Merge remote-tracking branch 'origin/released'

* origin/released:
  randrproto libXau libxcb mesa mkfontscale pixman git update 5 Jan 2014

3 files changed, 339 insertions, 1574 deletions

diff --git a/pixman/pixman/pixman-combine-float.c b/pixman/pixman/pixman-combine-float.c
index e9aab4833..f5145bc9d 100644
--- a/pixman/pixman/pixman-combine-float.c
+++ b/pixman/pixman/pixman-combine-float.c
@@ -319,23 +319,44 @@ MAKE_PD_COMBINERS (conjoint_xor,		ONE_MINUS_DA_OVER_SA,		ONE_MINUS_SA_OVER_DA)
  *
  * The following blend modes have been taken from the PDF ISO 32000
  * specification, which at this point in time is available from
- * http://www.adobe.com/devnet/acrobat/pdfs/PDF32000_2008.pdf
- * The relevant chapters are 11.3.5 and 11.3.6.
+ *
+ *     http://www.adobe.com/devnet/pdf/pdf_reference.html
+ *
+ * The specific documents of interest are the PDF spec itself:
+ *
+ *     http://wwwimages.adobe.com/www.adobe.com/content/dam/Adobe/en/devnet/pdf/pdfs/PDF32000_2008.pdf
+ *
+ * chapters 11.3.5 and 11.3.6 and a later supplement for Adobe Acrobat
+ * 9.1 and Reader 9.1:
+ *
+ *     http://wwwimages.adobe.com/www.adobe.com/content/dam/Adobe/en/devnet/pdf/pdfs/adobe_supplement_iso32000_1.pdf
+ *
+ * that clarifies the specifications for blend modes ColorDodge and
+ * ColorBurn.
+ *
  * The formula for computing the final pixel color given in 11.3.6 is:
- * αr × Cr = (1 – αs) × αb × Cb + (1 – αb) × αs × Cs + αb × αs × B(Cb, Cs)
- * with B() being the blend function.
- * Note that OVER is a special case of this operation, using B(Cb, Cs) = Cs
- *
- * These blend modes should match the SVG filter draft specification, as
- * it has been designed to mirror ISO 32000. Note that at the current point
- * no released draft exists that shows this, as the formulas have not been
- * updated yet after the release of ISO 32000.
- *
- * The default implementation here uses the PDF_SEPARABLE_BLEND_MODE and
- * PDF_NON_SEPARABLE_BLEND_MODE macros, which take the blend function as an
- * argument. Note that this implementation operates on premultiplied colors,
- * while the PDF specification does not. Therefore the code uses the formula
- * ar.Cra = (1 – as) . Dca + (1 – ad) . Sca + B(Dca, ad, Sca, as)
+ *
+ *     αr × Cr = (1 – αs) × αb × Cb + (1 – αb) × αs × Cs + αb × αs × B(Cb, Cs)
+ *
+ * with B() is the blend function. When B(Cb, Cs) = Cs, this formula
+ * reduces to the regular OVER operator.
+ *
+ * Cs and Cb are not premultiplied, so in our implementation we instead
+ * use:
+ *
+ *     cr = (1 – αs) × cb  +  (1 – αb) × cs  +  αb × αs × B (cb/αb, cs/αs)
+ *
+ * where cr, cs, and cb are premultiplied colors, and where the
+ *
+ *     αb × αs × B(cb/αb, cs/αs)
+ *
+ * part is first arithmetically simplified under the assumption that αb
+ * and αs are not 0, and then updated to produce a meaningful result when
+ * they are.
+ *
+ * For all the blend mode operators, the alpha channel is given by
+ *
+ *     αr = αs + αb + αb × αs
  */
 
 #define MAKE_SEPARABLE_PDF_COMBINERS(name)				\
@@ -355,18 +376,55 @@ MAKE_PD_COMBINERS (conjoint_xor,		ONE_MINUS_DA_OVER_SA,		ONE_MINUS_SA_OVER_DA)
     									\
     MAKE_COMBINERS (name, combine_ ## name ## _a, combine_ ## name ## _c)
 
+/*
+ * Multiply
+ *
+ *      ad * as * B(d / ad, s / as)
+ *    = ad * as * d/ad * s/as
+ *    = d * s
+ *
+ */
 static force_inline float
 blend_multiply (float sa, float s, float da, float d)
 {
     return d * s;
 }
 
+/*
+ * Screen
+ *
+ *      ad * as * B(d/ad, s/as)
+ *    = ad * as * (d/ad + s/as - s/as * d/ad)
+ *    = ad * s + as * d - s * d
+ */
 static force_inline float
 blend_screen (float sa, float s, float da, float d)
 {
     return d * sa + s * da - s * d;
 }
 
+/*
+ * Overlay
+ *
+ *     ad * as * B(d/ad, s/as)
+ *   = ad * as * Hardlight (s, d)
+ *   = if (d / ad < 0.5)
+ *         as * ad * Multiply (s/as, 2 * d/ad)
+ *     else
+ *         as * ad * Screen (s/as, 2 * d / ad - 1)
+ *   = if (d < 0.5 * ad)
+ *         as * ad * s/as * 2 * d /ad
+ *     else
+ *         as * ad * (s/as + 2 * d / ad - 1 - s / as * (2 * d / ad - 1))
+ *   = if (2 * d < ad)
+ *         2 * s * d
+ *     else
+ *         ad * s + 2 * as * d - as * ad - ad * s * (2 * d / ad - 1)
+ *   = if (2 * d < ad)
+ *         2 * s * d
+ *     else
+ *         as * ad - 2 * (ad - d) * (as - s)
+ */
 static force_inline float
 blend_overlay (float sa, float s, float da, float d)
 {
@@ -376,6 +434,13 @@ blend_overlay (float sa, float s, float da, float d)
 	return sa * da - 2 * (da - d) * (sa - s);
 }
 
+/*
+ * Darken
+ *
+ *     ad * as * B(d/ad, s/as)
+ *   = ad * as * MIN(d/ad, s/as)
+ *   = MIN (as * d, ad * s)
+ */
 static force_inline float
 blend_darken (float sa, float s, float da, float d)
 {
@@ -388,6 +453,13 @@ blend_darken (float sa, float s, float da, float d)
 	return s;
 }
 
+/*
+ * Lighten
+ *
+ *     ad * as * B(d/ad, s/as)
+ *   = ad * as * MAX(d/ad, s/as)
+ *   = MAX (as * d, ad * s)
+ */
 static force_inline float
 blend_lighten (float sa, float s, float da, float d)
 {
@@ -400,6 +472,24 @@ blend_lighten (float sa, float s, float da, float d)
 	return d;
 }
 
+/*
+ * Color dodge
+ *
+ *     ad * as * B(d/ad, s/as)
+ *   = if d/ad = 0
+ *         ad * as * 0
+ *     else if (d/ad >= (1 - s/as)
+ *         ad * as * 1
+ *     else
+ *         ad * as * ((d/ad) / (1 - s/as))
+ *   = if d = 0
+ *         0
+ *     elif as * d >= ad * (as - s)
+ *         ad * as
+ *     else
+ *         as * (as * d / (as - s))
+ *
+ */
 static force_inline float
 blend_color_dodge (float sa, float s, float da, float d)
 {
@@ -413,6 +503,26 @@ blend_color_dodge (float sa, float s, float da, float d)
 	return sa * sa * d / (sa - s);
 }
 
+/*
+ * Color burn
+ *
+ * We modify the first clause "if d = 1" to "if d >= 1" since with
+ * premultiplied colors d > 1 can actually happen.
+ *
+ *     ad * as * B(d/ad, s/as)
+ *   = if d/ad >= 1
+ *         ad * as * 1
+ *     elif (1 - d/ad) >= s/as
+ *         ad * as * 0
+ *     else
+ *         ad * as * (1 - ((1 - d/ad) / (s/as)))
+ *   = if d >= ad
+ *         ad * as
+ *     elif as * ad - as * d >= ad * s
+ *         0
+ *     else
+ *         ad * as  - as * as * (ad - d) / s
+ */
 static force_inline float
 blend_color_burn (float sa, float s, float da, float d)
 {
@@ -426,6 +536,23 @@ blend_color_burn (float sa, float s, float da, float d)
 	return sa * (da - sa * (da - d) / s);
 }
 
+/*
+ * Hard light
+ *
+ *     ad * as * B(d/ad, s/as)
+ *   = if (s/as <= 0.5)
+ *         ad * as * Multiply (d/ad, 2 * s/as)
+ *     else
+ *         ad * as * Screen (d/ad, 2 * s/as - 1)
+ *   = if 2 * s <= as
+ *         ad * as * d/ad * 2 * s / as
+ *     else
+ *         ad * as * (d/ad + (2 * s/as - 1) + d/ad * (2 * s/as - 1))
+ *   = if 2 * s <= as
+ *         2 * s * d
+ *     else
+ *         as * ad - 2 * (ad - d) * (as - s)
+ */
 static force_inline float
 blend_hard_light (float sa, float s, float da, float d)
 {
@@ -435,6 +562,23 @@ blend_hard_light (float sa, float s, float da, float d)
 	return sa * da - 2 * (da - d) * (sa - s);
 }
 
+/*
+ * Soft light
+ *
+ *     ad * as * B(d/ad, s/as)
+ *   = if (s/as <= 0.5)
+ *         ad * as * (d/ad - (1 - 2 * s/as) * d/ad * (1 - d/ad))
+ *     else if (d/ad <= 0.25)
+ *         ad * as * (d/ad + (2 * s/as - 1) * ((((16 * d/ad - 12) * d/ad + 4) * d/ad) - d/ad))
+ *     else
+ *         ad * as * (d/ad + (2 * s/as - 1) * sqrt (d/ad))
+ *   = if (2 * s <= as)
+ *         d * as - d * (ad - d) * (as - 2 * s) / ad;
+ *     else if (4 * d <= ad)
+ *         (2 * s - as) * d * ((16 * d / ad - 12) * d / ad + 3);
+ *     else
+ *         d * as + (sqrt (d * ad) - d) * (2 * s - as);
+ */
 static force_inline float
 blend_soft_light (float sa, float s, float da, float d)
 {
@@ -449,7 +593,7 @@ blend_soft_light (float sa, float s, float da, float d)
     {
 	if (FLOAT_IS_ZERO (da))
 	{
-	    return 0.0f;
+	    return d * sa;
 	}
 	else
 	{
@@ -461,6 +605,20 @@ blend_soft_light (float sa, float s, float da, float d)
     }
 }
 
+/*
+ * Difference
+ *
+ *     ad * as * B(s/as, d/ad)
+ *   = ad * as * abs (s/as - d/ad)
+ *   = if (s/as <= d/ad)
+ *         ad * as * (d/ad - s/as)
+ *     else
+ *         ad * as * (s/as - d/ad)
+ *   = if (ad * s <= as * d)
+ *        as * d - ad * s
+ *     else
+ *        ad * s - as * d
+ */
 static force_inline float
 blend_difference (float sa, float s, float da, float d)
 {
@@ -473,6 +631,13 @@ blend_difference (float sa, float s, float da, float d)
 	return sda - dsa;
 }
 
+/*
+ * Exclusion
+ *
+ *     ad * as * B(s/as, d/ad)
+ *   = ad * as * (d/ad + s/as - 2 * d/ad * s/as)
+ *   = as * d + ad * s - 2 * s * d
+ */
 static force_inline float
 blend_exclusion (float sa, float s, float da, float d)
 {
@@ -492,116 +657,79 @@ MAKE_SEPARABLE_PDF_COMBINERS (difference)
 MAKE_SEPARABLE_PDF_COMBINERS (exclusion)
 
