1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/media/libjpeg/jdct.h Wed Dec 31 06:09:35 2014 +0100
1.3 @@ -0,0 +1,232 @@
1.4 +/*
1.5 + * jdct.h
1.6 + *
1.7 + * Copyright (C) 1994-1996, Thomas G. Lane.
1.8 + * This file is part of the Independent JPEG Group's software.
1.9 + * For conditions of distribution and use, see the accompanying README file.
1.10 + *
1.11 + * This include file contains common declarations for the forward and
1.12 + * inverse DCT modules. These declarations are private to the DCT managers
1.13 + * (jcdctmgr.c, jddctmgr.c) and the individual DCT algorithms.
1.14 + * The individual DCT algorithms are kept in separate files to ease
1.15 + * machine-dependent tuning (e.g., assembly coding).
1.16 + */
1.17 +
1.18 +
1.19 +/*
1.20 + * A forward DCT routine is given a pointer to a work area of type DCTELEM[];
1.21 + * the DCT is to be performed in-place in that buffer. Type DCTELEM is int
1.22 + * for 8-bit samples, INT32 for 12-bit samples. (NOTE: Floating-point DCT
1.23 + * implementations use an array of type FAST_FLOAT, instead.)
1.24 + * The DCT inputs are expected to be signed (range +-CENTERJSAMPLE).
1.25 + * The DCT outputs are returned scaled up by a factor of 8; they therefore
1.26 + * have a range of +-8K for 8-bit data, +-128K for 12-bit data. This
1.27 + * convention improves accuracy in integer implementations and saves some
1.28 + * work in floating-point ones.
1.29 + * Quantization of the output coefficients is done by jcdctmgr.c. This
1.30 + * step requires an unsigned type and also one with twice the bits.
1.31 + */
1.32 +
1.33 +#if BITS_IN_JSAMPLE == 8
1.34 +#ifndef WITH_SIMD
1.35 +typedef int DCTELEM; /* 16 or 32 bits is fine */
1.36 +typedef unsigned int UDCTELEM;
1.37 +typedef unsigned long long UDCTELEM2;
1.38 +#else
1.39 +typedef short DCTELEM; /* prefer 16 bit with SIMD for parellelism */
1.40 +typedef unsigned short UDCTELEM;
1.41 +typedef unsigned int UDCTELEM2;
1.42 +#endif
1.43 +#else
1.44 +typedef INT32 DCTELEM; /* must have 32 bits */
1.45 +typedef UINT32 UDCTELEM;
1.46 +typedef unsigned long long UDCTELEM2;
1.47 +#endif
1.48 +
1.49 +
1.50 +/*
1.51 + * An inverse DCT routine is given a pointer to the input JBLOCK and a pointer
1.52 + * to an output sample array. The routine must dequantize the input data as
1.53 + * well as perform the IDCT; for dequantization, it uses the multiplier table
1.54 + * pointed to by compptr->dct_table. The output data is to be placed into the
1.55 + * sample array starting at a specified column. (Any row offset needed will
1.56 + * be applied to the array pointer before it is passed to the IDCT code.)
1.57 + * Note that the number of samples emitted by the IDCT routine is
1.58 + * DCT_scaled_size * DCT_scaled_size.
1.59 + */
1.60 +
1.61 +/* typedef inverse_DCT_method_ptr is declared in jpegint.h */
1.62 +
1.63 +/*
1.64 + * Each IDCT routine has its own ideas about the best dct_table element type.
