1 /* vim: set ts=8 sw=8 noexpandtab: */
2 // qcms
3 // Copyright (C) 2009 Mozilla Foundation
4 //
5 // Permission is hereby granted, free of charge, to any person obtaining
6 // a copy of this software and associated documentation files (the "Software"),
7 // to deal in the Software without restriction, including without limitation
8 // the rights to use, copy, modify, merge, publish, distribute, sublicense,
9 // and/or sell copies of the Software, and to permit persons to whom the Software
10 // is furnished to do so, subject to the following conditions:
11 //
12 // The above copyright notice and this permission notice shall be included in
13 // all copies or substantial portions of the Software.
14 //
15 // THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
16 // EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO
17 // THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
18 // NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
19 // LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
20 // OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
21 // WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
22
23 #include "qcms.h"
24 #include "qcmstypes.h"
25
26 /* used as a lookup table for the output transformation.
27 * we refcount them so we only need to have one around per output
28 * profile, instead of duplicating them per transform */
29 struct precache_output
30 {
31 int ref_count;
32 /* We previously used a count of 65536 here but that seems like more
33 * precision than we actually need. By reducing the size we can
34 * improve startup performance and reduce memory usage. ColorSync on
35 * 10.5 uses 4097 which is perhaps because they use a fixed point
36 * representation where 1. is represented by 0x1000. */
37 #define PRECACHE_OUTPUT_SIZE 8192
38 #define PRECACHE_OUTPUT_MAX (PRECACHE_OUTPUT_SIZE-1)
39 uint8_t data[PRECACHE_OUTPUT_SIZE];
40 };
41
42 #ifdef _MSC_VER
43 #define ALIGN __declspec(align(16))
44 #else
45 #define ALIGN __attribute__(( aligned (16) ))
46 #endif
47
48 typedef struct _qcms_format_type {
49 int r;
50 int b;
51 } qcms_format_type;
52
53 struct _qcms_transform {
54 float ALIGN matrix[3][4];
55 float *input_gamma_table_r;
56 float *input_gamma_table_g;
57 float *input_gamma_table_b;
58
59 float *input_clut_table_r;
60 float *input_clut_table_g;
61 float *input_clut_table_b;
62 uint16_t input_clut_table_length;
63 float *r_clut;
64 float *g_clut;
65 float *b_clut;
66 uint16_t grid_size;
67 float *output_clut_table_r;
68 float *output_clut_table_g;
69 float *output_clut_table_b;
70 uint16_t output_clut_table_length;
71
72 float *input_gamma_table_gray;
73
74 float out_gamma_r;
75 float out_gamma_g;
76 float out_gamma_b;
77
78 float out_gamma_gray;
79
80 uint16_t *output_gamma_lut_r;
81 uint16_t *output_gamma_lut_g;
82 uint16_t *output_gamma_lut_b;
83
84 uint16_t *output_gamma_lut_gray;
85
86 size_t output_gamma_lut_r_length;
87 size_t output_gamma_lut_g_length;
88 size_t output_gamma_lut_b_length;
89
90 size_t output_gamma_lut_gray_length;
91
92 struct precache_output *output_table_r;
93 struct precache_output *output_table_g;
94 struct precache_output *output_table_b;
95
96 void (*transform_fn)(struct _qcms_transform *transform, unsigned char *src, unsigned char *dest, size_t length, struct _qcms_format_type output_format);
97
98 unsigned char floor_cache[256];
99 unsigned char ceil_cache[256];
100 float r_cache[256];
101
102 #define TRANSFORM_FLAG_MATRIX 0x0001
103 #define TRANSFORM_FLAG_CLUT_CACHE 0x0002
104
105 uint16_t transform_flags;
106 };
107
108 struct matrix {
109 float m[3][3];
110 bool invalid;
111 };
112
113 struct qcms_modular_transform;
114
115 typedef void (*transform_module_fn_t)(struct qcms_modular_transform *transform, float *src, float *dest, size_t length);
116
117 struct qcms_modular_transform {
118 struct matrix matrix;
119 float tx, ty, tz;
120
121 float *input_clut_table_r;
122 float *input_clut_table_g;
123 float *input_clut_table_b;
124 uint16_t input_clut_table_length;
125 float *r_clut;
126 float *g_clut;
127 float *b_clut;
128 uint16_t grid_size;
129 float *output_clut_table_r;
130 float *output_clut_table_g;
131 float *output_clut_table_b;
132 uint16_t output_clut_table_length;
133
134 uint16_t *output_gamma_lut_r;
135 uint16_t *output_gamma_lut_g;
136 uint16_t *output_gamma_lut_b;
137
138 size_t output_gamma_lut_r_length;
139 size_t output_gamma_lut_g_length;
140 size_t output_gamma_lut_b_length;
141
142 transform_module_fn_t transform_module_fn;
143 struct qcms_modular_transform *next_transform;
144 };
145
146 typedef int32_t s15Fixed16Number;
147 typedef uint16_t uInt16Number;
148 typedef uint8_t uInt8Number;
149
150 struct XYZNumber {
151 s15Fixed16Number X;
152 s15Fixed16Number Y;
153 s15Fixed16Number Z;
154 };
155
156 struct curveType {
157 uint32_t type;
158 uint32_t count;
159 float parameter[7];
160 uInt16Number data[];
161 };
162
163 struct lutmABType {
164 uint8_t num_in_channels;
165 uint8_t num_out_channels;
166 // 16 is the upperbound, actual is 0..num_in_channels.