 /*
- * PDF nonseperable blend modes.
- *
- * These are implemented using the following functions to operate in Hsl
- * space, with Cmax, Cmid, Cmin referring to the max, mid and min value
- * of the red, green and blue components.
+ * PDF nonseperable blend modes are implemented using the following functions
+ * to operate in Hsl space, with Cmax, Cmid, Cmin referring to the max, mid
+ * and min value of the red, green and blue components.
  *
  * LUM (C) = 0.3 × Cred + 0.59 × Cgreen + 0.11 × Cblue
  *
  * clip_color (C):
- *   l = LUM (C)
- *   min = Cmin
- *   max = Cmax
- *   if n < 0.0
- *     C = l + (((C – l) × l) ⁄     (l – min))
- *   if x > 1.0
- *     C = l + (((C – l) × (1 – l)) (max – l))
- *   return C
+ *     l = LUM (C)
+ *     min = Cmin
+ *     max = Cmax
+ *     if n < 0.0
+ *         C = l + (((C – l) × l) ⁄ (l – min))
+ *     if x > 1.0
+ *         C = l + (((C – l) × (1 – l) ) ⁄ (max – l))
+ *     return C
  *
  * set_lum (C, l):
- *   d = l – LUM (C)
- *   C += d
- *   return clip_color (C)
+ *     d = l – LUM (C)
+ *     C += d
+ *     return clip_color (C)
  *
  * SAT (C) = CH_MAX (C) - CH_MIN (C)
  *
  * set_sat (C, s):
- *  if Cmax > Cmin
- *    Cmid = ( ( ( Cmid – Cmin ) × s ) ⁄ ( Cmax – Cmin ) )
- *    Cmax = s
- *  else
- *    Cmid = Cmax = 0.0
- *  Cmin = 0.0
- *  return C
+ *     if Cmax > Cmin
+ *         Cmid = ( ( ( Cmid – Cmin ) × s ) ⁄ ( Cmax – Cmin ) )
+ *         Cmax = s
+ *     else
+ *         Cmid = Cmax = 0.0
+ *         Cmin = 0.0
+ *     return C
  */
 
 /* For premultiplied colors, we need to know what happens when C is
  * multiplied by a real number. LUM and SAT are linear:
  *
- *    LUM (r × C) = r × LUM (C)		SAT (r × C) = r × SAT (C)
+ *     LUM (r × C) = r × LUM (C)	SAT (r * C) = r * SAT (C)
  *
  * If we extend clip_color with an extra argument a and change
  *
- *        if x >= 1.0
+ *     if x >= 1.0
  *
  * into
  *
- *        if x >= a
+ *     if x >= a
  *
  * then clip_color is also linear:
  *
- *     r * clip_color (C, a) = clip_color (r_c, ra);
+ *     r * clip_color (C, a) = clip_color (r * C, r * a);
  *
  * for positive r.
  *
  * Similarly, we can extend set_lum with an extra argument that is just passed
  * on to clip_color:
  *
- *     r × set_lum ( C, l, a)
+ *       r * set_lum (C, l, a)
  *
- *   = r × clip_color ( C + l - LUM (C), a)
+ *     = r × clip_color (C + l - LUM (C), a)
  *
- *   = clip_color ( r * C + r × l - LUM (r × C), r * a)
+ *     = clip_color (r * C + r × l - r * LUM (C), r * a)
  *
- *   = set_lum ( r * C, r * l, r * a)
+ *     = set_lum (r * C, r * l, r * a)
  *
  * Finally, set_sat:
  *
- *     r * set_sat (C, s) = set_sat (x * C, r * s)
+ *       r * set_sat (C, s) = set_sat (x * C, r * s)
  *
- * The above holds for all non-zero x because they x'es in the fraction for
+ * The above holds for all non-zero x, because the x'es in the fraction for
  * C_mid cancel out. Specifically, it holds for x = r:
  *
- *     r * set_sat (C, s) = set_sat (r_c, rs)
- *
- *
- *
- *
- * So, for the non-separable PDF blend modes, we have (using s, d for
- * non-premultiplied colors, and S, D for premultiplied:
- *
- *   Color:
- *
- *     a_s * a_d * B(s, d)
- *   = a_s * a_d * set_lum (S/a_s, LUM (D/a_d), 1)
- *   = set_lum (S * a_d, a_s * LUM (D), a_s * a_d)
- *
- *
- *   Luminosity:
- *
- *     a_s * a_d * B(s, d)
- *   = a_s * a_d * set_lum (D/a_d, LUM(S/a_s), 1)
- *   = set_lum (a_s * D, a_d * LUM(S), a_s * a_d)
- *
- *
- *   Saturation:
- *
- *     a_s * a_d * B(s, d)
- *   = a_s * a_d * set_lum (set_sat (D/a_d, SAT (S/a_s)), LUM (D/a_d), 1)
- *   = set_lum (a_s * a_d * set_sat (D/a_d, SAT (S/a_s)),
- *                                        a_s * LUM (D), a_s * a_d)
- *   = set_lum (set_sat (a_s * D, a_d * SAT (S), a_s * LUM (D), a_s * a_d))
- *
- *   Hue:
- *
- *     a_s * a_d * B(s, d)
- *   = a_s * a_d * set_lum (set_sat (S/a_s, SAT (D/a_d)), LUM (D/a_d), 1)
- *   = set_lum (set_sat (a_d * S, a_s * SAT (D)), a_s * LUM (D), a_s * a_d)
+ *       r * set_sat (C, s) = set_sat (r * C, r * s)
  *
  */
-
 typedef struct
 {
     float	r;
@@ -769,9 +897,12 @@ set_sat (rgb_t *src, float sat)
     *min = 0.0f;
 }
 
-/*
- * Hue:
- * B(Cb, Cs) = set_lum (set_sat (Cs, SAT (Cb)), LUM (Cb))
+/* Hue:
+ *
+ *       as * ad * B(s/as, d/as)
+ *     = as * ad * set_lum (set_sat (s/as, SAT (d/ad)), LUM (d/ad), 1)
+ *     = set_lum (set_sat (ad * s, as * SAT (d)), as * LUM (d), as * ad)
+ *
  */
 static force_inline void
 blend_hsl_hue (rgb_t *res,
@@ -786,9 +917,14 @@ blend_hsl_hue (rgb_t *res,
     set_lum (res, sa * da, get_lum (dest) * sa);
 }
 
-/*
- * Saturation:
- * B(Cb, Cs) = set_lum (set_sat (Cb, SAT (Cs)), LUM (Cb))
+/* 
+ * Saturation
+ *
+ *     as * ad * B(s/as, d/ad)
+ *   = as * ad * set_lum (set_sat (d/ad, SAT (s/as)), LUM (d/ad), 1)
+ *   = set_lum (as * ad * set_sat (d/ad, SAT (s/as)),
+ *                                       as * LUM (d), as * ad)
+ *   = set_lum (set_sat (as * d, ad * SAT (s), as * LUM (d), as * ad))
  */
 static force_inline void
 blend_hsl_saturation (rgb_t *res,
@@ -803,9 +939,12 @@ blend_hsl_saturation (rgb_t *res,
     set_lum (res, sa * da, get_lum (dest) * sa);
 }
 
-/*
- * Color:
- * B(Cb, Cs) = set_lum (Cs, LUM (Cb))
+/* 
+ * Color
+ *
+ *     as * ad * B(s/as, d/as)
+ *   = as * ad * set_lum (s/as, LUM (d/ad), 1)
+ *   = set_lum (s * ad, as * LUM (d), as * ad)
  */
 static force_inline void
 blend_hsl_color (rgb_t *res,
@@ -820,8 +959,11 @@ blend_hsl_color (rgb_t *res,
 }
 
 /*
- * Luminosity:
- * B(Cb, Cs) = set_lum (Cb, LUM (Cs))
+ * Luminosity
+ *
+ *     as * ad * B(s/as, d/ad)
+ *   = as * ad * set_lum (d/ad, LUM (s/as), 1)
+ *   = set_lum (as * d, ad * LUM (s), as * ad)
  */
 static force_inline void
 blend_hsl_luminosity (rgb_t *res,
diff --git a/pixman/pixman/pixman-combine32.c b/pixman/pixman/pixman-combine32.c
index 44ca7e857..4c484d3e3 100644
--- a/pixman/pixman/pixman-combine32.c
+++ b/pixman/pixman/pixman-combine32.c
@@ -434,36 +434,6 @@ combine_add_u (pixman_implementation_t *imp,
     }
 }
 
-static void
-combine_saturate_u (pixman_implementation_t *imp,
-                    pixman_op_t              op,
-                    uint32_t *               dest,
-                    const uint32_t *         src,
-                    const uint32_t *         mask,
-                    int                      width)
-{
-    int i;
-
-    for (i = 0; i < width; ++i)
-    {
-	uint32_t s = combine_mask (src, mask, i);
-	uint32_t d = *(dest + i);
-	uint16_t sa, da;
-
-	sa = s >> A_SHIFT;
-	da = ~d >> A_SHIFT;
-	if (sa > da)
-	{
-	    sa = DIV_UN8 (da, sa);
-	    UN8x4_MUL_UN8 (s, sa);
-	}
-	;
-	UN8x4_ADD_UN8x4 (d, s);
-	*(dest + i) = d;
-    }
-}
-
-
 /*
  * PDF blend modes:
  *
@@ -571,6 +541,15 @@ combine_multiply_ca (pixman_implementation_t *imp,
     }
 }
 
