1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/gfx/cairo/libpixman/src/pixman-combine64.c Wed Dec 31 06:09:35 2014 +0100
1.3 @@ -0,0 +1,2465 @@
1.4 +/* WARNING: This file is generated by combine.pl from combine.inc.
1.5 + Please edit one of those files rather than this one. */
1.6 +
1.7 +#line 1 "pixman-combine.c.template"
1.8 +#ifdef HAVE_CONFIG_H
1.9 +#include <config.h>
1.10 +#endif
1.11 +
1.12 +#include <math.h>
1.13 +#include <string.h>
1.14 +
1.15 +#include "pixman-private.h"
1.16 +
1.17 +#include "pixman-combine64.h"
1.18 +
1.19 +/*** per channel helper functions ***/
1.20 +
1.21 +static void
1.22 +combine_mask_ca (uint64_t *src, uint64_t *mask)
1.23 +{
1.24 + uint64_t a = *mask;
1.25 +
1.26 + uint64_t x;
1.27 + uint32_t xa;
1.28 +
1.29 + if (!a)
1.30 + {
1.31 + *(src) = 0;
1.32 + return;
1.33 + }
1.34 +
1.35 + x = *(src);
1.36 + if (a == ~0)
1.37 + {
1.38 + x = x >> A_SHIFT;
1.39 + x |= x << G_SHIFT;
1.40 + x |= x << R_SHIFT;
1.41 + *(mask) = x;
1.42 + return;
1.43 + }
1.44 +
1.45 + xa = x >> A_SHIFT;
1.46 + UN16x4_MUL_UN16x4 (x, a);
1.47 + *(src) = x;
1.48 +
1.49 + UN16x4_MUL_UN16 (a, xa);
1.50 + *(mask) = a;
1.51 +}
1.52 +
1.53 +static void
1.54 +combine_mask_value_ca (uint64_t *src, const uint64_t *mask)
1.55 +{
1.56 + uint64_t a = *mask;
1.57 + uint64_t x;
1.58 +
1.59 + if (!a)
1.60 + {
1.61 + *(src) = 0;
1.62 + return;
1.63 + }
1.64 +
1.65 + if (a == ~0)
1.66 + return;
1.67 +
1.68 + x = *(src);
1.69 + UN16x4_MUL_UN16x4 (x, a);
1.70 + *(src) = x;
1.71 +}
1.72 +
1.73 +static void
1.74 +combine_mask_alpha_ca (const uint64_t *src, uint64_t *mask)
1.75 +{
1.76 + uint64_t a = *(mask);
1.77 + uint64_t x;
1.78 +
1.79 + if (!a)
1.80 + return;
1.81 +
1.82 + x = *(src) >> A_SHIFT;
1.83 + if (x == MASK)
1.84 + return;
1.85 +
1.86 + if (a == ~0)
1.87 + {
1.88 + x |= x << G_SHIFT;
1.89 + x |= x << R_SHIFT;
1.90 + *(mask) = x;
1.91 + return;
1.92 + }
1.93 +
1.94 + UN16x4_MUL_UN16 (a, x);
1.95 + *(mask) = a;
1.96 +}
1.97 +
1.98 +/*
1.99 + * There are two ways of handling alpha -- either as a single unified value or
1.100 + * a separate value for each component, hence each macro must have two
1.101 + * versions. The unified alpha version has a 'U' at the end of the name,
1.102 + * the component version has a 'C'. Similarly, functions which deal with
1.103 + * this difference will have two versions using the same convention.
1.104 + */
1.105 +
1.106 +/*
1.107 + * All of the composing functions
1.108 + */
1.109 +
1.110 +static force_inline uint64_t
1.111 +combine_mask (const uint64_t *src, const uint64_t *mask, int i)
1.112 +{
1.113 + uint64_t s, m;
1.114 +
1.115 + if (mask)
1.116 + {
1.117 + m = *(mask + i) >> A_SHIFT;
1.118 +
1.119 + if (!m)
1.120 + return 0;
1.121 + }
1.122 +
1.123 + s = *(src + i);
1.124 +
1.125 + if (mask)
1.126 + UN16x4_MUL_UN16 (s, m);
1.127 +
1.128 + return s;
1.129 +}
1.130 +
1.131 +static void
1.132 +combine_clear (pixman_implementation_t *imp,
1.133 + pixman_op_t op,
1.134 + uint64_t * dest,
1.135 + const uint64_t * src,
1.136 + const uint64_t * mask,
1.137 + int width)
1.138 +{
1.139 + memset (dest, 0, width * sizeof(uint64_t));
1.140 +}
1.141 +
1.142 +static void
1.143 +combine_dst (pixman_implementation_t *imp,
1.144 + pixman_op_t op,
1.145 + uint64_t * dest,
1.146 + const uint64_t * src,
1.147 + const uint64_t * mask,
1.148 + int width)
1.149 +{
1.150 + return;
1.151 +}
1.152 +
1.153 +static void
1.154 +combine_src_u (pixman_implementation_t *imp,
1.155 + pixman_op_t op,
1.156 + uint64_t * dest,
1.157 + const uint64_t * src,
1.158 + const uint64_t * mask,
1.159 + int width)
1.160 +{
1.161 + int i;
1.162 +
1.163 + if (!mask)
1.164 + memcpy (dest, src, width * sizeof (uint64_t));
1.165 + else
1.166 + {
1.167 + for (i = 0; i < width; ++i)
1.168 + {
1.169 + uint64_t s = combine_mask (src, mask, i);
1.170 +
1.171 + *(dest + i) = s;
1.172 + }
1.173 + }
1.174 +}
1.175 +
1.176 +/* if the Src is opaque, call combine_src_u */
1.177 +static void
1.178 +combine_over_u (pixman_implementation_t *imp,
1.179 + pixman_op_t op,
1.180 + uint64_t * dest,
1.181 + const uint64_t * src,
1.182 + const uint64_t * mask,
1.183 + int width)
1.184 +{
1.185 + int i;
1.186 +
1.187 + for (i = 0; i < width; ++i)
1.188 + {
1.189 + uint64_t s = combine_mask (src, mask, i);
1.190 + uint64_t d = *(dest + i);
1.191 + uint64_t ia = ALPHA_16 (~s);
1.192 +
1.193 + UN16x4_MUL_UN16_ADD_UN16x4 (d, ia, s);
1.194 + *(dest + i) = d;
1.195 + }
1.196 +}
1.197 +
1.198 +/* if the Dst is opaque, this is a noop */
1.199 +static void
1.200 +combine_over_reverse_u (pixman_implementation_t *imp,
1.201 + pixman_op_t op,
1.202 + uint64_t * dest,
1.203 + const uint64_t * src,
1.204 + const uint64_t * mask,
1.205 + int width)
1.206 +{
1.207 + int i;
1.208 +
1.209 + for (i = 0; i < width; ++i)
1.210 + {
1.211 + uint64_t s = combine_mask (src, mask, i);
1.212 + uint64_t d = *(dest + i);
1.213 + uint64_t ia = ALPHA_16 (~*(dest + i));
1.214 + UN16x4_MUL_UN16_ADD_UN16x4 (s, ia, d);
1.215 + *(dest + i) = s;
1.216 + }
1.217 +}
1.218 +
1.219 +/* if the Dst is opaque, call combine_src_u */
1.220 +static void
1.221 +combine_in_u (pixman_implementation_t *imp,
1.222 + pixman_op_t op,
1.223 + uint64_t * dest,
1.224 + const uint64_t * src,
1.225 + const uint64_t * mask,
1.226 + int width)
1.227 +{
1.228 + int i;
1.229 +
1.230 + for (i = 0; i < width; ++i)
1.231 + {
1.232 + uint64_t s = combine_mask (src, mask, i);
1.233 + uint64_t a = ALPHA_16 (*(dest + i));
1.234 + UN16x4_MUL_UN16 (s, a);
1.235 + *(dest + i) = s;
1.236 + }
1.237 +}
1.238 +
1.239 +/* if the Src is opaque, this is a noop */
1.240 +static void
1.241 +combine_in_reverse_u (pixman_implementation_t *imp,
1.242 + pixman_op_t op,
1.243 + uint64_t * dest,
1.244 + const uint64_t * src,
1.245 + const uint64_t * mask,
1.246 + int width)
1.247 +{
1.248 + int i;
1.249 +
1.250 + for (i = 0; i < width; ++i)
1.251 + {
1.252 + uint64_t s = combine_mask (src, mask, i);
1.253 + uint64_t d = *(dest + i);
1.254 + uint64_t a = ALPHA_16 (s);
1.255 + UN16x4_MUL_UN16 (d, a);
1.256 + *(dest + i) = d;
1.257 + }
1.258 +}
1.259 +
1.260 +/* if the Dst is opaque, call combine_clear */
1.261 +static void
1.262 +combine_out_u (pixman_implementation_t *imp,
1.263 + pixman_op_t op,
1.264 + uint64_t * dest,
1.265 + const uint64_t * src,
1.266 + const uint64_t * mask,
1.267 + int width)
1.268 +{
1.269 + int i;
1.270 +
1.271 + for (i = 0; i < width; ++i)
1.272 + {
1.273 + uint64_t s = combine_mask (src, mask, i);
1.274 + uint64_t a = ALPHA_16 (~*(dest + i));
1.275 + UN16x4_MUL_UN16 (s, a);
1.276 + *(dest + i) = s;
1.277 + }
1.278 +}
1.279 +
1.280 +/* if the Src is opaque, call combine_clear */
1.281 +static void
1.282 +combine_out_reverse_u (pixman_implementation_t *imp,
1.283 + pixman_op_t op,
1.284 + uint64_t * dest,
1.285 + const uint64_t * src,
1.286 + const uint64_t * mask,
1.287 + int width)
1.288 +{
1.289 + int i;
1.290 +
1.291 + for (i = 0; i < width; ++i)
1.292 + {
1.293 + uint64_t s = combine_mask (src, mask, i);
1.294 + uint64_t d = *(dest + i);
1.295 + uint64_t a = ALPHA_16 (~s);
1.296 + UN16x4_MUL_UN16 (d, a);
1.297 + *(dest + i) = d;
1.298 + }
1.299 +}
1.300 +
1.301 +/* if the Src is opaque, call combine_in_u */
1.302 +/* if the Dst is opaque, call combine_over_u */
1.303 +/* if both the Src and Dst are opaque, call combine_src_u */
1.304 +static void
1.305 +combine_atop_u (pixman_implementation_t *imp,
1.306 + pixman_op_t op,
1.307 + uint64_t * dest,
1.308 + const uint64_t * src,
1.309 + const uint64_t * mask,
1.310 + int width)
1.311 +{
1.312 + int i;
1.313 +
1.314 + for (i = 0; i < width; ++i)
1.315 + {
1.316 + uint64_t s = combine_mask (src, mask, i);
1.317 + uint64_t d = *(dest + i);
1.318 + uint64_t dest_a = ALPHA_16 (d);
1.319 + uint64_t src_ia = ALPHA_16 (~s);
1.320 +
1.321 + UN16x4_MUL_UN16_ADD_UN16x4_MUL_UN16 (s, dest_a, d, src_ia);
1.322 + *(dest + i) = s;
1.323 + }
1.324 +}
1.325 +
1.326 +/* if the Src is opaque, call combine_over_reverse_u */
1.327 +/* if the Dst is opaque, call combine_in_reverse_u */
1.328 +/* if both the Src and Dst are opaque, call combine_dst_u */
1.329 +static void
1.330 +combine_atop_reverse_u (pixman_implementation_t *imp,
1.331 + pixman_op_t op,
1.332 + uint64_t * dest,
1.333 + const uint64_t * src,
1.334 + const uint64_t * mask,
1.335 + int width)
1.336 +{
1.337 + int i;
1.338 +
1.339 + for (i = 0; i < width; ++i)
1.340 + {
1.341 + uint64_t s = combine_mask (src, mask, i);
1.342 + uint64_t d = *(dest + i);
1.343 + uint64_t src_a = ALPHA_16 (s);
1.344 + uint64_t dest_ia = ALPHA_16 (~d);
1.345 +
1.346 + UN16x4_MUL_UN16_ADD_UN16x4_MUL_UN16 (s, dest_ia, d, src_a);
1.347 + *(dest + i) = s;
1.348 + }
1.349 +}
1.350 +
1.351 +/* if the Src is opaque, call combine_over_u */
1.352 +/* if the Dst is opaque, call combine_over_reverse_u */
1.353 +/* if both the Src and Dst are opaque, call combine_clear */
1.354 +static void
1.355 +combine_xor_u (pixman_implementation_t *imp,
1.356 + pixman_op_t op,
1.357 + uint64_t * dest,
1.358 + const uint64_t * src,
1.359 + const uint64_t * mask,
1.360 + int width)
1.361 +{
1.362 + int i;
1.363 +
1.364 + for (i = 0; i < width; ++i)
1.365 + {
1.366 + uint64_t s = combine_mask (src, mask, i);
1.367 + uint64_t d = *(dest + i);
1.368 + uint64_t src_ia = ALPHA_16 (~s);
1.369 + uint64_t dest_ia = ALPHA_16 (~d);
1.370 +
1.371 + UN16x4_MUL_UN16_ADD_UN16x4_MUL_UN16 (s, dest_ia, d, src_ia);
1.372 + *(dest + i) = s;
1.373 + }
1.374 +}
1.375 +
1.376 +static void
1.377 +combine_add_u (pixman_implementation_t *imp,
1.378 + pixman_op_t op,
1.379 + uint64_t * dest,
1.380 + const uint64_t * src,
1.381 + const uint64_t * mask,
1.382 + int width)
1.383 +{
1.384 + int i;
1.385 +
1.386 + for (i = 0; i < width; ++i)
1.387 + {
1.388 + uint64_t s = combine_mask (src, mask, i);
1.389 + uint64_t d = *(dest + i);
1.390 + UN16x4_ADD_UN16x4 (d, s);
1.391 + *(dest + i) = d;
1.392 + }
1.393 +}
1.394 +
1.395 +/* if the Src is opaque, call combine_add_u */
1.396 +/* if the Dst is opaque, call combine_add_u */
1.397 +/* if both the Src and Dst are opaque, call combine_add_u */
1.398 +static void
1.399 +combine_saturate_u (pixman_implementation_t *imp,
1.400 + pixman_op_t op,
1.401 + uint64_t * dest,
1.402 + const uint64_t * src,
1.403 + const uint64_t * mask,
1.404 + int width)
1.405 +{
1.406 + int i;
1.407 +
1.408 + for (i = 0; i < width; ++i)
1.409 + {
1.410 + uint64_t s = combine_mask (src, mask, i);
1.411 + uint64_t d = *(dest + i);
1.412 + uint32_t sa, da;
1.413 +
1.414 + sa = s >> A_SHIFT;
1.415 + da = ~d >> A_SHIFT;
1.416 + if (sa > da)
1.417 + {
1.418 + sa = DIV_UN16 (da, sa);
1.419 + UN16x4_MUL_UN16 (s, sa);
1.420 + }
1.421 + ;
1.422 + UN16x4_ADD_UN16x4 (d, s);
1.423 + *(dest + i) = d;
1.424 + }
1.425 +}
1.426 +
1.427 +/*
1.428 + * PDF blend modes:
1.429 + * The following blend modes have been taken from the PDF ISO 32000
1.430 + * specification, which at this point in time is available from
1.431 + * http://www.adobe.com/devnet/acrobat/pdfs/PDF32000_2008.pdf
1.432 + * The relevant chapters are 11.3.5 and 11.3.6.
