332 | tmp5 = MULTIPLY(tmp11 + tmp13, - FIX(1.399818907)); /* -c1 */ |
shoaib_ahmed | 0:791a779d6220 | 1333 | tmp1 += tmp4 + tmp5 + MULTIPLY(tmp11, FIX(1.276416582)) /* c9+c7+c1-c3 */ |
shoaib_ahmed | 0:791a779d6220 | 1334 | - MULTIPLY(tmp14, FIX(1.068791298)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 1335 | tmp10 = MULTIPLY(tmp12 + tmp13, FIX(0.398430003)); /* c9 */ |
shoaib_ahmed | 0:791a779d6220 | 1336 | tmp2 += tmp4 + tmp10 - MULTIPLY(tmp12, FIX(1.989053629)) /* c9+c5+c3-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1337 | + MULTIPLY(tmp14, FIX(1.399818907)); /* c1 */ |
shoaib_ahmed | 0:791a779d6220 | 1338 | tmp3 += tmp5 + tmp10 + MULTIPLY(tmp13, FIX(1.305598626)) /* c1+c5-c9-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1339 | - MULTIPLY(tmp14, FIX(1.286413905)); /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 1340 | |
shoaib_ahmed | 0:791a779d6220 | 1341 | dataptr[1] = (DCTELEM) DESCALE(tmp0, CONST_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 1342 | dataptr[3] = (DCTELEM) DESCALE(tmp1, CONST_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 1343 | dataptr[5] = (DCTELEM) DESCALE(tmp2, CONST_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 1344 | dataptr[7] = (DCTELEM) DESCALE(tmp3, CONST_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 1345 | |
shoaib_ahmed | 0:791a779d6220 | 1346 | ctr++; |
shoaib_ahmed | 0:791a779d6220 | 1347 | |
shoaib_ahmed | 0:791a779d6220 | 1348 | if (ctr != DCTSIZE) { |
shoaib_ahmed | 0:791a779d6220 | 1349 | if (ctr == 11) |
shoaib_ahmed | 0:791a779d6220 | 1350 | break; /* Done. */ |
shoaib_ahmed | 0:791a779d6220 | 1351 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 1352 | } else |
shoaib_ahmed | 0:791a779d6220 | 1353 | dataptr = workspace; /* switch pointer to extended workspace */ |
shoaib_ahmed | 0:791a779d6220 | 1354 | } |
shoaib_ahmed | 0:791a779d6220 | 1355 | |
shoaib_ahmed | 0:791a779d6220 | 1356 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 1357 | * We leave the results scaled up by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 1358 | * We must also scale the output by (8/11)**2 = 64/121, which we partially |
shoaib_ahmed | 0:791a779d6220 | 1359 | * fold into the constant multipliers and final/initial shifting: |
shoaib_ahmed | 0:791a779d6220 | 1360 | * cK now represents sqrt(2) * cos(K*pi/22) * 128/121. |
shoaib_ahmed | 0:791a779d6220 | 1361 | */ |
shoaib_ahmed | 0:791a779d6220 | 1362 | |
shoaib_ahmed | 0:791a779d6220 | 1363 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 1364 | wsptr = workspace; |
shoaib_ahmed | 0:791a779d6220 | 1365 | for (ctr = DCTSIZE-1; ctr >= 0; ctr--) { |
shoaib_ahmed | 0:791a779d6220 | 1366 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 1367 | |
shoaib_ahmed | 0:791a779d6220 | 1368 | tmp0 = dataptr[DCTSIZE*0] + wsptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 1369 | tmp1 = dataptr[DCTSIZE*1] + wsptr[DCTSIZE*1]; |
shoaib_ahmed | 0:791a779d6220 | 1370 | tmp2 = dataptr[DCTSIZE*2] + wsptr[DCTSIZE*0]; |
shoaib_ahmed | 0:791a779d6220 | 1371 | tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 1372 | tmp4 = dataptr[DCTSIZE*4] + dataptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 1373 | tmp5 = dataptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 1374 | |
shoaib_ahmed | 0:791a779d6220 | 1375 | tmp10 = dataptr[DCTSIZE*0] - wsptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 1376 | tmp11 = dataptr[DCTSIZE*1] - wsptr[DCTSIZE*1]; |
shoaib_ahmed | 0:791a779d6220 | 1377 | tmp12 = dataptr[DCTSIZE*2] - wsptr[DCTSIZE*0]; |
shoaib_ahmed | 0:791a779d6220 | 1378 | tmp13 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 1379 | tmp14 = dataptr[DCTSIZE*4] - dataptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 1380 | |
shoaib_ahmed | 0:791a779d6220 | 1381 | dataptr[DCTSIZE*0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1382 | DESCALE(MULTIPLY(tmp0 + tmp1 + tmp2 + tmp3 + tmp4 + tmp5, |
shoaib_ahmed | 0:791a779d6220 | 1383 | FIX(1.057851240)), /* 128/121 */ |
shoaib_ahmed | 0:791a779d6220 | 1384 | CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 1385 | tmp5 += tmp5; |
shoaib_ahmed | 0:791a779d6220 | 1386 | tmp0 -= tmp5; |
shoaib_ahmed | 0:791a779d6220 | 1387 | tmp1 -= tmp5; |
shoaib_ahmed | 0:791a779d6220 | 1388 | tmp2 -= tmp5; |
shoaib_ahmed | 0:791a779d6220 | 1389 | tmp3 -= tmp5; |
shoaib_ahmed | 0:791a779d6220 | 1390 | tmp4 -= tmp5; |
shoaib_ahmed | 0:791a779d6220 | 1391 | z1 = MULTIPLY(tmp0 + tmp3, FIX(1.435427942)) + /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 1392 | MULTIPLY(tmp2 + tmp4, FIX(0.212906922)); /* c10 */ |
shoaib_ahmed | 0:791a779d6220 | 1393 | z2 = MULTIPLY(tmp1 - tmp3, FIX(0.979689713)); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 1394 | z3 = MULTIPLY(tmp0 - tmp1, FIX(1.258538479)); /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 1395 | dataptr[DCTSIZE*2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1396 | DESCALE(z1 + z2 - MULTIPLY(tmp3, FIX(1.077210542)) /* c2+c8-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 1397 | - MULTIPLY(tmp4, FIX(1.471445400)), /* c4+c10 */ |
shoaib_ahmed | 0:791a779d6220 | 1398 | CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 1399 | dataptr[DCTSIZE*4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1400 | DESCALE(z2 + z3 + MULTIPLY(tmp1, FIX(0.065941844)) /* c4-c6-c10 */ |
shoaib_ahmed | 0:791a779d6220 | 1401 | - MULTIPLY(tmp2, FIX(1.435427942)) /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 1402 | + MULTIPLY(tmp4, FIX(0.621472312)), /* c8 */ |
shoaib_ahmed | 0:791a779d6220 | 1403 | CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 1404 | dataptr[DCTSIZE*6] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1405 | DESCALE(z1 + z3 - MULTIPLY(tmp0, FIX(1.714276708)) /* c2+c4-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 1406 | - MULTIPLY(tmp2, FIX(0.834379234)), /* c8+c10 */ |
shoaib_ahmed | 0:791a779d6220 | 1407 | CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 1408 | |
shoaib_ahmed | 0:791a779d6220 | 1409 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 1410 | |
shoaib_ahmed | 0:791a779d6220 | 1411 | tmp1 = MULTIPLY(tmp10 + tmp11, FIX(1.360834544)); /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 1412 | tmp2 = MULTIPLY(tmp10 + tmp12, FIX(1.130622199)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 1413 | tmp3 = MULTIPLY(tmp10 + tmp13, FIX(0.808813568)); /* c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1414 | tmp0 = tmp1 + tmp2 + tmp3 - MULTIPLY(tmp10, FIX(1.819470145)) /* c7+c5+c3-c1 */ |
shoaib_ahmed | 0:791a779d6220 | 1415 | + MULTIPLY(tmp14, FIX(0.421479672)); /* c9 */ |
shoaib_ahmed | 0:791a779d6220 | 1416 | tmp4 = MULTIPLY(tmp11 + tmp12, - FIX(0.808813568)); /* -c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1417 | tmp5 = MULTIPLY(tmp11 + tmp13, - FIX(1.480800167)); /* -c1 */ |
shoaib_ahmed | 0:791a779d6220 | 1418 | tmp1 += tmp4 + tmp5 + MULTIPLY(tmp11, FIX(1.350258864)) /* c9+c7+c1-c3 */ |
shoaib_ahmed | 0:791a779d6220 | 1419 | - MULTIPLY(tmp14, FIX(1.130622199)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 1420 | tmp10 = MULTIPLY(tmp12 + tmp13, FIX(0.421479672)); /* c9 */ |
shoaib_ahmed | 0:791a779d6220 | 1421 | tmp2 += tmp4 + tmp10 - MULTIPLY(tmp12, FIX(2.104122847)) /* c9+c5+c3-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1422 | + MULTIPLY(tmp14, FIX(1.480800167)); /* c1 */ |
shoaib_ahmed | 0:791a779d6220 | 1423 | tmp3 += tmp5 + tmp10 + MULTIPLY(tmp13, FIX(1.381129125)) /* c1+c5-c9-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1424 | - MULTIPLY(tmp14, FIX(1.360834544)); /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 1425 | |
shoaib_ahmed | 0:791a779d6220 | 1426 | dataptr[DCTSIZE*1] = (DCTELEM) DESCALE(tmp0, CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 1427 | dataptr[DCTSIZE*3] = (DCTELEM) DESCALE(tmp1, CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 1428 | dataptr[DCTSIZE*5] = (DCTELEM) DESCALE(tmp2, CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 1429 | dataptr[DCTSIZE*7] = (DCTELEM) DESCALE(tmp3, CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 1430 | |
shoaib_ahmed | 0:791a779d6220 | 1431 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 1432 | wsptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 1433 | } |
shoaib_ahmed | 0:791a779d6220 | 1434 | } |
shoaib_ahmed | 0:791a779d6220 | 1435 | |
shoaib_ahmed | 0:791a779d6220 | 1436 | |
shoaib_ahmed | 0:791a779d6220 | 1437 | /* |
shoaib_ahmed | 0:791a779d6220 | 1438 | * Perform the forward DCT on a 12x12 sample block. |
shoaib_ahmed | 0:791a779d6220 | 1439 | */ |
shoaib_ahmed | 0:791a779d6220 | 1440 | |
shoaib_ahmed | 0:791a779d6220 | 1441 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 1442 | jpeg_fdct_12x12 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 1443 | { |
shoaib_ahmed | 0:791a779d6220 | 1444 | INT32 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5; |
shoaib_ahmed | 0:791a779d6220 | 1445 | INT32 tmp10, tmp11, tmp12, tmp13, tmp14, tmp15; |
shoaib_ahmed | 0:791a779d6220 | 1446 | DCTELEM workspace[8*4]; |
shoaib_ahmed | 0:791a779d6220 | 1447 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 1448 | DCTELEM *wsptr; |
shoaib_ahmed | 0:791a779d6220 | 1449 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 1450 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 1451 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 1452 | |
shoaib_ahmed | 0:791a779d6220 | 1453 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 1454 | * Note results are scaled up by sqrt(8) compared to a true DCT. |
shoaib_ahmed | 0:791a779d6220 | 1455 | * cK represents sqrt(2) * cos(K*pi/24). |
shoaib_ahmed | 0:791a779d6220 | 1456 | */ |
shoaib_ahmed | 0:791a779d6220 | 1457 | |
shoaib_ahmed | 0:791a779d6220 | 1458 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 1459 | ctr = 0; |
shoaib_ahmed | 0:791a779d6220 | 1460 | for (;;) { |
shoaib_ahmed | 0:791a779d6220 | 1461 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 1462 | |
shoaib_ahmed | 0:791a779d6220 | 1463 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 1464 | |
shoaib_ahmed | 0:791a779d6220 | 1465 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[11]); |
shoaib_ahmed | 0:791a779d6220 | 1466 | tmp1 = GETJSAMPLE(elemptr[1]) + GETJSAMPLE(elemptr[10]); |
shoaib_ahmed | 0:791a779d6220 | 1467 | tmp2 = GETJSAMPLE(elemptr[2]) + GETJSAMPLE(elemptr[9]); |
shoaib_ahmed | 0:791a779d6220 | 1468 | tmp3 = GETJSAMPLE(elemptr[3]) + GETJSAMPLE(elemptr[8]); |
shoaib_ahmed | 0:791a779d6220 | 1469 | tmp4 = GETJSAMPLE(elemptr[4]) + GETJSAMPLE(elemptr[7]); |
shoaib_ahmed | 0:791a779d6220 | 1470 | tmp5 = GETJSAMPLE(elemptr[5]) + GETJSAMPLE(elemptr[6]); |
shoaib_ahmed | 0:791a779d6220 | 1471 | |
shoaib_ahmed | 0:791a779d6220 | 1472 | tmp10 = tmp0 + tmp5; |
shoaib_ahmed | 0:791a779d6220 | 1473 | tmp13 = tmp0 - tmp5; |
shoaib_ahmed | 0:791a779d6220 | 1474 | tmp11 = tmp1 + tmp4; |
shoaib_ahmed | 0:791a779d6220 | 1475 | tmp14 = tmp1 - tmp4; |
shoaib_ahmed | 0:791a779d6220 | 1476 | tmp12 = tmp2 + tmp3; |
shoaib_ahmed | 0:791a779d6220 | 1477 | tmp15 = tmp2 - tmp3; |
shoaib_ahmed | 0:791a779d6220 | 1478 | |
shoaib_ahmed | 0:791a779d6220 | 1479 | tmp0 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[11]); |
shoaib_ahmed | 0:791a779d6220 | 1480 | tmp1 = GETJSAMPLE(elemptr[1]) - GETJSAMPLE(elemptr[10]); |
shoaib_ahmed | 0:791a779d6220 | 1481 | tmp2 = GETJSAMPLE(elemptr[2]) - GETJSAMPLE(elemptr[9]); |
shoaib_ahmed | 0:791a779d6220 | 1482 | tmp3 = GETJSAMPLE(elemptr[3]) - GETJSAMPLE(elemptr[8]); |
shoaib_ahmed | 0:791a779d6220 | 1483 | tmp4 = GETJSAMPLE(elemptr[4]) - GETJSAMPLE(elemptr[7]); |
shoaib_ahmed | 0:791a779d6220 | 1484 | tmp5 = GETJSAMPLE(elemptr[5]) - GETJSAMPLE(elemptr[6]); |
shoaib_ahmed | 0:791a779d6220 | 1485 | |
shoaib_ahmed | 0:791a779d6220 | 1486 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 1487 | dataptr[0] = (DCTELEM) (tmp10 + tmp11 + tmp12 - 12 * CENTERJSAMPLE); |
shoaib_ahmed | 0:791a779d6220 | 1488 | dataptr[6] = (DCTELEM) (tmp13 - tmp14 - tmp15); |
shoaib_ahmed | 0:791a779d6220 | 1489 | dataptr[4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1490 | DESCALE(MULTIPLY(tmp10 - tmp12, FIX(1.224744871)), /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 1491 | CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 1492 | dataptr[2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1493 | DESCALE(tmp14 - tmp15 + MULTIPLY(tmp13 + tmp15, FIX(1.366025404)), /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 1494 | CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 1495 | |
shoaib_ahmed | 0:791a779d6220 | 1496 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 1497 | |
shoaib_ahmed | 0:791a779d6220 | 1498 | tmp10 = MULTIPLY(tmp1 + tmp4, FIX_0_541196100); /* c9 */ |
shoaib_ahmed | 0:791a779d6220 | 1499 | tmp14 = tmp10 + MULTIPLY(tmp1, FIX_0_765366865); /* c3-c9 */ |
shoaib_ahmed | 0:791a779d6220 | 1500 | tmp15 = tmp10 - MULTIPLY(tmp4, FIX_1_847759065); /* c3+c9 */ |
shoaib_ahmed | 0:791a779d6220 | 1501 | tmp12 = MULTIPLY(tmp0 + tmp2, FIX(1.121971054)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 1502 | tmp13 = MULTIPLY(tmp0 + tmp3, FIX(0.860918669)); /* c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1503 | tmp10 = tmp12 + tmp13 + tmp14 - MULTIPLY(tmp0, FIX(0.580774953)) /* c5+c7-c1 */ |
shoaib_ahmed | 0:791a779d6220 | 1504 | + MULTIPLY(tmp5, FIX(0.184591911)); /* c11 */ |
shoaib_ahmed | 0:791a779d6220 | 1505 | tmp11 = MULTIPLY(tmp2 + tmp3, - FIX(0.184591911)); /* -c11 */ |
shoaib_ahmed | 0:791a779d6220 | 1506 | tmp12 += tmp11 - tmp15 - MULTIPLY(tmp2, FIX(2.339493912)) /* c1+c5-c11 */ |
shoaib_ahmed | 0:791a779d6220 | 1507 | + MULTIPLY(tmp5, FIX(0.860918669)); /* c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1508 | tmp13 += tmp11 - tmp14 + MULTIPLY(tmp3, FIX(0.725788011)) /* c1+c11-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1509 | - MULTIPLY(tmp5, FIX(1.121971054)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 1510 | tmp11 = tmp15 + MULTIPLY(tmp0 - tmp3, FIX(1.306562965)) /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 1511 | - MULTIPLY(tmp2 + tmp5, FIX_0_541196100); /* c9 */ |
shoaib_ahmed | 0:791a779d6220 | 1512 | |
shoaib_ahmed | 0:791a779d6220 | 1513 | dataptr[1] = (DCTELEM) DESCALE(tmp10, CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 1514 | dataptr[3] = (DCTELEM) DESCALE(tmp11, CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 1515 | dataptr[5] = (DCTELEM) DESCALE(tmp12, CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 1516 | dataptr[7] = (DCTELEM) DESCALE(tmp13, CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 1517 | |
shoaib_ahmed | 0:791a779d6220 | 1518 | ctr++; |
shoaib_ahmed | 0:791a779d6220 | 1519 | |
shoaib_ahmed | 0:791a779d6220 | 1520 | if (ctr != DCTSIZE) { |
shoaib_ahmed | 0:791a779d6220 | 1521 | if (ctr == 12) |
shoaib_ahmed | 0:791a779d6220 | 1522 | break; /* Done. */ |
shoaib_ahmed | 0:791a779d6220 | 1523 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 1524 | } else |
shoaib_ahmed | 0:791a779d6220 | 1525 | dataptr = workspace; /* switch pointer to extended workspace */ |
shoaib_ahmed | 0:791a779d6220 | 1526 | } |
shoaib_ahmed | 0:791a779d6220 | 1527 | |
shoaib_ahmed | 0:791a779d6220 | 1528 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 1529 | * We leave the results scaled up by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 1530 | * We must also scale the output by (8/12)**2 = 4/9, which we partially |
shoaib_ahmed | 0:791a779d6220 | 1531 | * fold into the constant multipliers and final shifting: |
shoaib_ahmed | 0:791a779d6220 | 1532 | * cK now represents sqrt(2) * cos(K*pi/24) * 8/9. |
shoaib_ahmed | 0:791a779d6220 | 1533 | */ |
shoaib_ahmed | 0:791a779d6220 | 1534 | |
shoaib_ahmed | 0:791a779d6220 | 1535 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 1536 | wsptr = workspace; |
shoaib_ahmed | 0:791a779d6220 | 1537 | for (ctr = DCTSIZE-1; ctr >= 0; ctr--) { |
shoaib_ahmed | 0:791a779d6220 | 1538 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 1539 | |
shoaib_ahmed | 0:791a779d6220 | 1540 | tmp0 = dataptr[DCTSIZE*0] + wsptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 1541 | tmp1 = dataptr[DCTSIZE*1] + wsptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 1542 | tmp2 = dataptr[DCTSIZE*2] + wsptr[DCTSIZE*1]; |
shoaib_ahmed | 0:791a779d6220 | 1543 | tmp3 = dataptr[DCTSIZE*3] + wsptr[DCTSIZE*0]; |
shoaib_ahmed | 0:791a779d6220 | 1544 | tmp4 = dataptr[DCTSIZE*4] + dataptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 1545 | tmp5 = dataptr[DCTSIZE*5] + dataptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 1546 | |
shoaib_ahmed | 0:791a779d6220 | 1547 | tmp10 = tmp0 + tmp5; |
shoaib_ahmed | 0:791a779d6220 | 1548 | tmp13 = tmp0 - tmp5; |
shoaib_ahmed | 0:791a779d6220 | 1549 | tmp11 = tmp1 + tmp4; |
shoaib_ahmed | 0:791a779d6220 | 1550 | tmp14 = tmp1 - tmp4; |
shoaib_ahmed | 0:791a779d6220 | 1551 | tmp12 = tmp2 + tmp3; |
shoaib_ahmed | 0:791a779d6220 | 1552 | tmp15 = tmp2 - tmp3; |
shoaib_ahmed | 0:791a779d6220 | 1553 | |
shoaib_ahmed | 0:791a779d6220 | 1554 | tmp0 = dataptr[DCTSIZE*0] - wsptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 1555 | tmp1 = dataptr[DCTSIZE*1] - wsptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 1556 | tmp2 = dataptr[DCTSIZE*2] - wsptr[DCTSIZE*1]; |
shoaib_ahmed | 0:791a779d6220 | 1557 | tmp3 = dataptr[DCTSIZE*3] - wsptr[DCTSIZE*0]; |
shoaib_ahmed | 0:791a779d6220 | 1558 | tmp4 = dataptr[DCTSIZE*4] - dataptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 1559 | tmp5 = dataptr[DCTSIZE*5] - dataptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 1560 | |
shoaib_ahmed | 0:791a779d6220 | 1561 | dataptr[DCTSIZE*0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1562 | DESCALE(MULTIPLY(tmp10 + tmp11 + tmp12, FIX(0.888888889)), /* 8/9 */ |
shoaib_ahmed | 0:791a779d6220 | 1563 | CONST_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 1564 | dataptr[DCTSIZE*6] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1565 | DESCALE(MULTIPLY(tmp13 - tmp14 - tmp15, FIX(0.888888889)), /* 8/9 */ |
shoaib_ahmed | 0:791a779d6220 | 1566 | CONST_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 1567 | dataptr[DCTSIZE*4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1568 | DESCALE(MULTIPLY(tmp10 - tmp12, FIX(1.088662108)), /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 1569 | CONST_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 1570 | dataptr[DCTSIZE*2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1571 | DESCALE(MULTIPLY(tmp14 - tmp15, FIX(0.888888889)) + /* 8/9 */ |
shoaib_ahmed | 0:791a779d6220 | 1572 | MULTIPLY(tmp13 + tmp15, FIX(1.214244803)), /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 1573 | CONST_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 1574 | |
shoaib_ahmed | 0:791a779d6220 | 1575 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 1576 | |
shoaib_ahmed | 0:791a779d6220 | 1577 | tmp10 = MULTIPLY(tmp1 + tmp4, FIX(0.481063200)); /* c9 */ |
shoaib_ahmed | 0:791a779d6220 | 1578 | tmp14 = tmp10 + MULTIPLY(tmp1, FIX(0.680326102)); /* c3-c9 */ |
shoaib_ahmed | 0:791a779d6220 | 1579 | tmp15 = tmp10 - MULTIPLY(tmp4, FIX(1.642452502)); /* c3+c9 */ |
shoaib_ahmed | 0:791a779d6220 | 1580 | tmp12 = MULTIPLY(tmp0 + tmp2, FIX(0.997307603)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 1581 | tmp13 = MULTIPLY(tmp0 + tmp3, FIX(0.765261039)); /* c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1582 | tmp10 = tmp12 + tmp13 + tmp14 - MULTIPLY(tmp0, FIX(0.516244403)) /* c5+c7-c1 */ |
shoaib_ahmed | 0:791a779d6220 | 1583 | + MULTIPLY(tmp5, FIX(0.164081699)); /* c11 */ |
shoaib_ahmed | 0:791a779d6220 | 1584 | tmp11 = MULTIPLY(tmp2 + tmp3, - FIX(0.164081699)); /* -c11 */ |
shoaib_ahmed | 0:791a779d6220 | 1585 | tmp12 += tmp11 - tmp15 - MULTIPLY(tmp2, FIX(2.079550144)) /* c1+c5-c11 */ |
shoaib_ahmed | 0:791a779d6220 | 1586 | + MULTIPLY(tmp5, FIX(0.765261039)); /* c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1587 | tmp13 += tmp11 - tmp14 + MULTIPLY(tmp3, FIX(0.645144899)) /* c1+c11-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1588 | - MULTIPLY(tmp5, FIX(0.997307603)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 1589 | tmp11 = tmp15 + MULTIPLY(tmp0 - tmp3, FIX(1.161389302)) /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 1590 | - MULTIPLY(tmp2 + tmp5, FIX(0.481063200)); /* c9 */ |
shoaib_ahmed | 0:791a779d6220 | 1591 | |
shoaib_ahmed | 0:791a779d6220 | 1592 | dataptr[DCTSIZE*1] = (DCTELEM) DESCALE(tmp10, CONST_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 1593 | dataptr[DCTSIZE*3] = (DCTELEM) DESCALE(tmp11, CONST_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 1594 | dataptr[DCTSIZE*5] = (DCTELEM) DESCALE(tmp12, CONST_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 1595 | dataptr[DCTSIZE*7] = (DCTELEM) DESCALE(tmp13, CONST_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 1596 | |
shoaib_ahmed | 0:791a779d6220 | 1597 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 1598 | wsptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 1599 | } |
shoaib_ahmed | 0:791a779d6220 | 1600 | } |
shoaib_ahmed | 0:791a779d6220 | 1601 | |
shoaib_ahmed | 0:791a779d6220 | 1602 | |
shoaib_ahmed | 0:791a779d6220 | 1603 | /* |
shoaib_ahmed | 0:791a779d6220 | 1604 | * Perform the forward DCT on a 13x13 sample block. |
shoaib_ahmed | 0:791a779d6220 | 1605 | */ |
shoaib_ahmed | 0:791a779d6220 | 1606 | |
shoaib_ahmed | 0:791a779d6220 | 1607 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 1608 | jpeg_fdct_13x13 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 1609 | { |
shoaib_ahmed | 0:791a779d6220 | 1610 | INT32 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6; |
shoaib_ahmed | 0:791a779d6220 | 1611 | INT32 tmp10, tmp11, tmp12, tmp13, tmp14, tmp15; |
shoaib_ahmed | 0:791a779d6220 | 1612 | INT32 z1, z2; |
shoaib_ahmed | 0:791a779d6220 | 1613 | DCTELEM workspace[8*5]; |
shoaib_ahmed | 0:791a779d6220 | 1614 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 1615 | DCTELEM *wsptr; |
shoaib_ahmed | 0:791a779d6220 | 1616 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 1617 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 1618 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 1619 | |
shoaib_ahmed | 0:791a779d6220 | 1620 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 1621 | * Note results are scaled up by sqrt(8) compared to a true DCT. |
shoaib_ahmed | 0:791a779d6220 | 1622 | * cK represents sqrt(2) * cos(K*pi/26). |
shoaib_ahmed | 0:791a779d6220 | 1623 | */ |
shoaib_ahmed | 0:791a779d6220 | 1624 | |
shoaib_ahmed | 0:791a779d6220 | 1625 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 1626 | ctr = 0; |
shoaib_ahmed | 0:791a779d6220 | 1627 | for (;;) { |
shoaib_ahmed | 0:791a779d6220 | 1628 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 1629 | |
shoaib_ahmed | 0:791a779d6220 | 1630 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 1631 | |
shoaib_ahmed | 0:791a779d6220 | 1632 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[12]); |
shoaib_ahmed | 0:791a779d6220 | 1633 | tmp1 = GETJSAMPLE(elemptr[1]) + GETJSAMPLE(elemptr[11]); |
shoaib_ahmed | 0:791a779d6220 | 1634 | tmp2 = GETJSAMPLE(elemptr[2]) + GETJSAMPLE(elemptr[10]); |
shoaib_ahmed | 0:791a779d6220 | 1635 | tmp3 = GETJSAMPLE(elemptr[3]) + GETJSAMPLE(elemptr[9]); |
shoaib_ahmed | 0:791a779d6220 | 1636 | tmp4 = GETJSAMPLE(elemptr[4]) + GETJSAMPLE(elemptr[8]); |
shoaib_ahmed | 0:791a779d6220 | 1637 | tmp5 = GETJSAMPLE(elemptr[5]) + GETJSAMPLE(elemptr[7]); |
shoaib_ahmed | 0:791a779d6220 | 1638 | tmp6 = GETJSAMPLE(elemptr[6]); |
shoaib_ahmed | 0:791a779d6220 | 1639 | |
shoaib_ahmed | 0:791a779d6220 | 1640 | tmp10 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[12]); |
shoaib_ahmed | 0:791a779d6220 | 1641 | tmp11 = GETJSAMPLE(elemptr[1]) - GETJSAMPLE(elemptr[11]); |
shoaib_ahmed | 0:791a779d6220 | 1642 | tmp12 = GETJSAMPLE(elemptr[2]) - GETJSAMPLE(elemptr[10]); |
shoaib_ahmed | 0:791a779d6220 | 1643 | tmp13 = GETJSAMPLE(elemptr[3]) - GETJSAMPLE(elemptr[9]); |
shoaib_ahmed | 0:791a779d6220 | 1644 | tmp14 = GETJSAMPLE(elemptr[4]) - GETJSAMPLE(elemptr[8]); |
shoaib_ahmed | 0:791a779d6220 | 1645 | tmp15 = GETJSAMPLE(elemptr[5]) - GETJSAMPLE(elemptr[7]); |
shoaib_ahmed | 0:791a779d6220 | 1646 | |
shoaib_ahmed | 0:791a779d6220 | 1647 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 1648 | dataptr[0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1649 | (tmp0 + tmp1 + tmp2 + tmp3 + tmp4 + tmp5 + tmp6 - 13 * CENTERJSAMPLE); |
shoaib_ahmed | 0:791a779d6220 | 1650 | tmp6 += tmp6; |
shoaib_ahmed | 0:791a779d6220 | 1651 | tmp0 -= tmp6; |
shoaib_ahmed | 0:791a779d6220 | 1652 | tmp1 -= tmp6; |
shoaib_ahmed | 0:791a779d6220 | 1653 | tmp2 -= tmp6; |
shoaib_ahmed | 0:791a779d6220 | 1654 | tmp3 -= tmp6; |
shoaib_ahmed | 0:791a779d6220 | 1655 | tmp4 -= tmp6; |
shoaib_ahmed | 0:791a779d6220 | 1656 | tmp5 -= tmp6; |
shoaib_ahmed | 0:791a779d6220 | 1657 | dataptr[2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1658 | DESCALE(MULTIPLY(tmp0, FIX(1.373119086)) + /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 1659 | MULTIPLY(tmp1, FIX(1.058554052)) + /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 1660 | MULTIPLY(tmp2, FIX(0.501487041)) - /* c10 */ |
shoaib_ahmed | 0:791a779d6220 | 1661 | MULTIPLY(tmp3, FIX(0.170464608)) - /* c12 */ |
shoaib_ahmed | 0:791a779d6220 | 1662 | MULTIPLY(tmp4, FIX(0.803364869)) - /* c8 */ |
shoaib_ahmed | 0:791a779d6220 | 1663 | MULTIPLY(tmp5, FIX(1.252223920)), /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 1664 | CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 1665 | z1 = MULTIPLY(tmp0 - tmp2, FIX(1.155388986)) - /* (c4+c6)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 1666 | MULTIPLY(tmp3 - tmp4, FIX(0.435816023)) - /* (c2-c10)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 1667 | MULTIPLY(tmp1 - tmp5, FIX(0.316450131)); /* (c8-c12)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 1668 | z2 = MULTIPLY(tmp0 + tmp2, FIX(0.096834934)) - /* (c4-c6)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 1669 | MULTIPLY(tmp3 + tmp4, FIX(0.937303064)) + /* (c2+c10)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 1670 | MULTIPLY(tmp1 + tmp5, FIX(0.486914739)); /* (c8+c12)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 1671 | |
shoaib_ahmed | 0:791a779d6220 | 1672 | dataptr[4] = (DCTELEM) DESCALE(z1 + z2, CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 1673 | dataptr[6] = (DCTELEM) DESCALE(z1 - z2, CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 1674 | |
shoaib_ahmed | 0:791a779d6220 | 1675 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 1676 | |
shoaib_ahmed | 0:791a779d6220 | 1677 | tmp1 = MULTIPLY(tmp10 + tmp11, FIX(1.322312651)); /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 1678 | tmp2 = MULTIPLY(tmp10 + tmp12, FIX(1.163874945)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 1679 | tmp3 = MULTIPLY(tmp10 + tmp13, FIX(0.937797057)) + /* c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1680 | MULTIPLY(tmp14 + tmp15, FIX(0.338443458)); /* c11 */ |
shoaib_ahmed | 0:791a779d6220 | 1681 | tmp0 = tmp1 + tmp2 + tmp3 - |
shoaib_ahmed | 0:791a779d6220 | 1682 | MULTIPLY(tmp10, FIX(2.020082300)) + /* c3+c5+c7-c1 */ |
shoaib_ahmed | 0:791a779d6220 | 1683 | MULTIPLY(tmp14, FIX(0.318774355)); /* c9-c11 */ |
shoaib_ahmed | 0:791a779d6220 | 1684 | tmp4 = MULTIPLY(tmp14 - tmp15, FIX(0.937797057)) - /* c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1685 | MULTIPLY(tmp11 + tmp12, FIX(0.338443458)); /* c11 */ |
shoaib_ahmed | 0:791a779d6220 | 1686 | tmp5 = MULTIPLY(tmp11 + tmp13, - FIX(1.163874945)); /* -c5 */ |
shoaib_ahmed | 0:791a779d6220 | 1687 | tmp1 += tmp4 + tmp5 + |
shoaib_ahmed | 0:791a779d6220 | 1688 | MULTIPLY(tmp11, FIX(0.837223564)) - /* c5+c9+c11-c3 */ |
shoaib_ahmed | 0:791a779d6220 | 1689 | MULTIPLY(tmp14, FIX(2.341699410)); /* c1+c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1690 | tmp6 = MULTIPLY(tmp12 + tmp13, - FIX(0.657217813)); /* -c9 */ |
shoaib_ahmed | 0:791a779d6220 | 1691 | tmp2 += tmp4 + tmp6 - |
shoaib_ahmed | 0:791a779d6220 | 1692 | MULTIPLY(tmp12, FIX(1.572116027)) + /* c1+c5-c9-c11 */ |
shoaib_ahmed | 0:791a779d6220 | 1693 | MULTIPLY(tmp15, FIX(2.260109708)); /* c3+c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1694 | tmp3 += tmp5 + tmp6 + |
