m68k: Atari fb revival

Update the atari fb to 2.6 by Michael Schmitz,
Reformatting and rewrite of bit plane functions by Roman Zippel,
A few more fixes by Geert Uytterhoeven.
Signed-off-by: Michael Schmitz <schmitz@debian.org>
Signed-off-by: Roman Zippel <zippel@linux-m68k.org>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/video/Kconfig

@@ -389,7 +389,10 @@ config FB_ARC
 
 config FB_ATARI
 	bool "Atari native chipset support"
-	depends on (FB = y) && ATARI && BROKEN
+	depends on (FB = y) && ATARI
+	select FB_CFB_FILLRECT
+	select FB_CFB_COPYAREA
+	select FB_CFB_IMAGEBLIT
 	help
 	  This is the frame buffer device driver for the builtin graphics
 	  chipset found in Ataris.

drivers/video/Makefile

@@ -63,7 +63,8 @@ obj-$(CONFIG_FB_TCX)              += tcx.o sbuslib.o
 obj-$(CONFIG_FB_LEO)              += leo.o sbuslib.o
 obj-$(CONFIG_FB_SGIVW)            += sgivwfb.o
 obj-$(CONFIG_FB_ACORN)            += acornfb.o
-obj-$(CONFIG_FB_ATARI)            += atafb.o
+obj-$(CONFIG_FB_ATARI)            += atafb.o c2p.o atafb_mfb.o \
+                                     atafb_iplan2p2.o atafb_iplan2p4.o atafb_iplan2p8.o
 obj-$(CONFIG_FB_MAC)              += macfb.o
 obj-$(CONFIG_FB_HGA)              += hgafb.o
 obj-$(CONFIG_FB_IGA)              += igafb.o

drivers/video/atafb.h

+#ifndef _VIDEO_ATAFB_H
+#define _VIDEO_ATAFB_H
+
+void atafb_mfb_copyarea(struct fb_info *info, u_long next_line, int sy, int sx, int dy,
+			int dx, int height, int width);
+void atafb_mfb_fillrect(struct fb_info *info, u_long next_line, u32 color,
+			int sy, int sx, int height, int width);
+void atafb_mfb_linefill(struct fb_info *info, u_long next_line,
+			int dy, int dx, u32 width,
+			const u8 *data, u32 bgcolor, u32 fgcolor);
+
+void atafb_iplan2p2_copyarea(struct fb_info *info, u_long next_line, int sy, int sx, int dy,
+			     int dx, int height, int width);
+void atafb_iplan2p2_fillrect(struct fb_info *info, u_long next_line, u32 color,
+			     int sy, int sx, int height, int width);
+void atafb_iplan2p2_linefill(struct fb_info *info, u_long next_line,
+			     int dy, int dx, u32 width,
+			     const u8 *data, u32 bgcolor, u32 fgcolor);
+
+void atafb_iplan2p4_copyarea(struct fb_info *info, u_long next_line, int sy, int sx, int dy,
+			     int dx, int height, int width);
+void atafb_iplan2p4_fillrect(struct fb_info *info, u_long next_line, u32 color,
+			     int sy, int sx, int height, int width);
+void atafb_iplan2p4_linefill(struct fb_info *info, u_long next_line,
+			     int dy, int dx, u32 width,
+			     const u8 *data, u32 bgcolor, u32 fgcolor);
+
+void atafb_iplan2p8_copyarea(struct fb_info *info, u_long next_line, int sy, int sx, int dy,
+			     int dx, int height, int width);
+void atafb_iplan2p8_fillrect(struct fb_info *info, u_long next_line, u32 color,
+			     int sy, int sx, int height, int width);
+void atafb_iplan2p8_linefill(struct fb_info *info, u_long next_line,
+			     int dy, int dx, u32 width,
+			     const u8 *data, u32 bgcolor, u32 fgcolor);
+
+#endif /* _VIDEO_ATAFB_H */