1.433 + * The formula for computing the final pixel color given in 11.3.6 is:
1.434 + * αr × Cr = (1 – αs) × αb × Cb + (1 – αb) × αs × Cs + αb × αs × B(Cb, Cs)
1.435 + * with B() being the blend function.
1.436 + * Note that OVER is a special case of this operation, using B(Cb, Cs) = Cs
1.437 + *
1.438 + * These blend modes should match the SVG filter draft specification, as
1.439 + * it has been designed to mirror ISO 32000. Note that at the current point
1.440 + * no released draft exists that shows this, as the formulas have not been
1.441 + * updated yet after the release of ISO 32000.
1.442 + *
1.443 + * The default implementation here uses the PDF_SEPARABLE_BLEND_MODE and
1.444 + * PDF_NON_SEPARABLE_BLEND_MODE macros, which take the blend function as an
1.445 + * argument. Note that this implementation operates on premultiplied colors,
1.446 + * while the PDF specification does not. Therefore the code uses the formula
1.447 + * Cra = (1 – as) . Dca + (1 – ad) . Sca + B(Dca, ad, Sca, as)
1.448 + */
1.449 +
1.450 +/*
1.451 + * Multiply
1.452 + * B(Dca, ad, Sca, as) = Dca.Sca
1.453 + */
1.454 +
1.455 +static void
1.456 +combine_multiply_u (pixman_implementation_t *imp,
1.457 + pixman_op_t op,
1.458 + uint64_t * dest,
1.459 + const uint64_t * src,
1.460 + const uint64_t * mask,
1.461 + int width)
1.462 +{
1.463 + int i;
1.464 +
1.465 + for (i = 0; i < width; ++i)
1.466 + {
1.467 + uint64_t s = combine_mask (src, mask, i);
1.468 + uint64_t d = *(dest + i);
1.469 + uint64_t ss = s;
1.470 + uint64_t src_ia = ALPHA_16 (~s);
1.471 + uint64_t dest_ia = ALPHA_16 (~d);
1.472 +
1.473 + UN16x4_MUL_UN16_ADD_UN16x4_MUL_UN16 (ss, dest_ia, d, src_ia);
1.474 + UN16x4_MUL_UN16x4 (d, s);
1.475 + UN16x4_ADD_UN16x4 (d, ss);
1.476 +
1.477 + *(dest + i) = d;
1.478 + }
1.479 +}
1.480 +
1.481 +static void
1.482 +combine_multiply_ca (pixman_implementation_t *imp,
1.483 + pixman_op_t op,
1.484 + uint64_t * dest,
1.485 + const uint64_t * src,
1.486 + const uint64_t * mask,
1.487 + int width)
1.488 +{
1.489 + int i;
1.490 +
1.491 + for (i = 0; i < width; ++i)
1.492 + {
1.493 + uint64_t m = *(mask + i);
1.494 + uint64_t s = *(src + i);
1.495 + uint64_t d = *(dest + i);
1.496 + uint64_t r = d;
1.497 + uint64_t dest_ia = ALPHA_16 (~d);
1.498 +
1.499 + combine_mask_value_ca (&s, &m);
1.500 +
1.501 + UN16x4_MUL_UN16x4_ADD_UN16x4_MUL_UN16 (r, ~m, s, dest_ia);
1.502 + UN16x4_MUL_UN16x4 (d, s);
1.503 + UN16x4_ADD_UN16x4 (r, d);
1.504 +
1.505 + *(dest + i) = r;
1.506 + }
1.507 +}
1.508 +
1.509 +#define PDF_SEPARABLE_BLEND_MODE(name) \
1.510 + static void \
1.511 + combine_ ## name ## _u (pixman_implementation_t *imp, \
1.512 + pixman_op_t op, \
1.513 + uint64_t * dest, \
1.514 + const uint64_t * src, \
1.515 + const uint64_t * mask, \
1.516 + int width) \
1.517 + { \
1.518 + int i; \
1.519 + for (i = 0; i < width; ++i) { \
1.520 + uint64_t s = combine_mask (src, mask, i); \
1.521 + uint64_t d = *(dest + i); \
1.522 + uint16_t sa = ALPHA_16 (s); \
1.523 + uint16_t isa = ~sa; \
1.524 + uint16_t da = ALPHA_16 (d); \
1.525 + uint16_t ida = ~da; \
1.526 + uint64_t result; \
1.527 + \
1.528 + result = d; \
1.529 + UN16x4_MUL_UN16_ADD_UN16x4_MUL_UN16 (result, isa, s, ida); \
1.530 + \
1.531 + *(dest + i) = result + \
1.532 + (DIV_ONE_UN16 (sa * (uint64_t)da) << A_SHIFT) + \
1.533 + (blend_ ## name (RED_16 (d), da, RED_16 (s), sa) << R_SHIFT) + \
1.534 + (blend_ ## name (GREEN_16 (d), da, GREEN_16 (s), sa) << G_SHIFT) + \
1.535 + (blend_ ## name (BLUE_16 (d), da, BLUE_16 (s), sa)); \
1.536 + } \
1.537 + } \
1.538 + \
1.539 + static void \
1.540 + combine_ ## name ## _ca (pixman_implementation_t *imp, \
1.541 + pixman_op_t op, \
1.542 + uint64_t * dest, \
1.543 + const uint64_t * src, \
1.544 + const uint64_t * mask, \
1.545 + int width) \
1.546 + { \
1.547 + int i; \
1.548 + for (i = 0; i < width; ++i) { \
1.549 + uint64_t m = *(mask + i); \
1.550 + uint64_t s = *(src + i); \
1.551 + uint64_t d = *(dest + i); \
1.552 + uint16_t da = ALPHA_16 (d); \
1.553 + uint16_t ida = ~da; \
1.554 + uint64_t result; \
1.555 + \
1.556 + combine_mask_value_ca (&s, &m); \
1.557 + \
1.558 + result = d; \
1.559 + UN16x4_MUL_UN16x4_ADD_UN16x4_MUL_UN16 (result, ~m, s, ida); \
1.560 + \
1.561 + result += \
1.562 + (DIV_ONE_UN16 (ALPHA_16 (m) * (uint64_t)da) << A_SHIFT) + \
1.563 + (blend_ ## name (RED_16 (d), da, RED_16 (s), RED_16 (m)) << R_SHIFT) + \
1.564 + (blend_ ## name (GREEN_16 (d), da, GREEN_16 (s), GREEN_16 (m)) << G_SHIFT) + \
1.565 + (blend_ ## name (BLUE_16 (d), da, BLUE_16 (s), BLUE_16 (m))); \
1.566 + \
1.567 + *(dest + i) = result; \
1.568 + } \
1.569 + }
1.570 +
1.571 +/*
1.572 + * Screen
1.573 + * B(Dca, ad, Sca, as) = Dca.sa + Sca.da - Dca.Sca
1.574 + */
1.575 +static inline uint64_t
1.576 +blend_screen (uint64_t dca, uint64_t da, uint64_t sca, uint64_t sa)
1.577 +{
1.578 + return DIV_ONE_UN16 (sca * da + dca * sa - sca * dca);
1.579 +}
1.580 +
1.581 +PDF_SEPARABLE_BLEND_MODE (screen)
1.582 +
1.583 +/*
1.584 + * Overlay
1.585 + * B(Dca, Da, Sca, Sa) =
1.586 + * if 2.Dca < Da
1.587 + * 2.Sca.Dca
1.588 + * otherwise
1.589 + * Sa.Da - 2.(Da - Dca).(Sa - Sca)
1.590 + */
1.591 +static inline uint64_t
1.592 +blend_overlay (uint64_t dca, uint64_t da, uint64_t sca, uint64_t sa)
1.593 +{
1.594 + uint64_t rca;
1.595 +
1.596 + if (2 * dca < da)
1.597 + rca = 2 * sca * dca;
1.598 + else
1.599 + rca = sa * da - 2 * (da - dca) * (sa - sca);
1.600 + return DIV_ONE_UN16 (rca);
1.601 +}
1.602 +
1.603 +PDF_SEPARABLE_BLEND_MODE (overlay)
1.604 +
1.605 +/*
1.606 + * Darken
1.607 + * B(Dca, Da, Sca, Sa) = min (Sca.Da, Dca.Sa)
1.608 + */
1.609 +static inline uint64_t
1.610 +blend_darken (uint64_t dca, uint64_t da, uint64_t sca, uint64_t sa)
1.611 +{
1.612 + uint64_t s, d;
1.613 +
1.614 + s = sca * da;
1.615 + d = dca * sa;
1.616 + return DIV_ONE_UN16 (s > d ? d : s);
1.617 +}
1.618 +
1.619 +PDF_SEPARABLE_BLEND_MODE (darken)
1.620 +
1.621 +/*
1.622 + * Lighten
1.623 + * B(Dca, Da, Sca, Sa) = max (Sca.Da, Dca.Sa)
1.624 + */
1.625 +static inline uint64_t
1.626 +blend_lighten (uint64_t dca, uint64_t da, uint64_t sca, uint64_t sa)
1.627 +{
1.628 + uint64_t s, d;
1.629 +
1.630 + s = sca * da;
1.631 + d = dca * sa;
1.632 + return DIV_ONE_UN16 (s > d ? s : d);
1.633 +}
1.634 +
1.635 +PDF_SEPARABLE_BLEND_MODE (lighten)
1.636 +
1.637 +/*
1.638 + * Color dodge
1.639 + * B(Dca, Da, Sca, Sa) =
1.640 + * if Dca == 0
1.641 + * 0
1.642 + * if Sca == Sa
1.643 + * Sa.Da
1.644 + * otherwise
1.645 + * Sa.Da. min (1, Dca / Da / (1 - Sca/Sa))
1.646 + */
1.647 +static inline uint64_t
1.648 +blend_color_dodge (uint64_t dca, uint64_t da, uint64_t sca, uint64_t sa)
1.649 +{
1.650 + if (sca >= sa)
1.651 + {
1.652 + return dca == 0 ? 0 : DIV_ONE_UN16 (sa * da);
1.653 + }
1.654 + else
1.655 + {
1.656 + uint64_t rca = dca * sa / (sa - sca);
1.657 + return DIV_ONE_UN16 (sa * MIN (rca, da));
1.658 + }
1.659 +}
1.660 +
1.661 +PDF_SEPARABLE_BLEND_MODE (color_dodge)
1.662 +
1.663 +/*
1.664 + * Color burn
1.665 + * B(Dca, Da, Sca, Sa) =
1.666 + * if Dca == Da
1.667 + * Sa.Da
1.668 + * if Sca == 0
1.669 + * 0
1.670 + * otherwise
1.671 + * Sa.Da.(1 - min (1, (1 - Dca/Da).Sa / Sca))
1.672 + */
1.673 +static inline uint64_t
1.674 +blend_color_burn (uint64_t dca, uint64_t da, uint64_t sca, uint64_t sa)
1.675 +{
1.676 + if (sca == 0)
1.677 + {
1.678 + return dca < da ? 0 : DIV_ONE_UN16 (sa * da);
1.679 + }
1.680 + else
1.681 + {
1.682 + uint64_t rca = (da - dca) * sa / sca;
1.683 + return DIV_ONE_UN16 (sa * (MAX (rca, da) - rca));
1.684 + }
1.685 +}
1.686 +
1.687 +PDF_SEPARABLE_BLEND_MODE (color_burn)
1.688 +
1.689 +/*
1.690 + * Hard light
1.691 + * B(Dca, Da, Sca, Sa) =
1.692 + * if 2.Sca < Sa
1.693 + * 2.Sca.Dca
1.694 + * otherwise
1.695 + * Sa.Da - 2.(Da - Dca).(Sa - Sca)
1.696 + */
1.697 +static inline uint64_t
1.698 +blend_hard_light (uint64_t dca, uint64_t da, uint64_t sca, uint64_t sa)
1.699 +{
1.700 + if (2 * sca < sa)
1.701 + return DIV_ONE_UN16 (2 * sca * dca);
1.702 + else
1.703 + return DIV_ONE_UN16 (sa * da - 2 * (da - dca) * (sa - sca));
1.704 +}
1.705 +
1.706 +PDF_SEPARABLE_BLEND_MODE (hard_light)
1.707 +
1.708 +/*
1.709 + * Soft light
1.710 + * B(Dca, Da, Sca, Sa) =
1.711 + * if (2.Sca <= Sa)
1.712 + * Dca.(Sa - (1 - Dca/Da).(2.Sca - Sa))
1.713 + * otherwise if Dca.4 <= Da
1.714 + * Dca.(Sa + (2.Sca - Sa).((16.Dca/Da - 12).Dca/Da + 3)
1.715 + * otherwise
1.716 + * (Dca.Sa + (SQRT (Dca/Da).Da - Dca).(2.Sca - Sa))
1.717 + */
1.718 +static inline uint64_t
1.719 +blend_soft_light (uint64_t dca_org,
1.720 + uint64_t da_org,
1.721 + uint64_t sca_org,
1.722 + uint64_t sa_org)
1.723 +{
1.724 + double dca = dca_org * (1.0 / MASK);
1.725 + double da = da_org * (1.0 / MASK);
1.726 + double sca = sca_org * (1.0 / MASK);
1.727 + double sa = sa_org * (1.0 / MASK);
1.728 + double rca;
1.729 +
1.730 + if (2 * sca < sa)
1.731 + {
1.732 + if (da == 0)
1.733 + rca = dca * sa;
1.734 + else
1.735 + rca = dca * sa - dca * (da - dca) * (sa - 2 * sca) / da;
1.736 + }
1.737 + else if (da == 0)
1.738 + {
1.739 + rca = 0;
1.740 + }
1.741 + else if (4 * dca <= da)
1.742 + {
1.743 + rca = dca * sa +
1.744 + (2 * sca - sa) * dca * ((16 * dca / da - 12) * dca / da + 3);
1.745 + }
1.746 + else
1.747 + {
1.748 + rca = dca * sa + (sqrt (dca * da) - dca) * (2 * sca - sa);
1.749 + }
1.750 + return rca * MASK + 0.5;
1.751 +}
1.752 +
1.753 +PDF_SEPARABLE_BLEND_MODE (soft_light)
1.754 +
1.755 +/*
1.756 + * Difference
1.757 + * B(Dca, Da, Sca, Sa) = abs (Dca.Sa - Sca.Da)
1.758 + */
1.759 +static inline uint64_t
1.760 +blend_difference (uint64_t dca, uint64_t da, uint64_t sca, uint64_t sa)
1.761 +{
1.762 + uint64_t dcasa = dca * sa;
1.763 + uint64_t scada = sca * da;
1.764 +
1.765 + if (scada < dcasa)
1.766 + return DIV_ONE_UN16 (dcasa - scada);
1.767 + else
1.768 + return DIV_ONE_UN16 (scada - dcasa);
1.769 +}
1.770 +
1.771 +PDF_SEPARABLE_BLEND_MODE (difference)
1.772 +
1.773 +/*
1.774 + * Exclusion
1.775 + * B(Dca, Da, Sca, Sa) = (Sca.Da + Dca.Sa - 2.Sca.Dca)
1.776 + */
1.777 +
1.778 +/* This can be made faster by writing it directly and not using
1.779 + * PDF_SEPARABLE_BLEND_MODE, but that's a performance optimization */
1.780 +
1.781 +static inline uint64_t
1.782 +blend_exclusion (uint64_t dca, uint64_t da, uint64_t sca, uint64_t sa)
1.783 +{
1.784 + return DIV_ONE_UN16 (sca * da + dca * sa - 2 * dca * sca);
1.785 +}
1.786 +
1.787 +PDF_SEPARABLE_BLEND_MODE (exclusion)
1.788 +
1.789 +#undef PDF_SEPARABLE_BLEND_MODE
1.790 +
1.791 +/*
1.792 + * PDF nonseperable blend modes are implemented using the following functions
1.793 + * to operate in Hsl space, with Cmax, Cmid, Cmin referring to the max, mid
1.794 + * and min value of the red, green and blue components.
