michael@0: /* WARNING: This file is generated by combine.pl from combine.inc.
michael@0:    Please edit one of those files rather than this one. */
michael@0: 
michael@0: #line 1 "pixman-combine.c.template"
michael@0: #ifdef HAVE_CONFIG_H
michael@0: #include <config.h>
michael@0: #endif
michael@0: 
michael@0: #include <math.h>
michael@0: #include <string.h>
michael@0: 
michael@0: #include "pixman-private.h"
michael@0: 
michael@0: #include "pixman-combine64.h"
michael@0: 
michael@0: /*** per channel helper functions ***/
michael@0: 
michael@0: static void
michael@0: combine_mask_ca (uint64_t *src, uint64_t *mask)
michael@0: {
michael@0:     uint64_t a = *mask;
michael@0: 
michael@0:     uint64_t x;
michael@0:     uint32_t xa;
michael@0: 
michael@0:     if (!a)
michael@0:     {
michael@0: 	*(src) = 0;
michael@0: 	return;
michael@0:     }
michael@0: 
michael@0:     x = *(src);
michael@0:     if (a == ~0)
michael@0:     {
michael@0: 	x = x >> A_SHIFT;
michael@0: 	x |= x << G_SHIFT;
michael@0: 	x |= x << R_SHIFT;
michael@0: 	*(mask) = x;
michael@0: 	return;
michael@0:     }
michael@0: 
michael@0:     xa = x >> A_SHIFT;
michael@0:     UN16x4_MUL_UN16x4 (x, a);
michael@0:     *(src) = x;
michael@0:     
michael@0:     UN16x4_MUL_UN16 (a, xa);
michael@0:     *(mask) = a;
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_mask_value_ca (uint64_t *src, const uint64_t *mask)
michael@0: {
michael@0:     uint64_t a = *mask;
michael@0:     uint64_t x;
michael@0: 
michael@0:     if (!a)
michael@0:     {
michael@0: 	*(src) = 0;
michael@0: 	return;
michael@0:     }
michael@0: 
michael@0:     if (a == ~0)
michael@0: 	return;
michael@0: 
michael@0:     x = *(src);
michael@0:     UN16x4_MUL_UN16x4 (x, a);
michael@0:     *(src) = x;
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_mask_alpha_ca (const uint64_t *src, uint64_t *mask)
michael@0: {
michael@0:     uint64_t a = *(mask);
michael@0:     uint64_t x;
michael@0: 
michael@0:     if (!a)
michael@0: 	return;
michael@0: 
michael@0:     x = *(src) >> A_SHIFT;
michael@0:     if (x == MASK)
michael@0: 	return;
michael@0: 
michael@0:     if (a == ~0)
michael@0:     {
michael@0: 	x |= x << G_SHIFT;
michael@0: 	x |= x << R_SHIFT;
michael@0: 	*(mask) = x;
michael@0: 	return;
michael@0:     }
michael@0: 
michael@0:     UN16x4_MUL_UN16 (a, x);
michael@0:     *(mask) = a;
michael@0: }
michael@0: 
michael@0: /*
michael@0:  * There are two ways of handling alpha -- either as a single unified value or
michael@0:  * a separate value for each component, hence each macro must have two
michael@0:  * versions.  The unified alpha version has a 'U' at the end of the name,
michael@0:  * the component version has a 'C'.  Similarly, functions which deal with
michael@0:  * this difference will have two versions using the same convention.
michael@0:  */
michael@0: 
michael@0: /*
michael@0:  * All of the composing functions
michael@0:  */
michael@0: 
michael@0: static force_inline uint64_t
michael@0: combine_mask (const uint64_t *src, const uint64_t *mask, int i)
michael@0: {
michael@0:     uint64_t s, m;
michael@0: 
michael@0:     if (mask)
michael@0:     {
michael@0: 	m = *(mask + i) >> A_SHIFT;
michael@0: 
michael@0: 	if (!m)
michael@0: 	    return 0;
michael@0:     }
michael@0: 
michael@0:     s = *(src + i);
michael@0: 
michael@0:     if (mask)
michael@0: 	UN16x4_MUL_UN16 (s, m);
michael@0: 
michael@0:     return s;
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_clear (pixman_implementation_t *imp,
michael@0:                pixman_op_t              op,
michael@0:                uint64_t *                dest,
michael@0:                const uint64_t *          src,
michael@0:                const uint64_t *          mask,
michael@0:                int                      width)
michael@0: {
michael@0:     memset (dest, 0, width * sizeof(uint64_t));
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_dst (pixman_implementation_t *imp,
michael@0: 	     pixman_op_t	      op,
michael@0: 	     uint64_t *		      dest,
michael@0: 	     const uint64_t *	      src,
michael@0: 	     const uint64_t *          mask,
michael@0: 	     int		      width)
michael@0: {
michael@0:     return;
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_src_u (pixman_implementation_t *imp,
michael@0:                pixman_op_t              op,
michael@0:                uint64_t *                dest,
michael@0:                const uint64_t *          src,
michael@0:                const uint64_t *          mask,
michael@0:                int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     if (!mask)
michael@0: 	memcpy (dest, src, width * sizeof (uint64_t));
michael@0:     else
michael@0:     {
michael@0: 	for (i = 0; i < width; ++i)
michael@0: 	{
michael@0: 	    uint64_t s = combine_mask (src, mask, i);
michael@0: 
michael@0: 	    *(dest + i) = s;
michael@0: 	}
michael@0:     }
michael@0: }
michael@0: 
michael@0: /* if the Src is opaque, call combine_src_u */
michael@0: static void
michael@0: combine_over_u (pixman_implementation_t *imp,
michael@0:                 pixman_op_t              op,
michael@0:                 uint64_t *                dest,
michael@0:                 const uint64_t *          src,
michael@0:                 const uint64_t *          mask,
michael@0:                 int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t s = combine_mask (src, mask, i);
michael@0: 	uint64_t d = *(dest + i);
michael@0: 	uint64_t ia = ALPHA_16 (~s);
michael@0: 
michael@0: 	UN16x4_MUL_UN16_ADD_UN16x4 (d, ia, s);
michael@0: 	*(dest + i) = d;
michael@0:     }
michael@0: }
michael@0: 
michael@0: /* if the Dst is opaque, this is a noop */
michael@0: static void
michael@0: combine_over_reverse_u (pixman_implementation_t *imp,
michael@0:                         pixman_op_t              op,
michael@0:                         uint64_t *                dest,
michael@0:                         const uint64_t *          src,
michael@0:                         const uint64_t *          mask,
michael@0:                         int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t s = combine_mask (src, mask, i);
michael@0: 	uint64_t d = *(dest + i);
michael@0: 	uint64_t ia = ALPHA_16 (~*(dest + i));
michael@0: 	UN16x4_MUL_UN16_ADD_UN16x4 (s, ia, d);
michael@0: 	*(dest + i) = s;
michael@0:     }
michael@0: }
michael@0: 
michael@0: /* if the Dst is opaque, call combine_src_u */
michael@0: static void
michael@0: combine_in_u (pixman_implementation_t *imp,
michael@0:               pixman_op_t              op,
michael@0:               uint64_t *                dest,
michael@0:               const uint64_t *          src,
michael@0:               const uint64_t *          mask,
michael@0:               int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t s = combine_mask (src, mask, i);
michael@0: 	uint64_t a = ALPHA_16 (*(dest + i));
michael@0: 	UN16x4_MUL_UN16 (s, a);
michael@0: 	*(dest + i) = s;
michael@0:     }
michael@0: }
michael@0: 
michael@0: /* if the Src is opaque, this is a noop */
michael@0: static void
michael@0: combine_in_reverse_u (pixman_implementation_t *imp,
michael@0:                       pixman_op_t              op,
michael@0:                       uint64_t *                dest,
michael@0:                       const uint64_t *          src,
michael@0:                       const uint64_t *          mask,
michael@0:                       int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t s = combine_mask (src, mask, i);
michael@0: 	uint64_t d = *(dest + i);
michael@0: 	uint64_t a = ALPHA_16 (s);
michael@0: 	UN16x4_MUL_UN16 (d, a);
michael@0: 	*(dest + i) = d;
michael@0:     }
michael@0: }
michael@0: 
michael@0: /* if the Dst is opaque, call combine_clear */
michael@0: static void
michael@0: combine_out_u (pixman_implementation_t *imp,
michael@0:                pixman_op_t              op,
michael@0:                uint64_t *                dest,
michael@0:                const uint64_t *          src,
michael@0:                const uint64_t *          mask,
michael@0:                int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t s = combine_mask (src, mask, i);
michael@0: 	uint64_t a = ALPHA_16 (~*(dest + i));
michael@0: 	UN16x4_MUL_UN16 (s, a);
michael@0: 	*(dest + i) = s;
michael@0:     }
michael@0: }
michael@0: 
michael@0: /* if the Src is opaque, call combine_clear */
michael@0: static void
michael@0: combine_out_reverse_u (pixman_implementation_t *imp,
michael@0:                        pixman_op_t              op,
michael@0:                        uint64_t *                dest,
michael@0:                        const uint64_t *          src,
michael@0:                        const uint64_t *          mask,
michael@0:                        int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t s = combine_mask (src, mask, i);
michael@0: 	uint64_t d = *(dest + i);
michael@0: 	uint64_t a = ALPHA_16 (~s);
michael@0: 	UN16x4_MUL_UN16 (d, a);
michael@0: 	*(dest + i) = d;
michael@0:     }
michael@0: }
michael@0: 
michael@0: /* if the Src is opaque, call combine_in_u */
michael@0: /* if the Dst is opaque, call combine_over_u */
michael@0: /* if both the Src and Dst are opaque, call combine_src_u */
michael@0: static void
michael@0: combine_atop_u (pixman_implementation_t *imp,
michael@0:                 pixman_op_t              op,
michael@0:                 uint64_t *                dest,
michael@0:                 const uint64_t *          src,
michael@0:                 const uint64_t *          mask,
michael@0:                 int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t s = combine_mask (src, mask, i);
michael@0: 	uint64_t d = *(dest + i);
michael@0: 	uint64_t dest_a = ALPHA_16 (d);
michael@0: 	uint64_t src_ia = ALPHA_16 (~s);
michael@0: 
michael@0: 	UN16x4_MUL_UN16_ADD_UN16x4_MUL_UN16 (s, dest_a, d, src_ia);
michael@0: 	*(dest + i) = s;
michael@0:     }
michael@0: }
michael@0: 
michael@0: /* if the Src is opaque, call combine_over_reverse_u */
michael@0: /* if the Dst is opaque, call combine_in_reverse_u */
michael@0: /* if both the Src and Dst are opaque, call combine_dst_u */
michael@0: static void
michael@0: combine_atop_reverse_u (pixman_implementation_t *imp,
michael@0:                         pixman_op_t              op,
michael@0:                         uint64_t *                dest,
michael@0:                         const uint64_t *          src,
michael@0:                         const uint64_t *          mask,
michael@0:                         int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t s = combine_mask (src, mask, i);
michael@0: 	uint64_t d = *(dest + i);
michael@0: 	uint64_t src_a = ALPHA_16 (s);
michael@0: 	uint64_t dest_ia = ALPHA_16 (~d);
michael@0: 
michael@0: 	UN16x4_MUL_UN16_ADD_UN16x4_MUL_UN16 (s, dest_ia, d, src_a);
michael@0: 	*(dest + i) = s;
michael@0:     }
michael@0: }
michael@0: 
michael@0: /* if the Src is opaque, call combine_over_u */
michael@0: /* if the Dst is opaque, call combine_over_reverse_u */
michael@0: /* if both the Src and Dst are opaque, call combine_clear */
michael@0: static void
michael@0: combine_xor_u (pixman_implementation_t *imp,
michael@0:                pixman_op_t              op,
michael@0:                uint64_t *                dest,
michael@0:                const uint64_t *          src,
michael@0:                const uint64_t *          mask,
michael@0:                int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t s = combine_mask (src, mask, i);
michael@0: 	uint64_t d = *(dest + i);
michael@0: 	uint64_t src_ia = ALPHA_16 (~s);
michael@0: 	uint64_t dest_ia = ALPHA_16 (~d);
michael@0: 
michael@0: 	UN16x4_MUL_UN16_ADD_UN16x4_MUL_UN16 (s, dest_ia, d, src_ia);
michael@0: 	*(dest + i) = s;
michael@0:     }
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_add_u (pixman_implementation_t *imp,
michael@0:                pixman_op_t              op,
michael@0:                uint64_t *                dest,
michael@0:                const uint64_t *          src,
michael@0:                const uint64_t *          mask,
michael@0:                int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t s = combine_mask (src, mask, i);
michael@0: 	uint64_t d = *(dest + i);
michael@0: 	UN16x4_ADD_UN16x4 (d, s);
michael@0: 	*(dest + i) = d;
michael@0:     }
michael@0: }
michael@0: 
michael@0: /* if the Src is opaque, call combine_add_u */
michael@0: /* if the Dst is opaque, call combine_add_u */
michael@0: /* if both the Src and Dst are opaque, call combine_add_u */
michael@0: static void
michael@0: combine_saturate_u (pixman_implementation_t *imp,
michael@0:                     pixman_op_t              op,
michael@0:                     uint64_t *                dest,
michael@0:                     const uint64_t *          src,
michael@0:                     const uint64_t *          mask,
michael@0:                     int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t s = combine_mask (src, mask, i);
michael@0: 	uint64_t d = *(dest + i);
michael@0: 	uint32_t sa, da;
michael@0: 
michael@0: 	sa = s >> A_SHIFT;
michael@0: 	da = ~d >> A_SHIFT;
michael@0: 	if (sa > da)
michael@0: 	{
michael@0: 	    sa = DIV_UN16 (da, sa);
michael@0: 	    UN16x4_MUL_UN16 (s, sa);
michael@0: 	}
michael@0: 	;
michael@0: 	UN16x4_ADD_UN16x4 (d, s);
michael@0: 	*(dest + i) = d;
michael@0:     }
michael@0: }
michael@0: 
michael@0: /*
michael@0:  * PDF blend modes:
michael@0:  * The following blend modes have been taken from the PDF ISO 32000
michael@0:  * specification, which at this point in time is available from
michael@0:  * http://www.adobe.com/devnet/acrobat/pdfs/PDF32000_2008.pdf
michael@0:  * The relevant chapters are 11.3.5 and 11.3.6.
