1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/gfx/qcms/qcmsint.h Wed Dec 31 06:09:35 2014 +0100
1.3 @@ -0,0 +1,327 @@
1.4 +/* vim: set ts=8 sw=8 noexpandtab: */
1.5 +#include "qcms.h"
1.6 +#include "qcmstypes.h"
1.7 +
1.8 +/* used as a lookup table for the output transformation.
1.9 + * we refcount them so we only need to have one around per output
1.10 + * profile, instead of duplicating them per transform */
1.11 +struct precache_output
1.12 +{
1.13 + int ref_count;
1.14 + /* We previously used a count of 65536 here but that seems like more
1.15 + * precision than we actually need. By reducing the size we can
1.16 + * improve startup performance and reduce memory usage. ColorSync on
1.17 + * 10.5 uses 4097 which is perhaps because they use a fixed point
1.18 + * representation where 1. is represented by 0x1000. */
1.19 +#define PRECACHE_OUTPUT_SIZE 8192
1.20 +#define PRECACHE_OUTPUT_MAX (PRECACHE_OUTPUT_SIZE-1)
1.21 + uint8_t data[PRECACHE_OUTPUT_SIZE];
1.22 +};
1.23 +
1.24 +#ifdef _MSC_VER
1.25 +#define ALIGN __declspec(align(16))
1.26 +#else
1.27 +#define ALIGN __attribute__(( aligned (16) ))
1.28 +#endif
1.29 +
1.30 +struct _qcms_transform {
1.31 + float ALIGN matrix[3][4];
1.32 + float *input_gamma_table_r;
1.33 + float *input_gamma_table_g;
1.34 + float *input_gamma_table_b;
1.35 +
1.36 + float *input_clut_table_r;
1.37 + float *input_clut_table_g;
1.38 + float *input_clut_table_b;
1.39 + uint16_t input_clut_table_length;
1.40 + float *r_clut;
1.41 + float *g_clut;
1.42 + float *b_clut;
1.43 + uint16_t grid_size;
1.44 + float *output_clut_table_r;
1.45 + float *output_clut_table_g;
1.46 + float *output_clut_table_b;
1.47 + uint16_t output_clut_table_length;
1.48 +
1.49 + float *input_gamma_table_gray;
1.50 +
1.51 + float out_gamma_r;
1.52 + float out_gamma_g;
1.53 + float out_gamma_b;
1.54 +
1.55 + float out_gamma_gray;
1.56 +
1.57 + uint16_t *output_gamma_lut_r;
1.58 + uint16_t *output_gamma_lut_g;
1.59 + uint16_t *output_gamma_lut_b;
1.60 +
1.61 + uint16_t *output_gamma_lut_gray;
1.62 +
1.63 + size_t output_gamma_lut_r_length;
1.64 + size_t output_gamma_lut_g_length;
1.65 + size_t output_gamma_lut_b_length;
1.66 +
1.67 + size_t output_gamma_lut_gray_length;
1.68 +
1.69 + struct precache_output *output_table_r;
1.70 + struct precache_output *output_table_g;
1.71 + struct precache_output *output_table_b;
1.72 +
1.73 + void (*transform_fn)(struct _qcms_transform *transform, unsigned char *src, unsigned char *dest, size_t length);
1.74 +};
1.75 +
1.76 +struct matrix {
1.77 + float m[3][3];
1.78 + bool invalid;
1.79 +};
1.80 +
1.81 +struct qcms_modular_transform;
1.82 +
1.83 +typedef void (*transform_module_fn_t)(struct qcms_modular_transform *transform, float *src, float *dest, size_t length);
1.84 +
1.85 +struct qcms_modular_transform {
1.86 + struct matrix matrix;
1.87 + float tx, ty, tz;
1.88 +
1.89 + float *input_clut_table_r;
1.90 + float *input_clut_table_g;
1.91 + float *input_clut_table_b;
1.92 + uint16_t input_clut_table_length;
1.93 + float *r_clut;
1.94 + float *g_clut;
1.95 + float *b_clut;
1.96 + uint16_t grid_size;
1.97 + float *output_clut_table_r;
1.98 + float *output_clut_table_g;
1.99 + float *output_clut_table_b;
1.100 + uint16_t output_clut_table_length;
1.101 +
1.102 + uint16_t *output_gamma_lut_r;
1.103 + uint16_t *output_gamma_lut_g;
1.104 + uint16_t *output_gamma_lut_b;
1.105 +
1.106 + size_t output_gamma_lut_r_length;
1.107 + size_t output_gamma_lut_g_length;
1.108 + size_t output_gamma_lut_b_length;
1.109 +
1.110 + transform_module_fn_t transform_module_fn;
1.111 + struct qcms_modular_transform *next_transform;
1.112 +};
1.113 +
1.114 +typedef int32_t s15Fixed16Number;
1.115 +typedef uint16_t uInt16Number;
1.116 +typedef uint8_t uInt8Number;
1.117 +
1.118 +struct XYZNumber {
1.119 + s15Fixed16Number X;
1.120 + s15Fixed16Number Y;
1.121 + s15Fixed16Number Z;
1.122 +};
1.123 +
1.124 +struct curveType {
1.125 + uint32_t type;
1.126 + uint32_t count;
1.127 + float parameter[7];
1.128 + uInt16Number data[];
1.129 +};
1.130 +
1.131 +struct lutmABType {
1.132 + uint8_t num_in_channels;
1.133 + uint8_t num_out_channels;
1.134 + // 16 is the upperbound, actual is 0..num_in_channels.