drivers/video/atafb_iplan2p2.c

+/*
+ *  linux/drivers/video/iplan2p2.c -- Low level frame buffer operations for
+ *				      interleaved bitplanes à la Atari (2
+ *				      planes, 2 bytes interleave)
+ *
+ *	Created 5 Apr 1997 by Geert Uytterhoeven
+ *
+ *  This file is subject to the terms and conditions of the GNU General Public
+ *  License.  See the file COPYING in the main directory of this archive for
+ *  more details.
+ */
+
+#include <linux/module.h>
+#include <linux/string.h>
+#include <linux/fb.h>
+
+#include <asm/setup.h>
+
+#include "atafb.h"
+
+#define BPL	2
+#include "atafb_utils.h"
+
+void atafb_iplan2p2_copyarea(struct fb_info *info, u_long next_line,
+			     int sy, int sx, int dy, int dx,
+			     int height, int width)
+{
+	/*  bmove() has to distinguish two major cases: If both, source and
+	 *  destination, start at even addresses or both are at odd
+	 *  addresses, just the first odd and last even column (if present)
+	 *  require special treatment (memmove_col()). The rest between
+	 *  then can be copied by normal operations, because all adjacent
+	 *  bytes are affected and are to be stored in the same order.
+	 *    The pathological case is when the move should go from an odd
+	 *  address to an even or vice versa. Since the bytes in the plane
+	 *  words must be assembled in new order, it seems wisest to make
+	 *  all movements by memmove_col().
+	 */
+
+	u8 *src, *dst;
+	u32 *s, *d;
+	int w, l , i, j;
+	u_int colsize;
+	u_int upwards = (dy < sy) || (dy == sy && dx < sx);
+
+	colsize = height;
+	if (!((sx ^ dx) & 15)) {
+		/* odd->odd or even->even */
+
+		if (upwards) {
+			src = (u8 *)info->screen_base + sy * next_line + (sx & ~15) / (8 / BPL);
+			dst = (u8 *)info->screen_base + dy * next_line + (dx & ~15) / (8 / BPL);
+			if (sx & 15) {
+				memmove32_col(dst, src, 0xff00ff, height, next_line - BPL * 2);
+				src += BPL * 2;
+				dst += BPL * 2;
+				width -= 8;
+			}
+			w = width >> 4;
+			if (w) {
+				s = (u32 *)src;
+				d = (u32 *)dst;
+				w *= BPL / 2;
+				l = next_line - w * 4;
+				for (j = height; j > 0; j--) {
+					for (i = w; i > 0; i--)
+						*d++ = *s++;
+					s = (u32 *)((u8 *)s + l);
+					d = (u32 *)((u8 *)d + l);
+				}
+			}
+			if (width & 15)
+				memmove32_col(dst + width / (8 / BPL), src + width / (8 / BPL),
+					      0xff00ff00, height, next_line - BPL * 2);
+		} else {
+			src = (u8 *)info->screen_base + (sy - 1) * next_line + ((sx + width + 8) & ~15) / (8 / BPL);
+			dst = (u8 *)info->screen_base + (dy - 1) * next_line + ((dx + width + 8) & ~15) / (8 / BPL);
+
+			if ((sx + width) & 15) {
+				src -= BPL * 2;
+				dst -= BPL * 2;
+				memmove32_col(dst, src, 0xff00ff00, colsize, -next_line - BPL * 2);
+				width -= 8;
+			}
+			w = width >> 4;
+			if (w) {
+				s = (u32 *)src;
+				d = (u32 *)dst;
+				w *= BPL / 2;
+				l = next_line - w * 4;
+				for (j = height; j > 0; j--) {
+					for (i = w; i > 0; i--)
+						*--d = *--s;
+					s = (u32 *)((u8 *)s - l);
+					d = (u32 *)((u8 *)d - l);
+				}
+			}
+			if (sx & 15)
+				memmove32_col(dst - (width - 16) / (8 / BPL),
+					      src - (width - 16) / (8 / BPL),
+					      0xff00ff, colsize, -next_line - BPL * 2);
+		}
+	} else {
+		/* odd->even or even->odd */
+		if (upwards) {
+			u32 *src32, *dst32;
+			u32 pval[4], v, v1, mask;
+			int i, j, w, f;
+
+			src = (u8 *)info->screen_base + sy * next_line + (sx & ~15) / (8 / BPL);
+			dst = (u8 *)info->screen_base + dy * next_line + (dx & ~15) / (8 / BPL);
+
+			mask = 0xff00ff00;
+			f = 0;
+			w = width;
+			if (sx & 15) {
+				f = 1;