michael@0:  * The formula for computing the final pixel color given in 11.3.6 is:
michael@0:  * αr × Cr = (1 – αs) × αb × Cb + (1 – αb) × αs × Cs + αb × αs × B(Cb, Cs)
michael@0:  * with B() being the blend function.
michael@0:  * Note that OVER is a special case of this operation, using B(Cb, Cs) = Cs
michael@0:  *
michael@0:  * These blend modes should match the SVG filter draft specification, as
michael@0:  * it has been designed to mirror ISO 32000. Note that at the current point
michael@0:  * no released draft exists that shows this, as the formulas have not been
michael@0:  * updated yet after the release of ISO 32000.
michael@0:  *
michael@0:  * The default implementation here uses the PDF_SEPARABLE_BLEND_MODE and
michael@0:  * PDF_NON_SEPARABLE_BLEND_MODE macros, which take the blend function as an
michael@0:  * argument. Note that this implementation operates on premultiplied colors,
michael@0:  * while the PDF specification does not. Therefore the code uses the formula
michael@0:  * Cra = (1 – as) . Dca + (1 – ad) . Sca + B(Dca, ad, Sca, as)
michael@0:  */
michael@0: 
michael@0: /*
michael@0:  * Multiply
michael@0:  * B(Dca, ad, Sca, as) = Dca.Sca
michael@0:  */
michael@0: 
michael@0: static void
michael@0: combine_multiply_u (pixman_implementation_t *imp,
michael@0:                     pixman_op_t              op,
michael@0:                     uint64_t *                dest,
michael@0:                     const uint64_t *          src,
michael@0:                     const uint64_t *          mask,
michael@0:                     int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t s = combine_mask (src, mask, i);
michael@0: 	uint64_t d = *(dest + i);
michael@0: 	uint64_t ss = s;
michael@0: 	uint64_t src_ia = ALPHA_16 (~s);
michael@0: 	uint64_t dest_ia = ALPHA_16 (~d);
michael@0: 
michael@0: 	UN16x4_MUL_UN16_ADD_UN16x4_MUL_UN16 (ss, dest_ia, d, src_ia);
michael@0: 	UN16x4_MUL_UN16x4 (d, s);
michael@0: 	UN16x4_ADD_UN16x4 (d, ss);
michael@0: 
michael@0: 	*(dest + i) = d;
michael@0:     }
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_multiply_ca (pixman_implementation_t *imp,
michael@0:                      pixman_op_t              op,
michael@0:                      uint64_t *                dest,
michael@0:                      const uint64_t *          src,
michael@0:                      const uint64_t *          mask,
michael@0:                      int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t m = *(mask + i);
michael@0: 	uint64_t s = *(src + i);
michael@0: 	uint64_t d = *(dest + i);
michael@0: 	uint64_t r = d;
michael@0: 	uint64_t dest_ia = ALPHA_16 (~d);
michael@0: 
michael@0: 	combine_mask_value_ca (&s, &m);
michael@0: 
michael@0: 	UN16x4_MUL_UN16x4_ADD_UN16x4_MUL_UN16 (r, ~m, s, dest_ia);
michael@0: 	UN16x4_MUL_UN16x4 (d, s);
michael@0: 	UN16x4_ADD_UN16x4 (r, d);
michael@0: 
michael@0: 	*(dest + i) = r;
michael@0:     }
michael@0: }
michael@0: 
michael@0: #define PDF_SEPARABLE_BLEND_MODE(name)					\
michael@0:     static void								\
michael@0:     combine_ ## name ## _u (pixman_implementation_t *imp,		\
michael@0: 			    pixman_op_t              op,		\
michael@0:                             uint64_t *                dest,		\
michael@0: 			    const uint64_t *          src,		\
michael@0: 			    const uint64_t *          mask,		\
michael@0: 			    int                      width)		\
michael@0:     {									\
michael@0: 	int i;								\
michael@0: 	for (i = 0; i < width; ++i) {					\
michael@0: 	    uint64_t s = combine_mask (src, mask, i);			\
michael@0: 	    uint64_t d = *(dest + i);					\
michael@0: 	    uint16_t sa = ALPHA_16 (s);					\
michael@0: 	    uint16_t isa = ~sa;						\
michael@0: 	    uint16_t da = ALPHA_16 (d);					\
michael@0: 	    uint16_t ida = ~da;						\
michael@0: 	    uint64_t result;						\
michael@0: 									\
michael@0: 	    result = d;							\
michael@0: 	    UN16x4_MUL_UN16_ADD_UN16x4_MUL_UN16 (result, isa, s, ida);	\
michael@0: 	    								\
michael@0: 	    *(dest + i) = result +					\
michael@0: 		(DIV_ONE_UN16 (sa * (uint64_t)da) << A_SHIFT) +		\
michael@0: 		(blend_ ## name (RED_16 (d), da, RED_16 (s), sa) << R_SHIFT) + \
michael@0: 		(blend_ ## name (GREEN_16 (d), da, GREEN_16 (s), sa) << G_SHIFT) + \
michael@0: 		(blend_ ## name (BLUE_16 (d), da, BLUE_16 (s), sa));	\
michael@0: 	}								\
michael@0:     }									\
michael@0:     									\
michael@0:     static void								\
michael@0:     combine_ ## name ## _ca (pixman_implementation_t *imp,		\
michael@0: 			     pixman_op_t              op,		\
michael@0:                              uint64_t *                dest,		\
michael@0: 			     const uint64_t *          src,		\
michael@0: 			     const uint64_t *          mask,		\
michael@0: 			     int                     width)		\
michael@0:     {									\
michael@0: 	int i;								\
michael@0: 	for (i = 0; i < width; ++i) {					\
michael@0: 	    uint64_t m = *(mask + i);					\
michael@0: 	    uint64_t s = *(src + i);					\
michael@0: 	    uint64_t d = *(dest + i);					\
michael@0: 	    uint16_t da = ALPHA_16 (d);					\
michael@0: 	    uint16_t ida = ~da;						\
michael@0: 	    uint64_t result;						\
michael@0:             								\
michael@0: 	    combine_mask_value_ca (&s, &m);				\
michael@0:             								\
michael@0: 	    result = d;							\
michael@0: 	    UN16x4_MUL_UN16x4_ADD_UN16x4_MUL_UN16 (result, ~m, s, ida);     \
michael@0:             								\
michael@0: 	    result +=							\
michael@0: 	        (DIV_ONE_UN16 (ALPHA_16 (m) * (uint64_t)da) << A_SHIFT) +	\
michael@0: 	        (blend_ ## name (RED_16 (d), da, RED_16 (s), RED_16 (m)) << R_SHIFT) + \
michael@0: 	        (blend_ ## name (GREEN_16 (d), da, GREEN_16 (s), GREEN_16 (m)) << G_SHIFT) + \
michael@0: 	        (blend_ ## name (BLUE_16 (d), da, BLUE_16 (s), BLUE_16 (m))); \
michael@0: 	    								\
michael@0: 	    *(dest + i) = result;					\
michael@0: 	}								\
michael@0:     }
michael@0: 
michael@0: /*
michael@0:  * Screen
michael@0:  * B(Dca, ad, Sca, as) = Dca.sa + Sca.da - Dca.Sca
michael@0:  */
michael@0: static inline uint64_t
michael@0: blend_screen (uint64_t dca, uint64_t da, uint64_t sca, uint64_t sa)
michael@0: {
michael@0:     return DIV_ONE_UN16 (sca * da + dca * sa - sca * dca);
michael@0: }
michael@0: 
michael@0: PDF_SEPARABLE_BLEND_MODE (screen)
michael@0: 
michael@0: /*
michael@0:  * Overlay
michael@0:  * B(Dca, Da, Sca, Sa) =
michael@0:  *   if 2.Dca < Da
michael@0:  *     2.Sca.Dca
michael@0:  *   otherwise
michael@0:  *     Sa.Da - 2.(Da - Dca).(Sa - Sca)
michael@0:  */
michael@0: static inline uint64_t
michael@0: blend_overlay (uint64_t dca, uint64_t da, uint64_t sca, uint64_t sa)
michael@0: {
michael@0:     uint64_t rca;
michael@0: 
michael@0:     if (2 * dca < da)
michael@0: 	rca = 2 * sca * dca;
michael@0:     else
michael@0: 	rca = sa * da - 2 * (da - dca) * (sa - sca);
michael@0:     return DIV_ONE_UN16 (rca);
michael@0: }
michael@0: 
michael@0: PDF_SEPARABLE_BLEND_MODE (overlay)
michael@0: 
michael@0: /*
michael@0:  * Darken
michael@0:  * B(Dca, Da, Sca, Sa) = min (Sca.Da, Dca.Sa)
michael@0:  */
michael@0: static inline uint64_t
michael@0: blend_darken (uint64_t dca, uint64_t da, uint64_t sca, uint64_t sa)
michael@0: {
michael@0:     uint64_t s, d;
michael@0: 
michael@0:     s = sca * da;
michael@0:     d = dca * sa;
michael@0:     return DIV_ONE_UN16 (s > d ? d : s);
michael@0: }
michael@0: 
michael@0: PDF_SEPARABLE_BLEND_MODE (darken)
michael@0: 
michael@0: /*
michael@0:  * Lighten
michael@0:  * B(Dca, Da, Sca, Sa) = max (Sca.Da, Dca.Sa)
michael@0:  */
michael@0: static inline uint64_t
michael@0: blend_lighten (uint64_t dca, uint64_t da, uint64_t sca, uint64_t sa)
michael@0: {
michael@0:     uint64_t s, d;
michael@0: 
michael@0:     s = sca * da;
michael@0:     d = dca * sa;
michael@0:     return DIV_ONE_UN16 (s > d ? s : d);
michael@0: }
michael@0: 
michael@0: PDF_SEPARABLE_BLEND_MODE (lighten)
michael@0: 
michael@0: /*
michael@0:  * Color dodge
michael@0:  * B(Dca, Da, Sca, Sa) =
michael@0:  *   if Dca == 0
michael@0:  *     0
michael@0:  *   if Sca == Sa
michael@0:  *     Sa.Da
michael@0:  *   otherwise
michael@0:  *     Sa.Da. min (1, Dca / Da / (1 - Sca/Sa))
michael@0:  */
michael@0: static inline uint64_t
michael@0: blend_color_dodge (uint64_t dca, uint64_t da, uint64_t sca, uint64_t sa)
michael@0: {
michael@0:     if (sca >= sa)
michael@0:     {
michael@0: 	return dca == 0 ? 0 : DIV_ONE_UN16 (sa * da);
michael@0:     }
michael@0:     else
michael@0:     {
michael@0: 	uint64_t rca = dca * sa / (sa - sca);
michael@0: 	return DIV_ONE_UN16 (sa * MIN (rca, da));
michael@0:     }
michael@0: }
michael@0: 
michael@0: PDF_SEPARABLE_BLEND_MODE (color_dodge)
michael@0: 
michael@0: /*
michael@0:  * Color burn