1.135 + uint8_t num_grid_points[16];
1.136 +
1.137 + s15Fixed16Number e00;
1.138 + s15Fixed16Number e01;
1.139 + s15Fixed16Number e02;
1.140 + s15Fixed16Number e03;
1.141 + s15Fixed16Number e10;
1.142 + s15Fixed16Number e11;
1.143 + s15Fixed16Number e12;
1.144 + s15Fixed16Number e13;
1.145 + s15Fixed16Number e20;
1.146 + s15Fixed16Number e21;
1.147 + s15Fixed16Number e22;
1.148 + s15Fixed16Number e23;
1.149 +
1.150 + // reversed elements (for mBA)
1.151 + bool reversed;
1.152 +
1.153 + float *clut_table;
1.154 + struct curveType *a_curves[10];
1.155 + struct curveType *b_curves[10];
1.156 + struct curveType *m_curves[10];
1.157 + float clut_table_data[];
1.158 +};
1.159 +
1.160 +/* should lut8Type and lut16Type be different types? */
1.161 +struct lutType { // used by lut8Type/lut16Type (mft2) only
1.162 + uint8_t num_input_channels;
1.163 + uint8_t num_output_channels;
1.164 + uint8_t num_clut_grid_points;
1.165 +
1.166 + s15Fixed16Number e00;
1.167 + s15Fixed16Number e01;
1.168 + s15Fixed16Number e02;
1.169 + s15Fixed16Number e10;
1.170 + s15Fixed16Number e11;
1.171 + s15Fixed16Number e12;
1.172 + s15Fixed16Number e20;
1.173 + s15Fixed16Number e21;
1.174 + s15Fixed16Number e22;
1.175 +
1.176 + uint16_t num_input_table_entries;
1.177 + uint16_t num_output_table_entries;
1.178 +
1.179 + float *input_table;
1.180 + float *clut_table;
1.181 + float *output_table;
1.182 +
1.183 + float table_data[];
1.184 +};
1.185 +#if 0
1.186 +/* this is from an intial idea of having the struct correspond to the data in
1.187 + * the file. I decided that it wasn't a good idea.
1.188 + */
1.189 +struct tag_value {
1.190 + uint32_t type;
1.191 + union {
1.192 + struct {
1.193 + uint32_t reserved;
1.194 + struct {
1.195 + s15Fixed16Number X;
1.196 + s15Fixed16Number Y;
1.197 + s15Fixed16Number Z;
1.198 + } XYZNumber;
1.199 + } XYZType;
1.200 + };
1.201 +}; // I guess we need to pack this?