+#define CLAMP(v, low, high)						\
+    do									\
+    {									\
+	if (v < (low))							\
+	    v = (low);							\
+	if (v > (high))							\
+	    v = (high);							\
+    } while (0)
+
 #define PDF_SEPARABLE_BLEND_MODE(name)					\
     static void								\
     combine_ ## name ## _u (pixman_implementation_t *imp,		\
@@ -589,16 +568,28 @@ combine_multiply_ca (pixman_implementation_t *imp,
 	    uint8_t isa = ~sa;						\
 	    uint8_t da = ALPHA_8 (d);					\
 	    uint8_t ida = ~da;						\
-	    uint32_t result;						\
-									\
-	    result = d;							\
-	    UN8x4_MUL_UN8_ADD_UN8x4_MUL_UN8 (result, isa, s, ida);	\
+	    int32_t ra, rr, rg, rb;					\
 	    								\
-	    *(dest + i) = result +					\
-		(DIV_ONE_UN8 (sa * (uint32_t)da) << A_SHIFT) +		\
-		(blend_ ## name (RED_8 (d), da, RED_8 (s), sa) << R_SHIFT) + \
-		(blend_ ## name (GREEN_8 (d), da, GREEN_8 (s), sa) << G_SHIFT) + \
-		(blend_ ## name (BLUE_8 (d), da, BLUE_8 (s), sa));	\
+	    ra = da * 0xff + sa * 0xff - sa * da;			\
+	    rr = isa * RED_8 (d) + ida * RED_8 (s);			\
+	    rg = isa * GREEN_8 (d) + ida * GREEN_8 (s);			\
+	    rb = isa * BLUE_8 (d) + ida * BLUE_8 (s);			\
+									\
+	    rr += blend_ ## name (RED_8 (d), da, RED_8 (s), sa);	\
+	    rg += blend_ ## name (GREEN_8 (d), da, GREEN_8 (s), sa);    \
+	    rb += blend_ ## name (BLUE_8 (d), da, BLUE_8 (s), sa);	\
+                                                                        \
+	    CLAMP (ra, 0, 255 * 255);				        \
+	    CLAMP (rr, 0, 255 * 255);				        \
+	    CLAMP (rg, 0, 255 * 255);				        \
+	    CLAMP (rb, 0, 255 * 255);				        \
+									\
+	    ra = DIV_ONE_UN8 (ra);					\
+	    rr = DIV_ONE_UN8 (rr);					\
+	    rg = DIV_ONE_UN8 (rg);					\
+	    rb = DIV_ONE_UN8 (rb);					\
+									\
+	    *(dest + i) = ra << 24 | rr << 16 | rg << 8 | rb;		\
 	}								\
     }									\
     									\
@@ -618,20 +609,35 @@ combine_multiply_ca (pixman_implementation_t *imp,
 	    uint32_t d = *(dest + i);					\
 	    uint8_t da = ALPHA_8 (d);					\
 	    uint8_t ida = ~da;						\
-	    uint32_t result;						\
-            								\
+	    int32_t ra, rr, rg, rb;					\
+	    uint8_t ira, iga, iba;					\
+	    								\
 	    combine_mask_ca (&s, &m);					\
-            								\
-	    result = d;							\
-	    UN8x4_MUL_UN8x4_ADD_UN8x4_MUL_UN8 (result, ~m, s, ida);     \
-            								\
-	    result +=							\
-	        (DIV_ONE_UN8 (ALPHA_8 (m) * (uint32_t)da) << A_SHIFT) +	\
-	        (blend_ ## name (RED_8 (d), da, RED_8 (s), RED_8 (m)) << R_SHIFT) + \
-	        (blend_ ## name (GREEN_8 (d), da, GREEN_8 (s), GREEN_8 (m)) << G_SHIFT) + \
-	        (blend_ ## name (BLUE_8 (d), da, BLUE_8 (s), BLUE_8 (m))); \
 	    								\
-	    *(dest + i) = result;					\
+	    ira = ~RED_8 (m);						\
+	    iga = ~GREEN_8 (m);						\
+	    iba = ~BLUE_8 (m);						\
+									\
+	    ra = da * 0xff + ALPHA_8 (s) * 0xff - ALPHA_8 (s) * da;	\
+	    rr = ira * RED_8 (d) + ida * RED_8 (s);			\
+	    rg = iga * GREEN_8 (d) + ida * GREEN_8 (s);			\
+	    rb = iba * BLUE_8 (d) + ida * BLUE_8 (s);			\
+									\
+	    rr += blend_ ## name (RED_8 (d), da, RED_8 (s), RED_8 (m));	\
+	    rg += blend_ ## name (GREEN_8 (d), da, GREEN_8 (s), GREEN_8 (m)); \
+	    rb += blend_ ## name (BLUE_8 (d), da, BLUE_8 (s), BLUE_8 (m)); \
+									\
+	    CLAMP (ra, 0, 255 * 255);				        \
+	    CLAMP (rr, 0, 255 * 255);				        \
+	    CLAMP (rg, 0, 255 * 255);				        \
+	    CLAMP (rb, 0, 255 * 255);				        \
+									\
+	    ra = DIV_ONE_UN8 (ra);					\
+	    rr = DIV_ONE_UN8 (rr);					\
+	    rg = DIV_ONE_UN8 (rg);					\
+	    rb = DIV_ONE_UN8 (rb);					\
+									\
+	    *(dest + i) = ra << 24 | rr << 16 | rg << 8 | rb;		\
 	}								\
     }
 
@@ -642,10 +648,10 @@ combine_multiply_ca (pixman_implementation_t *imp,
  *    = ad * as * (d/ad + s/as - s/as * d/ad)
  *    = ad * s + as * d - s * d
  */
-static inline uint32_t
-blend_screen (uint32_t d, uint32_t ad, uint32_t s, uint32_t as)
+static inline int32_t
+blend_screen (int32_t d, int32_t ad, int32_t s, int32_t as)
 {
-    return DIV_ONE_UN8 (s * ad + d * as - s * d);
+    return s * ad + d * as - s * d;
 }
 
 PDF_SEPARABLE_BLEND_MODE (screen)
@@ -672,8 +678,8 @@ PDF_SEPARABLE_BLEND_MODE (screen)
  *     else
  *         as * ad - 2 * (ad - d) * (as - s)
  */
-static inline uint32_t
-blend_overlay (uint32_t d, uint32_t ad, uint32_t s, uint32_t as)
+static inline int32_t
+blend_overlay (int32_t d, int32_t ad, int32_t s, int32_t as)
 {
     uint32_t r;
 
@@ -682,7 +688,7 @@ blend_overlay (uint32_t d, uint32_t ad, uint32_t s, uint32_t as)
     else
 	r = as * ad - 2 * (ad - d) * (as - s);
 
-    return DIV_ONE_UN8 (r);
+    return r;
 }
 
 PDF_SEPARABLE_BLEND_MODE (overlay)
@@ -694,13 +700,13 @@ PDF_SEPARABLE_BLEND_MODE (overlay)
  *   = ad * as * MIN(d/ad, s/as)
  *   = MIN (as * d, ad * s)
  */
-static inline uint32_t
-blend_darken (uint32_t d, uint32_t ad, uint32_t s, uint32_t as)
+static inline int32_t
+blend_darken (int32_t d, int32_t ad, int32_t s, int32_t as)
 {
     s = ad * s;
     d = as * d;
 
-    return DIV_ONE_UN8 (s > d ? d : s);
+    return s > d ? d : s;
 }
 
 PDF_SEPARABLE_BLEND_MODE (darken)
@@ -712,86 +718,18 @@ PDF_SEPARABLE_BLEND_MODE (darken)
  *   = ad * as * MAX(d/ad, s/as)
  *   = MAX (as * d, ad * s)
  */
-static inline uint32_t
-blend_lighten (uint32_t d, uint32_t ad, uint32_t s, uint32_t as)
+static inline int32_t
+blend_lighten (int32_t d, int32_t ad, int32_t s, int32_t as)
 {
     s = ad * s;
     d = as * d;
     
-    return DIV_ONE_UN8 (s > d ? s : d);
+    return s > d ? s : d;
 }
 
 PDF_SEPARABLE_BLEND_MODE (lighten)
 
 /*
- * Color dodge
- *
- *     ad * as * B(d/ad, s/as)
- *   = if d/ad = 0
- *         ad * as * 0
- *     else if (d/ad >= (1 - s/as)
- *         ad * as * 1
- *     else
- *         ad * as * ((d/ad) / (1 - s/as))
- *   = if d = 0
- *         0
- *     elif as * d >= ad * (as - s)
- *         ad * as
- *     else
- *         as * (as * d / (as - s))
- *
- */
-static inline uint32_t
-blend_color_dodge (uint32_t d, uint32_t ad, uint32_t s, uint32_t as)
-{
-    if (d == 0)
-        return 0;
-    else if (as * d >= ad * (as - s))
-	return DIV_ONE_UN8 (as * ad);
-    else if (as - s == 0)
-        return DIV_ONE_UN8 (as * ad);
-    else
-        return DIV_ONE_UN8 (as * ((d * as) / ((as - s))));
-}
-
-PDF_SEPARABLE_BLEND_MODE (color_dodge)
-
-/*
- * Color burn
- *
- * We modify the first clause "if d = 1" to "if d >= 1" since with
- * premultiplied colors d > 1 can actually happen.
- *
- *     ad * as * B(d/ad, s/as)
- *   = if d/ad >= 1
- *         ad * as * 1
- *     elif (1 - d/ad) >= s/as
- *         ad * as * 0
- *     else
- *         ad * as * (1 - ((1 - d/ad) / (s/as)))
- *   = if d >= ad
- *         ad * as
- *     elif as * ad - as * d >= ad * s
- *         0
- *     else
- *         ad * as  - as * as * (ad - d) / s
- */
-static inline uint32_t
-blend_color_burn (uint32_t d, uint32_t ad, uint32_t s, uint32_t as)
-{
-    if (d >= ad)
-	return DIV_ONE_UN8 (ad * as);
-    else if (as * ad - as * d >= ad * s)
-	return 0;
-    else if (s == 0)
-	return 0;
-    else
-	return DIV_ONE_UN8 (ad * as - (as * as * (ad - d)) / s);
-}
-
-PDF_SEPARABLE_BLEND_MODE (color_burn)
-
-/*
  * Hard light
  *
  *     ad * as * B(d/ad, s/as)
@@ -808,72 +746,18 @@ PDF_SEPARABLE_BLEND_MODE (color_burn)
  *     else
  *         as * ad - 2 * (ad - d) * (as - s)
  */
-static inline uint32_t
-blend_hard_light (uint32_t d, uint32_t ad, uint32_t s, uint32_t as)
+static inline int32_t
+blend_hard_light (int32_t d, int32_t ad, int32_t s, int32_t as)
 {
     if (2 * s < as)
-	return DIV_ONE_UN8 (2 * s * d);
+	return 2 * s * d;
     else
-	return DIV_ONE_UN8 (as * ad - 2 * (ad - d) * (as - s));
+	return as * ad - 2 * (ad - d) * (as - s);
 }
 
 PDF_SEPARABLE_BLEND_MODE (hard_light)
 
 /*
- * Soft light
- *
- *     ad * as * B(d/ad, s/as)
- *   = if (s/as <= 0.5)
- *         ad * as * (d/ad - (1 - 2 * s/as) * d/ad * (1 - d/ad))
- *     else if (d/ad <= 0.25)
- *         ad * as * (d/ad + (2 * s/as - 1) * ((((16 * d/ad - 12) * d/ad + 4) * d/ad) - d/ad))
- *     else
- *         ad * as * (d/ad + (2 * s/as - 1) * sqrt (d/ad))
- *   = if (2 * s <= as)
- *         d * as - d * (ad - d) * (as - 2 * s) / ad;
- *     else if (4 * d <= ad)
- *         (2 * s - as) * d * ((16 * d / ad - 12) * d / ad + 3);
- *     else
- *         d * as + (sqrt (d * ad) - d) * (2 * s - as);
- */
-static inline uint32_t
-blend_soft_light (uint32_t d_org,
-		  uint32_t ad_org,
-		  uint32_t s_org,
-		  uint32_t as_org)
-{
-    double d = d_org * (1.0 / MASK);
-    double ad = ad_org * (1.0 / MASK);
-    double s = s_org * (1.0 / MASK);
-    double as = as_org * (1.0 / MASK);
-    double r;
-
-    if (2 * s <= as)
-    {
-	if (ad == 0)
-	    r = d * as;
-	else
-	    r = d * as - d * (ad - d) * (as - 2 * s) / ad;
-    }
-    else if (ad == 0)
-    {
-	r = 0;
-    }
-    else if (4 * d <= ad)
-    {
-	r = d * as +
-	    (2 * s - as) * d * ((16 * d / ad - 12) * d / ad + 3);
-    }
-    else
-    {
-	r = d * as + (sqrt (d * ad) - d) * (2 * s - as);
-    }
-    return r * MASK + 0.5;
-}
-
-PDF_SEPARABLE_BLEND_MODE (soft_light)
-
-/*
  * Difference
  *
  *     ad * as * B(s/as, d/ad)
@@ -887,16 +771,16 @@ PDF_SEPARABLE_BLEND_MODE (soft_light)
  *     else
  *        ad * s - as * d
  */
-static inline uint32_t
-blend_difference (uint32_t d, uint32_t ad, uint32_t s, uint32_t as)
+static inline int32_t
+blend_difference (int32_t d, int32_t ad, int32_t s, int32_t as)
 {
-    uint32_t das = d * as;
-    uint32_t sad = s * ad;
+    int32_t das = d * as;
+    int32_t sad = s * ad;
 
     if (sad < das)
-	return DIV_ONE_UN8 (das - sad);
+	return das - sad;
     else
-	return DIV_ONE_UN8 (sad - das);
+	return sad - das;
 }
 
 PDF_SEPARABLE_BLEND_MODE (difference)
@@ -912,796 +796,16 @@ PDF_SEPARABLE_BLEND_MODE (difference)
 /* This can be made faster by writing it directly and not using
  * PDF_SEPARABLE_BLEND_MODE, but that's a performance optimization */
 
-static inline uint32_t
-blend_exclusion (uint32_t d, uint32_t ad, uint32_t s, uint32_t as)
+static inline int32_t
+blend_exclusion (int32_t d, int32_t ad, int32_t s, int32_t as)
 {
-    return DIV_ONE_UN8 (s * ad + d * as - 2 * d * s);
+    return s * ad + d * as - 2 * d * s;
 }
 
 PDF_SEPARABLE_BLEND_MODE (exclusion)
 
 #undef PDF_SEPARABLE_BLEND_MODE
 
-/*
- * PDF nonseperable blend modes are implemented using the following functions
- * to operate in Hsl space, with Cmax, Cmid, Cmin referring to the max, mid
- * and min value of the red, green and blue components.
- *
- * LUM (C) = 0.3 × Cred + 0.59 × Cgreen + 0.11 × Cblue
- *
- * clip_color (C):
- *     l = LUM (C)
- *     min = Cmin
- *     max = Cmax
- *     if n < 0.0
- *         C = l + (((C – l) × l) ⁄ (l – min))
- *     if x > 1.0
- *         C = l + (((C – l) × (1 – l) ) ⁄ (max – l))
- *     return C
- *
- * set_lum (C, l):
- *     d = l – LUM (C)
- *     C += d
- *     return clip_color (C)
- *
- * SAT (C) = CH_MAX (C) - CH_MIN (C)
- *
- * set_sat (C, s):
- *     if Cmax > Cmin
- *         Cmid = ( ( ( Cmid – Cmin ) × s ) ⁄ ( Cmax – Cmin ) )
- *         Cmax = s
- *     else
- *         Cmid = Cmax = 0.0
- *         Cmin = 0.0
- *     return C
- */
-
-/* For premultiplied colors, we need to know what happens when C is
- * multiplied by a real number. LUM and SAT are linear:
- *
- *     LUM (r × C) = r × LUM (C)	SAT (r * C) = r * SAT (C)
- *
- * If we extend clip_color with an extra argument a and change
- *
- *     if x >= 1.0
- *
- * into
- *
- *     if x >= a
- *
- * then clip_color is also linear:
- *
- *     r * clip_color (C, a) = clip_color (r * C, r * a);
- *
- * for positive r.
- *
- * Similarly, we can extend set_lum with an extra argument that is just passed
- * on to clip_color:
- *
- *       r * set_lum (C, l, a)
- *
- *     = r × clip_color (C + l - LUM (C), a)
- *
- *     = clip_color (r * C + r × l - r * LUM (C), r * a)
- *
- *     = set_lum (r * C, r * l, r * a)
- *
- * Finally, set_sat:
- *
- *       r * set_sat (C, s) = set_sat (x * C, r * s)
- *
- * The above holds for all non-zero x, because the x'es in the fraction for
- * C_mid cancel out. Specifically, it holds for x = r:
- *
- *       r * set_sat (C, s) = set_sat (r * C, r * s)
- *
- */
-
-#define CH_MIN(c) (c[0] < c[1] ? (c[0] < c[2] ? c[0] : c[2]) : (c[1] < c[2] ? c[1] : c[2]))
-#define CH_MAX(c) (c[0] > c[1] ? (c[0] > c[2] ? c[0] : c[2]) : (c[1] > c[2] ? c[1] : c[2]))
-#define LUM(c) ((c[0] * 30 + c[1] * 59 + c[2] * 11) / 100)
-#define SAT(c) (CH_MAX (c) - CH_MIN (c))
-
-#define PDF_NON_SEPARABLE_BLEND_MODE(name)				\
-    static void								\
-    combine_ ## name ## _u (pixman_implementation_t *imp,		\
-			    pixman_op_t		     op,		\
-                            uint32_t *               dest,		\
-			    const uint32_t *         src,		\
-			    const uint32_t *         mask,		\
-			    int                      width)		\
-    {									\
-	int i;								\
-	for (i = 0; i < width; ++i)					\
-	{								\
-	    uint32_t s = combine_mask (src, mask, i);			\
-	    uint32_t d = *(dest + i);					\
-	    uint8_t sa = ALPHA_8 (s);					\
-	    uint8_t isa = ~sa;						\
-	    uint8_t da = ALPHA_8 (d);					\
-	    uint8_t ida = ~da;						\
-	    uint32_t result;						\
-	    uint32_t sc[3], dc[3], c[3];				\
-            								\
-	    result = d;							\
-	    UN8x4_MUL_UN8_ADD_UN8x4_MUL_UN8 (result, isa, s, ida);	\
-	    dc[0] = RED_8 (d);						\
-	    sc[0] = RED_8 (s);						\
-	    dc[1] = GREEN_8 (d);					\
-	    sc[1] = GREEN_8 (s);					\
-	    dc[2] = BLUE_8 (d);						\
-	    sc[2] = BLUE_8 (s);						\
-	    blend_ ## name (c, dc, da, sc, sa);				\
-            								\
-	    *(dest + i) = result +					\
-		(DIV_ONE_UN8 (sa * (uint32_t)da) << A_SHIFT) +		\
-		(DIV_ONE_UN8 (c[0]) << R_SHIFT) +			\
-		(DIV_ONE_UN8 (c[1]) << G_SHIFT) +			\
-		(DIV_ONE_UN8 (c[2]));					\
-	}								\
-    }
-
-static void
-set_lum (uint32_t dest[3], uint32_t src[3], uint32_t sa, uint32_t lum)
-{
-    double a, l, min, max;
-    double tmp[3];
-
-    a = sa * (1.0 / MASK);
-
-    l = lum * (1.0 / MASK);
-    tmp[0] = src[0] * (1.0 / MASK);
-    tmp[1] = src[1] * (1.0 / MASK);
-    tmp[2] = src[2] * (1.0 / MASK);
-
-    l = l - LUM (tmp);
-    tmp[0] += l;
-    tmp[1] += l;
-    tmp[2] += l;
-
-    /* clip_color */
-    l = LUM (tmp);
-    min = CH_MIN (tmp);
-    max = CH_MAX (tmp);
-
-    if (min < 0)
-    {
-	if (l - min == 0.0)
-	{
-	    tmp[0] = 0;
-	    tmp[1] = 0;
-	    tmp[2] = 0;
-	}
-	else
-	{
-	    tmp[0] = l + (tmp[0] - l) * l / (l - min);
-	    tmp[1] = l + (tmp[1] - l) * l / (l - min);
-	    tmp[2] = l + (tmp[2] - l) * l / (l - min);
-	}
-    }
-    if (max > a)
-    {
-	if (max - l == 0.0)
-	{
-	    tmp[0] = a;
-	    tmp[1] = a;
-	    tmp[2] = a;
-	}
-	else
-	{
-	    tmp[0] = l + (tmp[0] - l) * (a - l) / (max - l);
-	    tmp[1] = l + (tmp[1] - l) * (a - l) / (max - l);
-	    tmp[2] = l + (tmp[2] - l) * (a - l) / (max - l);
-	}
-    }
-
-    dest[0] = tmp[0] * MASK + 0.5;
-    dest[1] = tmp[1] * MASK + 0.5;
-    dest[2] = tmp[2] * MASK + 0.5;
-}
-
-static void
-set_sat (uint32_t dest[3], uint32_t src[3], uint32_t sat)
-{
-    int id[3];
-    uint32_t min, max;
-
-    if (src[0] > src[1])
-    {
-	if (src[0] > src[2])
-	{
-	    id[0] = 0;
-	    if (src[1] > src[2])
-	    {
-		id[1] = 1;
-		id[2] = 2;
-	    }
-	    else
-	    {
-		id[1] = 2;
-		id[2] = 1;
-	    }
-	}
-	else
-	{
-	    id[0] = 2;
-	    id[1] = 0;
-	    id[2] = 1;
-	}
-    }
-    else
-    {
-	if (src[0] > src[2])
-	{
-	    id[0] = 1;
-	    id[1] = 0;
-	    id[2] = 2;
-	}
-	else
-	{
-	    id[2] = 0;
-	    if (src[1] > src[2])
-	    {
-		id[0] = 1;
-		id[1] = 2;
-	    }
-	    else
-	    {
-		id[0] = 2;
-		id[1] = 1;
-	    }
-	}
-    }
-
-    max = dest[id[0]];
-    min = dest[id[2]];
-    if (max > min)
-    {
-	dest[id[1]] = (dest[id[1]] - min) * sat / (max - min);
-	dest[id[0]] = sat;
-	dest[id[2]] = 0;
-    }
-    else
-    {
-	dest[0] = dest[1] = dest[2] = 0;
-    }
-}
-
-/* Hue:
- *
- *       as * ad * B(s/as, d/as)
- *     = as * ad * set_lum (set_sat (s/as, SAT (d/ad)), LUM (d/ad), 1)
- *     = set_lum (set_sat (ad * s, as * SAT (d)), as * LUM (d), as * ad)
- *
- */
-static inline void
-blend_hsl_hue (uint32_t r[3],
-               uint32_t d[3],
-               uint32_t ad,
-               uint32_t s[3],
-               uint32_t as)
-{
-    r[0] = s[0] * ad;
-    r[1] = s[1] * ad;
-    r[2] = s[2] * ad;
-    set_sat (r, r, SAT (d) * as);
-    set_lum (r, r, as * ad, LUM (d) * as);
-}
-
-PDF_NON_SEPARABLE_BLEND_MODE (hsl_hue)
-
-/* 
- * Saturation
- *
- *     as * ad * B(s/as, d/ad)
- *   = as * ad * set_lum (set_sat (d/ad, SAT (s/as)), LUM (d/ad), 1)
- *   = set_lum (as * ad * set_sat (d/ad, SAT (s/as)),
- *                                       as * LUM (d), as * ad)
- *   = set_lum (set_sat (as * d, ad * SAT (s), as * LUM (d), as * ad))
- */
-static inline void
-blend_hsl_saturation (uint32_t r[3],
-                      uint32_t d[3],
-                      uint32_t ad,
-                      uint32_t s[3],
-                      uint32_t as)
-{
-    r[0] = d[0] * as;
-    r[1] = d[1] * as;
-    r[2] = d[2] * as;
-    set_sat (r, r, SAT (s) * ad);
-    set_lum (r, r, as * ad, LUM (d) * as);
-}
-
-PDF_NON_SEPARABLE_BLEND_MODE (hsl_saturation)
-
-/* 
- * Color
- *
- *     as * ad * B(s/as, d/as)
- *   = as * ad * set_lum (s/as, LUM (d/ad), 1)
- *   = set_lum (s * ad, as * LUM (d), as * ad)
- */
-static inline void
-blend_hsl_color (uint32_t r[3],
-                 uint32_t d[3],
-                 uint32_t ad,
-                 uint32_t s[3],
-                 uint32_t as)
-{
-    r[0] = s[0] * ad;
-    r[1] = s[1] * ad;
-    r[2] = s[2] * ad;
-    set_lum (r, r, as * ad, LUM (d) * as);
-}
-
-PDF_NON_SEPARABLE_BLEND_MODE (hsl_color)
-
-/*
- * Luminosity
- *
- *     as * ad * B(s/as, d/ad)
- *   = as * ad * set_lum (d/ad, LUM (s/as), 1)
- *   = set_lum (as * d, ad * LUM (s), as * ad)
- */
-static inline void
-blend_hsl_luminosity (uint32_t r[3],
-                      uint32_t d[3],
-                      uint32_t ad,
-                      uint32_t s[3],
-                      uint32_t as)
-{
-    r[0] = d[0] * as;
-    r[1] = d[1] * as;
-    r[2] = d[2] * as;
-    set_lum (r, r, as * ad, LUM (s) * ad);
-}
-
-PDF_NON_SEPARABLE_BLEND_MODE (hsl_luminosity)
-
-#undef SAT
-#undef LUM
-#undef CH_MAX
-#undef CH_MIN
-#undef PDF_NON_SEPARABLE_BLEND_MODE
-
-/* All of the disjoint/conjoint composing functions
- *
- * The four entries in the first column indicate what source contributions
- * come from each of the four areas of the picture -- areas covered by neither
- * A nor B, areas covered only by A, areas covered only by B and finally
- * areas covered by both A and B.
- *
- * Disjoint			Conjoint
- * Fa		Fb		Fa		Fb
- * (0,0,0,0)	0		0		0		0
- * (0,A,0,A)	1		0		1		0
- * (0,0,B,B)	0		1		0		1
- * (0,A,B,A)	1		min((1-a)/b,1)	1		max(1-a/b,0)
- * (0,A,B,B)	min((1-b)/a,1)	1		max(1-b/a,0)	1
- * (0,0,0,A)	max(1-(1-b)/a,0) 0		min(1,b/a)	0
- * (0,0,0,B)	0		max(1-(1-a)/b,0) 0		min(a/b,1)
- * (0,A,0,0)	min(1,(1-b)/a)	0		max(1-b/a,0)	0
- * (0,0,B,0)	0		min(1,(1-a)/b)	0		max(1-a/b,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)
- * (0,A,0,B)	min(1,(1-b)/a)	max(1-(1-a)/b,0) max(1-b/a,0)	min(1,a/b)
- * (0,A,B,0)	min(1,(1-b)/a)	min(1,(1-a)/b)	max(1-b/a,0)	max(1-a/b,0)
- *
- * See  http://marc.info/?l=xfree-render&m=99792000027857&w=2  for more
- * information about these operators.
- */
-
-#define COMBINE_A_OUT 1
-#define COMBINE_A_IN  2
-#define COMBINE_B_OUT 4
-#define COMBINE_B_IN  8
-
-#define COMBINE_CLEAR   0
-#define COMBINE_A       (COMBINE_A_OUT | COMBINE_A_IN)
-#define COMBINE_B       (COMBINE_B_OUT | COMBINE_B_IN)
-#define COMBINE_A_OVER  (COMBINE_A_OUT | COMBINE_B_OUT | COMBINE_A_IN)
-#define COMBINE_B_OVER  (COMBINE_A_OUT | COMBINE_B_OUT | COMBINE_B_IN)
-#define COMBINE_A_ATOP  (COMBINE_B_OUT | COMBINE_A_IN)
-#define COMBINE_B_ATOP  (COMBINE_A_OUT | COMBINE_B_IN)
-#define COMBINE_XOR     (COMBINE_A_OUT | COMBINE_B_OUT)
-
-/* portion covered by a but not b */
-static uint8_t
-combine_disjoint_out_part (uint8_t a, uint8_t b)
-{
-    /* min (1, (1-b) / a) */
-
-    b = ~b;                 /* 1 - b */
-    if (b >= a)             /* 1 - b >= a -> (1-b)/a >= 1 */
-	return MASK;        /* 1 */
-    return DIV_UN8 (b, a);     /* (1-b) / a */
-}
-
-/* portion covered by both a and b */
-static uint8_t
-combine_disjoint_in_part (uint8_t a, uint8_t b)
-{
-    /* max (1-(1-b)/a,0) */
-    /*  = - min ((1-b)/a - 1, 0) */
-    /*  = 1 - min (1, (1-b)/a) */
-
-    b = ~b;                 /* 1 - b */
-    if (b >= a)             /* 1 - b >= a -> (1-b)/a >= 1 */
-	return 0;           /* 1 - 1 */
-    return ~DIV_UN8(b, a);    /* 1 - (1-b) / a */
-}
-
-/* portion covered by a but not b */
-static uint8_t
-combine_conjoint_out_part (uint8_t a, uint8_t b)
-{
-    /* max (1-b/a,0) */
-    /* = 1-min(b/a,1) */
-
-    /* min (1, (1-b) / a) */
-
-    if (b >= a)             /* b >= a -> b/a >= 1 */
-	return 0x00;        /* 0 */
-    return ~DIV_UN8(b, a);    /* 1 - b/a */
-}
-
-/* portion covered by both a and b */
-static uint8_t
-combine_conjoint_in_part (uint8_t a, uint8_t b)
-{
-    /* min (1,b/a) */
-
-    if (b >= a)             /* b >= a -> b/a >= 1 */
-	return MASK;        /* 1 */
-    return DIV_UN8 (b, a);     /* b/a */
-}
-
-#define GET_COMP(v, i)   ((uint16_t) (uint8_t) ((v) >> i))
-
-#define ADD(x, y, i, t)							\
-    ((t) = GET_COMP (x, i) + GET_COMP (y, i),				\
-     (uint32_t) ((uint8_t) ((t) | (0 - ((t) >> G_SHIFT)))) << (i))
-
-#define GENERIC(x, y, i, ax, ay, t, u, v)				\
-    ((t) = (MUL_UN8 (GET_COMP (y, i), ay, (u)) +			\
-            MUL_UN8 (GET_COMP (x, i), ax, (v))),			\
-     (uint32_t) ((uint8_t) ((t) |					\
-                           (0 - ((t) >> G_SHIFT)))) << (i))
-
-static void
-combine_disjoint_general_u (uint32_t *      dest,
-                            const uint32_t *src,
-                            const uint32_t *mask,
-                            int            width,
-                            uint8_t        combine)
-{
-    int i;
-
-    for (i = 0; i < width; ++i)
-    {
-	uint32_t s = combine_mask (src, mask, i);
-	uint32_t d = *(dest + i);
-	uint32_t m, n, o, p;
-	uint16_t Fa, Fb, t, u, v;
-	uint8_t sa = s >> A_SHIFT;
-	uint8_t da = d >> A_SHIFT;
-
-	switch (combine & COMBINE_A)
-	{
-	default:
-	    Fa = 0;
-	    break;
-
-	case COMBINE_A_OUT:
-	    Fa = combine_disjoint_out_part (sa, da);
-	    break;
-
-	case COMBINE_A_IN:
-	    Fa = combine_disjoint_in_part (sa, da);
-	    break;
-
-	case COMBINE_A:
-	    Fa = MASK;
-	    break;
-	}
-
-	switch (combine & COMBINE_B)
-	{
-	default:
-	    Fb = 0;
-	    break;
-
-	case COMBINE_B_OUT:
-	    Fb = combine_disjoint_out_part (da, sa);
-	    break;
-
-	case COMBINE_B_IN:
-	    Fb = combine_disjoint_in_part (da, sa);
-	    break;
-
-	case COMBINE_B:
-	    Fb = MASK;
-	    break;
-	}
-	m = GENERIC (s, d, 0, Fa, Fb, t, u, v);
-	n = GENERIC (s, d, G_SHIFT, Fa, Fb, t, u, v);
-	o = GENERIC (s, d, R_SHIFT, Fa, Fb, t, u, v);
-	p = GENERIC (s, d, A_SHIFT, Fa, Fb, t, u, v);
-	s = m | n | o | p;
-	*(dest + i) = s;
-    }
-}
-
-static void
-combine_disjoint_over_u (pixman_implementation_t *imp,
-                         pixman_op_t              op,
-                         uint32_t *                dest,
-                         const uint32_t *          src,
-                         const uint32_t *          mask,
-                         int                      width)
-{
-    int i;
-
-    for (i = 0; i < width; ++i)
-    {
-	uint32_t s = combine_mask (src, mask, i);
-	uint16_t a = s >> A_SHIFT;
-
-	if (s != 0x00)
-	{
-	    uint32_t d = *(dest + i);
-	    a = combine_disjoint_out_part (d >> A_SHIFT, a);
-	    UN8x4_MUL_UN8_ADD_UN8x4 (d, a, s);
-
-	    *(dest + i) = d;
-	}
-    }
-}
-
-static void
-combine_disjoint_in_u (pixman_implementation_t *imp,
-                       pixman_op_t              op,
-                       uint32_t *                dest,
-                       const uint32_t *          src,
-                       const uint32_t *          mask,
-                       int                      width)
-{
-    combine_disjoint_general_u (dest, src, mask, width, COMBINE_A_IN);
-}
-
-static void
-combine_disjoint_in_reverse_u (pixman_implementation_t *imp,
-                               pixman_op_t              op,
-                               uint32_t *                dest,
-                               const uint32_t *          src,
-                               const uint32_t *          mask,
-                               int                      width)
-{
-    combine_disjoint_general_u (dest, src, mask, width, COMBINE_B_IN);
-}
-
-static void
-combine_disjoint_out_u (pixman_implementation_t *imp,
-                        pixman_op_t              op,
-                        uint32_t *                dest,
-                        const uint32_t *          src,
-                        const uint32_t *          mask,
-                        int                      width)
-{
-    combine_disjoint_general_u (dest, src, mask, width, COMBINE_A_OUT);
-}
-
-static void
-combine_disjoint_out_reverse_u (pixman_implementation_t *imp,
-                                pixman_op_t              op,
-                                uint32_t *                dest,
-                                const uint32_t *          src,
-                                const uint32_t *          mask,
-                                int                      width)
-{
-    combine_disjoint_general_u (dest, src, mask, width, COMBINE_B_OUT);
-}
-
-static void
-combine_disjoint_atop_u (pixman_implementation_t *imp,
-                         pixman_op_t              op,
-                         uint32_t *                dest,
-                         const uint32_t *          src,
-                         const uint32_t *          mask,
-                         int                      width)
-{
-    combine_disjoint_general_u (dest, src, mask, width, COMBINE_A_ATOP);
-}
-
-static void
-combine_disjoint_atop_reverse_u (pixman_implementation_t *imp,
-                                 pixman_op_t              op,
-                                 uint32_t *                dest,
-                                 const uint32_t *          src,
-                                 const uint32_t *          mask,
-                                 int                      width)
-{
-    combine_disjoint_general_u (dest, src, mask, width, COMBINE_B_ATOP);
-}
-
-static void
-combine_disjoint_xor_u (pixman_implementation_t *imp,
-                        pixman_op_t              op,
-                        uint32_t *                dest,
-                        const uint32_t *          src,
-                        const uint32_t *          mask,
-                        int                      width)
-{
-    combine_disjoint_general_u (dest, src, mask, width, COMBINE_XOR);
-}
-
-static void
-combine_conjoint_general_u (uint32_t *      dest,
-                            const uint32_t *src,
-                            const uint32_t *mask,
-                            int            width,
-                            uint8_t        combine)
-{
-    int i;
-
-    for (i = 0; i < width; ++i)
-    {
-	uint32_t s = combine_mask (src, mask, i);
-	uint32_t d = *(dest + i);
-	uint32_t m, n, o, p;
-	uint16_t Fa, Fb, t, u, v;
-	uint8_t sa = s >> A_SHIFT;
-	uint8_t da = d >> A_SHIFT;
-
-	switch (combine & COMBINE_A)
-	{
-	default:
-	    Fa = 0;
-	    break;
-
-	case COMBINE_A_OUT:
-	    Fa = combine_conjoint_out_part (sa, da);
-	    break;
-
-	case COMBINE_A_IN:
-	    Fa = combine_conjoint_in_part (sa, da);
-	    break;
-
-	case COMBINE_A:
-	    Fa = MASK;
-	    break;
-	}
-
-	switch (combine & COMBINE_B)
-	{
-	default:
-	    Fb = 0;
-	    break;
-
-	case COMBINE_B_OUT:
-	    Fb = combine_conjoint_out_part (da, sa);
-	    break;
-
-	case COMBINE_B_IN:
-	    Fb = combine_conjoint_in_part (da, sa);
-	    break;
-
-	case COMBINE_B:
-	    Fb = MASK;
-	    break;
-	}
-
-	m = GENERIC (s, d, 0, Fa, Fb, t, u, v);
-	n = GENERIC (s, d, G_SHIFT, Fa, Fb, t, u, v);
-	o = GENERIC (s, d, R_SHIFT, Fa, Fb, t, u, v);
-	p = GENERIC (s, d, A_SHIFT, Fa, Fb, t, u, v);
-
-	s = m | n | o | p;
-
-	*(dest + i) = s;
-    }
-}
-
-static void
-combine_conjoint_over_u (pixman_implementation_t *imp,
-                         pixman_op_t              op,
-                         uint32_t *                dest,
-                         const uint32_t *          src,
-                         const uint32_t *          mask,
-                         int                      width)
-{
-    combine_conjoint_general_u (dest, src, mask, width, COMBINE_A_OVER);
-}
-
-static void
-combine_conjoint_over_reverse_u (pixman_implementation_t *imp,
-                                 pixman_op_t              op,
-                                 uint32_t *                dest,
-                                 const uint32_t *          src,
-                                 const uint32_t *          mask,
-                                 int                      width)
-{
-    combine_conjoint_general_u (dest, src, mask, width, COMBINE_B_OVER);
-}
-
-static void
-combine_conjoint_in_u (pixman_implementation_t *imp,
-                       pixman_op_t              op,
-                       uint32_t *                dest,
-                       const uint32_t *          src,
-                       const uint32_t *          mask,
-                       int                      width)
-{
-    combine_conjoint_general_u (dest, src, mask, width, COMBINE_A_IN);
-}
-
-static void
-combine_conjoint_in_reverse_u (pixman_implementation_t *imp,
-                               pixman_op_t              op,
-                               uint32_t *                dest,
-                               const uint32_t *          src,
-                               const uint32_t *          mask,
-                               int                      width)
-{
-    combine_conjoint_general_u (dest, src, mask, width, COMBINE_B_IN);
-}
-
-static void
-combine_conjoint_out_u (pixman_implementation_t *imp,
-                        pixman_op_t              op,
-                        uint32_t *                dest,
-                        const uint32_t *          src,
-                        const uint32_t *          mask,
-                        int                      width)
-{
-    combine_conjoint_general_u (dest, src, mask, width, COMBINE_A_OUT);
-}
-
-static void
-combine_conjoint_out_reverse_u (pixman_implementation_t *imp,
-                                pixman_op_t              op,
-                                uint32_t *                dest,
-                                const uint32_t *          src,
-                                const uint32_t *          mask,
-                                int                      width)
-{
-    combine_conjoint_general_u (dest, src, mask, width, COMBINE_B_OUT);
-}
-
-static void
-combine_conjoint_atop_u (pixman_implementation_t *imp,
-                         pixman_op_t              op,
-                         uint32_t *                dest,
-                         const uint32_t *          src,
-                         const uint32_t *          mask,
-                         int                      width)
-{
-    combine_conjoint_general_u (dest, src, mask, width, COMBINE_A_ATOP);
-}
-
-static void
-combine_conjoint_atop_reverse_u (pixman_implementation_t *imp,
-                                 pixman_op_t              op,
-                                 uint32_t *                dest,
-                                 const uint32_t *          src,
-                                 const uint32_t *          mask,
-                                 int                      width)
-{
-    combine_conjoint_general_u (dest, src, mask, width, COMBINE_B_ATOP);
-}
-
-static void
-combine_conjoint_xor_u (pixman_implementation_t *imp,
-                        pixman_op_t              op,
-                        uint32_t *                dest,
-                        const uint32_t *          src,
-                        const uint32_t *          mask,
-                        int                      width)
-{
-    combine_conjoint_general_u (dest, src, mask, width, COMBINE_XOR);
-}
-
-
 /* Component alpha combiners */
 
 static void
@@ -2032,428 +1136,6 @@ combine_add_ca (pixman_implementation_t *imp,
     }
 }
 
-static void
-combine_saturate_ca (pixman_implementation_t *imp,
-                     pixman_op_t              op,
-                     uint32_t *                dest,
-                     const uint32_t *          src,
-                     const uint32_t *          mask,
-                     int                      width)
-{
-    int i;
-
-    for (i = 0; i < width; ++i)
-    {
-	uint32_t s, d;
-	uint16_t sa, sr, sg, sb, da;
-	uint16_t t, u, v;
-	uint32_t m, n, o, p;
-
-	d = *(dest + i);
-	s = *(src + i);
-	m = *(mask + i);
-
-	combine_mask_ca (&s, &m);
-
-	sa = (m >> A_SHIFT);
-	sr = (m >> R_SHIFT) & MASK;
-	sg = (m >> G_SHIFT) & MASK;
-	sb =  m             & MASK;
-	da = ~d >> A_SHIFT;
-
-	if (sb <= da)
-	    m = ADD (s, d, 0, t);
-	else
-	    m = GENERIC (s, d, 0, (da << G_SHIFT) / sb, MASK, t, u, v);
-
-	if (sg <= da)
-	    n = ADD (s, d, G_SHIFT, t);
-	else
-	    n = GENERIC (s, d, G_SHIFT, (da << G_SHIFT) / sg, MASK, t, u, v);
-
-	if (sr <= da)
-	    o = ADD (s, d, R_SHIFT, t);
-	else
-	    o = GENERIC (s, d, R_SHIFT, (da << G_SHIFT) / sr, MASK, t, u, v);
-
-	if (sa <= da)
-	    p = ADD (s, d, A_SHIFT, t);
-	else
-	    p = GENERIC (s, d, A_SHIFT, (da << G_SHIFT) / sa, MASK, t, u, v);
-
-	*(dest + i) = m | n | o | p;
-    }
-}
-
-static void
-combine_disjoint_general_ca (uint32_t *      dest,
-                             const uint32_t *src,
-                             const uint32_t *mask,
-                             int            width,
-                             uint8_t        combine)
-{
-    int i;
-
-    for (i = 0; i < width; ++i)
-    {
-	uint32_t s, d;
-	uint32_t m, n, o, p;
-	uint32_t Fa, Fb;
-	uint16_t t, u, v;
-	uint32_t sa;
-	uint8_t da;
-
-	s = *(src + i);
-	m = *(mask + i);
-	d = *(dest + i);
-	da = d >> A_SHIFT;
-
-	combine_mask_ca (&s, &m);
-
-	sa = m;
-
-	switch (combine & COMBINE_A)
-	{
-	default:
-	    Fa = 0;
-	    break;
-
-	case COMBINE_A_OUT:
-	    m = (uint32_t)combine_disjoint_out_part ((uint8_t) (sa >> 0), da);
-	    n = (uint32_t)combine_disjoint_out_part ((uint8_t) (sa >> G_SHIFT), da) << G_SHIFT;
-	    o = (uint32_t)combine_disjoint_out_part ((uint8_t) (sa >> R_SHIFT), da) << R_SHIFT;
-	    p = (uint32_t)combine_disjoint_out_part ((uint8_t) (sa >> A_SHIFT), da) << A_SHIFT;
-	    Fa = m | n | o | p;
-	    break;
-
-	case COMBINE_A_IN:
-	    m = (uint32_t)combine_disjoint_in_part ((uint8_t) (sa >> 0), da);
-	    n = (uint32_t)combine_disjoint_in_part ((uint8_t) (sa >> G_SHIFT), da) << G_SHIFT;
-	    o = (uint32_t)combine_disjoint_in_part ((uint8_t) (sa >> R_SHIFT), da) << R_SHIFT;
-	    p = (uint32_t)combine_disjoint_in_part ((uint8_t) (sa >> A_SHIFT), da) << A_SHIFT;
-	    Fa = m | n | o | p;
-	    break;
-
-	case COMBINE_A:
-	    Fa = ~0;
-	    break;
-	}
-
-	switch (combine & COMBINE_B)
-	{
-	default:
-	    Fb = 0;
-	    break;
-
-	case COMBINE_B_OUT:
-	    m = (uint32_t)combine_disjoint_out_part (da, (uint8_t) (sa >> 0));
-	    n = (uint32_t)combine_disjoint_out_part (da, (uint8_t) (sa >> G_SHIFT)) << G_SHIFT;
-	    o = (uint32_t)combine_disjoint_out_part (da, (uint8_t) (sa >> R_SHIFT)) << R_SHIFT;
-	    p = (uint32_t)combine_disjoint_out_part (da, (uint8_t) (sa >> A_SHIFT)) << A_SHIFT;
-	    Fb = m | n | o | p;
-	    break;
-
-	case COMBINE_B_IN:
-	    m = (uint32_t)combine_disjoint_in_part (da, (uint8_t) (sa >> 0));
-	    n = (uint32_t)combine_disjoint_in_part (da, (uint8_t) (sa >> G_SHIFT)) << G_SHIFT;
-	    o = (uint32_t)combine_disjoint_in_part (da, (uint8_t) (sa >> R_SHIFT)) << R_SHIFT;
-	    p = (uint32_t)combine_disjoint_in_part (da, (uint8_t) (sa >> A_SHIFT)) << A_SHIFT;
-	    Fb = m | n | o | p;
-	    break;
-
-	case COMBINE_B:
-	    Fb = ~0;
-	    break;
-	}
-	m = GENERIC (s, d, 0, GET_COMP (Fa, 0), GET_COMP (Fb, 0), t, u, v);
-	n = GENERIC (s, d, G_SHIFT, GET_COMP (Fa, G_SHIFT), GET_COMP (Fb, G_SHIFT), t, u, v);
-	o = GENERIC (s, d, R_SHIFT, GET_COMP (Fa, R_SHIFT), GET_COMP (Fb, R_SHIFT), t, u, v);
-	p = GENERIC (s, d, A_SHIFT, GET_COMP (Fa, A_SHIFT), GET_COMP (Fb, A_SHIFT), t, u, v);
-
-	s = m | n | o | p;
-
-	*(dest + i) = s;
-    }
-}
-
-static void
-combine_disjoint_over_ca (pixman_implementation_t *imp,
-                          pixman_op_t              op,
-                          uint32_t *                dest,
-                          const uint32_t *          src,
-                          const uint32_t *          mask,
-                          int                      width)
-{
-    combine_disjoint_general_ca (dest, src, mask, width, COMBINE_A_OVER);
-}
-
-static void
-combine_disjoint_in_ca (pixman_implementation_t *imp,
-                        pixman_op_t              op,
-                        uint32_t *                dest,
-                        const uint32_t *          src,
-                        const uint32_t *          mask,
-                        int                      width)
-{
-    combine_disjoint_general_ca (dest, src, mask, width, COMBINE_A_IN);
-}
-
-static void
-combine_disjoint_in_reverse_ca (pixman_implementation_t *imp,
-                                pixman_op_t              op,
-                                uint32_t *                dest,
-                                const uint32_t *          src,
-                                const uint32_t *          mask,
-                                int                      width)
-{
-    combine_disjoint_general_ca (dest, src, mask, width, COMBINE_B_IN);
-}
-
-static void
-combine_disjoint_out_ca (pixman_implementation_t *imp,
-                         pixman_op_t              op,
-                         uint32_t *                dest,
-                         const uint32_t *          src,
-                         const uint32_t *          mask,
-                         int                      width)
-{
-    combine_disjoint_general_ca (dest, src, mask, width, COMBINE_A_OUT);
-}
-
-static void
-combine_disjoint_out_reverse_ca (pixman_implementation_t *imp,
-                                 pixman_op_t              op,
-                                 uint32_t *                dest,
-                                 const uint32_t *          src,
-                                 const uint32_t *          mask,
-                                 int                      width)
-{
-    combine_disjoint_general_ca (dest, src, mask, width, COMBINE_B_OUT);
-}
-
-static void
-combine_disjoint_atop_ca (pixman_implementation_t *imp,
-                          pixman_op_t              op,
-                          uint32_t *                dest,
-                          const uint32_t *          src,
-                          const uint32_t *          mask,
-                          int                      width)
-{
-    combine_disjoint_general_ca (dest, src, mask, width, COMBINE_A_ATOP);
-}
-
-static void
-combine_disjoint_atop_reverse_ca (pixman_implementation_t *imp,
-                                  pixman_op_t              op,
-                                  uint32_t *                dest,
-                                  const uint32_t *          src,
-                                  const uint32_t *          mask,
-                                  int                      width)
-{
-    combine_disjoint_general_ca (dest, src, mask, width, COMBINE_B_ATOP);
-}
-
-static void
-combine_disjoint_xor_ca (pixman_implementation_t *imp,
-                         pixman_op_t              op,
-                         uint32_t *                dest,
-                         const uint32_t *          src,
-                         const uint32_t *          mask,
-                         int                      width)
-{
-    combine_disjoint_general_ca (dest, src, mask, width, COMBINE_XOR);
-}
-
-static void
-combine_conjoint_general_ca (uint32_t *      dest,
-                             const uint32_t *src,
-                             const uint32_t *mask,
-                             int            width,
-                             uint8_t        combine)
-{
-    int i;
-
-    for (i = 0; i < width; ++i)
-    {
-	uint32_t s, d;
-	uint32_t m, n, o, p;
-	uint32_t Fa, Fb;
-	uint16_t t, u, v;
-	uint32_t sa;
-	uint8_t da;
-
-	s = *(src + i);
-	m = *(mask + i);
-	d = *(dest + i);
-	da = d >> A_SHIFT;
-
-	combine_mask_ca (&s, &m);
-
-	sa = m;
-
-	switch (combine & COMBINE_A)
-	{
-	default:
-	    Fa = 0;
-	    break;
-
-	case COMBINE_A_OUT:
-	    m = (uint32_t)combine_conjoint_out_part ((uint8_t) (sa >> 0), da);
-	    n = (uint32_t)combine_conjoint_out_part ((uint8_t) (sa >> G_SHIFT), da) << G_SHIFT;
-	    o = (uint32_t)combine_conjoint_out_part ((uint8_t) (sa >> R_SHIFT), da) << R_SHIFT;
-	    p = (uint32_t)combine_conjoint_out_part ((uint8_t) (sa >> A_SHIFT), da) << A_SHIFT;
-	    Fa = m | n | o | p;
-	    break;
-
-	case COMBINE_A_IN:
-	    m = (uint32_t)combine_conjoint_in_part ((uint8_t) (sa >> 0), da);
-	    n = (uint32_t)combine_conjoint_in_part ((uint8_t) (sa >> G_SHIFT), da) << G_SHIFT;
-	    o = (uint32_t)combine_conjoint_in_part ((uint8_t) (sa >> R_SHIFT), da) << R_SHIFT;
-	    p = (uint32_t)combine_conjoint_in_part ((uint8_t) (sa >> A_SHIFT), da) << A_SHIFT;
-	    Fa = m | n | o | p;
-	    break;
-
-	case COMBINE_A:
-	    Fa = ~0;
-	    break;
-	}
-
-	switch (combine & COMBINE_B)
-	{
-	default:
-	    Fb = 0;
-	    break;
-
-	case COMBINE_B_OUT:
-	    m = (uint32_t)combine_conjoint_out_part (da, (uint8_t) (sa >> 0));
-	    n = (uint32_t)combine_conjoint_out_part (da, (uint8_t) (sa >> G_SHIFT)) << G_SHIFT;
-	    o = (uint32_t)combine_conjoint_out_part (da, (uint8_t) (sa >> R_SHIFT)) << R_SHIFT;
-	    p = (uint32_t)combine_conjoint_out_part (da, (uint8_t) (sa >> A_SHIFT)) << A_SHIFT;
-	    Fb = m | n | o | p;
-	    break;
-
-	case COMBINE_B_IN:
-	    m = (uint32_t)combine_conjoint_in_part (da, (uint8_t) (sa >> 0));
-	    n = (uint32_t)combine_conjoint_in_part (da, (uint8_t) (sa >> G_SHIFT)) << G_SHIFT;
-	    o = (uint32_t)combine_conjoint_in_part (da, (uint8_t) (sa >> R_SHIFT)) << R_SHIFT;
-	    p = (uint32_t)combine_conjoint_in_part (da, (uint8_t) (sa >> A_SHIFT)) << A_SHIFT;
-	    Fb = m | n | o | p;
-	    break;
-
-	case COMBINE_B:
-	    Fb = ~0;
-	    break;
-	}
-	m = GENERIC (s, d, 0, GET_COMP (Fa, 0), GET_COMP (Fb, 0), t, u, v);
-	n = GENERIC (s, d, G_SHIFT, GET_COMP (Fa, G_SHIFT), GET_COMP (Fb, G_SHIFT), t, u, v);
-	o = GENERIC (s, d, R_SHIFT, GET_COMP (Fa, R_SHIFT), GET_COMP (Fb, R_SHIFT), t, u, v);
-	p = GENERIC (s, d, A_SHIFT, GET_COMP (Fa, A_SHIFT), GET_COMP (Fb, A_SHIFT), t, u, v);
-
-	s = m | n | o | p;
-
-	*(dest + i) = s;
-    }
-}
-
-static void
-combine_conjoint_over_ca (pixman_implementation_t *imp,
-                          pixman_op_t              op,
-                          uint32_t *                dest,
-                          const uint32_t *          src,
-                          const uint32_t *          mask,
-                          int                      width)
-{
-    combine_conjoint_general_ca (dest, src, mask, width, COMBINE_A_OVER);
-}
-
-static void
-combine_conjoint_over_reverse_ca (pixman_implementation_t *imp,
-                                  pixman_op_t              op,
-                                  uint32_t *                dest,
-                                  const uint32_t *          src,
-                                  const uint32_t *          mask,
-                                  int                      width)
-{
-    combine_conjoint_general_ca (dest, src, mask, width, COMBINE_B_OVER);
-}
-
-static void
-combine_conjoint_in_ca (pixman_implementation_t *imp,
-                        pixman_op_t              op,
-                        uint32_t *                dest,
-                        const uint32_t *          src,
-                        const uint32_t *          mask,
-                        int                      width)
-{
-    combine_conjoint_general_ca (dest, src, mask, width, COMBINE_A_IN);
-}
-
-static void
-combine_conjoint_in_reverse_ca (pixman_implementation_t *imp,
-                                pixman_op_t              op,
-                                uint32_t *                dest,
-                                const uint32_t *          src,
-                                const uint32_t *          mask,
-                                int                      width)
-{
-    combine_conjoint_general_ca (dest, src, mask, width, COMBINE_B_IN);
-}
-
-static void
-combine_conjoint_out_ca (pixman_implementation_t *imp,
-                         pixman_op_t              op,
-                         uint32_t *                dest,
-                         const uint32_t *          src,
-                         const uint32_t *          mask,
-                         int                      width)
-{
-    combine_conjoint_general_ca (dest, src, mask, width, COMBINE_A_OUT);
-}
-
-static void
-combine_conjoint_out_reverse_ca (pixman_implementation_t *imp,
-                                 pixman_op_t              op,
-                                 uint32_t *                dest,
-                                 const uint32_t *          src,
-                                 const uint32_t *          mask,
-                                 int                      width)
-{
-    combine_conjoint_general_ca (dest, src, mask, width, COMBINE_B_OUT);
-}
-
-static void
-combine_conjoint_atop_ca (pixman_implementation_t *imp,
-                          pixman_op_t              op,
-                          uint32_t *                dest,
-                          const uint32_t *          src,
-                          const uint32_t *          mask,
-                          int                      width)
-{
-    combine_conjoint_general_ca (dest, src, mask, width, COMBINE_A_ATOP);
-}
-
-static void
-combine_conjoint_atop_reverse_ca (pixman_implementation_t *imp,
-                                  pixman_op_t              op,
-                                  uint32_t *                dest,
-                                  const uint32_t *          src,
-                                  const uint32_t *          mask,
-                                  int                      width)
-{
-    combine_conjoint_general_ca (dest, src, mask, width, COMBINE_B_ATOP);
-}
-
-static void
-combine_conjoint_xor_ca (pixman_implementation_t *imp,
-                         pixman_op_t              op,
-                         uint32_t *                dest,
-                         const uint32_t *          src,
-                         const uint32_t *          mask,
-                         int                      width)
-{
-    combine_conjoint_general_ca (dest, src, mask, width, COMBINE_XOR);
-}
-
 void
 _pixman_setup_combiner_functions_32 (pixman_implementation_t *imp)
 {
@@ -2471,51 +1153,15 @@ _pixman_setup_combiner_functions_32 (pixman_implementation_t *imp)
     imp->combine_32[PIXMAN_OP_ATOP_REVERSE] = combine_atop_reverse_u;
     imp->combine_32[PIXMAN_OP_XOR] = combine_xor_u;
     imp->combine_32[PIXMAN_OP_ADD] = combine_add_u;
-    imp->combine_32[PIXMAN_OP_SATURATE] = combine_saturate_u;
-
-    /* Disjoint, unified */
-    imp->combine_32[PIXMAN_OP_DISJOINT_CLEAR] = combine_clear;
-    imp->combine_32[PIXMAN_OP_DISJOINT_SRC] = combine_src_u;
-    imp->combine_32[PIXMAN_OP_DISJOINT_DST] = combine_dst;
-    imp->combine_32[PIXMAN_OP_DISJOINT_OVER] = combine_disjoint_over_u;
-    imp->combine_32[PIXMAN_OP_DISJOINT_OVER_REVERSE] = combine_saturate_u;
-    imp->combine_32[PIXMAN_OP_DISJOINT_IN] = combine_disjoint_in_u;
-    imp->combine_32[PIXMAN_OP_DISJOINT_IN_REVERSE] = combine_disjoint_in_reverse_u;
-    imp->combine_32[PIXMAN_OP_DISJOINT_OUT] = combine_disjoint_out_u;
-    imp->combine_32[PIXMAN_OP_DISJOINT_OUT_REVERSE] = combine_disjoint_out_reverse_u;
-    imp->combine_32[PIXMAN_OP_DISJOINT_ATOP] = combine_disjoint_atop_u;
-    imp->combine_32[PIXMAN_OP_DISJOINT_ATOP_REVERSE] = combine_disjoint_atop_reverse_u;
-    imp->combine_32[PIXMAN_OP_DISJOINT_XOR] = combine_disjoint_xor_u;
-
-    /* Conjoint, unified */
-    imp->combine_32[PIXMAN_OP_CONJOINT_CLEAR] = combine_clear;
-    imp->combine_32[PIXMAN_OP_CONJOINT_SRC] = combine_src_u;
-    imp->combine_32[PIXMAN_OP_CONJOINT_DST] = combine_dst;
-    imp->combine_32[PIXMAN_OP_CONJOINT_OVER] = combine_conjoint_over_u;
-    imp->combine_32[PIXMAN_OP_CONJOINT_OVER_REVERSE] = combine_conjoint_over_reverse_u;
-    imp->combine_32[PIXMAN_OP_CONJOINT_IN] = combine_conjoint_in_u;
-    imp->combine_32[PIXMAN_OP_CONJOINT_IN_REVERSE] = combine_conjoint_in_reverse_u;
-    imp->combine_32[PIXMAN_OP_CONJOINT_OUT] = combine_conjoint_out_u;
-    imp->combine_32[PIXMAN_OP_CONJOINT_OUT_REVERSE] = combine_conjoint_out_reverse_u;
-    imp->combine_32[PIXMAN_OP_CONJOINT_ATOP] = combine_conjoint_atop_u;
-    imp->combine_32[PIXMAN_OP_CONJOINT_ATOP_REVERSE] = combine_conjoint_atop_reverse_u;
-    imp->combine_32[PIXMAN_OP_CONJOINT_XOR] = combine_conjoint_xor_u;
 
     imp->combine_32[PIXMAN_OP_MULTIPLY] = combine_multiply_u;
     imp->combine_32[PIXMAN_OP_SCREEN] = combine_screen_u;
     imp->combine_32[PIXMAN_OP_OVERLAY] = combine_overlay_u;
     imp->combine_32[PIXMAN_OP_DARKEN] = combine_darken_u;
     imp->combine_32[PIXMAN_OP_LIGHTEN] = combine_lighten_u;
-    imp->combine_32[PIXMAN_OP_COLOR_DODGE] = combine_color_dodge_u;
-    imp->combine_32[PIXMAN_OP_COLOR_BURN] = combine_color_burn_u;
     imp->combine_32[PIXMAN_OP_HARD_LIGHT] = combine_hard_light_u;
-    imp->combine_32[PIXMAN_OP_SOFT_LIGHT] = combine_soft_light_u;
     imp->combine_32[PIXMAN_OP_DIFFERENCE] = combine_difference_u;
     imp->combine_32[PIXMAN_OP_EXCLUSION] = combine_exclusion_u;
-    imp->combine_32[PIXMAN_OP_HSL_HUE] = combine_hsl_hue_u;
-    imp->combine_32[PIXMAN_OP_HSL_SATURATION] = combine_hsl_saturation_u;
-    imp->combine_32[PIXMAN_OP_HSL_COLOR] = combine_hsl_color_u;
-    imp->combine_32[PIXMAN_OP_HSL_LUMINOSITY] = combine_hsl_luminosity_u;
 
     /* Component alpha combiners */
     imp->combine_32_ca[PIXMAN_OP_CLEAR] = combine_clear_ca;
@@ -2531,51 +1177,13 @@ _pixman_setup_combiner_functions_32 (pixman_implementation_t *imp)
     imp->combine_32_ca[PIXMAN_OP_ATOP_REVERSE] = combine_atop_reverse_ca;
     imp->combine_32_ca[PIXMAN_OP_XOR] = combine_xor_ca;
     imp->combine_32_ca[PIXMAN_OP_ADD] = combine_add_ca;
-    imp->combine_32_ca[PIXMAN_OP_SATURATE] = combine_saturate_ca;
-
-    /* Disjoint CA */
-    imp->combine_32_ca[PIXMAN_OP_DISJOINT_CLEAR] = combine_clear_ca;
-    imp->combine_32_ca[PIXMAN_OP_DISJOINT_SRC] = combine_src_ca;
-    imp->combine_32_ca[PIXMAN_OP_DISJOINT_DST] = combine_dst;
-    imp->combine_32_ca[PIXMAN_OP_DISJOINT_OVER] = combine_disjoint_over_ca;
-    imp->combine_32_ca[PIXMAN_OP_DISJOINT_OVER_REVERSE] = combine_saturate_ca;
-    imp->combine_32_ca[PIXMAN_OP_DISJOINT_IN] = combine_disjoint_in_ca;
-    imp->combine_32_ca[PIXMAN_OP_DISJOINT_IN_REVERSE] = combine_disjoint_in_reverse_ca;
-    imp->combine_32_ca[PIXMAN_OP_DISJOINT_OUT] = combine_disjoint_out_ca;
-    imp->combine_32_ca[PIXMAN_OP_DISJOINT_OUT_REVERSE] = combine_disjoint_out_reverse_ca;
-    imp->combine_32_ca[PIXMAN_OP_DISJOINT_ATOP] = combine_disjoint_atop_ca;
-    imp->combine_32_ca[PIXMAN_OP_DISJOINT_ATOP_REVERSE] = combine_disjoint_atop_reverse_ca;
-    imp->combine_32_ca[PIXMAN_OP_DISJOINT_XOR] = combine_disjoint_xor_ca;
-
-    /* Conjoint CA */
-    imp->combine_32_ca[PIXMAN_OP_CONJOINT_CLEAR] = combine_clear_ca;
-    imp->combine_32_ca[PIXMAN_OP_CONJOINT_SRC] = combine_src_ca;
-    imp->combine_32_ca[PIXMAN_OP_CONJOINT_DST] = combine_dst;
-    imp->combine_32_ca[PIXMAN_OP_CONJOINT_OVER] = combine_conjoint_over_ca;
-    imp->combine_32_ca[PIXMAN_OP_CONJOINT_OVER_REVERSE] = combine_conjoint_over_reverse_ca;
-    imp->combine_32_ca[PIXMAN_OP_CONJOINT_IN] = combine_conjoint_in_ca;
-    imp->combine_32_ca[PIXMAN_OP_CONJOINT_IN_REVERSE] = combine_conjoint_in_reverse_ca;
-    imp->combine_32_ca[PIXMAN_OP_CONJOINT_OUT] = combine_conjoint_out_ca;
-    imp->combine_32_ca[PIXMAN_OP_CONJOINT_OUT_REVERSE] = combine_conjoint_out_reverse_ca;
-    imp->combine_32_ca[PIXMAN_OP_CONJOINT_ATOP] = combine_conjoint_atop_ca;
-    imp->combine_32_ca[PIXMAN_OP_CONJOINT_ATOP_REVERSE] = combine_conjoint_atop_reverse_ca;
-    imp->combine_32_ca[PIXMAN_OP_CONJOINT_XOR] = combine_conjoint_xor_ca;
 
     imp->combine_32_ca[PIXMAN_OP_MULTIPLY] = combine_multiply_ca;
     imp->combine_32_ca[PIXMAN_OP_SCREEN] = combine_screen_ca;
     imp->combine_32_ca[PIXMAN_OP_OVERLAY] = combine_overlay_ca;
     imp->combine_32_ca[PIXMAN_OP_DARKEN] = combine_darken_ca;
     imp->combine_32_ca[PIXMAN_OP_LIGHTEN] = combine_lighten_ca;
-    imp->combine_32_ca[PIXMAN_OP_COLOR_DODGE] = combine_color_dodge_ca;
-    imp->combine_32_ca[PIXMAN_OP_COLOR_BURN] = combine_color_burn_ca;
     imp->combine_32_ca[PIXMAN_OP_HARD_LIGHT] = combine_hard_light_ca;
-    imp->combine_32_ca[PIXMAN_OP_SOFT_LIGHT] = combine_soft_light_ca;
     imp->combine_32_ca[PIXMAN_OP_DIFFERENCE] = combine_difference_ca;
     imp->combine_32_ca[PIXMAN_OP_EXCLUSION] = combine_exclusion_ca;
-
-    /* It is not clear that these make sense, so make them noops for now */
-    imp->combine_32_ca[PIXMAN_OP_HSL_HUE] = combine_dst;
-    imp->combine_32_ca[PIXMAN_OP_HSL_SATURATION] = combine_dst;
-    imp->combine_32_ca[PIXMAN_OP_HSL_COLOR] = combine_dst;
-    imp->combine_32_ca[PIXMAN_OP_HSL_LUMINOSITY] = combine_dst;
 }
diff --git a/pixman/pixman/pixman-general.c b/pixman/pixman/pixman-general.c
index 8bce7c006..7cdea2956 100644
--- a/pixman/pixman/pixman-general.c
+++ b/pixman/pixman/pixman-general.c
@@ -109,6 +109,20 @@ static const op_info_t op_flags[PIXMAN_N_OPERATORS] =
 
 #define SCANLINE_BUFFER_LENGTH 8192
 
+static pixman_bool_t
+operator_needs_division (pixman_op_t op)
+{
+    static const uint8_t needs_division[] =
+    {
+	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, /* SATURATE */
+	1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, /* DISJOINT */
+	1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, /* CONJOINT */
+	0, 0, 0, 0, 0, 1, 1, 0, 1, 0, 0, 1, 1, 1, 1, 0, /* blend ops */
+    };
+
+    return needs_division[op];
+}
+
 static void
 general_composite_rect  (pixman_implementation_t *imp,
                          pixman_composite_info_t *info)
@@ -124,9 +138,10 @@ general_composite_rect  (pixman_implementation_t *imp,
     int Bpp;
     int i;
 
-    if ((src_image->common.flags & FAST_PATH_NARROW_FORMAT)		    &&
-	(!mask_image || mask_image->common.flags & FAST_PATH_NARROW_FORMAT) &&
-	(dest_image->common.flags & FAST_PATH_NARROW_FORMAT))
+    if ((src_image->common.flags & FAST_PATH_NARROW_FORMAT)		     &&
+	(!mask_image || mask_image->common.flags & FAST_PATH_NARROW_FORMAT)  &&
+	(dest_image->common.flags & FAST_PATH_NARROW_FORMAT)		     &&
+	!(operator_needs_division (op)))
     {
 	width_flag = ITER_NARROW;
 	Bpp = 4;