michael@0:  * B(Dca, Da, Sca, Sa) =
michael@0:  *   if Dca == Da
michael@0:  *     Sa.Da
michael@0:  *   if Sca == 0
michael@0:  *     0
michael@0:  *   otherwise
michael@0:  *     Sa.Da.(1 - min (1, (1 - Dca/Da).Sa / Sca))
michael@0:  */
michael@0: static inline uint64_t
michael@0: blend_color_burn (uint64_t dca, uint64_t da, uint64_t sca, uint64_t sa)
michael@0: {
michael@0:     if (sca == 0)
michael@0:     {
michael@0: 	return dca < da ? 0 : DIV_ONE_UN16 (sa * da);
michael@0:     }
michael@0:     else
michael@0:     {
michael@0: 	uint64_t rca = (da - dca) * sa / sca;
michael@0: 	return DIV_ONE_UN16 (sa * (MAX (rca, da) - rca));
michael@0:     }
michael@0: }
michael@0: 
michael@0: PDF_SEPARABLE_BLEND_MODE (color_burn)
michael@0: 
michael@0: /*
michael@0:  * Hard light
michael@0:  * B(Dca, Da, Sca, Sa) =
michael@0:  *   if 2.Sca < Sa
michael@0:  *     2.Sca.Dca
michael@0:  *   otherwise
michael@0:  *     Sa.Da - 2.(Da - Dca).(Sa - Sca)
michael@0:  */
michael@0: static inline uint64_t
michael@0: blend_hard_light (uint64_t dca, uint64_t da, uint64_t sca, uint64_t sa)
michael@0: {
michael@0:     if (2 * sca < sa)
michael@0: 	return DIV_ONE_UN16 (2 * sca * dca);
michael@0:     else
michael@0: 	return DIV_ONE_UN16 (sa * da - 2 * (da - dca) * (sa - sca));
michael@0: }
michael@0: 
michael@0: PDF_SEPARABLE_BLEND_MODE (hard_light)
michael@0: 
michael@0: /*
michael@0:  * Soft light
michael@0:  * B(Dca, Da, Sca, Sa) =
michael@0:  *   if (2.Sca <= Sa)
michael@0:  *     Dca.(Sa - (1 - Dca/Da).(2.Sca - Sa))
michael@0:  *   otherwise if Dca.4 <= Da
michael@0:  *     Dca.(Sa + (2.Sca - Sa).((16.Dca/Da - 12).Dca/Da + 3)
michael@0:  *   otherwise
michael@0:  *     (Dca.Sa + (SQRT (Dca/Da).Da - Dca).(2.Sca - Sa))
michael@0:  */
michael@0: static inline uint64_t
michael@0: blend_soft_light (uint64_t dca_org,
michael@0: 		  uint64_t da_org,
michael@0: 		  uint64_t sca_org,
michael@0: 		  uint64_t sa_org)
michael@0: {
michael@0:     double dca = dca_org * (1.0 / MASK);
michael@0:     double da = da_org * (1.0 / MASK);
michael@0:     double sca = sca_org * (1.0 / MASK);
michael@0:     double sa = sa_org * (1.0 / MASK);
michael@0:     double rca;
michael@0: 
michael@0:     if (2 * sca < sa)
michael@0:     {
michael@0: 	if (da == 0)
michael@0: 	    rca = dca * sa;
michael@0: 	else
michael@0: 	    rca = dca * sa - dca * (da - dca) * (sa - 2 * sca) / da;
michael@0:     }
michael@0:     else if (da == 0)
michael@0:     {
michael@0: 	rca = 0;
michael@0:     }
michael@0:     else if (4 * dca <= da)
michael@0:     {
michael@0: 	rca = dca * sa +
michael@0: 	    (2 * sca - sa) * dca * ((16 * dca / da - 12) * dca / da + 3);
michael@0:     }
michael@0:     else
michael@0:     {
michael@0: 	rca = dca * sa + (sqrt (dca * da) - dca) * (2 * sca - sa);
michael@0:     }
michael@0:     return rca * MASK + 0.5;
michael@0: }
michael@0: 
michael@0: PDF_SEPARABLE_BLEND_MODE (soft_light)
michael@0: 
michael@0: /*
michael@0:  * Difference
michael@0:  * B(Dca, Da, Sca, Sa) = abs (Dca.Sa - Sca.Da)
michael@0:  */
michael@0: static inline uint64_t
michael@0: blend_difference (uint64_t dca, uint64_t da, uint64_t sca, uint64_t sa)
michael@0: {
michael@0:     uint64_t dcasa = dca * sa;
michael@0:     uint64_t scada = sca * da;
michael@0: 
michael@0:     if (scada < dcasa)
michael@0: 	return DIV_ONE_UN16 (dcasa - scada);
michael@0:     else
michael@0: 	return DIV_ONE_UN16 (scada - dcasa);
michael@0: }
michael@0: 
michael@0: PDF_SEPARABLE_BLEND_MODE (difference)
michael@0: 
michael@0: /*
michael@0:  * Exclusion
michael@0:  * B(Dca, Da, Sca, Sa) = (Sca.Da + Dca.Sa - 2.Sca.Dca)
michael@0:  */
michael@0: 
michael@0: /* This can be made faster by writing it directly and not using
michael@0:  * PDF_SEPARABLE_BLEND_MODE, but that's a performance optimization */
michael@0: 
michael@0: static inline uint64_t
michael@0: blend_exclusion (uint64_t dca, uint64_t da, uint64_t sca, uint64_t sa)
michael@0: {
michael@0:     return DIV_ONE_UN16 (sca * da + dca * sa - 2 * dca * sca);
michael@0: }
michael@0: 
michael@0: PDF_SEPARABLE_BLEND_MODE (exclusion)
michael@0: 
michael@0: #undef PDF_SEPARABLE_BLEND_MODE
michael@0: 
michael@0: /*
michael@0:  * PDF nonseperable blend modes are implemented using the following functions
michael@0:  * to operate in Hsl space, with Cmax, Cmid, Cmin referring to the max, mid
michael@0:  * and min value of the red, green and blue components.
michael@0:  *
michael@0:  * LUM (C) = 0.3 × Cred + 0.59 × Cgreen + 0.11 × Cblue
michael@0:  *
michael@0:  * clip_color (C):
michael@0:  *   l = LUM (C)
michael@0:  *   min = Cmin
michael@0:  *   max = Cmax
michael@0:  *   if n < 0.0
michael@0:  *     C = l + ( ( ( C – l ) × l ) ⁄ ( l – min ) )
michael@0:  *   if x > 1.0
michael@0:  *     C = l + ( ( ( C – l ) × ( 1 – l ) ) ⁄ ( max – l ) )
michael@0:  *   return C
michael@0:  *
michael@0:  * set_lum (C, l):
michael@0:  *   d = l – LUM (C)
michael@0:  *   C += d
michael@0:  *   return clip_color (C)
michael@0:  *
michael@0:  * SAT (C) = CH_MAX (C) - CH_MIN (C)
michael@0:  *
michael@0:  * set_sat (C, s):
michael@0:  *  if Cmax > Cmin
michael@0:  *    Cmid = ( ( ( Cmid – Cmin ) × s ) ⁄ ( Cmax – Cmin ) )
michael@0:  *    Cmax = s
michael@0:  *  else
michael@0:  *    Cmid = Cmax = 0.0
michael@0:  *  Cmin = 0.0
michael@0:  *  return C
michael@0:  */
michael@0: 
michael@0: /* For premultiplied colors, we need to know what happens when C is
michael@0:  * multiplied by a real number. LUM and SAT are linear:
michael@0:  *
michael@0:  *    LUM (r × C) = r × LUM (C)		SAT (r * C) = r * SAT (C)
michael@0:  *
michael@0:  * If we extend clip_color with an extra argument a and change
michael@0:  *
michael@0:  *        if x >= 1.0
michael@0:  *
michael@0:  * into
michael@0:  *
michael@0:  *        if x >= a
michael@0:  *
michael@0:  * then clip_color is also linear:
michael@0:  *
michael@0:  *    r * clip_color (C, a) = clip_color (r_c, ra);
michael@0:  *
michael@0:  * for positive r.
michael@0:  *
michael@0:  * Similarly, we can extend set_lum with an extra argument that is just passed
michael@0:  * on to clip_color:
michael@0:  *
michael@0:  *   r * set_lum ( C, l, a)
michael@0:  *
michael@0:  *   = r × clip_color ( C + l - LUM (C), a)
michael@0:  *
michael@0:  *   = clip_color ( r * C + r × l - r * LUM (C), r * a)
michael@0:  *
michael@0:  *   = set_lum ( r * C, r * l, r * a)
michael@0:  *
michael@0:  * Finally, set_sat:
michael@0:  *
michael@0:  *    r * set_sat (C, s) = set_sat (x * C, r * s)
michael@0:  *
michael@0:  * The above holds for all non-zero x, because the x'es in the fraction for
michael@0:  * C_mid cancel out. Specifically, it holds for x = r:
michael@0:  *
michael@0:  *    r * set_sat (C, s) = set_sat (r_c, rs)
michael@0:  *
michael@0:  */
michael@0: 
michael@0: /* So, for the non-separable PDF blend modes, we have (using s, d for
michael@0:  * non-premultiplied colors, and S, D for premultiplied:
michael@0:  *
michael@0:  *   Color:
michael@0:  *
michael@0:  *     a_s * a_d * B(s, d)
michael@0:  *   = a_s * a_d * set_lum (S/a_s, LUM (D/a_d), 1)
michael@0:  *   = set_lum (S * a_d, a_s * LUM (D), a_s * a_d)
michael@0:  *
michael@0:  *
michael@0:  *   Luminosity:
michael@0:  *
michael@0:  *     a_s * a_d * B(s, d)
michael@0:  *   = a_s * a_d * set_lum (D/a_d, LUM(S/a_s), 1)
michael@0:  *   = set_lum (a_s * D, a_d * LUM(S), a_s * a_d)
michael@0:  *
michael@0:  *
michael@0:  *   Saturation:
michael@0:  *
michael@0:  *     a_s * a_d * B(s, d)
michael@0:  *   = a_s * a_d * set_lum (set_sat (D/a_d, SAT (S/a_s)), LUM (D/a_d), 1)
michael@0:  *   = set_lum (a_s * a_d * set_sat (D/a_d, SAT (S/a_s)),
michael@0:  *                                        a_s * LUM (D), a_s * a_d)
michael@0:  *   = set_lum (set_sat (a_s * D, a_d * SAT (S), a_s * LUM (D), a_s * a_d))
michael@0:  *
michael@0:  *   Hue:
michael@0:  *
michael@0:  *     a_s * a_d * B(s, d)
michael@0:  *   = a_s * a_d * set_lum (set_sat (S/a_s, SAT (D/a_d)), LUM (D/a_d), 1)
michael@0:  *   = set_lum (set_sat (a_d * S, a_s * SAT (D)), a_s * LUM (D), a_s * a_d)
michael@0:  *
michael@0:  */
michael@0: 
michael@0: #define CH_MIN(c) (c[0] < c[1] ? (c[0] < c[2] ? c[0] : c[2]) : (c[1] < c[2] ? c[1] : c[2]))
michael@0: #define CH_MAX(c) (c[0] > c[1] ? (c[0] > c[2] ? c[0] : c[2]) : (c[1] > c[2] ? c[1] : c[2]))
michael@0: #define LUM(c) ((c[0] * 30 + c[1] * 59 + c[2] * 11) / 100)
michael@0: #define SAT(c) (CH_MAX (c) - CH_MIN (c))
michael@0: 
michael@0: #define PDF_NON_SEPARABLE_BLEND_MODE(name)				\
michael@0:     static void								\
michael@0:     combine_ ## name ## _u (pixman_implementation_t *imp,		\
michael@0: 			    pixman_op_t op,				\
michael@0:                             uint64_t *dest,				\
michael@0: 			    const uint64_t *src,				\
michael@0: 			    const uint64_t *mask,			\