1.795 + *
1.796 + * LUM (C) = 0.3 × Cred + 0.59 × Cgreen + 0.11 × Cblue
1.797 + *
1.798 + * clip_color (C):
1.799 + * l = LUM (C)
1.800 + * min = Cmin
1.801 + * max = Cmax
1.802 + * if n < 0.0
1.803 + * C = l + ( ( ( C – l ) × l ) ⁄ ( l – min ) )
1.804 + * if x > 1.0
1.805 + * C = l + ( ( ( C – l ) × ( 1 – l ) ) ⁄ ( max – l ) )
1.806 + * return C
1.807 + *
1.808 + * set_lum (C, l):
1.809 + * d = l – LUM (C)
1.810 + * C += d
1.811 + * return clip_color (C)
1.812 + *
1.813 + * SAT (C) = CH_MAX (C) - CH_MIN (C)
1.814 + *
1.815 + * set_sat (C, s):
1.816 + * if Cmax > Cmin
1.817 + * Cmid = ( ( ( Cmid – Cmin ) × s ) ⁄ ( Cmax – Cmin ) )
1.818 + * Cmax = s
1.819 + * else
1.820 + * Cmid = Cmax = 0.0
1.821 + * Cmin = 0.0
1.822 + * return C
1.823 + */
1.824 +
1.825 +/* For premultiplied colors, we need to know what happens when C is
1.826 + * multiplied by a real number. LUM and SAT are linear:
1.827 + *
1.828 + * LUM (r × C) = r × LUM (C) SAT (r * C) = r * SAT (C)
1.829 + *
1.830 + * If we extend clip_color with an extra argument a and change
1.831 + *
1.832 + * if x >= 1.0
1.833 + *
1.834 + * into
1.835 + *
1.836 + * if x >= a
1.837 + *
1.838 + * then clip_color is also linear:
1.839 + *
1.840 + * r * clip_color (C, a) = clip_color (r_c, ra);
1.841 + *
1.842 + * for positive r.
1.843 + *
1.844 + * Similarly, we can extend set_lum with an extra argument that is just passed
1.845 + * on to clip_color:
1.846 + *
1.847 + * r * set_lum ( C, l, a)
1.848 + *
1.849 + * = r × clip_color ( C + l - LUM (C), a)
1.850 + *
1.851 + * = clip_color ( r * C + r × l - r * LUM (C), r * a)
1.852 + *
1.853 + * = set_lum ( r * C, r * l, r * a)
1.854 + *
1.855 + * Finally, set_sat:
1.856 + *
1.857 + * r * set_sat (C, s) = set_sat (x * C, r * s)
1.858 + *
1.859 + * The above holds for all non-zero x, because the x'es in the fraction for
1.860 + * C_mid cancel out. Specifically, it holds for x = r:
1.861 + *
1.862 + * r * set_sat (C, s) = set_sat (r_c, rs)
1.863 + *
1.864 + */
1.865 +
1.866 +/* So, for the non-separable PDF blend modes, we have (using s, d for
1.867 + * non-premultiplied colors, and S, D for premultiplied:
1.868 + *
1.869 + * Color:
1.870 + *
1.871 + * a_s * a_d * B(s, d)
1.872 + * = a_s * a_d * set_lum (S/a_s, LUM (D/a_d), 1)
1.873 + * = set_lum (S * a_d, a_s * LUM (D), a_s * a_d)
1.874 + *
1.875 + *
1.876 + * Luminosity:
1.877 + *
1.878 + * a_s * a_d * B(s, d)
1.879 + * = a_s * a_d * set_lum (D/a_d, LUM(S/a_s), 1)
1.880 + * = set_lum (a_s * D, a_d * LUM(S), a_s * a_d)
1.881 + *
1.882 + *
1.883 + * Saturation:
1.884 + *
1.885 + * a_s * a_d * B(s, d)
1.886 + * = a_s * a_d * set_lum (set_sat (D/a_d, SAT (S/a_s)), LUM (D/a_d), 1)
1.887 + * = set_lum (a_s * a_d * set_sat (D/a_d, SAT (S/a_s)),
1.888 + * a_s * LUM (D), a_s * a_d)
1.889 + * = set_lum (set_sat (a_s * D, a_d * SAT (S), a_s * LUM (D), a_s * a_d))
1.890 + *
1.891 + * Hue:
1.892 + *
1.893 + * a_s * a_d * B(s, d)
1.894 + * = a_s * a_d * set_lum (set_sat (S/a_s, SAT (D/a_d)), LUM (D/a_d), 1)
1.895 + * = set_lum (set_sat (a_d * S, a_s * SAT (D)), a_s * LUM (D), a_s * a_d)
1.896 + *
1.897 + */
1.898 +
1.899 +#define CH_MIN(c) (c[0] < c[1] ? (c[0] < c[2] ? c[0] : c[2]) : (c[1] < c[2] ? c[1] : c[2]))
1.900 +#define CH_MAX(c) (c[0] > c[1] ? (c[0] > c[2] ? c[0] : c[2]) : (c[1] > c[2] ? c[1] : c[2]))
1.901 +#define LUM(c) ((c[0] * 30 + c[1] * 59 + c[2] * 11) / 100)
1.902 +#define SAT(c) (CH_MAX (c) - CH_MIN (c))
1.903 +
1.904 +#define PDF_NON_SEPARABLE_BLEND_MODE(name) \
1.905 + static void \
1.906 + combine_ ## name ## _u (pixman_implementation_t *imp, \
1.907 + pixman_op_t op, \
1.908 + uint64_t *dest, \
1.909 + const uint64_t *src, \
1.910 + const uint64_t *mask, \
1.911 + int width) \
1.912 + { \
1.913 + int i; \
1.914 + for (i = 0; i < width; ++i) \
1.915 + { \
1.916 + uint64_t s = combine_mask (src, mask, i); \
1.917 + uint64_t d = *(dest + i); \
1.918 + uint16_t sa = ALPHA_16 (s); \
1.919 + uint16_t isa = ~sa; \
1.920 + uint16_t da = ALPHA_16 (d); \
1.921 + uint16_t ida = ~da; \
1.922 + uint64_t result; \
1.923 + uint64_t sc[3], dc[3], c[3]; \
1.924 + \
1.925 + result = d; \
1.926 + UN16x4_MUL_UN16_ADD_UN16x4_MUL_UN16 (result, isa, s, ida); \
1.927 + dc[0] = RED_16 (d); \
1.928 + sc[0] = RED_16 (s); \
1.929 + dc[1] = GREEN_16 (d); \
1.930 + sc[1] = GREEN_16 (s); \
1.931 + dc[2] = BLUE_16 (d); \
1.932 + sc[2] = BLUE_16 (s); \
1.933 + blend_ ## name (c, dc, da, sc, sa); \
1.934 + \
1.935 + *(dest + i) = result + \
1.936 + (DIV_ONE_UN16 (sa * (uint64_t)da) << A_SHIFT) + \
1.937 + (DIV_ONE_UN16 (c[0]) << R_SHIFT) + \
1.938 + (DIV_ONE_UN16 (c[1]) << G_SHIFT) + \
1.939 + (DIV_ONE_UN16 (c[2])); \
1.940 + } \
1.941 + }
1.942 +
1.943 +static void
1.944 +set_lum (uint64_t dest[3], uint64_t src[3], uint64_t sa, uint64_t lum)
1.945 +{
1.946 + double a, l, min, max;
1.947 + double tmp[3];
1.948 +
1.949 + a = sa * (1.0 / MASK);
1.950 +
1.951 + l = lum * (1.0 / MASK);
1.952 + tmp[0] = src[0] * (1.0 / MASK);
1.953 + tmp[1] = src[1] * (1.0 / MASK);
1.954 + tmp[2] = src[2] * (1.0 / MASK);
1.955 +
1.956 + l = l - LUM (tmp);
1.957 + tmp[0] += l;
1.958 + tmp[1] += l;
1.959 + tmp[2] += l;
1.960 +
1.961 + /* clip_color */
1.962 + l = LUM (tmp);
1.963 + min = CH_MIN (tmp);
1.964 + max = CH_MAX (tmp);
1.965 +
1.966 + if (min < 0)
1.967 + {
1.968 + if (l - min == 0.0)
1.969 + {
1.970 + tmp[0] = 0;
1.971 + tmp[1] = 0;
1.972 + tmp[2] = 0;
1.973 + }
1.974 + else
1.975 + {
1.976 + tmp[0] = l + (tmp[0] - l) * l / (l - min);
1.977 + tmp[1] = l + (tmp[1] - l) * l / (l - min);
1.978 + tmp[2] = l + (tmp[2] - l) * l / (l - min);
1.979 + }
1.980 + }
1.981 + if (max > a)
1.982 + {
1.983 + if (max - l == 0.0)
1.984 + {
1.985 + tmp[0] = a;
1.986 + tmp[1] = a;
1.987 + tmp[2] = a;
1.988 + }
1.989 + else
1.990 + {
1.991 + tmp[0] = l + (tmp[0] - l) * (a - l) / (max - l);
1.992 + tmp[1] = l + (tmp[1] - l) * (a - l) / (max - l);
1.993 + tmp[2] = l + (tmp[2] - l) * (a - l) / (max - l);
1.994 + }
1.995 + }
1.996 +
1.997 + dest[0] = tmp[0] * MASK + 0.5;
1.998 + dest[1] = tmp[1] * MASK + 0.5;
1.999 + dest[2] = tmp[2] * MASK + 0.5;
1.1000 +}
1.1001 +
1.1002 +static void
1.1003 +set_sat (uint64_t dest[3], uint64_t src[3], uint64_t sat)
1.1004 +{
1.1005 + int id[3];
1.1006 + uint64_t min, max;
1.1007 +
1.1008 + if (src[0] > src[1])
1.1009 + {
1.1010 + if (src[0] > src[2])
1.1011 + {
1.1012 + id[0] = 0;
1.1013 + if (src[1] > src[2])
1.1014 + {
1.1015 + id[1] = 1;
1.1016 + id[2] = 2;
1.1017 + }
1.1018 + else
1.1019 + {
1.1020 + id[1] = 2;
1.1021 + id[2] = 1;
1.1022 + }
1.1023 + }
1.1024 + else
1.1025 + {
1.1026 + id[0] = 2;
1.1027 + id[1] = 0;
1.1028 + id[2] = 1;
1.1029 + }
1.1030 + }
1.1031 + else
1.1032 + {
1.1033 + if (src[0] > src[2])
1.1034 + {
1.1035 + id[0] = 1;
1.1036 + id[1] = 0;
1.1037 + id[2] = 2;
1.1038 + }
1.1039 + else
1.1040 + {
1.1041 + id[2] = 0;
1.1042 + if (src[1] > src[2])
1.1043 + {
1.1044 + id[0] = 1;
1.1045 + id[1] = 2;
1.1046 + }
1.1047 + else
1.1048 + {
1.1049 + id[0] = 2;
1.1050 + id[1] = 1;
1.1051 + }
1.1052 + }
1.1053 + }
1.1054 +
1.1055 + max = dest[id[0]];
1.1056 + min = dest[id[2]];
1.1057 + if (max > min)
1.1058 + {
1.1059 + dest[id[1]] = (dest[id[1]] - min) * sat / (max - min);
1.1060 + dest[id[0]] = sat;
1.1061 + dest[id[2]] = 0;
1.1062 + }
1.1063 + else
1.1064 + {
1.1065 + dest[0] = dest[1] = dest[2] = 0;
1.1066 + }
1.1067 +}
1.1068 +
1.1069 +/*
1.1070 + * Hue:
1.1071 + * B(Cb, Cs) = set_lum (set_sat (Cs, SAT (Cb)), LUM (Cb))
1.1072 + */
1.1073 +static inline void
1.1074 +blend_hsl_hue (uint64_t c[3],
1.1075 + uint64_t dc[3],
1.1076 + uint64_t da,
1.1077 + uint64_t sc[3],
1.1078 + uint64_t sa)
1.1079 +{
1.1080 + c[0] = sc[0] * da;
1.1081 + c[1] = sc[1] * da;
1.1082 + c[2] = sc[2] * da;
1.1083 + set_sat (c, c, SAT (dc) * sa);
1.1084 + set_lum (c, c, sa * da, LUM (dc) * sa);
1.1085 +}
1.1086 +
1.1087 +PDF_NON_SEPARABLE_BLEND_MODE (hsl_hue)
1.1088 +
1.1089 +/*
1.1090 + * Saturation:
1.1091 + * B(Cb, Cs) = set_lum (set_sat (Cb, SAT (Cs)), LUM (Cb))
1.1092 + */
1.1093 +static inline void
1.1094 +blend_hsl_saturation (uint64_t c[3],
1.1095 + uint64_t dc[3],
1.1096 + uint64_t da,
1.1097 + uint64_t sc[3],
1.1098 + uint64_t sa)
1.1099 +{
1.1100 + c[0] = dc[0] * sa;
1.1101 + c[1] = dc[1] * sa;
1.1102 + c[2] = dc[2] * sa;
1.1103 + set_sat (c, c, SAT (sc) * da);
1.1104 + set_lum (c, c, sa * da, LUM (dc) * sa);
1.1105 +}
1.1106 +
1.1107 +PDF_NON_SEPARABLE_BLEND_MODE (hsl_saturation)
1.1108 +
1.1109 +/*
1.1110 + * Color:
1.1111 + * B(Cb, Cs) = set_lum (Cs, LUM (Cb))
1.1112 + */
1.1113 +static inline void
1.1114 +blend_hsl_color (uint64_t c[3],
1.1115 + uint64_t dc[3],
1.1116 + uint64_t da,
1.1117 + uint64_t sc[3],