shoaib_ahmed | 0:791a779d6220 | 1695 | MULTIPLY(tmp13, FIX(2.205608352)) - /* c3+c5+c9-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1696 | MULTIPLY(tmp15, FIX(1.742345811)); /* c1+c11 */ |
shoaib_ahmed | 0:791a779d6220 | 1697 | |
shoaib_ahmed | 0:791a779d6220 | 1698 | dataptr[1] = (DCTELEM) DESCALE(tmp0, CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 1699 | dataptr[3] = (DCTELEM) DESCALE(tmp1, CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 1700 | dataptr[5] = (DCTELEM) DESCALE(tmp2, CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 1701 | dataptr[7] = (DCTELEM) DESCALE(tmp3, CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 1702 | |
shoaib_ahmed | 0:791a779d6220 | 1703 | ctr++; |
shoaib_ahmed | 0:791a779d6220 | 1704 | |
shoaib_ahmed | 0:791a779d6220 | 1705 | if (ctr != DCTSIZE) { |
shoaib_ahmed | 0:791a779d6220 | 1706 | if (ctr == 13) |
shoaib_ahmed | 0:791a779d6220 | 1707 | break; /* Done. */ |
shoaib_ahmed | 0:791a779d6220 | 1708 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 1709 | } else |
shoaib_ahmed | 0:791a779d6220 | 1710 | dataptr = workspace; /* switch pointer to extended workspace */ |
shoaib_ahmed | 0:791a779d6220 | 1711 | } |
shoaib_ahmed | 0:791a779d6220 | 1712 | |
shoaib_ahmed | 0:791a779d6220 | 1713 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 1714 | * We leave the results scaled up by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 1715 | * We must also scale the output by (8/13)**2 = 64/169, which we partially |
shoaib_ahmed | 0:791a779d6220 | 1716 | * fold into the constant multipliers and final shifting: |
shoaib_ahmed | 0:791a779d6220 | 1717 | * cK now represents sqrt(2) * cos(K*pi/26) * 128/169. |
shoaib_ahmed | 0:791a779d6220 | 1718 | */ |
shoaib_ahmed | 0:791a779d6220 | 1719 | |
shoaib_ahmed | 0:791a779d6220 | 1720 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 1721 | wsptr = workspace; |
shoaib_ahmed | 0:791a779d6220 | 1722 | for (ctr = DCTSIZE-1; ctr >= 0; ctr--) { |
shoaib_ahmed | 0:791a779d6220 | 1723 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 1724 | |
shoaib_ahmed | 0:791a779d6220 | 1725 | tmp0 = dataptr[DCTSIZE*0] + wsptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 1726 | tmp1 = dataptr[DCTSIZE*1] + wsptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 1727 | tmp2 = dataptr[DCTSIZE*2] + wsptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 1728 | tmp3 = dataptr[DCTSIZE*3] + wsptr[DCTSIZE*1]; |
shoaib_ahmed | 0:791a779d6220 | 1729 | tmp4 = dataptr[DCTSIZE*4] + wsptr[DCTSIZE*0]; |
shoaib_ahmed | 0:791a779d6220 | 1730 | tmp5 = dataptr[DCTSIZE*5] + dataptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 1731 | tmp6 = dataptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 1732 | |
shoaib_ahmed | 0:791a779d6220 | 1733 | tmp10 = dataptr[DCTSIZE*0] - wsptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 1734 | tmp11 = dataptr[DCTSIZE*1] - wsptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 1735 | tmp12 = dataptr[DCTSIZE*2] - wsptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 1736 | tmp13 = dataptr[DCTSIZE*3] - wsptr[DCTSIZE*1]; |
shoaib_ahmed | 0:791a779d6220 | 1737 | tmp14 = dataptr[DCTSIZE*4] - wsptr[DCTSIZE*0]; |
shoaib_ahmed | 0:791a779d6220 | 1738 | tmp15 = dataptr[DCTSIZE*5] - dataptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 1739 | |
shoaib_ahmed | 0:791a779d6220 | 1740 | dataptr[DCTSIZE*0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1741 | DESCALE(MULTIPLY(tmp0 + tmp1 + tmp2 + tmp3 + tmp4 + tmp5 + tmp6, |
shoaib_ahmed | 0:791a779d6220 | 1742 | FIX(0.757396450)), /* 128/169 */ |
shoaib_ahmed | 0:791a779d6220 | 1743 | CONST_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 1744 | tmp6 += tmp6; |
shoaib_ahmed | 0:791a779d6220 | 1745 | tmp0 -= tmp6; |
shoaib_ahmed | 0:791a779d6220 | 1746 | tmp1 -= tmp6; |
shoaib_ahmed | 0:791a779d6220 | 1747 | tmp2 -= tmp6; |
shoaib_ahmed | 0:791a779d6220 | 1748 | tmp3 -= tmp6; |
shoaib_ahmed | 0:791a779d6220 | 1749 | tmp4 -= tmp6; |
shoaib_ahmed | 0:791a779d6220 | 1750 | tmp5 -= tmp6; |
shoaib_ahmed | 0:791a779d6220 | 1751 | dataptr[DCTSIZE*2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1752 | DESCALE(MULTIPLY(tmp0, FIX(1.039995521)) + /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 1753 | MULTIPLY(tmp1, FIX(0.801745081)) + /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 1754 | MULTIPLY(tmp2, FIX(0.379824504)) - /* c10 */ |
shoaib_ahmed | 0:791a779d6220 | 1755 | MULTIPLY(tmp3, FIX(0.129109289)) - /* c12 */ |
shoaib_ahmed | 0:791a779d6220 | 1756 | MULTIPLY(tmp4, FIX(0.608465700)) - /* c8 */ |
shoaib_ahmed | 0:791a779d6220 | 1757 | MULTIPLY(tmp5, FIX(0.948429952)), /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 1758 | CONST_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 1759 | z1 = MULTIPLY(tmp0 - tmp2, FIX(0.875087516)) - /* (c4+c6)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 1760 | MULTIPLY(tmp3 - tmp4, FIX(0.330085509)) - /* (c2-c10)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 1761 | MULTIPLY(tmp1 - tmp5, FIX(0.239678205)); /* (c8-c12)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 1762 | z2 = MULTIPLY(tmp0 + tmp2, FIX(0.073342435)) - /* (c4-c6)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 1763 | MULTIPLY(tmp3 + tmp4, FIX(0.709910013)) + /* (c2+c10)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 1764 | MULTIPLY(tmp1 + tmp5, FIX(0.368787494)); /* (c8+c12)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 1765 | |
shoaib_ahmed | 0:791a779d6220 | 1766 | dataptr[DCTSIZE*4] = (DCTELEM) DESCALE(z1 + z2, CONST_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 1767 | dataptr[DCTSIZE*6] = (DCTELEM) DESCALE(z1 - z2, CONST_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 1768 | |
shoaib_ahmed | 0:791a779d6220 | 1769 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 1770 | |
shoaib_ahmed | 0:791a779d6220 | 1771 | tmp1 = MULTIPLY(tmp10 + tmp11, FIX(1.001514908)); /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 1772 | tmp2 = MULTIPLY(tmp10 + tmp12, FIX(0.881514751)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 1773 | tmp3 = MULTIPLY(tmp10 + tmp13, FIX(0.710284161)) + /* c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1774 | MULTIPLY(tmp14 + tmp15, FIX(0.256335874)); /* c11 */ |
shoaib_ahmed | 0:791a779d6220 | 1775 | tmp0 = tmp1 + tmp2 + tmp3 - |
shoaib_ahmed | 0:791a779d6220 | 1776 | MULTIPLY(tmp10, FIX(1.530003162)) + /* c3+c5+c7-c1 */ |
shoaib_ahmed | 0:791a779d6220 | 1777 | MULTIPLY(tmp14, FIX(0.241438564)); /* c9-c11 */ |
shoaib_ahmed | 0:791a779d6220 | 1778 | tmp4 = MULTIPLY(tmp14 - tmp15, FIX(0.710284161)) - /* c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1779 | MULTIPLY(tmp11 + tmp12, FIX(0.256335874)); /* c11 */ |
shoaib_ahmed | 0:791a779d6220 | 1780 | tmp5 = MULTIPLY(tmp11 + tmp13, - FIX(0.881514751)); /* -c5 */ |
shoaib_ahmed | 0:791a779d6220 | 1781 | tmp1 += tmp4 + tmp5 + |
shoaib_ahmed | 0:791a779d6220 | 1782 | MULTIPLY(tmp11, FIX(0.634110155)) - /* c5+c9+c11-c3 */ |
shoaib_ahmed | 0:791a779d6220 | 1783 | MULTIPLY(tmp14, FIX(1.773594819)); /* c1+c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1784 | tmp6 = MULTIPLY(tmp12 + tmp13, - FIX(0.497774438)); /* -c9 */ |
shoaib_ahmed | 0:791a779d6220 | 1785 | tmp2 += tmp4 + tmp6 - |
shoaib_ahmed | 0:791a779d6220 | 1786 | MULTIPLY(tmp12, FIX(1.190715098)) + /* c1+c5-c9-c11 */ |
shoaib_ahmed | 0:791a779d6220 | 1787 | MULTIPLY(tmp15, FIX(1.711799069)); /* c3+c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1788 | tmp3 += tmp5 + tmp6 + |
shoaib_ahmed | 0:791a779d6220 | 1789 | MULTIPLY(tmp13, FIX(1.670519935)) - /* c3+c5+c9-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 1790 | MULTIPLY(tmp15, FIX(1.319646532)); /* c1+c11 */ |
shoaib_ahmed | 0:791a779d6220 | 1791 | |
shoaib_ahmed | 0:791a779d6220 | 1792 | dataptr[DCTSIZE*1] = (DCTELEM) DESCALE(tmp0, CONST_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 1793 | dataptr[DCTSIZE*3] = (DCTELEM) DESCALE(tmp1, CONST_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 1794 | dataptr[DCTSIZE*5] = (DCTELEM) DESCALE(tmp2, CONST_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 1795 | dataptr[DCTSIZE*7] = (DCTELEM) DESCALE(tmp3, CONST_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 1796 | |
shoaib_ahmed | 0:791a779d6220 | 1797 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 1798 | wsptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 1799 | } |
shoaib_ahmed | 0:791a779d6220 | 1800 | } |
shoaib_ahmed | 0:791a779d6220 | 1801 | |
shoaib_ahmed | 0:791a779d6220 | 1802 | |
shoaib_ahmed | 0:791a779d6220 | 1803 | /* |
shoaib_ahmed | 0:791a779d6220 | 1804 | * Perform the forward DCT on a 14x14 sample block. |
shoaib_ahmed | 0:791a779d6220 | 1805 | */ |
shoaib_ahmed | 0:791a779d6220 | 1806 | |
shoaib_ahmed | 0:791a779d6220 | 1807 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 1808 | jpeg_fdct_14x14 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 1809 | { |
shoaib_ahmed | 0:791a779d6220 | 1810 | INT32 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6; |
shoaib_ahmed | 0:791a779d6220 | 1811 | INT32 tmp10, tmp11, tmp12, tmp13, tmp14, tmp15, tmp16; |
shoaib_ahmed | 0:791a779d6220 | 1812 | DCTELEM workspace[8*6]; |
shoaib_ahmed | 0:791a779d6220 | 1813 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 1814 | DCTELEM *wsptr; |
shoaib_ahmed | 0:791a779d6220 | 1815 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 1816 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 1817 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 1818 | |
shoaib_ahmed | 0:791a779d6220 | 1819 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 1820 | * Note results are scaled up by sqrt(8) compared to a true DCT. |
shoaib_ahmed | 0:791a779d6220 | 1821 | * cK represents sqrt(2) * cos(K*pi/28). |
shoaib_ahmed | 0:791a779d6220 | 1822 | */ |
shoaib_ahmed | 0:791a779d6220 | 1823 | |
shoaib_ahmed | 0:791a779d6220 | 1824 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 1825 | ctr = 0; |
shoaib_ahmed | 0:791a779d6220 | 1826 | for (;;) { |
shoaib_ahmed | 0:791a779d6220 | 1827 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 1828 | |
shoaib_ahmed | 0:791a779d6220 | 1829 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 1830 | |
shoaib_ahmed | 0:791a779d6220 | 1831 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[13]); |
shoaib_ahmed | 0:791a779d6220 | 1832 | tmp1 = GETJSAMPLE(elemptr[1]) + GETJSAMPLE(elemptr[12]); |
shoaib_ahmed | 0:791a779d6220 | 1833 | tmp2 = GETJSAMPLE(elemptr[2]) + GETJSAMPLE(elemptr[11]); |
shoaib_ahmed | 0:791a779d6220 | 1834 | tmp13 = GETJSAMPLE(elemptr[3]) + GETJSAMPLE(elemptr[10]); |
shoaib_ahmed | 0:791a779d6220 | 1835 | tmp4 = GETJSAMPLE(elemptr[4]) + GETJSAMPLE(elemptr[9]); |
shoaib_ahmed | 0:791a779d6220 | 1836 | tmp5 = GETJSAMPLE(elemptr[5]) + GETJSAMPLE(elemptr[8]); |
shoaib_ahmed | 0:791a779d6220 | 1837 | tmp6 = GETJSAMPLE(elemptr[6]) + GETJSAMPLE(elemptr[7]); |
shoaib_ahmed | 0:791a779d6220 | 1838 | |
shoaib_ahmed | 0:791a779d6220 | 1839 | tmp10 = tmp0 + tmp6; |
shoaib_ahmed | 0:791a779d6220 | 1840 | tmp14 = tmp0 - tmp6; |
shoaib_ahmed | 0:791a779d6220 | 1841 | tmp11 = tmp1 + tmp5; |
shoaib_ahmed | 0:791a779d6220 | 1842 | tmp15 = tmp1 - tmp5; |
shoaib_ahmed | 0:791a779d6220 | 1843 | tmp12 = tmp2 + tmp4; |
shoaib_ahmed | 0:791a779d6220 | 1844 | tmp16 = tmp2 - tmp4; |
shoaib_ahmed | 0:791a779d6220 | 1845 | |
shoaib_ahmed | 0:791a779d6220 | 1846 | tmp0 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[13]); |
shoaib_ahmed | 0:791a779d6220 | 1847 | tmp1 = GETJSAMPLE(elemptr[1]) - GETJSAMPLE(elemptr[12]); |
shoaib_ahmed | 0:791a779d6220 | 1848 | tmp2 = GETJSAMPLE(elemptr[2]) - GETJSAMPLE(elemptr[11]); |
shoaib_ahmed | 0:791a779d6220 | 1849 | tmp3 = GETJSAMPLE(elemptr[3]) - GETJSAMPLE(elemptr[10]); |
shoaib_ahmed | 0:791a779d6220 | 1850 | tmp4 = GETJSAMPLE(elemptr[4]) - GETJSAMPLE(elemptr[9]); |
shoaib_ahmed | 0:791a779d6220 | 1851 | tmp5 = GETJSAMPLE(elemptr[5]) - GETJSAMPLE(elemptr[8]); |
shoaib_ahmed | 0:791a779d6220 | 1852 | tmp6 = GETJSAMPLE(elemptr[6]) - GETJSAMPLE(elemptr[7]); |
shoaib_ahmed | 0:791a779d6220 | 1853 | |
shoaib_ahmed | 0:791a779d6220 | 1854 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 1855 | dataptr[0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1856 | (tmp10 + tmp11 + tmp12 + tmp13 - 14 * CENTERJSAMPLE); |
shoaib_ahmed | 0:791a779d6220 | 1857 | tmp13 += tmp13; |
shoaib_ahmed | 0:791a779d6220 | 1858 | dataptr[4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1859 | DESCALE(MULTIPLY(tmp10 - tmp13, FIX(1.274162392)) + /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 1860 | MULTIPLY(tmp11 - tmp13, FIX(0.314692123)) - /* c12 */ |
shoaib_ahmed | 0:791a779d6220 | 1861 | MULTIPLY(tmp12 - tmp13, FIX(0.881747734)), /* c8 */ |
shoaib_ahmed | 0:791a779d6220 | 1862 | CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 1863 | |
shoaib_ahmed | 0:791a779d6220 | 1864 | tmp10 = MULTIPLY(tmp14 + tmp15, FIX(1.105676686)); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 1865 | |
shoaib_ahmed | 0:791a779d6220 | 1866 | dataptr[2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1867 | DESCALE(tmp10 + MULTIPLY(tmp14, FIX(0.273079590)) /* c2-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 1868 | + MULTIPLY(tmp16, FIX(0.613604268)), /* c10 */ |
shoaib_ahmed | 0:791a779d6220 | 1869 | CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 1870 | dataptr[6] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1871 | DESCALE(tmp10 - MULTIPLY(tmp15, FIX(1.719280954)) /* c6+c10 */ |
shoaib_ahmed | 0:791a779d6220 | 1872 | - MULTIPLY(tmp16, FIX(1.378756276)), /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 1873 | CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 1874 | |
shoaib_ahmed | 0:791a779d6220 | 1875 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 1876 | |
shoaib_ahmed | 0:791a779d6220 | 1877 | tmp10 = tmp1 + tmp2; |
shoaib_ahmed | 0:791a779d6220 | 1878 | tmp11 = tmp5 - tmp4; |
shoaib_ahmed | 0:791a779d6220 | 1879 | dataptr[7] = (DCTELEM) (tmp0 - tmp10 + tmp3 - tmp11 - tmp6); |
shoaib_ahmed | 0:791a779d6220 | 1880 | tmp3 <<= CONST_BITS; |
shoaib_ahmed | 0:791a779d6220 | 1881 | tmp10 = MULTIPLY(tmp10, - FIX(0.158341681)); /* -c13 */ |
shoaib_ahmed | 0:791a779d6220 | 1882 | tmp11 = MULTIPLY(tmp11, FIX(1.405321284)); /* c1 */ |
shoaib_ahmed | 0:791a779d6220 | 1883 | tmp10 += tmp11 - tmp3; |
shoaib_ahmed | 0:791a779d6220 | 1884 | tmp11 = MULTIPLY(tmp0 + tmp2, FIX(1.197448846)) + /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 1885 | MULTIPLY(tmp4 + tmp6, FIX(0.752406978)); /* c9 */ |
shoaib_ahmed | 0:791a779d6220 | 1886 | dataptr[5] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1887 | DESCALE(tmp10 + tmp11 - MULTIPLY(tmp2, FIX(2.373959773)) /* c3+c5-c13 */ |
shoaib_ahmed | 0:791a779d6220 | 1888 | + MULTIPLY(tmp4, FIX(1.119999435)), /* c1+c11-c9 */ |
shoaib_ahmed | 0:791a779d6220 | 1889 | CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 1890 | tmp12 = MULTIPLY(tmp0 + tmp1, FIX(1.334852607)) + /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 1891 | MULTIPLY(tmp5 - tmp6, FIX(0.467085129)); /* c11 */ |
shoaib_ahmed | 0:791a779d6220 | 1892 | dataptr[3] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1893 | DESCALE(tmp10 + tmp12 - MULTIPLY(tmp1, FIX(0.424103948)) /* c3-c9-c13 */ |
shoaib_ahmed | 0:791a779d6220 | 1894 | - MULTIPLY(tmp5, FIX(3.069855259)), /* c1+c5+c11 */ |
shoaib_ahmed | 0:791a779d6220 | 1895 | CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 1896 | dataptr[1] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1897 | DESCALE(tmp11 + tmp12 + tmp3 + tmp6 - |
shoaib_ahmed | 0:791a779d6220 | 1898 | MULTIPLY(tmp0 + tmp6, FIX(1.126980169)), /* c3+c5-c1 */ |
shoaib_ahmed | 0:791a779d6220 | 1899 | CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 1900 | |
shoaib_ahmed | 0:791a779d6220 | 1901 | ctr++; |
shoaib_ahmed | 0:791a779d6220 | 1902 | |
shoaib_ahmed | 0:791a779d6220 | 1903 | if (ctr != DCTSIZE) { |
shoaib_ahmed | 0:791a779d6220 | 1904 | if (ctr == 14) |
shoaib_ahmed | 0:791a779d6220 | 1905 | break; /* Done. */ |
shoaib_ahmed | 0:791a779d6220 | 1906 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 1907 | } else |
shoaib_ahmed | 0:791a779d6220 | 1908 | dataptr = workspace; /* switch pointer to extended workspace */ |
shoaib_ahmed | 0:791a779d6220 | 1909 | } |
shoaib_ahmed | 0:791a779d6220 | 1910 | |
shoaib_ahmed | 0:791a779d6220 | 1911 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 1912 | * We leave the results scaled up by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 1913 | * We must also scale the output by (8/14)**2 = 16/49, which we partially |
shoaib_ahmed | 0:791a779d6220 | 1914 | * fold into the constant multipliers and final shifting: |
shoaib_ahmed | 0:791a779d6220 | 1915 | * cK now represents sqrt(2) * cos(K*pi/28) * 32/49. |
shoaib_ahmed | 0:791a779d6220 | 1916 | */ |
shoaib_ahmed | 0:791a779d6220 | 1917 | |
shoaib_ahmed | 0:791a779d6220 | 1918 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 1919 | wsptr = workspace; |
shoaib_ahmed | 0:791a779d6220 | 1920 | for (ctr = DCTSIZE-1; ctr >= 0; ctr--) { |
shoaib_ahmed | 0:791a779d6220 | 1921 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 1922 | |
shoaib_ahmed | 0:791a779d6220 | 1923 | tmp0 = dataptr[DCTSIZE*0] + wsptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 1924 | tmp1 = dataptr[DCTSIZE*1] + wsptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 1925 | tmp2 = dataptr[DCTSIZE*2] + wsptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 1926 | tmp13 = dataptr[DCTSIZE*3] + wsptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 1927 | tmp4 = dataptr[DCTSIZE*4] + wsptr[DCTSIZE*1]; |
shoaib_ahmed | 0:791a779d6220 | 1928 | tmp5 = dataptr[DCTSIZE*5] + wsptr[DCTSIZE*0]; |
shoaib_ahmed | 0:791a779d6220 | 1929 | tmp6 = dataptr[DCTSIZE*6] + dataptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 1930 | |
shoaib_ahmed | 0:791a779d6220 | 1931 | tmp10 = tmp0 + tmp6; |
shoaib_ahmed | 0:791a779d6220 | 1932 | tmp14 = tmp0 - tmp6; |
shoaib_ahmed | 0:791a779d6220 | 1933 | tmp11 = tmp1 + tmp5; |
shoaib_ahmed | 0:791a779d6220 | 1934 | tmp15 = tmp1 - tmp5; |
shoaib_ahmed | 0:791a779d6220 | 1935 | tmp12 = tmp2 + tmp4; |
shoaib_ahmed | 0:791a779d6220 | 1936 | tmp16 = tmp2 - tmp4; |
shoaib_ahmed | 0:791a779d6220 | 1937 | |
shoaib_ahmed | 0:791a779d6220 | 1938 | tmp0 = dataptr[DCTSIZE*0] - wsptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 1939 | tmp1 = dataptr[DCTSIZE*1] - wsptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 1940 | tmp2 = dataptr[DCTSIZE*2] - wsptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 1941 | tmp3 = dataptr[DCTSIZE*3] - wsptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 1942 | tmp4 = dataptr[DCTSIZE*4] - wsptr[DCTSIZE*1]; |
shoaib_ahmed | 0:791a779d6220 | 1943 | tmp5 = dataptr[DCTSIZE*5] - wsptr[DCTSIZE*0]; |
shoaib_ahmed | 0:791a779d6220 | 1944 | tmp6 = dataptr[DCTSIZE*6] - dataptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 1945 | |
shoaib_ahmed | 0:791a779d6220 | 1946 | dataptr[DCTSIZE*0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1947 | DESCALE(MULTIPLY(tmp10 + tmp11 + tmp12 + tmp13, |
shoaib_ahmed | 0:791a779d6220 | 1948 | FIX(0.653061224)), /* 32/49 */ |
shoaib_ahmed | 0:791a779d6220 | 1949 | CONST_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 1950 | tmp13 += tmp13; |
shoaib_ahmed | 0:791a779d6220 | 1951 | dataptr[DCTSIZE*4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1952 | DESCALE(MULTIPLY(tmp10 - tmp13, FIX(0.832106052)) + /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 1953 | MULTIPLY(tmp11 - tmp13, FIX(0.205513223)) - /* c12 */ |
shoaib_ahmed | 0:791a779d6220 | 1954 | MULTIPLY(tmp12 - tmp13, FIX(0.575835255)), /* c8 */ |
shoaib_ahmed | 0:791a779d6220 | 1955 | CONST_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 1956 | |
shoaib_ahmed | 0:791a779d6220 | 1957 | tmp10 = MULTIPLY(tmp14 + tmp15, FIX(0.722074570)); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 1958 | |
shoaib_ahmed | 0:791a779d6220 | 1959 | dataptr[DCTSIZE*2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1960 | DESCALE(tmp10 + MULTIPLY(tmp14, FIX(0.178337691)) /* c2-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 1961 | + MULTIPLY(tmp16, FIX(0.400721155)), /* c10 */ |
shoaib_ahmed | 0:791a779d6220 | 1962 | CONST_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 1963 | dataptr[DCTSIZE*6] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1964 | DESCALE(tmp10 - MULTIPLY(tmp15, FIX(1.122795725)) /* c6+c10 */ |
shoaib_ahmed | 0:791a779d6220 | 1965 | - MULTIPLY(tmp16, FIX(0.900412262)), /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 1966 | CONST_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 1967 | |
shoaib_ahmed | 0:791a779d6220 | 1968 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 1969 | |
shoaib_ahmed | 0:791a779d6220 | 1970 | tmp10 = tmp1 + tmp2; |
shoaib_ahmed | 0:791a779d6220 | 1971 | tmp11 = tmp5 - tmp4; |
shoaib_ahmed | 0:791a779d6220 | 1972 | dataptr[DCTSIZE*7] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1973 | DESCALE(MULTIPLY(tmp0 - tmp10 + tmp3 - tmp11 - tmp6, |
shoaib_ahmed | 0:791a779d6220 | 1974 | FIX(0.653061224)), /* 32/49 */ |
shoaib_ahmed | 0:791a779d6220 | 1975 | CONST_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 1976 | tmp3 = MULTIPLY(tmp3 , FIX(0.653061224)); /* 32/49 */ |
shoaib_ahmed | 0:791a779d6220 | 1977 | tmp10 = MULTIPLY(tmp10, - FIX(0.103406812)); /* -c13 */ |
shoaib_ahmed | 0:791a779d6220 | 1978 | tmp11 = MULTIPLY(tmp11, FIX(0.917760839)); /* c1 */ |
shoaib_ahmed | 0:791a779d6220 | 1979 | tmp10 += tmp11 - tmp3; |
shoaib_ahmed | 0:791a779d6220 | 1980 | tmp11 = MULTIPLY(tmp0 + tmp2, FIX(0.782007410)) + /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 1981 | MULTIPLY(tmp4 + tmp6, FIX(0.491367823)); /* c9 */ |
shoaib_ahmed | 0:791a779d6220 | 1982 | dataptr[DCTSIZE*5] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1983 | DESCALE(tmp10 + tmp11 - MULTIPLY(tmp2, FIX(1.550341076)) /* c3+c5-c13 */ |
shoaib_ahmed | 0:791a779d6220 | 1984 | + MULTIPLY(tmp4, FIX(0.731428202)), /* c1+c11-c9 */ |
shoaib_ahmed | 0:791a779d6220 | 1985 | CONST_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 1986 | tmp12 = MULTIPLY(tmp0 + tmp1, FIX(0.871740478)) + /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 1987 | MULTIPLY(tmp5 - tmp6, FIX(0.305035186)); /* c11 */ |
shoaib_ahmed | 0:791a779d6220 | 1988 | dataptr[DCTSIZE*3] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1989 | DESCALE(tmp10 + tmp12 - MULTIPLY(tmp1, FIX(0.276965844)) /* c3-c9-c13 */ |
shoaib_ahmed | 0:791a779d6220 | 1990 | - MULTIPLY(tmp5, FIX(2.004803435)), /* c1+c5+c11 */ |
shoaib_ahmed | 0:791a779d6220 | 1991 | CONST_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 1992 | dataptr[DCTSIZE*1] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 1993 | DESCALE(tmp11 + tmp12 + tmp3 |
shoaib_ahmed | 0:791a779d6220 | 1994 | - MULTIPLY(tmp0, FIX(0.735987049)) /* c3+c5-c1 */ |
shoaib_ahmed | 0:791a779d6220 | 1995 | - MULTIPLY(tmp6, FIX(0.082925825)), /* c9-c11-c13 */ |
shoaib_ahmed | 0:791a779d6220 | 1996 | CONST_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 1997 | |
shoaib_ahmed | 0:791a779d6220 | 1998 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 1999 | wsptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 2000 | } |
shoaib_ahmed | 0:791a779d6220 | 2001 | } |
shoaib_ahmed | 0:791a779d6220 | 2002 | |
shoaib_ahmed | 0:791a779d6220 | 2003 | |
shoaib_ahmed | 0:791a779d6220 | 2004 | /* |
shoaib_ahmed | 0:791a779d6220 | 2005 | * Perform the forward DCT on a 15x15 sample block. |
shoaib_ahmed | 0:791a779d6220 | 2006 | */ |
shoaib_ahmed | 0:791a779d6220 | 2007 | |
shoaib_ahmed | 0:791a779d6220 | 2008 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 2009 | jpeg_fdct_15x15 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 2010 | { |
shoaib_ahmed | 0:791a779d6220 | 2011 | INT32 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7; |
shoaib_ahmed | 0:791a779d6220 | 2012 | INT32 tmp10, tmp11, tmp12, tmp13, tmp14, tmp15, tmp16; |
shoaib_ahmed | 0:791a779d6220 | 2013 | INT32 z1, z2, z3; |
shoaib_ahmed | 0:791a779d6220 | 2014 | DCTELEM workspace[8*7]; |
shoaib_ahmed | 0:791a779d6220 | 2015 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 2016 | DCTELEM *wsptr; |
shoaib_ahmed | 0:791a779d6220 | 2017 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 2018 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 2019 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 2020 | |
shoaib_ahmed | 0:791a779d6220 | 2021 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 2022 | * Note results are scaled up by sqrt(8) compared to a true DCT. |
shoaib_ahmed | 0:791a779d6220 | 2023 | * cK represents sqrt(2) * cos(K*pi/30). |
shoaib_ahmed | 0:791a779d6220 | 2024 | */ |
shoaib_ahmed | 0:791a779d6220 | 2025 | |
shoaib_ahmed | 0:791a779d6220 | 2026 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 2027 | ctr = 0; |
shoaib_ahmed | 0:791a779d6220 | 2028 | for (;;) { |
shoaib_ahmed | 0:791a779d6220 | 2029 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 2030 | |
shoaib_ahmed | 0:791a779d6220 | 2031 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 2032 | |
shoaib_ahmed | 0:791a779d6220 | 2033 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[14]); |
shoaib_ahmed | 0:791a779d6220 | 2034 | tmp1 = GETJSAMPLE(elemptr[1]) + GETJSAMPLE(elemptr[13]); |
shoaib_ahmed | 0:791a779d6220 | 2035 | tmp2 = GETJSAMPLE(elemptr[2]) + GETJSAMPLE(elemptr[12]); |
shoaib_ahmed | 0:791a779d6220 | 2036 | tmp3 = GETJSAMPLE(elemptr[3]) + GETJSAMPLE(elemptr[11]); |
shoaib_ahmed | 0:791a779d6220 | 2037 | tmp4 = GETJSAMPLE(elemptr[4]) + GETJSAMPLE(elemptr[10]); |
shoaib_ahmed | 0:791a779d6220 | 2038 | tmp5 = GETJSAMPLE(elemptr[5]) + GETJSAMPLE(elemptr[9]); |
shoaib_ahmed | 0:791a779d6220 | 2039 | tmp6 = GETJSAMPLE(elemptr[6]) + GETJSAMPLE(elemptr[8]); |
shoaib_ahmed | 0:791a779d6220 | 2040 | tmp7 = GETJSAMPLE(elemptr[7]); |
shoaib_ahmed | 0:791a779d6220 | 2041 | |
shoaib_ahmed | 0:791a779d6220 | 2042 | tmp10 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[14]); |
shoaib_ahmed | 0:791a779d6220 | 2043 | tmp11 = GETJSAMPLE(elemptr[1]) - GETJSAMPLE(elemptr[13]); |
shoaib_ahmed | 0:791a779d6220 | 2044 | tmp12 = GETJSAMPLE(elemptr[2]) - GETJSAMPLE(elemptr[12]); |
shoaib_ahmed | 0:791a779d6220 | 2045 | tmp13 = GETJSAMPLE(elemptr[3]) - GETJSAMPLE(elemptr[11]); |
shoaib_ahmed | 0:791a779d6220 | 2046 | tmp14 = GETJSAMPLE(elemptr[4]) - GETJSAMPLE(elemptr[10]); |
shoaib_ahmed | 0:791a779d6220 | 2047 | tmp15 = GETJSAMPLE(elemptr[5]) - GETJSAMPLE(elemptr[9]); |
shoaib_ahmed | 0:791a779d6220 | 2048 | tmp16 = GETJSAMPLE(elemptr[6]) - GETJSAMPLE(elemptr[8]); |
shoaib_ahmed | 0:791a779d6220 | 2049 | |
shoaib_ahmed | 0:791a779d6220 | 2050 | z1 = tmp0 + tmp4 + tmp5; |
shoaib_ahmed | 0:791a779d6220 | 2051 | z2 = tmp1 + tmp3 + tmp6; |
shoaib_ahmed | 0:791a779d6220 | 2052 | z3 = tmp2 + tmp7; |
shoaib_ahmed | 0:791a779d6220 | 2053 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 2054 | dataptr[0] = (DCTELEM) (z1 + z2 + z3 - 15 * CENTERJSAMPLE); |
shoaib_ahmed | 0:791a779d6220 | 2055 | z3 += z3; |
shoaib_ahmed | 0:791a779d6220 | 2056 | dataptr[6] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2057 | DESCALE(MULTIPLY(z1 - z3, FIX(1.144122806)) - /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 2058 | MULTIPLY(z2 - z3, FIX(0.437016024)), /* c12 */ |
shoaib_ahmed | 0:791a779d6220 | 2059 | CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2060 | tmp2 += ((tmp1 + tmp4) >> 1) - tmp7 - tmp7; |
shoaib_ahmed | 0:791a779d6220 | 2061 | z1 = MULTIPLY(tmp3 - tmp2, FIX(1.531135173)) - /* c2+c14 */ |
shoaib_ahmed | 0:791a779d6220 | 2062 | MULTIPLY(tmp6 - tmp2, FIX(2.238241955)); /* c4+c8 */ |
shoaib_ahmed | 0:791a779d6220 | 2063 | z2 = MULTIPLY(tmp5 - tmp2, FIX(0.798468008)) - /* c8-c14 */ |
shoaib_ahmed | 0:791a779d6220 | 2064 | MULTIPLY(tmp0 - tmp2, FIX(0.091361227)); /* c2-c4 */ |
shoaib_ahmed | 0:791a779d6220 | 2065 | z3 = MULTIPLY(tmp0 - tmp3, FIX(1.383309603)) + /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 2066 | MULTIPLY(tmp6 - tmp5, FIX(0.946293579)) + /* c8 */ |
shoaib_ahmed | 0:791a779d6220 | 2067 | MULTIPLY(tmp1 - tmp4, FIX(0.790569415)); /* (c6+c12)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 2068 | |