michael@0: 			    int width)					\
michael@0:     {									\
michael@0: 	int i;								\
michael@0: 	for (i = 0; i < width; ++i)					\
michael@0: 	{								\
michael@0: 	    uint64_t s = combine_mask (src, mask, i);			\
michael@0: 	    uint64_t d = *(dest + i);					\
michael@0: 	    uint16_t sa = ALPHA_16 (s);					\
michael@0: 	    uint16_t isa = ~sa;						\
michael@0: 	    uint16_t da = ALPHA_16 (d);					\
michael@0: 	    uint16_t ida = ~da;						\
michael@0: 	    uint64_t result;						\
michael@0: 	    uint64_t sc[3], dc[3], c[3];					\
michael@0:             								\
michael@0: 	    result = d;							\
michael@0: 	    UN16x4_MUL_UN16_ADD_UN16x4_MUL_UN16 (result, isa, s, ida);	\
michael@0: 	    dc[0] = RED_16 (d);						\
michael@0: 	    sc[0] = RED_16 (s);						\
michael@0: 	    dc[1] = GREEN_16 (d);					\
michael@0: 	    sc[1] = GREEN_16 (s);					\
michael@0: 	    dc[2] = BLUE_16 (d);						\
michael@0: 	    sc[2] = BLUE_16 (s);						\
michael@0: 	    blend_ ## name (c, dc, da, sc, sa);				\
michael@0:             								\
michael@0: 	    *(dest + i) = result +					\
michael@0: 		(DIV_ONE_UN16 (sa * (uint64_t)da) << A_SHIFT) +		\
michael@0: 		(DIV_ONE_UN16 (c[0]) << R_SHIFT) +			\
michael@0: 		(DIV_ONE_UN16 (c[1]) << G_SHIFT) +			\
michael@0: 		(DIV_ONE_UN16 (c[2]));					\
michael@0: 	}								\
michael@0:     }
michael@0: 
michael@0: static void
michael@0: set_lum (uint64_t dest[3], uint64_t src[3], uint64_t sa, uint64_t lum)
michael@0: {
michael@0:     double a, l, min, max;
michael@0:     double tmp[3];
michael@0: 
michael@0:     a = sa * (1.0 / MASK);
michael@0: 
michael@0:     l = lum * (1.0 / MASK);
michael@0:     tmp[0] = src[0] * (1.0 / MASK);
michael@0:     tmp[1] = src[1] * (1.0 / MASK);
michael@0:     tmp[2] = src[2] * (1.0 / MASK);
michael@0: 
michael@0:     l = l - LUM (tmp);
michael@0:     tmp[0] += l;
michael@0:     tmp[1] += l;
michael@0:     tmp[2] += l;
michael@0: 
michael@0:     /* clip_color */
michael@0:     l = LUM (tmp);
michael@0:     min = CH_MIN (tmp);
michael@0:     max = CH_MAX (tmp);
michael@0: 
michael@0:     if (min < 0)
michael@0:     {
michael@0: 	if (l - min == 0.0)
michael@0: 	{
michael@0: 	    tmp[0] = 0;
michael@0: 	    tmp[1] = 0;
michael@0: 	    tmp[2] = 0;
michael@0: 	}
michael@0: 	else
michael@0: 	{
michael@0: 	    tmp[0] = l + (tmp[0] - l) * l / (l - min);
michael@0: 	    tmp[1] = l + (tmp[1] - l) * l / (l - min);
michael@0: 	    tmp[2] = l + (tmp[2] - l) * l / (l - min);
michael@0: 	}
michael@0:     }
michael@0:     if (max > a)
michael@0:     {
michael@0: 	if (max - l == 0.0)
michael@0: 	{
michael@0: 	    tmp[0] = a;
michael@0: 	    tmp[1] = a;
michael@0: 	    tmp[2] = a;
michael@0: 	}
michael@0: 	else
michael@0: 	{
michael@0: 	    tmp[0] = l + (tmp[0] - l) * (a - l) / (max - l);
michael@0: 	    tmp[1] = l + (tmp[1] - l) * (a - l) / (max - l);
michael@0: 	    tmp[2] = l + (tmp[2] - l) * (a - l) / (max - l);
michael@0: 	}
michael@0:     }
michael@0: 
michael@0:     dest[0] = tmp[0] * MASK + 0.5;
michael@0:     dest[1] = tmp[1] * MASK + 0.5;
michael@0:     dest[2] = tmp[2] * MASK + 0.5;
michael@0: }
michael@0: 
michael@0: static void
michael@0: set_sat (uint64_t dest[3], uint64_t src[3], uint64_t sat)
michael@0: {
michael@0:     int id[3];
michael@0:     uint64_t min, max;
michael@0: 
michael@0:     if (src[0] > src[1])
michael@0:     {
michael@0: 	if (src[0] > src[2])
michael@0: 	{
michael@0: 	    id[0] = 0;
michael@0: 	    if (src[1] > src[2])
michael@0: 	    {
michael@0: 		id[1] = 1;
michael@0: 		id[2] = 2;
michael@0: 	    }
michael@0: 	    else
michael@0: 	    {
michael@0: 		id[1] = 2;
michael@0: 		id[2] = 1;
michael@0: 	    }
michael@0: 	}
michael@0: 	else
michael@0: 	{
michael@0: 	    id[0] = 2;
michael@0: 	    id[1] = 0;
michael@0: 	    id[2] = 1;
michael@0: 	}
michael@0:     }
michael@0:     else
michael@0:     {
michael@0: 	if (src[0] > src[2])
michael@0: 	{
michael@0: 	    id[0] = 1;
michael@0: 	    id[1] = 0;
michael@0: 	    id[2] = 2;
michael@0: 	}
michael@0: 	else
michael@0: 	{
michael@0: 	    id[2] = 0;
michael@0: 	    if (src[1] > src[2])
michael@0: 	    {
michael@0: 		id[0] = 1;
michael@0: 		id[1] = 2;
michael@0: 	    }
michael@0: 	    else
michael@0: 	    {
michael@0: 		id[0] = 2;
michael@0: 		id[1] = 1;
michael@0: 	    }
michael@0: 	}
michael@0:     }
michael@0: 
michael@0:     max = dest[id[0]];
michael@0:     min = dest[id[2]];
michael@0:     if (max > min)
michael@0:     {
michael@0: 	dest[id[1]] = (dest[id[1]] - min) * sat / (max - min);
michael@0: 	dest[id[0]] = sat;
michael@0: 	dest[id[2]] = 0;
michael@0:     }
michael@0:     else
michael@0:     {
michael@0: 	dest[0] = dest[1] = dest[2] = 0;
michael@0:     }
michael@0: }
michael@0: 
michael@0: /*
michael@0:  * Hue:
michael@0:  * B(Cb, Cs) = set_lum (set_sat (Cs, SAT (Cb)), LUM (Cb))
michael@0:  */
michael@0: static inline void
michael@0: blend_hsl_hue (uint64_t c[3],
michael@0:                uint64_t dc[3],
michael@0:                uint64_t da,
michael@0:                uint64_t sc[3],
michael@0:                uint64_t sa)
michael@0: {
michael@0:     c[0] = sc[0] * da;
michael@0:     c[1] = sc[1] * da;
michael@0:     c[2] = sc[2] * da;
michael@0:     set_sat (c, c, SAT (dc) * sa);
michael@0:     set_lum (c, c, sa * da, LUM (dc) * sa);
michael@0: }
michael@0: 
michael@0: PDF_NON_SEPARABLE_BLEND_MODE (hsl_hue)
michael@0: 
michael@0: /*
michael@0:  * Saturation:
michael@0:  * B(Cb, Cs) = set_lum (set_sat (Cb, SAT (Cs)), LUM (Cb))
michael@0:  */
michael@0: static inline void
michael@0: blend_hsl_saturation (uint64_t c[3],
michael@0:                       uint64_t dc[3],
michael@0:                       uint64_t da,
michael@0:                       uint64_t sc[3],
michael@0:                       uint64_t sa)
michael@0: {
michael@0:     c[0] = dc[0] * sa;
michael@0:     c[1] = dc[1] * sa;
michael@0:     c[2] = dc[2] * sa;
michael@0:     set_sat (c, c, SAT (sc) * da);
michael@0:     set_lum (c, c, sa * da, LUM (dc) * sa);
michael@0: }
michael@0: 
michael@0: PDF_NON_SEPARABLE_BLEND_MODE (hsl_saturation)
michael@0: 
michael@0: /*
michael@0:  * Color:
michael@0:  * B(Cb, Cs) = set_lum (Cs, LUM (Cb))
michael@0:  */
michael@0: static inline void
michael@0: blend_hsl_color (uint64_t c[3],
michael@0:                  uint64_t dc[3],
michael@0:                  uint64_t da,
michael@0:                  uint64_t sc[3],
michael@0:                  uint64_t sa)
michael@0: {
michael@0:     c[0] = sc[0] * da;
michael@0:     c[1] = sc[1] * da;
michael@0:     c[2] = sc[2] * da;
michael@0:     set_lum (c, c, sa * da, LUM (dc) * sa);
michael@0: }
michael@0: 
michael@0: PDF_NON_SEPARABLE_BLEND_MODE (hsl_color)
michael@0: 
michael@0: /*
michael@0:  * Luminosity:
michael@0:  * B(Cb, Cs) = set_lum (Cb, LUM (Cs))
michael@0:  */
michael@0: static inline void
michael@0: blend_hsl_luminosity (uint64_t c[3],
michael@0:                       uint64_t dc[3],
michael@0:                       uint64_t da,
michael@0:                       uint64_t sc[3],
michael@0:                       uint64_t sa)
michael@0: {
michael@0:     c[0] = dc[0] * sa;
michael@0:     c[1] = dc[1] * sa;
michael@0:     c[2] = dc[2] * sa;
michael@0:     set_lum (c, c, sa * da, LUM (sc) * da);
michael@0: }
michael@0: 
michael@0: PDF_NON_SEPARABLE_BLEND_MODE (hsl_luminosity)
michael@0: 
michael@0: #undef SAT
michael@0: #undef LUM
michael@0: #undef CH_MAX
michael@0: #undef CH_MIN
michael@0: #undef PDF_NON_SEPARABLE_BLEND_MODE
michael@0: 
michael@0: /* All of the disjoint/conjoint composing functions
michael@0:  *
michael@0:  * The four entries in the first column indicate what source contributions
michael@0:  * come from each of the four areas of the picture -- areas covered by neither
michael@0:  * A nor B, areas covered only by A, areas covered only by B and finally
michael@0:  * areas covered by both A and B.
michael@0:  * 
michael@0:  * Disjoint			Conjoint
michael@0:  * Fa		Fb		Fa		Fb
michael@0:  * (0,0,0,0)	0		0		0		0
michael@0:  * (0,A,0,A)	1		0		1		0
michael@0:  * (0,0,B,B)	0		1		0		1
michael@0:  * (0,A,B,A)	1		min((1-a)/b,1)	1		max(1-a/b,0)
michael@0:  * (0,A,B,B)	min((1-b)/a,1)	1		max(1-b/a,0)	1
michael@0:  * (0,0,0,A)	max(1-(1-b)/a,0) 0		min(1,b/a)	0
michael@0:  * (0,0,0,B)	0		max(1-(1-a)/b,0) 0		min(a/b,1)
michael@0:  * (0,A,0,0)	min(1,(1-b)/a)	0		max(1-b/a,0)	0
michael@0:  * (0,0,B,0)	0		min(1,(1-a)/b)	0		max(1-a/b,0)
michael@0:  * (0,0,B,A)	max(1-(1-b)/a,0) min(1,(1-a)/b)	 min(1,b/a)	max(1-a/b,0)
michael@0:  * (0,A,0,B)	min(1,(1-b)/a)	max(1-(1-a)/b,0) max(1-b/a,0)	min(1,a/b)
michael@0:  * (0,A,B,0)	min(1,(1-b)/a)	min(1,(1-a)/b)	max(1-b/a,0)	max(1-a/b,0)
michael@0:  *
michael@0:  * See  http://marc.info/?l=xfree-render&m=99792000027857&w=2  for more
michael@0:  * information about these operators.
