1 /*
2  * jdct.h
3  *
4  * Copyright (C) 1994-1996, Thomas G. Lane.
5  * This file is part of the Independent JPEG Group's software.
6  * For conditions of distribution and use, see the accompanying README file.
7  *
8  * This include file contains common declarations for the forward and
9  * inverse DCT modules.  These declarations are private to the DCT managers
10  * (jcdctmgr.c, jddctmgr.c) and the individual DCT algorithms.
11  * The individual DCT algorithms are kept in separate files to ease
12  * machine-dependent tuning (e.g., assembly coding).
13  */
14 
15 
16 /*
17  * A forward DCT routine is given a pointer to an input sample array and
18  * a pointer to a work area of type DCTELEM[]; the DCT is to be performed
19  * in-place in that buffer.  Type DCTELEM is int for 8-bit samples, INT32
20  * for 12-bit samples.  (NOTE: Floating-point DCT implementations use an
21  * array of type FAST_FLOAT, instead.)
22  * The input data is to be fetched from the sample array starting at a
23  * specified column.  (Any row offset needed will be applied to the array
24  * pointer before it is passed to the FDCT code.)
25  * Note that the number of samples fetched by the FDCT routine is
26  * DCT_h_scaled_size * DCT_v_scaled_size.
27  * The DCT outputs are returned scaled up by a factor of 8; they therefore
28  * have a range of +-8K for 8-bit data, +-128K for 12-bit data.  This
29  * convention improves accuracy in integer implementations and saves some
30  * work in floating-point ones.
31  * Quantization of the output coefficients is done by jcdctmgr.c.
32  */
33 
34 #if BITS_IN_JSAMPLE == 8
35 typedef int DCTELEM;		/* 16 or 32 bits is fine */
36 #else
37 typedef INT32 DCTELEM;		/* must have 32 bits */
38 #endif
39 
40 typedef JMETHOD(void, forward_DCT_method_ptr, (DCTELEM * data,
41 					       JSAMPARRAY sample_data,
42 					       JDIMENSION start_col));
43 typedef JMETHOD(void, float_DCT_method_ptr, (FAST_FLOAT * data,
44 					     JSAMPARRAY sample_data,
45 					     JDIMENSION start_col));
46 
47 
48 /*
49  * An inverse DCT routine is given a pointer to the input JBLOCK and a pointer
50  * to an output sample array.  The routine must dequantize the input data as
51  * well as perform the IDCT; for dequantization, it uses the multiplier table
52  * pointed to by compptr->dct_table.  The output data is to be placed into the
53  * sample array starting at a specified column.  (Any row offset needed will
54  * be applied to the array pointer before it is passed to the IDCT code.)
55  * Note that the number of samples emitted by the IDCT routine is
56  * DCT_h_scaled_size * DCT_v_scaled_size.
57  */
58 
59 /* typedef inverse_DCT_method_ptr is declared in jpegint.h */
60 
61 /*
62  * Each IDCT routine has its own ideas about the best dct_table element type.
63  */
64 
65 typedef MULTIPLIER ISLOW_MULT_TYPE; /* short or int, whichever is faster */
66 #if BITS_IN_JSAMPLE == 8
67 typedef MULTIPLIER IFAST_MULT_TYPE; /* 16 bits is OK, use short if faster */
68 #define IFAST_SCALE_BITS  2	/* fractional bits in scale factors */
69 #else
70 typedef INT32 IFAST_MULT_TYPE;	/* need 32 bits for scaled quantizers */
71 #define IFAST_SCALE_BITS  13	/* fractional bits in scale factors */
72 #endif
73 typedef FAST_FLOAT FLOAT_MULT_TYPE; /* preferred floating type */
74 
75 
76 /*
77  * Each IDCT routine is responsible for range-limiting its results and
78  * converting them to unsigned form (0..MAXJSAMPLE).  The raw outputs could
79  * be quite far out of range if the input data is corrupt, so a bulletproof
80  * range-limiting step is required.  We use a mask-and-table-lookup method
81  * to do the combined operations quickly.  See the comments with
82  * prepare_range_limit_table (in jdmaster.c) for more info.