1.1118 + uint64_t sa)
1.1119 +{
1.1120 + c[0] = sc[0] * da;
1.1121 + c[1] = sc[1] * da;
1.1122 + c[2] = sc[2] * da;
1.1123 + set_lum (c, c, sa * da, LUM (dc) * sa);
1.1124 +}
1.1125 +
1.1126 +PDF_NON_SEPARABLE_BLEND_MODE (hsl_color)
1.1127 +
1.1128 +/*
1.1129 + * Luminosity:
1.1130 + * B(Cb, Cs) = set_lum (Cb, LUM (Cs))
1.1131 + */
1.1132 +static inline void
1.1133 +blend_hsl_luminosity (uint64_t c[3],
1.1134 + uint64_t dc[3],
1.1135 + uint64_t da,
1.1136 + uint64_t sc[3],
1.1137 + uint64_t sa)
1.1138 +{
1.1139 + c[0] = dc[0] * sa;
1.1140 + c[1] = dc[1] * sa;
1.1141 + c[2] = dc[2] * sa;
1.1142 + set_lum (c, c, sa * da, LUM (sc) * da);
1.1143 +}
1.1144 +
1.1145 +PDF_NON_SEPARABLE_BLEND_MODE (hsl_luminosity)
1.1146 +
1.1147 +#undef SAT
1.1148 +#undef LUM
1.1149 +#undef CH_MAX
1.1150 +#undef CH_MIN
1.1151 +#undef PDF_NON_SEPARABLE_BLEND_MODE
1.1152 +
1.1153 +/* All of the disjoint/conjoint composing functions
1.1154 + *
1.1155 + * The four entries in the first column indicate what source contributions
1.1156 + * come from each of the four areas of the picture -- areas covered by neither
1.1157 + * A nor B, areas covered only by A, areas covered only by B and finally
1.1158 + * areas covered by both A and B.
1.1159 + *
1.1160 + * Disjoint Conjoint
1.1161 + * Fa Fb Fa Fb
1.1162 + * (0,0,0,0) 0 0 0 0
1.1163 + * (0,A,0,A) 1 0 1 0
1.1164 + * (0,0,B,B) 0 1 0 1
1.1165 + * (0,A,B,A) 1 min((1-a)/b,1) 1 max(1-a/b,0)
1.1166 + * (0,A,B,B) min((1-b)/a,1) 1 max(1-b/a,0) 1
1.1167 + * (0,0,0,A) max(1-(1-b)/a,0) 0 min(1,b/a) 0
1.1168 + * (0,0,0,B) 0 max(1-(1-a)/b,0) 0 min(a/b,1)
1.1169 + * (0,A,0,0) min(1,(1-b)/a) 0 max(1-b/a,0) 0
1.1170 + * (0,0,B,0) 0 min(1,(1-a)/b) 0 max(1-a/b,0)
1.1171 + * (0,0,B,A) max(1-(1-b)/a,0) min(1,(1-a)/b) min(1,b/a) max(1-a/b,0)
1.1172 + * (0,A,0,B) min(1,(1-b)/a) max(1-(1-a)/b,0) max(1-b/a,0) min(1,a/b)
1.1173 + * (0,A,B,0) min(1,(1-b)/a) min(1,(1-a)/b) max(1-b/a,0) max(1-a/b,0)
1.1174 + *
1.1175 + * See http://marc.info/?l=xfree-render&m=99792000027857&w=2 for more
1.1176 + * information about these operators.
1.1177 + */
1.1178 +
1.1179 +#define COMBINE_A_OUT 1
1.1180 +#define COMBINE_A_IN 2
1.1181 +#define COMBINE_B_OUT 4
1.1182 +#define COMBINE_B_IN 8
1.1183 +
1.1184 +#define COMBINE_CLEAR 0
1.1185 +#define COMBINE_A (COMBINE_A_OUT | COMBINE_A_IN)
1.1186 +#define COMBINE_B (COMBINE_B_OUT | COMBINE_B_IN)
1.1187 +#define COMBINE_A_OVER (COMBINE_A_OUT | COMBINE_B_OUT | COMBINE_A_IN)
1.1188 +#define COMBINE_B_OVER (COMBINE_A_OUT | COMBINE_B_OUT | COMBINE_B_IN)
1.1189 +#define COMBINE_A_ATOP (COMBINE_B_OUT | COMBINE_A_IN)
1.1190 +#define COMBINE_B_ATOP (COMBINE_A_OUT | COMBINE_B_IN)
1.1191 +#define COMBINE_XOR (COMBINE_A_OUT | COMBINE_B_OUT)
1.1192 +
1.1193 +/* portion covered by a but not b */
1.1194 +static uint16_t
1.1195 +combine_disjoint_out_part (uint16_t a, uint16_t b)
1.1196 +{
1.1197 + /* min (1, (1-b) / a) */
1.1198 +
1.1199 + b = ~b; /* 1 - b */
1.1200 + if (b >= a) /* 1 - b >= a -> (1-b)/a >= 1 */
1.1201 + return MASK; /* 1 */
1.1202 + return DIV_UN16 (b, a); /* (1-b) / a */
1.1203 +}
1.1204 +
1.1205 +/* portion covered by both a and b */
1.1206 +static uint16_t
1.1207 +combine_disjoint_in_part (uint16_t a, uint16_t b)
1.1208 +{
1.1209 + /* max (1-(1-b)/a,0) */
1.1210 + /* = - min ((1-b)/a - 1, 0) */
1.1211 + /* = 1 - min (1, (1-b)/a) */
1.1212 +
1.1213 + b = ~b; /* 1 - b */
1.1214 + if (b >= a) /* 1 - b >= a -> (1-b)/a >= 1 */
1.1215 + return 0; /* 1 - 1 */
1.1216 + return ~DIV_UN16(b, a); /* 1 - (1-b) / a */
1.1217 +}
1.1218 +
1.1219 +/* portion covered by a but not b */
1.1220 +static uint16_t
1.1221 +combine_conjoint_out_part (uint16_t a, uint16_t b)
1.1222 +{
1.1223 + /* max (1-b/a,0) */
1.1224 + /* = 1-min(b/a,1) */
1.1225 +
1.1226 + /* min (1, (1-b) / a) */
1.1227 +
1.1228 + if (b >= a) /* b >= a -> b/a >= 1 */
1.1229 + return 0x00; /* 0 */
1.1230 + return ~DIV_UN16(b, a); /* 1 - b/a */
1.1231 +}
1.1232 +
1.1233 +/* portion covered by both a and b */
1.1234 +static uint16_t
1.1235 +combine_conjoint_in_part (uint16_t a, uint16_t b)
1.1236 +{
1.1237 + /* min (1,b/a) */
1.1238 +
1.1239 + if (b >= a) /* b >= a -> b/a >= 1 */
1.1240 + return MASK; /* 1 */
1.1241 + return DIV_UN16 (b, a); /* b/a */
1.1242 +}
1.1243 +
1.1244 +#define GET_COMP(v, i) ((uint32_t) (uint16_t) ((v) >> i))
1.1245 +
1.1246 +#define ADD(x, y, i, t) \
1.1247 + ((t) = GET_COMP (x, i) + GET_COMP (y, i), \
1.1248 + (uint64_t) ((uint16_t) ((t) | (0 - ((t) >> G_SHIFT)))) << (i))
1.1249 +
1.1250 +#define GENERIC(x, y, i, ax, ay, t, u, v) \
1.1251 + ((t) = (MUL_UN16 (GET_COMP (y, i), ay, (u)) + \
1.1252 + MUL_UN16 (GET_COMP (x, i), ax, (v))), \
1.1253 + (uint64_t) ((uint16_t) ((t) | \
1.1254 + (0 - ((t) >> G_SHIFT)))) << (i))
1.1255 +
1.1256 +static void
1.1257 +combine_disjoint_general_u (uint64_t * dest,
1.1258 + const uint64_t *src,
1.1259 + const uint64_t *mask,
1.1260 + int width,
1.1261 + uint16_t combine)
1.1262 +{
1.1263 + int i;
1.1264 +
1.1265 + for (i = 0; i < width; ++i)
1.1266 + {
1.1267 + uint64_t s = combine_mask (src, mask, i);
1.1268 + uint64_t d = *(dest + i);
1.1269 + uint64_t m, n, o, p;
1.1270 + uint32_t Fa, Fb, t, u, v;
1.1271 + uint16_t sa = s >> A_SHIFT;
1.1272 + uint16_t da = d >> A_SHIFT;
1.1273 +
1.1274 + switch (combine & COMBINE_A)
1.1275 + {
1.1276 + default:
1.1277 + Fa = 0;
1.1278 + break;
1.1279 +
1.1280 + case COMBINE_A_OUT:
1.1281 + Fa = combine_disjoint_out_part (sa, da);
1.1282 + break;
1.1283 +
1.1284 + case COMBINE_A_IN:
1.1285 + Fa = combine_disjoint_in_part (sa, da);
1.1286 + break;
1.1287 +
1.1288 + case COMBINE_A:
1.1289 + Fa = MASK;
1.1290 + break;
1.1291 + }
1.1292 +
1.1293 + switch (combine & COMBINE_B)
1.1294 + {
1.1295 + default:
1.1296 + Fb = 0;
1.1297 + break;
1.1298 +
1.1299 + case COMBINE_B_OUT:
1.1300 + Fb = combine_disjoint_out_part (da, sa);
1.1301 + break;
1.1302 +
1.1303 + case COMBINE_B_IN:
1.1304 + Fb = combine_disjoint_in_part (da, sa);
1.1305 + break;
1.1306 +
1.1307 + case COMBINE_B:
1.1308 + Fb = MASK;
1.1309 + break;
1.1310 + }
1.1311 + m = GENERIC (s, d, 0, Fa, Fb, t, u, v);
1.1312 + n = GENERIC (s, d, G_SHIFT, Fa, Fb, t, u, v);
1.1313 + o = GENERIC (s, d, R_SHIFT, Fa, Fb, t, u, v);
1.1314 + p = GENERIC (s, d, A_SHIFT, Fa, Fb, t, u, v);
1.1315 + s = m | n | o | p;
1.1316 + *(dest + i) = s;
1.1317 + }
1.1318 +}
1.1319 +
1.1320 +static void
1.1321 +combine_disjoint_over_u (pixman_implementation_t *imp,
1.1322 + pixman_op_t op,
1.1323 + uint64_t * dest,
1.1324 + const uint64_t * src,
1.1325 + const uint64_t * mask,
1.1326 + int width)
1.1327 +{
1.1328 + int i;
1.1329 +
1.1330 + for (i = 0; i < width; ++i)
1.1331 + {
1.1332 + uint64_t s = combine_mask (src, mask, i);
1.1333 + uint32_t a = s >> A_SHIFT;
1.1334 +
1.1335 + if (s != 0x00)
1.1336 + {
1.1337 + uint64_t d = *(dest + i);
1.1338 + a = combine_disjoint_out_part (d >> A_SHIFT, a);
1.1339 + UN16x4_MUL_UN16_ADD_UN16x4 (d, a, s);
1.1340 +
1.1341 + *(dest + i) = d;
1.1342 + }
1.1343 + }
1.1344 +}
1.1345 +
1.1346 +static void
1.1347 +combine_disjoint_in_u (pixman_implementation_t *imp,
1.1348 + pixman_op_t op,
1.1349 + uint64_t * dest,
1.1350 + const uint64_t * src,
1.1351 + const uint64_t * mask,
1.1352 + int width)
1.1353 +{
1.1354 + combine_disjoint_general_u (dest, src, mask, width, COMBINE_A_IN);
1.1355 +}
1.1356 +
1.1357 +static void
1.1358 +combine_disjoint_in_reverse_u (pixman_implementation_t *imp,
1.1359 + pixman_op_t op,
1.1360 + uint64_t * dest,
1.1361 + const uint64_t * src,
1.1362 + const uint64_t * mask,
1.1363 + int width)
1.1364 +{
1.1365 + combine_disjoint_general_u (dest, src, mask, width, COMBINE_B_IN);
1.1366 +}
1.1367 +
1.1368 +static void
1.1369 +combine_disjoint_out_u (pixman_implementation_t *imp,
1.1370 + pixman_op_t op,
1.1371 + uint64_t * dest,
1.1372 + const uint64_t * src,
1.1373 + const uint64_t * mask,
1.1374 + int width)
1.1375 +{
1.1376 + combine_disjoint_general_u (dest, src, mask, width, COMBINE_A_OUT);
1.1377 +}
1.1378 +
1.1379 +static void
1.1380 +combine_disjoint_out_reverse_u (pixman_implementation_t *imp,
1.1381 + pixman_op_t op,
1.1382 + uint64_t * dest,
1.1383 + const uint64_t * src,
1.1384 + const uint64_t * mask,
1.1385 + int width)
1.1386 +{
1.1387 + combine_disjoint_general_u (dest, src, mask, width, COMBINE_B_OUT);
1.1388 +}
1.1389 +
1.1390 +static void
1.1391 +combine_disjoint_atop_u (pixman_implementation_t *imp,
1.1392 + pixman_op_t op,
1.1393 + uint64_t * dest,
1.1394 + const uint64_t * src,
1.1395 + const uint64_t * mask,
1.1396 + int width)
1.1397 +{
1.1398 + combine_disjoint_general_u (dest, src, mask, width, COMBINE_A_ATOP);
1.1399 +}
1.1400 +
1.1401 +static void