+				w += 8;
+			}
+			if ((sx + width) & 15)
+				f |= 2;
+			w >>= 4;
+			for (i = height; i; i--) {
+				src32 = (u32 *)src;
+				dst32 = (u32 *)dst;
+
+				if (f & 1) {
+					pval[0] = (*src32++ << 8) & mask;
+				} else {
+					pval[0] = dst32[0] & mask;
+				}
+
+				for (j = w; j > 0; j--) {
+					v = *src32++;
+					v1 = v & mask;
+					*dst32++ = pval[0] | (v1 >> 8);
+					pval[0] = (v ^ v1) << 8;
+				}
+
+				if (f & 2) {
+					dst32[0] = (dst32[0] & mask) | pval[0];
+				}
+
+				src += next_line;
+				dst += next_line;
+			}
+		} else {
+			u32 *src32, *dst32;
+			u32 pval[4], v, v1, mask;
+			int i, j, w, f;
+
+			src = (u8 *)info->screen_base + (sy - 1) * next_line + ((sx + width + 8) & ~15) / (8 / BPL);
+			dst = (u8 *)info->screen_base + (dy - 1) * next_line + ((dx + width + 8) & ~15) / (8 / BPL);
+
+			mask = 0xff00ff;
+			f = 0;
+			w = width;
+			if ((dx + width) & 15)
+				f = 1;
+			if (sx & 15) {
+				f |= 2;
+				w += 8;
+			}
+			w >>= 4;
+			for (i = height; i; i--) {
+				src32 = (u32 *)src;
+				dst32 = (u32 *)dst;
+
+				if (f & 1) {
+					pval[0] = dst32[-1] & mask;
+				} else {
+					pval[0] = (*--src32 >> 8) & mask;
+				}
+
+				for (j = w; j > 0; j--) {
+					v = *--src32;
+					v1 = v & mask;
+					*--dst32 = pval[0] | (v1 << 8);
+					pval[0] = (v ^ v1) >> 8;
+				}
+
+				if (!(f & 2)) {
+					dst32[-1] = (dst32[-1] & mask) | pval[0];
+				}
+
+				src -= next_line;
+				dst -= next_line;
+			}
+		}
+	}
+}
+
+void atafb_iplan2p2_fillrect(struct fb_info *info, u_long next_line, u32 color,
+                             int sy, int sx, int height, int width)
+{
+	u32 *dest;
+	int rows, i;
+	u32 cval[4];
+
+	dest = (u32 *)(info->screen_base + sy * next_line + (sx & ~15) / (8 / BPL));
+	if (sx & 15) {
+		u8 *dest8 = (u8 *)dest + 1;
+
+		expand8_col2mask(color, cval);
+
+		for (i = height; i; i--) {
+			fill8_col(dest8, cval);
+			dest8 += next_line;
+		}
+		dest += BPL / 2;
+		width -= 8;
+	}
+
+	expand16_col2mask(color, cval);
+	rows = width >> 4;
+	if (rows) {
+		u32 *d = dest;
+		u32 off = next_line - rows * BPL * 2;
+		for (i = height; i; i--) {
+			d = fill16_col(d, rows, cval);
+			d = (u32 *)((long)d + off);
+		}
+		dest += rows * BPL / 2;
+		width &= 15;
+	}
+
+	if (width) {
+		u8 *dest8 = (u8 *)dest;
+
+		expand8_col2mask(color, cval);
+
+		for (i = height; i; i--) {
+			fill8_col(dest8, cval);
+			dest8 += next_line;
+		}
+	}
+}
+
+void atafb_iplan2p2_linefill(struct fb_info *info, u_long next_line,
+                             int dy, int dx, u32 width,
+                             const u8 *data, u32 bgcolor, u32 fgcolor)
+{
+	u32 *dest;
+	const u16 *data16;
+	int rows;
+	u32 fgm[4], bgm[4], m;
+
+	dest = (u32 *)(info->screen_base + dy * next_line + (dx & ~15) / (8 / BPL));
+	if (dx & 15) {
+		fill8_2col((u8 *)dest + 1, fgcolor, bgcolor, *data++);
+		dest += BPL / 2;
+		width -= 8;
+	}
+
+	if (width >= 16) {
+		data16 = (const u16 *)data;
+		expand16_2col2mask(fgcolor, bgcolor, fgm, bgm);
+
+		for (rows = width / 16; rows; rows--) {
+			u16 d = *data16++;
+			m = d | ((u32)d << 16);
+			*dest++ = (m & fgm[0]) ^ bgm[0];
+		}
+
+		data = (const u8 *)data16;
+		width &= 15;
+	}
+
+	if (width)
+		fill8_2col((u8 *)dest, fgcolor, bgcolor, *data);
+}
+
+#ifdef MODULE
+MODULE_LICENSE("GPL");
+
+int init_module(void)
+{
+	return 0;
+}
+
+void cleanup_module(void)
+{
+}
+#endif /* MODULE */
+
+
+    /*
+     *  Visible symbols for modules
+     */
+
+EXPORT_SYMBOL(atafb_iplan2p2_copyarea);
+EXPORT_SYMBOL(atafb_iplan2p2_fillrect);
+EXPORT_SYMBOL(atafb_iplan2p2_linefill);