1.202 +#endif
1.203 +
1.204 +#define RGB_SIGNATURE 0x52474220
1.205 +#define GRAY_SIGNATURE 0x47524159
1.206 +#define XYZ_SIGNATURE 0x58595A20
1.207 +#define LAB_SIGNATURE 0x4C616220
1.208 +
1.209 +struct _qcms_profile {
1.210 + uint32_t class;
1.211 + uint32_t color_space;
1.212 + uint32_t pcs;
1.213 + qcms_intent rendering_intent;
1.214 + struct XYZNumber redColorant;
1.215 + struct XYZNumber blueColorant;
1.216 + struct XYZNumber greenColorant;
1.217 + struct curveType *redTRC;
1.218 + struct curveType *blueTRC;
1.219 + struct curveType *greenTRC;
1.220 + struct curveType *grayTRC;
1.221 + struct lutType *A2B0;
1.222 + struct lutType *B2A0;
1.223 + struct lutmABType *mAB;
1.224 + struct lutmABType *mBA;
1.225 + struct matrix chromaticAdaption;
1.226 +
1.227 + struct precache_output *output_table_r;
1.228 + struct precache_output *output_table_g;
1.229 + struct precache_output *output_table_b;
1.230 +};
1.231 +
1.232 +#ifdef _MSC_VER
1.233 +#define inline _inline
1.234 +#endif
1.235 +
1.236 +/* produces the nearest float to 'a' with a maximum error
1.237 + * of 1/1024 which happens for large values like 0x40000040 */
1.238 +static inline float s15Fixed16Number_to_float(s15Fixed16Number a)
1.239 +{
1.240 + return ((int32_t)a)/65536.f;
1.241 +}
1.242 +
1.243 +static inline s15Fixed16Number double_to_s15Fixed16Number(double v)
1.244 +{
1.245 + return (int32_t)(v*65536);
1.246 +}
1.247 +
1.248 +static inline float uInt8Number_to_float(uInt8Number a)
1.249 +{
1.250 + return ((int32_t)a)/255.f;
1.251 +}
1.252 +
1.253 +static inline float uInt16Number_to_float(uInt16Number a)
1.254 +{
1.255 + return ((int32_t)a)/65535.f;
1.256 +}
1.257 +
1.258 +
1.259 +void precache_release(struct precache_output *p);
1.260 +qcms_bool set_rgb_colorants(qcms_profile *profile, qcms_CIE_xyY white_point, qcms_CIE_xyYTRIPLE primaries);
1.261 +qcms_bool get_rgb_colorants(struct matrix *colorants, qcms_CIE_xyY white_point, qcms_CIE_xyYTRIPLE primaries);
1.262 +
1.263 +void qcms_transform_data_rgb_out_lut_sse2(qcms_transform *transform,
1.264 + unsigned char *src,
1.265 + unsigned char *dest,
1.266 + size_t length);
1.267 +void qcms_transform_data_rgba_out_lut_sse2(qcms_transform *transform,
1.268 + unsigned char *src,
1.269 + unsigned char *dest,
1.270 + size_t length);
1.271 +void qcms_transform_data_rgb_out_lut_sse1(qcms_transform *transform,
1.272 + unsigned char *src,
1.273 + unsigned char *dest,
1.274 + size_t length);
1.275 +void qcms_transform_data_rgba_out_lut_sse1(qcms_transform *transform,
1.276 + unsigned char *src,
1.277 + unsigned char *dest,
1.278 + size_t length);
1.279 +
1.280 +void qcms_transform_data_rgb_out_lut_altivec(qcms_transform *transform,
1.281 + unsigned char *src,
1.282 + unsigned char *dest,
1.283 + size_t length);
1.284 +void qcms_transform_data_rgba_out_lut_altivec(qcms_transform *transform,
1.285 + unsigned char *src,
1.286 + unsigned char *dest,
1.287 + size_t length);
1.288 +
1.289 +extern qcms_bool qcms_supports_iccv4;
1.290 +
1.291 +#ifdef _MSC_VER
1.292 +
1.293 +long __cdecl _InterlockedIncrement(long volatile *);
1.294 +long __cdecl _InterlockedDecrement(long volatile *);
1.295 +#pragma intrinsic(_InterlockedIncrement)
1.296 +#pragma intrinsic(_InterlockedDecrement)
1.297 +
1.298 +#define qcms_atomic_increment(x) _InterlockedIncrement((long volatile *)&x)
1.299 +#define qcms_atomic_decrement(x) _InterlockedDecrement((long volatile*)&x)
1.300 +
1.301 +#else
1.302 +
1.303 +#define qcms_atomic_increment(x) __sync_add_and_fetch(&x, 1)
1.304 +#define qcms_atomic_decrement(x) __sync_sub_and_fetch(&x, 1)
1.305 +
1.306 +#endif
1.307 +
1.308 +
1.309 +#ifdef NATIVE_OUTPUT
1.310 +# define RGB_OUTPUT_COMPONENTS 4
1.311 +# define RGBA_OUTPUT_COMPONENTS 4
1.312 +# ifdef IS_LITTLE_ENDIAN
1.313 +# define OUTPUT_A_INDEX 3
1.314 +# define OUTPUT_R_INDEX 2
1.315 +# define OUTPUT_G_INDEX 1
1.316 +# define OUTPUT_B_INDEX 0
1.317 +# else
1.318 +# define OUTPUT_A_INDEX 0
1.319 +# define OUTPUT_R_INDEX 1
1.320 +# define OUTPUT_G_INDEX 2
1.321 +# define OUTPUT_B_INDEX 3
1.322 +# endif
1.323 +#else
1.324 +# define RGB_OUTPUT_COMPONENTS 3
1.325 +# define RGBA_OUTPUT_COMPONENTS 4
1.326 +# define OUTPUT_R_INDEX 0
1.327 +# define OUTPUT_G_INDEX 1
1.328 +# define OUTPUT_B_INDEX 2
1.329 +# define OUTPUT_A_INDEX 3
1.330 +#endif