michael@0:  */
michael@0: 
michael@0: #define COMBINE_A_OUT 1
michael@0: #define COMBINE_A_IN  2
michael@0: #define COMBINE_B_OUT 4
michael@0: #define COMBINE_B_IN  8
michael@0: 
michael@0: #define COMBINE_CLEAR   0
michael@0: #define COMBINE_A       (COMBINE_A_OUT | COMBINE_A_IN)
michael@0: #define COMBINE_B       (COMBINE_B_OUT | COMBINE_B_IN)
michael@0: #define COMBINE_A_OVER  (COMBINE_A_OUT | COMBINE_B_OUT | COMBINE_A_IN)
michael@0: #define COMBINE_B_OVER  (COMBINE_A_OUT | COMBINE_B_OUT | COMBINE_B_IN)
michael@0: #define COMBINE_A_ATOP  (COMBINE_B_OUT | COMBINE_A_IN)
michael@0: #define COMBINE_B_ATOP  (COMBINE_A_OUT | COMBINE_B_IN)
michael@0: #define COMBINE_XOR     (COMBINE_A_OUT | COMBINE_B_OUT)
michael@0: 
michael@0: /* portion covered by a but not b */
michael@0: static uint16_t
michael@0: combine_disjoint_out_part (uint16_t a, uint16_t b)
michael@0: {
michael@0:     /* min (1, (1-b) / a) */
michael@0: 
michael@0:     b = ~b;                 /* 1 - b */
michael@0:     if (b >= a)             /* 1 - b >= a -> (1-b)/a >= 1 */
michael@0: 	return MASK;        /* 1 */
michael@0:     return DIV_UN16 (b, a);     /* (1-b) / a */
michael@0: }
michael@0: 
michael@0: /* portion covered by both a and b */
michael@0: static uint16_t
michael@0: combine_disjoint_in_part (uint16_t a, uint16_t b)
michael@0: {
michael@0:     /* max (1-(1-b)/a,0) */
michael@0:     /*  = - min ((1-b)/a - 1, 0) */
michael@0:     /*  = 1 - min (1, (1-b)/a) */
michael@0: 
michael@0:     b = ~b;                 /* 1 - b */
michael@0:     if (b >= a)             /* 1 - b >= a -> (1-b)/a >= 1 */
michael@0: 	return 0;           /* 1 - 1 */
michael@0:     return ~DIV_UN16(b, a);    /* 1 - (1-b) / a */
michael@0: }
michael@0: 
michael@0: /* portion covered by a but not b */
michael@0: static uint16_t
michael@0: combine_conjoint_out_part (uint16_t a, uint16_t b)
michael@0: {
michael@0:     /* max (1-b/a,0) */
michael@0:     /* = 1-min(b/a,1) */
michael@0: 
michael@0:     /* min (1, (1-b) / a) */
michael@0: 
michael@0:     if (b >= a)             /* b >= a -> b/a >= 1 */
michael@0: 	return 0x00;        /* 0 */
michael@0:     return ~DIV_UN16(b, a);    /* 1 - b/a */
michael@0: }
michael@0: 
michael@0: /* portion covered by both a and b */
michael@0: static uint16_t
michael@0: combine_conjoint_in_part (uint16_t a, uint16_t b)
michael@0: {
michael@0:     /* min (1,b/a) */
michael@0: 
michael@0:     if (b >= a)             /* b >= a -> b/a >= 1 */
michael@0: 	return MASK;        /* 1 */
michael@0:     return DIV_UN16 (b, a);     /* b/a */
michael@0: }
michael@0: 
michael@0: #define GET_COMP(v, i)   ((uint32_t) (uint16_t) ((v) >> i))
michael@0: 
michael@0: #define ADD(x, y, i, t)							\
michael@0:     ((t) = GET_COMP (x, i) + GET_COMP (y, i),				\
michael@0:      (uint64_t) ((uint16_t) ((t) | (0 - ((t) >> G_SHIFT)))) << (i))
michael@0: 
michael@0: #define GENERIC(x, y, i, ax, ay, t, u, v)				\
michael@0:     ((t) = (MUL_UN16 (GET_COMP (y, i), ay, (u)) +			\
michael@0:             MUL_UN16 (GET_COMP (x, i), ax, (v))),			\
michael@0:      (uint64_t) ((uint16_t) ((t) |					\
michael@0:                            (0 - ((t) >> G_SHIFT)))) << (i))
michael@0: 
michael@0: static void
michael@0: combine_disjoint_general_u (uint64_t *      dest,
michael@0:                             const uint64_t *src,
michael@0:                             const uint64_t *mask,
michael@0:                             int            width,
michael@0:                             uint16_t        combine)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t s = combine_mask (src, mask, i);
michael@0: 	uint64_t d = *(dest + i);
michael@0: 	uint64_t m, n, o, p;
michael@0: 	uint32_t Fa, Fb, t, u, v;
michael@0: 	uint16_t sa = s >> A_SHIFT;
michael@0: 	uint16_t da = d >> A_SHIFT;
michael@0: 
michael@0: 	switch (combine & COMBINE_A)
michael@0: 	{
michael@0: 	default:
michael@0: 	    Fa = 0;
michael@0: 	    break;
michael@0: 
michael@0: 	case COMBINE_A_OUT:
michael@0: 	    Fa = combine_disjoint_out_part (sa, da);
michael@0: 	    break;
michael@0: 
michael@0: 	case COMBINE_A_IN:
michael@0: 	    Fa = combine_disjoint_in_part (sa, da);
michael@0: 	    break;
michael@0: 
michael@0: 	case COMBINE_A:
michael@0: 	    Fa = MASK;
michael@0: 	    break;
michael@0: 	}
michael@0: 
michael@0: 	switch (combine & COMBINE_B)
michael@0: 	{
michael@0: 	default:
michael@0: 	    Fb = 0;
michael@0: 	    break;
michael@0: 
michael@0: 	case COMBINE_B_OUT:
michael@0: 	    Fb = combine_disjoint_out_part (da, sa);
michael@0: 	    break;
michael@0: 
michael@0: 	case COMBINE_B_IN:
michael@0: 	    Fb = combine_disjoint_in_part (da, sa);
michael@0: 	    break;
michael@0: 
michael@0: 	case COMBINE_B:
michael@0: 	    Fb = MASK;
michael@0: 	    break;
michael@0: 	}
michael@0: 	m = GENERIC (s, d, 0, Fa, Fb, t, u, v);
michael@0: 	n = GENERIC (s, d, G_SHIFT, Fa, Fb, t, u, v);
michael@0: 	o = GENERIC (s, d, R_SHIFT, Fa, Fb, t, u, v);
michael@0: 	p = GENERIC (s, d, A_SHIFT, Fa, Fb, t, u, v);
michael@0: 	s = m | n | o | p;
michael@0: 	*(dest + i) = s;
michael@0:     }
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_disjoint_over_u (pixman_implementation_t *imp,
michael@0:                          pixman_op_t              op,
michael@0:                          uint64_t *                dest,
michael@0:                          const uint64_t *          src,
michael@0:                          const uint64_t *          mask,
michael@0:                          int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t s = combine_mask (src, mask, i);
michael@0: 	uint32_t a = s >> A_SHIFT;
michael@0: 
michael@0: 	if (s != 0x00)
michael@0: 	{
michael@0: 	    uint64_t d = *(dest + i);
michael@0: 	    a = combine_disjoint_out_part (d >> A_SHIFT, a);
michael@0: 	    UN16x4_MUL_UN16_ADD_UN16x4 (d, a, s);
michael@0: 
michael@0: 	    *(dest + i) = d;
michael@0: 	}
michael@0:     }
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_disjoint_in_u (pixman_implementation_t *imp,
michael@0:                        pixman_op_t              op,
michael@0:                        uint64_t *                dest,
michael@0:                        const uint64_t *          src,
michael@0:                        const uint64_t *          mask,
michael@0:                        int                      width)
michael@0: {
michael@0:     combine_disjoint_general_u (dest, src, mask, width, COMBINE_A_IN);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_disjoint_in_reverse_u (pixman_implementation_t *imp,
michael@0:                                pixman_op_t              op,
michael@0:                                uint64_t *                dest,
michael@0:                                const uint64_t *          src,
michael@0:                                const uint64_t *          mask,
michael@0:                                int                      width)
michael@0: {
michael@0:     combine_disjoint_general_u (dest, src, mask, width, COMBINE_B_IN);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_disjoint_out_u (pixman_implementation_t *imp,
michael@0:                         pixman_op_t              op,
michael@0:                         uint64_t *                dest,
michael@0:                         const uint64_t *          src,
michael@0:                         const uint64_t *          mask,
michael@0:                         int                      width)
michael@0: {
michael@0:     combine_disjoint_general_u (dest, src, mask, width, COMBINE_A_OUT);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_disjoint_out_reverse_u (pixman_implementation_t *imp,
michael@0:                                 pixman_op_t              op,
michael@0:                                 uint64_t *                dest,
michael@0:                                 const uint64_t *          src,
michael@0:                                 const uint64_t *          mask,
michael@0:                                 int                      width)
michael@0: {
michael@0:     combine_disjoint_general_u (dest, src, mask, width, COMBINE_B_OUT);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_disjoint_atop_u (pixman_implementation_t *imp,
michael@0:                          pixman_op_t              op,
michael@0:                          uint64_t *                dest,
michael@0:                          const uint64_t *          src,
michael@0:                          const uint64_t *          mask,
michael@0:                          int                      width)
michael@0: {
michael@0:     combine_disjoint_general_u (dest, src, mask, width, COMBINE_A_ATOP);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_disjoint_atop_reverse_u (pixman_implementation_t *imp,
michael@0:                                  pixman_op_t              op,
michael@0:                                  uint64_t *                dest,
michael@0:                                  const uint64_t *          src,
michael@0:                                  const uint64_t *          mask,
michael@0:                                  int                      width)
michael@0: {
michael@0:     combine_disjoint_general_u (dest, src, mask, width, COMBINE_B_ATOP);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_disjoint_xor_u (pixman_implementation_t *imp,
michael@0:                         pixman_op_t              op,
michael@0:                         uint64_t *                dest,
michael@0:                         const uint64_t *          src,
michael@0:                         const uint64_t *          mask,
michael@0:                         int                      width)
michael@0: {
michael@0:     combine_disjoint_general_u (dest, src, mask, width, COMBINE_XOR);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_conjoint_general_u (uint64_t *      dest,
michael@0:                             const uint64_t *src,
michael@0:                             const uint64_t *mask,
michael@0:                             int            width,
michael@0:                             uint16_t        combine)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t s = combine_mask (src, mask, i);
michael@0: 	uint64_t d = *(dest + i);
michael@0: 	uint64_t m, n, o, p;
michael@0: 	uint32_t Fa, Fb, t, u, v;
michael@0: 	uint16_t sa = s >> A_SHIFT;
michael@0: 	uint16_t da = d >> A_SHIFT;
michael@0: 
michael@0: 	switch (combine & COMBINE_A)
michael@0: 	{
michael@0: 	default:
michael@0: 	    Fa = 0;
michael@0: 	    break;
michael@0: 
michael@0: 	case COMBINE_A_OUT:
michael@0: 	    Fa = combine_conjoint_out_part (sa, da);
michael@0: 	    break;
michael@0: 
michael@0: 	case COMBINE_A_IN:
michael@0: 	    Fa = combine_conjoint_in_part (sa, da);
michael@0: 	    break;
michael@0: 
michael@0: 	case COMBINE_A:
michael@0: 	    Fa = MASK;
michael@0: 	    break;
michael@0: 	}
michael@0: 
michael@0: 	switch (combine & COMBINE_B)
michael@0: 	{
michael@0: 	default:
michael@0: 	    Fb = 0;
michael@0: 	    break;
michael@0: 
michael@0: 	case COMBINE_B_OUT:
michael@0: 	    Fb = combine_conjoint_out_part (da, sa);
michael@0: 	    break;
michael@0: 
michael@0: 	case COMBINE_B_IN:
michael@0: 	    Fb = combine_conjoint_in_part (da, sa);
michael@0: 	    break;
michael@0: 
michael@0: 	case COMBINE_B:
michael@0: 	    Fb = MASK;
michael@0: 	    break;
michael@0: 	}
michael@0: 
michael@0: 	m = GENERIC (s, d, 0, Fa, Fb, t, u, v);
michael@0: 	n = GENERIC (s, d, G_SHIFT, Fa, Fb, t, u, v);
michael@0: 	o = GENERIC (s, d, R_SHIFT, Fa, Fb, t, u, v);
michael@0: 	p = GENERIC (s, d, A_SHIFT, Fa, Fb, t, u, v);
michael@0: 
michael@0: 	s = m | n | o | p;
michael@0: 
michael@0: 	*(dest + i) = s;
michael@0:     }
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_conjoint_over_u (pixman_implementation_t *imp,
michael@0:                          pixman_op_t              op,
michael@0:                          uint64_t *                dest,
michael@0:                          const uint64_t *          src,
michael@0:                          const uint64_t *          mask,
michael@0:                          int                      width)
michael@0: {
michael@0:     combine_conjoint_general_u (dest, src, mask, width, COMBINE_A_OVER);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_conjoint_over_reverse_u (pixman_implementation_t *imp,
michael@0:                                  pixman_op_t              op,
michael@0:                                  uint64_t *                dest,
michael@0:                                  const uint64_t *          src,
michael@0:                                  const uint64_t *          mask,
michael@0:                                  int                      width)
michael@0: {
michael@0:     combine_conjoint_general_u (dest, src, mask, width, COMBINE_B_OVER);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_conjoint_in_u (pixman_implementation_t *imp,
michael@0:                        pixman_op_t              op,
michael@0:                        uint64_t *                dest,
michael@0:                        const uint64_t *          src,
michael@0:                        const uint64_t *          mask,
michael@0:                        int                      width)
michael@0: {
michael@0:     combine_conjoint_general_u (dest, src, mask, width, COMBINE_A_IN);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_conjoint_in_reverse_u (pixman_implementation_t *imp,
michael@0:                                pixman_op_t              op,