83  */
84 
85 #define IDCT_range_limit(cinfo)  ((cinfo)->sample_range_limit + CENTERJSAMPLE)
86 
87 #define RANGE_MASK  (MAXJSAMPLE * 4 + 3) /* 2 bits wider than legal samples */
88 
89 
90 /* Short forms of external names for systems with brain-damaged linkers. */
91 
92 #ifdef NEED_SHORT_EXTERNAL_NAMES
93 #define jpeg_fdct_islow		jFDislow
94 #define jpeg_fdct_ifast		jFDifast
95 #define jpeg_fdct_float		jFDfloat
96 #define jpeg_fdct_7x7		jFD7x7
97 #define jpeg_fdct_6x6		jFD6x6
98 #define jpeg_fdct_5x5		jFD5x5
99 #define jpeg_fdct_4x4		jFD4x4
100 #define jpeg_fdct_3x3		jFD3x3
101 #define jpeg_fdct_2x2		jFD2x2
102 #define jpeg_fdct_1x1		jFD1x1
103 #define jpeg_fdct_9x9		jFD9x9
104 #define jpeg_fdct_10x10		jFD10x10
105 #define jpeg_fdct_11x11		jFD11x11
106 #define jpeg_fdct_12x12		jFD12x12
107 #define jpeg_fdct_13x13		jFD13x13
108 #define jpeg_fdct_14x14		jFD14x14
109 #define jpeg_fdct_15x15		jFD15x15
110 #define jpeg_fdct_16x16		jFD16x16
111 #define jpeg_fdct_16x8		jFD16x8
112 #define jpeg_fdct_14x7		jFD14x7
113 #define jpeg_fdct_12x6		jFD12x6
114 #define jpeg_fdct_10x5		jFD10x5
115 #define jpeg_fdct_8x4		jFD8x4
116 #define jpeg_fdct_6x3		jFD6x3
117 #define jpeg_fdct_4x2		jFD4x2
118 #define jpeg_fdct_2x1		jFD2x1
119 #define jpeg_fdct_8x16		jFD8x16
120 #define jpeg_fdct_7x14		jFD7x14
121 #define jpeg_fdct_6x12		jFD6x12
122 #define jpeg_fdct_5x10		jFD5x10
123 #define jpeg_fdct_4x8		jFD4x8
124 #define jpeg_fdct_3x6		jFD3x6
125 #define jpeg_fdct_2x4		jFD2x4
126 #define jpeg_fdct_1x2		jFD1x2
127 #define jpeg_idct_islow		jRDislow
128 #define jpeg_idct_ifast		jRDifast
129 #define jpeg_idct_float		jRDfloat
130 #define jpeg_idct_7x7		jRD7x7
131 #define jpeg_idct_6x6		jRD6x6
132 #define jpeg_idct_5x5		jRD5x5
133 #define jpeg_idct_4x4		jRD4x4
134 #define jpeg_idct_3x3		jRD3x3
135 #define jpeg_idct_2x2		jRD2x2
136 #define jpeg_idct_1x1		jRD1x1
137 #define jpeg_idct_9x9		jRD9x9
138 #define jpeg_idct_10x10		jRD10x10
139 #define jpeg_idct_11x11		jRD11x11
140 #define jpeg_idct_12x12		jRD12x12
141 #define jpeg_idct_13x13		jRD13x13
142 #define jpeg_idct_14x14		jRD14x14
143 #define jpeg_idct_15x15		jRD15x15
144 #define jpeg_idct_16x16		jRD16x16
145 #define jpeg_idct_16x8		jRD16x8
146 #define jpeg_idct_14x7		jRD14x7
147 #define jpeg_idct_12x6		jRD12x6
148 #define jpeg_idct_10x5		jRD10x5
149 #define jpeg_idct_8x4		jRD8x4
150 #define jpeg_idct_6x3		jRD6x3
151 #define jpeg_idct_4x2		jRD4x2
152 #define jpeg_idct_2x1		jRD2x1
153 #define jpeg_idct_8x16		jRD8x16
154 #define jpeg_idct_7x14		jRD7x14
155 #define jpeg_idct_6x12		jRD6x12
156 #define jpeg_idct_5x10		jRD5x10
157 #define jpeg_idct_4x8		jRD4x8
158 #define jpeg_idct_3x6		jRD3x8
159 #define jpeg_idct_2x4		jRD2x4
160 #define jpeg_idct_1x2		jRD1x2
161 #endif /* NEED_SHORT_EXTERNAL_NAMES */
162 
163 /* Extern declarations for the forward and inverse DCT routines. */
164 
165 EXTERN(void) jpeg_fdct_islow
166     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
167 EXTERN(void) jpeg_fdct_ifast
168     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
169 EXTERN(void) jpeg_fdct_float
170     JPP((FAST_FLOAT * data, JSAMPARRAY sample_data, JDIMENSION start_col));
171 EXTERN(void) jpeg_fdct_7x7
172     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
173 EXTERN(void) jpeg_fdct_6x6
174     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
175 EXTERN(void) jpeg_fdct_5x5
176     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
177 EXTERN(void) jpeg_fdct_4x4
178     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
179 EXTERN(void) jpeg_fdct_3x3
180     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
181 EXTERN(void) jpeg_fdct_2x2
182     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
183 EXTERN(void) jpeg_fdct_1x1
184     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
185 EXTERN(void) jpeg_fdct_9x9
186     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
187 EXTERN(void) jpeg_fdct_10x10
188     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
189 EXTERN(void) jpeg_fdct_11x11
190     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
191 EXTERN(void) jpeg_fdct_12x12
192     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
193 EXTERN(void) jpeg_fdct_13x13
194     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
195 EXTERN(void) jpeg_fdct_14x14
196     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
197 EXTERN(void) jpeg_fdct_15x15
198     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
199 EXTERN(void) jpeg_fdct_16x16