shoaib_ahmed | 0:791a779d6220 | 2069 | dataptr[2] = (DCTELEM) DESCALE(z1 + z3, CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2070 | dataptr[4] = (DCTELEM) DESCALE(z2 + z3, CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2071 | |
shoaib_ahmed | 0:791a779d6220 | 2072 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 2073 | |
shoaib_ahmed | 0:791a779d6220 | 2074 | tmp2 = MULTIPLY(tmp10 - tmp12 - tmp13 + tmp15 + tmp16, |
shoaib_ahmed | 0:791a779d6220 | 2075 | FIX(1.224744871)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 2076 | tmp1 = MULTIPLY(tmp10 - tmp14 - tmp15, FIX(1.344997024)) + /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 2077 | MULTIPLY(tmp11 - tmp13 - tmp16, FIX(0.831253876)); /* c9 */ |
shoaib_ahmed | 0:791a779d6220 | 2078 | tmp12 = MULTIPLY(tmp12, FIX(1.224744871)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 2079 | tmp4 = MULTIPLY(tmp10 - tmp16, FIX(1.406466353)) + /* c1 */ |
shoaib_ahmed | 0:791a779d6220 | 2080 | MULTIPLY(tmp11 + tmp14, FIX(1.344997024)) + /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 2081 | MULTIPLY(tmp13 + tmp15, FIX(0.575212477)); /* c11 */ |
shoaib_ahmed | 0:791a779d6220 | 2082 | tmp0 = MULTIPLY(tmp13, FIX(0.475753014)) - /* c7-c11 */ |
shoaib_ahmed | 0:791a779d6220 | 2083 | MULTIPLY(tmp14, FIX(0.513743148)) + /* c3-c9 */ |
shoaib_ahmed | 0:791a779d6220 | 2084 | MULTIPLY(tmp16, FIX(1.700497885)) + tmp4 + tmp12; /* c1+c13 */ |
shoaib_ahmed | 0:791a779d6220 | 2085 | tmp3 = MULTIPLY(tmp10, - FIX(0.355500862)) - /* -(c1-c7) */ |
shoaib_ahmed | 0:791a779d6220 | 2086 | MULTIPLY(tmp11, FIX(2.176250899)) - /* c3+c9 */ |
shoaib_ahmed | 0:791a779d6220 | 2087 | MULTIPLY(tmp15, FIX(0.869244010)) + tmp4 - tmp12; /* c11+c13 */ |
shoaib_ahmed | 0:791a779d6220 | 2088 | |
shoaib_ahmed | 0:791a779d6220 | 2089 | dataptr[1] = (DCTELEM) DESCALE(tmp0, CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2090 | dataptr[3] = (DCTELEM) DESCALE(tmp1, CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2091 | dataptr[5] = (DCTELEM) DESCALE(tmp2, CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2092 | dataptr[7] = (DCTELEM) DESCALE(tmp3, CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2093 | |
shoaib_ahmed | 0:791a779d6220 | 2094 | ctr++; |
shoaib_ahmed | 0:791a779d6220 | 2095 | |
shoaib_ahmed | 0:791a779d6220 | 2096 | if (ctr != DCTSIZE) { |
shoaib_ahmed | 0:791a779d6220 | 2097 | if (ctr == 15) |
shoaib_ahmed | 0:791a779d6220 | 2098 | break; /* Done. */ |
shoaib_ahmed | 0:791a779d6220 | 2099 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 2100 | } else |
shoaib_ahmed | 0:791a779d6220 | 2101 | dataptr = workspace; /* switch pointer to extended workspace */ |
shoaib_ahmed | 0:791a779d6220 | 2102 | } |
shoaib_ahmed | 0:791a779d6220 | 2103 | |
shoaib_ahmed | 0:791a779d6220 | 2104 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 2105 | * We leave the results scaled up by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 2106 | * We must also scale the output by (8/15)**2 = 64/225, which we partially |
shoaib_ahmed | 0:791a779d6220 | 2107 | * fold into the constant multipliers and final shifting: |
shoaib_ahmed | 0:791a779d6220 | 2108 | * cK now represents sqrt(2) * cos(K*pi/30) * 256/225. |
shoaib_ahmed | 0:791a779d6220 | 2109 | */ |
shoaib_ahmed | 0:791a779d6220 | 2110 | |
shoaib_ahmed | 0:791a779d6220 | 2111 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 2112 | wsptr = workspace; |
shoaib_ahmed | 0:791a779d6220 | 2113 | for (ctr = DCTSIZE-1; ctr >= 0; ctr--) { |
shoaib_ahmed | 0:791a779d6220 | 2114 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 2115 | |
shoaib_ahmed | 0:791a779d6220 | 2116 | tmp0 = dataptr[DCTSIZE*0] + wsptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 2117 | tmp1 = dataptr[DCTSIZE*1] + wsptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 2118 | tmp2 = dataptr[DCTSIZE*2] + wsptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 2119 | tmp3 = dataptr[DCTSIZE*3] + wsptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 2120 | tmp4 = dataptr[DCTSIZE*4] + wsptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 2121 | tmp5 = dataptr[DCTSIZE*5] + wsptr[DCTSIZE*1]; |
shoaib_ahmed | 0:791a779d6220 | 2122 | tmp6 = dataptr[DCTSIZE*6] + wsptr[DCTSIZE*0]; |
shoaib_ahmed | 0:791a779d6220 | 2123 | tmp7 = dataptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 2124 | |
shoaib_ahmed | 0:791a779d6220 | 2125 | tmp10 = dataptr[DCTSIZE*0] - wsptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 2126 | tmp11 = dataptr[DCTSIZE*1] - wsptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 2127 | tmp12 = dataptr[DCTSIZE*2] - wsptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 2128 | tmp13 = dataptr[DCTSIZE*3] - wsptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 2129 | tmp14 = dataptr[DCTSIZE*4] - wsptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 2130 | tmp15 = dataptr[DCTSIZE*5] - wsptr[DCTSIZE*1]; |
shoaib_ahmed | 0:791a779d6220 | 2131 | tmp16 = dataptr[DCTSIZE*6] - wsptr[DCTSIZE*0]; |
shoaib_ahmed | 0:791a779d6220 | 2132 | |
shoaib_ahmed | 0:791a779d6220 | 2133 | z1 = tmp0 + tmp4 + tmp5; |
shoaib_ahmed | 0:791a779d6220 | 2134 | z2 = tmp1 + tmp3 + tmp6; |
shoaib_ahmed | 0:791a779d6220 | 2135 | z3 = tmp2 + tmp7; |
shoaib_ahmed | 0:791a779d6220 | 2136 | dataptr[DCTSIZE*0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2137 | DESCALE(MULTIPLY(z1 + z2 + z3, FIX(1.137777778)), /* 256/225 */ |
shoaib_ahmed | 0:791a779d6220 | 2138 | CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 2139 | z3 += z3; |
shoaib_ahmed | 0:791a779d6220 | 2140 | dataptr[DCTSIZE*6] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2141 | DESCALE(MULTIPLY(z1 - z3, FIX(1.301757503)) - /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 2142 | MULTIPLY(z2 - z3, FIX(0.497227121)), /* c12 */ |
shoaib_ahmed | 0:791a779d6220 | 2143 | CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 2144 | tmp2 += ((tmp1 + tmp4) >> 1) - tmp7 - tmp7; |
shoaib_ahmed | 0:791a779d6220 | 2145 | z1 = MULTIPLY(tmp3 - tmp2, FIX(1.742091575)) - /* c2+c14 */ |
shoaib_ahmed | 0:791a779d6220 | 2146 | MULTIPLY(tmp6 - tmp2, FIX(2.546621957)); /* c4+c8 */ |
shoaib_ahmed | 0:791a779d6220 | 2147 | z2 = MULTIPLY(tmp5 - tmp2, FIX(0.908479156)) - /* c8-c14 */ |
shoaib_ahmed | 0:791a779d6220 | 2148 | MULTIPLY(tmp0 - tmp2, FIX(0.103948774)); /* c2-c4 */ |
shoaib_ahmed | 0:791a779d6220 | 2149 | z3 = MULTIPLY(tmp0 - tmp3, FIX(1.573898926)) + /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 2150 | MULTIPLY(tmp6 - tmp5, FIX(1.076671805)) + /* c8 */ |
shoaib_ahmed | 0:791a779d6220 | 2151 | MULTIPLY(tmp1 - tmp4, FIX(0.899492312)); /* (c6+c12)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 2152 | |
shoaib_ahmed | 0:791a779d6220 | 2153 | dataptr[DCTSIZE*2] = (DCTELEM) DESCALE(z1 + z3, CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 2154 | dataptr[DCTSIZE*4] = (DCTELEM) DESCALE(z2 + z3, CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 2155 | |
shoaib_ahmed | 0:791a779d6220 | 2156 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 2157 | |
shoaib_ahmed | 0:791a779d6220 | 2158 | tmp2 = MULTIPLY(tmp10 - tmp12 - tmp13 + tmp15 + tmp16, |
shoaib_ahmed | 0:791a779d6220 | 2159 | FIX(1.393487498)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 2160 | tmp1 = MULTIPLY(tmp10 - tmp14 - tmp15, FIX(1.530307725)) + /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 2161 | MULTIPLY(tmp11 - tmp13 - tmp16, FIX(0.945782187)); /* c9 */ |
shoaib_ahmed | 0:791a779d6220 | 2162 | tmp12 = MULTIPLY(tmp12, FIX(1.393487498)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 2163 | tmp4 = MULTIPLY(tmp10 - tmp16, FIX(1.600246161)) + /* c1 */ |
shoaib_ahmed | 0:791a779d6220 | 2164 | MULTIPLY(tmp11 + tmp14, FIX(1.530307725)) + /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 2165 | MULTIPLY(tmp13 + tmp15, FIX(0.654463974)); /* c11 */ |
shoaib_ahmed | 0:791a779d6220 | 2166 | tmp0 = MULTIPLY(tmp13, FIX(0.541301207)) - /* c7-c11 */ |
shoaib_ahmed | 0:791a779d6220 | 2167 | MULTIPLY(tmp14, FIX(0.584525538)) + /* c3-c9 */ |
shoaib_ahmed | 0:791a779d6220 | 2168 | MULTIPLY(tmp16, FIX(1.934788705)) + tmp4 + tmp12; /* c1+c13 */ |
shoaib_ahmed | 0:791a779d6220 | 2169 | tmp3 = MULTIPLY(tmp10, - FIX(0.404480980)) - /* -(c1-c7) */ |
shoaib_ahmed | 0:791a779d6220 | 2170 | MULTIPLY(tmp11, FIX(2.476089912)) - /* c3+c9 */ |
shoaib_ahmed | 0:791a779d6220 | 2171 | MULTIPLY(tmp15, FIX(0.989006518)) + tmp4 - tmp12; /* c11+c13 */ |
shoaib_ahmed | 0:791a779d6220 | 2172 | |
shoaib_ahmed | 0:791a779d6220 | 2173 | dataptr[DCTSIZE*1] = (DCTELEM) DESCALE(tmp0, CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 2174 | dataptr[DCTSIZE*3] = (DCTELEM) DESCALE(tmp1, CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 2175 | dataptr[DCTSIZE*5] = (DCTELEM) DESCALE(tmp2, CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 2176 | dataptr[DCTSIZE*7] = (DCTELEM) DESCALE(tmp3, CONST_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 2177 | |
shoaib_ahmed | 0:791a779d6220 | 2178 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 2179 | wsptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 2180 | } |
shoaib_ahmed | 0:791a779d6220 | 2181 | } |
shoaib_ahmed | 0:791a779d6220 | 2182 | |
shoaib_ahmed | 0:791a779d6220 | 2183 | |
shoaib_ahmed | 0:791a779d6220 | 2184 | /* |
shoaib_ahmed | 0:791a779d6220 | 2185 | * Perform the forward DCT on a 16x16 sample block. |
shoaib_ahmed | 0:791a779d6220 | 2186 | */ |
shoaib_ahmed | 0:791a779d6220 | 2187 | |
shoaib_ahmed | 0:791a779d6220 | 2188 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 2189 | jpeg_fdct_16x16 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 2190 | { |
shoaib_ahmed | 0:791a779d6220 | 2191 | INT32 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7; |
shoaib_ahmed | 0:791a779d6220 | 2192 | INT32 tmp10, tmp11, tmp12, tmp13, tmp14, tmp15, tmp16, tmp17; |
shoaib_ahmed | 0:791a779d6220 | 2193 | DCTELEM workspace[DCTSIZE2]; |
shoaib_ahmed | 0:791a779d6220 | 2194 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 2195 | DCTELEM *wsptr; |
shoaib_ahmed | 0:791a779d6220 | 2196 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 2197 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 2198 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 2199 | |
shoaib_ahmed | 0:791a779d6220 | 2200 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 2201 | * Note results are scaled up by sqrt(8) compared to a true DCT; |
shoaib_ahmed | 0:791a779d6220 | 2202 | * furthermore, we scale the results by 2**PASS1_BITS. |
shoaib_ahmed | 0:791a779d6220 | 2203 | * cK represents sqrt(2) * cos(K*pi/32). |
shoaib_ahmed | 0:791a779d6220 | 2204 | */ |
shoaib_ahmed | 0:791a779d6220 | 2205 | |
shoaib_ahmed | 0:791a779d6220 | 2206 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 2207 | ctr = 0; |
shoaib_ahmed | 0:791a779d6220 | 2208 | for (;;) { |
shoaib_ahmed | 0:791a779d6220 | 2209 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 2210 | |
shoaib_ahmed | 0:791a779d6220 | 2211 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 2212 | |
shoaib_ahmed | 0:791a779d6220 | 2213 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[15]); |
shoaib_ahmed | 0:791a779d6220 | 2214 | tmp1 = GETJSAMPLE(elemptr[1]) + GETJSAMPLE(elemptr[14]); |
shoaib_ahmed | 0:791a779d6220 | 2215 | tmp2 = GETJSAMPLE(elemptr[2]) + GETJSAMPLE(elemptr[13]); |
shoaib_ahmed | 0:791a779d6220 | 2216 | tmp3 = GETJSAMPLE(elemptr[3]) + GETJSAMPLE(elemptr[12]); |
shoaib_ahmed | 0:791a779d6220 | 2217 | tmp4 = GETJSAMPLE(elemptr[4]) + GETJSAMPLE(elemptr[11]); |
shoaib_ahmed | 0:791a779d6220 | 2218 | tmp5 = GETJSAMPLE(elemptr[5]) + GETJSAMPLE(elemptr[10]); |
shoaib_ahmed | 0:791a779d6220 | 2219 | tmp6 = GETJSAMPLE(elemptr[6]) + GETJSAMPLE(elemptr[9]); |
shoaib_ahmed | 0:791a779d6220 | 2220 | tmp7 = GETJSAMPLE(elemptr[7]) + GETJSAMPLE(elemptr[8]); |
shoaib_ahmed | 0:791a779d6220 | 2221 | |
shoaib_ahmed | 0:791a779d6220 | 2222 | tmp10 = tmp0 + tmp7; |
shoaib_ahmed | 0:791a779d6220 | 2223 | tmp14 = tmp0 - tmp7; |
shoaib_ahmed | 0:791a779d6220 | 2224 | tmp11 = tmp1 + tmp6; |
shoaib_ahmed | 0:791a779d6220 | 2225 | tmp15 = tmp1 - tmp6; |
shoaib_ahmed | 0:791a779d6220 | 2226 | tmp12 = tmp2 + tmp5; |
shoaib_ahmed | 0:791a779d6220 | 2227 | tmp16 = tmp2 - tmp5; |
shoaib_ahmed | 0:791a779d6220 | 2228 | tmp13 = tmp3 + tmp4; |
shoaib_ahmed | 0:791a779d6220 | 2229 | tmp17 = tmp3 - tmp4; |
shoaib_ahmed | 0:791a779d6220 | 2230 | |
shoaib_ahmed | 0:791a779d6220 | 2231 | tmp0 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[15]); |
shoaib_ahmed | 0:791a779d6220 | 2232 | tmp1 = GETJSAMPLE(elemptr[1]) - GETJSAMPLE(elemptr[14]); |
shoaib_ahmed | 0:791a779d6220 | 2233 | tmp2 = GETJSAMPLE(elemptr[2]) - GETJSAMPLE(elemptr[13]); |
shoaib_ahmed | 0:791a779d6220 | 2234 | tmp3 = GETJSAMPLE(elemptr[3]) - GETJSAMPLE(elemptr[12]); |
shoaib_ahmed | 0:791a779d6220 | 2235 | tmp4 = GETJSAMPLE(elemptr[4]) - GETJSAMPLE(elemptr[11]); |
shoaib_ahmed | 0:791a779d6220 | 2236 | tmp5 = GETJSAMPLE(elemptr[5]) - GETJSAMPLE(elemptr[10]); |
shoaib_ahmed | 0:791a779d6220 | 2237 | tmp6 = GETJSAMPLE(elemptr[6]) - GETJSAMPLE(elemptr[9]); |
shoaib_ahmed | 0:791a779d6220 | 2238 | tmp7 = GETJSAMPLE(elemptr[7]) - GETJSAMPLE(elemptr[8]); |
shoaib_ahmed | 0:791a779d6220 | 2239 | |
shoaib_ahmed | 0:791a779d6220 | 2240 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 2241 | dataptr[0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2242 | ((tmp10 + tmp11 + tmp12 + tmp13 - 16 * CENTERJSAMPLE) << PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2243 | dataptr[4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2244 | DESCALE(MULTIPLY(tmp10 - tmp13, FIX(1.306562965)) + /* c4[16] = c2[8] */ |
shoaib_ahmed | 0:791a779d6220 | 2245 | MULTIPLY(tmp11 - tmp12, FIX_0_541196100), /* c12[16] = c6[8] */ |
shoaib_ahmed | 0:791a779d6220 | 2246 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2247 | |
shoaib_ahmed | 0:791a779d6220 | 2248 | tmp10 = MULTIPLY(tmp17 - tmp15, FIX(0.275899379)) + /* c14[16] = c7[8] */ |
shoaib_ahmed | 0:791a779d6220 | 2249 | MULTIPLY(tmp14 - tmp16, FIX(1.387039845)); /* c2[16] = c1[8] */ |
shoaib_ahmed | 0:791a779d6220 | 2250 | |
shoaib_ahmed | 0:791a779d6220 | 2251 | dataptr[2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2252 | DESCALE(tmp10 + MULTIPLY(tmp15, FIX(1.451774982)) /* c6+c14 */ |
shoaib_ahmed | 0:791a779d6220 | 2253 | + MULTIPLY(tmp16, FIX(2.172734804)), /* c2+c10 */ |
shoaib_ahmed | 0:791a779d6220 | 2254 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2255 | dataptr[6] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2256 | DESCALE(tmp10 - MULTIPLY(tmp14, FIX(0.211164243)) /* c2-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 2257 | - MULTIPLY(tmp17, FIX(1.061594338)), /* c10+c14 */ |
shoaib_ahmed | 0:791a779d6220 | 2258 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2259 | |
shoaib_ahmed | 0:791a779d6220 | 2260 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 2261 | |
shoaib_ahmed | 0:791a779d6220 | 2262 | tmp11 = MULTIPLY(tmp0 + tmp1, FIX(1.353318001)) + /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 2263 | MULTIPLY(tmp6 - tmp7, FIX(0.410524528)); /* c13 */ |
shoaib_ahmed | 0:791a779d6220 | 2264 | tmp12 = MULTIPLY(tmp0 + tmp2, FIX(1.247225013)) + /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 2265 | MULTIPLY(tmp5 + tmp7, FIX(0.666655658)); /* c11 */ |
shoaib_ahmed | 0:791a779d6220 | 2266 | tmp13 = MULTIPLY(tmp0 + tmp3, FIX(1.093201867)) + /* c7 */ |
shoaib_ahmed | 0:791a779d6220 | 2267 | MULTIPLY(tmp4 - tmp7, FIX(0.897167586)); /* c9 */ |
shoaib_ahmed | 0:791a779d6220 | 2268 | tmp14 = MULTIPLY(tmp1 + tmp2, FIX(0.138617169)) + /* c15 */ |
shoaib_ahmed | 0:791a779d6220 | 2269 | MULTIPLY(tmp6 - tmp5, FIX(1.407403738)); /* c1 */ |
shoaib_ahmed | 0:791a779d6220 | 2270 | tmp15 = MULTIPLY(tmp1 + tmp3, - FIX(0.666655658)) + /* -c11 */ |
shoaib_ahmed | 0:791a779d6220 | 2271 | MULTIPLY(tmp4 + tmp6, - FIX(1.247225013)); /* -c5 */ |
shoaib_ahmed | 0:791a779d6220 | 2272 | tmp16 = MULTIPLY(tmp2 + tmp3, - FIX(1.353318001)) + /* -c3 */ |
shoaib_ahmed | 0:791a779d6220 | 2273 | MULTIPLY(tmp5 - tmp4, FIX(0.410524528)); /* c13 */ |
shoaib_ahmed | 0:791a779d6220 | 2274 | tmp10 = tmp11 + tmp12 + tmp13 - |
shoaib_ahmed | 0:791a779d6220 | 2275 | MULTIPLY(tmp0, FIX(2.286341144)) + /* c7+c5+c3-c1 */ |
shoaib_ahmed | 0:791a779d6220 | 2276 | MULTIPLY(tmp7, FIX(0.779653625)); /* c15+c13-c11+c9 */ |
shoaib_ahmed | 0:791a779d6220 | 2277 | tmp11 += tmp14 + tmp15 + MULTIPLY(tmp1, FIX(0.071888074)) /* c9-c3-c15+c11 */ |
shoaib_ahmed | 0:791a779d6220 | 2278 | - MULTIPLY(tmp6, FIX(1.663905119)); /* c7+c13+c1-c5 */ |
shoaib_ahmed | 0:791a779d6220 | 2279 | tmp12 += tmp14 + tmp16 - MULTIPLY(tmp2, FIX(1.125726048)) /* c7+c5+c15-c3 */ |
shoaib_ahmed | 0:791a779d6220 | 2280 | + MULTIPLY(tmp5, FIX(1.227391138)); /* c9-c11+c1-c13 */ |
shoaib_ahmed | 0:791a779d6220 | 2281 | tmp13 += tmp15 + tmp16 + MULTIPLY(tmp3, FIX(1.065388962)) /* c15+c3+c11-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 2282 | + MULTIPLY(tmp4, FIX(2.167985692)); /* c1+c13+c5-c9 */ |
shoaib_ahmed | 0:791a779d6220 | 2283 | |
shoaib_ahmed | 0:791a779d6220 | 2284 | dataptr[1] = (DCTELEM) DESCALE(tmp10, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2285 | dataptr[3] = (DCTELEM) DESCALE(tmp11, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2286 | dataptr[5] = (DCTELEM) DESCALE(tmp12, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2287 | dataptr[7] = (DCTELEM) DESCALE(tmp13, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2288 | |
shoaib_ahmed | 0:791a779d6220 | 2289 | ctr++; |
shoaib_ahmed | 0:791a779d6220 | 2290 | |
shoaib_ahmed | 0:791a779d6220 | 2291 | if (ctr != DCTSIZE) { |
shoaib_ahmed | 0:791a779d6220 | 2292 | if (ctr == DCTSIZE * 2) |
shoaib_ahmed | 0:791a779d6220 | 2293 | break; /* Done. */ |
shoaib_ahmed | 0:791a779d6220 | 2294 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 2295 | } else |
shoaib_ahmed | 0:791a779d6220 | 2296 | dataptr = workspace; /* switch pointer to extended workspace */ |
shoaib_ahmed | 0:791a779d6220 | 2297 | } |
shoaib_ahmed | 0:791a779d6220 | 2298 | |
shoaib_ahmed | 0:791a779d6220 | 2299 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 2300 | * We remove the PASS1_BITS scaling, but leave the results scaled up |
shoaib_ahmed | 0:791a779d6220 | 2301 | * by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 2302 | * We must also scale the output by (8/16)**2 = 1/2**2. |
shoaib_ahmed | 0:791a779d6220 | 2303 | * cK represents sqrt(2) * cos(K*pi/32). |
shoaib_ahmed | 0:791a779d6220 | 2304 | */ |
shoaib_ahmed | 0:791a779d6220 | 2305 | |
shoaib_ahmed | 0:791a779d6220 | 2306 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 2307 | wsptr = workspace; |
shoaib_ahmed | 0:791a779d6220 | 2308 | for (ctr = DCTSIZE-1; ctr >= 0; ctr--) { |
shoaib_ahmed | 0:791a779d6220 | 2309 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 2310 | |
shoaib_ahmed | 0:791a779d6220 | 2311 | tmp0 = dataptr[DCTSIZE*0] + wsptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 2312 | tmp1 = dataptr[DCTSIZE*1] + wsptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 2313 | tmp2 = dataptr[DCTSIZE*2] + wsptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 2314 | tmp3 = dataptr[DCTSIZE*3] + wsptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 2315 | tmp4 = dataptr[DCTSIZE*4] + wsptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 2316 | tmp5 = dataptr[DCTSIZE*5] + wsptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 2317 | tmp6 = dataptr[DCTSIZE*6] + wsptr[DCTSIZE*1]; |
shoaib_ahmed | 0:791a779d6220 | 2318 | tmp7 = dataptr[DCTSIZE*7] + wsptr[DCTSIZE*0]; |
shoaib_ahmed | 0:791a779d6220 | 2319 | |
shoaib_ahmed | 0:791a779d6220 | 2320 | tmp10 = tmp0 + tmp7; |
shoaib_ahmed | 0:791a779d6220 | 2321 | tmp14 = tmp0 - tmp7; |
shoaib_ahmed | 0:791a779d6220 | 2322 | tmp11 = tmp1 + tmp6; |
shoaib_ahmed | 0:791a779d6220 | 2323 | tmp15 = tmp1 - tmp6; |
shoaib_ahmed | 0:791a779d6220 | 2324 | tmp12 = tmp2 + tmp5; |
shoaib_ahmed | 0:791a779d6220 | 2325 | tmp16 = tmp2 - tmp5; |
shoaib_ahmed | 0:791a779d6220 | 2326 | tmp13 = tmp3 + tmp4; |
shoaib_ahmed | 0:791a779d6220 | 2327 | tmp17 = tmp3 - tmp4; |
shoaib_ahmed | 0:791a779d6220 | 2328 | |
shoaib_ahmed | 0:791a779d6220 | 2329 | tmp0 = dataptr[DCTSIZE*0] - wsptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 2330 | tmp1 = dataptr[DCTSIZE*1] - wsptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 2331 | tmp2 = dataptr[DCTSIZE*2] - wsptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 2332 | tmp3 = dataptr[DCTSIZE*3] - wsptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 2333 | tmp4 = dataptr[DCTSIZE*4] - wsptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 2334 | tmp5 = dataptr[DCTSIZE*5] - wsptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 2335 | tmp6 = dataptr[DCTSIZE*6] - wsptr[DCTSIZE*1]; |
shoaib_ahmed | 0:791a779d6220 | 2336 | tmp7 = dataptr[DCTSIZE*7] - wsptr[DCTSIZE*0]; |
shoaib_ahmed | 0:791a779d6220 | 2337 | |
shoaib_ahmed | 0:791a779d6220 | 2338 | dataptr[DCTSIZE*0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2339 | DESCALE(tmp10 + tmp11 + tmp12 + tmp13, PASS1_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 2340 | dataptr[DCTSIZE*4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2341 | DESCALE(MULTIPLY(tmp10 - tmp13, FIX(1.306562965)) + /* c4[16] = c2[8] */ |
shoaib_ahmed | 0:791a779d6220 | 2342 | MULTIPLY(tmp11 - tmp12, FIX_0_541196100), /* c12[16] = c6[8] */ |
shoaib_ahmed | 0:791a779d6220 | 2343 | CONST_BITS+PASS1_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 2344 | |
shoaib_ahmed | 0:791a779d6220 | 2345 | tmp10 = MULTIPLY(tmp17 - tmp15, FIX(0.275899379)) + /* c14[16] = c7[8] */ |
shoaib_ahmed | 0:791a779d6220 | 2346 | MULTIPLY(tmp14 - tmp16, FIX(1.387039845)); /* c2[16] = c1[8] */ |
shoaib_ahmed | 0:791a779d6220 | 2347 | |
shoaib_ahmed | 0:791a779d6220 | 2348 | dataptr[DCTSIZE*2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2349 | DESCALE(tmp10 + MULTIPLY(tmp15, FIX(1.451774982)) /* c6+c14 */ |
shoaib_ahmed | 0:791a779d6220 | 2350 | + MULTIPLY(tmp16, FIX(2.172734804)), /* c2+10 */ |
shoaib_ahmed | 0:791a779d6220 | 2351 | CONST_BITS+PASS1_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 2352 | dataptr[DCTSIZE*6] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2353 | DESCALE(tmp10 - MULTIPLY(tmp14, FIX(0.211164243)) /* c2-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 2354 | - MULTIPLY(tmp17, FIX(1.061594338)), /* c10+c14 */ |
shoaib_ahmed | 0:791a779d6220 | 2355 | CONST_BITS+PASS1_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 2356 | |
shoaib_ahmed | 0:791a779d6220 | 2357 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 2358 | |
shoaib_ahmed | 0:791a779d6220 | 2359 | tmp11 = MULTIPLY(tmp0 + tmp1, FIX(1.353318001)) + /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 2360 | MULTIPLY(tmp6 - tmp7, FIX(0.410524528)); /* c13 */ |
shoaib_ahmed | 0:791a779d6220 | 2361 | tmp12 = MULTIPLY(tmp0 + tmp2, FIX(1.247225013)) + /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 2362 | MULTIPLY(tmp5 + tmp7, FIX(0.666655658)); /* c11 */ |
shoaib_ahmed | 0:791a779d6220 | 2363 | tmp13 = MULTIPLY(tmp0 + tmp3, FIX(1.093201867)) + /* c7 */ |
shoaib_ahmed | 0:791a779d6220 | 2364 | MULTIPLY(tmp4 - tmp7, FIX(0.897167586)); /* c9 */ |
shoaib_ahmed | 0:791a779d6220 | 2365 | tmp14 = MULTIPLY(tmp1 + tmp2, FIX(0.138617169)) + /* c15 */ |
shoaib_ahmed | 0:791a779d6220 | 2366 | MULTIPLY(tmp6 - tmp5, FIX(1.407403738)); /* c1 */ |
shoaib_ahmed | 0:791a779d6220 | 2367 | tmp15 = MULTIPLY(tmp1 + tmp3, - FIX(0.666655658)) + /* -c11 */ |
shoaib_ahmed | 0:791a779d6220 | 2368 | MULTIPLY(tmp4 + tmp6, - FIX(1.247225013)); /* -c5 */ |
shoaib_ahmed | 0:791a779d6220 | 2369 | tmp16 = MULTIPLY(tmp2 + tmp3, - FIX(1.353318001)) + /* -c3 */ |
shoaib_ahmed | 0:791a779d6220 | 2370 | MULTIPLY(tmp5 - tmp4, FIX(0.410524528)); /* c13 */ |
shoaib_ahmed | 0:791a779d6220 | 2371 | tmp10 = tmp11 + tmp12 + tmp13 - |
shoaib_ahmed | 0:791a779d6220 | 2372 | MULTIPLY(tmp0, FIX(2.286341144)) + /* c7+c5+c3-c1 */ |
shoaib_ahmed | 0:791a779d6220 | 2373 | MULTIPLY(tmp7, FIX(0.779653625)); /* c15+c13-c11+c9 */ |
shoaib_ahmed | 0:791a779d6220 | 2374 | tmp11 += tmp14 + tmp15 + MULTIPLY(tmp1, FIX(0.071888074)) /* c9-c3-c15+c11 */ |
shoaib_ahmed | 0:791a779d6220 | 2375 | - MULTIPLY(tmp6, FIX(1.663905119)); /* c7+c13+c1-c5 */ |
shoaib_ahmed | 0:791a779d6220 | 2376 | tmp12 += tmp14 + tmp16 - MULTIPLY(tmp2, FIX(1.125726048)) /* c7+c5+c15-c3 */ |
shoaib_ahmed | 0:791a779d6220 | 2377 | + MULTIPLY(tmp5, FIX(1.227391138)); /* c9-c11+c1-c13 */ |
shoaib_ahmed | 0:791a779d6220 | 2378 | tmp13 += tmp15 + tmp16 + MULTIPLY(tmp3, FIX(1.065388962)) /* c15+c3+c11-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 2379 | + MULTIPLY(tmp4, FIX(2.167985692)); /* c1+c13+c5-c9 */ |
shoaib_ahmed | 0:791a779d6220 | 2380 | |
shoaib_ahmed | 0:791a779d6220 | 2381 | dataptr[DCTSIZE*1] = (DCTELEM) DESCALE(tmp10, CONST_BITS+PASS1_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 2382 | dataptr[DCTSIZE*3] = (DCTELEM) DESCALE(tmp11, CONST_BITS+PASS1_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 2383 | dataptr[DCTSIZE*5] = (DCTELEM) DESCALE(tmp12, CONST_BITS+PASS1_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 2384 | dataptr[DCTSIZE*7] = (DCTELEM) DESCALE(tmp13, CONST_BITS+PASS1_BITS+2); |
shoaib_ahmed | 0:791a779d6220 | 2385 | |
shoaib_ahmed | 0:791a779d6220 | 2386 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 2387 | wsptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 2388 | } |
shoaib_ahmed | 0:791a779d6220 | 2389 | } |
shoaib_ahmed | 0:791a779d6220 | 2390 | |
shoaib_ahmed | 0:791a779d6220 | 2391 | |
shoaib_ahmed | 0:791a779d6220 | 2392 | /* |
shoaib_ahmed | 0:791a779d6220 | 2393 | * Perform the forward DCT on a 16x8 sample block. |
shoaib_ahmed | 0:791a779d6220 | 2394 | * |
shoaib_ahmed | 0:791a779d6220 | 2395 | * 16-point FDCT in pass 1 (rows), 8-point in pass 2 (columns). |
shoaib_ahmed | 0:791a779d6220 | 2396 | */ |
shoaib_ahmed | 0:791a779d6220 | 2397 | |
shoaib_ahmed | 0:791a779d6220 | 2398 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 2399 | jpeg_fdct_16x8 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 2400 | { |
shoaib_ahmed | 0:791a779d6220 | 2401 | INT32 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7; |
shoaib_ahmed | 0:791a779d6220 | 2402 | INT32 tmp10, tmp11, tmp12, tmp13, tmp14, tmp15, tmp16, tmp17; |
shoaib_ahmed | 0:791a779d6220 | 2403 | INT32 z1; |
shoaib_ahmed | 0:791a779d6220 | 2404 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 2405 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 2406 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 2407 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 2408 | |
shoaib_ahmed | 0:791a779d6220 | 2409 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 2410 | * Note results are scaled up by sqrt(8) compared to a true DCT; |
shoaib_ahmed | 0:791a779d6220 | 2411 | * furthermore, we scale the results by 2**PASS1_BITS. |
shoaib_ahmed | 0:791a779d6220 | 2412 | * 16-point FDCT kernel, cK represents sqrt(2) * cos(K*pi/32). |
shoaib_ahmed | 0:791a779d6220 | 2413 | */ |
shoaib_ahmed | 0:791a779d6220 | 2414 | |
shoaib_ahmed | 0:791a779d6220 | 2415 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 2416 | ctr = 0; |
shoaib_ahmed | 0:791a779d6220 | 2417 | for (ctr = 0; ctr < DCTSIZE; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 2418 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 2419 | |
shoaib_ahmed | 0:791a779d6220 | 2420 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 2421 | |
shoaib_ahmed | 0:791a779d6220 | 2422 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[15]); |
shoaib_ahmed | 0:791a779d6220 | 2423 | tmp1 = GETJSAMPLE(elemptr[1]) + GETJSAMPLE(elemptr[14]); |
shoaib_ahmed | 0:791a779d6220 | 2424 | tmp2 = GETJSAMPLE(elemptr[2]) + GETJSAMPLE(elemptr[13]); |
shoaib_ahmed | 0:791a779d6220 | 2425 | tmp3 = GETJSAMPLE(elemptr[3]) + GETJSAMPLE(elemptr[12]); |
shoaib_ahmed | 0:791a779d6220 | 2426 | tmp4 = GETJSAMPLE(elemptr[4]) + GETJSAMPLE(elemptr[11]); |
shoaib_ahmed | 0:791a779d6220 | 2427 | tmp5 = GETJSAMPLE(elemptr[5]) + GETJSAMPLE(elemptr[10]); |
shoaib_ahmed | 0:791a779d6220 | 2428 | tmp6 = GETJSAMPLE(elemptr[6]) + GETJSAMPLE(elemptr[9]); |
shoaib_ahmed | 0:791a779d6220 | 2429 | tmp7 = GETJSAMPLE(elemptr[7]) + GETJSAMPLE(elemptr[8]); |
shoaib_ahmed | 0:791a779d6220 | 2430 | |
shoaib_ahmed | 0:791a779d6220 | 2431 | tmp10 = tmp0 + tmp7; |