1.1402 +combine_disjoint_atop_reverse_u (pixman_implementation_t *imp,
1.1403 + pixman_op_t op,
1.1404 + uint64_t * dest,
1.1405 + const uint64_t * src,
1.1406 + const uint64_t * mask,
1.1407 + int width)
1.1408 +{
1.1409 + combine_disjoint_general_u (dest, src, mask, width, COMBINE_B_ATOP);
1.1410 +}
1.1411 +
1.1412 +static void
1.1413 +combine_disjoint_xor_u (pixman_implementation_t *imp,
1.1414 + pixman_op_t op,
1.1415 + uint64_t * dest,
1.1416 + const uint64_t * src,
1.1417 + const uint64_t * mask,
1.1418 + int width)
1.1419 +{
1.1420 + combine_disjoint_general_u (dest, src, mask, width, COMBINE_XOR);
1.1421 +}
1.1422 +
1.1423 +static void
1.1424 +combine_conjoint_general_u (uint64_t * dest,
1.1425 + const uint64_t *src,
1.1426 + const uint64_t *mask,
1.1427 + int width,
1.1428 + uint16_t combine)
1.1429 +{
1.1430 + int i;
1.1431 +
1.1432 + for (i = 0; i < width; ++i)
1.1433 + {
1.1434 + uint64_t s = combine_mask (src, mask, i);
1.1435 + uint64_t d = *(dest + i);
1.1436 + uint64_t m, n, o, p;
1.1437 + uint32_t Fa, Fb, t, u, v;
1.1438 + uint16_t sa = s >> A_SHIFT;
1.1439 + uint16_t da = d >> A_SHIFT;
1.1440 +
1.1441 + switch (combine & COMBINE_A)
1.1442 + {
1.1443 + default:
1.1444 + Fa = 0;
1.1445 + break;
1.1446 +
1.1447 + case COMBINE_A_OUT:
1.1448 + Fa = combine_conjoint_out_part (sa, da);
1.1449 + break;
1.1450 +
1.1451 + case COMBINE_A_IN:
1.1452 + Fa = combine_conjoint_in_part (sa, da);
1.1453 + break;
1.1454 +
1.1455 + case COMBINE_A:
1.1456 + Fa = MASK;
1.1457 + break;
1.1458 + }
1.1459 +
1.1460 + switch (combine & COMBINE_B)
1.1461 + {
1.1462 + default:
1.1463 + Fb = 0;
1.1464 + break;
1.1465 +
1.1466 + case COMBINE_B_OUT:
1.1467 + Fb = combine_conjoint_out_part (da, sa);
1.1468 + break;
1.1469 +
1.1470 + case COMBINE_B_IN:
1.1471 + Fb = combine_conjoint_in_part (da, sa);
1.1472 + break;
1.1473 +
1.1474 + case COMBINE_B:
1.1475 + Fb = MASK;
1.1476 + break;
1.1477 + }
1.1478 +
1.1479 + m = GENERIC (s, d, 0, Fa, Fb, t, u, v);
1.1480 + n = GENERIC (s, d, G_SHIFT, Fa, Fb, t, u, v);
1.1481 + o = GENERIC (s, d, R_SHIFT, Fa, Fb, t, u, v);
1.1482 + p = GENERIC (s, d, A_SHIFT, Fa, Fb, t, u, v);
1.1483 +
1.1484 + s = m | n | o | p;
1.1485 +
1.1486 + *(dest + i) = s;
1.1487 + }
1.1488 +}
1.1489 +
1.1490 +static void
1.1491 +combine_conjoint_over_u (pixman_implementation_t *imp,
1.1492 + pixman_op_t op,
1.1493 + uint64_t * dest,
1.1494 + const uint64_t * src,
1.1495 + const uint64_t * mask,
1.1496 + int width)
1.1497 +{
1.1498 + combine_conjoint_general_u (dest, src, mask, width, COMBINE_A_OVER);
1.1499 +}
1.1500 +
1.1501 +static void
1.1502 +combine_conjoint_over_reverse_u (pixman_implementation_t *imp,
1.1503 + pixman_op_t op,
1.1504 + uint64_t * dest,
1.1505 + const uint64_t * src,
1.1506 + const uint64_t * mask,
1.1507 + int width)
1.1508 +{
1.1509 + combine_conjoint_general_u (dest, src, mask, width, COMBINE_B_OVER);
1.1510 +}
1.1511 +
1.1512 +static void
1.1513 +combine_conjoint_in_u (pixman_implementation_t *imp,
1.1514 + pixman_op_t op,
1.1515 + uint64_t * dest,
1.1516 + const uint64_t * src,
1.1517 + const uint64_t * mask,
1.1518 + int width)
1.1519 +{
1.1520 + combine_conjoint_general_u (dest, src, mask, width, COMBINE_A_IN);
1.1521 +}
1.1522 +
1.1523 +static void
1.1524 +combine_conjoint_in_reverse_u (pixman_implementation_t *imp,
1.1525 + pixman_op_t op,
1.1526 + uint64_t * dest,
1.1527 + const uint64_t * src,
1.1528 + const uint64_t * mask,
1.1529 + int width)
1.1530 +{
1.1531 + combine_conjoint_general_u (dest, src, mask, width, COMBINE_B_IN);
1.1532 +}
1.1533 +
1.1534 +static void
1.1535 +combine_conjoint_out_u (pixman_implementation_t *imp,
1.1536 + pixman_op_t op,
1.1537 + uint64_t * dest,
1.1538 + const uint64_t * src,
1.1539 + const uint64_t * mask,
1.1540 + int width)
1.1541 +{
1.1542 + combine_conjoint_general_u (dest, src, mask, width, COMBINE_A_OUT);
1.1543 +}
1.1544 +
1.1545 +static void
1.1546 +combine_conjoint_out_reverse_u (pixman_implementation_t *imp,
1.1547 + pixman_op_t op,
1.1548 + uint64_t * dest,
1.1549 + const uint64_t * src,
1.1550 + const uint64_t * mask,
1.1551 + int width)
1.1552 +{
1.1553 + combine_conjoint_general_u (dest, src, mask, width, COMBINE_B_OUT);
1.1554 +}
1.1555 +
1.1556 +static void
1.1557 +combine_conjoint_atop_u (pixman_implementation_t *imp,
1.1558 + pixman_op_t op,
1.1559 + uint64_t * dest,
1.1560 + const uint64_t * src,
1.1561 + const uint64_t * mask,
1.1562 + int width)
1.1563 +{
1.1564 + combine_conjoint_general_u (dest, src, mask, width, COMBINE_A_ATOP);
1.1565 +}
1.1566 +
1.1567 +static void
1.1568 +combine_conjoint_atop_reverse_u (pixman_implementation_t *imp,
1.1569 + pixman_op_t op,
1.1570 + uint64_t * dest,
1.1571 + const uint64_t * src,
1.1572 + const uint64_t * mask,
1.1573 + int width)
1.1574 +{
1.1575 + combine_conjoint_general_u (dest, src, mask, width, COMBINE_B_ATOP);
1.1576 +}
1.1577 +
1.1578 +static void
1.1579 +combine_conjoint_xor_u (pixman_implementation_t *imp,
1.1580 + pixman_op_t op,
1.1581 + uint64_t * dest,
1.1582 + const uint64_t * src,
1.1583 + const uint64_t * mask,
1.1584 + int width)
1.1585 +{
1.1586 + combine_conjoint_general_u (dest, src, mask, width, COMBINE_XOR);
1.1587 +}
1.1588 +
1.1589 +/************************************************************************/
1.1590 +/*********************** Per Channel functions **************************/
1.1591 +/************************************************************************/
1.1592 +
1.1593 +static void
1.1594 +combine_clear_ca (pixman_implementation_t *imp,
1.1595 + pixman_op_t op,
1.1596 + uint64_t * dest,
1.1597 + const uint64_t * src,
1.1598 + const uint64_t * mask,
1.1599 + int width)
1.1600 +{
1.1601 + memset (dest, 0, width * sizeof(uint64_t));
1.1602 +}
1.1603 +
1.1604 +static void
1.1605 +combine_src_ca (pixman_implementation_t *imp,
1.1606 + pixman_op_t op,
1.1607 + uint64_t * dest,
1.1608 + const uint64_t * src,
1.1609 + const uint64_t * mask,
1.1610 + int width)
1.1611 +{
1.1612 + int i;
1.1613 +
1.1614 + for (i = 0; i < width; ++i)
1.1615 + {
1.1616 + uint64_t s = *(src + i);
1.1617 + uint64_t m = *(mask + i);
1.1618 +
1.1619 + combine_mask_value_ca (&s, &m);
1.1620 +
1.1621 + *(dest + i) = s;
1.1622 + }
1.1623 +}
1.1624 +
1.1625 +static void
1.1626 +combine_over_ca (pixman_implementation_t *imp,
1.1627 + pixman_op_t op,
1.1628 + uint64_t * dest,
1.1629 + const uint64_t * src,
1.1630 + const uint64_t * mask,
1.1631 + int width)
1.1632 +{
1.1633 + int i;
1.1634 +
1.1635 + for (i = 0; i < width; ++i)
1.1636 + {
1.1637 + uint64_t s = *(src + i);
1.1638 + uint64_t m = *(mask + i);
1.1639 + uint64_t a;
1.1640 +
1.1641 + combine_mask_ca (&s, &m);
1.1642 +
1.1643 + a = ~m;
1.1644 + if (a)
1.1645 + {
1.1646 + uint64_t d = *(dest + i);
1.1647 + UN16x4_MUL_UN16x4_ADD_UN16x4 (d, a, s);
1.1648 + s = d;
1.1649 + }
1.1650 +
1.1651 + *(dest + i) = s;
1.1652 + }
1.1653 +}
1.1654 +
1.1655 +static void
1.1656 +combine_over_reverse_ca (pixman_implementation_t *imp,
1.1657 + pixman_op_t op,
1.1658 + uint64_t * dest,
1.1659 + const uint64_t * src,
1.1660 + const uint64_t * mask,
1.1661 + int width)
1.1662 +{
1.1663 + int i;
1.1664 +
1.1665 + for (i = 0; i < width; ++i)
1.1666 + {
1.1667 + uint64_t d = *(dest + i);
1.1668 + uint64_t a = ~d >> A_SHIFT;
1.1669 +
1.1670 + if (a)
1.1671 + {
1.1672 + uint64_t s = *(src + i);
1.1673 + uint64_t m = *(mask + i);
1.1674 +
1.1675 + UN16x4_MUL_UN16x4 (s, m);
1.1676 + UN16x4_MUL_UN16_ADD_UN16x4 (s, a, d);
1.1677 +
1.1678 + *(dest + i) = s;
1.1679 + }
1.1680 + }
1.1681 +}
1.1682 +
1.1683 +static void
1.1684 +combine_in_ca (pixman_implementation_t *imp,
1.1685 + pixman_op_t op,
1.1686 + uint64_t * dest,
1.1687 + const uint64_t * src,
1.1688 + const uint64_t * mask,
1.1689 + int width)
1.1690 +{
1.1691 + int i;
1.1692 +
1.1693 + for (i = 0; i < width; ++i)
1.1694 + {
1.1695 + uint64_t d = *(dest + i);
1.1696 + uint32_t a = d >> A_SHIFT;
1.1697 + uint64_t s = 0;
1.1698 +
1.1699 + if (a)
1.1700 + {
1.1701 + uint64_t m = *(mask + i);
1.1702 +
1.1703 + s = *(src + i);
1.1704 + combine_mask_value_ca (&s, &m);
1.1705 +
1.1706 + if (a != MASK)
1.1707 + UN16x4_MUL_UN16 (s, a);
1.1708 + }
1.1709 +
1.1710 + *(dest + i) = s;
1.1711 + }
1.1712 +}
1.1713 +
1.1714 +static void
1.1715 +combine_in_reverse_ca (pixman_implementation_t *imp,
1.1716 + pixman_op_t op,
1.1717 + uint64_t * dest,
1.1718 + const uint64_t * src,
1.1719 + const uint64_t * mask,
1.1720 + int width)
1.1721 +{
1.1722 + int i;
1.1723 +
1.1724 + for (i = 0; i < width; ++i)
1.1725 + {
1.1726 + uint64_t s = *(src + i);
1.1727 + uint64_t m = *(mask + i);
1.1728 + uint64_t a;
1.1729 +
1.1730 + combine_mask_alpha_ca (&s, &m);