drivers/video/atafb_iplan2p4.c

+/*
+ *  linux/drivers/video/iplan2p4.c -- Low level frame buffer operations for
+ *				      interleaved bitplanes à la Atari (4
+ *				      planes, 2 bytes interleave)
+ *
+ *	Created 5 Apr 1997 by Geert Uytterhoeven
+ *
+ *  This file is subject to the terms and conditions of the GNU General Public
+ *  License.  See the file COPYING in the main directory of this archive for
+ *  more details.
+ */
+
+#include <linux/module.h>
+#include <linux/string.h>
+#include <linux/fb.h>
+
+#include <asm/setup.h>
+
+#include "atafb.h"
+
+#define BPL	4
+#include "atafb_utils.h"
+
+void atafb_iplan2p4_copyarea(struct fb_info *info, u_long next_line,
+			     int sy, int sx, int dy, int dx,
+			     int height, int width)
+{
+	/*  bmove() has to distinguish two major cases: If both, source and
+	 *  destination, start at even addresses or both are at odd
+	 *  addresses, just the first odd and last even column (if present)
+	 *  require special treatment (memmove_col()). The rest between
+	 *  then can be copied by normal operations, because all adjacent
+	 *  bytes are affected and are to be stored in the same order.
+	 *    The pathological case is when the move should go from an odd
+	 *  address to an even or vice versa. Since the bytes in the plane
+	 *  words must be assembled in new order, it seems wisest to make
+	 *  all movements by memmove_col().
+	 */
+
+	u8 *src, *dst;
+	u32 *s, *d;
+	int w, l , i, j;
+	u_int colsize;
+	u_int upwards = (dy < sy) || (dy == sy && dx < sx);
+
+	colsize = height;
+	if (!((sx ^ dx) & 15)) {
+		/* odd->odd or even->even */
+
+		if (upwards) {
+			src = (u8 *)info->screen_base + sy * next_line + (sx & ~15) / (8 / BPL);
+			dst = (u8 *)info->screen_base + dy * next_line + (dx & ~15) / (8 / BPL);
+			if (sx & 15) {
+				memmove32_col(dst, src, 0xff00ff, height, next_line - BPL * 2);
+				src += BPL * 2;
+				dst += BPL * 2;
+				width -= 8;
+			}
+			w = width >> 4;
+			if (w) {
+				s = (u32 *)src;
+				d = (u32 *)dst;
+				w *= BPL / 2;
+				l = next_line - w * 4;
+				for (j = height; j > 0; j--) {
+					for (i = w; i > 0; i--)
+						*d++ = *s++;
+					s = (u32 *)((u8 *)s + l);
+					d = (u32 *)((u8 *)d + l);
+				}
+			}
+			if (width & 15)
+				memmove32_col(dst + width / (8 / BPL), src + width / (8 / BPL),
+					      0xff00ff00, height, next_line - BPL * 2);
+		} else {
+			src = (u8 *)info->screen_base + (sy - 1) * next_line + ((sx + width + 8) & ~15) / (8 / BPL);
+			dst = (u8 *)info->screen_base + (dy - 1) * next_line + ((dx + width + 8) & ~15) / (8 / BPL);
+
+			if ((sx + width) & 15) {
+				src -= BPL * 2;
+				dst -= BPL * 2;
+				memmove32_col(dst, src, 0xff00ff00, colsize, -next_line - BPL * 2);
+				width -= 8;
+			}
+			w = width >> 4;
+			if (w) {
+				s = (u32 *)src;
+				d = (u32 *)dst;
+				w *= BPL / 2;
+				l = next_line - w * 4;
+				for (j = height; j > 0; j--) {
+					for (i = w; i > 0; i--)
+						*--d = *--s;
+					s = (u32 *)((u8 *)s - l);
+					d = (u32 *)((u8 *)d - l);
+				}
+			}
+			if (sx & 15)
+				memmove32_col(dst - (width - 16) / (8 / BPL),
+					      src - (width - 16) / (8 / BPL),
+					      0xff00ff, colsize, -next_line - BPL * 2);
+		}
+	} else {
+		/* odd->even or even->odd */
+		if (upwards) {
+			u32 *src32, *dst32;
+			u32 pval[4], v, v1, mask;
+			int i, j, w, f;
+
+			src = (u8 *)info->screen_base + sy * next_line + (sx & ~15) / (8 / BPL);
+			dst = (u8 *)info->screen_base + dy * next_line + (dx & ~15) / (8 / BPL);
+
+			mask = 0xff00ff00;
+			f = 0;
+			w = width;
+			if (sx & 15) {
+				f = 1;
+				w += 8;
+			}
+			if ((sx + width) & 15)
+				f |= 2;
+			w >>= 4;
+			for (i = height; i; i--) {
+				src32 = (u32 *)src;
+				dst32 = (u32 *)dst;
+
+				if (f & 1) {
+					pval[0] = (*src32++ << 8) & mask;
+					pval[1] = (*src32++ << 8) & mask;