167 uint8_t num_grid_points[16];
168
169 s15Fixed16Number e00;
170 s15Fixed16Number e01;
171 s15Fixed16Number e02;
172 s15Fixed16Number e03;
173 s15Fixed16Number e10;
174 s15Fixed16Number e11;
175 s15Fixed16Number e12;
176 s15Fixed16Number e13;
177 s15Fixed16Number e20;
178 s15Fixed16Number e21;
179 s15Fixed16Number e22;
180 s15Fixed16Number e23;
181
182 // reversed elements (for mBA)
183 bool reversed;
184
185 float *clut_table;
186 struct curveType *a_curves[10];
187 struct curveType *b_curves[10];
188 struct curveType *m_curves[10];
189 float clut_table_data[];
190 };
191
192 /* should lut8Type and lut16Type be different types? */
193 struct lutType { // used by lut8Type/lut16Type (mft2) only
194 uint8_t num_input_channels;
195 uint8_t num_output_channels;
196 uint8_t num_clut_grid_points;
197
198 s15Fixed16Number e00;
199 s15Fixed16Number e01;
200 s15Fixed16Number e02;
201 s15Fixed16Number e10;
202 s15Fixed16Number e11;
203 s15Fixed16Number e12;
204 s15Fixed16Number e20;
205 s15Fixed16Number e21;
206 s15Fixed16Number e22;
207
208 uint16_t num_input_table_entries;
209 uint16_t num_output_table_entries;
210
211 float *input_table;
212 float *clut_table;
213 float *output_table;
214
215 float table_data[];
216 };
217
218 struct vcgtType {
219 /* data contains three gamma channels: R[length], then G[length], then
220 * B[length]. */
221 uint16_t *data;
222 size_t length;
223 };
224
225 #if 0
226 /* this is from an intial idea of having the struct correspond to the data in
227 * the file. I decided that it wasn't a good idea.
228 */
229 struct tag_value {
230 uint32_t type;
231 union {
232 struct {
233 uint32_t reserved;
234 struct {
235 s15Fixed16Number X;
236 s15Fixed16Number Y;
237 s15Fixed16Number Z;
238 } XYZNumber;
239 } XYZType;
240 };
241 }; // I guess we need to pack this?
242 #endif
243
244 #define RGB_SIGNATURE 0x52474220
245 #define GRAY_SIGNATURE 0x47524159
246 #define XYZ_SIGNATURE 0x58595A20
247 #define LAB_SIGNATURE 0x4C616220
248
249 struct _qcms_profile {
250 uint32_t icc_version;
251 char description[64];
252 uint32_t class;
253 uint32_t color_space;
254 uint32_t pcs;
255 qcms_intent rendering_intent;
256 struct XYZNumber mediaWhitePoint;
257 struct XYZNumber redColorant;
258 struct XYZNumber blueColorant;
259 struct XYZNumber greenColorant;
260 struct curveType *redTRC;
261 struct curveType *blueTRC;
262 struct curveType *greenTRC;
263 struct curveType *grayTRC;
264 struct lutType *A2B0;
265 struct lutType *B2A0;
266 struct lutmABType *mAB;
267 struct lutmABType *mBA;
268 struct matrix chromaticAdaption;
269 struct vcgtType vcgt;
270
271 struct precache_output *output_table_r;
272 struct precache_output *output_table_g;
273 struct precache_output *output_table_b;
274 };
275
276 #ifdef _MSC_VER
277 #define inline _inline
278 #endif
279
280 /* produces the nearest float to 'a' with a maximum error
281 * of 1/1024 which happens for large values like 0x40000040 */
s15Fixed16Number_to_float(s15Fixed16Number a)282 static inline float s15Fixed16Number_to_float(s15Fixed16Number a)
283 {
284 return ((int32_t)a)/65536.f;
285 }
286
double_to_s15Fixed16Number(double v)287 static inline s15Fixed16Number double_to_s15Fixed16Number(double v)
288 {
289 return (int32_t)(v*65536);
290 }
291
uInt8Number_to_float(uInt8Number a)292 static inline float uInt8Number_to_float(uInt8Number a)
293 {
294 return ((int32_t)a)/255.f;
295 }
296
uInt16Number_to_float(uInt16Number a)297 static inline float uInt16Number_to_float(uInt16Number a)
298 {
299 return ((int32_t)a)/65535.f;
300 }
301
302
303 void precache_release(struct precache_output *p);
304 qcms_bool set_rgb_colorants(qcms_profile *profile, qcms_CIE_xyY white_point, qcms_CIE_xyYTRIPLE primaries);
305
306 void qcms_transform_data_rgb_out_lut_sse2(qcms_transform *transform,
307 unsigned char *src,
308 unsigned char *dest,
309 size_t length,
310 qcms_format_type output_format);
311 void qcms_transform_data_rgba_out_lut_sse2(qcms_transform *transform,
312 unsigned char *src,
313 unsigned char *dest,
314 size_t length,
315 qcms_format_type output_format);
316 void qcms_transform_data_tetra_clut_rgba_sse2(qcms_transform* transform,
317 unsigned char* src,
318 unsigned char* dest,
319 size_t length,
320 qcms_format_type output_format);
321
322 void qcms_transform_build_clut_cache(qcms_transform* transform);
323
324 extern qcms_bool qcms_supports_iccv4;
325
326
327 #ifdef _MSC_VER
328
329 long __cdecl _InterlockedIncrement(long volatile *);
330 long __cdecl _InterlockedDecrement(long volatile *);
331 #pragma intrinsic(_InterlockedIncrement)
332 #pragma intrinsic(_InterlockedDecrement)
333
334 #define qcms_atomic_increment(x) _InterlockedIncrement((long volatile *)&x)
335 #define qcms_atomic_decrement(x) _InterlockedDecrement((long volatile*)&x)
336
337 #else
338
339 #define qcms_atomic_increment(x) __sync_add_and_fetch(&x, 1)
340 #define qcms_atomic_decrement(x) __sync_sub_and_fetch(&x, 1)
341
342 #endif
343