1.1731 +
1.1732 + a = m;
1.1733 + if (a != ~0)
1.1734 + {
1.1735 + uint64_t d = 0;
1.1736 +
1.1737 + if (a)
1.1738 + {
1.1739 + d = *(dest + i);
1.1740 + UN16x4_MUL_UN16x4 (d, a);
1.1741 + }
1.1742 +
1.1743 + *(dest + i) = d;
1.1744 + }
1.1745 + }
1.1746 +}
1.1747 +
1.1748 +static void
1.1749 +combine_out_ca (pixman_implementation_t *imp,
1.1750 + pixman_op_t op,
1.1751 + uint64_t * dest,
1.1752 + const uint64_t * src,
1.1753 + const uint64_t * mask,
1.1754 + int width)
1.1755 +{
1.1756 + int i;
1.1757 +
1.1758 + for (i = 0; i < width; ++i)
1.1759 + {
1.1760 + uint64_t d = *(dest + i);
1.1761 + uint32_t a = ~d >> A_SHIFT;
1.1762 + uint64_t s = 0;
1.1763 +
1.1764 + if (a)
1.1765 + {
1.1766 + uint64_t m = *(mask + i);
1.1767 +
1.1768 + s = *(src + i);
1.1769 + combine_mask_value_ca (&s, &m);
1.1770 +
1.1771 + if (a != MASK)
1.1772 + UN16x4_MUL_UN16 (s, a);
1.1773 + }
1.1774 +
1.1775 + *(dest + i) = s;
1.1776 + }
1.1777 +}
1.1778 +
1.1779 +static void
1.1780 +combine_out_reverse_ca (pixman_implementation_t *imp,
1.1781 + pixman_op_t op,
1.1782 + uint64_t * dest,
1.1783 + const uint64_t * src,
1.1784 + const uint64_t * mask,
1.1785 + int width)
1.1786 +{
1.1787 + int i;
1.1788 +
1.1789 + for (i = 0; i < width; ++i)
1.1790 + {
1.1791 + uint64_t s = *(src + i);
1.1792 + uint64_t m = *(mask + i);
1.1793 + uint64_t a;
1.1794 +
1.1795 + combine_mask_alpha_ca (&s, &m);
1.1796 +
1.1797 + a = ~m;
1.1798 + if (a != ~0)
1.1799 + {
1.1800 + uint64_t d = 0;
1.1801 +
1.1802 + if (a)
1.1803 + {
1.1804 + d = *(dest + i);
1.1805 + UN16x4_MUL_UN16x4 (d, a);
1.1806 + }
1.1807 +
1.1808 + *(dest + i) = d;
1.1809 + }
1.1810 + }
1.1811 +}
1.1812 +
1.1813 +static void
1.1814 +combine_atop_ca (pixman_implementation_t *imp,
1.1815 + pixman_op_t op,
1.1816 + uint64_t * dest,
1.1817 + const uint64_t * src,
1.1818 + const uint64_t * mask,
1.1819 + int width)
1.1820 +{
1.1821 + int i;
1.1822 +
1.1823 + for (i = 0; i < width; ++i)
1.1824 + {
1.1825 + uint64_t d = *(dest + i);
1.1826 + uint64_t s = *(src + i);
1.1827 + uint64_t m = *(mask + i);
1.1828 + uint64_t ad;
1.1829 + uint32_t as = d >> A_SHIFT;
1.1830 +
1.1831 + combine_mask_ca (&s, &m);
1.1832 +
1.1833 + ad = ~m;
1.1834 +
1.1835 + UN16x4_MUL_UN16x4_ADD_UN16x4_MUL_UN16 (d, ad, s, as);
1.1836 +
1.1837 + *(dest + i) = d;
1.1838 + }
1.1839 +}
1.1840 +
1.1841 +static void
1.1842 +combine_atop_reverse_ca (pixman_implementation_t *imp,
1.1843 + pixman_op_t op,
1.1844 + uint64_t * dest,
1.1845 + const uint64_t * src,
1.1846 + const uint64_t * mask,
1.1847 + int width)
1.1848 +{
1.1849 + int i;
1.1850 +
1.1851 + for (i = 0; i < width; ++i)
1.1852 + {
1.1853 + uint64_t d = *(dest + i);
1.1854 + uint64_t s = *(src + i);
1.1855 + uint64_t m = *(mask + i);
1.1856 + uint64_t ad;
1.1857 + uint32_t as = ~d >> A_SHIFT;
1.1858 +
1.1859 + combine_mask_ca (&s, &m);
1.1860 +
1.1861 + ad = m;
1.1862 +
1.1863 + UN16x4_MUL_UN16x4_ADD_UN16x4_MUL_UN16 (d, ad, s, as);
1.1864 +
1.1865 + *(dest + i) = d;
1.1866 + }
1.1867 +}
1.1868 +
1.1869 +static void
1.1870 +combine_xor_ca (pixman_implementation_t *imp,
1.1871 + pixman_op_t op,
1.1872 + uint64_t * dest,
1.1873 + const uint64_t * src,
1.1874 + const uint64_t * mask,
1.1875 + int width)
1.1876 +{
1.1877 + int i;
1.1878 +
1.1879 + for (i = 0; i < width; ++i)
1.1880 + {
1.1881 + uint64_t d = *(dest + i);
1.1882 + uint64_t s = *(src + i);
1.1883 + uint64_t m = *(mask + i);
1.1884 + uint64_t ad;
1.1885 + uint32_t as = ~d >> A_SHIFT;
1.1886 +
1.1887 + combine_mask_ca (&s, &m);
1.1888 +
1.1889 + ad = ~m;
1.1890 +
1.1891 + UN16x4_MUL_UN16x4_ADD_UN16x4_MUL_UN16 (d, ad, s, as);
1.1892 +
1.1893 + *(dest + i) = d;
1.1894 + }
1.1895 +}
1.1896 +
1.1897 +static void
1.1898 +combine_add_ca (pixman_implementation_t *imp,
1.1899 + pixman_op_t op,
1.1900 + uint64_t * dest,
1.1901 + const uint64_t * src,
1.1902 + const uint64_t * mask,
1.1903 + int width)
1.1904 +{
1.1905 + int i;
1.1906 +
1.1907 + for (i = 0; i < width; ++i)
1.1908 + {
1.1909 + uint64_t s = *(src + i);
1.1910 + uint64_t m = *(mask + i);
1.1911 + uint64_t d = *(dest + i);
1.1912 +
1.1913 + combine_mask_value_ca (&s, &m);
1.1914 +
1.1915 + UN16x4_ADD_UN16x4 (d, s);
1.1916 +
1.1917 + *(dest + i) = d;
1.1918 + }
1.1919 +}
1.1920 +
1.1921 +static void
1.1922 +combine_saturate_ca (pixman_implementation_t *imp,
1.1923 + pixman_op_t op,
1.1924 + uint64_t * dest,
1.1925 + const uint64_t * src,
1.1926 + const uint64_t * mask,
1.1927 + int width)
1.1928 +{
1.1929 + int i;
1.1930 +
1.1931 + for (i = 0; i < width; ++i)
1.1932 + {
1.1933 + uint64_t s, d;
1.1934 + uint32_t sa, sr, sg, sb, da;
1.1935 + uint32_t t, u, v;
1.1936 + uint64_t m, n, o, p;
1.1937 +
1.1938 + d = *(dest + i);
1.1939 + s = *(src + i);
1.1940 + m = *(mask + i);
1.1941 +
1.1942 + combine_mask_ca (&s, &m);
1.1943 +
1.1944 + sa = (m >> A_SHIFT);
1.1945 + sr = (m >> R_SHIFT) & MASK;
1.1946 + sg = (m >> G_SHIFT) & MASK;
1.1947 + sb = m & MASK;
1.1948 + da = ~d >> A_SHIFT;
1.1949 +
1.1950 + if (sb <= da)
1.1951 + m = ADD (s, d, 0, t);
1.1952 + else
1.1953 + m = GENERIC (s, d, 0, (da << G_SHIFT) / sb, MASK, t, u, v);
1.1954 +
1.1955 + if (sg <= da)
1.1956 + n = ADD (s, d, G_SHIFT, t);
1.1957 + else
1.1958 + n = GENERIC (s, d, G_SHIFT, (da << G_SHIFT) / sg, MASK, t, u, v);
1.1959 +
1.1960 + if (sr <= da)
1.1961 + o = ADD (s, d, R_SHIFT, t);
1.1962 + else
1.1963 + o = GENERIC (s, d, R_SHIFT, (da << G_SHIFT) / sr, MASK, t, u, v);
1.1964 +
1.1965 + if (sa <= da)
1.1966 + p = ADD (s, d, A_SHIFT, t);
1.1967 + else
1.1968 + p = GENERIC (s, d, A_SHIFT, (da << G_SHIFT) / sa, MASK, t, u, v);
1.1969 +
1.1970 + *(dest + i) = m | n | o | p;
1.1971 + }
1.1972 +}
1.1973 +
1.1974 +static void
1.1975 +combine_disjoint_general_ca (uint64_t * dest,
1.1976 + const uint64_t *src,
1.1977 + const uint64_t *mask,
1.1978 + int width,
1.1979 + uint16_t combine)
1.1980 +{
1.1981 + int i;
1.1982 +
1.1983 + for (i = 0; i < width; ++i)
1.1984 + {
1.1985 + uint64_t s, d;
1.1986 + uint64_t m, n, o, p;
1.1987 + uint64_t Fa, Fb;
1.1988 + uint32_t t, u, v;
1.1989 + uint64_t sa;
1.1990 + uint16_t da;
1.1991 +
1.1992 + s = *(src + i);
1.1993 + m = *(mask + i);
1.1994 + d = *(dest + i);
1.1995 + da = d >> A_SHIFT;
1.1996 +
1.1997 + combine_mask_ca (&s, &m);
1.1998 +
1.1999 + sa = m;
1.2000 +
1.2001 + switch (combine & COMBINE_A)
1.2002 + {
1.2003 + default:
1.2004 + Fa = 0;
1.2005 + break;
1.2006 +
1.2007 + case COMBINE_A_OUT:
1.2008 + m = (uint64_t)combine_disjoint_out_part ((uint16_t) (sa >> 0), da);
1.2009 + n = (uint64_t)combine_disjoint_out_part ((uint16_t) (sa >> G_SHIFT), da) << G_SHIFT;
1.2010 + o = (uint64_t)combine_disjoint_out_part ((uint16_t) (sa >> R_SHIFT), da) << R_SHIFT;
1.2011 + p = (uint64_t)combine_disjoint_out_part ((uint16_t) (sa >> A_SHIFT), da) << A_SHIFT;
1.2012 + Fa = m | n | o | p;
1.2013 + break;
1.2014 +
1.2015 + case COMBINE_A_IN:
1.2016 + m = (uint64_t)combine_disjoint_in_part ((uint16_t) (sa >> 0), da);
1.2017 + n = (uint64_t)combine_disjoint_in_part ((uint16_t) (sa >> G_SHIFT), da) << G_SHIFT;
1.2018 + o = (uint64_t)combine_disjoint_in_part ((uint16_t) (sa >> R_SHIFT), da) << R_SHIFT;
1.2019 + p = (uint64_t)combine_disjoint_in_part ((uint16_t) (sa >> A_SHIFT), da) << A_SHIFT;
1.2020 + Fa = m | n | o | p;
1.2021 + break;
1.2022 +
1.2023 + case COMBINE_A:
1.2024 + Fa = ~0;
1.2025 + break;
1.2026 + }
1.2027 +
1.2028 + switch (combine & COMBINE_B)
1.2029 + {
1.2030 + default:
1.2031 + Fb = 0;
1.2032 + break;
1.2033 +
1.2034 + case COMBINE_B_OUT:
1.2035 + m = (uint64_t)combine_disjoint_out_part (da, (uint16_t) (sa >> 0));
1.2036 + n = (uint64_t)combine_disjoint_out_part (da, (uint16_t) (sa >> G_SHIFT)) << G_SHIFT;
1.2037 + o = (uint64_t)combine_disjoint_out_part (da, (uint16_t) (sa >> R_SHIFT)) << R_SHIFT;
1.2038 + p = (uint64_t)combine_disjoint_out_part (da, (uint16_t) (sa >> A_SHIFT)) << A_SHIFT;
1.2039 + Fb = m | n | o | p;
1.2040 + break;
1.2041 +
1.2042 + case COMBINE_B_IN:
1.2043 + m = (uint64_t)combine_disjoint_in_part (da, (uint16_t) (sa >> 0));
1.2044 + n = (uint64_t)combine_disjoint_in_part (da, (uint16_t) (sa >> G_SHIFT)) << G_SHIFT;
1.2045 + o = (uint64_t)combine_disjoint_in_part (da, (uint16_t) (sa >> R_SHIFT)) << R_SHIFT;
1.2046 + p = (uint64_t)combine_disjoint_in_part (da, (uint16_t) (sa >> A_SHIFT)) << A_SHIFT;
1.2047 + Fb = m | n | o | p;
1.2048 + break;
1.2049 +
1.2050 + case COMBINE_B:
1.2051 + Fb = ~0;
1.2052 + break;
1.2053 + }
1.2054 + m = GENERIC (s, d, 0, GET_COMP (Fa, 0), GET_COMP (Fb, 0), t, u, v);
1.2055 + n = GENERIC (s, d, G_SHIFT, GET_COMP (Fa, G_SHIFT), GET_COMP (Fb, G_SHIFT), t, u, v);
1.2056 + o = GENERIC (s, d, R_SHIFT, GET_COMP (Fa, R_SHIFT), GET_COMP (Fb, R_SHIFT), t, u, v);