1.65 + */
1.66 +
1.67 +typedef MULTIPLIER ISLOW_MULT_TYPE; /* short or int, whichever is faster */
1.68 +#if BITS_IN_JSAMPLE == 8
1.69 +typedef MULTIPLIER IFAST_MULT_TYPE; /* 16 bits is OK, use short if faster */
1.70 +#define IFAST_SCALE_BITS 2 /* fractional bits in scale factors */
1.71 +#else
1.72 +typedef INT32 IFAST_MULT_TYPE; /* need 32 bits for scaled quantizers */
1.73 +#define IFAST_SCALE_BITS 13 /* fractional bits in scale factors */
1.74 +#endif
1.75 +typedef FAST_FLOAT FLOAT_MULT_TYPE; /* preferred floating type */
1.76 +
1.77 +
1.78 +/*
1.79 + * Each IDCT routine is responsible for range-limiting its results and
1.80 + * converting them to unsigned form (0..MAXJSAMPLE). The raw outputs could
1.81 + * be quite far out of range if the input data is corrupt, so a bulletproof
1.82 + * range-limiting step is required. We use a mask-and-table-lookup method
1.83 + * to do the combined operations quickly. See the comments with
1.84 + * prepare_range_limit_table (in jdmaster.c) for more info.
1.85 + */
1.86 +
1.87 +#define IDCT_range_limit(cinfo) ((cinfo)->sample_range_limit + CENTERJSAMPLE)
1.88 +
1.89 +#define RANGE_MASK (MAXJSAMPLE * 4 + 3) /* 2 bits wider than legal samples */
1.90 +
1.91 +
1.92 +/* Short forms of external names for systems with brain-damaged linkers. */
1.93 +
1.94 +#ifdef NEED_SHORT_EXTERNAL_NAMES
1.95 +#define jpeg_fdct_islow jFDislow
1.96 +#define jpeg_fdct_ifast jFDifast
1.97 +#define jpeg_fdct_float jFDfloat
1.98 +#define jpeg_idct_islow jRDislow
1.99 +#define jpeg_idct_ifast jRDifast
1.100 +#define jpeg_idct_float jRDfloat
1.101 +#define jpeg_idct_7x7 jRD7x7
1.102 +#define jpeg_idct_6x6 jRD6x6
1.103 +#define jpeg_idct_5x5 jRD5x5
1.104 +#define jpeg_idct_4x4 jRD4x4
1.105 +#define jpeg_idct_3x3 jRD3x3
1.106 +#define jpeg_idct_2x2 jRD2x2
1.107 +#define jpeg_idct_1x1 jRD1x1
1.108 +#define jpeg_idct_9x9 jRD9x9
1.109 +#define jpeg_idct_10x10 jRD10x10
1.110 +#define jpeg_idct_11x11 jRD11x11
1.111 +#define jpeg_idct_12x12 jRD12x12
1.112 +#define jpeg_idct_13x13 jRD13x13
1.113 +#define jpeg_idct_14x14 jRD14x14
1.114 +#define jpeg_idct_15x15 jRD15x15
1.115 +#define jpeg_idct_16x16 jRD16x16
1.116 +#endif /* NEED_SHORT_EXTERNAL_NAMES */
1.117 +
1.118 +/* Extern declarations for the forward and inverse DCT routines. */
1.119 +
1.120 +EXTERN(void) jpeg_fdct_islow JPP((DCTELEM * data));
1.121 +EXTERN(void) jpeg_fdct_ifast JPP((DCTELEM * data));
1.122 +EXTERN(void) jpeg_fdct_float JPP((FAST_FLOAT * data));
1.123 +
1.124 +EXTERN(void) jpeg_idct_islow
1.125 + JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
1.126 + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
1.127 +EXTERN(void) jpeg_idct_ifast
1.128 + JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
1.129 + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
1.130 +EXTERN(void) jpeg_idct_float
1.131 + JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
1.132 + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
1.133 +EXTERN(void) jpeg_idct_7x7
1.134 + JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
1.135 + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
1.136 +EXTERN(void) jpeg_idct_6x6
1.137 + JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
1.138 + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
1.139 +EXTERN(void) jpeg_idct_5x5
1.140 + JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
1.141 + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
1.142 +EXTERN(void) jpeg_idct_4x4
1.143 + JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
1.144 + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
1.145 +EXTERN(void) jpeg_idct_3x3
1.146 + JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
1.147 + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
1.148 +EXTERN(void) jpeg_idct_2x2
1.149 + JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
1.150 + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
1.151 +EXTERN(void) jpeg_idct_1x1
1.152 + JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
1.153 + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
1.154 +EXTERN(void) jpeg_idct_9x9
1.155 + JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
1.156 + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
1.157 +EXTERN(void) jpeg_idct_10x10
1.158 + JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
1.159 + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
1.160 +EXTERN(void) jpeg_idct_11x11
1.161 + JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
1.162 + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
1.163 +EXTERN(void) jpeg_idct_12x12
1.164 + JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
1.165 + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
1.166 +EXTERN(void) jpeg_idct_13x13
1.167 + JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
1.168 + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
1.169 +EXTERN(void) jpeg_idct_14x14
1.170 + JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
1.171 + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
1.172 +EXTERN(void) jpeg_idct_15x15
1.173 + JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
1.174 + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
1.175 +EXTERN(void) jpeg_idct_16x16
1.176 + JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
1.177 + JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
1.178 +
1.179 +
1.180 +/*
1.181 + * Macros for handling fixed-point arithmetic; these are used by many
1.182 + * but not all of the DCT/IDCT modules.