200     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
201 EXTERN(void) jpeg_fdct_16x8
202     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
203 EXTERN(void) jpeg_fdct_14x7
204     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
205 EXTERN(void) jpeg_fdct_12x6
206     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
207 EXTERN(void) jpeg_fdct_10x5
208     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
209 EXTERN(void) jpeg_fdct_8x4
210     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
211 EXTERN(void) jpeg_fdct_6x3
212     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
213 EXTERN(void) jpeg_fdct_4x2
214     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
215 EXTERN(void) jpeg_fdct_2x1
216     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
217 EXTERN(void) jpeg_fdct_8x16
218     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
219 EXTERN(void) jpeg_fdct_7x14
220     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
221 EXTERN(void) jpeg_fdct_6x12
222     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
223 EXTERN(void) jpeg_fdct_5x10
224     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
225 EXTERN(void) jpeg_fdct_4x8
226     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
227 EXTERN(void) jpeg_fdct_3x6
228     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
229 EXTERN(void) jpeg_fdct_2x4
230     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
231 EXTERN(void) jpeg_fdct_1x2
232     JPP((DCTELEM * data, JSAMPARRAY sample_data, JDIMENSION start_col));
233 
234 EXTERN(void) jpeg_idct_islow
235     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
236 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
237 EXTERN(void) jpeg_idct_ifast
238     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
239 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
240 EXTERN(void) jpeg_idct_float
241     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
242 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
243 EXTERN(void) jpeg_idct_7x7
244     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
245 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
246 EXTERN(void) jpeg_idct_6x6
247     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
248 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
249 EXTERN(void) jpeg_idct_5x5
250     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
251 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
252 EXTERN(void) jpeg_idct_4x4
253     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
254 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
255 EXTERN(void) jpeg_idct_3x3
256     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
257 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
258 EXTERN(void) jpeg_idct_2x2
259     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
260 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
261 EXTERN(void) jpeg_idct_1x1
262     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
263 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
264 EXTERN(void) jpeg_idct_9x9
265     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
266 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
267 EXTERN(void) jpeg_idct_10x10
268     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
269 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
270 EXTERN(void) jpeg_idct_11x11
271     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
272 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
273 EXTERN(void) jpeg_idct_12x12
274     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
275 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
276 EXTERN(void) jpeg_idct_13x13
277     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
278 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
279 EXTERN(void) jpeg_idct_14x14
280     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
281 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
282 EXTERN(void) jpeg_idct_15x15
283     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
284 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
285 EXTERN(void) jpeg_idct_16x16
286     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
287 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
288 EXTERN(void) jpeg_idct_16x8
289     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
290 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
291 EXTERN(void) jpeg_idct_14x7
292     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