michael@0:                                uint64_t *                dest,
michael@0:                                const uint64_t *          src,
michael@0:                                const uint64_t *          mask,
michael@0:                                int                      width)
michael@0: {
michael@0:     combine_conjoint_general_u (dest, src, mask, width, COMBINE_B_IN);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_conjoint_out_u (pixman_implementation_t *imp,
michael@0:                         pixman_op_t              op,
michael@0:                         uint64_t *                dest,
michael@0:                         const uint64_t *          src,
michael@0:                         const uint64_t *          mask,
michael@0:                         int                      width)
michael@0: {
michael@0:     combine_conjoint_general_u (dest, src, mask, width, COMBINE_A_OUT);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_conjoint_out_reverse_u (pixman_implementation_t *imp,
michael@0:                                 pixman_op_t              op,
michael@0:                                 uint64_t *                dest,
michael@0:                                 const uint64_t *          src,
michael@0:                                 const uint64_t *          mask,
michael@0:                                 int                      width)
michael@0: {
michael@0:     combine_conjoint_general_u (dest, src, mask, width, COMBINE_B_OUT);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_conjoint_atop_u (pixman_implementation_t *imp,
michael@0:                          pixman_op_t              op,
michael@0:                          uint64_t *                dest,
michael@0:                          const uint64_t *          src,
michael@0:                          const uint64_t *          mask,
michael@0:                          int                      width)
michael@0: {
michael@0:     combine_conjoint_general_u (dest, src, mask, width, COMBINE_A_ATOP);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_conjoint_atop_reverse_u (pixman_implementation_t *imp,
michael@0:                                  pixman_op_t              op,
michael@0:                                  uint64_t *                dest,
michael@0:                                  const uint64_t *          src,
michael@0:                                  const uint64_t *          mask,
michael@0:                                  int                      width)
michael@0: {
michael@0:     combine_conjoint_general_u (dest, src, mask, width, COMBINE_B_ATOP);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_conjoint_xor_u (pixman_implementation_t *imp,
michael@0:                         pixman_op_t              op,
michael@0:                         uint64_t *                dest,
michael@0:                         const uint64_t *          src,
michael@0:                         const uint64_t *          mask,
michael@0:                         int                      width)
michael@0: {
michael@0:     combine_conjoint_general_u (dest, src, mask, width, COMBINE_XOR);
michael@0: }
michael@0: 
michael@0: /************************************************************************/
michael@0: /*********************** Per Channel functions **************************/
michael@0: /************************************************************************/
michael@0: 
michael@0: static void
michael@0: combine_clear_ca (pixman_implementation_t *imp,
michael@0:                   pixman_op_t              op,
michael@0:                   uint64_t *                dest,
michael@0:                   const uint64_t *          src,
michael@0:                   const uint64_t *          mask,
michael@0:                   int                      width)
michael@0: {
michael@0:     memset (dest, 0, width * sizeof(uint64_t));
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_src_ca (pixman_implementation_t *imp,
michael@0:                 pixman_op_t              op,
michael@0:                 uint64_t *                dest,
michael@0:                 const uint64_t *          src,
michael@0:                 const uint64_t *          mask,
michael@0:                 int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t s = *(src + i);
michael@0: 	uint64_t m = *(mask + i);
michael@0: 
michael@0: 	combine_mask_value_ca (&s, &m);
michael@0: 
michael@0: 	*(dest + i) = s;
michael@0:     }
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_over_ca (pixman_implementation_t *imp,
michael@0:                  pixman_op_t              op,
michael@0:                  uint64_t *                dest,
michael@0:                  const uint64_t *          src,
michael@0:                  const uint64_t *          mask,
michael@0:                  int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t s = *(src + i);
michael@0: 	uint64_t m = *(mask + i);
michael@0: 	uint64_t a;
michael@0: 
michael@0: 	combine_mask_ca (&s, &m);
michael@0: 
michael@0: 	a = ~m;
michael@0: 	if (a)
michael@0: 	{
michael@0: 	    uint64_t d = *(dest + i);
michael@0: 	    UN16x4_MUL_UN16x4_ADD_UN16x4 (d, a, s);
michael@0: 	    s = d;
michael@0: 	}
michael@0: 
michael@0: 	*(dest + i) = s;
michael@0:     }
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_over_reverse_ca (pixman_implementation_t *imp,
michael@0:                          pixman_op_t              op,
michael@0:                          uint64_t *                dest,
michael@0:                          const uint64_t *          src,
michael@0:                          const uint64_t *          mask,
michael@0:                          int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t d = *(dest + i);
michael@0: 	uint64_t a = ~d >> A_SHIFT;
michael@0: 
michael@0: 	if (a)
michael@0: 	{
michael@0: 	    uint64_t s = *(src + i);
michael@0: 	    uint64_t m = *(mask + i);
michael@0: 
michael@0: 	    UN16x4_MUL_UN16x4 (s, m);
michael@0: 	    UN16x4_MUL_UN16_ADD_UN16x4 (s, a, d);
michael@0: 
michael@0: 	    *(dest + i) = s;
michael@0: 	}
michael@0:     }
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_in_ca (pixman_implementation_t *imp,
michael@0:                pixman_op_t              op,
michael@0:                uint64_t *                dest,
michael@0:                const uint64_t *          src,
michael@0:                const uint64_t *          mask,
michael@0:                int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t d = *(dest + i);
michael@0: 	uint32_t a = d >> A_SHIFT;
michael@0: 	uint64_t s = 0;
michael@0: 
michael@0: 	if (a)
michael@0: 	{
michael@0: 	    uint64_t m = *(mask + i);
michael@0: 
michael@0: 	    s = *(src + i);
michael@0: 	    combine_mask_value_ca (&s, &m);
michael@0: 
michael@0: 	    if (a != MASK)
michael@0: 		UN16x4_MUL_UN16 (s, a);
michael@0: 	}
michael@0: 
michael@0: 	*(dest + i) = s;
michael@0:     }
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_in_reverse_ca (pixman_implementation_t *imp,
michael@0:                        pixman_op_t              op,
michael@0:                        uint64_t *                dest,
michael@0:                        const uint64_t *          src,
michael@0:                        const uint64_t *          mask,
michael@0:                        int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t s = *(src + i);
michael@0: 	uint64_t m = *(mask + i);
michael@0: 	uint64_t a;
michael@0: 
michael@0: 	combine_mask_alpha_ca (&s, &m);
michael@0: 
michael@0: 	a = m;
michael@0: 	if (a != ~0)
michael@0: 	{
michael@0: 	    uint64_t d = 0;
michael@0: 
michael@0: 	    if (a)
michael@0: 	    {
michael@0: 		d = *(dest + i);
michael@0: 		UN16x4_MUL_UN16x4 (d, a);
michael@0: 	    }
michael@0: 
michael@0: 	    *(dest + i) = d;
michael@0: 	}
michael@0:     }
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_out_ca (pixman_implementation_t *imp,
michael@0:                 pixman_op_t              op,
michael@0:                 uint64_t *                dest,
michael@0:                 const uint64_t *          src,
michael@0:                 const uint64_t *          mask,
michael@0:                 int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t d = *(dest + i);
michael@0: 	uint32_t a = ~d >> A_SHIFT;
michael@0: 	uint64_t s = 0;
michael@0: 
michael@0: 	if (a)
michael@0: 	{
michael@0: 	    uint64_t m = *(mask + i);
michael@0: 
michael@0: 	    s = *(src + i);
michael@0: 	    combine_mask_value_ca (&s, &m);
michael@0: 
michael@0: 	    if (a != MASK)
michael@0: 		UN16x4_MUL_UN16 (s, a);
michael@0: 	}
michael@0: 
michael@0: 	*(dest + i) = s;
michael@0:     }
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_out_reverse_ca (pixman_implementation_t *imp,
michael@0:                         pixman_op_t              op,
michael@0:                         uint64_t *                dest,
michael@0:                         const uint64_t *          src,
michael@0:                         const uint64_t *          mask,
michael@0:                         int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t s = *(src + i);
michael@0: 	uint64_t m = *(mask + i);
michael@0: 	uint64_t a;
michael@0: 
michael@0: 	combine_mask_alpha_ca (&s, &m);
michael@0: 
michael@0: 	a = ~m;
michael@0: 	if (a != ~0)
michael@0: 	{
michael@0: 	    uint64_t d = 0;
michael@0: 
michael@0: 	    if (a)
michael@0: 	    {
michael@0: 		d = *(dest + i);
michael@0: 		UN16x4_MUL_UN16x4 (d, a);
michael@0: 	    }
michael@0: 
michael@0: 	    *(dest + i) = d;
michael@0: 	}
michael@0:     }
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_atop_ca (pixman_implementation_t *imp,
michael@0:                  pixman_op_t              op,
michael@0:                  uint64_t *                dest,
michael@0:                  const uint64_t *          src,
michael@0:                  const uint64_t *          mask,
michael@0:                  int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t d = *(dest + i);
michael@0: 	uint64_t s = *(src + i);
michael@0: 	uint64_t m = *(mask + i);
michael@0: 	uint64_t ad;
michael@0: 	uint32_t as = d >> A_SHIFT;
michael@0: 
michael@0: 	combine_mask_ca (&s, &m);
michael@0: 
michael@0: 	ad = ~m;
michael@0: 
michael@0: 	UN16x4_MUL_UN16x4_ADD_UN16x4_MUL_UN16 (d, ad, s, as);
michael@0: 
michael@0: 	*(dest + i) = d;
michael@0:     }
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_atop_reverse_ca (pixman_implementation_t *imp,
michael@0:                          pixman_op_t              op,
michael@0:                          uint64_t *                dest,
michael@0:                          const uint64_t *          src,
michael@0:                          const uint64_t *          mask,
michael@0:                          int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t d = *(dest + i);
michael@0: 	uint64_t s = *(src + i);
michael@0: 	uint64_t m = *(mask + i);
michael@0: 	uint64_t ad;
michael@0: 	uint32_t as = ~d >> A_SHIFT;
michael@0: 
michael@0: 	combine_mask_ca (&s, &m);
michael@0: 
michael@0: 	ad = m;
michael@0: 
michael@0: 	UN16x4_MUL_UN16x4_ADD_UN16x4_MUL_UN16 (d, ad, s, as);
michael@0: 
michael@0: 	*(dest + i) = d;
michael@0:     }
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_xor_ca (pixman_implementation_t *imp,
michael@0:                 pixman_op_t              op,
michael@0:                 uint64_t *                dest,
michael@0:                 const uint64_t *          src,
michael@0:                 const uint64_t *          mask,
michael@0:                 int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t d = *(dest + i);
michael@0: 	uint64_t s = *(src + i);
michael@0: 	uint64_t m = *(mask + i);
michael@0: 	uint64_t ad;
michael@0: 	uint32_t as = ~d >> A_SHIFT;
michael@0: 
michael@0: 	combine_mask_ca (&s, &m);
michael@0: 
michael@0: 	ad = ~m;
michael@0: 
michael@0: 	UN16x4_MUL_UN16x4_ADD_UN16x4_MUL_UN16 (d, ad, s, as);
michael@0: 
michael@0: 	*(dest + i) = d;
michael@0:     }
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_add_ca (pixman_implementation_t *imp,
michael@0:                 pixman_op_t              op,
michael@0:                 uint64_t *                dest,
michael@0:                 const uint64_t *          src,
michael@0:                 const uint64_t *          mask,
michael@0:                 int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t s = *(src + i);
michael@0: 	uint64_t m = *(mask + i);
michael@0: 	uint64_t d = *(dest + i);
michael@0: 
michael@0: 	combine_mask_value_ca (&s, &m);
michael@0: 
michael@0: 	UN16x4_ADD_UN16x4 (d, s);
michael@0: 
michael@0: 	*(dest + i) = d;
michael@0:     }
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_saturate_ca (pixman_implementation_t *imp,
michael@0:                      pixman_op_t              op,
michael@0:                      uint64_t *                dest,
michael@0:                      const uint64_t *          src,
michael@0:                      const uint64_t *          mask,
michael@0:                      int                      width)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t s, d;
michael@0: 	uint32_t sa, sr, sg, sb, da;
michael@0: 	uint32_t t, u, v;
michael@0: 	uint64_t m, n, o, p;
michael@0: 
michael@0: 	d = *(dest + i);
michael@0: 	s = *(src + i);
michael@0: 	m = *(mask + i);
michael@0: 
michael@0: 	combine_mask_ca (&s, &m);
michael@0: 
michael@0: 	sa = (m >> A_SHIFT);
michael@0: 	sr = (m >> R_SHIFT) & MASK;
