lib/art/art_rgb.c

   1 /* Libart_LGPL - library of basic graphic primitives
   2  * Copyright (C) 1998 Raph Levien
   3  *
   4  * This library is free software; you can redistribute it and/or
   5  * modify it under the terms of the GNU Library General Public
   6  * License as published by the Free Software Foundation; either
   7  * version 2 of the License, or (at your option) any later version.
   8  *
   9  * This library is distributed in the hope that it will be useful,
  10  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  11  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  12  * Library General Public License for more details.
  13  *
  14  * You should have received a copy of the GNU Library General Public
  15  * License along with this library; if not, write to the
  16  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  17  * Boston, MA 02111-1307, USA.
  18  */
  19
  20 #include "config.h"
  21 #include "art_rgb.h"
  22
  23 #include <string.h>     /* for memset */
  24
  25 /* Basic operators for manipulating 24-bit packed RGB buffers. */
  26
  27 #define COLOR_RUN_COMPLEX
  28
  29 #ifdef COLOR_RUN_SIMPLE
  30 /* This is really slow. Is there any way we might speed it up?
  31    Two ideas:
  32
  33    First, maybe we should be working at 32-bit alignment. Then,
  34    this can be a simple loop over word stores.
  35
  36    Second, we can keep working at 24-bit alignment, but have some
  37    intelligence about storing. For example, we can iterate over
  38    4-pixel chunks (aligned at 4 pixels), with an inner loop
  39    something like:
  40
  41    *buf++ = v1;
  42    *buf++ = v2;
  43    *buf++ = v3;
  44
  45    One source of extra complexity is the need to make sure linebuf is
  46    aligned to a 32-bit boundary.
  47
  48    This second alternative has some complexity to it, but is
  49    appealing because it really minimizes the memory bandwidth. */
  50 void
  51 art_rgb_fill_run (art_u8 *buf, art_u8 r, art_u8 g, art_u8 b, gint n)
  52 {
  53   int i;
  54
  55   if (r == g && g == b)
  56     {
  57       memset (buf, g, n + n + n);
  58     }
  59   else
  60     {
  61       for (i = 0; i < n; i++)
  62         {
  63           *buf++ = r;
  64           *buf++ = g;
  65           *buf++ = b;
  66         }
  67     }
  68 }
  69 #endif
  70
  71 #ifdef COLOR_RUN_COMPLEX
  72 /* This implements the second of the two ideas above. The test results
  73    are _very_ encouraging - it seems the speed is within 10% of
  74    memset, which is quite good! */
  75 /**
  76  * art_rgb_fill_run: fill a buffer a solid RGB color.
  77  * @buf: Buffer to fill.
  78  * @r: Red, range 0..255.
  79  * @g: Green, range 0..255.
  80  * @b: Blue, range 0..255.
  81  * @n: Number of RGB triples to fill.
  82  *
  83  * Fills a buffer with @n copies of the (@r, @g, @b) triple. Thus,
  84  * locations @buf (inclusive) through @buf + 3 * @n (exclusive) are
  85  * written.
  86  *
  87  * The implementation of this routine is very highly optimized.
  88  **/
  89 void
  90 art_rgb_fill_run (art_u8 *buf, art_u8 r, art_u8 g, art_u8 b, int n)
  91 {
  92   int i;
  93   unsigned int v1, v2, v3;
  94
  95   if (r == g && g == b)
  96     {
  97       memset (buf, g, n + n + n);
  98     }
  99   else
 100     {
 101       if (n < 8)
 102         {
 103           for (i = 0; i < n; i++)
 104             {
 105               *buf++ = r;
 106               *buf++ = g;
 107               *buf++ = b;
 108             }
 109         } else {
 110           /* handle prefix up to byte alignment */
 111           /* I'm worried about this cast on sizeof(long) != sizeof(uchar *)
 112              architectures, but it _should_ work. */
 113           for (i = 0; ((unsigned long)buf) & 3; i++)
 114             {
 115               *buf++ = r;
 116               *buf++ = g;
 117               *buf++ = b;
 118             }
 119 #ifndef WORDS_BIGENDIAN
 120           v1 = r | (g << 8) | (b << 16) | (r << 24);
 121           v3 = (v1 << 8) | b;
 122           v2 = (v3 << 8) | g;
 123 #else
 124           v1 = (r << 24) | (g << 16) | (b << 8) | r;
 125           v2 = (v1 << 8) | g;
 126           v3 = (v2 << 8) | b;
 127 #endif
 128           for (; i < n - 3; i += 4)
 129             {
 130               ((art_u32 *)buf)[0] = v1;
 131               ((art_u32 *)buf)[1] = v2;
 132               ((art_u32 *)buf)[2] = v3;
 133               buf += 12;
 134             }
 135           /* handle postfix */
 136           for (; i < n; i++)
 137             {
 138               *buf++ = r;
 139               *buf++ = g;
 140               *buf++ = b;
 141             }
 142         }
 143     }
 144 }
 145 #endif
 146
 147 /**
 148  * art_rgb_run_alpha: Render semitransparent color over RGB buffer.
 149  * @buf: Buffer for rendering.
 150  * @r: Red, range 0..255.
 151  * @g: Green, range 0..255.
 152  * @b: Blue, range 0..255.
 153  * @alpha: Alpha, range 0..256.
 154  * @n: Number of RGB triples to render.
 155  *
 156  * Renders a sequential run of solid (@r, @g, @b) color over @buf with
 157  * opacity @alpha.
 158  **/
 159 void
 160 art_rgb_run_alpha (art_u8 *buf, art_u8 r, art_u8 g, art_u8 b, int alpha, int n)
 161 {
 162   int i;
 163   int v;
 164
 165   for (i = 0; i < n; i++)
 166     {
 167       v = *buf;
 168       *buf++ = v + (((r - v) * alpha + 0x80) >> 8);
 169       v = *buf;
 170       *buf++ = v + (((g - v) * alpha + 0x80) >> 8);
 171       v = *buf;
 172       *buf++ = v + (((b - v) * alpha + 0x80) >> 8);
 173     }
 174 }
 175