shoaib_ahmed | 0:791a779d6220 | 2432 | tmp14 = tmp0 - tmp7; |
shoaib_ahmed | 0:791a779d6220 | 2433 | tmp11 = tmp1 + tmp6; |
shoaib_ahmed | 0:791a779d6220 | 2434 | tmp15 = tmp1 - tmp6; |
shoaib_ahmed | 0:791a779d6220 | 2435 | tmp12 = tmp2 + tmp5; |
shoaib_ahmed | 0:791a779d6220 | 2436 | tmp16 = tmp2 - tmp5; |
shoaib_ahmed | 0:791a779d6220 | 2437 | tmp13 = tmp3 + tmp4; |
shoaib_ahmed | 0:791a779d6220 | 2438 | tmp17 = tmp3 - tmp4; |
shoaib_ahmed | 0:791a779d6220 | 2439 | |
shoaib_ahmed | 0:791a779d6220 | 2440 | tmp0 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[15]); |
shoaib_ahmed | 0:791a779d6220 | 2441 | tmp1 = GETJSAMPLE(elemptr[1]) - GETJSAMPLE(elemptr[14]); |
shoaib_ahmed | 0:791a779d6220 | 2442 | tmp2 = GETJSAMPLE(elemptr[2]) - GETJSAMPLE(elemptr[13]); |
shoaib_ahmed | 0:791a779d6220 | 2443 | tmp3 = GETJSAMPLE(elemptr[3]) - GETJSAMPLE(elemptr[12]); |
shoaib_ahmed | 0:791a779d6220 | 2444 | tmp4 = GETJSAMPLE(elemptr[4]) - GETJSAMPLE(elemptr[11]); |
shoaib_ahmed | 0:791a779d6220 | 2445 | tmp5 = GETJSAMPLE(elemptr[5]) - GETJSAMPLE(elemptr[10]); |
shoaib_ahmed | 0:791a779d6220 | 2446 | tmp6 = GETJSAMPLE(elemptr[6]) - GETJSAMPLE(elemptr[9]); |
shoaib_ahmed | 0:791a779d6220 | 2447 | tmp7 = GETJSAMPLE(elemptr[7]) - GETJSAMPLE(elemptr[8]); |
shoaib_ahmed | 0:791a779d6220 | 2448 | |
shoaib_ahmed | 0:791a779d6220 | 2449 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 2450 | dataptr[0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2451 | ((tmp10 + tmp11 + tmp12 + tmp13 - 16 * CENTERJSAMPLE) << PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2452 | dataptr[4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2453 | DESCALE(MULTIPLY(tmp10 - tmp13, FIX(1.306562965)) + /* c4[16] = c2[8] */ |
shoaib_ahmed | 0:791a779d6220 | 2454 | MULTIPLY(tmp11 - tmp12, FIX_0_541196100), /* c12[16] = c6[8] */ |
shoaib_ahmed | 0:791a779d6220 | 2455 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2456 | |
shoaib_ahmed | 0:791a779d6220 | 2457 | tmp10 = MULTIPLY(tmp17 - tmp15, FIX(0.275899379)) + /* c14[16] = c7[8] */ |
shoaib_ahmed | 0:791a779d6220 | 2458 | MULTIPLY(tmp14 - tmp16, FIX(1.387039845)); /* c2[16] = c1[8] */ |
shoaib_ahmed | 0:791a779d6220 | 2459 | |
shoaib_ahmed | 0:791a779d6220 | 2460 | dataptr[2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2461 | DESCALE(tmp10 + MULTIPLY(tmp15, FIX(1.451774982)) /* c6+c14 */ |
shoaib_ahmed | 0:791a779d6220 | 2462 | + MULTIPLY(tmp16, FIX(2.172734804)), /* c2+c10 */ |
shoaib_ahmed | 0:791a779d6220 | 2463 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2464 | dataptr[6] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2465 | DESCALE(tmp10 - MULTIPLY(tmp14, FIX(0.211164243)) /* c2-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 2466 | - MULTIPLY(tmp17, FIX(1.061594338)), /* c10+c14 */ |
shoaib_ahmed | 0:791a779d6220 | 2467 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2468 | |
shoaib_ahmed | 0:791a779d6220 | 2469 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 2470 | |
shoaib_ahmed | 0:791a779d6220 | 2471 | tmp11 = MULTIPLY(tmp0 + tmp1, FIX(1.353318001)) + /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 2472 | MULTIPLY(tmp6 - tmp7, FIX(0.410524528)); /* c13 */ |
shoaib_ahmed | 0:791a779d6220 | 2473 | tmp12 = MULTIPLY(tmp0 + tmp2, FIX(1.247225013)) + /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 2474 | MULTIPLY(tmp5 + tmp7, FIX(0.666655658)); /* c11 */ |
shoaib_ahmed | 0:791a779d6220 | 2475 | tmp13 = MULTIPLY(tmp0 + tmp3, FIX(1.093201867)) + /* c7 */ |
shoaib_ahmed | 0:791a779d6220 | 2476 | MULTIPLY(tmp4 - tmp7, FIX(0.897167586)); /* c9 */ |
shoaib_ahmed | 0:791a779d6220 | 2477 | tmp14 = MULTIPLY(tmp1 + tmp2, FIX(0.138617169)) + /* c15 */ |
shoaib_ahmed | 0:791a779d6220 | 2478 | MULTIPLY(tmp6 - tmp5, FIX(1.407403738)); /* c1 */ |
shoaib_ahmed | 0:791a779d6220 | 2479 | tmp15 = MULTIPLY(tmp1 + tmp3, - FIX(0.666655658)) + /* -c11 */ |
shoaib_ahmed | 0:791a779d6220 | 2480 | MULTIPLY(tmp4 + tmp6, - FIX(1.247225013)); /* -c5 */ |
shoaib_ahmed | 0:791a779d6220 | 2481 | tmp16 = MULTIPLY(tmp2 + tmp3, - FIX(1.353318001)) + /* -c3 */ |
shoaib_ahmed | 0:791a779d6220 | 2482 | MULTIPLY(tmp5 - tmp4, FIX(0.410524528)); /* c13 */ |
shoaib_ahmed | 0:791a779d6220 | 2483 | tmp10 = tmp11 + tmp12 + tmp13 - |
shoaib_ahmed | 0:791a779d6220 | 2484 | MULTIPLY(tmp0, FIX(2.286341144)) + /* c7+c5+c3-c1 */ |
shoaib_ahmed | 0:791a779d6220 | 2485 | MULTIPLY(tmp7, FIX(0.779653625)); /* c15+c13-c11+c9 */ |
shoaib_ahmed | 0:791a779d6220 | 2486 | tmp11 += tmp14 + tmp15 + MULTIPLY(tmp1, FIX(0.071888074)) /* c9-c3-c15+c11 */ |
shoaib_ahmed | 0:791a779d6220 | 2487 | - MULTIPLY(tmp6, FIX(1.663905119)); /* c7+c13+c1-c5 */ |
shoaib_ahmed | 0:791a779d6220 | 2488 | tmp12 += tmp14 + tmp16 - MULTIPLY(tmp2, FIX(1.125726048)) /* c7+c5+c15-c3 */ |
shoaib_ahmed | 0:791a779d6220 | 2489 | + MULTIPLY(tmp5, FIX(1.227391138)); /* c9-c11+c1-c13 */ |
shoaib_ahmed | 0:791a779d6220 | 2490 | tmp13 += tmp15 + tmp16 + MULTIPLY(tmp3, FIX(1.065388962)) /* c15+c3+c11-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 2491 | + MULTIPLY(tmp4, FIX(2.167985692)); /* c1+c13+c5-c9 */ |
shoaib_ahmed | 0:791a779d6220 | 2492 | |
shoaib_ahmed | 0:791a779d6220 | 2493 | dataptr[1] = (DCTELEM) DESCALE(tmp10, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2494 | dataptr[3] = (DCTELEM) DESCALE(tmp11, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2495 | dataptr[5] = (DCTELEM) DESCALE(tmp12, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2496 | dataptr[7] = (DCTELEM) DESCALE(tmp13, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2497 | |
shoaib_ahmed | 0:791a779d6220 | 2498 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 2499 | } |
shoaib_ahmed | 0:791a779d6220 | 2500 | |
shoaib_ahmed | 0:791a779d6220 | 2501 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 2502 | * We remove the PASS1_BITS scaling, but leave the results scaled up |
shoaib_ahmed | 0:791a779d6220 | 2503 | * by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 2504 | * We must also scale the output by 8/16 = 1/2. |
shoaib_ahmed | 0:791a779d6220 | 2505 | * 8-point FDCT kernel, cK represents sqrt(2) * cos(K*pi/16). |
shoaib_ahmed | 0:791a779d6220 | 2506 | */ |
shoaib_ahmed | 0:791a779d6220 | 2507 | |
shoaib_ahmed | 0:791a779d6220 | 2508 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 2509 | for (ctr = DCTSIZE-1; ctr >= 0; ctr--) { |
shoaib_ahmed | 0:791a779d6220 | 2510 | /* Even part per LL&M figure 1 --- note that published figure is faulty; |
shoaib_ahmed | 0:791a779d6220 | 2511 | * rotator "c1" should be "c6". |
shoaib_ahmed | 0:791a779d6220 | 2512 | */ |
shoaib_ahmed | 0:791a779d6220 | 2513 | |
shoaib_ahmed | 0:791a779d6220 | 2514 | tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 2515 | tmp1 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 2516 | tmp2 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 2517 | tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 2518 | |
shoaib_ahmed | 0:791a779d6220 | 2519 | tmp10 = tmp0 + tmp3; |
shoaib_ahmed | 0:791a779d6220 | 2520 | tmp12 = tmp0 - tmp3; |
shoaib_ahmed | 0:791a779d6220 | 2521 | tmp11 = tmp1 + tmp2; |
shoaib_ahmed | 0:791a779d6220 | 2522 | tmp13 = tmp1 - tmp2; |
shoaib_ahmed | 0:791a779d6220 | 2523 | |
shoaib_ahmed | 0:791a779d6220 | 2524 | tmp0 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 2525 | tmp1 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 2526 | tmp2 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 2527 | tmp3 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 2528 | |
shoaib_ahmed | 0:791a779d6220 | 2529 | dataptr[DCTSIZE*0] = (DCTELEM) DESCALE(tmp10 + tmp11, PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 2530 | dataptr[DCTSIZE*4] = (DCTELEM) DESCALE(tmp10 - tmp11, PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 2531 | |
shoaib_ahmed | 0:791a779d6220 | 2532 | z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 2533 | dataptr[DCTSIZE*2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2534 | DESCALE(z1 + MULTIPLY(tmp12, FIX_0_765366865), /* c2-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 2535 | CONST_BITS+PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 2536 | dataptr[DCTSIZE*6] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2537 | DESCALE(z1 - MULTIPLY(tmp13, FIX_1_847759065), /* c2+c6 */ |
shoaib_ahmed | 0:791a779d6220 | 2538 | CONST_BITS+PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 2539 | |
shoaib_ahmed | 0:791a779d6220 | 2540 | /* Odd part per figure 8 --- note paper omits factor of sqrt(2). |
shoaib_ahmed | 0:791a779d6220 | 2541 | * i0..i3 in the paper are tmp0..tmp3 here. |
shoaib_ahmed | 0:791a779d6220 | 2542 | */ |
shoaib_ahmed | 0:791a779d6220 | 2543 | |
shoaib_ahmed | 0:791a779d6220 | 2544 | tmp12 = tmp0 + tmp2; |
shoaib_ahmed | 0:791a779d6220 | 2545 | tmp13 = tmp1 + tmp3; |
shoaib_ahmed | 0:791a779d6220 | 2546 | |
shoaib_ahmed | 0:791a779d6220 | 2547 | z1 = MULTIPLY(tmp12 + tmp13, FIX_1_175875602); /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 2548 | tmp12 = MULTIPLY(tmp12, - FIX_0_390180644); /* -c3+c5 */ |
shoaib_ahmed | 0:791a779d6220 | 2549 | tmp13 = MULTIPLY(tmp13, - FIX_1_961570560); /* -c3-c5 */ |
shoaib_ahmed | 0:791a779d6220 | 2550 | tmp12 += z1; |
shoaib_ahmed | 0:791a779d6220 | 2551 | tmp13 += z1; |
shoaib_ahmed | 0:791a779d6220 | 2552 | |
shoaib_ahmed | 0:791a779d6220 | 2553 | z1 = MULTIPLY(tmp0 + tmp3, - FIX_0_899976223); /* -c3+c7 */ |
shoaib_ahmed | 0:791a779d6220 | 2554 | tmp0 = MULTIPLY(tmp0, FIX_1_501321110); /* c1+c3-c5-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 2555 | tmp3 = MULTIPLY(tmp3, FIX_0_298631336); /* -c1+c3+c5-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 2556 | tmp0 += z1 + tmp12; |
shoaib_ahmed | 0:791a779d6220 | 2557 | tmp3 += z1 + tmp13; |
shoaib_ahmed | 0:791a779d6220 | 2558 | |
shoaib_ahmed | 0:791a779d6220 | 2559 | z1 = MULTIPLY(tmp1 + tmp2, - FIX_2_562915447); /* -c1-c3 */ |
shoaib_ahmed | 0:791a779d6220 | 2560 | tmp1 = MULTIPLY(tmp1, FIX_3_072711026); /* c1+c3+c5-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 2561 | tmp2 = MULTIPLY(tmp2, FIX_2_053119869); /* c1+c3-c5+c7 */ |
shoaib_ahmed | 0:791a779d6220 | 2562 | tmp1 += z1 + tmp13; |
shoaib_ahmed | 0:791a779d6220 | 2563 | tmp2 += z1 + tmp12; |
shoaib_ahmed | 0:791a779d6220 | 2564 | |
shoaib_ahmed | 0:791a779d6220 | 2565 | dataptr[DCTSIZE*1] = (DCTELEM) DESCALE(tmp0, CONST_BITS+PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 2566 | dataptr[DCTSIZE*3] = (DCTELEM) DESCALE(tmp1, CONST_BITS+PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 2567 | dataptr[DCTSIZE*5] = (DCTELEM) DESCALE(tmp2, CONST_BITS+PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 2568 | dataptr[DCTSIZE*7] = (DCTELEM) DESCALE(tmp3, CONST_BITS+PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 2569 | |
shoaib_ahmed | 0:791a779d6220 | 2570 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 2571 | } |
shoaib_ahmed | 0:791a779d6220 | 2572 | } |
shoaib_ahmed | 0:791a779d6220 | 2573 | |
shoaib_ahmed | 0:791a779d6220 | 2574 | |
shoaib_ahmed | 0:791a779d6220 | 2575 | /* |
shoaib_ahmed | 0:791a779d6220 | 2576 | * Perform the forward DCT on a 14x7 sample block. |
shoaib_ahmed | 0:791a779d6220 | 2577 | * |
shoaib_ahmed | 0:791a779d6220 | 2578 | * 14-point FDCT in pass 1 (rows), 7-point in pass 2 (columns). |
shoaib_ahmed | 0:791a779d6220 | 2579 | */ |
shoaib_ahmed | 0:791a779d6220 | 2580 | |
shoaib_ahmed | 0:791a779d6220 | 2581 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 2582 | jpeg_fdct_14x7 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 2583 | { |
shoaib_ahmed | 0:791a779d6220 | 2584 | INT32 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6; |
shoaib_ahmed | 0:791a779d6220 | 2585 | INT32 tmp10, tmp11, tmp12, tmp13, tmp14, tmp15, tmp16; |
shoaib_ahmed | 0:791a779d6220 | 2586 | INT32 z1, z2, z3; |
shoaib_ahmed | 0:791a779d6220 | 2587 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 2588 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 2589 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 2590 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 2591 | |
shoaib_ahmed | 0:791a779d6220 | 2592 | /* Zero bottom row of output coefficient block. */ |
shoaib_ahmed | 0:791a779d6220 | 2593 | MEMZERO(&data[DCTSIZE*7], SIZEOF(DCTELEM) * DCTSIZE); |
shoaib_ahmed | 0:791a779d6220 | 2594 | |
shoaib_ahmed | 0:791a779d6220 | 2595 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 2596 | * Note results are scaled up by sqrt(8) compared to a true DCT; |
shoaib_ahmed | 0:791a779d6220 | 2597 | * furthermore, we scale the results by 2**PASS1_BITS. |
shoaib_ahmed | 0:791a779d6220 | 2598 | * 14-point FDCT kernel, cK represents sqrt(2) * cos(K*pi/28). |
shoaib_ahmed | 0:791a779d6220 | 2599 | */ |
shoaib_ahmed | 0:791a779d6220 | 2600 | |
shoaib_ahmed | 0:791a779d6220 | 2601 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 2602 | for (ctr = 0; ctr < 7; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 2603 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 2604 | |
shoaib_ahmed | 0:791a779d6220 | 2605 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 2606 | |
shoaib_ahmed | 0:791a779d6220 | 2607 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[13]); |
shoaib_ahmed | 0:791a779d6220 | 2608 | tmp1 = GETJSAMPLE(elemptr[1]) + GETJSAMPLE(elemptr[12]); |
shoaib_ahmed | 0:791a779d6220 | 2609 | tmp2 = GETJSAMPLE(elemptr[2]) + GETJSAMPLE(elemptr[11]); |
shoaib_ahmed | 0:791a779d6220 | 2610 | tmp13 = GETJSAMPLE(elemptr[3]) + GETJSAMPLE(elemptr[10]); |
shoaib_ahmed | 0:791a779d6220 | 2611 | tmp4 = GETJSAMPLE(elemptr[4]) + GETJSAMPLE(elemptr[9]); |
shoaib_ahmed | 0:791a779d6220 | 2612 | tmp5 = GETJSAMPLE(elemptr[5]) + GETJSAMPLE(elemptr[8]); |
shoaib_ahmed | 0:791a779d6220 | 2613 | tmp6 = GETJSAMPLE(elemptr[6]) + GETJSAMPLE(elemptr[7]); |
shoaib_ahmed | 0:791a779d6220 | 2614 | |
shoaib_ahmed | 0:791a779d6220 | 2615 | tmp10 = tmp0 + tmp6; |
shoaib_ahmed | 0:791a779d6220 | 2616 | tmp14 = tmp0 - tmp6; |
shoaib_ahmed | 0:791a779d6220 | 2617 | tmp11 = tmp1 + tmp5; |
shoaib_ahmed | 0:791a779d6220 | 2618 | tmp15 = tmp1 - tmp5; |
shoaib_ahmed | 0:791a779d6220 | 2619 | tmp12 = tmp2 + tmp4; |
shoaib_ahmed | 0:791a779d6220 | 2620 | tmp16 = tmp2 - tmp4; |
shoaib_ahmed | 0:791a779d6220 | 2621 | |
shoaib_ahmed | 0:791a779d6220 | 2622 | tmp0 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[13]); |
shoaib_ahmed | 0:791a779d6220 | 2623 | tmp1 = GETJSAMPLE(elemptr[1]) - GETJSAMPLE(elemptr[12]); |
shoaib_ahmed | 0:791a779d6220 | 2624 | tmp2 = GETJSAMPLE(elemptr[2]) - GETJSAMPLE(elemptr[11]); |
shoaib_ahmed | 0:791a779d6220 | 2625 | tmp3 = GETJSAMPLE(elemptr[3]) - GETJSAMPLE(elemptr[10]); |
shoaib_ahmed | 0:791a779d6220 | 2626 | tmp4 = GETJSAMPLE(elemptr[4]) - GETJSAMPLE(elemptr[9]); |
shoaib_ahmed | 0:791a779d6220 | 2627 | tmp5 = GETJSAMPLE(elemptr[5]) - GETJSAMPLE(elemptr[8]); |
shoaib_ahmed | 0:791a779d6220 | 2628 | tmp6 = GETJSAMPLE(elemptr[6]) - GETJSAMPLE(elemptr[7]); |
shoaib_ahmed | 0:791a779d6220 | 2629 | |
shoaib_ahmed | 0:791a779d6220 | 2630 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 2631 | dataptr[0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2632 | ((tmp10 + tmp11 + tmp12 + tmp13 - 14 * CENTERJSAMPLE) << PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2633 | tmp13 += tmp13; |
shoaib_ahmed | 0:791a779d6220 | 2634 | dataptr[4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2635 | DESCALE(MULTIPLY(tmp10 - tmp13, FIX(1.274162392)) + /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 2636 | MULTIPLY(tmp11 - tmp13, FIX(0.314692123)) - /* c12 */ |
shoaib_ahmed | 0:791a779d6220 | 2637 | MULTIPLY(tmp12 - tmp13, FIX(0.881747734)), /* c8 */ |
shoaib_ahmed | 0:791a779d6220 | 2638 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2639 | |
shoaib_ahmed | 0:791a779d6220 | 2640 | tmp10 = MULTIPLY(tmp14 + tmp15, FIX(1.105676686)); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 2641 | |
shoaib_ahmed | 0:791a779d6220 | 2642 | dataptr[2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2643 | DESCALE(tmp10 + MULTIPLY(tmp14, FIX(0.273079590)) /* c2-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 2644 | + MULTIPLY(tmp16, FIX(0.613604268)), /* c10 */ |
shoaib_ahmed | 0:791a779d6220 | 2645 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2646 | dataptr[6] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2647 | DESCALE(tmp10 - MULTIPLY(tmp15, FIX(1.719280954)) /* c6+c10 */ |
shoaib_ahmed | 0:791a779d6220 | 2648 | - MULTIPLY(tmp16, FIX(1.378756276)), /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 2649 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2650 | |
shoaib_ahmed | 0:791a779d6220 | 2651 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 2652 | |
shoaib_ahmed | 0:791a779d6220 | 2653 | tmp10 = tmp1 + tmp2; |
shoaib_ahmed | 0:791a779d6220 | 2654 | tmp11 = tmp5 - tmp4; |
shoaib_ahmed | 0:791a779d6220 | 2655 | dataptr[7] = (DCTELEM) ((tmp0 - tmp10 + tmp3 - tmp11 - tmp6) << PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2656 | tmp3 <<= CONST_BITS; |
shoaib_ahmed | 0:791a779d6220 | 2657 | tmp10 = MULTIPLY(tmp10, - FIX(0.158341681)); /* -c13 */ |
shoaib_ahmed | 0:791a779d6220 | 2658 | tmp11 = MULTIPLY(tmp11, FIX(1.405321284)); /* c1 */ |
shoaib_ahmed | 0:791a779d6220 | 2659 | tmp10 += tmp11 - tmp3; |
shoaib_ahmed | 0:791a779d6220 | 2660 | tmp11 = MULTIPLY(tmp0 + tmp2, FIX(1.197448846)) + /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 2661 | MULTIPLY(tmp4 + tmp6, FIX(0.752406978)); /* c9 */ |
shoaib_ahmed | 0:791a779d6220 | 2662 | dataptr[5] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2663 | DESCALE(tmp10 + tmp11 - MULTIPLY(tmp2, FIX(2.373959773)) /* c3+c5-c13 */ |
shoaib_ahmed | 0:791a779d6220 | 2664 | + MULTIPLY(tmp4, FIX(1.119999435)), /* c1+c11-c9 */ |
shoaib_ahmed | 0:791a779d6220 | 2665 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2666 | tmp12 = MULTIPLY(tmp0 + tmp1, FIX(1.334852607)) + /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 2667 | MULTIPLY(tmp5 - tmp6, FIX(0.467085129)); /* c11 */ |
shoaib_ahmed | 0:791a779d6220 | 2668 | dataptr[3] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2669 | DESCALE(tmp10 + tmp12 - MULTIPLY(tmp1, FIX(0.424103948)) /* c3-c9-c13 */ |
shoaib_ahmed | 0:791a779d6220 | 2670 | - MULTIPLY(tmp5, FIX(3.069855259)), /* c1+c5+c11 */ |
shoaib_ahmed | 0:791a779d6220 | 2671 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2672 | dataptr[1] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2673 | DESCALE(tmp11 + tmp12 + tmp3 + tmp6 - |
shoaib_ahmed | 0:791a779d6220 | 2674 | MULTIPLY(tmp0 + tmp6, FIX(1.126980169)), /* c3+c5-c1 */ |
shoaib_ahmed | 0:791a779d6220 | 2675 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2676 | |
shoaib_ahmed | 0:791a779d6220 | 2677 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 2678 | } |
shoaib_ahmed | 0:791a779d6220 | 2679 | |
shoaib_ahmed | 0:791a779d6220 | 2680 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 2681 | * We remove the PASS1_BITS scaling, but leave the results scaled up |
shoaib_ahmed | 0:791a779d6220 | 2682 | * by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 2683 | * We must also scale the output by (8/14)*(8/7) = 32/49, which we |
shoaib_ahmed | 0:791a779d6220 | 2684 | * partially fold into the constant multipliers and final shifting: |
shoaib_ahmed | 0:791a779d6220 | 2685 | * 7-point FDCT kernel, cK represents sqrt(2) * cos(K*pi/14) * 64/49. |
shoaib_ahmed | 0:791a779d6220 | 2686 | */ |
shoaib_ahmed | 0:791a779d6220 | 2687 | |
shoaib_ahmed | 0:791a779d6220 | 2688 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 2689 | for (ctr = DCTSIZE-1; ctr >= 0; ctr--) { |
shoaib_ahmed | 0:791a779d6220 | 2690 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 2691 | |
shoaib_ahmed | 0:791a779d6220 | 2692 | tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 2693 | tmp1 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 2694 | tmp2 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 2695 | tmp3 = dataptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 2696 | |
shoaib_ahmed | 0:791a779d6220 | 2697 | tmp10 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 2698 | tmp11 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 2699 | tmp12 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 2700 | |
shoaib_ahmed | 0:791a779d6220 | 2701 | z1 = tmp0 + tmp2; |
shoaib_ahmed | 0:791a779d6220 | 2702 | dataptr[DCTSIZE*0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2703 | DESCALE(MULTIPLY(z1 + tmp1 + tmp3, FIX(1.306122449)), /* 64/49 */ |
shoaib_ahmed | 0:791a779d6220 | 2704 | CONST_BITS+PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 2705 | tmp3 += tmp3; |
shoaib_ahmed | 0:791a779d6220 | 2706 | z1 -= tmp3; |
shoaib_ahmed | 0:791a779d6220 | 2707 | z1 -= tmp3; |
shoaib_ahmed | 0:791a779d6220 | 2708 | z1 = MULTIPLY(z1, FIX(0.461784020)); /* (c2+c6-c4)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 2709 | z2 = MULTIPLY(tmp0 - tmp2, FIX(1.202428084)); /* (c2+c4-c6)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 2710 | z3 = MULTIPLY(tmp1 - tmp2, FIX(0.411026446)); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 2711 | dataptr[DCTSIZE*2] = (DCTELEM) DESCALE(z1 + z2 + z3, CONST_BITS+PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 2712 | z1 -= z2; |
shoaib_ahmed | 0:791a779d6220 | 2713 | z2 = MULTIPLY(tmp0 - tmp1, FIX(1.151670509)); /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 2714 | dataptr[DCTSIZE*4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2715 | DESCALE(z2 + z3 - MULTIPLY(tmp1 - tmp3, FIX(0.923568041)), /* c2+c6-c4 */ |
shoaib_ahmed | 0:791a779d6220 | 2716 | CONST_BITS+PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 2717 | dataptr[DCTSIZE*6] = (DCTELEM) DESCALE(z1 + z2, CONST_BITS+PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 2718 | |
shoaib_ahmed | 0:791a779d6220 | 2719 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 2720 | |
shoaib_ahmed | 0:791a779d6220 | 2721 | tmp1 = MULTIPLY(tmp10 + tmp11, FIX(1.221765677)); /* (c3+c1-c5)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 2722 | tmp2 = MULTIPLY(tmp10 - tmp11, FIX(0.222383464)); /* (c3+c5-c1)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 2723 | tmp0 = tmp1 - tmp2; |
shoaib_ahmed | 0:791a779d6220 | 2724 | tmp1 += tmp2; |
shoaib_ahmed | 0:791a779d6220 | 2725 | tmp2 = MULTIPLY(tmp11 + tmp12, - FIX(1.800824523)); /* -c1 */ |
shoaib_ahmed | 0:791a779d6220 | 2726 | tmp1 += tmp2; |
shoaib_ahmed | 0:791a779d6220 | 2727 | tmp3 = MULTIPLY(tmp10 + tmp12, FIX(0.801442310)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 2728 | tmp0 += tmp3; |
shoaib_ahmed | 0:791a779d6220 | 2729 | tmp2 += tmp3 + MULTIPLY(tmp12, FIX(2.443531355)); /* c3+c1-c5 */ |
shoaib_ahmed | 0:791a779d6220 | 2730 | |
shoaib_ahmed | 0:791a779d6220 | 2731 | dataptr[DCTSIZE*1] = (DCTELEM) DESCALE(tmp0, CONST_BITS+PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 2732 | dataptr[DCTSIZE*3] = (DCTELEM) DESCALE(tmp1, CONST_BITS+PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 2733 | dataptr[DCTSIZE*5] = (DCTELEM) DESCALE(tmp2, CONST_BITS+PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 2734 | |
shoaib_ahmed | 0:791a779d6220 | 2735 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 2736 | } |
shoaib_ahmed | 0:791a779d6220 | 2737 | } |
shoaib_ahmed | 0:791a779d6220 | 2738 | |
shoaib_ahmed | 0:791a779d6220 | 2739 | |
shoaib_ahmed | 0:791a779d6220 | 2740 | /* |
shoaib_ahmed | 0:791a779d6220 | 2741 | * Perform the forward DCT on a 12x6 sample block. |
shoaib_ahmed | 0:791a779d6220 | 2742 | * |
shoaib_ahmed | 0:791a779d6220 | 2743 | * 12-point FDCT in pass 1 (rows), 6-point in pass 2 (columns). |
shoaib_ahmed | 0:791a779d6220 | 2744 | */ |
shoaib_ahmed | 0:791a779d6220 | 2745 | |
shoaib_ahmed | 0:791a779d6220 | 2746 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 2747 | jpeg_fdct_12x6 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 2748 | { |
shoaib_ahmed | 0:791a779d6220 | 2749 | INT32 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5; |
shoaib_ahmed | 0:791a779d6220 | 2750 | INT32 tmp10, tmp11, tmp12, tmp13, tmp14, tmp15; |
shoaib_ahmed | 0:791a779d6220 | 2751 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 2752 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 2753 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 2754 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 2755 | |
shoaib_ahmed | 0:791a779d6220 | 2756 | /* Zero 2 bottom rows of output coefficient block. */ |
shoaib_ahmed | 0:791a779d6220 | 2757 | MEMZERO(&data[DCTSIZE*6], SIZEOF(DCTELEM) * DCTSIZE * 2); |
shoaib_ahmed | 0:791a779d6220 | 2758 | |
shoaib_ahmed | 0:791a779d6220 | 2759 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 2760 | * Note results are scaled up by sqrt(8) compared to a true DCT; |
shoaib_ahmed | 0:791a779d6220 | 2761 | * furthermore, we scale the results by 2**PASS1_BITS. |
shoaib_ahmed | 0:791a779d6220 | 2762 | * 12-point FDCT kernel, cK represents sqrt(2) * cos(K*pi/24). |
shoaib_ahmed | 0:791a779d6220 | 2763 | */ |
shoaib_ahmed | 0:791a779d6220 | 2764 | |
shoaib_ahmed | 0:791a779d6220 | 2765 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 2766 | for (ctr = 0; ctr < 6; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 2767 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 2768 | |
shoaib_ahmed | 0:791a779d6220 | 2769 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 2770 | |
shoaib_ahmed | 0:791a779d6220 | 2771 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[11]); |
shoaib_ahmed | 0:791a779d6220 | 2772 | tmp1 = GETJSAMPLE(elemptr[1]) + GETJSAMPLE(elemptr[10]); |
shoaib_ahmed | 0:791a779d6220 | 2773 | tmp2 = GETJSAMPLE(elemptr[2]) + GETJSAMPLE(elemptr[9]); |
shoaib_ahmed | 0:791a779d6220 | 2774 | tmp3 = GETJSAMPLE(elemptr[3]) + GETJSAMPLE(elemptr[8]); |
shoaib_ahmed | 0:791a779d6220 | 2775 | tmp4 = GETJSAMPLE(elemptr[4]) + GETJSAMPLE(elemptr[7]); |
shoaib_ahmed | 0:791a779d6220 | 2776 | tmp5 = GETJSAMPLE(elemptr[5]) + GETJSAMPLE(elemptr[6]); |
shoaib_ahmed | 0:791a779d6220 | 2777 | |
shoaib_ahmed | 0:791a779d6220 | 2778 | tmp10 = tmp0 + tmp5; |
shoaib_ahmed | 0:791a779d6220 | 2779 | tmp13 = tmp0 - tmp5; |
shoaib_ahmed | 0:791a779d6220 | 2780 | tmp11 = tmp1 + tmp4; |
shoaib_ahmed | 0:791a779d6220 | 2781 | tmp14 = tmp1 - tmp4; |
shoaib_ahmed | 0:791a779d6220 | 2782 | tmp12 = tmp2 + tmp3; |
shoaib_ahmed | 0:791a779d6220 | 2783 | tmp15 = tmp2 - tmp3; |
shoaib_ahmed | 0:791a779d6220 | 2784 | |
shoaib_ahmed | 0:791a779d6220 | 2785 | tmp0 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[11]); |
shoaib_ahmed | 0:791a779d6220 | 2786 | tmp1 = GETJSAMPLE(elemptr[1]) - GETJSAMPLE(elemptr[10]); |
shoaib_ahmed | 0:791a779d6220 | 2787 | tmp2 = GETJSAMPLE(elemptr[2]) - GETJSAMPLE(elemptr[9]); |
shoaib_ahmed | 0:791a779d6220 | 2788 | tmp3 = GETJSAMPLE(elemptr[3]) - GETJSAMPLE(elemptr[8]); |
shoaib_ahmed | 0:791a779d6220 | 2789 | tmp4 = GETJSAMPLE(elemptr[4]) - GETJSAMPLE(elemptr[7]); |
shoaib_ahmed | 0:791a779d6220 | 2790 | tmp5 = GETJSAMPLE(elemptr[5]) - GETJSAMPLE(elemptr[6]); |
shoaib_ahmed | 0:791a779d6220 | 2791 | |
shoaib_ahmed | 0:791a779d6220 | 2792 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 2793 | dataptr[0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2794 | ((tmp10 + tmp11 + tmp12 - 12 * CENTERJSAMPLE) << PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2795 | dataptr[6] = (DCTELEM) ((tmp13 - tmp14 - tmp15) << PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2796 | dataptr[4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2797 | DESCALE(MULTIPLY(tmp10 - tmp12, FIX(1.224744871)), /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 2798 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2799 | dataptr[2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2800 | DESCALE(tmp14 - tmp15 + MULTIPLY(tmp13 + tmp15, FIX(1.366025404)), /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 2801 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2802 | |
shoaib_ahmed | 0:791a779d6220 | 2803 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 2804 | |
shoaib_ahmed | 0:791a779d6220 | 2805 | tmp10 = MULTIPLY(tmp1 + tmp4, FIX_0_541196100); /* c9 */ |
shoaib_ahmed | 0:791a779d6220 | 2806 | tmp14 = tmp10 + MULTIPLY(tmp1, FIX_0_765366865); /* c3-c9 */ |
shoaib_ahmed | 0:791a779d6220 | 2807 | tmp15 = tmp10 - MULTIPLY(tmp4, FIX_1_847759065); /* c3+c9 */ |