1.2057 + p = GENERIC (s, d, A_SHIFT, GET_COMP (Fa, A_SHIFT), GET_COMP (Fb, A_SHIFT), t, u, v);
1.2058 +
1.2059 + s = m | n | o | p;
1.2060 +
1.2061 + *(dest + i) = s;
1.2062 + }
1.2063 +}
1.2064 +
1.2065 +static void
1.2066 +combine_disjoint_over_ca (pixman_implementation_t *imp,
1.2067 + pixman_op_t op,
1.2068 + uint64_t * dest,
1.2069 + const uint64_t * src,
1.2070 + const uint64_t * mask,
1.2071 + int width)
1.2072 +{
1.2073 + combine_disjoint_general_ca (dest, src, mask, width, COMBINE_A_OVER);
1.2074 +}
1.2075 +
1.2076 +static void
1.2077 +combine_disjoint_in_ca (pixman_implementation_t *imp,
1.2078 + pixman_op_t op,
1.2079 + uint64_t * dest,
1.2080 + const uint64_t * src,
1.2081 + const uint64_t * mask,
1.2082 + int width)
1.2083 +{
1.2084 + combine_disjoint_general_ca (dest, src, mask, width, COMBINE_A_IN);
1.2085 +}
1.2086 +
1.2087 +static void
1.2088 +combine_disjoint_in_reverse_ca (pixman_implementation_t *imp,
1.2089 + pixman_op_t op,
1.2090 + uint64_t * dest,
1.2091 + const uint64_t * src,
1.2092 + const uint64_t * mask,
1.2093 + int width)
1.2094 +{
1.2095 + combine_disjoint_general_ca (dest, src, mask, width, COMBINE_B_IN);
1.2096 +}
1.2097 +
1.2098 +static void
1.2099 +combine_disjoint_out_ca (pixman_implementation_t *imp,
1.2100 + pixman_op_t op,
1.2101 + uint64_t * dest,
1.2102 + const uint64_t * src,
1.2103 + const uint64_t * mask,
1.2104 + int width)
1.2105 +{
1.2106 + combine_disjoint_general_ca (dest, src, mask, width, COMBINE_A_OUT);
1.2107 +}
1.2108 +
1.2109 +static void
1.2110 +combine_disjoint_out_reverse_ca (pixman_implementation_t *imp,
1.2111 + pixman_op_t op,
1.2112 + uint64_t * dest,
1.2113 + const uint64_t * src,
1.2114 + const uint64_t * mask,
1.2115 + int width)
1.2116 +{
1.2117 + combine_disjoint_general_ca (dest, src, mask, width, COMBINE_B_OUT);
1.2118 +}
1.2119 +
1.2120 +static void
1.2121 +combine_disjoint_atop_ca (pixman_implementation_t *imp,
1.2122 + pixman_op_t op,
1.2123 + uint64_t * dest,
1.2124 + const uint64_t * src,
1.2125 + const uint64_t * mask,
1.2126 + int width)
1.2127 +{
1.2128 + combine_disjoint_general_ca (dest, src, mask, width, COMBINE_A_ATOP);
1.2129 +}
1.2130 +
1.2131 +static void
1.2132 +combine_disjoint_atop_reverse_ca (pixman_implementation_t *imp,
1.2133 + pixman_op_t op,
1.2134 + uint64_t * dest,
1.2135 + const uint64_t * src,
1.2136 + const uint64_t * mask,
1.2137 + int width)
1.2138 +{
1.2139 + combine_disjoint_general_ca (dest, src, mask, width, COMBINE_B_ATOP);
1.2140 +}
1.2141 +
1.2142 +static void
1.2143 +combine_disjoint_xor_ca (pixman_implementation_t *imp,
1.2144 + pixman_op_t op,
1.2145 + uint64_t * dest,
1.2146 + const uint64_t * src,
1.2147 + const uint64_t * mask,
1.2148 + int width)
1.2149 +{
1.2150 + combine_disjoint_general_ca (dest, src, mask, width, COMBINE_XOR);
1.2151 +}
1.2152 +
1.2153 +static void
1.2154 +combine_conjoint_general_ca (uint64_t * dest,
1.2155 + const uint64_t *src,
1.2156 + const uint64_t *mask,
1.2157 + int width,
1.2158 + uint16_t combine)
1.2159 +{
1.2160 + int i;
1.2161 +
1.2162 + for (i = 0; i < width; ++i)
1.2163 + {
1.2164 + uint64_t s, d;
1.2165 + uint64_t m, n, o, p;
1.2166 + uint64_t Fa, Fb;
1.2167 + uint32_t t, u, v;
1.2168 + uint64_t sa;
1.2169 + uint16_t da;
1.2170 +
1.2171 + s = *(src + i);
1.2172 + m = *(mask + i);
1.2173 + d = *(dest + i);
1.2174 + da = d >> A_SHIFT;
1.2175 +
1.2176 + combine_mask_ca (&s, &m);
1.2177 +
1.2178 + sa = m;
1.2179 +
1.2180 + switch (combine & COMBINE_A)
1.2181 + {
1.2182 + default:
1.2183 + Fa = 0;
1.2184 + break;
1.2185 +
1.2186 + case COMBINE_A_OUT:
1.2187 + m = (uint64_t)combine_conjoint_out_part ((uint16_t) (sa >> 0), da);
1.2188 + n = (uint64_t)combine_conjoint_out_part ((uint16_t) (sa >> G_SHIFT), da) << G_SHIFT;
1.2189 + o = (uint64_t)combine_conjoint_out_part ((uint16_t) (sa >> R_SHIFT), da) << R_SHIFT;
1.2190 + p = (uint64_t)combine_conjoint_out_part ((uint16_t) (sa >> A_SHIFT), da) << A_SHIFT;
1.2191 + Fa = m | n | o | p;
1.2192 + break;
1.2193 +
1.2194 + case COMBINE_A_IN:
1.2195 + m = (uint64_t)combine_conjoint_in_part ((uint16_t) (sa >> 0), da);
1.2196 + n = (uint64_t)combine_conjoint_in_part ((uint16_t) (sa >> G_SHIFT), da) << G_SHIFT;
1.2197 + o = (uint64_t)combine_conjoint_in_part ((uint16_t) (sa >> R_SHIFT), da) << R_SHIFT;
1.2198 + p = (uint64_t)combine_conjoint_in_part ((uint16_t) (sa >> A_SHIFT), da) << A_SHIFT;
1.2199 + Fa = m | n | o | p;
1.2200 + break;
1.2201 +
1.2202 + case COMBINE_A:
1.2203 + Fa = ~0;
1.2204 + break;
1.2205 + }
1.2206 +
1.2207 + switch (combine & COMBINE_B)
1.2208 + {
1.2209 + default:
1.2210 + Fb = 0;
1.2211 + break;
1.2212 +
1.2213 + case COMBINE_B_OUT:
1.2214 + m = (uint64_t)combine_conjoint_out_part (da, (uint16_t) (sa >> 0));
1.2215 + n = (uint64_t)combine_conjoint_out_part (da, (uint16_t) (sa >> G_SHIFT)) << G_SHIFT;
1.2216 + o = (uint64_t)combine_conjoint_out_part (da, (uint16_t) (sa >> R_SHIFT)) << R_SHIFT;
1.2217 + p = (uint64_t)combine_conjoint_out_part (da, (uint16_t) (sa >> A_SHIFT)) << A_SHIFT;
1.2218 + Fb = m | n | o | p;
1.2219 + break;
1.2220 +
1.2221 + case COMBINE_B_IN:
1.2222 + m = (uint64_t)combine_conjoint_in_part (da, (uint16_t) (sa >> 0));
1.2223 + n = (uint64_t)combine_conjoint_in_part (da, (uint16_t) (sa >> G_SHIFT)) << G_SHIFT;
1.2224 + o = (uint64_t)combine_conjoint_in_part (da, (uint16_t) (sa >> R_SHIFT)) << R_SHIFT;
1.2225 + p = (uint64_t)combine_conjoint_in_part (da, (uint16_t) (sa >> A_SHIFT)) << A_SHIFT;
1.2226 + Fb = m | n | o | p;
1.2227 + break;
1.2228 +
1.2229 + case COMBINE_B:
1.2230 + Fb = ~0;
1.2231 + break;
1.2232 + }
1.2233 + m = GENERIC (s, d, 0, GET_COMP (Fa, 0), GET_COMP (Fb, 0), t, u, v);
1.2234 + n = GENERIC (s, d, G_SHIFT, GET_COMP (Fa, G_SHIFT), GET_COMP (Fb, G_SHIFT), t, u, v);
1.2235 + o = GENERIC (s, d, R_SHIFT, GET_COMP (Fa, R_SHIFT), GET_COMP (Fb, R_SHIFT), t, u, v);
1.2236 + p = GENERIC (s, d, A_SHIFT, GET_COMP (Fa, A_SHIFT), GET_COMP (Fb, A_SHIFT), t, u, v);
1.2237 +
1.2238 + s = m | n | o | p;
1.2239 +
1.2240 + *(dest + i) = s;
1.2241 + }
1.2242 +}
1.2243 +
1.2244 +static void
1.2245 +combine_conjoint_over_ca (pixman_implementation_t *imp,
1.2246 + pixman_op_t op,
1.2247 + uint64_t * dest,
1.2248 + const uint64_t * src,
1.2249 + const uint64_t * mask,
1.2250 + int width)
1.2251 +{
1.2252 + combine_conjoint_general_ca (dest, src, mask, width, COMBINE_A_OVER);
1.2253 +}
1.2254 +
1.2255 +static void
1.2256 +combine_conjoint_over_reverse_ca (pixman_implementation_t *imp,
1.2257 + pixman_op_t op,
1.2258 + uint64_t * dest,
1.2259 + const uint64_t * src,
1.2260 + const uint64_t * mask,
1.2261 + int width)
1.2262 +{
1.2263 + combine_conjoint_general_ca (dest, src, mask, width, COMBINE_B_OVER);
1.2264 +}
1.2265 +
1.2266 +static void
1.2267 +combine_conjoint_in_ca (pixman_implementation_t *imp,
1.2268 + pixman_op_t op,
1.2269 + uint64_t * dest,
1.2270 + const uint64_t * src,
1.2271 + const uint64_t * mask,
1.2272 + int width)
1.2273 +{
1.2274 + combine_conjoint_general_ca (dest, src, mask, width, COMBINE_A_IN);
1.2275 +}
1.2276 +
1.2277 +static void
1.2278 +combine_conjoint_in_reverse_ca (pixman_implementation_t *imp,
1.2279 + pixman_op_t op,
1.2280 + uint64_t * dest,
1.2281 + const uint64_t * src,
1.2282 + const uint64_t * mask,
1.2283 + int width)
1.2284 +{
1.2285 + combine_conjoint_general_ca (dest, src, mask, width, COMBINE_B_IN);
1.2286 +}
1.2287 +
1.2288 +static void
1.2289 +combine_conjoint_out_ca (pixman_implementation_t *imp,
1.2290 + pixman_op_t op,
1.2291 + uint64_t * dest,
1.2292 + const uint64_t * src,
1.2293 + const uint64_t * mask,
1.2294 + int width)
1.2295 +{
1.2296 + combine_conjoint_general_ca (dest, src, mask, width, COMBINE_A_OUT);
1.2297 +}
1.2298 +
1.2299 +static void
1.2300 +combine_conjoint_out_reverse_ca (pixman_implementation_t *imp,
1.2301 + pixman_op_t op,
1.2302 + uint64_t * dest,
1.2303 + const uint64_t * src,
1.2304 + const uint64_t * mask,
1.2305 + int width)
1.2306 +{
1.2307 + combine_conjoint_general_ca (dest, src, mask, width, COMBINE_B_OUT);
1.2308 +}
1.2309 +
1.2310 +static void
1.2311 +combine_conjoint_atop_ca (pixman_implementation_t *imp,
1.2312 + pixman_op_t op,
1.2313 + uint64_t * dest,
1.2314 + const uint64_t * src,
1.2315 + const uint64_t * mask,
1.2316 + int width)
1.2317 +{
1.2318 + combine_conjoint_general_ca (dest, src, mask, width, COMBINE_A_ATOP);
1.2319 +}
1.2320 +
1.2321 +static void
1.2322 +combine_conjoint_atop_reverse_ca (pixman_implementation_t *imp,
1.2323 + pixman_op_t op,
1.2324 + uint64_t * dest,
1.2325 + const uint64_t * src,
1.2326 + const uint64_t * mask,
1.2327 + int width)
1.2328 +{
