/* Libart_LGPL - library of basic graphic primitives
* Copyright (C) 1998 Raph Levien
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Library General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Library General Public License for more details.
*
* You should have received a copy of the GNU Library General Public
* License along with this library; if not, write to the
* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
* Boston, MA 02111-1307, USA.
*/
#include "config.h"
#include "art_rgb.h"
#include <string.h> /* for memset */
/* Basic operators for manipulating 24-bit packed RGB buffers. */
#define COLOR_RUN_COMPLEX
#ifdef COLOR_RUN_SIMPLE
/* This is really slow. Is there any way we might speed it up?
Two ideas:
First, maybe we should be working at 32-bit alignment. Then,
this can be a simple loop over word stores.
Second, we can keep working at 24-bit alignment, but have some
intelligence about storing. For example, we can iterate over
4-pixel chunks (aligned at 4 pixels), with an inner loop
something like:
*buf++ = v1;
*buf++ = v2;
*buf++ = v3;
One source of extra complexity is the need to make sure linebuf is
aligned to a 32-bit boundary.
This second alternative has some complexity to it, but is
appealing because it really minimizes the memory bandwidth. */
void
art_rgb_fill_run (art_u8 *buf, art_u8 r, art_u8 g, art_u8 b, gint n)
{
int i;
if (r == g && g == b)
{
memset (buf, g, n + n + n);
}
else
{
for (i = 0; i < n; i++)
{
*buf++ = r;
*buf++ = g;
*buf++ = b;
}
}
}
#endif
#ifdef COLOR_RUN_COMPLEX
/* This implements the second of the two ideas above. The test results
are _very_ encouraging - it seems the speed is within 10% of
memset, which is quite good! */
/**
* art_rgb_fill_run: fill a buffer a solid RGB color.
* @buf: Buffer to fill.
* @r: Red, range 0..255.
* @g: Green, range 0..255.
* @b: Blue, range 0..255.
* @n: Number of RGB triples to fill.
*
* Fills a buffer with @n copies of the (@r, @g, @b) triple. Thus,
* locations @buf (inclusive) through @buf + 3 * @n (exclusive) are
* written.
*
* The implementation of this routine is very highly optimized.
**/
void
art_rgb_fill_run (art_u8 *buf, art_u8 r, art_u8 g, art_u8 b, int n)
{
int i;
unsigned int v1, v2, v3;
if (r == g && g == b)
{
memset (buf, g, n + n + n);
}
else
{
if (n < 8)
{
for (i = 0; i < n; i++)
{
*buf++ = r;
*buf++ = g;
*buf++ = b;
}
} else {
/* handle prefix up to byte alignment */
/* I'm worried about this cast on sizeof(long) != sizeof(uchar *)
architectures, but it _should_ work. */
for (i = 0; ((unsigned long)buf) & 3; i++)
{
*buf++ = r;
*buf++ = g;
*buf++ = b;
}
#ifndef WORDS_BIGENDIAN
v1 = r | (g << 8) | (b << 16) | (r << 24);
v3 = (v1 << 8) | b;
v2 = (v3 << 8) | g;
#else
v1 = (r << 24) | (g << 16) | (b << 8) | r;
v2 = (v1 << 8) | g;
v3 = (v2 << 8) | b;
#endif
for (; i < n - 3; i += 4)
{
((art_u32 *)buf)[0] = v1;
((art_u32 *)buf)[1] = v2;
((art_u32 *)buf)[2] = v3;
buf += 12;
}
/* handle postfix */
for (; i < n; i++)
{
*buf++ = r;
*buf++ = g;
*buf++ = b;
}
}
}
}
#endif
/**
* art_rgb_run_alpha: Render semitransparent color over RGB buffer.
* @buf: Buffer for rendering.
* @r: Red, range 0..255.
* @g: Green, range 0..255.
* @b: Blue, range 0..255.
* @alpha: Alpha, range 0..256.
* @n: Number of RGB triples to render.
*
* Renders a sequential run of solid (@r, @g, @b) color over @buf with
* opacity @alpha.
**/
void
art_rgb_run_alpha (art_u8 *buf, art_u8 r, art_u8 g, art_u8 b, int alpha, int n)
{
int i;
int v;
for (i = 0; i < n; i++)
{
v = *buf;
*buf++ = v + (((r - v) * alpha + 0x80) >> 8);
v = *buf;
*buf++ = v + (((g - v) * alpha + 0x80) >> 8);
v = *buf;
*buf++ = v + (((b - v) * alpha + 0x80) >> 8);
}
}