shoaib_ahmed | 0:791a779d6220 | 2808 | tmp12 = MULTIPLY(tmp0 + tmp2, FIX(1.121971054)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 2809 | tmp13 = MULTIPLY(tmp0 + tmp3, FIX(0.860918669)); /* c7 */ |
shoaib_ahmed | 0:791a779d6220 | 2810 | tmp10 = tmp12 + tmp13 + tmp14 - MULTIPLY(tmp0, FIX(0.580774953)) /* c5+c7-c1 */ |
shoaib_ahmed | 0:791a779d6220 | 2811 | + MULTIPLY(tmp5, FIX(0.184591911)); /* c11 */ |
shoaib_ahmed | 0:791a779d6220 | 2812 | tmp11 = MULTIPLY(tmp2 + tmp3, - FIX(0.184591911)); /* -c11 */ |
shoaib_ahmed | 0:791a779d6220 | 2813 | tmp12 += tmp11 - tmp15 - MULTIPLY(tmp2, FIX(2.339493912)) /* c1+c5-c11 */ |
shoaib_ahmed | 0:791a779d6220 | 2814 | + MULTIPLY(tmp5, FIX(0.860918669)); /* c7 */ |
shoaib_ahmed | 0:791a779d6220 | 2815 | tmp13 += tmp11 - tmp14 + MULTIPLY(tmp3, FIX(0.725788011)) /* c1+c11-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 2816 | - MULTIPLY(tmp5, FIX(1.121971054)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 2817 | tmp11 = tmp15 + MULTIPLY(tmp0 - tmp3, FIX(1.306562965)) /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 2818 | - MULTIPLY(tmp2 + tmp5, FIX_0_541196100); /* c9 */ |
shoaib_ahmed | 0:791a779d6220 | 2819 | |
shoaib_ahmed | 0:791a779d6220 | 2820 | dataptr[1] = (DCTELEM) DESCALE(tmp10, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2821 | dataptr[3] = (DCTELEM) DESCALE(tmp11, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2822 | dataptr[5] = (DCTELEM) DESCALE(tmp12, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2823 | dataptr[7] = (DCTELEM) DESCALE(tmp13, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2824 | |
shoaib_ahmed | 0:791a779d6220 | 2825 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 2826 | } |
shoaib_ahmed | 0:791a779d6220 | 2827 | |
shoaib_ahmed | 0:791a779d6220 | 2828 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 2829 | * We remove the PASS1_BITS scaling, but leave the results scaled up |
shoaib_ahmed | 0:791a779d6220 | 2830 | * by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 2831 | * We must also scale the output by (8/12)*(8/6) = 8/9, which we |
shoaib_ahmed | 0:791a779d6220 | 2832 | * partially fold into the constant multipliers and final shifting: |
shoaib_ahmed | 0:791a779d6220 | 2833 | * 6-point FDCT kernel, cK represents sqrt(2) * cos(K*pi/12) * 16/9. |
shoaib_ahmed | 0:791a779d6220 | 2834 | */ |
shoaib_ahmed | 0:791a779d6220 | 2835 | |
shoaib_ahmed | 0:791a779d6220 | 2836 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 2837 | for (ctr = DCTSIZE-1; ctr >= 0; ctr--) { |
shoaib_ahmed | 0:791a779d6220 | 2838 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 2839 | |
shoaib_ahmed | 0:791a779d6220 | 2840 | tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 2841 | tmp11 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 2842 | tmp2 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 2843 | |
shoaib_ahmed | 0:791a779d6220 | 2844 | tmp10 = tmp0 + tmp2; |
shoaib_ahmed | 0:791a779d6220 | 2845 | tmp12 = tmp0 - tmp2; |
shoaib_ahmed | 0:791a779d6220 | 2846 | |
shoaib_ahmed | 0:791a779d6220 | 2847 | tmp0 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 2848 | tmp1 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 2849 | tmp2 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 2850 | |
shoaib_ahmed | 0:791a779d6220 | 2851 | dataptr[DCTSIZE*0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2852 | DESCALE(MULTIPLY(tmp10 + tmp11, FIX(1.777777778)), /* 16/9 */ |
shoaib_ahmed | 0:791a779d6220 | 2853 | CONST_BITS+PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 2854 | dataptr[DCTSIZE*2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2855 | DESCALE(MULTIPLY(tmp12, FIX(2.177324216)), /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 2856 | CONST_BITS+PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 2857 | dataptr[DCTSIZE*4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2858 | DESCALE(MULTIPLY(tmp10 - tmp11 - tmp11, FIX(1.257078722)), /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 2859 | CONST_BITS+PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 2860 | |
shoaib_ahmed | 0:791a779d6220 | 2861 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 2862 | |
shoaib_ahmed | 0:791a779d6220 | 2863 | tmp10 = MULTIPLY(tmp0 + tmp2, FIX(0.650711829)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 2864 | |
shoaib_ahmed | 0:791a779d6220 | 2865 | dataptr[DCTSIZE*1] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2866 | DESCALE(tmp10 + MULTIPLY(tmp0 + tmp1, FIX(1.777777778)), /* 16/9 */ |
shoaib_ahmed | 0:791a779d6220 | 2867 | CONST_BITS+PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 2868 | dataptr[DCTSIZE*3] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2869 | DESCALE(MULTIPLY(tmp0 - tmp1 - tmp2, FIX(1.777777778)), /* 16/9 */ |
shoaib_ahmed | 0:791a779d6220 | 2870 | CONST_BITS+PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 2871 | dataptr[DCTSIZE*5] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2872 | DESCALE(tmp10 + MULTIPLY(tmp2 - tmp1, FIX(1.777777778)), /* 16/9 */ |
shoaib_ahmed | 0:791a779d6220 | 2873 | CONST_BITS+PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 2874 | |
shoaib_ahmed | 0:791a779d6220 | 2875 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 2876 | } |
shoaib_ahmed | 0:791a779d6220 | 2877 | } |
shoaib_ahmed | 0:791a779d6220 | 2878 | |
shoaib_ahmed | 0:791a779d6220 | 2879 | |
shoaib_ahmed | 0:791a779d6220 | 2880 | /* |
shoaib_ahmed | 0:791a779d6220 | 2881 | * Perform the forward DCT on a 10x5 sample block. |
shoaib_ahmed | 0:791a779d6220 | 2882 | * |
shoaib_ahmed | 0:791a779d6220 | 2883 | * 10-point FDCT in pass 1 (rows), 5-point in pass 2 (columns). |
shoaib_ahmed | 0:791a779d6220 | 2884 | */ |
shoaib_ahmed | 0:791a779d6220 | 2885 | |
shoaib_ahmed | 0:791a779d6220 | 2886 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 2887 | jpeg_fdct_10x5 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 2888 | { |
shoaib_ahmed | 0:791a779d6220 | 2889 | INT32 tmp0, tmp1, tmp2, tmp3, tmp4; |
shoaib_ahmed | 0:791a779d6220 | 2890 | INT32 tmp10, tmp11, tmp12, tmp13, tmp14; |
shoaib_ahmed | 0:791a779d6220 | 2891 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 2892 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 2893 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 2894 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 2895 | |
shoaib_ahmed | 0:791a779d6220 | 2896 | /* Zero 3 bottom rows of output coefficient block. */ |
shoaib_ahmed | 0:791a779d6220 | 2897 | MEMZERO(&data[DCTSIZE*5], SIZEOF(DCTELEM) * DCTSIZE * 3); |
shoaib_ahmed | 0:791a779d6220 | 2898 | |
shoaib_ahmed | 0:791a779d6220 | 2899 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 2900 | * Note results are scaled up by sqrt(8) compared to a true DCT; |
shoaib_ahmed | 0:791a779d6220 | 2901 | * furthermore, we scale the results by 2**PASS1_BITS. |
shoaib_ahmed | 0:791a779d6220 | 2902 | * 10-point FDCT kernel, cK represents sqrt(2) * cos(K*pi/20). |
shoaib_ahmed | 0:791a779d6220 | 2903 | */ |
shoaib_ahmed | 0:791a779d6220 | 2904 | |
shoaib_ahmed | 0:791a779d6220 | 2905 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 2906 | for (ctr = 0; ctr < 5; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 2907 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 2908 | |
shoaib_ahmed | 0:791a779d6220 | 2909 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 2910 | |
shoaib_ahmed | 0:791a779d6220 | 2911 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[9]); |
shoaib_ahmed | 0:791a779d6220 | 2912 | tmp1 = GETJSAMPLE(elemptr[1]) + GETJSAMPLE(elemptr[8]); |
shoaib_ahmed | 0:791a779d6220 | 2913 | tmp12 = GETJSAMPLE(elemptr[2]) + GETJSAMPLE(elemptr[7]); |
shoaib_ahmed | 0:791a779d6220 | 2914 | tmp3 = GETJSAMPLE(elemptr[3]) + GETJSAMPLE(elemptr[6]); |
shoaib_ahmed | 0:791a779d6220 | 2915 | tmp4 = GETJSAMPLE(elemptr[4]) + GETJSAMPLE(elemptr[5]); |
shoaib_ahmed | 0:791a779d6220 | 2916 | |
shoaib_ahmed | 0:791a779d6220 | 2917 | tmp10 = tmp0 + tmp4; |
shoaib_ahmed | 0:791a779d6220 | 2918 | tmp13 = tmp0 - tmp4; |
shoaib_ahmed | 0:791a779d6220 | 2919 | tmp11 = tmp1 + tmp3; |
shoaib_ahmed | 0:791a779d6220 | 2920 | tmp14 = tmp1 - tmp3; |
shoaib_ahmed | 0:791a779d6220 | 2921 | |
shoaib_ahmed | 0:791a779d6220 | 2922 | tmp0 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[9]); |
shoaib_ahmed | 0:791a779d6220 | 2923 | tmp1 = GETJSAMPLE(elemptr[1]) - GETJSAMPLE(elemptr[8]); |
shoaib_ahmed | 0:791a779d6220 | 2924 | tmp2 = GETJSAMPLE(elemptr[2]) - GETJSAMPLE(elemptr[7]); |
shoaib_ahmed | 0:791a779d6220 | 2925 | tmp3 = GETJSAMPLE(elemptr[3]) - GETJSAMPLE(elemptr[6]); |
shoaib_ahmed | 0:791a779d6220 | 2926 | tmp4 = GETJSAMPLE(elemptr[4]) - GETJSAMPLE(elemptr[5]); |
shoaib_ahmed | 0:791a779d6220 | 2927 | |
shoaib_ahmed | 0:791a779d6220 | 2928 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 2929 | dataptr[0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2930 | ((tmp10 + tmp11 + tmp12 - 10 * CENTERJSAMPLE) << PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2931 | tmp12 += tmp12; |
shoaib_ahmed | 0:791a779d6220 | 2932 | dataptr[4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2933 | DESCALE(MULTIPLY(tmp10 - tmp12, FIX(1.144122806)) - /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 2934 | MULTIPLY(tmp11 - tmp12, FIX(0.437016024)), /* c8 */ |
shoaib_ahmed | 0:791a779d6220 | 2935 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2936 | tmp10 = MULTIPLY(tmp13 + tmp14, FIX(0.831253876)); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 2937 | dataptr[2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2938 | DESCALE(tmp10 + MULTIPLY(tmp13, FIX(0.513743148)), /* c2-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 2939 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2940 | dataptr[6] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2941 | DESCALE(tmp10 - MULTIPLY(tmp14, FIX(2.176250899)), /* c2+c6 */ |
shoaib_ahmed | 0:791a779d6220 | 2942 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2943 | |
shoaib_ahmed | 0:791a779d6220 | 2944 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 2945 | |
shoaib_ahmed | 0:791a779d6220 | 2946 | tmp10 = tmp0 + tmp4; |
shoaib_ahmed | 0:791a779d6220 | 2947 | tmp11 = tmp1 - tmp3; |
shoaib_ahmed | 0:791a779d6220 | 2948 | dataptr[5] = (DCTELEM) ((tmp10 - tmp11 - tmp2) << PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2949 | tmp2 <<= CONST_BITS; |
shoaib_ahmed | 0:791a779d6220 | 2950 | dataptr[1] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2951 | DESCALE(MULTIPLY(tmp0, FIX(1.396802247)) + /* c1 */ |
shoaib_ahmed | 0:791a779d6220 | 2952 | MULTIPLY(tmp1, FIX(1.260073511)) + tmp2 + /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 2953 | MULTIPLY(tmp3, FIX(0.642039522)) + /* c7 */ |
shoaib_ahmed | 0:791a779d6220 | 2954 | MULTIPLY(tmp4, FIX(0.221231742)), /* c9 */ |
shoaib_ahmed | 0:791a779d6220 | 2955 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2956 | tmp12 = MULTIPLY(tmp0 - tmp4, FIX(0.951056516)) - /* (c3+c7)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 2957 | MULTIPLY(tmp1 + tmp3, FIX(0.587785252)); /* (c1-c9)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 2958 | tmp13 = MULTIPLY(tmp10 + tmp11, FIX(0.309016994)) + /* (c3-c7)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 2959 | (tmp11 << (CONST_BITS - 1)) - tmp2; |
shoaib_ahmed | 0:791a779d6220 | 2960 | dataptr[3] = (DCTELEM) DESCALE(tmp12 + tmp13, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2961 | dataptr[7] = (DCTELEM) DESCALE(tmp12 - tmp13, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2962 | |
shoaib_ahmed | 0:791a779d6220 | 2963 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 2964 | } |
shoaib_ahmed | 0:791a779d6220 | 2965 | |
shoaib_ahmed | 0:791a779d6220 | 2966 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 2967 | * We remove the PASS1_BITS scaling, but leave the results scaled up |
shoaib_ahmed | 0:791a779d6220 | 2968 | * by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 2969 | * We must also scale the output by (8/10)*(8/5) = 32/25, which we |
shoaib_ahmed | 0:791a779d6220 | 2970 | * fold into the constant multipliers: |
shoaib_ahmed | 0:791a779d6220 | 2971 | * 5-point FDCT kernel, cK represents sqrt(2) * cos(K*pi/10) * 32/25. |
shoaib_ahmed | 0:791a779d6220 | 2972 | */ |
shoaib_ahmed | 0:791a779d6220 | 2973 | |
shoaib_ahmed | 0:791a779d6220 | 2974 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 2975 | for (ctr = DCTSIZE-1; ctr >= 0; ctr--) { |
shoaib_ahmed | 0:791a779d6220 | 2976 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 2977 | |
shoaib_ahmed | 0:791a779d6220 | 2978 | tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 2979 | tmp1 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 2980 | tmp2 = dataptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 2981 | |
shoaib_ahmed | 0:791a779d6220 | 2982 | tmp10 = tmp0 + tmp1; |
shoaib_ahmed | 0:791a779d6220 | 2983 | tmp11 = tmp0 - tmp1; |
shoaib_ahmed | 0:791a779d6220 | 2984 | |
shoaib_ahmed | 0:791a779d6220 | 2985 | tmp0 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 2986 | tmp1 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 2987 | |
shoaib_ahmed | 0:791a779d6220 | 2988 | dataptr[DCTSIZE*0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 2989 | DESCALE(MULTIPLY(tmp10 + tmp2, FIX(1.28)), /* 32/25 */ |
shoaib_ahmed | 0:791a779d6220 | 2990 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2991 | tmp11 = MULTIPLY(tmp11, FIX(1.011928851)); /* (c2+c4)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 2992 | tmp10 -= tmp2 << 2; |
shoaib_ahmed | 0:791a779d6220 | 2993 | tmp10 = MULTIPLY(tmp10, FIX(0.452548340)); /* (c2-c4)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 2994 | dataptr[DCTSIZE*2] = (DCTELEM) DESCALE(tmp11 + tmp10, CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2995 | dataptr[DCTSIZE*4] = (DCTELEM) DESCALE(tmp11 - tmp10, CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 2996 | |
shoaib_ahmed | 0:791a779d6220 | 2997 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 2998 | |
shoaib_ahmed | 0:791a779d6220 | 2999 | tmp10 = MULTIPLY(tmp0 + tmp1, FIX(1.064004961)); /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 3000 | |
shoaib_ahmed | 0:791a779d6220 | 3001 | dataptr[DCTSIZE*1] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3002 | DESCALE(tmp10 + MULTIPLY(tmp0, FIX(0.657591230)), /* c1-c3 */ |
shoaib_ahmed | 0:791a779d6220 | 3003 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3004 | dataptr[DCTSIZE*3] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3005 | DESCALE(tmp10 - MULTIPLY(tmp1, FIX(2.785601151)), /* c1+c3 */ |
shoaib_ahmed | 0:791a779d6220 | 3006 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3007 | |
shoaib_ahmed | 0:791a779d6220 | 3008 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 3009 | } |
shoaib_ahmed | 0:791a779d6220 | 3010 | } |
shoaib_ahmed | 0:791a779d6220 | 3011 | |
shoaib_ahmed | 0:791a779d6220 | 3012 | |
shoaib_ahmed | 0:791a779d6220 | 3013 | /* |
shoaib_ahmed | 0:791a779d6220 | 3014 | * Perform the forward DCT on an 8x4 sample block. |
shoaib_ahmed | 0:791a779d6220 | 3015 | * |
shoaib_ahmed | 0:791a779d6220 | 3016 | * 8-point FDCT in pass 1 (rows), 4-point in pass 2 (columns). |
shoaib_ahmed | 0:791a779d6220 | 3017 | */ |
shoaib_ahmed | 0:791a779d6220 | 3018 | |
shoaib_ahmed | 0:791a779d6220 | 3019 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 3020 | jpeg_fdct_8x4 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 3021 | { |
shoaib_ahmed | 0:791a779d6220 | 3022 | INT32 tmp0, tmp1, tmp2, tmp3; |
shoaib_ahmed | 0:791a779d6220 | 3023 | INT32 tmp10, tmp11, tmp12, tmp13; |
shoaib_ahmed | 0:791a779d6220 | 3024 | INT32 z1; |
shoaib_ahmed | 0:791a779d6220 | 3025 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 3026 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 3027 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 3028 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 3029 | |
shoaib_ahmed | 0:791a779d6220 | 3030 | /* Zero 4 bottom rows of output coefficient block. */ |
shoaib_ahmed | 0:791a779d6220 | 3031 | MEMZERO(&data[DCTSIZE*4], SIZEOF(DCTELEM) * DCTSIZE * 4); |
shoaib_ahmed | 0:791a779d6220 | 3032 | |
shoaib_ahmed | 0:791a779d6220 | 3033 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 3034 | * Note results are scaled up by sqrt(8) compared to a true DCT; |
shoaib_ahmed | 0:791a779d6220 | 3035 | * furthermore, we scale the results by 2**PASS1_BITS. |
shoaib_ahmed | 0:791a779d6220 | 3036 | * We must also scale the output by 8/4 = 2, which we add here. |
shoaib_ahmed | 0:791a779d6220 | 3037 | * 8-point FDCT kernel, cK represents sqrt(2) * cos(K*pi/16). |
shoaib_ahmed | 0:791a779d6220 | 3038 | */ |
shoaib_ahmed | 0:791a779d6220 | 3039 | |
shoaib_ahmed | 0:791a779d6220 | 3040 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 3041 | for (ctr = 0; ctr < 4; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 3042 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 3043 | |
shoaib_ahmed | 0:791a779d6220 | 3044 | /* Even part per LL&M figure 1 --- note that published figure is faulty; |
shoaib_ahmed | 0:791a779d6220 | 3045 | * rotator "c1" should be "c6". |
shoaib_ahmed | 0:791a779d6220 | 3046 | */ |
shoaib_ahmed | 0:791a779d6220 | 3047 | |
shoaib_ahmed | 0:791a779d6220 | 3048 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[7]); |
shoaib_ahmed | 0:791a779d6220 | 3049 | tmp1 = GETJSAMPLE(elemptr[1]) + GETJSAMPLE(elemptr[6]); |
shoaib_ahmed | 0:791a779d6220 | 3050 | tmp2 = GETJSAMPLE(elemptr[2]) + GETJSAMPLE(elemptr[5]); |
shoaib_ahmed | 0:791a779d6220 | 3051 | tmp3 = GETJSAMPLE(elemptr[3]) + GETJSAMPLE(elemptr[4]); |
shoaib_ahmed | 0:791a779d6220 | 3052 | |
shoaib_ahmed | 0:791a779d6220 | 3053 | tmp10 = tmp0 + tmp3; |
shoaib_ahmed | 0:791a779d6220 | 3054 | tmp12 = tmp0 - tmp3; |
shoaib_ahmed | 0:791a779d6220 | 3055 | tmp11 = tmp1 + tmp2; |
shoaib_ahmed | 0:791a779d6220 | 3056 | tmp13 = tmp1 - tmp2; |
shoaib_ahmed | 0:791a779d6220 | 3057 | |
shoaib_ahmed | 0:791a779d6220 | 3058 | tmp0 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[7]); |
shoaib_ahmed | 0:791a779d6220 | 3059 | tmp1 = GETJSAMPLE(elemptr[1]) - GETJSAMPLE(elemptr[6]); |
shoaib_ahmed | 0:791a779d6220 | 3060 | tmp2 = GETJSAMPLE(elemptr[2]) - GETJSAMPLE(elemptr[5]); |
shoaib_ahmed | 0:791a779d6220 | 3061 | tmp3 = GETJSAMPLE(elemptr[3]) - GETJSAMPLE(elemptr[4]); |
shoaib_ahmed | 0:791a779d6220 | 3062 | |
shoaib_ahmed | 0:791a779d6220 | 3063 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 3064 | dataptr[0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3065 | ((tmp10 + tmp11 - 8 * CENTERJSAMPLE) << (PASS1_BITS+1)); |
shoaib_ahmed | 0:791a779d6220 | 3066 | dataptr[4] = (DCTELEM) ((tmp10 - tmp11) << (PASS1_BITS+1)); |
shoaib_ahmed | 0:791a779d6220 | 3067 | |
shoaib_ahmed | 0:791a779d6220 | 3068 | z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 3069 | /* Add fudge factor here for final descale. */ |
shoaib_ahmed | 0:791a779d6220 | 3070 | z1 += ONE << (CONST_BITS-PASS1_BITS-2); |
shoaib_ahmed | 0:791a779d6220 | 3071 | |
shoaib_ahmed | 0:791a779d6220 | 3072 | dataptr[2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3073 | RIGHT_SHIFT(z1 + MULTIPLY(tmp12, FIX_0_765366865), /* c2-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 3074 | CONST_BITS-PASS1_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 3075 | dataptr[6] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3076 | RIGHT_SHIFT(z1 - MULTIPLY(tmp13, FIX_1_847759065), /* c2+c6 */ |
shoaib_ahmed | 0:791a779d6220 | 3077 | CONST_BITS-PASS1_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 3078 | |
shoaib_ahmed | 0:791a779d6220 | 3079 | /* Odd part per figure 8 --- note paper omits factor of sqrt(2). |
shoaib_ahmed | 0:791a779d6220 | 3080 | * i0..i3 in the paper are tmp0..tmp3 here. |
shoaib_ahmed | 0:791a779d6220 | 3081 | */ |
shoaib_ahmed | 0:791a779d6220 | 3082 | |
shoaib_ahmed | 0:791a779d6220 | 3083 | tmp12 = tmp0 + tmp2; |
shoaib_ahmed | 0:791a779d6220 | 3084 | tmp13 = tmp1 + tmp3; |
shoaib_ahmed | 0:791a779d6220 | 3085 | |
shoaib_ahmed | 0:791a779d6220 | 3086 | z1 = MULTIPLY(tmp12 + tmp13, FIX_1_175875602); /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 3087 | /* Add fudge factor here for final descale. */ |
shoaib_ahmed | 0:791a779d6220 | 3088 | z1 += ONE << (CONST_BITS-PASS1_BITS-2); |
shoaib_ahmed | 0:791a779d6220 | 3089 | |
shoaib_ahmed | 0:791a779d6220 | 3090 | tmp12 = MULTIPLY(tmp12, - FIX_0_390180644); /* -c3+c5 */ |
shoaib_ahmed | 0:791a779d6220 | 3091 | tmp13 = MULTIPLY(tmp13, - FIX_1_961570560); /* -c3-c5 */ |
shoaib_ahmed | 0:791a779d6220 | 3092 | tmp12 += z1; |
shoaib_ahmed | 0:791a779d6220 | 3093 | tmp13 += z1; |
shoaib_ahmed | 0:791a779d6220 | 3094 | |
shoaib_ahmed | 0:791a779d6220 | 3095 | z1 = MULTIPLY(tmp0 + tmp3, - FIX_0_899976223); /* -c3+c7 */ |
shoaib_ahmed | 0:791a779d6220 | 3096 | tmp0 = MULTIPLY(tmp0, FIX_1_501321110); /* c1+c3-c5-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 3097 | tmp3 = MULTIPLY(tmp3, FIX_0_298631336); /* -c1+c3+c5-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 3098 | tmp0 += z1 + tmp12; |
shoaib_ahmed | 0:791a779d6220 | 3099 | tmp3 += z1 + tmp13; |
shoaib_ahmed | 0:791a779d6220 | 3100 | |
shoaib_ahmed | 0:791a779d6220 | 3101 | z1 = MULTIPLY(tmp1 + tmp2, - FIX_2_562915447); /* -c1-c3 */ |
shoaib_ahmed | 0:791a779d6220 | 3102 | tmp1 = MULTIPLY(tmp1, FIX_3_072711026); /* c1+c3+c5-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 3103 | tmp2 = MULTIPLY(tmp2, FIX_2_053119869); /* c1+c3-c5+c7 */ |
shoaib_ahmed | 0:791a779d6220 | 3104 | tmp1 += z1 + tmp13; |
shoaib_ahmed | 0:791a779d6220 | 3105 | tmp2 += z1 + tmp12; |
shoaib_ahmed | 0:791a779d6220 | 3106 | |
shoaib_ahmed | 0:791a779d6220 | 3107 | dataptr[1] = (DCTELEM) RIGHT_SHIFT(tmp0, CONST_BITS-PASS1_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 3108 | dataptr[3] = (DCTELEM) RIGHT_SHIFT(tmp1, CONST_BITS-PASS1_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 3109 | dataptr[5] = (DCTELEM) RIGHT_SHIFT(tmp2, CONST_BITS-PASS1_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 3110 | dataptr[7] = (DCTELEM) RIGHT_SHIFT(tmp3, CONST_BITS-PASS1_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 3111 | |
shoaib_ahmed | 0:791a779d6220 | 3112 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 3113 | } |
shoaib_ahmed | 0:791a779d6220 | 3114 | |
shoaib_ahmed | 0:791a779d6220 | 3115 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 3116 | * We remove the PASS1_BITS scaling, but leave the results scaled up |
shoaib_ahmed | 0:791a779d6220 | 3117 | * by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 3118 | * 4-point FDCT kernel, |
shoaib_ahmed | 0:791a779d6220 | 3119 | * cK represents sqrt(2) * cos(K*pi/16) [refers to 8-point FDCT]. |
shoaib_ahmed | 0:791a779d6220 | 3120 | */ |
shoaib_ahmed | 0:791a779d6220 | 3121 | |
shoaib_ahmed | 0:791a779d6220 | 3122 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 3123 | for (ctr = DCTSIZE-1; ctr >= 0; ctr--) { |
shoaib_ahmed | 0:791a779d6220 | 3124 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 3125 | |
shoaib_ahmed | 0:791a779d6220 | 3126 | /* Add fudge factor here for final descale. */ |
shoaib_ahmed | 0:791a779d6220 | 3127 | tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*3] + (ONE << (PASS1_BITS-1)); |
shoaib_ahmed | 0:791a779d6220 | 3128 | tmp1 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 3129 | |
shoaib_ahmed | 0:791a779d6220 | 3130 | tmp10 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 3131 | tmp11 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 3132 | |
shoaib_ahmed | 0:791a779d6220 | 3133 | dataptr[DCTSIZE*0] = (DCTELEM) RIGHT_SHIFT(tmp0 + tmp1, PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3134 | dataptr[DCTSIZE*2] = (DCTELEM) RIGHT_SHIFT(tmp0 - tmp1, PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3135 | |
shoaib_ahmed | 0:791a779d6220 | 3136 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 3137 | |
shoaib_ahmed | 0:791a779d6220 | 3138 | tmp0 = MULTIPLY(tmp10 + tmp11, FIX_0_541196100); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 3139 | /* Add fudge factor here for final descale. */ |
shoaib_ahmed | 0:791a779d6220 | 3140 | tmp0 += ONE << (CONST_BITS+PASS1_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 3141 | |
shoaib_ahmed | 0:791a779d6220 | 3142 | dataptr[DCTSIZE*1] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3143 | RIGHT_SHIFT(tmp0 + MULTIPLY(tmp10, FIX_0_765366865), /* c2-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 3144 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3145 | dataptr[DCTSIZE*3] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3146 | RIGHT_SHIFT(tmp0 - MULTIPLY(tmp11, FIX_1_847759065), /* c2+c6 */ |
shoaib_ahmed | 0:791a779d6220 | 3147 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3148 | |
shoaib_ahmed | 0:791a779d6220 | 3149 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 3150 | } |
shoaib_ahmed | 0:791a779d6220 | 3151 | } |
shoaib_ahmed | 0:791a779d6220 | 3152 | |
shoaib_ahmed | 0:791a779d6220 | 3153 | |
shoaib_ahmed | 0:791a779d6220 | 3154 | /* |
shoaib_ahmed | 0:791a779d6220 | 3155 | * Perform the forward DCT on a 6x3 sample block. |
shoaib_ahmed | 0:791a779d6220 | 3156 | * |
shoaib_ahmed | 0:791a779d6220 | 3157 | * 6-point FDCT in pass 1 (rows), 3-point in pass 2 (columns). |
shoaib_ahmed | 0:791a779d6220 | 3158 | */ |
shoaib_ahmed | 0:791a779d6220 | 3159 | |
shoaib_ahmed | 0:791a779d6220 | 3160 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 3161 | jpeg_fdct_6x3 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 3162 | { |
shoaib_ahmed | 0:791a779d6220 | 3163 | INT32 tmp0, tmp1, tmp2; |
shoaib_ahmed | 0:791a779d6220 | 3164 | INT32 tmp10, tmp11, tmp12; |
shoaib_ahmed | 0:791a779d6220 | 3165 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 3166 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 3167 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 3168 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 3169 | |
shoaib_ahmed | 0:791a779d6220 | 3170 | /* Pre-zero output coefficient block. */ |
shoaib_ahmed | 0:791a779d6220 | 3171 | MEMZERO(data, SIZEOF(DCTELEM) * DCTSIZE2); |
shoaib_ahmed | 0:791a779d6220 | 3172 | |
shoaib_ahmed | 0:791a779d6220 | 3173 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 3174 | * Note results are scaled up by sqrt(8) compared to a true DCT; |
shoaib_ahmed | 0:791a779d6220 | 3175 | * furthermore, we scale the results by 2**PASS1_BITS. |
shoaib_ahmed | 0:791a779d6220 | 3176 | * We scale the results further by 2 as part of output adaption |
shoaib_ahmed | 0:791a779d6220 | 3177 | * scaling for different DCT size. |
shoaib_ahmed | 0:791a779d6220 | 3178 | * 6-point FDCT kernel, cK represents sqrt(2) * cos(K*pi/12). |
shoaib_ahmed | 0:791a779d6220 | 3179 | */ |
shoaib_ahmed | 0:791a779d6220 | 3180 | |
shoaib_ahmed | 0:791a779d6220 | 3181 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 3182 | for (ctr = 0; ctr < 3; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 3183 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 3184 | |
shoaib_ahmed | 0:791a779d6220 | 3185 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 3186 | |
shoaib_ahmed | 0:791a779d6220 | 3187 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[5]); |
shoaib_ahmed | 0:791a779d6220 | 3188 | tmp11 = GETJSAMPLE(elemptr[1]) + GETJSAMPLE(elemptr[4]); |
shoaib_ahmed | 0:791a779d6220 | 3189 | tmp2 = GETJSAMPLE(elemptr[2]) + GETJSAMPLE(elemptr[3]); |
shoaib_ahmed | 0:791a779d6220 | 3190 | |
shoaib_ahmed | 0:791a779d6220 | 3191 | tmp10 = tmp0 + tmp2; |
shoaib_ahmed | 0:791a779d6220 | 3192 | tmp12 = tmp0 - tmp2; |
shoaib_ahmed | 0:791a779d6220 | 3193 | |
shoaib_ahmed | 0:791a779d6220 | 3194 | tmp0 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[5]); |
shoaib_ahmed | 0:791a779d6220 | 3195 | tmp1 = GETJSAMPLE(elemptr[1]) - GETJSAMPLE(elemptr[4]); |
shoaib_ahmed | 0:791a779d6220 | 3196 | tmp2 = GETJSAMPLE(elemptr[2]) - GETJSAMPLE(elemptr[3]); |
shoaib_ahmed | 0:791a779d6220 | 3197 | |
shoaib_ahmed | 0:791a779d6220 | 3198 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 3199 | dataptr[0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3200 | ((tmp10 + tmp11 - 6 * CENTERJSAMPLE) << (PASS1_BITS+1)); |