michael@0: 	sg = (m >> G_SHIFT) & MASK;
michael@0: 	sb =  m             & MASK;
michael@0: 	da = ~d >> A_SHIFT;
michael@0: 
michael@0: 	if (sb <= da)
michael@0: 	    m = ADD (s, d, 0, t);
michael@0: 	else
michael@0: 	    m = GENERIC (s, d, 0, (da << G_SHIFT) / sb, MASK, t, u, v);
michael@0: 
michael@0: 	if (sg <= da)
michael@0: 	    n = ADD (s, d, G_SHIFT, t);
michael@0: 	else
michael@0: 	    n = GENERIC (s, d, G_SHIFT, (da << G_SHIFT) / sg, MASK, t, u, v);
michael@0: 
michael@0: 	if (sr <= da)
michael@0: 	    o = ADD (s, d, R_SHIFT, t);
michael@0: 	else
michael@0: 	    o = GENERIC (s, d, R_SHIFT, (da << G_SHIFT) / sr, MASK, t, u, v);
michael@0: 
michael@0: 	if (sa <= da)
michael@0: 	    p = ADD (s, d, A_SHIFT, t);
michael@0: 	else
michael@0: 	    p = GENERIC (s, d, A_SHIFT, (da << G_SHIFT) / sa, MASK, t, u, v);
michael@0: 
michael@0: 	*(dest + i) = m | n | o | p;
michael@0:     }
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_disjoint_general_ca (uint64_t *      dest,
michael@0:                              const uint64_t *src,
michael@0:                              const uint64_t *mask,
michael@0:                              int            width,
michael@0:                              uint16_t        combine)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t s, d;
michael@0: 	uint64_t m, n, o, p;
michael@0: 	uint64_t Fa, Fb;
michael@0: 	uint32_t t, u, v;
michael@0: 	uint64_t sa;
michael@0: 	uint16_t da;
michael@0: 
michael@0: 	s = *(src + i);
michael@0: 	m = *(mask + i);
michael@0: 	d = *(dest + i);
michael@0: 	da = d >> A_SHIFT;
michael@0: 
michael@0: 	combine_mask_ca (&s, &m);
michael@0: 
michael@0: 	sa = m;
michael@0: 
michael@0: 	switch (combine & COMBINE_A)
michael@0: 	{
michael@0: 	default:
michael@0: 	    Fa = 0;
michael@0: 	    break;
michael@0: 
michael@0: 	case COMBINE_A_OUT:
michael@0: 	    m = (uint64_t)combine_disjoint_out_part ((uint16_t) (sa >> 0), da);
michael@0: 	    n = (uint64_t)combine_disjoint_out_part ((uint16_t) (sa >> G_SHIFT), da) << G_SHIFT;
michael@0: 	    o = (uint64_t)combine_disjoint_out_part ((uint16_t) (sa >> R_SHIFT), da) << R_SHIFT;
michael@0: 	    p = (uint64_t)combine_disjoint_out_part ((uint16_t) (sa >> A_SHIFT), da) << A_SHIFT;
michael@0: 	    Fa = m | n | o | p;
michael@0: 	    break;
michael@0: 
michael@0: 	case COMBINE_A_IN:
michael@0: 	    m = (uint64_t)combine_disjoint_in_part ((uint16_t) (sa >> 0), da);
michael@0: 	    n = (uint64_t)combine_disjoint_in_part ((uint16_t) (sa >> G_SHIFT), da) << G_SHIFT;
michael@0: 	    o = (uint64_t)combine_disjoint_in_part ((uint16_t) (sa >> R_SHIFT), da) << R_SHIFT;
michael@0: 	    p = (uint64_t)combine_disjoint_in_part ((uint16_t) (sa >> A_SHIFT), da) << A_SHIFT;
michael@0: 	    Fa = m | n | o | p;
michael@0: 	    break;
michael@0: 
michael@0: 	case COMBINE_A:
michael@0: 	    Fa = ~0;
michael@0: 	    break;
michael@0: 	}
michael@0: 
michael@0: 	switch (combine & COMBINE_B)
michael@0: 	{
michael@0: 	default:
michael@0: 	    Fb = 0;
michael@0: 	    break;
michael@0: 
michael@0: 	case COMBINE_B_OUT:
michael@0: 	    m = (uint64_t)combine_disjoint_out_part (da, (uint16_t) (sa >> 0));
michael@0: 	    n = (uint64_t)combine_disjoint_out_part (da, (uint16_t) (sa >> G_SHIFT)) << G_SHIFT;
michael@0: 	    o = (uint64_t)combine_disjoint_out_part (da, (uint16_t) (sa >> R_SHIFT)) << R_SHIFT;
michael@0: 	    p = (uint64_t)combine_disjoint_out_part (da, (uint16_t) (sa >> A_SHIFT)) << A_SHIFT;
michael@0: 	    Fb = m | n | o | p;
michael@0: 	    break;
michael@0: 
michael@0: 	case COMBINE_B_IN:
michael@0: 	    m = (uint64_t)combine_disjoint_in_part (da, (uint16_t) (sa >> 0));
michael@0: 	    n = (uint64_t)combine_disjoint_in_part (da, (uint16_t) (sa >> G_SHIFT)) << G_SHIFT;
michael@0: 	    o = (uint64_t)combine_disjoint_in_part (da, (uint16_t) (sa >> R_SHIFT)) << R_SHIFT;
michael@0: 	    p = (uint64_t)combine_disjoint_in_part (da, (uint16_t) (sa >> A_SHIFT)) << A_SHIFT;
michael@0: 	    Fb = m | n | o | p;
michael@0: 	    break;
michael@0: 
michael@0: 	case COMBINE_B:
michael@0: 	    Fb = ~0;
michael@0: 	    break;
michael@0: 	}
michael@0: 	m = GENERIC (s, d, 0, GET_COMP (Fa, 0), GET_COMP (Fb, 0), t, u, v);
michael@0: 	n = GENERIC (s, d, G_SHIFT, GET_COMP (Fa, G_SHIFT), GET_COMP (Fb, G_SHIFT), t, u, v);
michael@0: 	o = GENERIC (s, d, R_SHIFT, GET_COMP (Fa, R_SHIFT), GET_COMP (Fb, R_SHIFT), t, u, v);
michael@0: 	p = GENERIC (s, d, A_SHIFT, GET_COMP (Fa, A_SHIFT), GET_COMP (Fb, A_SHIFT), t, u, v);
michael@0: 
michael@0: 	s = m | n | o | p;
michael@0: 
michael@0: 	*(dest + i) = s;
michael@0:     }
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_disjoint_over_ca (pixman_implementation_t *imp,
michael@0:                           pixman_op_t              op,
michael@0:                           uint64_t *                dest,
michael@0:                           const uint64_t *          src,
michael@0:                           const uint64_t *          mask,
michael@0:                           int                      width)
michael@0: {
michael@0:     combine_disjoint_general_ca (dest, src, mask, width, COMBINE_A_OVER);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_disjoint_in_ca (pixman_implementation_t *imp,
michael@0:                         pixman_op_t              op,
michael@0:                         uint64_t *                dest,
michael@0:                         const uint64_t *          src,
michael@0:                         const uint64_t *          mask,
michael@0:                         int                      width)
michael@0: {
michael@0:     combine_disjoint_general_ca (dest, src, mask, width, COMBINE_A_IN);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_disjoint_in_reverse_ca (pixman_implementation_t *imp,
michael@0:                                 pixman_op_t              op,
michael@0:                                 uint64_t *                dest,
michael@0:                                 const uint64_t *          src,
michael@0:                                 const uint64_t *          mask,
michael@0:                                 int                      width)
michael@0: {
michael@0:     combine_disjoint_general_ca (dest, src, mask, width, COMBINE_B_IN);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_disjoint_out_ca (pixman_implementation_t *imp,
michael@0:                          pixman_op_t              op,
michael@0:                          uint64_t *                dest,
michael@0:                          const uint64_t *          src,
michael@0:                          const uint64_t *          mask,
michael@0:                          int                      width)
michael@0: {
michael@0:     combine_disjoint_general_ca (dest, src, mask, width, COMBINE_A_OUT);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_disjoint_out_reverse_ca (pixman_implementation_t *imp,
michael@0:                                  pixman_op_t              op,
michael@0:                                  uint64_t *                dest,
michael@0:                                  const uint64_t *          src,
michael@0:                                  const uint64_t *          mask,
michael@0:                                  int                      width)
michael@0: {
michael@0:     combine_disjoint_general_ca (dest, src, mask, width, COMBINE_B_OUT);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_disjoint_atop_ca (pixman_implementation_t *imp,
michael@0:                           pixman_op_t              op,
michael@0:                           uint64_t *                dest,
michael@0:                           const uint64_t *          src,
michael@0:                           const uint64_t *          mask,
michael@0:                           int                      width)
michael@0: {
michael@0:     combine_disjoint_general_ca (dest, src, mask, width, COMBINE_A_ATOP);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_disjoint_atop_reverse_ca (pixman_implementation_t *imp,
michael@0:                                   pixman_op_t              op,
michael@0:                                   uint64_t *                dest,
michael@0:                                   const uint64_t *          src,
michael@0:                                   const uint64_t *          mask,
michael@0:                                   int                      width)
michael@0: {
michael@0:     combine_disjoint_general_ca (dest, src, mask, width, COMBINE_B_ATOP);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_disjoint_xor_ca (pixman_implementation_t *imp,
michael@0:                          pixman_op_t              op,
michael@0:                          uint64_t *                dest,
michael@0:                          const uint64_t *          src,
michael@0:                          const uint64_t *          mask,
michael@0:                          int                      width)
michael@0: {
michael@0:     combine_disjoint_general_ca (dest, src, mask, width, COMBINE_XOR);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_conjoint_general_ca (uint64_t *      dest,
michael@0:                              const uint64_t *src,
michael@0:                              const uint64_t *mask,
michael@0:                              int            width,
michael@0:                              uint16_t        combine)
michael@0: {
michael@0:     int i;
michael@0: 
michael@0:     for (i = 0; i < width; ++i)
michael@0:     {
michael@0: 	uint64_t s, d;
michael@0: 	uint64_t m, n, o, p;
michael@0: 	uint64_t Fa, Fb;
michael@0: 	uint32_t t, u, v;
michael@0: 	uint64_t sa;
michael@0: 	uint16_t da;
michael@0: 
michael@0: 	s = *(src + i);
michael@0: 	m = *(mask + i);
michael@0: 	d = *(dest + i);
michael@0: 	da = d >> A_SHIFT;
michael@0: 
michael@0: 	combine_mask_ca (&s, &m);
michael@0: 
michael@0: 	sa = m;
michael@0: 
michael@0: 	switch (combine & COMBINE_A)
michael@0: 	{
michael@0: 	default:
michael@0: 	    Fa = 0;
michael@0: 	    break;
michael@0: 
michael@0: 	case COMBINE_A_OUT:
michael@0: 	    m = (uint64_t)combine_conjoint_out_part ((uint16_t) (sa >> 0), da);
michael@0: 	    n = (uint64_t)combine_conjoint_out_part ((uint16_t) (sa >> G_SHIFT), da) << G_SHIFT;
michael@0: 	    o = (uint64_t)combine_conjoint_out_part ((uint16_t) (sa >> R_SHIFT), da) << R_SHIFT;
michael@0: 	    p = (uint64_t)combine_conjoint_out_part ((uint16_t) (sa >> A_SHIFT), da) << A_SHIFT;
michael@0: 	    Fa = m | n | o | p;
michael@0: 	    break;
michael@0: 
michael@0: 	case COMBINE_A_IN:
michael@0: 	    m = (uint64_t)combine_conjoint_in_part ((uint16_t) (sa >> 0), da);
michael@0: 	    n = (uint64_t)combine_conjoint_in_part ((uint16_t) (sa >> G_SHIFT), da) << G_SHIFT;
michael@0: 	    o = (uint64_t)combine_conjoint_in_part ((uint16_t) (sa >> R_SHIFT), da) << R_SHIFT;
michael@0: 	    p = (uint64_t)combine_conjoint_in_part ((uint16_t) (sa >> A_SHIFT), da) << A_SHIFT;
michael@0: 	    Fa = m | n | o | p;
michael@0: 	    break;
michael@0: 
michael@0: 	case COMBINE_A:
michael@0: 	    Fa = ~0;
michael@0: 	    break;
michael@0: 	}
michael@0: 
michael@0: 	switch (combine & COMBINE_B)
michael@0: 	{
michael@0: 	default:
michael@0: 	    Fb = 0;
michael@0: 	    break;
michael@0: 
michael@0: 	case COMBINE_B_OUT:
michael@0: 	    m = (uint64_t)combine_conjoint_out_part (da, (uint16_t) (sa >> 0));
michael@0: 	    n = (uint64_t)combine_conjoint_out_part (da, (uint16_t) (sa >> G_SHIFT)) << G_SHIFT;
michael@0: 	    o = (uint64_t)combine_conjoint_out_part (da, (uint16_t) (sa >> R_SHIFT)) << R_SHIFT;
michael@0: 	    p = (uint64_t)combine_conjoint_out_part (da, (uint16_t) (sa >> A_SHIFT)) << A_SHIFT;
michael@0: 	    Fb = m | n | o | p;
michael@0: 	    break;
michael@0: 
michael@0: 	case COMBINE_B_IN:
michael@0: 	    m = (uint64_t)combine_conjoint_in_part (da, (uint16_t) (sa >> 0));
michael@0: 	    n = (uint64_t)combine_conjoint_in_part (da, (uint16_t) (sa >> G_SHIFT)) << G_SHIFT;
michael@0: 	    o = (uint64_t)combine_conjoint_in_part (da, (uint16_t) (sa >> R_SHIFT)) << R_SHIFT;
michael@0: 	    p = (uint64_t)combine_conjoint_in_part (da, (uint16_t) (sa >> A_SHIFT)) << A_SHIFT;
michael@0: 	    Fb = m | n | o | p;
michael@0: 	    break;
michael@0: 
michael@0: 	case COMBINE_B:
michael@0: 	    Fb = ~0;
michael@0: 	    break;
michael@0: 	}
michael@0: 	m = GENERIC (s, d, 0, GET_COMP (Fa, 0), GET_COMP (Fb, 0), t, u, v);
michael@0: 	n = GENERIC (s, d, G_SHIFT, GET_COMP (Fa, G_SHIFT), GET_COMP (Fb, G_SHIFT), t, u, v);
michael@0: 	o = GENERIC (s, d, R_SHIFT, GET_COMP (Fa, R_SHIFT), GET_COMP (Fb, R_SHIFT), t, u, v);
michael@0: 	p = GENERIC (s, d, A_SHIFT, GET_COMP (Fa, A_SHIFT), GET_COMP (Fb, A_SHIFT), t, u, v);
michael@0: 
michael@0: 	s = m | n | o | p;
michael@0: 
michael@0: 	*(dest + i) = s;
michael@0:     }
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_conjoint_over_ca (pixman_implementation_t *imp,
michael@0:                           pixman_op_t              op,
michael@0:                           uint64_t *                dest,
michael@0:                           const uint64_t *          src,
michael@0:                           const uint64_t *          mask,
michael@0:                           int                      width)
michael@0: {
michael@0:     combine_conjoint_general_ca (dest, src, mask, width, COMBINE_A_OVER);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_conjoint_over_reverse_ca (pixman_implementation_t *imp,
michael@0:                                   pixman_op_t              op,