293 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
294 EXTERN(void) jpeg_idct_12x6
295     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
296 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
297 EXTERN(void) jpeg_idct_10x5
298     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
299 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
300 EXTERN(void) jpeg_idct_8x4
301     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
302 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
303 EXTERN(void) jpeg_idct_6x3
304     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
305 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
306 EXTERN(void) jpeg_idct_4x2
307     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
308 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
309 EXTERN(void) jpeg_idct_2x1
310     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
311 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
312 EXTERN(void) jpeg_idct_8x16
313     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
314 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
315 EXTERN(void) jpeg_idct_7x14
316     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
317 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
318 EXTERN(void) jpeg_idct_6x12
319     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
320 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
321 EXTERN(void) jpeg_idct_5x10
322     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
323 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
324 EXTERN(void) jpeg_idct_4x8
325     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
326 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
327 EXTERN(void) jpeg_idct_3x6
328     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
329 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
330 EXTERN(void) jpeg_idct_2x4
331     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
332 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
333 EXTERN(void) jpeg_idct_1x2
334     JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
335 	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
336 
337 
338 /*
339  * Macros for handling fixed-point arithmetic; these are used by many
340  * but not all of the DCT/IDCT modules.
341  *
342  * All values are expected to be of type INT32.
343  * Fractional constants are scaled left by CONST_BITS bits.
344  * CONST_BITS is defined within each module using these macros,
345  * and may differ from one module to the next.
346  */
347 
348 #define ONE	((INT32) 1)
349 #define CONST_SCALE (ONE << CONST_BITS)
350 
351 /* Convert a positive real constant to an integer scaled by CONST_SCALE.
352  * Caution: some C compilers fail to reduce "FIX(constant)" at compile time,
353  * thus causing a lot of useless floating-point operations at run time.
354  */
355 
356 #define FIX(x)	((INT32) ((x) * CONST_SCALE + 0.5))
357 
358 /* Descale and correctly round an INT32 value that's scaled by N bits.
359  * We assume RIGHT_SHIFT rounds towards minus infinity, so adding
360  * the fudge factor is correct for either sign of X.
361  */
362 
363 #define DESCALE(x,n)  RIGHT_SHIFT((x) + (ONE << ((n)-1)), n)
364 
365 /* Multiply an INT32 variable by an INT32 constant to yield an INT32 result.
366  * This macro is used only when the two inputs will actually be no more than
367  * 16 bits wide, so that a 16x16->32 bit multiply can be used instead of a
368  * full 32x32 multiply.  This provides a useful speedup on many machines.
369  * Unfortunately there is no way to specify a 16x16->32 multiply portably
370  * in C, but some C compilers will do the right thing if you provide the
371  * correct combination of casts.
372  */
373 
374 #ifdef SHORTxSHORT_32		/* may work if 'int' is 32 bits */
375 #define MULTIPLY16C16(var,const)  (((INT16) (var)) * ((INT16) (const)))
376 #endif
377 #ifdef SHORTxLCONST_32		/* known to work with Microsoft C 6.0 */
378 #define MULTIPLY16C16(var,const)  (((INT16) (var)) * ((INT32) (const)))
379 #endif
380 
381 #ifndef MULTIPLY16C16		/* default definition */
382 #define MULTIPLY16C16(var,const)  ((var) * (const))
383 #endif
384 
385 /* Same except both inputs are variables. */
386 
387 #ifdef SHORTxSHORT_32		/* may work if 'int' is 32 bits */
388 #define MULTIPLY16V16(var1,var2)  (((INT16) (var1)) * ((INT16) (var2)))
389 #endif
390 
391 #ifndef MULTIPLY16V16		/* default definition */
392 #define MULTIPLY16V16(var1,var2)  ((var1) * (var2))
393 #endif
394