+				} else {
+					pval[0] = dst32[0] & mask;
+					pval[1] = dst32[1] & mask;
+				}
+
+				for (j = w; j > 0; j--) {
+					v = *src32++;
+					v1 = v & mask;
+					*dst32++ = pval[0] | (v1 >> 8);
+					pval[0] = (v ^ v1) << 8;
+					v = *src32++;
+					v1 = v & mask;
+					*dst32++ = pval[1] | (v1 >> 8);
+					pval[1] = (v ^ v1) << 8;
+				}
+
+				if (f & 2) {
+					dst32[0] = (dst32[0] & mask) | pval[0];
+					dst32[1] = (dst32[1] & mask) | pval[1];
+				}
+
+				src += next_line;
+				dst += next_line;
+			}
+		} else {
+			u32 *src32, *dst32;
+			u32 pval[4], v, v1, mask;
+			int i, j, w, f;
+
+			src = (u8 *)info->screen_base + (sy - 1) * next_line + ((sx + width + 8) & ~15) / (8 / BPL);
+			dst = (u8 *)info->screen_base + (dy - 1) * next_line + ((dx + width + 8) & ~15) / (8 / BPL);
+
+			mask = 0xff00ff;
+			f = 0;
+			w = width;
+			if ((dx + width) & 15)
+				f = 1;
+			if (sx & 15) {
+				f |= 2;
+				w += 8;
+			}
+			w >>= 4;
+			for (i = height; i; i--) {
+				src32 = (u32 *)src;
+				dst32 = (u32 *)dst;
+
+				if (f & 1) {
+					pval[0] = dst32[-1] & mask;
+					pval[1] = dst32[-2] & mask;
+				} else {
+					pval[0] = (*--src32 >> 8) & mask;
+					pval[1] = (*--src32 >> 8) & mask;
+				}
+
+				for (j = w; j > 0; j--) {
+					v = *--src32;
+					v1 = v & mask;
+					*--dst32 = pval[0] | (v1 << 8);
+					pval[0] = (v ^ v1) >> 8;
+					v = *--src32;
+					v1 = v & mask;
+					*--dst32 = pval[1] | (v1 << 8);
+					pval[1] = (v ^ v1) >> 8;
+				}
+
+				if (!(f & 2)) {
+					dst32[-1] = (dst32[-1] & mask) | pval[0];
+					dst32[-2] = (dst32[-2] & mask) | pval[1];
+				}
+
+				src -= next_line;
+				dst -= next_line;
+			}
+		}
+	}
+}
+
+void atafb_iplan2p4_fillrect(struct fb_info *info, u_long next_line, u32 color,
+                             int sy, int sx, int height, int width)
+{
+	u32 *dest;
+	int rows, i;
+	u32 cval[4];
+
+	dest = (u32 *)(info->screen_base + sy * next_line + (sx & ~15) / (8 / BPL));
+	if (sx & 15) {
+		u8 *dest8 = (u8 *)dest + 1;
+
+		expand8_col2mask(color, cval);
+
+		for (i = height; i; i--) {
+			fill8_col(dest8, cval);
+			dest8 += next_line;
+		}
+		dest += BPL / 2;
+		width -= 8;
+	}
+
+	expand16_col2mask(color, cval);
+	rows = width >> 4;
+	if (rows) {
+		u32 *d = dest;
+		u32 off = next_line - rows * BPL * 2;
+		for (i = height; i; i--) {
+			d = fill16_col(d, rows, cval);
+			d = (u32 *)((long)d + off);
+		}
+		dest += rows * BPL / 2;
+		width &= 15;
+	}
+
+	if (width) {
+		u8 *dest8 = (u8 *)dest;
+
+		expand8_col2mask(color, cval);
+
+		for (i = height; i; i--) {
+			fill8_col(dest8, cval);
+			dest8 += next_line;
+		}
+	}
+}
+
+void atafb_iplan2p4_linefill(struct fb_info *info, u_long next_line,
+                             int dy, int dx, u32 width,
+                             const u8 *data, u32 bgcolor, u32 fgcolor)
+{
+	u32 *dest;
+	const u16 *data16;
+	int rows;
+	u32 fgm[4], bgm[4], m;
+
+	dest = (u32 *)(info->screen_base + dy * next_line + (dx & ~15) / (8 / BPL));
+	if (dx & 15) {
+		fill8_2col((u8 *)dest + 1, fgcolor, bgcolor, *data++);
+		dest += BPL / 2;
+		width -= 8;
+	}
+
+	if (width >= 16) {
+		data16 = (const u16 *)data;
+		expand16_2col2mask(fgcolor, bgcolor, fgm, bgm);
+
+		for (rows = width / 16; rows; rows--) {
+			u16 d = *data16++;
+			m = d | ((u32)d << 16);
+			*dest++ = (m & fgm[0]) ^ bgm[0];
+			*dest++ = (m & fgm[1]) ^ bgm[1];
+		}
+
+		data = (const u8 *)data16;
+		width &= 15;
+	}
+
+	if (width)
+		fill8_2col((u8 *)dest, fgcolor, bgcolor, *data);
+}
+
+#ifdef MODULE
+MODULE_LICENSE("GPL");
+
+int init_module(void)
+{
+	return 0;
+}
+
+void cleanup_module(void)
+{
+}
+#endif /* MODULE */
+
+
+    /*
+     *  Visible symbols for modules
+     */
+
+EXPORT_SYMBOL(atafb_iplan2p4_copyarea);
+EXPORT_SYMBOL(atafb_iplan2p4_fillrect);
+EXPORT_SYMBOL(atafb_iplan2p4_linefill);