shoaib_ahmed | 0:791a779d6220 | 3201 | dataptr[2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3202 | DESCALE(MULTIPLY(tmp12, FIX(1.224744871)), /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 3203 | CONST_BITS-PASS1_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 3204 | dataptr[4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3205 | DESCALE(MULTIPLY(tmp10 - tmp11 - tmp11, FIX(0.707106781)), /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 3206 | CONST_BITS-PASS1_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 3207 | |
shoaib_ahmed | 0:791a779d6220 | 3208 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 3209 | |
shoaib_ahmed | 0:791a779d6220 | 3210 | tmp10 = DESCALE(MULTIPLY(tmp0 + tmp2, FIX(0.366025404)), /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 3211 | CONST_BITS-PASS1_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 3212 | |
shoaib_ahmed | 0:791a779d6220 | 3213 | dataptr[1] = (DCTELEM) (tmp10 + ((tmp0 + tmp1) << (PASS1_BITS+1))); |
shoaib_ahmed | 0:791a779d6220 | 3214 | dataptr[3] = (DCTELEM) ((tmp0 - tmp1 - tmp2) << (PASS1_BITS+1)); |
shoaib_ahmed | 0:791a779d6220 | 3215 | dataptr[5] = (DCTELEM) (tmp10 + ((tmp2 - tmp1) << (PASS1_BITS+1))); |
shoaib_ahmed | 0:791a779d6220 | 3216 | |
shoaib_ahmed | 0:791a779d6220 | 3217 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 3218 | } |
shoaib_ahmed | 0:791a779d6220 | 3219 | |
shoaib_ahmed | 0:791a779d6220 | 3220 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 3221 | * We remove the PASS1_BITS scaling, but leave the results scaled up |
shoaib_ahmed | 0:791a779d6220 | 3222 | * by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 3223 | * We must also scale the output by (8/6)*(8/3) = 32/9, which we partially |
shoaib_ahmed | 0:791a779d6220 | 3224 | * fold into the constant multipliers (other part was done in pass 1): |
shoaib_ahmed | 0:791a779d6220 | 3225 | * 3-point FDCT kernel, cK represents sqrt(2) * cos(K*pi/6) * 16/9. |
shoaib_ahmed | 0:791a779d6220 | 3226 | */ |
shoaib_ahmed | 0:791a779d6220 | 3227 | |
shoaib_ahmed | 0:791a779d6220 | 3228 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 3229 | for (ctr = 0; ctr < 6; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 3230 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 3231 | |
shoaib_ahmed | 0:791a779d6220 | 3232 | tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 3233 | tmp1 = dataptr[DCTSIZE*1]; |
shoaib_ahmed | 0:791a779d6220 | 3234 | |
shoaib_ahmed | 0:791a779d6220 | 3235 | tmp2 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 3236 | |
shoaib_ahmed | 0:791a779d6220 | 3237 | dataptr[DCTSIZE*0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3238 | DESCALE(MULTIPLY(tmp0 + tmp1, FIX(1.777777778)), /* 16/9 */ |
shoaib_ahmed | 0:791a779d6220 | 3239 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3240 | dataptr[DCTSIZE*2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3241 | DESCALE(MULTIPLY(tmp0 - tmp1 - tmp1, FIX(1.257078722)), /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 3242 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3243 | |
shoaib_ahmed | 0:791a779d6220 | 3244 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 3245 | |
shoaib_ahmed | 0:791a779d6220 | 3246 | dataptr[DCTSIZE*1] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3247 | DESCALE(MULTIPLY(tmp2, FIX(2.177324216)), /* c1 */ |
shoaib_ahmed | 0:791a779d6220 | 3248 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3249 | |
shoaib_ahmed | 0:791a779d6220 | 3250 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 3251 | } |
shoaib_ahmed | 0:791a779d6220 | 3252 | } |
shoaib_ahmed | 0:791a779d6220 | 3253 | |
shoaib_ahmed | 0:791a779d6220 | 3254 | |
shoaib_ahmed | 0:791a779d6220 | 3255 | /* |
shoaib_ahmed | 0:791a779d6220 | 3256 | * Perform the forward DCT on a 4x2 sample block. |
shoaib_ahmed | 0:791a779d6220 | 3257 | * |
shoaib_ahmed | 0:791a779d6220 | 3258 | * 4-point FDCT in pass 1 (rows), 2-point in pass 2 (columns). |
shoaib_ahmed | 0:791a779d6220 | 3259 | */ |
shoaib_ahmed | 0:791a779d6220 | 3260 | |
shoaib_ahmed | 0:791a779d6220 | 3261 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 3262 | jpeg_fdct_4x2 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 3263 | { |
shoaib_ahmed | 0:791a779d6220 | 3264 | INT32 tmp0, tmp1; |
shoaib_ahmed | 0:791a779d6220 | 3265 | INT32 tmp10, tmp11; |
shoaib_ahmed | 0:791a779d6220 | 3266 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 3267 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 3268 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 3269 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 3270 | |
shoaib_ahmed | 0:791a779d6220 | 3271 | /* Pre-zero output coefficient block. */ |
shoaib_ahmed | 0:791a779d6220 | 3272 | MEMZERO(data, SIZEOF(DCTELEM) * DCTSIZE2); |
shoaib_ahmed | 0:791a779d6220 | 3273 | |
shoaib_ahmed | 0:791a779d6220 | 3274 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 3275 | * Note results are scaled up by sqrt(8) compared to a true DCT; |
shoaib_ahmed | 0:791a779d6220 | 3276 | * furthermore, we scale the results by 2**PASS1_BITS. |
shoaib_ahmed | 0:791a779d6220 | 3277 | * We must also scale the output by (8/4)*(8/2) = 2**3, which we add here. |
shoaib_ahmed | 0:791a779d6220 | 3278 | * 4-point FDCT kernel, |
shoaib_ahmed | 0:791a779d6220 | 3279 | * cK represents sqrt(2) * cos(K*pi/16) [refers to 8-point FDCT]. |
shoaib_ahmed | 0:791a779d6220 | 3280 | */ |
shoaib_ahmed | 0:791a779d6220 | 3281 | |
shoaib_ahmed | 0:791a779d6220 | 3282 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 3283 | for (ctr = 0; ctr < 2; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 3284 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 3285 | |
shoaib_ahmed | 0:791a779d6220 | 3286 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 3287 | |
shoaib_ahmed | 0:791a779d6220 | 3288 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[3]); |
shoaib_ahmed | 0:791a779d6220 | 3289 | tmp1 = GETJSAMPLE(elemptr[1]) + GETJSAMPLE(elemptr[2]); |
shoaib_ahmed | 0:791a779d6220 | 3290 | |
shoaib_ahmed | 0:791a779d6220 | 3291 | tmp10 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[3]); |
shoaib_ahmed | 0:791a779d6220 | 3292 | tmp11 = GETJSAMPLE(elemptr[1]) - GETJSAMPLE(elemptr[2]); |
shoaib_ahmed | 0:791a779d6220 | 3293 | |
shoaib_ahmed | 0:791a779d6220 | 3294 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 3295 | dataptr[0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3296 | ((tmp0 + tmp1 - 4 * CENTERJSAMPLE) << (PASS1_BITS+3)); |
shoaib_ahmed | 0:791a779d6220 | 3297 | dataptr[2] = (DCTELEM) ((tmp0 - tmp1) << (PASS1_BITS+3)); |
shoaib_ahmed | 0:791a779d6220 | 3298 | |
shoaib_ahmed | 0:791a779d6220 | 3299 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 3300 | |
shoaib_ahmed | 0:791a779d6220 | 3301 | tmp0 = MULTIPLY(tmp10 + tmp11, FIX_0_541196100); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 3302 | /* Add fudge factor here for final descale. */ |
shoaib_ahmed | 0:791a779d6220 | 3303 | tmp0 += ONE << (CONST_BITS-PASS1_BITS-4); |
shoaib_ahmed | 0:791a779d6220 | 3304 | |
shoaib_ahmed | 0:791a779d6220 | 3305 | dataptr[1] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3306 | RIGHT_SHIFT(tmp0 + MULTIPLY(tmp10, FIX_0_765366865), /* c2-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 3307 | CONST_BITS-PASS1_BITS-3); |
shoaib_ahmed | 0:791a779d6220 | 3308 | dataptr[3] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3309 | RIGHT_SHIFT(tmp0 - MULTIPLY(tmp11, FIX_1_847759065), /* c2+c6 */ |
shoaib_ahmed | 0:791a779d6220 | 3310 | CONST_BITS-PASS1_BITS-3); |
shoaib_ahmed | 0:791a779d6220 | 3311 | |
shoaib_ahmed | 0:791a779d6220 | 3312 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 3313 | } |
shoaib_ahmed | 0:791a779d6220 | 3314 | |
shoaib_ahmed | 0:791a779d6220 | 3315 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 3316 | * We remove the PASS1_BITS scaling, but leave the results scaled up |
shoaib_ahmed | 0:791a779d6220 | 3317 | * by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 3318 | */ |
shoaib_ahmed | 0:791a779d6220 | 3319 | |
shoaib_ahmed | 0:791a779d6220 | 3320 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 3321 | for (ctr = 0; ctr < 4; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 3322 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 3323 | |
shoaib_ahmed | 0:791a779d6220 | 3324 | /* Add fudge factor here for final descale. */ |
shoaib_ahmed | 0:791a779d6220 | 3325 | tmp0 = dataptr[DCTSIZE*0] + (ONE << (PASS1_BITS-1)); |
shoaib_ahmed | 0:791a779d6220 | 3326 | tmp1 = dataptr[DCTSIZE*1]; |
shoaib_ahmed | 0:791a779d6220 | 3327 | |
shoaib_ahmed | 0:791a779d6220 | 3328 | dataptr[DCTSIZE*0] = (DCTELEM) RIGHT_SHIFT(tmp0 + tmp1, PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3329 | |
shoaib_ahmed | 0:791a779d6220 | 3330 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 3331 | |
shoaib_ahmed | 0:791a779d6220 | 3332 | dataptr[DCTSIZE*1] = (DCTELEM) RIGHT_SHIFT(tmp0 - tmp1, PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3333 | |
shoaib_ahmed | 0:791a779d6220 | 3334 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 3335 | } |
shoaib_ahmed | 0:791a779d6220 | 3336 | } |
shoaib_ahmed | 0:791a779d6220 | 3337 | |
shoaib_ahmed | 0:791a779d6220 | 3338 | |
shoaib_ahmed | 0:791a779d6220 | 3339 | /* |
shoaib_ahmed | 0:791a779d6220 | 3340 | * Perform the forward DCT on a 2x1 sample block. |
shoaib_ahmed | 0:791a779d6220 | 3341 | * |
shoaib_ahmed | 0:791a779d6220 | 3342 | * 2-point FDCT in pass 1 (rows), 1-point in pass 2 (columns). |
shoaib_ahmed | 0:791a779d6220 | 3343 | */ |
shoaib_ahmed | 0:791a779d6220 | 3344 | |
shoaib_ahmed | 0:791a779d6220 | 3345 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 3346 | jpeg_fdct_2x1 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 3347 | { |
shoaib_ahmed | 0:791a779d6220 | 3348 | DCTELEM tmp0, tmp1; |
shoaib_ahmed | 0:791a779d6220 | 3349 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 3350 | |
shoaib_ahmed | 0:791a779d6220 | 3351 | /* Pre-zero output coefficient block. */ |
shoaib_ahmed | 0:791a779d6220 | 3352 | MEMZERO(data, SIZEOF(DCTELEM) * DCTSIZE2); |
shoaib_ahmed | 0:791a779d6220 | 3353 | |
shoaib_ahmed | 0:791a779d6220 | 3354 | elemptr = sample_data[0] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 3355 | |
shoaib_ahmed | 0:791a779d6220 | 3356 | tmp0 = GETJSAMPLE(elemptr[0]); |
shoaib_ahmed | 0:791a779d6220 | 3357 | tmp1 = GETJSAMPLE(elemptr[1]); |
shoaib_ahmed | 0:791a779d6220 | 3358 | |
shoaib_ahmed | 0:791a779d6220 | 3359 | /* We leave the results scaled up by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 3360 | * We must also scale the output by (8/2)*(8/1) = 2**5. |
shoaib_ahmed | 0:791a779d6220 | 3361 | */ |
shoaib_ahmed | 0:791a779d6220 | 3362 | |
shoaib_ahmed | 0:791a779d6220 | 3363 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 3364 | |
shoaib_ahmed | 0:791a779d6220 | 3365 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 3366 | data[0] = (tmp0 + tmp1 - 2 * CENTERJSAMPLE) << 5; |
shoaib_ahmed | 0:791a779d6220 | 3367 | |
shoaib_ahmed | 0:791a779d6220 | 3368 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 3369 | |
shoaib_ahmed | 0:791a779d6220 | 3370 | data[1] = (tmp0 - tmp1) << 5; |
shoaib_ahmed | 0:791a779d6220 | 3371 | } |
shoaib_ahmed | 0:791a779d6220 | 3372 | |
shoaib_ahmed | 0:791a779d6220 | 3373 | |
shoaib_ahmed | 0:791a779d6220 | 3374 | /* |
shoaib_ahmed | 0:791a779d6220 | 3375 | * Perform the forward DCT on an 8x16 sample block. |
shoaib_ahmed | 0:791a779d6220 | 3376 | * |
shoaib_ahmed | 0:791a779d6220 | 3377 | * 8-point FDCT in pass 1 (rows), 16-point in pass 2 (columns). |
shoaib_ahmed | 0:791a779d6220 | 3378 | */ |
shoaib_ahmed | 0:791a779d6220 | 3379 | |
shoaib_ahmed | 0:791a779d6220 | 3380 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 3381 | jpeg_fdct_8x16 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 3382 | { |
shoaib_ahmed | 0:791a779d6220 | 3383 | INT32 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7; |
shoaib_ahmed | 0:791a779d6220 | 3384 | INT32 tmp10, tmp11, tmp12, tmp13, tmp14, tmp15, tmp16, tmp17; |
shoaib_ahmed | 0:791a779d6220 | 3385 | INT32 z1; |
shoaib_ahmed | 0:791a779d6220 | 3386 | DCTELEM workspace[DCTSIZE2]; |
shoaib_ahmed | 0:791a779d6220 | 3387 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 3388 | DCTELEM *wsptr; |
shoaib_ahmed | 0:791a779d6220 | 3389 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 3390 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 3391 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 3392 | |
shoaib_ahmed | 0:791a779d6220 | 3393 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 3394 | * Note results are scaled up by sqrt(8) compared to a true DCT; |
shoaib_ahmed | 0:791a779d6220 | 3395 | * furthermore, we scale the results by 2**PASS1_BITS. |
shoaib_ahmed | 0:791a779d6220 | 3396 | * 8-point FDCT kernel, cK represents sqrt(2) * cos(K*pi/16). |
shoaib_ahmed | 0:791a779d6220 | 3397 | */ |
shoaib_ahmed | 0:791a779d6220 | 3398 | |
shoaib_ahmed | 0:791a779d6220 | 3399 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 3400 | ctr = 0; |
shoaib_ahmed | 0:791a779d6220 | 3401 | for (;;) { |
shoaib_ahmed | 0:791a779d6220 | 3402 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 3403 | |
shoaib_ahmed | 0:791a779d6220 | 3404 | /* Even part per LL&M figure 1 --- note that published figure is faulty; |
shoaib_ahmed | 0:791a779d6220 | 3405 | * rotator "c1" should be "c6". |
shoaib_ahmed | 0:791a779d6220 | 3406 | */ |
shoaib_ahmed | 0:791a779d6220 | 3407 | |
shoaib_ahmed | 0:791a779d6220 | 3408 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[7]); |
shoaib_ahmed | 0:791a779d6220 | 3409 | tmp1 = GETJSAMPLE(elemptr[1]) + GETJSAMPLE(elemptr[6]); |
shoaib_ahmed | 0:791a779d6220 | 3410 | tmp2 = GETJSAMPLE(elemptr[2]) + GETJSAMPLE(elemptr[5]); |
shoaib_ahmed | 0:791a779d6220 | 3411 | tmp3 = GETJSAMPLE(elemptr[3]) + GETJSAMPLE(elemptr[4]); |
shoaib_ahmed | 0:791a779d6220 | 3412 | |
shoaib_ahmed | 0:791a779d6220 | 3413 | tmp10 = tmp0 + tmp3; |
shoaib_ahmed | 0:791a779d6220 | 3414 | tmp12 = tmp0 - tmp3; |
shoaib_ahmed | 0:791a779d6220 | 3415 | tmp11 = tmp1 + tmp2; |
shoaib_ahmed | 0:791a779d6220 | 3416 | tmp13 = tmp1 - tmp2; |
shoaib_ahmed | 0:791a779d6220 | 3417 | |
shoaib_ahmed | 0:791a779d6220 | 3418 | tmp0 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[7]); |
shoaib_ahmed | 0:791a779d6220 | 3419 | tmp1 = GETJSAMPLE(elemptr[1]) - GETJSAMPLE(elemptr[6]); |
shoaib_ahmed | 0:791a779d6220 | 3420 | tmp2 = GETJSAMPLE(elemptr[2]) - GETJSAMPLE(elemptr[5]); |
shoaib_ahmed | 0:791a779d6220 | 3421 | tmp3 = GETJSAMPLE(elemptr[3]) - GETJSAMPLE(elemptr[4]); |
shoaib_ahmed | 0:791a779d6220 | 3422 | |
shoaib_ahmed | 0:791a779d6220 | 3423 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 3424 | dataptr[0] = (DCTELEM) ((tmp10 + tmp11 - 8 * CENTERJSAMPLE) << PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3425 | dataptr[4] = (DCTELEM) ((tmp10 - tmp11) << PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3426 | |
shoaib_ahmed | 0:791a779d6220 | 3427 | z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 3428 | dataptr[2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3429 | DESCALE(z1 + MULTIPLY(tmp12, FIX_0_765366865), /* c2-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 3430 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3431 | dataptr[6] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3432 | DESCALE(z1 - MULTIPLY(tmp13, FIX_1_847759065), /* c2+c6 */ |
shoaib_ahmed | 0:791a779d6220 | 3433 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3434 | |
shoaib_ahmed | 0:791a779d6220 | 3435 | /* Odd part per figure 8 --- note paper omits factor of sqrt(2). |
shoaib_ahmed | 0:791a779d6220 | 3436 | * i0..i3 in the paper are tmp0..tmp3 here. |
shoaib_ahmed | 0:791a779d6220 | 3437 | */ |
shoaib_ahmed | 0:791a779d6220 | 3438 | |
shoaib_ahmed | 0:791a779d6220 | 3439 | tmp12 = tmp0 + tmp2; |
shoaib_ahmed | 0:791a779d6220 | 3440 | tmp13 = tmp1 + tmp3; |
shoaib_ahmed | 0:791a779d6220 | 3441 | |
shoaib_ahmed | 0:791a779d6220 | 3442 | z1 = MULTIPLY(tmp12 + tmp13, FIX_1_175875602); /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 3443 | tmp12 = MULTIPLY(tmp12, - FIX_0_390180644); /* -c3+c5 */ |
shoaib_ahmed | 0:791a779d6220 | 3444 | tmp13 = MULTIPLY(tmp13, - FIX_1_961570560); /* -c3-c5 */ |
shoaib_ahmed | 0:791a779d6220 | 3445 | tmp12 += z1; |
shoaib_ahmed | 0:791a779d6220 | 3446 | tmp13 += z1; |
shoaib_ahmed | 0:791a779d6220 | 3447 | |
shoaib_ahmed | 0:791a779d6220 | 3448 | z1 = MULTIPLY(tmp0 + tmp3, - FIX_0_899976223); /* -c3+c7 */ |
shoaib_ahmed | 0:791a779d6220 | 3449 | tmp0 = MULTIPLY(tmp0, FIX_1_501321110); /* c1+c3-c5-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 3450 | tmp3 = MULTIPLY(tmp3, FIX_0_298631336); /* -c1+c3+c5-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 3451 | tmp0 += z1 + tmp12; |
shoaib_ahmed | 0:791a779d6220 | 3452 | tmp3 += z1 + tmp13; |
shoaib_ahmed | 0:791a779d6220 | 3453 | |
shoaib_ahmed | 0:791a779d6220 | 3454 | z1 = MULTIPLY(tmp1 + tmp2, - FIX_2_562915447); /* -c1-c3 */ |
shoaib_ahmed | 0:791a779d6220 | 3455 | tmp1 = MULTIPLY(tmp1, FIX_3_072711026); /* c1+c3+c5-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 3456 | tmp2 = MULTIPLY(tmp2, FIX_2_053119869); /* c1+c3-c5+c7 */ |
shoaib_ahmed | 0:791a779d6220 | 3457 | tmp1 += z1 + tmp13; |
shoaib_ahmed | 0:791a779d6220 | 3458 | tmp2 += z1 + tmp12; |
shoaib_ahmed | 0:791a779d6220 | 3459 | |
shoaib_ahmed | 0:791a779d6220 | 3460 | dataptr[1] = (DCTELEM) DESCALE(tmp0, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3461 | dataptr[3] = (DCTELEM) DESCALE(tmp1, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3462 | dataptr[5] = (DCTELEM) DESCALE(tmp2, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3463 | dataptr[7] = (DCTELEM) DESCALE(tmp3, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3464 | |
shoaib_ahmed | 0:791a779d6220 | 3465 | ctr++; |
shoaib_ahmed | 0:791a779d6220 | 3466 | |
shoaib_ahmed | 0:791a779d6220 | 3467 | if (ctr != DCTSIZE) { |
shoaib_ahmed | 0:791a779d6220 | 3468 | if (ctr == DCTSIZE * 2) |
shoaib_ahmed | 0:791a779d6220 | 3469 | break; /* Done. */ |
shoaib_ahmed | 0:791a779d6220 | 3470 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 3471 | } else |
shoaib_ahmed | 0:791a779d6220 | 3472 | dataptr = workspace; /* switch pointer to extended workspace */ |
shoaib_ahmed | 0:791a779d6220 | 3473 | } |
shoaib_ahmed | 0:791a779d6220 | 3474 | |
shoaib_ahmed | 0:791a779d6220 | 3475 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 3476 | * We remove the PASS1_BITS scaling, but leave the results scaled up |
shoaib_ahmed | 0:791a779d6220 | 3477 | * by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 3478 | * We must also scale the output by 8/16 = 1/2. |
shoaib_ahmed | 0:791a779d6220 | 3479 | * 16-point FDCT kernel, cK represents sqrt(2) * cos(K*pi/32). |
shoaib_ahmed | 0:791a779d6220 | 3480 | */ |
shoaib_ahmed | 0:791a779d6220 | 3481 | |
shoaib_ahmed | 0:791a779d6220 | 3482 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 3483 | wsptr = workspace; |
shoaib_ahmed | 0:791a779d6220 | 3484 | for (ctr = DCTSIZE-1; ctr >= 0; ctr--) { |
shoaib_ahmed | 0:791a779d6220 | 3485 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 3486 | |
shoaib_ahmed | 0:791a779d6220 | 3487 | tmp0 = dataptr[DCTSIZE*0] + wsptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 3488 | tmp1 = dataptr[DCTSIZE*1] + wsptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 3489 | tmp2 = dataptr[DCTSIZE*2] + wsptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 3490 | tmp3 = dataptr[DCTSIZE*3] + wsptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 3491 | tmp4 = dataptr[DCTSIZE*4] + wsptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 3492 | tmp5 = dataptr[DCTSIZE*5] + wsptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 3493 | tmp6 = dataptr[DCTSIZE*6] + wsptr[DCTSIZE*1]; |
shoaib_ahmed | 0:791a779d6220 | 3494 | tmp7 = dataptr[DCTSIZE*7] + wsptr[DCTSIZE*0]; |
shoaib_ahmed | 0:791a779d6220 | 3495 | |
shoaib_ahmed | 0:791a779d6220 | 3496 | tmp10 = tmp0 + tmp7; |
shoaib_ahmed | 0:791a779d6220 | 3497 | tmp14 = tmp0 - tmp7; |
shoaib_ahmed | 0:791a779d6220 | 3498 | tmp11 = tmp1 + tmp6; |
shoaib_ahmed | 0:791a779d6220 | 3499 | tmp15 = tmp1 - tmp6; |
shoaib_ahmed | 0:791a779d6220 | 3500 | tmp12 = tmp2 + tmp5; |
shoaib_ahmed | 0:791a779d6220 | 3501 | tmp16 = tmp2 - tmp5; |
shoaib_ahmed | 0:791a779d6220 | 3502 | tmp13 = tmp3 + tmp4; |
shoaib_ahmed | 0:791a779d6220 | 3503 | tmp17 = tmp3 - tmp4; |
shoaib_ahmed | 0:791a779d6220 | 3504 | |
shoaib_ahmed | 0:791a779d6220 | 3505 | tmp0 = dataptr[DCTSIZE*0] - wsptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 3506 | tmp1 = dataptr[DCTSIZE*1] - wsptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 3507 | tmp2 = dataptr[DCTSIZE*2] - wsptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 3508 | tmp3 = dataptr[DCTSIZE*3] - wsptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 3509 | tmp4 = dataptr[DCTSIZE*4] - wsptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 3510 | tmp5 = dataptr[DCTSIZE*5] - wsptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 3511 | tmp6 = dataptr[DCTSIZE*6] - wsptr[DCTSIZE*1]; |
shoaib_ahmed | 0:791a779d6220 | 3512 | tmp7 = dataptr[DCTSIZE*7] - wsptr[DCTSIZE*0]; |
shoaib_ahmed | 0:791a779d6220 | 3513 | |
shoaib_ahmed | 0:791a779d6220 | 3514 | dataptr[DCTSIZE*0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3515 | DESCALE(tmp10 + tmp11 + tmp12 + tmp13, PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 3516 | dataptr[DCTSIZE*4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3517 | DESCALE(MULTIPLY(tmp10 - tmp13, FIX(1.306562965)) + /* c4[16] = c2[8] */ |
shoaib_ahmed | 0:791a779d6220 | 3518 | MULTIPLY(tmp11 - tmp12, FIX_0_541196100), /* c12[16] = c6[8] */ |
shoaib_ahmed | 0:791a779d6220 | 3519 | CONST_BITS+PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 3520 | |
shoaib_ahmed | 0:791a779d6220 | 3521 | tmp10 = MULTIPLY(tmp17 - tmp15, FIX(0.275899379)) + /* c14[16] = c7[8] */ |
shoaib_ahmed | 0:791a779d6220 | 3522 | MULTIPLY(tmp14 - tmp16, FIX(1.387039845)); /* c2[16] = c1[8] */ |
shoaib_ahmed | 0:791a779d6220 | 3523 | |
shoaib_ahmed | 0:791a779d6220 | 3524 | dataptr[DCTSIZE*2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3525 | DESCALE(tmp10 + MULTIPLY(tmp15, FIX(1.451774982)) /* c6+c14 */ |
shoaib_ahmed | 0:791a779d6220 | 3526 | + MULTIPLY(tmp16, FIX(2.172734804)), /* c2+c10 */ |
shoaib_ahmed | 0:791a779d6220 | 3527 | CONST_BITS+PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 3528 | dataptr[DCTSIZE*6] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3529 | DESCALE(tmp10 - MULTIPLY(tmp14, FIX(0.211164243)) /* c2-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 3530 | - MULTIPLY(tmp17, FIX(1.061594338)), /* c10+c14 */ |
shoaib_ahmed | 0:791a779d6220 | 3531 | CONST_BITS+PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 3532 | |
shoaib_ahmed | 0:791a779d6220 | 3533 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 3534 | |
shoaib_ahmed | 0:791a779d6220 | 3535 | tmp11 = MULTIPLY(tmp0 + tmp1, FIX(1.353318001)) + /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 3536 | MULTIPLY(tmp6 - tmp7, FIX(0.410524528)); /* c13 */ |
shoaib_ahmed | 0:791a779d6220 | 3537 | tmp12 = MULTIPLY(tmp0 + tmp2, FIX(1.247225013)) + /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 3538 | MULTIPLY(tmp5 + tmp7, FIX(0.666655658)); /* c11 */ |
shoaib_ahmed | 0:791a779d6220 | 3539 | tmp13 = MULTIPLY(tmp0 + tmp3, FIX(1.093201867)) + /* c7 */ |
shoaib_ahmed | 0:791a779d6220 | 3540 | MULTIPLY(tmp4 - tmp7, FIX(0.897167586)); /* c9 */ |
shoaib_ahmed | 0:791a779d6220 | 3541 | tmp14 = MULTIPLY(tmp1 + tmp2, FIX(0.138617169)) + /* c15 */ |
shoaib_ahmed | 0:791a779d6220 | 3542 | MULTIPLY(tmp6 - tmp5, FIX(1.407403738)); /* c1 */ |
shoaib_ahmed | 0:791a779d6220 | 3543 | tmp15 = MULTIPLY(tmp1 + tmp3, - FIX(0.666655658)) + /* -c11 */ |
shoaib_ahmed | 0:791a779d6220 | 3544 | MULTIPLY(tmp4 + tmp6, - FIX(1.247225013)); /* -c5 */ |
shoaib_ahmed | 0:791a779d6220 | 3545 | tmp16 = MULTIPLY(tmp2 + tmp3, - FIX(1.353318001)) + /* -c3 */ |
shoaib_ahmed | 0:791a779d6220 | 3546 | MULTIPLY(tmp5 - tmp4, FIX(0.410524528)); /* c13 */ |
shoaib_ahmed | 0:791a779d6220 | 3547 | tmp10 = tmp11 + tmp12 + tmp13 - |
shoaib_ahmed | 0:791a779d6220 | 3548 | MULTIPLY(tmp0, FIX(2.286341144)) + /* c7+c5+c3-c1 */ |
shoaib_ahmed | 0:791a779d6220 | 3549 | MULTIPLY(tmp7, FIX(0.779653625)); /* c15+c13-c11+c9 */ |
shoaib_ahmed | 0:791a779d6220 | 3550 | tmp11 += tmp14 + tmp15 + MULTIPLY(tmp1, FIX(0.071888074)) /* c9-c3-c15+c11 */ |
shoaib_ahmed | 0:791a779d6220 | 3551 | - MULTIPLY(tmp6, FIX(1.663905119)); /* c7+c13+c1-c5 */ |
shoaib_ahmed | 0:791a779d6220 | 3552 | tmp12 += tmp14 + tmp16 - MULTIPLY(tmp2, FIX(1.125726048)) /* c7+c5+c15-c3 */ |
shoaib_ahmed | 0:791a779d6220 | 3553 | + MULTIPLY(tmp5, FIX(1.227391138)); /* c9-c11+c1-c13 */ |
shoaib_ahmed | 0:791a779d6220 | 3554 | tmp13 += tmp15 + tmp16 + MULTIPLY(tmp3, FIX(1.065388962)) /* c15+c3+c11-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 3555 | + MULTIPLY(tmp4, FIX(2.167985692)); /* c1+c13+c5-c9 */ |
shoaib_ahmed | 0:791a779d6220 | 3556 | |
shoaib_ahmed | 0:791a779d6220 | 3557 | dataptr[DCTSIZE*1] = (DCTELEM) DESCALE(tmp10, CONST_BITS+PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 3558 | dataptr[DCTSIZE*3] = (DCTELEM) DESCALE(tmp11, CONST_BITS+PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 3559 | dataptr[DCTSIZE*5] = (DCTELEM) DESCALE(tmp12, CONST_BITS+PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 3560 | dataptr[DCTSIZE*7] = (DCTELEM) DESCALE(tmp13, CONST_BITS+PASS1_BITS+1); |
shoaib_ahmed | 0:791a779d6220 | 3561 | |
shoaib_ahmed | 0:791a779d6220 | 3562 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 3563 | wsptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 3564 | } |
shoaib_ahmed | 0:791a779d6220 | 3565 | } |
shoaib_ahmed | 0:791a779d6220 | 3566 | |
shoaib_ahmed | 0:791a779d6220 | 3567 | |
shoaib_ahmed | 0:791a779d6220 | 3568 | /* |
shoaib_ahmed | 0:791a779d6220 | 3569 | * Perform the forward DCT on a 7x14 sample block. |
shoaib_ahmed | 0:791a779d6220 | 3570 | * |
shoaib_ahmed | 0:791a779d6220 | 3571 | * 7-point FDCT in pass 1 (rows), 14-point in pass 2 (columns). |
shoaib_ahmed | 0:791a779d6220 | 3572 | */ |
shoaib_ahmed | 0:791a779d6220 | 3573 | |
shoaib_ahmed | 0:791a779d6220 | 3574 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 3575 | jpeg_fdct_7x14 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 3576 | { |
shoaib_ahmed | 0:791a779d6220 | 3577 | INT32 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6; |
shoaib_ahmed | 0:791a779d6220 | 3578 | INT32 tmp10, tmp11, tmp12, tmp13, tmp14, tmp15, tmp16; |
shoaib_ahmed | 0:791a779d6220 | 3579 | INT32 z1, z2, z3; |
shoaib_ahmed | 0:791a779d6220 | 3580 | DCTELEM workspace[8*6]; |
shoaib_ahmed | 0:791a779d6220 | 3581 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 3582 | DCTELEM *wsptr; |
shoaib_ahmed | 0:791a779d6220 | 3583 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 3584 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 3585 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 3586 | |
shoaib_ahmed | 0:791a779d6220 | 3587 | /* Pre-zero output coefficient block. */ |
shoaib_ahmed | 0:791a779d6220 | 3588 | MEMZERO(data, SIZEOF(DCTELEM) * DCTSIZE2); |
shoaib_ahmed | 0:791a779d6220 | 3589 | |
shoaib_ahmed | 0:791a779d6220 | 3590 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 3591 | * Note results are scaled up by sqrt(8) compared to a true DCT; |
shoaib_ahmed | 0:791a779d6220 | 3592 | * furthermore, we scale the results by 2**PASS1_BITS. |
shoaib_ahmed | 0:791a779d6220 | 3593 | * 7-point FDCT kernel, cK represents sqrt(2) * cos(K*pi/14). |
shoaib_ahmed | 0:791a779d6220 | 3594 | */ |
shoaib_ahmed | 0:791a779d6220 | 3595 | |
shoaib_ahmed | 0:791a779d6220 | 3596 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 3597 | ctr = 0; |
shoaib_ahmed | 0:791a779d6220 | 3598 | for (;;) { |
shoaib_ahmed | 0:791a779d6220 | 3599 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 3600 | |
shoaib_ahmed | 0:791a779d6220 | 3601 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 3602 | |
shoaib_ahmed | 0:791a779d6220 | 3603 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[6]); |
shoaib_ahmed | 0:791a779d6220 | 3604 | tmp1 = GETJSAMPLE(elemptr[1]) + GETJSAMPLE(elemptr[5]); |
shoaib_ahmed | 0:791a779d6220 | 3605 | tmp2 = GETJSAMPLE(elemptr[2]) + GETJSAMPLE(elemptr[4]); |