1.183 + *
1.184 + * All values are expected to be of type INT32.
1.185 + * Fractional constants are scaled left by CONST_BITS bits.
1.186 + * CONST_BITS is defined within each module using these macros,
1.187 + * and may differ from one module to the next.
1.188 + */
1.189 +
1.190 +#define ONE ((INT32) 1)
1.191 +#define CONST_SCALE (ONE << CONST_BITS)
1.192 +
1.193 +/* Convert a positive real constant to an integer scaled by CONST_SCALE.
1.194 + * Caution: some C compilers fail to reduce "FIX(constant)" at compile time,
1.195 + * thus causing a lot of useless floating-point operations at run time.
1.196 + */
1.197 +
1.198 +#define FIX(x) ((INT32) ((x) * CONST_SCALE + 0.5))
1.199 +
1.200 +/* Descale and correctly round an INT32 value that's scaled by N bits.
1.201 + * We assume RIGHT_SHIFT rounds towards minus infinity, so adding
1.202 + * the fudge factor is correct for either sign of X.
1.203 + */
1.204 +
1.205 +#define DESCALE(x,n) RIGHT_SHIFT((x) + (ONE << ((n)-1)), n)
1.206 +
1.207 +/* Multiply an INT32 variable by an INT32 constant to yield an INT32 result.
1.208 + * This macro is used only when the two inputs will actually be no more than
1.209 + * 16 bits wide, so that a 16x16->32 bit multiply can be used instead of a
1.210 + * full 32x32 multiply. This provides a useful speedup on many machines.
1.211 + * Unfortunately there is no way to specify a 16x16->32 multiply portably
1.212 + * in C, but some C compilers will do the right thing if you provide the
1.213 + * correct combination of casts.
1.214 + */
1.215 +
1.216 +#ifdef SHORTxSHORT_32 /* may work if 'int' is 32 bits */
1.217 +#define MULTIPLY16C16(var,const) (((INT16) (var)) * ((INT16) (const)))
1.218 +#endif
1.219 +#ifdef SHORTxLCONST_32 /* known to work with Microsoft C 6.0 */
1.220 +#define MULTIPLY16C16(var,const) (((INT16) (var)) * ((INT32) (const)))
1.221 +#endif
1.222 +
1.223 +#ifndef MULTIPLY16C16 /* default definition */
1.224 +#define MULTIPLY16C16(var,const) ((var) * (const))
1.225 +#endif
1.226 +
1.227 +/* Same except both inputs are variables. */
1.228 +
1.229 +#ifdef SHORTxSHORT_32 /* may work if 'int' is 32 bits */
1.230 +#define MULTIPLY16V16(var1,var2) (((INT16) (var1)) * ((INT16) (var2)))
1.231 +#endif
1.232 +
1.233 +#ifndef MULTIPLY16V16 /* default definition */
1.234 +#define MULTIPLY16V16(var1,var2) ((var1) * (var2))
1.235 +#endif