1.2329 + combine_conjoint_general_ca (dest, src, mask, width, COMBINE_B_ATOP);
1.2330 +}
1.2331 +
1.2332 +static void
1.2333 +combine_conjoint_xor_ca (pixman_implementation_t *imp,
1.2334 + pixman_op_t op,
1.2335 + uint64_t * dest,
1.2336 + const uint64_t * src,
1.2337 + const uint64_t * mask,
1.2338 + int width)
1.2339 +{
1.2340 + combine_conjoint_general_ca (dest, src, mask, width, COMBINE_XOR);
1.2341 +}
1.2342 +
1.2343 +void
1.2344 +_pixman_setup_combiner_functions_64 (pixman_implementation_t *imp)
1.2345 +{
1.2346 + /* Unified alpha */
1.2347 + imp->combine_64[PIXMAN_OP_CLEAR] = combine_clear;
1.2348 + imp->combine_64[PIXMAN_OP_SRC] = combine_src_u;
1.2349 + imp->combine_64[PIXMAN_OP_DST] = combine_dst;
1.2350 + imp->combine_64[PIXMAN_OP_OVER] = combine_over_u;
1.2351 + imp->combine_64[PIXMAN_OP_OVER_REVERSE] = combine_over_reverse_u;
1.2352 + imp->combine_64[PIXMAN_OP_IN] = combine_in_u;
1.2353 + imp->combine_64[PIXMAN_OP_IN_REVERSE] = combine_in_reverse_u;
1.2354 + imp->combine_64[PIXMAN_OP_OUT] = combine_out_u;
1.2355 + imp->combine_64[PIXMAN_OP_OUT_REVERSE] = combine_out_reverse_u;
1.2356 + imp->combine_64[PIXMAN_OP_ATOP] = combine_atop_u;
1.2357 + imp->combine_64[PIXMAN_OP_ATOP_REVERSE] = combine_atop_reverse_u;
1.2358 + imp->combine_64[PIXMAN_OP_XOR] = combine_xor_u;
1.2359 + imp->combine_64[PIXMAN_OP_ADD] = combine_add_u;
1.2360 + imp->combine_64[PIXMAN_OP_SATURATE] = combine_saturate_u;
1.2361 +
1.2362 + /* Disjoint, unified */
1.2363 + imp->combine_64[PIXMAN_OP_DISJOINT_CLEAR] = combine_clear;
1.2364 + imp->combine_64[PIXMAN_OP_DISJOINT_SRC] = combine_src_u;
1.2365 + imp->combine_64[PIXMAN_OP_DISJOINT_DST] = combine_dst;
1.2366 + imp->combine_64[PIXMAN_OP_DISJOINT_OVER] = combine_disjoint_over_u;
1.2367 + imp->combine_64[PIXMAN_OP_DISJOINT_OVER_REVERSE] = combine_saturate_u;
1.2368 + imp->combine_64[PIXMAN_OP_DISJOINT_IN] = combine_disjoint_in_u;
1.2369 + imp->combine_64[PIXMAN_OP_DISJOINT_IN_REVERSE] = combine_disjoint_in_reverse_u;
1.2370 + imp->combine_64[PIXMAN_OP_DISJOINT_OUT] = combine_disjoint_out_u;
1.2371 + imp->combine_64[PIXMAN_OP_DISJOINT_OUT_REVERSE] = combine_disjoint_out_reverse_u;
1.2372 + imp->combine_64[PIXMAN_OP_DISJOINT_ATOP] = combine_disjoint_atop_u;
1.2373 + imp->combine_64[PIXMAN_OP_DISJOINT_ATOP_REVERSE] = combine_disjoint_atop_reverse_u;
1.2374 + imp->combine_64[PIXMAN_OP_DISJOINT_XOR] = combine_disjoint_xor_u;
1.2375 +
1.2376 + /* Conjoint, unified */
1.2377 + imp->combine_64[PIXMAN_OP_CONJOINT_CLEAR] = combine_clear;
1.2378 + imp->combine_64[PIXMAN_OP_CONJOINT_SRC] = combine_src_u;
1.2379 + imp->combine_64[PIXMAN_OP_CONJOINT_DST] = combine_dst;
1.2380 + imp->combine_64[PIXMAN_OP_CONJOINT_OVER] = combine_conjoint_over_u;
1.2381 + imp->combine_64[PIXMAN_OP_CONJOINT_OVER_REVERSE] = combine_conjoint_over_reverse_u;
1.2382 + imp->combine_64[PIXMAN_OP_CONJOINT_IN] = combine_conjoint_in_u;
1.2383 + imp->combine_64[PIXMAN_OP_CONJOINT_IN_REVERSE] = combine_conjoint_in_reverse_u;
1.2384 + imp->combine_64[PIXMAN_OP_CONJOINT_OUT] = combine_conjoint_out_u;
1.2385 + imp->combine_64[PIXMAN_OP_CONJOINT_OUT_REVERSE] = combine_conjoint_out_reverse_u;
1.2386 + imp->combine_64[PIXMAN_OP_CONJOINT_ATOP] = combine_conjoint_atop_u;
1.2387 + imp->combine_64[PIXMAN_OP_CONJOINT_ATOP_REVERSE] = combine_conjoint_atop_reverse_u;
1.2388 + imp->combine_64[PIXMAN_OP_CONJOINT_XOR] = combine_conjoint_xor_u;
1.2389 +
1.2390 + imp->combine_64[PIXMAN_OP_MULTIPLY] = combine_multiply_u;
1.2391 + imp->combine_64[PIXMAN_OP_SCREEN] = combine_screen_u;
1.2392 + imp->combine_64[PIXMAN_OP_OVERLAY] = combine_overlay_u;
1.2393 + imp->combine_64[PIXMAN_OP_DARKEN] = combine_darken_u;
1.2394 + imp->combine_64[PIXMAN_OP_LIGHTEN] = combine_lighten_u;
1.2395 + imp->combine_64[PIXMAN_OP_COLOR_DODGE] = combine_color_dodge_u;
1.2396 + imp->combine_64[PIXMAN_OP_COLOR_BURN] = combine_color_burn_u;
1.2397 + imp->combine_64[PIXMAN_OP_HARD_LIGHT] = combine_hard_light_u;
1.2398 + imp->combine_64[PIXMAN_OP_SOFT_LIGHT] = combine_soft_light_u;
1.2399 + imp->combine_64[PIXMAN_OP_DIFFERENCE] = combine_difference_u;
1.2400 + imp->combine_64[PIXMAN_OP_EXCLUSION] = combine_exclusion_u;
1.2401 + imp->combine_64[PIXMAN_OP_HSL_HUE] = combine_hsl_hue_u;
1.2402 + imp->combine_64[PIXMAN_OP_HSL_SATURATION] = combine_hsl_saturation_u;
1.2403 + imp->combine_64[PIXMAN_OP_HSL_COLOR] = combine_hsl_color_u;
1.2404 + imp->combine_64[PIXMAN_OP_HSL_LUMINOSITY] = combine_hsl_luminosity_u;
1.2405 +
1.2406 + /* Component alpha combiners */
1.2407 + imp->combine_64_ca[PIXMAN_OP_CLEAR] = combine_clear_ca;
1.2408 + imp->combine_64_ca[PIXMAN_OP_SRC] = combine_src_ca;
1.2409 + /* dest */
1.2410 + imp->combine_64_ca[PIXMAN_OP_OVER] = combine_over_ca;
1.2411 + imp->combine_64_ca[PIXMAN_OP_OVER_REVERSE] = combine_over_reverse_ca;
1.2412 + imp->combine_64_ca[PIXMAN_OP_IN] = combine_in_ca;
1.2413 + imp->combine_64_ca[PIXMAN_OP_IN_REVERSE] = combine_in_reverse_ca;
1.2414 + imp->combine_64_ca[PIXMAN_OP_OUT] = combine_out_ca;
1.2415 + imp->combine_64_ca[PIXMAN_OP_OUT_REVERSE] = combine_out_reverse_ca;
1.2416 + imp->combine_64_ca[PIXMAN_OP_ATOP] = combine_atop_ca;
1.2417 + imp->combine_64_ca[PIXMAN_OP_ATOP_REVERSE] = combine_atop_reverse_ca;
1.2418 + imp->combine_64_ca[PIXMAN_OP_XOR] = combine_xor_ca;
1.2419 + imp->combine_64_ca[PIXMAN_OP_ADD] = combine_add_ca;
1.2420 + imp->combine_64_ca[PIXMAN_OP_SATURATE] = combine_saturate_ca;
1.2421 +
1.2422 + /* Disjoint CA */
1.2423 + imp->combine_64_ca[PIXMAN_OP_DISJOINT_CLEAR] = combine_clear_ca;
1.2424 + imp->combine_64_ca[PIXMAN_OP_DISJOINT_SRC] = combine_src_ca;
1.2425 + imp->combine_64_ca[PIXMAN_OP_DISJOINT_DST] = combine_dst;
1.2426 + imp->combine_64_ca[PIXMAN_OP_DISJOINT_OVER] = combine_disjoint_over_ca;
1.2427 + imp->combine_64_ca[PIXMAN_OP_DISJOINT_OVER_REVERSE] = combine_saturate_ca;
1.2428 + imp->combine_64_ca[PIXMAN_OP_DISJOINT_IN] = combine_disjoint_in_ca;
1.2429 + imp->combine_64_ca[PIXMAN_OP_DISJOINT_IN_REVERSE] = combine_disjoint_in_reverse_ca;
1.2430 + imp->combine_64_ca[PIXMAN_OP_DISJOINT_OUT] = combine_disjoint_out_ca;
1.2431 + imp->combine_64_ca[PIXMAN_OP_DISJOINT_OUT_REVERSE] = combine_disjoint_out_reverse_ca;
1.2432 + imp->combine_64_ca[PIXMAN_OP_DISJOINT_ATOP] = combine_disjoint_atop_ca;
1.2433 + imp->combine_64_ca[PIXMAN_OP_DISJOINT_ATOP_REVERSE] = combine_disjoint_atop_reverse_ca;
1.2434 + imp->combine_64_ca[PIXMAN_OP_DISJOINT_XOR] = combine_disjoint_xor_ca;
1.2435 +
1.2436 + /* Conjoint CA */
1.2437 + imp->combine_64_ca[PIXMAN_OP_CONJOINT_CLEAR] = combine_clear_ca;
1.2438 + imp->combine_64_ca[PIXMAN_OP_CONJOINT_SRC] = combine_src_ca;
1.2439 + imp->combine_64_ca[PIXMAN_OP_CONJOINT_DST] = combine_dst;
1.2440 + imp->combine_64_ca[PIXMAN_OP_CONJOINT_OVER] = combine_conjoint_over_ca;
1.2441 + imp->combine_64_ca[PIXMAN_OP_CONJOINT_OVER_REVERSE] = combine_conjoint_over_reverse_ca;
1.2442 + imp->combine_64_ca[PIXMAN_OP_CONJOINT_IN] = combine_conjoint_in_ca;
1.2443 + imp->combine_64_ca[PIXMAN_OP_CONJOINT_IN_REVERSE] = combine_conjoint_in_reverse_ca;
1.2444 + imp->combine_64_ca[PIXMAN_OP_CONJOINT_OUT] = combine_conjoint_out_ca;
1.2445 + imp->combine_64_ca[PIXMAN_OP_CONJOINT_OUT_REVERSE] = combine_conjoint_out_reverse_ca;
1.2446 + imp->combine_64_ca[PIXMAN_OP_CONJOINT_ATOP] = combine_conjoint_atop_ca;
1.2447 + imp->combine_64_ca[PIXMAN_OP_CONJOINT_ATOP_REVERSE] = combine_conjoint_atop_reverse_ca;
1.2448 + imp->combine_64_ca[PIXMAN_OP_CONJOINT_XOR] = combine_conjoint_xor_ca;
1.2449 +
1.2450 + imp->combine_64_ca[PIXMAN_OP_MULTIPLY] = combine_multiply_ca;
1.2451 + imp->combine_64_ca[PIXMAN_OP_SCREEN] = combine_screen_ca;
1.2452 + imp->combine_64_ca[PIXMAN_OP_OVERLAY] = combine_overlay_ca;
1.2453 + imp->combine_64_ca[PIXMAN_OP_DARKEN] = combine_darken_ca;
1.2454 + imp->combine_64_ca[PIXMAN_OP_LIGHTEN] = combine_lighten_ca;
1.2455 + imp->combine_64_ca[PIXMAN_OP_COLOR_DODGE] = combine_color_dodge_ca;
1.2456 + imp->combine_64_ca[PIXMAN_OP_COLOR_BURN] = combine_color_burn_ca;
1.2457 + imp->combine_64_ca[PIXMAN_OP_HARD_LIGHT] = combine_hard_light_ca;
1.2458 + imp->combine_64_ca[PIXMAN_OP_SOFT_LIGHT] = combine_soft_light_ca;
1.2459 + imp->combine_64_ca[PIXMAN_OP_DIFFERENCE] = combine_difference_ca;
1.2460 + imp->combine_64_ca[PIXMAN_OP_EXCLUSION] = combine_exclusion_ca;
1.2461 +
1.2462 + /* It is not clear that these make sense, so make them noops for now */
1.2463 + imp->combine_64_ca[PIXMAN_OP_HSL_HUE] = combine_dst;
1.2464 + imp->combine_64_ca[PIXMAN_OP_HSL_SATURATION] = combine_dst;
1.2465 + imp->combine_64_ca[PIXMAN_OP_HSL_COLOR] = combine_dst;
1.2466 + imp->combine_64_ca[PIXMAN_OP_HSL_LUMINOSITY] = combine_dst;
1.2467 +}
1.2468 +