shoaib_ahmed | 0:791a779d6220 | 3606 | tmp3 = GETJSAMPLE(elemptr[3]); |
shoaib_ahmed | 0:791a779d6220 | 3607 | |
shoaib_ahmed | 0:791a779d6220 | 3608 | tmp10 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[6]); |
shoaib_ahmed | 0:791a779d6220 | 3609 | tmp11 = GETJSAMPLE(elemptr[1]) - GETJSAMPLE(elemptr[5]); |
shoaib_ahmed | 0:791a779d6220 | 3610 | tmp12 = GETJSAMPLE(elemptr[2]) - GETJSAMPLE(elemptr[4]); |
shoaib_ahmed | 0:791a779d6220 | 3611 | |
shoaib_ahmed | 0:791a779d6220 | 3612 | z1 = tmp0 + tmp2; |
shoaib_ahmed | 0:791a779d6220 | 3613 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 3614 | dataptr[0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3615 | ((z1 + tmp1 + tmp3 - 7 * CENTERJSAMPLE) << PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3616 | tmp3 += tmp3; |
shoaib_ahmed | 0:791a779d6220 | 3617 | z1 -= tmp3; |
shoaib_ahmed | 0:791a779d6220 | 3618 | z1 -= tmp3; |
shoaib_ahmed | 0:791a779d6220 | 3619 | z1 = MULTIPLY(z1, FIX(0.353553391)); /* (c2+c6-c4)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 3620 | z2 = MULTIPLY(tmp0 - tmp2, FIX(0.920609002)); /* (c2+c4-c6)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 3621 | z3 = MULTIPLY(tmp1 - tmp2, FIX(0.314692123)); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 3622 | dataptr[2] = (DCTELEM) DESCALE(z1 + z2 + z3, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3623 | z1 -= z2; |
shoaib_ahmed | 0:791a779d6220 | 3624 | z2 = MULTIPLY(tmp0 - tmp1, FIX(0.881747734)); /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 3625 | dataptr[4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3626 | DESCALE(z2 + z3 - MULTIPLY(tmp1 - tmp3, FIX(0.707106781)), /* c2+c6-c4 */ |
shoaib_ahmed | 0:791a779d6220 | 3627 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3628 | dataptr[6] = (DCTELEM) DESCALE(z1 + z2, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3629 | |
shoaib_ahmed | 0:791a779d6220 | 3630 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 3631 | |
shoaib_ahmed | 0:791a779d6220 | 3632 | tmp1 = MULTIPLY(tmp10 + tmp11, FIX(0.935414347)); /* (c3+c1-c5)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 3633 | tmp2 = MULTIPLY(tmp10 - tmp11, FIX(0.170262339)); /* (c3+c5-c1)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 3634 | tmp0 = tmp1 - tmp2; |
shoaib_ahmed | 0:791a779d6220 | 3635 | tmp1 += tmp2; |
shoaib_ahmed | 0:791a779d6220 | 3636 | tmp2 = MULTIPLY(tmp11 + tmp12, - FIX(1.378756276)); /* -c1 */ |
shoaib_ahmed | 0:791a779d6220 | 3637 | tmp1 += tmp2; |
shoaib_ahmed | 0:791a779d6220 | 3638 | tmp3 = MULTIPLY(tmp10 + tmp12, FIX(0.613604268)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 3639 | tmp0 += tmp3; |
shoaib_ahmed | 0:791a779d6220 | 3640 | tmp2 += tmp3 + MULTIPLY(tmp12, FIX(1.870828693)); /* c3+c1-c5 */ |
shoaib_ahmed | 0:791a779d6220 | 3641 | |
shoaib_ahmed | 0:791a779d6220 | 3642 | dataptr[1] = (DCTELEM) DESCALE(tmp0, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3643 | dataptr[3] = (DCTELEM) DESCALE(tmp1, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3644 | dataptr[5] = (DCTELEM) DESCALE(tmp2, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3645 | |
shoaib_ahmed | 0:791a779d6220 | 3646 | ctr++; |
shoaib_ahmed | 0:791a779d6220 | 3647 | |
shoaib_ahmed | 0:791a779d6220 | 3648 | if (ctr != DCTSIZE) { |
shoaib_ahmed | 0:791a779d6220 | 3649 | if (ctr == 14) |
shoaib_ahmed | 0:791a779d6220 | 3650 | break; /* Done. */ |
shoaib_ahmed | 0:791a779d6220 | 3651 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 3652 | } else |
shoaib_ahmed | 0:791a779d6220 | 3653 | dataptr = workspace; /* switch pointer to extended workspace */ |
shoaib_ahmed | 0:791a779d6220 | 3654 | } |
shoaib_ahmed | 0:791a779d6220 | 3655 | |
shoaib_ahmed | 0:791a779d6220 | 3656 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 3657 | * We remove the PASS1_BITS scaling, but leave the results scaled up |
shoaib_ahmed | 0:791a779d6220 | 3658 | * by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 3659 | * We must also scale the output by (8/7)*(8/14) = 32/49, which we |
shoaib_ahmed | 0:791a779d6220 | 3660 | * fold into the constant multipliers: |
shoaib_ahmed | 0:791a779d6220 | 3661 | * 14-point FDCT kernel, cK represents sqrt(2) * cos(K*pi/28) * 32/49. |
shoaib_ahmed | 0:791a779d6220 | 3662 | */ |
shoaib_ahmed | 0:791a779d6220 | 3663 | |
shoaib_ahmed | 0:791a779d6220 | 3664 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 3665 | wsptr = workspace; |
shoaib_ahmed | 0:791a779d6220 | 3666 | for (ctr = 0; ctr < 7; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 3667 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 3668 | |
shoaib_ahmed | 0:791a779d6220 | 3669 | tmp0 = dataptr[DCTSIZE*0] + wsptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 3670 | tmp1 = dataptr[DCTSIZE*1] + wsptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 3671 | tmp2 = dataptr[DCTSIZE*2] + wsptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 3672 | tmp13 = dataptr[DCTSIZE*3] + wsptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 3673 | tmp4 = dataptr[DCTSIZE*4] + wsptr[DCTSIZE*1]; |
shoaib_ahmed | 0:791a779d6220 | 3674 | tmp5 = dataptr[DCTSIZE*5] + wsptr[DCTSIZE*0]; |
shoaib_ahmed | 0:791a779d6220 | 3675 | tmp6 = dataptr[DCTSIZE*6] + dataptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 3676 | |
shoaib_ahmed | 0:791a779d6220 | 3677 | tmp10 = tmp0 + tmp6; |
shoaib_ahmed | 0:791a779d6220 | 3678 | tmp14 = tmp0 - tmp6; |
shoaib_ahmed | 0:791a779d6220 | 3679 | tmp11 = tmp1 + tmp5; |
shoaib_ahmed | 0:791a779d6220 | 3680 | tmp15 = tmp1 - tmp5; |
shoaib_ahmed | 0:791a779d6220 | 3681 | tmp12 = tmp2 + tmp4; |
shoaib_ahmed | 0:791a779d6220 | 3682 | tmp16 = tmp2 - tmp4; |
shoaib_ahmed | 0:791a779d6220 | 3683 | |
shoaib_ahmed | 0:791a779d6220 | 3684 | tmp0 = dataptr[DCTSIZE*0] - wsptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 3685 | tmp1 = dataptr[DCTSIZE*1] - wsptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 3686 | tmp2 = dataptr[DCTSIZE*2] - wsptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 3687 | tmp3 = dataptr[DCTSIZE*3] - wsptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 3688 | tmp4 = dataptr[DCTSIZE*4] - wsptr[DCTSIZE*1]; |
shoaib_ahmed | 0:791a779d6220 | 3689 | tmp5 = dataptr[DCTSIZE*5] - wsptr[DCTSIZE*0]; |
shoaib_ahmed | 0:791a779d6220 | 3690 | tmp6 = dataptr[DCTSIZE*6] - dataptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 3691 | |
shoaib_ahmed | 0:791a779d6220 | 3692 | dataptr[DCTSIZE*0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3693 | DESCALE(MULTIPLY(tmp10 + tmp11 + tmp12 + tmp13, |
shoaib_ahmed | 0:791a779d6220 | 3694 | FIX(0.653061224)), /* 32/49 */ |
shoaib_ahmed | 0:791a779d6220 | 3695 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3696 | tmp13 += tmp13; |
shoaib_ahmed | 0:791a779d6220 | 3697 | dataptr[DCTSIZE*4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3698 | DESCALE(MULTIPLY(tmp10 - tmp13, FIX(0.832106052)) + /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 3699 | MULTIPLY(tmp11 - tmp13, FIX(0.205513223)) - /* c12 */ |
shoaib_ahmed | 0:791a779d6220 | 3700 | MULTIPLY(tmp12 - tmp13, FIX(0.575835255)), /* c8 */ |
shoaib_ahmed | 0:791a779d6220 | 3701 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3702 | |
shoaib_ahmed | 0:791a779d6220 | 3703 | tmp10 = MULTIPLY(tmp14 + tmp15, FIX(0.722074570)); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 3704 | |
shoaib_ahmed | 0:791a779d6220 | 3705 | dataptr[DCTSIZE*2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3706 | DESCALE(tmp10 + MULTIPLY(tmp14, FIX(0.178337691)) /* c2-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 3707 | + MULTIPLY(tmp16, FIX(0.400721155)), /* c10 */ |
shoaib_ahmed | 0:791a779d6220 | 3708 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3709 | dataptr[DCTSIZE*6] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3710 | DESCALE(tmp10 - MULTIPLY(tmp15, FIX(1.122795725)) /* c6+c10 */ |
shoaib_ahmed | 0:791a779d6220 | 3711 | - MULTIPLY(tmp16, FIX(0.900412262)), /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 3712 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3713 | |
shoaib_ahmed | 0:791a779d6220 | 3714 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 3715 | |
shoaib_ahmed | 0:791a779d6220 | 3716 | tmp10 = tmp1 + tmp2; |
shoaib_ahmed | 0:791a779d6220 | 3717 | tmp11 = tmp5 - tmp4; |
shoaib_ahmed | 0:791a779d6220 | 3718 | dataptr[DCTSIZE*7] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3719 | DESCALE(MULTIPLY(tmp0 - tmp10 + tmp3 - tmp11 - tmp6, |
shoaib_ahmed | 0:791a779d6220 | 3720 | FIX(0.653061224)), /* 32/49 */ |
shoaib_ahmed | 0:791a779d6220 | 3721 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3722 | tmp3 = MULTIPLY(tmp3 , FIX(0.653061224)); /* 32/49 */ |
shoaib_ahmed | 0:791a779d6220 | 3723 | tmp10 = MULTIPLY(tmp10, - FIX(0.103406812)); /* -c13 */ |
shoaib_ahmed | 0:791a779d6220 | 3724 | tmp11 = MULTIPLY(tmp11, FIX(0.917760839)); /* c1 */ |
shoaib_ahmed | 0:791a779d6220 | 3725 | tmp10 += tmp11 - tmp3; |
shoaib_ahmed | 0:791a779d6220 | 3726 | tmp11 = MULTIPLY(tmp0 + tmp2, FIX(0.782007410)) + /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 3727 | MULTIPLY(tmp4 + tmp6, FIX(0.491367823)); /* c9 */ |
shoaib_ahmed | 0:791a779d6220 | 3728 | dataptr[DCTSIZE*5] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3729 | DESCALE(tmp10 + tmp11 - MULTIPLY(tmp2, FIX(1.550341076)) /* c3+c5-c13 */ |
shoaib_ahmed | 0:791a779d6220 | 3730 | + MULTIPLY(tmp4, FIX(0.731428202)), /* c1+c11-c9 */ |
shoaib_ahmed | 0:791a779d6220 | 3731 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3732 | tmp12 = MULTIPLY(tmp0 + tmp1, FIX(0.871740478)) + /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 3733 | MULTIPLY(tmp5 - tmp6, FIX(0.305035186)); /* c11 */ |
shoaib_ahmed | 0:791a779d6220 | 3734 | dataptr[DCTSIZE*3] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3735 | DESCALE(tmp10 + tmp12 - MULTIPLY(tmp1, FIX(0.276965844)) /* c3-c9-c13 */ |
shoaib_ahmed | 0:791a779d6220 | 3736 | - MULTIPLY(tmp5, FIX(2.004803435)), /* c1+c5+c11 */ |
shoaib_ahmed | 0:791a779d6220 | 3737 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3738 | dataptr[DCTSIZE*1] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3739 | DESCALE(tmp11 + tmp12 + tmp3 |
shoaib_ahmed | 0:791a779d6220 | 3740 | - MULTIPLY(tmp0, FIX(0.735987049)) /* c3+c5-c1 */ |
shoaib_ahmed | 0:791a779d6220 | 3741 | - MULTIPLY(tmp6, FIX(0.082925825)), /* c9-c11-c13 */ |
shoaib_ahmed | 0:791a779d6220 | 3742 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3743 | |
shoaib_ahmed | 0:791a779d6220 | 3744 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 3745 | wsptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 3746 | } |
shoaib_ahmed | 0:791a779d6220 | 3747 | } |
shoaib_ahmed | 0:791a779d6220 | 3748 | |
shoaib_ahmed | 0:791a779d6220 | 3749 | |
shoaib_ahmed | 0:791a779d6220 | 3750 | /* |
shoaib_ahmed | 0:791a779d6220 | 3751 | * Perform the forward DCT on a 6x12 sample block. |
shoaib_ahmed | 0:791a779d6220 | 3752 | * |
shoaib_ahmed | 0:791a779d6220 | 3753 | * 6-point FDCT in pass 1 (rows), 12-point in pass 2 (columns). |
shoaib_ahmed | 0:791a779d6220 | 3754 | */ |
shoaib_ahmed | 0:791a779d6220 | 3755 | |
shoaib_ahmed | 0:791a779d6220 | 3756 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 3757 | jpeg_fdct_6x12 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 3758 | { |
shoaib_ahmed | 0:791a779d6220 | 3759 | INT32 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5; |
shoaib_ahmed | 0:791a779d6220 | 3760 | INT32 tmp10, tmp11, tmp12, tmp13, tmp14, tmp15; |
shoaib_ahmed | 0:791a779d6220 | 3761 | DCTELEM workspace[8*4]; |
shoaib_ahmed | 0:791a779d6220 | 3762 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 3763 | DCTELEM *wsptr; |
shoaib_ahmed | 0:791a779d6220 | 3764 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 3765 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 3766 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 3767 | |
shoaib_ahmed | 0:791a779d6220 | 3768 | /* Pre-zero output coefficient block. */ |
shoaib_ahmed | 0:791a779d6220 | 3769 | MEMZERO(data, SIZEOF(DCTELEM) * DCTSIZE2); |
shoaib_ahmed | 0:791a779d6220 | 3770 | |
shoaib_ahmed | 0:791a779d6220 | 3771 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 3772 | * Note results are scaled up by sqrt(8) compared to a true DCT; |
shoaib_ahmed | 0:791a779d6220 | 3773 | * furthermore, we scale the results by 2**PASS1_BITS. |
shoaib_ahmed | 0:791a779d6220 | 3774 | * 6-point FDCT kernel, cK represents sqrt(2) * cos(K*pi/12). |
shoaib_ahmed | 0:791a779d6220 | 3775 | */ |
shoaib_ahmed | 0:791a779d6220 | 3776 | |
shoaib_ahmed | 0:791a779d6220 | 3777 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 3778 | ctr = 0; |
shoaib_ahmed | 0:791a779d6220 | 3779 | for (;;) { |
shoaib_ahmed | 0:791a779d6220 | 3780 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 3781 | |
shoaib_ahmed | 0:791a779d6220 | 3782 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 3783 | |
shoaib_ahmed | 0:791a779d6220 | 3784 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[5]); |
shoaib_ahmed | 0:791a779d6220 | 3785 | tmp11 = GETJSAMPLE(elemptr[1]) + GETJSAMPLE(elemptr[4]); |
shoaib_ahmed | 0:791a779d6220 | 3786 | tmp2 = GETJSAMPLE(elemptr[2]) + GETJSAMPLE(elemptr[3]); |
shoaib_ahmed | 0:791a779d6220 | 3787 | |
shoaib_ahmed | 0:791a779d6220 | 3788 | tmp10 = tmp0 + tmp2; |
shoaib_ahmed | 0:791a779d6220 | 3789 | tmp12 = tmp0 - tmp2; |
shoaib_ahmed | 0:791a779d6220 | 3790 | |
shoaib_ahmed | 0:791a779d6220 | 3791 | tmp0 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[5]); |
shoaib_ahmed | 0:791a779d6220 | 3792 | tmp1 = GETJSAMPLE(elemptr[1]) - GETJSAMPLE(elemptr[4]); |
shoaib_ahmed | 0:791a779d6220 | 3793 | tmp2 = GETJSAMPLE(elemptr[2]) - GETJSAMPLE(elemptr[3]); |
shoaib_ahmed | 0:791a779d6220 | 3794 | |
shoaib_ahmed | 0:791a779d6220 | 3795 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 3796 | dataptr[0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3797 | ((tmp10 + tmp11 - 6 * CENTERJSAMPLE) << PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3798 | dataptr[2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3799 | DESCALE(MULTIPLY(tmp12, FIX(1.224744871)), /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 3800 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3801 | dataptr[4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3802 | DESCALE(MULTIPLY(tmp10 - tmp11 - tmp11, FIX(0.707106781)), /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 3803 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3804 | |
shoaib_ahmed | 0:791a779d6220 | 3805 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 3806 | |
shoaib_ahmed | 0:791a779d6220 | 3807 | tmp10 = DESCALE(MULTIPLY(tmp0 + tmp2, FIX(0.366025404)), /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 3808 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3809 | |
shoaib_ahmed | 0:791a779d6220 | 3810 | dataptr[1] = (DCTELEM) (tmp10 + ((tmp0 + tmp1) << PASS1_BITS)); |
shoaib_ahmed | 0:791a779d6220 | 3811 | dataptr[3] = (DCTELEM) ((tmp0 - tmp1 - tmp2) << PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3812 | dataptr[5] = (DCTELEM) (tmp10 + ((tmp2 - tmp1) << PASS1_BITS)); |
shoaib_ahmed | 0:791a779d6220 | 3813 | |
shoaib_ahmed | 0:791a779d6220 | 3814 | ctr++; |
shoaib_ahmed | 0:791a779d6220 | 3815 | |
shoaib_ahmed | 0:791a779d6220 | 3816 | if (ctr != DCTSIZE) { |
shoaib_ahmed | 0:791a779d6220 | 3817 | if (ctr == 12) |
shoaib_ahmed | 0:791a779d6220 | 3818 | break; /* Done. */ |
shoaib_ahmed | 0:791a779d6220 | 3819 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 3820 | } else |
shoaib_ahmed | 0:791a779d6220 | 3821 | dataptr = workspace; /* switch pointer to extended workspace */ |
shoaib_ahmed | 0:791a779d6220 | 3822 | } |
shoaib_ahmed | 0:791a779d6220 | 3823 | |
shoaib_ahmed | 0:791a779d6220 | 3824 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 3825 | * We remove the PASS1_BITS scaling, but leave the results scaled up |
shoaib_ahmed | 0:791a779d6220 | 3826 | * by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 3827 | * We must also scale the output by (8/6)*(8/12) = 8/9, which we |
shoaib_ahmed | 0:791a779d6220 | 3828 | * fold into the constant multipliers: |
shoaib_ahmed | 0:791a779d6220 | 3829 | * 12-point FDCT kernel, cK represents sqrt(2) * cos(K*pi/24) * 8/9. |
shoaib_ahmed | 0:791a779d6220 | 3830 | */ |
shoaib_ahmed | 0:791a779d6220 | 3831 | |
shoaib_ahmed | 0:791a779d6220 | 3832 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 3833 | wsptr = workspace; |
shoaib_ahmed | 0:791a779d6220 | 3834 | for (ctr = 0; ctr < 6; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 3835 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 3836 | |
shoaib_ahmed | 0:791a779d6220 | 3837 | tmp0 = dataptr[DCTSIZE*0] + wsptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 3838 | tmp1 = dataptr[DCTSIZE*1] + wsptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 3839 | tmp2 = dataptr[DCTSIZE*2] + wsptr[DCTSIZE*1]; |
shoaib_ahmed | 0:791a779d6220 | 3840 | tmp3 = dataptr[DCTSIZE*3] + wsptr[DCTSIZE*0]; |
shoaib_ahmed | 0:791a779d6220 | 3841 | tmp4 = dataptr[DCTSIZE*4] + dataptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 3842 | tmp5 = dataptr[DCTSIZE*5] + dataptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 3843 | |
shoaib_ahmed | 0:791a779d6220 | 3844 | tmp10 = tmp0 + tmp5; |
shoaib_ahmed | 0:791a779d6220 | 3845 | tmp13 = tmp0 - tmp5; |
shoaib_ahmed | 0:791a779d6220 | 3846 | tmp11 = tmp1 + tmp4; |
shoaib_ahmed | 0:791a779d6220 | 3847 | tmp14 = tmp1 - tmp4; |
shoaib_ahmed | 0:791a779d6220 | 3848 | tmp12 = tmp2 + tmp3; |
shoaib_ahmed | 0:791a779d6220 | 3849 | tmp15 = tmp2 - tmp3; |
shoaib_ahmed | 0:791a779d6220 | 3850 | |
shoaib_ahmed | 0:791a779d6220 | 3851 | tmp0 = dataptr[DCTSIZE*0] - wsptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 3852 | tmp1 = dataptr[DCTSIZE*1] - wsptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 3853 | tmp2 = dataptr[DCTSIZE*2] - wsptr[DCTSIZE*1]; |
shoaib_ahmed | 0:791a779d6220 | 3854 | tmp3 = dataptr[DCTSIZE*3] - wsptr[DCTSIZE*0]; |
shoaib_ahmed | 0:791a779d6220 | 3855 | tmp4 = dataptr[DCTSIZE*4] - dataptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 3856 | tmp5 = dataptr[DCTSIZE*5] - dataptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 3857 | |
shoaib_ahmed | 0:791a779d6220 | 3858 | dataptr[DCTSIZE*0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3859 | DESCALE(MULTIPLY(tmp10 + tmp11 + tmp12, FIX(0.888888889)), /* 8/9 */ |
shoaib_ahmed | 0:791a779d6220 | 3860 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3861 | dataptr[DCTSIZE*6] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3862 | DESCALE(MULTIPLY(tmp13 - tmp14 - tmp15, FIX(0.888888889)), /* 8/9 */ |
shoaib_ahmed | 0:791a779d6220 | 3863 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3864 | dataptr[DCTSIZE*4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3865 | DESCALE(MULTIPLY(tmp10 - tmp12, FIX(1.088662108)), /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 3866 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3867 | dataptr[DCTSIZE*2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3868 | DESCALE(MULTIPLY(tmp14 - tmp15, FIX(0.888888889)) + /* 8/9 */ |
shoaib_ahmed | 0:791a779d6220 | 3869 | MULTIPLY(tmp13 + tmp15, FIX(1.214244803)), /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 3870 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3871 | |
shoaib_ahmed | 0:791a779d6220 | 3872 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 3873 | |
shoaib_ahmed | 0:791a779d6220 | 3874 | tmp10 = MULTIPLY(tmp1 + tmp4, FIX(0.481063200)); /* c9 */ |
shoaib_ahmed | 0:791a779d6220 | 3875 | tmp14 = tmp10 + MULTIPLY(tmp1, FIX(0.680326102)); /* c3-c9 */ |
shoaib_ahmed | 0:791a779d6220 | 3876 | tmp15 = tmp10 - MULTIPLY(tmp4, FIX(1.642452502)); /* c3+c9 */ |
shoaib_ahmed | 0:791a779d6220 | 3877 | tmp12 = MULTIPLY(tmp0 + tmp2, FIX(0.997307603)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 3878 | tmp13 = MULTIPLY(tmp0 + tmp3, FIX(0.765261039)); /* c7 */ |
shoaib_ahmed | 0:791a779d6220 | 3879 | tmp10 = tmp12 + tmp13 + tmp14 - MULTIPLY(tmp0, FIX(0.516244403)) /* c5+c7-c1 */ |
shoaib_ahmed | 0:791a779d6220 | 3880 | + MULTIPLY(tmp5, FIX(0.164081699)); /* c11 */ |
shoaib_ahmed | 0:791a779d6220 | 3881 | tmp11 = MULTIPLY(tmp2 + tmp3, - FIX(0.164081699)); /* -c11 */ |
shoaib_ahmed | 0:791a779d6220 | 3882 | tmp12 += tmp11 - tmp15 - MULTIPLY(tmp2, FIX(2.079550144)) /* c1+c5-c11 */ |
shoaib_ahmed | 0:791a779d6220 | 3883 | + MULTIPLY(tmp5, FIX(0.765261039)); /* c7 */ |
shoaib_ahmed | 0:791a779d6220 | 3884 | tmp13 += tmp11 - tmp14 + MULTIPLY(tmp3, FIX(0.645144899)) /* c1+c11-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 3885 | - MULTIPLY(tmp5, FIX(0.997307603)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 3886 | tmp11 = tmp15 + MULTIPLY(tmp0 - tmp3, FIX(1.161389302)) /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 3887 | - MULTIPLY(tmp2 + tmp5, FIX(0.481063200)); /* c9 */ |
shoaib_ahmed | 0:791a779d6220 | 3888 | |
shoaib_ahmed | 0:791a779d6220 | 3889 | dataptr[DCTSIZE*1] = (DCTELEM) DESCALE(tmp10, CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3890 | dataptr[DCTSIZE*3] = (DCTELEM) DESCALE(tmp11, CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3891 | dataptr[DCTSIZE*5] = (DCTELEM) DESCALE(tmp12, CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3892 | dataptr[DCTSIZE*7] = (DCTELEM) DESCALE(tmp13, CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3893 | |
shoaib_ahmed | 0:791a779d6220 | 3894 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 3895 | wsptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 3896 | } |
shoaib_ahmed | 0:791a779d6220 | 3897 | } |
shoaib_ahmed | 0:791a779d6220 | 3898 | |
shoaib_ahmed | 0:791a779d6220 | 3899 | |
shoaib_ahmed | 0:791a779d6220 | 3900 | /* |
shoaib_ahmed | 0:791a779d6220 | 3901 | * Perform the forward DCT on a 5x10 sample block. |
shoaib_ahmed | 0:791a779d6220 | 3902 | * |
shoaib_ahmed | 0:791a779d6220 | 3903 | * 5-point FDCT in pass 1 (rows), 10-point in pass 2 (columns). |
shoaib_ahmed | 0:791a779d6220 | 3904 | */ |
shoaib_ahmed | 0:791a779d6220 | 3905 | |
shoaib_ahmed | 0:791a779d6220 | 3906 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 3907 | jpeg_fdct_5x10 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 3908 | { |
shoaib_ahmed | 0:791a779d6220 | 3909 | INT32 tmp0, tmp1, tmp2, tmp3, tmp4; |
shoaib_ahmed | 0:791a779d6220 | 3910 | INT32 tmp10, tmp11, tmp12, tmp13, tmp14; |
shoaib_ahmed | 0:791a779d6220 | 3911 | DCTELEM workspace[8*2]; |
shoaib_ahmed | 0:791a779d6220 | 3912 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 3913 | DCTELEM *wsptr; |
shoaib_ahmed | 0:791a779d6220 | 3914 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 3915 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 3916 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 3917 | |
shoaib_ahmed | 0:791a779d6220 | 3918 | /* Pre-zero output coefficient block. */ |
shoaib_ahmed | 0:791a779d6220 | 3919 | MEMZERO(data, SIZEOF(DCTELEM) * DCTSIZE2); |
shoaib_ahmed | 0:791a779d6220 | 3920 | |
shoaib_ahmed | 0:791a779d6220 | 3921 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 3922 | * Note results are scaled up by sqrt(8) compared to a true DCT; |
shoaib_ahmed | 0:791a779d6220 | 3923 | * furthermore, we scale the results by 2**PASS1_BITS. |
shoaib_ahmed | 0:791a779d6220 | 3924 | * 5-point FDCT kernel, cK represents sqrt(2) * cos(K*pi/10). |
shoaib_ahmed | 0:791a779d6220 | 3925 | */ |
shoaib_ahmed | 0:791a779d6220 | 3926 | |
shoaib_ahmed | 0:791a779d6220 | 3927 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 3928 | ctr = 0; |
shoaib_ahmed | 0:791a779d6220 | 3929 | for (;;) { |
shoaib_ahmed | 0:791a779d6220 | 3930 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 3931 | |
shoaib_ahmed | 0:791a779d6220 | 3932 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 3933 | |
shoaib_ahmed | 0:791a779d6220 | 3934 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[4]); |
shoaib_ahmed | 0:791a779d6220 | 3935 | tmp1 = GETJSAMPLE(elemptr[1]) + GETJSAMPLE(elemptr[3]); |
shoaib_ahmed | 0:791a779d6220 | 3936 | tmp2 = GETJSAMPLE(elemptr[2]); |
shoaib_ahmed | 0:791a779d6220 | 3937 | |
shoaib_ahmed | 0:791a779d6220 | 3938 | tmp10 = tmp0 + tmp1; |
shoaib_ahmed | 0:791a779d6220 | 3939 | tmp11 = tmp0 - tmp1; |
shoaib_ahmed | 0:791a779d6220 | 3940 | |
shoaib_ahmed | 0:791a779d6220 | 3941 | tmp0 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[4]); |
shoaib_ahmed | 0:791a779d6220 | 3942 | tmp1 = GETJSAMPLE(elemptr[1]) - GETJSAMPLE(elemptr[3]); |
shoaib_ahmed | 0:791a779d6220 | 3943 | |
shoaib_ahmed | 0:791a779d6220 | 3944 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 3945 | dataptr[0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3946 | ((tmp10 + tmp2 - 5 * CENTERJSAMPLE) << PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3947 | tmp11 = MULTIPLY(tmp11, FIX(0.790569415)); /* (c2+c4)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 3948 | tmp10 -= tmp2 << 2; |
shoaib_ahmed | 0:791a779d6220 | 3949 | tmp10 = MULTIPLY(tmp10, FIX(0.353553391)); /* (c2-c4)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 3950 | dataptr[2] = (DCTELEM) DESCALE(tmp11 + tmp10, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3951 | dataptr[4] = (DCTELEM) DESCALE(tmp11 - tmp10, CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3952 | |
shoaib_ahmed | 0:791a779d6220 | 3953 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 3954 | |
shoaib_ahmed | 0:791a779d6220 | 3955 | tmp10 = MULTIPLY(tmp0 + tmp1, FIX(0.831253876)); /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 3956 | |
shoaib_ahmed | 0:791a779d6220 | 3957 | dataptr[1] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3958 | DESCALE(tmp10 + MULTIPLY(tmp0, FIX(0.513743148)), /* c1-c3 */ |
shoaib_ahmed | 0:791a779d6220 | 3959 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3960 | dataptr[3] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 3961 | DESCALE(tmp10 - MULTIPLY(tmp1, FIX(2.176250899)), /* c1+c3 */ |
shoaib_ahmed | 0:791a779d6220 | 3962 | CONST_BITS-PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 3963 | |
shoaib_ahmed | 0:791a779d6220 | 3964 | ctr++; |
shoaib_ahmed | 0:791a779d6220 | 3965 | |
shoaib_ahmed | 0:791a779d6220 | 3966 | if (ctr != DCTSIZE) { |
shoaib_ahmed | 0:791a779d6220 | 3967 | if (ctr == 10) |
shoaib_ahmed | 0:791a779d6220 | 3968 | break; /* Done. */ |
shoaib_ahmed | 0:791a779d6220 | 3969 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 3970 | } else |
shoaib_ahmed | 0:791a779d6220 | 3971 | dataptr = workspace; /* switch pointer to extended workspace */ |
shoaib_ahmed | 0:791a779d6220 | 3972 | } |
shoaib_ahmed | 0:791a779d6220 | 3973 | |
shoaib_ahmed | 0:791a779d6220 | 3974 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 3975 | * We remove the PASS1_BITS scaling, but leave the results scaled up |
shoaib_ahmed | 0:791a779d6220 | 3976 | * by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 3977 | * We must also scale the output by (8/5)*(8/10) = 32/25, which we |
shoaib_ahmed | 0:791a779d6220 | 3978 | * fold into the constant multipliers: |
shoaib_ahmed | 0:791a779d6220 | 3979 | * 10-point FDCT kernel, cK represents sqrt(2) * cos(K*pi/20) * 32/25. |
shoaib_ahmed | 0:791a779d6220 | 3980 | */ |
shoaib_ahmed | 0:791a779d6220 | 3981 | |
shoaib_ahmed | 0:791a779d6220 | 3982 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 3983 | wsptr = workspace; |
shoaib_ahmed | 0:791a779d6220 | 3984 | for (ctr = 0; ctr < 5; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 3985 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 3986 | |
shoaib_ahmed | 0:791a779d6220 | 3987 | tmp0 = dataptr[DCTSIZE*0] + wsptr[DCTSIZE*1]; |
shoaib_ahmed | 0:791a779d6220 | 3988 | tmp1 = dataptr[DCTSIZE*1] + wsptr[DCTSIZE*0]; |
shoaib_ahmed | 0:791a779d6220 | 3989 | tmp12 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 3990 | tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 3991 | tmp4 = dataptr[DCTSIZE*4] + dataptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 3992 | |
shoaib_ahmed | 0:791a779d6220 | 3993 | tmp10 = tmp0 + tmp4; |
shoaib_ahmed | 0:791a779d6220 | 3994 | tmp13 = tmp0 - tmp4; |
shoaib_ahmed | 0:791a779d6220 | 3995 | tmp11 = tmp1 + tmp3; |
shoaib_ahmed | 0:791a779d6220 | 3996 | tmp14 = tmp1 - tmp3; |
shoaib_ahmed | 0:791a779d6220 | 3997 | |
shoaib_ahmed | 0:791a779d6220 | 3998 | tmp0 = dataptr[DCTSIZE*0] - wsptr[DCTSIZE*1]; |
shoaib_ahmed | 0:791a779d6220 | 3999 | tmp1 = dataptr[DCTSIZE*1] - wsptr[DCTSIZE*0]; |
shoaib_ahmed | 0:791a779d6220 | 4000 | tmp2 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 4001 | tmp3 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 4002 | tmp4 = dataptr[DCTSIZE*4] - dataptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 4003 | |
shoaib_ahmed | 0:791a779d6220 | 4004 | dataptr[DCTSIZE*0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 4005 | DESCALE(MULTIPLY(tmp10 + tmp11 + tmp12, FIX(1.28)), /* 32/25 */ |
shoaib_ahmed | 0:791a779d6220 | 4006 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 4007 | tmp12 += tmp12; |
shoaib_ahmed | 0:791a779d6220 | 4008 | dataptr[DCTSIZE*4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 4009 | DESCALE(MULTIPLY(tmp10 - tmp12, FIX(1.464477191)) - /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 4010 | MULTIPLY(tmp11 - tmp12, FIX(0.559380511)), /* c8 */ |
shoaib_ahmed | 0:791a779d6220 | 4011 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 4012 | tmp10 = MULTIPLY(tmp13 + tmp14, FIX(1.064004961)); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 4013 | dataptr[DCTSIZE*2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 4014 | DESCALE(tmp10 + MULTIPLY(tmp13, FIX(0.657591230)), /* c2-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 4015 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 4016 | dataptr[DCTSIZE*6] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 4017 | DESCALE(tmp10 - MULTIPLY(tmp14, FIX(2.785601151)), /* c2+c6 */ |
shoaib_ahmed | 0:791a779d6220 | 4018 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 4019 | |
shoaib_ahmed | 0:791a779d6220 | 4020 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 4021 | |
shoaib_ahmed | 0:791a779d6220 | 4022 | tmp10 = tmp0 + tmp4; |
shoaib_ahmed | 0:791a779d6220 | 4023 | tmp11 = tmp1 - tmp3; |
shoaib_ahmed | 0:791a779d6220 | 4024 | dataptr[DCTSIZE*5] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 4025 | DESCALE(MULTIPLY(tmp10 - tmp11 - tmp2, FIX(1.28)), /* 32/25 */ |
shoaib_ahmed | 0:791a779d6220 | 4026 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 4027 | tmp2 = MULTIPLY(tmp2, FIX(1.28)); /* 32/25 */ |
shoaib_ahmed | 0:791a779d6220 | 4028 | dataptr[DCTSIZE*1] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 4029 | DESCALE(MULTIPLY(tmp0, FIX(1.787906876)) + /* c1 */ |
shoaib_ahmed | 0:791a779d6220 | 4030 | MULTIPLY(tmp1, FIX(1.612894094)) + tmp2 + /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 4031 | MULTIPLY(tmp3, FIX(0.821810588)) + /* c7 */ |
shoaib_ahmed | 0:791a779d6220 | 4032 | MULTIPLY(tmp4, FIX(0.283176630)), /* c9 */ |
shoaib_ahmed | 0:791a779d6220 | 4033 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 4034 | tmp12 = MULTIPLY(tmp0 - tmp4, FIX(1.217352341)) - /* (c3+c7)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 4035 | MULTIPLY(tmp1 + tmp3, FIX(0.752365123)); /* (c1-c9)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 4036 | tmp13 = MULTIPLY(tmp10 + tmp11, FIX(0.395541753)) + /* (c3-c7)/2 */ |
shoaib_ahmed | 0:791a779d6220 | 4037 | MULTIPLY(tmp11, FIX(0.64)) - tmp2; /* 16/25 */ |
shoaib_ahmed | 0:791a779d6220 | 4038 | dataptr[DCTSIZE*3] = (DCTELEM) DESCALE(tmp12 + tmp13, CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 4039 | dataptr[DCTSIZE*7] = (DCTELEM) DESCALE(tmp12 - tmp13, CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 4040 | |
shoaib_ahmed | 0:791a779d6220 | 4041 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 4042 | wsptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 4043 | } |
shoaib_ahmed | 0:791a779d6220 | 4044 | } |
shoaib_ahmed | 0:791a779d6220 | 4045 | |
shoaib_ahmed | 0:791a779d6220 | 4046 | |
shoaib_ahmed | 0:791a779d6220 | 4047 | /* |
shoaib_ahmed | 0:791a779d6220 | 4048 | * Perform the forward DCT on a 4x8 sample block. |
shoaib_ahmed | 0:791a779d6220 | 4049 | * |
shoaib_ahmed | 0:791a779d6220 | 4050 | * 4-point FDCT in pass 1 (rows), 8-point in pass 2 (columns). |
shoaib_ahmed | 0:791a779d6220 | 4051 | */ |
shoaib_ahmed | 0:791a779d6220 | 4052 | |
shoaib_ahmed | 0:791a779d6220 | 4053 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 4054 | jpeg_fdct_4x8 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 4055 | { |
shoaib_ahmed | 0:791a779d6220 | 4056 | INT32 tmp0, tmp1, tmp2, tmp3; |
shoaib_ahmed | 0:791a779d6220 | 4057 | INT32 tmp10, tmp11, tmp12, tmp13; |
shoaib_ahmed | 0:791a779d6220 | 4058 | INT32 z1; |
shoaib_ahmed | 0:791a779d6220 | 4059 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 4060 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 4061 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 4062 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 4063 | |
shoaib_ahmed | 0:791a779d6220 | 4064 | /* Pre-zero output coefficient block. */ |
shoaib_ahmed | 0:791a779d6220 | 4065 | MEMZERO(data, SIZEOF(DCTELEM) * DCTSIZE2); |
shoaib_ahmed | 0:791a779d6220 | 4066 | |
shoaib_ahmed | 0:791a779d6220 | 4067 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 4068 | * Note results are scaled up by sqrt(8) compared to a true DCT; |
shoaib_ahmed | 0:791a779d6220 | 4069 | * furthermore, we scale the results by 2**PASS1_BITS. |
shoaib_ahmed | 0:791a779d6220 | 4070 | * We must also scale the output by 8/4 = 2, which we add here. |
shoaib_ahmed | 0:791a779d6220 | 4071 | * 4-point FDCT kernel, |
shoaib_ahmed | 0:791a779d6220 | 4072 | * cK represents sqrt(2) * cos(K*pi/16) [refers to 8-point FDCT]. |
shoaib_ahmed | 0:791a779d6220 | 4073 | */ |
shoaib_ahmed | 0:791a779d6220 | 4074 | |
shoaib_ahmed | 0:791a779d6220 | 4075 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 4076 | for (ctr = 0; ctr < DCTSIZE; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 4077 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 4078 | |
shoaib_ahmed | 0:791a779d6220 | 4079 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 4080 | |
shoaib_ahmed | 0:791a779d6220 | 4081 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[3]); |
shoaib_ahmed | 0:791a779d6220 | 4082 | tmp1 = GETJSAMPLE(elemptr[1]) + GETJSAMPLE(elemptr[2]); |
shoaib_ahmed | 0:791a779d6220 | 4083 | |
shoaib_ahmed | 0:791a779d6220 | 4084 | tmp10 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[3]); |
shoaib_ahmed | 0:791a779d6220 | 4085 | tmp11 = GETJSAMPLE(elemptr[1]) - GETJSAMPLE(elemptr[2]); |
shoaib_ahmed | 0:791a779d6220 | 4086 | |
shoaib_ahmed | 0:791a779d6220 | 4087 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 4088 | dataptr[0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 4089 | ((tmp0 + tmp1 - 4 * CENTERJSAMPLE) << (PASS1_BITS+1)); |
shoaib_ahmed | 0:791a779d6220 | 4090 | dataptr[2] = (DCTELEM) ((tmp0 - tmp1) << (PASS1_BITS+1)); |
shoaib_ahmed | 0:791a779d6220 | 4091 | |
shoaib_ahmed | 0:791a779d6220 | 4092 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 4093 | |
shoaib_ahmed | 0:791a779d6220 | 4094 | tmp0 = MULTIPLY(tmp10 + tmp11, FIX_0_541196100); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 4095 | /* Add fudge factor here for final descale. */ |
shoaib_ahmed | 0:791a779d6220 | 4096 | tmp0 += ONE << (CONST_BITS-PASS1_BITS-2); |
shoaib_ahmed | 0:791a779d6220 | 4097 | |
shoaib_ahmed | 0:791a779d6220 | 4098 | dataptr[1] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 4099 | RIGHT_SHIFT(tmp0 + MULTIPLY(tmp10, FIX_0_765366865), /* c2-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 4100 | CONST_BITS-PASS1_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 4101 | dataptr[3] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 4102 | RIGHT_SHIFT(tmp0 - MULTIPLY(tmp11, FIX_1_847759065), /* c2+c6 */ |
shoaib_ahmed | 0:791a779d6220 | 4103 | CONST_BITS-PASS1_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 4104 | |
shoaib_ahmed | 0:791a779d6220 | 4105 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 4106 | } |
shoaib_ahmed | 0:791a779d6220 | 4107 | |
shoaib_ahmed | 0:791a779d6220 | 4108 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 4109 | * We remove the PASS1_BITS scaling, but leave the results scaled up |
shoaib_ahmed | 0:791a779d6220 | 4110 | * by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 4111 | * 8-point FDCT kernel, cK represents sqrt(2) * cos(K*pi/16). |
shoaib_ahmed | 0:791a779d6220 | 4112 | */ |
shoaib_ahmed | 0:791a779d6220 | 4113 | |
shoaib_ahmed | 0:791a779d6220 | 4114 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 4115 | for (ctr = 0; ctr < 4; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 4116 | /* Even part per LL&M figure 1 --- note that published figure is faulty; |
shoaib_ahmed | 0:791a779d6220 | 4117 | * rotator "c1" should be "c6". |
shoaib_ahmed | 0:791a779d6220 | 4118 | */ |
shoaib_ahmed | 0:791a779d6220 | 4119 | |
shoaib_ahmed | 0:791a779d6220 | 4120 | tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 4121 | tmp1 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 4122 | tmp2 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 4123 | tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 4124 | |
shoaib_ahmed | 0:791a779d6220 | 4125 | /* Add fudge factor here for final descale. */ |
shoaib_ahmed | 0:791a779d6220 | 4126 | tmp10 = tmp0 + tmp3 + (ONE << (PASS1_BITS-1)); |
shoaib_ahmed | 0:791a779d6220 | 4127 | tmp12 = tmp0 - tmp3; |
shoaib_ahmed | 0:791a779d6220 | 4128 | tmp11 = tmp1 + tmp2; |
shoaib_ahmed | 0:791a779d6220 | 4129 | tmp13 = tmp1 - tmp2; |
shoaib_ahmed | 0:791a779d6220 | 4130 | |
shoaib_ahmed | 0:791a779d6220 | 4131 | tmp0 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*7]; |
shoaib_ahmed | 0:791a779d6220 | 4132 | tmp1 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*6]; |
shoaib_ahmed | 0:791a779d6220 | 4133 | tmp2 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 4134 | tmp3 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 4135 | |
shoaib_ahmed | 0:791a779d6220 | 4136 | dataptr[DCTSIZE*0] = (DCTELEM) RIGHT_SHIFT(tmp10 + tmp11, PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 4137 | dataptr[DCTSIZE*4] = (DCTELEM) RIGHT_SHIFT(tmp10 - tmp11, PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 4138 | |
shoaib_ahmed | 0:791a779d6220 | 4139 | z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 4140 | /* Add fudge factor here for final descale. */ |
shoaib_ahmed | 0:791a779d6220 | 4141 | z1 += ONE << (CONST_BITS+PASS1_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 4142 | |
shoaib_ahmed | 0:791a779d6220 | 4143 | dataptr[DCTSIZE*2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 4144 | RIGHT_SHIFT(z1 + MULTIPLY(tmp12, FIX_0_765366865), /* c2-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 4145 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 4146 | dataptr[DCTSIZE*6] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 4147 | RIGHT_SHIFT(z1 - MULTIPLY(tmp13, FIX_1_847759065), /* c2+c6 */ |
shoaib_ahmed | 0:791a779d6220 | 4148 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 4149 | |
shoaib_ahmed | 0:791a779d6220 | 4150 | /* Odd part per figure 8 --- note paper omits factor of sqrt(2). |
shoaib_ahmed | 0:791a779d6220 | 4151 | * i0..i3 in the paper are tmp0..tmp3 here. |
shoaib_ahmed | 0:791a779d6220 | 4152 | */ |
shoaib_ahmed | 0:791a779d6220 | 4153 | |
shoaib_ahmed | 0:791a779d6220 | 4154 | tmp12 = tmp0 + tmp2; |
shoaib_ahmed | 0:791a779d6220 | 4155 | tmp13 = tmp1 + tmp3; |
shoaib_ahmed | 0:791a779d6220 | 4156 | |
shoaib_ahmed | 0:791a779d6220 | 4157 | z1 = MULTIPLY(tmp12 + tmp13, FIX_1_175875602); /* c3 */ |
shoaib_ahmed | 0:791a779d6220 | 4158 | /* Add fudge factor here for final descale. */ |
shoaib_ahmed | 0:791a779d6220 | 4159 | z1 += ONE << (CONST_BITS+PASS1_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 4160 | |
shoaib_ahmed | 0:791a779d6220 | 4161 | tmp12 = MULTIPLY(tmp12, - FIX_0_390180644); /* -c3+c5 */ |
shoaib_ahmed | 0:791a779d6220 | 4162 | tmp13 = MULTIPLY(tmp13, - FIX_1_961570560); /* -c3-c5 */ |
shoaib_ahmed | 0:791a779d6220 | 4163 | tmp12 += z1; |
shoaib_ahmed | 0:791a779d6220 | 4164 | tmp13 += z1; |
shoaib_ahmed | 0:791a779d6220 | 4165 | |
shoaib_ahmed | 0:791a779d6220 | 4166 | z1 = MULTIPLY(tmp0 + tmp3, - FIX_0_899976223); /* -c3+c7 */ |
shoaib_ahmed | 0:791a779d6220 | 4167 | tmp0 = MULTIPLY(tmp0, FIX_1_501321110); /* c1+c3-c5-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 4168 | tmp3 = MULTIPLY(tmp3, FIX_0_298631336); /* -c1+c3+c5-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 4169 | tmp0 += z1 + tmp12; |
shoaib_ahmed | 0:791a779d6220 | 4170 | tmp3 += z1 + tmp13; |
shoaib_ahmed | 0:791a779d6220 | 4171 | |
shoaib_ahmed | 0:791a779d6220 | 4172 | z1 = MULTIPLY(tmp1 + tmp2, - FIX_2_562915447); /* -c1-c3 */ |
shoaib_ahmed | 0:791a779d6220 | 4173 | tmp1 = MULTIPLY(tmp1, FIX_3_072711026); /* c1+c3+c5-c7 */ |
shoaib_ahmed | 0:791a779d6220 | 4174 | tmp2 = MULTIPLY(tmp2, FIX_2_053119869); /* c1+c3-c5+c7 */ |
shoaib_ahmed | 0:791a779d6220 | 4175 | tmp1 += z1 + tmp13; |
shoaib_ahmed | 0:791a779d6220 | 4176 | tmp2 += z1 + tmp12; |
shoaib_ahmed | 0:791a779d6220 | 4177 | |
shoaib_ahmed | 0:791a779d6220 | 4178 | dataptr[DCTSIZE*1] = (DCTELEM) RIGHT_SHIFT(tmp0, CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 4179 | dataptr[DCTSIZE*3] = (DCTELEM) RIGHT_SHIFT(tmp1, CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 4180 | dataptr[DCTSIZE*5] = (DCTELEM) RIGHT_SHIFT(tmp2, CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 4181 | dataptr[DCTSIZE*7] = (DCTELEM) RIGHT_SHIFT(tmp3, CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 4182 | |
shoaib_ahmed | 0:791a779d6220 | 4183 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 4184 | } |
shoaib_ahmed | 0:791a779d6220 | 4185 | } |
shoaib_ahmed | 0:791a779d6220 | 4186 | |
shoaib_ahmed | 0:791a779d6220 | 4187 | |
shoaib_ahmed | 0:791a779d6220 | 4188 | /* |
shoaib_ahmed | 0:791a779d6220 | 4189 | * Perform the forward DCT on a 3x6 sample block. |
shoaib_ahmed | 0:791a779d6220 | 4190 | * |
shoaib_ahmed | 0:791a779d6220 | 4191 | * 3-point FDCT in pass 1 (rows), 6-point in pass 2 (columns). |
shoaib_ahmed | 0:791a779d6220 | 4192 | */ |
shoaib_ahmed | 0:791a779d6220 | 4193 | |
shoaib_ahmed | 0:791a779d6220 | 4194 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 4195 | jpeg_fdct_3x6 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 4196 | { |
shoaib_ahmed | 0:791a779d6220 | 4197 | INT32 tmp0, tmp1, tmp2; |
shoaib_ahmed | 0:791a779d6220 | 4198 | INT32 tmp10, tmp11, tmp12; |
shoaib_ahmed | 0:791a779d6220 | 4199 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 4200 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 4201 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 4202 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 4203 | |
shoaib_ahmed | 0:791a779d6220 | 4204 | /* Pre-zero output coefficient block. */ |
shoaib_ahmed | 0:791a779d6220 | 4205 | MEMZERO(data, SIZEOF(DCTELEM) * DCTSIZE2); |
shoaib_ahmed | 0:791a779d6220 | 4206 | |
shoaib_ahmed | 0:791a779d6220 | 4207 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 4208 | * Note results are scaled up by sqrt(8) compared to a true DCT; |
shoaib_ahmed | 0:791a779d6220 | 4209 | * furthermore, we scale the results by 2**PASS1_BITS. |
shoaib_ahmed | 0:791a779d6220 | 4210 | * We scale the results further by 2 as part of output adaption |
shoaib_ahmed | 0:791a779d6220 | 4211 | * scaling for different DCT size. |
shoaib_ahmed | 0:791a779d6220 | 4212 | * 3-point FDCT kernel, cK represents sqrt(2) * cos(K*pi/6). |
shoaib_ahmed | 0:791a779d6220 | 4213 | */ |
shoaib_ahmed | 0:791a779d6220 | 4214 | |
shoaib_ahmed | 0:791a779d6220 | 4215 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 4216 | for (ctr = 0; ctr < 6; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 4217 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 4218 | |
shoaib_ahmed | 0:791a779d6220 | 4219 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 4220 | |
shoaib_ahmed | 0:791a779d6220 | 4221 | tmp0 = GETJSAMPLE(elemptr[0]) + GETJSAMPLE(elemptr[2]); |
shoaib_ahmed | 0:791a779d6220 | 4222 | tmp1 = GETJSAMPLE(elemptr[1]); |
shoaib_ahmed | 0:791a779d6220 | 4223 | |
shoaib_ahmed | 0:791a779d6220 | 4224 | tmp2 = GETJSAMPLE(elemptr[0]) - GETJSAMPLE(elemptr[2]); |
shoaib_ahmed | 0:791a779d6220 | 4225 | |
shoaib_ahmed | 0:791a779d6220 | 4226 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 4227 | dataptr[0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 4228 | ((tmp0 + tmp1 - 3 * CENTERJSAMPLE) << (PASS1_BITS+1)); |
shoaib_ahmed | 0:791a779d6220 | 4229 | dataptr[2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 4230 | DESCALE(MULTIPLY(tmp0 - tmp1 - tmp1, FIX(0.707106781)), /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 4231 | CONST_BITS-PASS1_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 4232 | |
shoaib_ahmed | 0:791a779d6220 | 4233 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 4234 | |
shoaib_ahmed | 0:791a779d6220 | 4235 | dataptr[1] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 4236 | DESCALE(MULTIPLY(tmp2, FIX(1.224744871)), /* c1 */ |
shoaib_ahmed | 0:791a779d6220 | 4237 | CONST_BITS-PASS1_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 4238 | |
shoaib_ahmed | 0:791a779d6220 | 4239 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 4240 | } |
shoaib_ahmed | 0:791a779d6220 | 4241 | |
shoaib_ahmed | 0:791a779d6220 | 4242 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 4243 | * We remove the PASS1_BITS scaling, but leave the results scaled up |
shoaib_ahmed | 0:791a779d6220 | 4244 | * by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 4245 | * We must also scale the output by (8/6)*(8/3) = 32/9, which we partially |
shoaib_ahmed | 0:791a779d6220 | 4246 | * fold into the constant multipliers (other part was done in pass 1): |
shoaib_ahmed | 0:791a779d6220 | 4247 | * 6-point FDCT kernel, cK represents sqrt(2) * cos(K*pi/12) * 16/9. |
shoaib_ahmed | 0:791a779d6220 | 4248 | */ |
shoaib_ahmed | 0:791a779d6220 | 4249 | |
shoaib_ahmed | 0:791a779d6220 | 4250 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 4251 | for (ctr = 0; ctr < 3; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 4252 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 4253 | |
shoaib_ahmed | 0:791a779d6220 | 4254 | tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 4255 | tmp11 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 4256 | tmp2 = dataptr[DCTSIZE*2] + dataptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 4257 | |
shoaib_ahmed | 0:791a779d6220 | 4258 | tmp10 = tmp0 + tmp2; |
shoaib_ahmed | 0:791a779d6220 | 4259 | tmp12 = tmp0 - tmp2; |
shoaib_ahmed | 0:791a779d6220 | 4260 | |
shoaib_ahmed | 0:791a779d6220 | 4261 | tmp0 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*5]; |
shoaib_ahmed | 0:791a779d6220 | 4262 | tmp1 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*4]; |
shoaib_ahmed | 0:791a779d6220 | 4263 | tmp2 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 4264 | |
shoaib_ahmed | 0:791a779d6220 | 4265 | dataptr[DCTSIZE*0] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 4266 | DESCALE(MULTIPLY(tmp10 + tmp11, FIX(1.777777778)), /* 16/9 */ |
shoaib_ahmed | 0:791a779d6220 | 4267 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 4268 | dataptr[DCTSIZE*2] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 4269 | DESCALE(MULTIPLY(tmp12, FIX(2.177324216)), /* c2 */ |
shoaib_ahmed | 0:791a779d6220 | 4270 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 4271 | dataptr[DCTSIZE*4] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 4272 | DESCALE(MULTIPLY(tmp10 - tmp11 - tmp11, FIX(1.257078722)), /* c4 */ |
shoaib_ahmed | 0:791a779d6220 | 4273 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 4274 | |
shoaib_ahmed | 0:791a779d6220 | 4275 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 4276 | |
shoaib_ahmed | 0:791a779d6220 | 4277 | tmp10 = MULTIPLY(tmp0 + tmp2, FIX(0.650711829)); /* c5 */ |
shoaib_ahmed | 0:791a779d6220 | 4278 | |
shoaib_ahmed | 0:791a779d6220 | 4279 | dataptr[DCTSIZE*1] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 4280 | DESCALE(tmp10 + MULTIPLY(tmp0 + tmp1, FIX(1.777777778)), /* 16/9 */ |
shoaib_ahmed | 0:791a779d6220 | 4281 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 4282 | dataptr[DCTSIZE*3] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 4283 | DESCALE(MULTIPLY(tmp0 - tmp1 - tmp2, FIX(1.777777778)), /* 16/9 */ |
shoaib_ahmed | 0:791a779d6220 | 4284 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 4285 | dataptr[DCTSIZE*5] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 4286 | DESCALE(tmp10 + MULTIPLY(tmp2 - tmp1, FIX(1.777777778)), /* 16/9 */ |
shoaib_ahmed | 0:791a779d6220 | 4287 | CONST_BITS+PASS1_BITS); |
shoaib_ahmed | 0:791a779d6220 | 4288 | |
shoaib_ahmed | 0:791a779d6220 | 4289 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 4290 | } |
shoaib_ahmed | 0:791a779d6220 | 4291 | } |
shoaib_ahmed | 0:791a779d6220 | 4292 | |
shoaib_ahmed | 0:791a779d6220 | 4293 | |
shoaib_ahmed | 0:791a779d6220 | 4294 | /* |
shoaib_ahmed | 0:791a779d6220 | 4295 | * Perform the forward DCT on a 2x4 sample block. |
shoaib_ahmed | 0:791a779d6220 | 4296 | * |
shoaib_ahmed | 0:791a779d6220 | 4297 | * 2-point FDCT in pass 1 (rows), 4-point in pass 2 (columns). |
shoaib_ahmed | 0:791a779d6220 | 4298 | */ |
shoaib_ahmed | 0:791a779d6220 | 4299 | |
shoaib_ahmed | 0:791a779d6220 | 4300 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 4301 | jpeg_fdct_2x4 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 4302 | { |
shoaib_ahmed | 0:791a779d6220 | 4303 | INT32 tmp0, tmp1; |
shoaib_ahmed | 0:791a779d6220 | 4304 | INT32 tmp10, tmp11; |
shoaib_ahmed | 0:791a779d6220 | 4305 | DCTELEM *dataptr; |
shoaib_ahmed | 0:791a779d6220 | 4306 | JSAMPROW elemptr; |
shoaib_ahmed | 0:791a779d6220 | 4307 | int ctr; |
shoaib_ahmed | 0:791a779d6220 | 4308 | SHIFT_TEMPS |
shoaib_ahmed | 0:791a779d6220 | 4309 | |
shoaib_ahmed | 0:791a779d6220 | 4310 | /* Pre-zero output coefficient block. */ |
shoaib_ahmed | 0:791a779d6220 | 4311 | MEMZERO(data, SIZEOF(DCTELEM) * DCTSIZE2); |
shoaib_ahmed | 0:791a779d6220 | 4312 | |
shoaib_ahmed | 0:791a779d6220 | 4313 | /* Pass 1: process rows. |
shoaib_ahmed | 0:791a779d6220 | 4314 | * Note results are scaled up by sqrt(8) compared to a true DCT. |
shoaib_ahmed | 0:791a779d6220 | 4315 | * We must also scale the output by (8/2)*(8/4) = 2**3, which we add here. |
shoaib_ahmed | 0:791a779d6220 | 4316 | */ |
shoaib_ahmed | 0:791a779d6220 | 4317 | |
shoaib_ahmed | 0:791a779d6220 | 4318 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 4319 | for (ctr = 0; ctr < 4; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 4320 | elemptr = sample_data[ctr] + start_col; |
shoaib_ahmed | 0:791a779d6220 | 4321 | |
shoaib_ahmed | 0:791a779d6220 | 4322 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 4323 | |
shoaib_ahmed | 0:791a779d6220 | 4324 | tmp0 = GETJSAMPLE(elemptr[0]); |
shoaib_ahmed | 0:791a779d6220 | 4325 | tmp1 = GETJSAMPLE(elemptr[1]); |
shoaib_ahmed | 0:791a779d6220 | 4326 | |
shoaib_ahmed | 0:791a779d6220 | 4327 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 4328 | dataptr[0] = (DCTELEM) ((tmp0 + tmp1 - 2 * CENTERJSAMPLE) << 3); |
shoaib_ahmed | 0:791a779d6220 | 4329 | |
shoaib_ahmed | 0:791a779d6220 | 4330 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 4331 | |
shoaib_ahmed | 0:791a779d6220 | 4332 | dataptr[1] = (DCTELEM) ((tmp0 - tmp1) << 3); |
shoaib_ahmed | 0:791a779d6220 | 4333 | |
shoaib_ahmed | 0:791a779d6220 | 4334 | dataptr += DCTSIZE; /* advance pointer to next row */ |
shoaib_ahmed | 0:791a779d6220 | 4335 | } |
shoaib_ahmed | 0:791a779d6220 | 4336 | |
shoaib_ahmed | 0:791a779d6220 | 4337 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 4338 | * We leave the results scaled up by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 4339 | * 4-point FDCT kernel, |
shoaib_ahmed | 0:791a779d6220 | 4340 | * cK represents sqrt(2) * cos(K*pi/16) [refers to 8-point FDCT]. |
shoaib_ahmed | 0:791a779d6220 | 4341 | */ |
shoaib_ahmed | 0:791a779d6220 | 4342 | |
shoaib_ahmed | 0:791a779d6220 | 4343 | dataptr = data; |
shoaib_ahmed | 0:791a779d6220 | 4344 | for (ctr = 0; ctr < 2; ctr++) { |
shoaib_ahmed | 0:791a779d6220 | 4345 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 4346 | |
shoaib_ahmed | 0:791a779d6220 | 4347 | tmp0 = dataptr[DCTSIZE*0] + dataptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 4348 | tmp1 = dataptr[DCTSIZE*1] + dataptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 4349 | |
shoaib_ahmed | 0:791a779d6220 | 4350 | tmp10 = dataptr[DCTSIZE*0] - dataptr[DCTSIZE*3]; |
shoaib_ahmed | 0:791a779d6220 | 4351 | tmp11 = dataptr[DCTSIZE*1] - dataptr[DCTSIZE*2]; |
shoaib_ahmed | 0:791a779d6220 | 4352 | |
shoaib_ahmed | 0:791a779d6220 | 4353 | dataptr[DCTSIZE*0] = (DCTELEM) (tmp0 + tmp1); |
shoaib_ahmed | 0:791a779d6220 | 4354 | dataptr[DCTSIZE*2] = (DCTELEM) (tmp0 - tmp1); |
shoaib_ahmed | 0:791a779d6220 | 4355 | |
shoaib_ahmed | 0:791a779d6220 | 4356 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 4357 | |
shoaib_ahmed | 0:791a779d6220 | 4358 | tmp0 = MULTIPLY(tmp10 + tmp11, FIX_0_541196100); /* c6 */ |
shoaib_ahmed | 0:791a779d6220 | 4359 | /* Add fudge factor here for final descale. */ |
shoaib_ahmed | 0:791a779d6220 | 4360 | tmp0 += ONE << (CONST_BITS-1); |
shoaib_ahmed | 0:791a779d6220 | 4361 | |
shoaib_ahmed | 0:791a779d6220 | 4362 | dataptr[DCTSIZE*1] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 4363 | RIGHT_SHIFT(tmp0 + MULTIPLY(tmp10, FIX_0_765366865), /* c2-c6 */ |
shoaib_ahmed | 0:791a779d6220 | 4364 | CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 4365 | dataptr[DCTSIZE*3] = (DCTELEM) |
shoaib_ahmed | 0:791a779d6220 | 4366 | RIGHT_SHIFT(tmp0 - MULTIPLY(tmp11, FIX_1_847759065), /* c2+c6 */ |
shoaib_ahmed | 0:791a779d6220 | 4367 | CONST_BITS); |
shoaib_ahmed | 0:791a779d6220 | 4368 | |
shoaib_ahmed | 0:791a779d6220 | 4369 | dataptr++; /* advance pointer to next column */ |
shoaib_ahmed | 0:791a779d6220 | 4370 | } |
shoaib_ahmed | 0:791a779d6220 | 4371 | } |
shoaib_ahmed | 0:791a779d6220 | 4372 | |
shoaib_ahmed | 0:791a779d6220 | 4373 | |
shoaib_ahmed | 0:791a779d6220 | 4374 | /* |
shoaib_ahmed | 0:791a779d6220 | 4375 | * Perform the forward DCT on a 1x2 sample block. |
shoaib_ahmed | 0:791a779d6220 | 4376 | * |
shoaib_ahmed | 0:791a779d6220 | 4377 | * 1-point FDCT in pass 1 (rows), 2-point in pass 2 (columns). |
shoaib_ahmed | 0:791a779d6220 | 4378 | */ |
shoaib_ahmed | 0:791a779d6220 | 4379 | |
shoaib_ahmed | 0:791a779d6220 | 4380 | GLOBAL(void) |
shoaib_ahmed | 0:791a779d6220 | 4381 | jpeg_fdct_1x2 (DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col) |
shoaib_ahmed | 0:791a779d6220 | 4382 | { |
shoaib_ahmed | 0:791a779d6220 | 4383 | DCTELEM tmp0, tmp1; |
shoaib_ahmed | 0:791a779d6220 | 4384 | |
shoaib_ahmed | 0:791a779d6220 | 4385 | /* Pre-zero output coefficient block. */ |
shoaib_ahmed | 0:791a779d6220 | 4386 | MEMZERO(data, SIZEOF(DCTELEM) * DCTSIZE2); |
shoaib_ahmed | 0:791a779d6220 | 4387 | |
shoaib_ahmed | 0:791a779d6220 | 4388 | /* Pass 1: empty. */ |
shoaib_ahmed | 0:791a779d6220 | 4389 | |
shoaib_ahmed | 0:791a779d6220 | 4390 | /* Pass 2: process columns. |
shoaib_ahmed | 0:791a779d6220 | 4391 | * We leave the results scaled up by an overall factor of 8. |
shoaib_ahmed | 0:791a779d6220 | 4392 | * We must also scale the output by (8/1)*(8/2) = 2**5. |
shoaib_ahmed | 0:791a779d6220 | 4393 | */ |
shoaib_ahmed | 0:791a779d6220 | 4394 | |
shoaib_ahmed | 0:791a779d6220 | 4395 | /* Even part */ |
shoaib_ahmed | 0:791a779d6220 | 4396 | |
shoaib_ahmed | 0:791a779d6220 | 4397 | tmp0 = GETJSAMPLE(sample_data[0][start_col]); |
shoaib_ahmed | 0:791a779d6220 | 4398 | tmp1 = GETJSAMPLE(sample_data[1][start_col]); |
shoaib_ahmed | 0:791a779d6220 | 4399 | |
shoaib_ahmed | 0:791a779d6220 | 4400 | /* Apply unsigned->signed conversion. */ |
shoaib_ahmed | 0:791a779d6220 | 4401 | data[DCTSIZE*0] = (tmp0 + tmp1 - 2 * CENTERJSAMPLE) << 5; |
shoaib_ahmed | 0:791a779d6220 | 4402 | |
shoaib_ahmed | 0:791a779d6220 | 4403 | /* Odd part */ |
shoaib_ahmed | 0:791a779d6220 | 4404 | |
shoaib_ahmed | 0:791a779d6220 | 4405 | data[DCTSIZE*1] = (tmp0 - tmp1) << 5; |
shoaib_ahmed | 0:791a779d6220 | 4406 | } |
shoaib_ahmed | 0:791a779d6220 | 4407 | |
shoaib_ahmed | 0:791a779d6220 | 4408 | #endif /* DCT_SCALING_SUPPORTED */ |
shoaib_ahmed | 0:791a779d6220 | 4409 | #endif /* DCT_ISLOW_SUPPORTED */ |