michael@0:                                   uint64_t *                dest,
michael@0:                                   const uint64_t *          src,
michael@0:                                   const uint64_t *          mask,
michael@0:                                   int                      width)
michael@0: {
michael@0:     combine_conjoint_general_ca (dest, src, mask, width, COMBINE_B_OVER);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_conjoint_in_ca (pixman_implementation_t *imp,
michael@0:                         pixman_op_t              op,
michael@0:                         uint64_t *                dest,
michael@0:                         const uint64_t *          src,
michael@0:                         const uint64_t *          mask,
michael@0:                         int                      width)
michael@0: {
michael@0:     combine_conjoint_general_ca (dest, src, mask, width, COMBINE_A_IN);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_conjoint_in_reverse_ca (pixman_implementation_t *imp,
michael@0:                                 pixman_op_t              op,
michael@0:                                 uint64_t *                dest,
michael@0:                                 const uint64_t *          src,
michael@0:                                 const uint64_t *          mask,
michael@0:                                 int                      width)
michael@0: {
michael@0:     combine_conjoint_general_ca (dest, src, mask, width, COMBINE_B_IN);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_conjoint_out_ca (pixman_implementation_t *imp,
michael@0:                          pixman_op_t              op,
michael@0:                          uint64_t *                dest,
michael@0:                          const uint64_t *          src,
michael@0:                          const uint64_t *          mask,
michael@0:                          int                      width)
michael@0: {
michael@0:     combine_conjoint_general_ca (dest, src, mask, width, COMBINE_A_OUT);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_conjoint_out_reverse_ca (pixman_implementation_t *imp,
michael@0:                                  pixman_op_t              op,
michael@0:                                  uint64_t *                dest,
michael@0:                                  const uint64_t *          src,
michael@0:                                  const uint64_t *          mask,
michael@0:                                  int                      width)
michael@0: {
michael@0:     combine_conjoint_general_ca (dest, src, mask, width, COMBINE_B_OUT);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_conjoint_atop_ca (pixman_implementation_t *imp,
michael@0:                           pixman_op_t              op,
michael@0:                           uint64_t *                dest,
michael@0:                           const uint64_t *          src,
michael@0:                           const uint64_t *          mask,
michael@0:                           int                      width)
michael@0: {
michael@0:     combine_conjoint_general_ca (dest, src, mask, width, COMBINE_A_ATOP);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_conjoint_atop_reverse_ca (pixman_implementation_t *imp,
michael@0:                                   pixman_op_t              op,
michael@0:                                   uint64_t *                dest,
michael@0:                                   const uint64_t *          src,
michael@0:                                   const uint64_t *          mask,
michael@0:                                   int                      width)
michael@0: {
michael@0:     combine_conjoint_general_ca (dest, src, mask, width, COMBINE_B_ATOP);
michael@0: }
michael@0: 
michael@0: static void
michael@0: combine_conjoint_xor_ca (pixman_implementation_t *imp,
michael@0:                          pixman_op_t              op,
michael@0:                          uint64_t *                dest,
michael@0:                          const uint64_t *          src,
michael@0:                          const uint64_t *          mask,
michael@0:                          int                      width)
michael@0: {
michael@0:     combine_conjoint_general_ca (dest, src, mask, width, COMBINE_XOR);
michael@0: }
michael@0: 
michael@0: void
michael@0: _pixman_setup_combiner_functions_64 (pixman_implementation_t *imp)
michael@0: {
michael@0:     /* Unified alpha */
michael@0:     imp->combine_64[PIXMAN_OP_CLEAR] = combine_clear;
michael@0:     imp->combine_64[PIXMAN_OP_SRC] = combine_src_u;
michael@0:     imp->combine_64[PIXMAN_OP_DST] = combine_dst;
michael@0:     imp->combine_64[PIXMAN_OP_OVER] = combine_over_u;
michael@0:     imp->combine_64[PIXMAN_OP_OVER_REVERSE] = combine_over_reverse_u;
michael@0:     imp->combine_64[PIXMAN_OP_IN] = combine_in_u;
michael@0:     imp->combine_64[PIXMAN_OP_IN_REVERSE] = combine_in_reverse_u;
michael@0:     imp->combine_64[PIXMAN_OP_OUT] = combine_out_u;
michael@0:     imp->combine_64[PIXMAN_OP_OUT_REVERSE] = combine_out_reverse_u;
michael@0:     imp->combine_64[PIXMAN_OP_ATOP] = combine_atop_u;
michael@0:     imp->combine_64[PIXMAN_OP_ATOP_REVERSE] = combine_atop_reverse_u;
michael@0:     imp->combine_64[PIXMAN_OP_XOR] = combine_xor_u;
michael@0:     imp->combine_64[PIXMAN_OP_ADD] = combine_add_u;
michael@0:     imp->combine_64[PIXMAN_OP_SATURATE] = combine_saturate_u;
michael@0: 
michael@0:     /* Disjoint, unified */
michael@0:     imp->combine_64[PIXMAN_OP_DISJOINT_CLEAR] = combine_clear;
michael@0:     imp->combine_64[PIXMAN_OP_DISJOINT_SRC] = combine_src_u;
michael@0:     imp->combine_64[PIXMAN_OP_DISJOINT_DST] = combine_dst;
michael@0:     imp->combine_64[PIXMAN_OP_DISJOINT_OVER] = combine_disjoint_over_u;
michael@0:     imp->combine_64[PIXMAN_OP_DISJOINT_OVER_REVERSE] = combine_saturate_u;
michael@0:     imp->combine_64[PIXMAN_OP_DISJOINT_IN] = combine_disjoint_in_u;
michael@0:     imp->combine_64[PIXMAN_OP_DISJOINT_IN_REVERSE] = combine_disjoint_in_reverse_u;
michael@0:     imp->combine_64[PIXMAN_OP_DISJOINT_OUT] = combine_disjoint_out_u;
michael@0:     imp->combine_64[PIXMAN_OP_DISJOINT_OUT_REVERSE] = combine_disjoint_out_reverse_u;
michael@0:     imp->combine_64[PIXMAN_OP_DISJOINT_ATOP] = combine_disjoint_atop_u;
michael@0:     imp->combine_64[PIXMAN_OP_DISJOINT_ATOP_REVERSE] = combine_disjoint_atop_reverse_u;
michael@0:     imp->combine_64[PIXMAN_OP_DISJOINT_XOR] = combine_disjoint_xor_u;
michael@0: 
michael@0:     /* Conjoint, unified */
michael@0:     imp->combine_64[PIXMAN_OP_CONJOINT_CLEAR] = combine_clear;
michael@0:     imp->combine_64[PIXMAN_OP_CONJOINT_SRC] = combine_src_u;
michael@0:     imp->combine_64[PIXMAN_OP_CONJOINT_DST] = combine_dst;
michael@0:     imp->combine_64[PIXMAN_OP_CONJOINT_OVER] = combine_conjoint_over_u;
michael@0:     imp->combine_64[PIXMAN_OP_CONJOINT_OVER_REVERSE] = combine_conjoint_over_reverse_u;
michael@0:     imp->combine_64[PIXMAN_OP_CONJOINT_IN] = combine_conjoint_in_u;
michael@0:     imp->combine_64[PIXMAN_OP_CONJOINT_IN_REVERSE] = combine_conjoint_in_reverse_u;
michael@0:     imp->combine_64[PIXMAN_OP_CONJOINT_OUT] = combine_conjoint_out_u;
michael@0:     imp->combine_64[PIXMAN_OP_CONJOINT_OUT_REVERSE] = combine_conjoint_out_reverse_u;
michael@0:     imp->combine_64[PIXMAN_OP_CONJOINT_ATOP] = combine_conjoint_atop_u;
michael@0:     imp->combine_64[PIXMAN_OP_CONJOINT_ATOP_REVERSE] = combine_conjoint_atop_reverse_u;
michael@0:     imp->combine_64[PIXMAN_OP_CONJOINT_XOR] = combine_conjoint_xor_u;
michael@0: 
michael@0:     imp->combine_64[PIXMAN_OP_MULTIPLY] = combine_multiply_u;
michael@0:     imp->combine_64[PIXMAN_OP_SCREEN] = combine_screen_u;
michael@0:     imp->combine_64[PIXMAN_OP_OVERLAY] = combine_overlay_u;
michael@0:     imp->combine_64[PIXMAN_OP_DARKEN] = combine_darken_u;
michael@0:     imp->combine_64[PIXMAN_OP_LIGHTEN] = combine_lighten_u;
michael@0:     imp->combine_64[PIXMAN_OP_COLOR_DODGE] = combine_color_dodge_u;
michael@0:     imp->combine_64[PIXMAN_OP_COLOR_BURN] = combine_color_burn_u;
michael@0:     imp->combine_64[PIXMAN_OP_HARD_LIGHT] = combine_hard_light_u;
michael@0:     imp->combine_64[PIXMAN_OP_SOFT_LIGHT] = combine_soft_light_u;
michael@0:     imp->combine_64[PIXMAN_OP_DIFFERENCE] = combine_difference_u;
michael@0:     imp->combine_64[PIXMAN_OP_EXCLUSION] = combine_exclusion_u;
michael@0:     imp->combine_64[PIXMAN_OP_HSL_HUE] = combine_hsl_hue_u;
michael@0:     imp->combine_64[PIXMAN_OP_HSL_SATURATION] = combine_hsl_saturation_u;
michael@0:     imp->combine_64[PIXMAN_OP_HSL_COLOR] = combine_hsl_color_u;
michael@0:     imp->combine_64[PIXMAN_OP_HSL_LUMINOSITY] = combine_hsl_luminosity_u;
michael@0: 
michael@0:     /* Component alpha combiners */
michael@0:     imp->combine_64_ca[PIXMAN_OP_CLEAR] = combine_clear_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_SRC] = combine_src_ca;
michael@0:     /* dest */
michael@0:     imp->combine_64_ca[PIXMAN_OP_OVER] = combine_over_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_OVER_REVERSE] = combine_over_reverse_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_IN] = combine_in_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_IN_REVERSE] = combine_in_reverse_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_OUT] = combine_out_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_OUT_REVERSE] = combine_out_reverse_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_ATOP] = combine_atop_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_ATOP_REVERSE] = combine_atop_reverse_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_XOR] = combine_xor_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_ADD] = combine_add_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_SATURATE] = combine_saturate_ca;
michael@0: 
michael@0:     /* Disjoint CA */
michael@0:     imp->combine_64_ca[PIXMAN_OP_DISJOINT_CLEAR] = combine_clear_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_DISJOINT_SRC] = combine_src_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_DISJOINT_DST] = combine_dst;
michael@0:     imp->combine_64_ca[PIXMAN_OP_DISJOINT_OVER] = combine_disjoint_over_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_DISJOINT_OVER_REVERSE] = combine_saturate_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_DISJOINT_IN] = combine_disjoint_in_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_DISJOINT_IN_REVERSE] = combine_disjoint_in_reverse_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_DISJOINT_OUT] = combine_disjoint_out_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_DISJOINT_OUT_REVERSE] = combine_disjoint_out_reverse_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_DISJOINT_ATOP] = combine_disjoint_atop_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_DISJOINT_ATOP_REVERSE] = combine_disjoint_atop_reverse_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_DISJOINT_XOR] = combine_disjoint_xor_ca;
michael@0: 
michael@0:     /* Conjoint CA */
michael@0:     imp->combine_64_ca[PIXMAN_OP_CONJOINT_CLEAR] = combine_clear_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_CONJOINT_SRC] = combine_src_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_CONJOINT_DST] = combine_dst;
michael@0:     imp->combine_64_ca[PIXMAN_OP_CONJOINT_OVER] = combine_conjoint_over_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_CONJOINT_OVER_REVERSE] = combine_conjoint_over_reverse_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_CONJOINT_IN] = combine_conjoint_in_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_CONJOINT_IN_REVERSE] = combine_conjoint_in_reverse_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_CONJOINT_OUT] = combine_conjoint_out_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_CONJOINT_OUT_REVERSE] = combine_conjoint_out_reverse_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_CONJOINT_ATOP] = combine_conjoint_atop_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_CONJOINT_ATOP_REVERSE] = combine_conjoint_atop_reverse_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_CONJOINT_XOR] = combine_conjoint_xor_ca;
michael@0: 
michael@0:     imp->combine_64_ca[PIXMAN_OP_MULTIPLY] = combine_multiply_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_SCREEN] = combine_screen_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_OVERLAY] = combine_overlay_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_DARKEN] = combine_darken_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_LIGHTEN] = combine_lighten_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_COLOR_DODGE] = combine_color_dodge_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_COLOR_BURN] = combine_color_burn_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_HARD_LIGHT] = combine_hard_light_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_SOFT_LIGHT] = combine_soft_light_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_DIFFERENCE] = combine_difference_ca;
michael@0:     imp->combine_64_ca[PIXMAN_OP_EXCLUSION] = combine_exclusion_ca;
michael@0: 
michael@0:     /* It is not clear that these make sense, so make them noops for now */
michael@0:     imp->combine_64_ca[PIXMAN_OP_HSL_HUE] = combine_dst;
michael@0:     imp->combine_64_ca[PIXMAN_OP_HSL_SATURATION] = combine_dst;
michael@0:     imp->combine_64_ca[PIXMAN_OP_HSL_COLOR] = combine_dst;
michael@0:     imp->combine_64_ca[PIXMAN_OP_HSL_LUMINOSITY] = combine_dst;
michael@0: }
michael@0: