Add pixman-combine-float.c

This file contains floating point implementations of combiners for all
pixman operators. These combiners operate on buffers containing single
precision floating point pixels stored in (a, r, g, b) order.

The combiners are added to the pixman_implementation_t struct, but
nothing uses them yet.

This commit incorporates a number of bug fixes contributed by Andrea
Canciani.

4 files changed, 968 insertions, 0 deletions

diff --git a/pixman/Makefile.sources b/pixman/Makefile.sources
index cf7040f3..96540ec1 100644
--- a/pixman/Makefile.sources
+++ b/pixman/Makefile.sources
@@ -5,6 +5,7 @@ libpixman_sources =			\
 	pixman-bits-image.c		\
 	pixman-combine32.c		\
 	pixman-combine64.c		\
+	pixman-combine-float.c		\
 	pixman-conical-gradient.c	\
 	pixman-x86.c			\
 	pixman-mips.c			\
diff --git a/pixman/pixman-combine-float.c b/pixman/pixman-combine-float.c
new file mode 100644
index 00000000..9617e240
--- /dev/null
+++ b/pixman/pixman-combine-float.c
@@ -0,0 +1,956 @@
+/* -*- Mode: c; c-basic-offset: 4; tab-width: 8; indent-tabs-mode: t; -*- */
+/*
+ * Copyright © 2010, 2012 Soren Sandmann Pedersen
+ * Copyright © 2010, 2012 Red Hat, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ * Author: Soren Sandmann Pedersen (sandmann@cs.au.dk)
+ */
+
+#ifdef HAVE_CONFIG_H
+#include <config.h>
+#endif
+
+#include <math.h>
+#include <string.h>
+#include <float.h>
+
+#include "pixman-private.h"
+
+typedef float (* combine_channel_t) (float sa, float s, float da, float d);
+
+static force_inline void
+combine_inner (pixman_bool_t component,
+	       float *dest, const float *src, const float *mask, int n_pixels,
+	       combine_channel_t combine_a, combine_channel_t combine_c)
+{
+    int i;
+
+    if (!mask)
+    {
+	for (i = 0; i < 4 * n_pixels; i += 4)
+	{
+	    float sa = src[i + 0];
+	    float sr = src[i + 1];
+	    float sg = src[i + 2];
+	    float sb = src[i + 3];
+	    
+	    float da = dest[i + 0];
+	    float dr = dest[i + 1];
+	    float dg = dest[i + 2];
+	    float db = dest[i + 3];					
+	    
+	    dest[i + 0] = combine_a (sa, sa, da, da);
+	    dest[i + 1] = combine_c (sa, sr, da, dr);
+	    dest[i + 2] = combine_c (sa, sg, da, dg);
+	    dest[i + 3] = combine_c (sa, sb, da, db);
+	}
+    }
+    else
+    {
+	for (i = 0; i < 4 * n_pixels; i += 4)
+	{
+	    float sa, sr, sg, sb;
+	    float ma, mr, mg, mb;
+	    float da, dr, dg, db;
+	    
+	    sa = src[i + 0];
+	    sr = src[i + 1];
+	    sg = src[i + 2];
+	    sb = src[i + 3];
+	    
+	    if (component)
+	    {
+		ma = mask[i + 0];
+		mr = mask[i + 1];
+		mg = mask[i + 2];
+		mb = mask[i + 3];
+
+		sr *= mr;
+		sg *= mg;
+		sb *= mb;
+
+		ma *= sa;
+		mr *= sa;
+		mg *= sa;
+		mb *= sa;
+		
+		sa = ma;
+	    }
+	    else
+	    {
+		ma = mask[i + 0];
+
+		sa *= ma;
+		sr *= ma;
+		sg *= ma;
+		sb *= ma;
+
+		ma = mr = mg = mb = sa;
+	    }
+	    
+	    da = dest[i + 0];
+	    dr = dest[i + 1];
+	    dg = dest[i + 2];
+	    db = dest[i + 3];
+	    
+	    dest[i + 0] = combine_a (ma, sa, da, da);
+	    dest[i + 1] = combine_c (mr, sr, da, dr);
+	    dest[i + 2] = combine_c (mg, sg, da, dg);
+	    dest[i + 3] = combine_c (mb, sb, da, db);
+	}
+    }
+}
+
+#define MAKE_COMBINER(name, component, combine_a, combine_c)		\
+    static void								\
+    combine_ ## name ## _float (pixman_implementation_t *imp,		\
+				pixman_op_t              op,		\
+				float                   *dest,		\
+				const float             *src,		\
+				const float             *mask,		\
+				int		         n_pixels)	\
+    {									\
+	combine_inner (component, dest, src, mask, n_pixels,		\
+		       combine_a, combine_c);				\
+    }
+
+#define MAKE_COMBINERS(name, combine_a, combine_c)			\
+    MAKE_COMBINER(name ## _ca, TRUE, combine_a, combine_c)		\
+    MAKE_COMBINER(name ## _u, FALSE, combine_a, combine_c)
+
+
+/*
+ * Porter/Duff operators
+ */
+typedef enum
+{
+    ZERO,
+    ONE,
+    SRC_ALPHA,
+    DEST_ALPHA,
+    INV_SA,
+    INV_DA,
+    SA_OVER_DA,
+    DA_OVER_SA,
+    INV_SA_OVER_DA,
+    INV_DA_OVER_SA,
+    ONE_MINUS_SA_OVER_DA,
+    ONE_MINUS_DA_OVER_SA,
+    ONE_MINUS_INV_DA_OVER_SA,
+    ONE_MINUS_INV_SA_OVER_DA
+} combine_factor_t;
+
+static force_inline float
+get_factor (combine_factor_t factor, float sa, float da)
+{
+    switch (factor)
+    {
+    case ZERO:
+	return 0.0f;
+
+    case ONE:
+	return 1.0f;
+
+    case SRC_ALPHA:
+	return sa;
+
+    case DEST_ALPHA:
+	return da;
+
+    case INV_SA:
+	return 1 - sa;
+
+    case INV_DA:
+	return 1 - da;
+
+    case SA_OVER_DA:
+	return da <= sa ? 1.0f : sa / da;
+
+    case DA_OVER_SA:
+	return sa <= da ? 1.0f : da / sa;
+
+    case INV_SA_OVER_DA:
+	return da <= 1.0f - sa ? 1.0f : (1.0f - sa) / da;
+
+    case INV_DA_OVER_SA:
+	return sa <= 1.0f - da ? 1.0f : (1.0f - da) / sa;
+
+    case ONE_MINUS_SA_OVER_DA:
+	return da <= sa ? 0.0f : 1.0f - sa / da;
+
+    case ONE_MINUS_DA_OVER_SA:
+	return sa <= da ? 0.0f : 1.0f - da / sa;
+
+    case ONE_MINUS_INV_DA_OVER_SA:
+	return sa <= 1.0f - da ? 0.0f : 1.0f - (1.0f - da) / sa;
+
+    case ONE_MINUS_INV_SA_OVER_DA:
+	return da <= 1.0f - sa ? 0.0f : 1.0f - (1.0f - sa) / da;
+    }
+
+    return -1;
+}
+
+#define MAKE_PD_COMBINERS(name, a, b)					\
+    static float force_inline						\
+    pd_combine_ ## name (float sa, float s, float da, float d)		\
+    {									\
+	const float fa = get_factor (a, sa, da);			\
+	const float fb = get_factor (b, sa, da);			\
+									\
+	return MIN (1.0f, s * fa + d * fb);				\
+    }									\
+    									\
+    MAKE_COMBINERS(name, pd_combine_ ## name, pd_combine_ ## name)
+
+MAKE_PD_COMBINERS (clear,			ZERO,				ZERO)
+MAKE_PD_COMBINERS (src,				ONE,				ZERO)
+MAKE_PD_COMBINERS (dst,				ZERO,				ONE)
+MAKE_PD_COMBINERS (over,			ONE,				INV_SA)
+MAKE_PD_COMBINERS (over_reverse,		INV_DA,				ONE)
+MAKE_PD_COMBINERS (in,				DEST_ALPHA,			ZERO)
+MAKE_PD_COMBINERS (in_reverse,			ZERO,				SRC_ALPHA)
+MAKE_PD_COMBINERS (out,				INV_DA,				ZERO)
+MAKE_PD_COMBINERS (out_reverse,			ZERO,				INV_SA)
+MAKE_PD_COMBINERS (atop,			DEST_ALPHA,			INV_SA)
+MAKE_PD_COMBINERS (atop_reverse,		INV_DA,				SRC_ALPHA)
+MAKE_PD_COMBINERS (xor,				INV_DA,				INV_SA)
+MAKE_PD_COMBINERS (add,				ONE,				ONE)
+
+MAKE_PD_COMBINERS (saturate,			INV_DA_OVER_SA,			ONE)
+
+MAKE_PD_COMBINERS (disjoint_clear,		ZERO,				ZERO)
+MAKE_PD_COMBINERS (disjoint_src,		ONE,				ZERO)
+MAKE_PD_COMBINERS (disjoint_dst,		ZERO,				ONE)
+MAKE_PD_COMBINERS (disjoint_over,		ONE,				INV_SA_OVER_DA)
+MAKE_PD_COMBINERS (disjoint_over_reverse,	INV_DA_OVER_SA,			ONE)
+MAKE_PD_COMBINERS (disjoint_in,			ONE_MINUS_INV_DA_OVER_SA,	ZERO)
+MAKE_PD_COMBINERS (disjoint_in_reverse,		ZERO,				ONE_MINUS_INV_SA_OVER_DA)
+MAKE_PD_COMBINERS (disjoint_out,		INV_DA_OVER_SA,			ZERO)
+MAKE_PD_COMBINERS (disjoint_out_reverse,	ZERO,				INV_SA_OVER_DA)
+MAKE_PD_COMBINERS (disjoint_atop,		ONE_MINUS_INV_DA_OVER_SA,	INV_SA_OVER_DA)
+MAKE_PD_COMBINERS (disjoint_atop_reverse,	INV_DA_OVER_SA,			ONE_MINUS_INV_SA_OVER_DA)
+MAKE_PD_COMBINERS (disjoint_xor,		INV_DA_OVER_SA,			INV_SA_OVER_DA)
+
+MAKE_PD_COMBINERS (conjoint_clear,		ZERO,				ZERO)
+MAKE_PD_COMBINERS (conjoint_src,		ONE,				ZERO)
+MAKE_PD_COMBINERS (conjoint_dst,		ZERO,				ONE)
+MAKE_PD_COMBINERS (conjoint_over,		ONE,				ONE_MINUS_SA_OVER_DA)
+MAKE_PD_COMBINERS (conjoint_over_reverse,	ONE_MINUS_DA_OVER_SA,		ONE)
+MAKE_PD_COMBINERS (conjoint_in,			DA_OVER_SA,			ZERO)
+MAKE_PD_COMBINERS (conjoint_in_reverse,		ZERO,				SA_OVER_DA)
+MAKE_PD_COMBINERS (conjoint_out,		ONE_MINUS_DA_OVER_SA,		ZERO)
+MAKE_PD_COMBINERS (conjoint_out_reverse,	ZERO,				ONE_MINUS_SA_OVER_DA)
+MAKE_PD_COMBINERS (conjoint_atop,		DA_OVER_SA,			ONE_MINUS_SA_OVER_DA)
+MAKE_PD_COMBINERS (conjoint_atop_reverse,	ONE_MINUS_DA_OVER_SA,		SA_OVER_DA)
+MAKE_PD_COMBINERS (conjoint_xor,		ONE_MINUS_DA_OVER_SA,		ONE_MINUS_SA_OVER_DA)
+
+/*
+ * PDF blend modes:
+ *
+ * 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.
+ * 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)
+ */
+
+#define MAKE_SEPARABLE_PDF_COMBINERS(name)				\
+    static force_inline float						\
+    combine_ ## name ## _a (float sa, float s, float da, float d)	\
+    {									\
+	return da + sa - da * sa;					\
+    }									\
+    									\
+    static force_inline float						\
+    combine_ ## name ## _c (float sa, float s, float da, float d)	\
+    {									\
+	float f = (1 - sa) * d + (1 - da) * s;				\
+									\
+	return f + blend_ ## name (sa, s, da, d);			\
+    }									\
+    									\
+    MAKE_COMBINERS (name, combine_ ## name ## _a, combine_ ## name ## _c)
+
+static force_inline float
+blend_multiply (float sa, float s, float da, float d)
+{
+    return d * s;
+}
+
+static force_inline float
+blend_screen (float sa, float s, float da, float d)
+{
+    return d * sa + s * da - s * d;
+}
+
+static force_inline float
+blend_overlay (float sa, float s, float da, float d)
+{
+    if (2 * d < da)
+	return 2 * s * d;
+    else
+	return sa * da - 2 * (da - d) * (sa - s);
+}
+
+static force_inline float
+blend_darken (float sa, float s, float da, float d)
+{
+    s = s * da;
+    d = d * sa;
+
+    if (s > d)
+	return d;
+    else
+	return s;
+}
+
+static force_inline float
+blend_lighten (float sa, float s, float da, float d)
+{
+    s = s * da;
+    d = d * sa;
+
+    if (s > d)
+	return s;
+    else
+	return d;
+}
+
+static force_inline float
+blend_color_dodge (float sa, float s, float da, float d)
+{
+    if (d == 0)
+	return 0.0f;
+    else if (d * sa >= sa * da - s * da)
+	return sa * da;
+    else
+	return sa * sa * d / (sa - s);
+}
+
+static force_inline float
+blend_color_burn (float sa, float s, float da, float d)
+{
+    if (d >= da)
+	return sa * da;
+    else if (sa * da - d * sa >= s * da)
+	return 0.0f;
+    else
+	return sa * (da - sa * (da - d) / s);
+}
+
+static force_inline float
+blend_hard_light (float sa, float s, float da, float d)
+{
+    if (2 * s < sa)
+	return 2 * s * d;
+    else
+	return sa * da - 2 * (da - d) * (sa - s);
+}
+
+#define IS_ZERO(f) ((f) == 0.0f) /* FIXME? */
+
+static force_inline float
+blend_soft_light (float sa, float s, float da, float d)
+{
+    if (2 * s < sa)
+    {
+	if (IS_ZERO (da))
+	    return d * sa;
+	else
+	    return d * sa - d * (da - d) * (sa - 2 * s) / da;
+    }
+    else if (IS_ZERO (da))
+    {
+	return 0.0f;
+    }
+    else if (4 * d <= da)
+    {
+	return d * sa + (2 * s - sa) * d * ((16 * d / da - 12) * d / da + 3);
+    }
+    else
+    {
+	return d * sa + (sqrtf (d * da) - d) * (2 * s - sa);
+    }
+}
+
+static force_inline float
+blend_difference (float sa, float s, float da, float d)
+{
+    float dsa = d * sa;
+    float sda = s * da;
+
+    if (sda < dsa)
+	return dsa - sda;
+    else
+	return sda - dsa;
+}
+
+static force_inline float
+blend_exclusion (float sa, float s, float da, float d)
+{
+    return s * da + d * sa - 2 * d * s;
+}
+
+MAKE_SEPARABLE_PDF_COMBINERS (multiply)
+MAKE_SEPARABLE_PDF_COMBINERS (screen)
+MAKE_SEPARABLE_PDF_COMBINERS (overlay)
+MAKE_SEPARABLE_PDF_COMBINERS (darken)
+MAKE_SEPARABLE_PDF_COMBINERS (lighten)
+MAKE_SEPARABLE_PDF_COMBINERS (color_dodge)
+MAKE_SEPARABLE_PDF_COMBINERS (color_burn)
+MAKE_SEPARABLE_PDF_COMBINERS (hard_light)
+MAKE_SEPARABLE_PDF_COMBINERS (soft_light)
+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.
+ *
+ * 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, ra);
+ *
+ * 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 - LUM (r × 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 they 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)
+ *
+ */
+
+typedef struct
+{
+    float	r;
+    float	g;
+    float	b;
+} rgb_t;
+
+static force_inline float
+minf (float a, float b)
+{
+    return a < b? a : b;
+}
+
+static force_inline float
+maxf (float a, float b)
+{
+    return a > b? a : b;
+}
+
+static force_inline float
+channel_min (rgb_t *c)
+{
+    return minf (minf (c->r, c->g), c->b);
+}
+
+static force_inline float
+channel_max (rgb_t *c)
+{
+    return maxf (maxf (c->r, c->g), c->b);
+}
+
+static force_inline float
+get_lum (rgb_t *c)
+{
+    return c->r * 0.3f + c->g * 0.59f + c->b * 0.11f;
+}
+
+static force_inline float
+get_sat (rgb_t *c)
+{
+    return channel_max (c) - channel_min (c);
+}
+
+static void
+clip_color (rgb_t *color, float a)
+{
+    float l = get_lum (color);
+    float n = channel_min (color);
+    float x = channel_max (color);
+
+    if (n < 0.0f)
+    {
+	if ((l - n) < 4 * FLT_EPSILON)
+	{
+	    color->r = 0.0f;
+	    color->g = 0.0f;
+	    color->b = 0.0f;
+	}
+	else
+	{
+	    color->r = l + (((color->r - l) * l) / (l - n));
+	    color->g = l + (((color->g - l) * l) / (l - n));
+	    color->b = l + (((color->b - l) * l) / (l - n));
+	}
+    }
+    if (x > a)
+    {
+	if ((x - l) < 4 * FLT_EPSILON)
+	{
+	    color->r = a;
+	    color->g = a;
+	    color->b = a;
+	}
+	else
+	{
+	    color->r = l + (((color->r - l) * (a - l) / (x - l)));
+	    color->g = l + (((color->g - l) * (a - l) / (x - l)));
+	    color->b = l + (((color->b - l) * (a - l) / (x - l)));
+	}
+    }
+}
+
+static void
+set_lum (rgb_t *color, float sa, float l)
+{
+    float d = l - get_lum (color);
+
+    color->r = color->r + d;
+    color->g = color->g + d;
+    color->b = color->b + d;
+
+    clip_color (color, sa);
+}
+
+static void
+set_sat (rgb_t *src, float sat)
+{
+    float *max, *mid, *min;
+
+    if (src->r > src->g)
+    {
+	if (src->r > src->b)
+	{
+	    max = &(src->r);
+
+	    if (src->g > src->b)
+	    {
+		mid = &(src->g);
+		min = &(src->b);
+	    }
+	    else
+	    {
+		mid = &(src->b);
+		min = &(src->g);
+	    }
+	}
+	else
+	{
+	    max = &(src->b);
+	    mid = &(src->r);
+	    min = &(src->g);
+	}
+    }
+    else
+    {
+	if (src->r > src->b)
+	{
+	    max = &(src->g);
+	    mid = &(src->r);
+	    min = &(src->b);
+	}
+	else
+	{
+	    min = &(src->r);
+
+	    if (src->g > src->b)
+	    {
+		max = &(src->g);
+		mid = &(src->b);
+	    }
+	    else
+	    {
+		max = &(src->b);
+		mid = &(src->g);
+	    }
+	}
+    }
+
+    if (*max > *min)
+    {
+	*mid = (((*mid - *min) * sat) / (*max - *min));
+	*max = sat;
+    }
+    else
+    {
+	*mid = *max = 0.0f;
+    }
+
+    *min = 0.0f;
+}
+
+/*
+ * Hue:
+ * B(Cb, Cs) = set_lum (set_sat (Cs, SAT (Cb)), LUM (Cb))
+ */
+static force_inline void
+blend_hsl_hue (rgb_t *res,
+	       rgb_t *dest, float da,
+	       rgb_t *src, float sa)
+{
+    res->r = src->r * da;
+    res->g = src->g * da;
+    res->b = src->b * da;
+
+    set_sat (res, get_sat (dest) * sa);
+    set_lum (res, sa * da, get_lum (dest) * sa);
+}
+
+/*
+ * Saturation:
+ * B(Cb, Cs) = set_lum (set_sat (Cb, SAT (Cs)), LUM (Cb))
+ */
+static force_inline void
+blend_hsl_saturation (rgb_t *res,
+		      rgb_t *dest, float da,
+		      rgb_t *src, float sa)
+{
+    res->r = dest->r * sa;
+    res->g = dest->g * sa;
+    res->b = dest->b * sa;
+
+    set_sat (res, get_sat (src) * da);
+    set_lum (res, sa * da, get_lum (dest) * sa);
+}
+
+/*
+ * Color:
+ * B(Cb, Cs) = set_lum (Cs, LUM (Cb))
+ */
+static force_inline void
+blend_hsl_color (rgb_t *res,
+		 rgb_t *dest, float da,
+		 rgb_t *src, float sa)
+{
+    res->r = src->r * da;
+    res->g = src->g * da;
+    res->b = src->b * da;
+
+    set_lum (res, sa * da, get_lum (dest) * sa);
+}
+
+/*
+ * Luminosity:
+ * B(Cb, Cs) = set_lum (Cb, LUM (Cs))
+ */
+static force_inline void
+blend_hsl_luminosity (rgb_t *res,
+		      rgb_t *dest, float da,
+		      rgb_t *src, float sa)
+{
+    res->r = dest->r * sa;
+    res->g = dest->g * sa;
+    res->b = dest->b * sa;
+
+    set_lum (res, sa * da, get_lum (src) * da);
+}
+
+#define MAKE_NON_SEPARABLE_PDF_COMBINERS(name)				\
+    static void								\
+    combine_ ## name ## _u_float (pixman_implementation_t *imp,		\
+				  pixman_op_t              op,		\
+				  float                   *dest,	\
+				  const float             *src,		\
+				  const float             *mask,	\
+				  int		           n_pixels)	\
+    {									\
+    	int i;								\
+									\
+	for (i = 0; i < 4 * n_pixels; i += 4)				\
+	{								\
+	    float sa, da;						\
+	    rgb_t sc, dc, rc;						\
+									\
+	    sa = src[i + 0];						\
+	    sc.r = src[i + 1];						\
+	    sc.g = src[i + 2];						\
+	    sc.b = src[i + 3];						\
+									\
+	    da = dest[i + 0];						\
+	    dc.r = dest[i + 1];						\
+	    dc.g = dest[i + 2];						\
+	    dc.b = dest[i + 3];						\
+									\
+	    if (mask)							\
+	    {								\
+		float ma = mask[i + 0];					\
+									\
+		/* Component alpha is not supported for HSL modes */	\
+		sa *= ma;						\
+		sc.r *= ma;						\
+		sc.g *= ma;						\
+		sc.g *= ma;						\
+	    }								\
+									\
+	    blend_ ## name (&rc, &dc, da, &sc, sa);			\
+									\
+	    dest[i + 0] = sa + da - sa * da;				\
+	    dest[i + 1] = (1-sa)*dc.r + (1-da)*sc.r + rc.r;		\
+	    dest[i + 2] = (1-sa)*dc.g + (1-da)*sc.g + rc.g;		\
+	    dest[i + 3] = (1-sa)*dc.b + (1-da)*sc.b + rc.b;		\
+	}								\
+    }
+
+MAKE_NON_SEPARABLE_PDF_COMBINERS(hsl_hue)
+MAKE_NON_SEPARABLE_PDF_COMBINERS(hsl_saturation)
+MAKE_NON_SEPARABLE_PDF_COMBINERS(hsl_color)
+MAKE_NON_SEPARABLE_PDF_COMBINERS(hsl_luminosity)
+
+void
+_pixman_setup_combiner_functions_float (pixman_implementation_t *imp)
+{
+    /* Unified alpha */
+    imp->combine_float[PIXMAN_OP_CLEAR] = combine_clear_u_float;
+    imp->combine_float[PIXMAN_OP_SRC] = combine_src_u_float;
+    imp->combine_float[PIXMAN_OP_DST] = combine_dst_u_float;
+    imp->combine_float[PIXMAN_OP_OVER] = combine_over_u_float;
+    imp->combine_float[PIXMAN_OP_OVER_REVERSE] = combine_over_reverse_u_float;
+    imp->combine_float[PIXMAN_OP_IN] = combine_in_u_float;
+    imp->combine_float[PIXMAN_OP_IN_REVERSE] = combine_in_reverse_u_float;
+    imp->combine_float[PIXMAN_OP_OUT] = combine_out_u_float;
+    imp->combine_float[PIXMAN_OP_OUT_REVERSE] = combine_out_reverse_u_float;
+    imp->combine_float[PIXMAN_OP_ATOP] = combine_atop_u_float;
+    imp->combine_float[PIXMAN_OP_ATOP_REVERSE] = combine_atop_reverse_u_float;
+    imp->combine_float[PIXMAN_OP_XOR] = combine_xor_u_float;
+    imp->combine_float[PIXMAN_OP_ADD] = combine_add_u_float;
+    imp->combine_float[PIXMAN_OP_SATURATE] = combine_saturate_u_float;
+
+    /* Disjoint, unified */
+    imp->combine_float[PIXMAN_OP_DISJOINT_CLEAR] = combine_disjoint_clear_u_float;
+    imp->combine_float[PIXMAN_OP_DISJOINT_SRC] = combine_disjoint_src_u_float;
+    imp->combine_float[PIXMAN_OP_DISJOINT_DST] = combine_disjoint_dst_u_float;
+    imp->combine_float[PIXMAN_OP_DISJOINT_OVER] = combine_disjoint_over_u_float;
+    imp->combine_float[PIXMAN_OP_DISJOINT_OVER_REVERSE] = combine_disjoint_over_reverse_u_float;
+    imp->combine_float[PIXMAN_OP_DISJOINT_IN] = combine_disjoint_in_u_float;
+    imp->combine_float[PIXMAN_OP_DISJOINT_IN_REVERSE] = combine_disjoint_in_reverse_u_float;
+    imp->combine_float[PIXMAN_OP_DISJOINT_OUT] = combine_disjoint_out_u_float;
+    imp->combine_float[PIXMAN_OP_DISJOINT_OUT_REVERSE] = combine_disjoint_out_reverse_u_float;
+    imp->combine_float[PIXMAN_OP_DISJOINT_ATOP] = combine_disjoint_atop_u_float;
+    imp->combine_float[PIXMAN_OP_DISJOINT_ATOP_REVERSE] = combine_disjoint_atop_reverse_u_float;
+    imp->combine_float[PIXMAN_OP_DISJOINT_XOR] = combine_disjoint_xor_u_float;
+
+    /* Conjoint, unified */
+    imp->combine_float[PIXMAN_OP_CONJOINT_CLEAR] = combine_conjoint_clear_u_float;
+    imp->combine_float[PIXMAN_OP_CONJOINT_SRC] = combine_conjoint_src_u_float;
+    imp->combine_float[PIXMAN_OP_CONJOINT_DST] = combine_conjoint_dst_u_float;
+    imp->combine_float[PIXMAN_OP_CONJOINT_OVER] = combine_conjoint_over_u_float;
+    imp->combine_float[PIXMAN_OP_CONJOINT_OVER_REVERSE] = combine_conjoint_over_reverse_u_float;
+    imp->combine_float[PIXMAN_OP_CONJOINT_IN] = combine_conjoint_in_u_float;
+    imp->combine_float[PIXMAN_OP_CONJOINT_IN_REVERSE] = combine_conjoint_in_reverse_u_float;
+    imp->combine_float[PIXMAN_OP_CONJOINT_OUT] = combine_conjoint_out_u_float;
+    imp->combine_float[PIXMAN_OP_CONJOINT_OUT_REVERSE] = combine_conjoint_out_reverse_u_float;
+    imp->combine_float[PIXMAN_OP_CONJOINT_ATOP] = combine_conjoint_atop_u_float;
+    imp->combine_float[PIXMAN_OP_CONJOINT_ATOP_REVERSE] = combine_conjoint_atop_reverse_u_float;
+    imp->combine_float[PIXMAN_OP_CONJOINT_XOR] = combine_conjoint_xor_u_float;
+
+    /* PDF operators, unified */
+    imp->combine_float[PIXMAN_OP_MULTIPLY] = combine_multiply_u_float;
+    imp->combine_float[PIXMAN_OP_SCREEN] = combine_screen_u_float;
+    imp->combine_float[PIXMAN_OP_OVERLAY] = combine_overlay_u_float;
+    imp->combine_float[PIXMAN_OP_DARKEN] = combine_darken_u_float;
+    imp->combine_float[PIXMAN_OP_LIGHTEN] = combine_lighten_u_float;
+    imp->combine_float[PIXMAN_OP_COLOR_DODGE] = combine_color_dodge_u_float;
+    imp->combine_float[PIXMAN_OP_COLOR_BURN] = combine_color_burn_u_float;
+    imp->combine_float[PIXMAN_OP_HARD_LIGHT] = combine_hard_light_u_float;
+    imp->combine_float[PIXMAN_OP_SOFT_LIGHT] = combine_soft_light_u_float;
+    imp->combine_float[PIXMAN_OP_DIFFERENCE] = combine_difference_u_float;
+    imp->combine_float[PIXMAN_OP_EXCLUSION] = combine_exclusion_u_float;
+
+    imp->combine_float[PIXMAN_OP_HSL_HUE] = combine_hsl_hue_u_float;
+    imp->combine_float[PIXMAN_OP_HSL_SATURATION] = combine_hsl_saturation_u_float;
+    imp->combine_float[PIXMAN_OP_HSL_COLOR] = combine_hsl_color_u_float;
+    imp->combine_float[PIXMAN_OP_HSL_LUMINOSITY] = combine_hsl_luminosity_u_float;
+
+    /* Component alpha combiners */
+    imp->combine_float_ca[PIXMAN_OP_CLEAR] = combine_clear_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_SRC] = combine_src_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DST] = combine_dst_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_OVER] = combine_over_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_OVER_REVERSE] = combine_over_reverse_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_IN] = combine_in_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_IN_REVERSE] = combine_in_reverse_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_OUT] = combine_out_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_OUT_REVERSE] = combine_out_reverse_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_ATOP] = combine_atop_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_ATOP_REVERSE] = combine_atop_reverse_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_XOR] = combine_xor_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_ADD] = combine_add_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_SATURATE] = combine_saturate_ca_float;
+
+    /* Disjoint CA */
+    imp->combine_float_ca[PIXMAN_OP_DISJOINT_CLEAR] = combine_disjoint_clear_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DISJOINT_SRC] = combine_disjoint_src_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DISJOINT_DST] = combine_disjoint_dst_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DISJOINT_OVER] = combine_disjoint_over_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DISJOINT_OVER_REVERSE] = combine_disjoint_over_reverse_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DISJOINT_IN] = combine_disjoint_in_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DISJOINT_IN_REVERSE] = combine_disjoint_in_reverse_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DISJOINT_OUT] = combine_disjoint_out_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DISJOINT_OUT_REVERSE] = combine_disjoint_out_reverse_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DISJOINT_ATOP] = combine_disjoint_atop_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DISJOINT_ATOP_REVERSE] = combine_disjoint_atop_reverse_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DISJOINT_XOR] = combine_disjoint_xor_ca_float;
+
+    /* Conjoint CA */
+    imp->combine_float_ca[PIXMAN_OP_CONJOINT_CLEAR] = combine_conjoint_clear_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_CONJOINT_SRC] = combine_conjoint_src_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_CONJOINT_DST] = combine_conjoint_dst_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_CONJOINT_OVER] = combine_conjoint_over_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_CONJOINT_OVER_REVERSE] = combine_conjoint_over_reverse_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_CONJOINT_IN] = combine_conjoint_in_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_CONJOINT_IN_REVERSE] = combine_conjoint_in_reverse_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_CONJOINT_OUT] = combine_conjoint_out_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_CONJOINT_OUT_REVERSE] = combine_conjoint_out_reverse_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_CONJOINT_ATOP] = combine_conjoint_atop_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_CONJOINT_ATOP_REVERSE] = combine_conjoint_atop_reverse_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_CONJOINT_XOR] = combine_conjoint_xor_ca_float;
+
+    /* PDF operators CA */
+    imp->combine_float_ca[PIXMAN_OP_MULTIPLY] = combine_multiply_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_SCREEN] = combine_screen_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_OVERLAY] = combine_overlay_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DARKEN] = combine_darken_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_LIGHTEN] = combine_lighten_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_COLOR_DODGE] = combine_color_dodge_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_COLOR_BURN] = combine_color_burn_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_HARD_LIGHT] = combine_hard_light_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_SOFT_LIGHT] = combine_soft_light_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_DIFFERENCE] = combine_difference_ca_float;
+    imp->combine_float_ca[PIXMAN_OP_EXCLUSION] = combine_exclusion_ca_float;
+
+    /* It is not clear that these make sense, so make them noops for now */
+    imp->combine_float_ca[PIXMAN_OP_HSL_HUE] = combine_dst_u_float;
+    imp->combine_float_ca[PIXMAN_OP_HSL_SATURATION] = combine_dst_u_float;
+    imp->combine_float_ca[PIXMAN_OP_HSL_COLOR] = combine_dst_u_float;
+    imp->combine_float_ca[PIXMAN_OP_HSL_LUMINOSITY] = combine_dst_u_float;
+}
diff --git a/pixman/pixman-general.c b/pixman/pixman-general.c
index 42a84a0b..2d920145 100644
--- a/pixman/pixman-general.c
+++ b/pixman/pixman-general.c
@@ -213,6 +213,7 @@ _pixman_implementation_create_general (void)
 
     _pixman_setup_combiner_functions_32 (imp);
     _pixman_setup_combiner_functions_64 (imp);
+    _pixman_setup_combiner_functions_float (imp);
 
     imp->src_iter_init = general_src_iter_init;
     imp->dest_iter_init = general_dest_iter_init;
diff --git a/pixman/pixman-private.h b/pixman/pixman-private.h
index b9c8319d..949d384f 100644
--- a/pixman/pixman-private.h
+++ b/pixman/pixman-private.h
@@ -421,6 +421,13 @@ typedef void (*pixman_combine_64_func_t) (pixman_implementation_t *imp,
 					  const uint64_t *         mask,
 					  int                      width);
 
+typedef void (*pixman_combine_float_func_t) (pixman_implementation_t *imp,
+					     pixman_op_t	      op,
+					     float *		      dest,
+					     const float *	      src,
+					     const float *	      mask,
+					     int		      n_pixels);
+
 typedef void (*pixman_composite_func_t) (pixman_implementation_t *imp,
 					 pixman_composite_info_t *info);
 typedef pixman_bool_t (*pixman_blt_func_t) (pixman_implementation_t *imp,
@@ -450,6 +457,7 @@ typedef pixman_bool_t (*pixman_iter_init_func_t) (pixman_implementation_t *imp,
 
 void _pixman_setup_combiner_functions_32 (pixman_implementation_t *imp);
 void _pixman_setup_combiner_functions_64 (pixman_implementation_t *imp);
+void _pixman_setup_combiner_functions_float (pixman_implementation_t *imp);
 
 typedef struct
 {
@@ -478,6 +486,8 @@ struct pixman_implementation_t
     pixman_combine_32_func_t	combine_32_ca[PIXMAN_N_OPERATORS];
     pixman_combine_64_func_t	combine_64[PIXMAN_N_OPERATORS];
     pixman_combine_64_func_t	combine_64_ca[PIXMAN_N_OPERATORS];
+    pixman_combine_float_func_t	combine_float[PIXMAN_N_OPERATORS];
+    pixman_combine_float_func_t	combine_float_ca[PIXMAN_N_OPERATORS];
 };
 
 uint32_t