drivers/video/atafb_iplan2p8.c

+/*
+ *  linux/drivers/video/iplan2p8.c -- Low level frame buffer operations for
+ *				      interleaved bitplanes à la Atari (8
+ *				      planes, 2 bytes interleave)
+ *
+ *	Created 5 Apr 1997 by Geert Uytterhoeven
+ *
+ *  This file is subject to the terms and conditions of the GNU General Public
+ *  License.  See the file COPYING in the main directory of this archive for
+ *  more details.
+ */
+
+#include <linux/module.h>
+#include <linux/string.h>
+#include <linux/fb.h>
+
+#include <asm/setup.h>
+
+#include "atafb.h"
+
+#define BPL	8
+#include "atafb_utils.h"
+
+
+/* Copies a 8 plane column from 's', height 'h', to 'd'. */
+
+/* This expands a 8 bit color into two longs for two movepl (8 plane)
+ * operations.
+ */
+
+void atafb_iplan2p8_copyarea(struct fb_info *info, u_long next_line,
+			     int sy, int sx, int dy, int dx,
+			     int height, int width)
+{
+	/*  bmove() has to distinguish two major cases: If both, source and
+	 *  destination, start at even addresses or both are at odd
+	 *  addresses, just the first odd and last even column (if present)
+	 *  require special treatment (memmove_col()). The rest between
+	 *  then can be copied by normal operations, because all adjacent
+	 *  bytes are affected and are to be stored in the same order.
+	 *    The pathological case is when the move should go from an odd
+	 *  address to an even or vice versa. Since the bytes in the plane
+	 *  words must be assembled in new order, it seems wisest to make
+	 *  all movements by memmove_col().
+	 */
+
+	u8 *src, *dst;
+	u32 *s, *d;
+	int w, l , i, j;
+	u_int colsize;
+	u_int upwards = (dy < sy) || (dy == sy && dx < sx);
+
+	colsize = height;
+	if (!((sx ^ dx) & 15)) {
+		/* odd->odd or even->even */
+
+		if (upwards) {
+			src = (u8 *)info->screen_base + sy * next_line + (sx & ~15) / (8 / BPL);
+			dst = (u8 *)info->screen_base + dy * next_line + (dx & ~15) / (8 / BPL);
+			if (sx & 15) {
+				memmove32_col(dst, src, 0xff00ff, height, next_line - BPL * 2);
+				src += BPL * 2;
+				dst += BPL * 2;
+				width -= 8;
+			}
+			w = width >> 4;
+			if (w) {
+				s = (u32 *)src;
+				d = (u32 *)dst;
+				w *= BPL / 2;
+				l = next_line - w * 4;
+				for (j = height; j > 0; j--) {
+					for (i = w; i > 0; i--)
+						*d++ = *s++;
+					s = (u32 *)((u8 *)s + l);
+					d = (u32 *)((u8 *)d + l);
+				}
+			}
+			if (width & 15)
+				memmove32_col(dst + width / (8 / BPL), src + width / (8 / BPL),
+					      0xff00ff00, height, next_line - BPL * 2);
+		} else {
+			src = (u8 *)info->screen_base + (sy - 1) * next_line + ((sx + width + 8) & ~15) / (8 / BPL);
+			dst = (u8 *)info->screen_base + (dy - 1) * next_line + ((dx + width + 8) & ~15) / (8 / BPL);
+
+			if ((sx + width) & 15) {
+				src -= BPL * 2;
+				dst -= BPL * 2;
+				memmove32_col(dst, src, 0xff00ff00, colsize, -next_line - BPL * 2);
+				width -= 8;
+			}
+			w = width >> 4;
+			if (w) {
+				s = (u32 *)src;
+				d = (u32 *)dst;
+				w *= BPL / 2;
+				l = next_line - w * 4;
+				for (j = height; j > 0; j--) {
+					for (i = w; i > 0; i--)
+						*--d = *--s;
+					s = (u32 *)((u8 *)s - l);
+					d = (u32 *)((u8 *)d - l);
+				}
+			}
+			if (sx & 15)
+				memmove32_col(dst - (width - 16) / (8 / BPL),
+					      src - (width - 16) / (8 / BPL),
+					      0xff00ff, colsize, -next_line - BPL * 2);
+		}
+	} else {
+		/* odd->even or even->odd */
+		if (upwards) {
+			u32 *src32, *dst32;
+			u32 pval[4], v, v1, mask;
+			int i, j, w, f;
+
+			src = (u8 *)info->screen_base + sy * next_line + (sx & ~15) / (8 / BPL);
+			dst = (u8 *)info->screen_base + dy * next_line + (dx & ~15) / (8 / BPL);
+
+			mask = 0xff00ff00;
+			f = 0;
+			w = width;
+			if (sx & 15) {
+				f = 1;
+				w += 8;
+			}
+			if ((sx + width) & 15)
+				f |= 2;
+			w >>= 4;
+			for (i = height; i; i--) {
+				src32 = (u32 *)src;
+				dst32 = (u32 *)dst;
+
+				if (f & 1) {
+					pval[0] = (*src32++ << 8) & mask;
+					pval[1] = (*src32++ << 8) & mask;
+					pval[2] = (*src32++ << 8) & mask;
+					pval[3] = (*src32++ << 8) & mask;
+				} else {
+					pval[0] = dst32[0] & mask;
+					pval[1] = dst32[1] & mask;
+					pval[2] = dst32[2] & mask;
+					pval[3] = dst32[3] & mask;
+				}
+
+				for (j = w; j > 0; j--) {
+					v = *src32++;
+					v1 = v & mask;
+					*dst32++ = pval[0] | (v1 >> 8);
+					pval[0] = (v ^ v1) << 8;
+					v = *src32++;
+					v1 = v & mask;
+					*dst32++ = pval[1] | (v1 >> 8);
+					pval[1] = (v ^ v1) << 8;
+					v = *src32++;
+					v1 = v & mask;
+					*dst32++ = pval[2] | (v1 >> 8);
+					pval[2] = (v ^ v1) << 8;
+					v = *src32++;
+					v1 = v & mask;
+					*dst32++ = pval[3] | (v1 >> 8);
+					pval[3] = (v ^ v1) << 8;
+				}
+
+				if (f & 2) {
+					dst32[0] = (dst32[0] & mask) | pval[0];
+					dst32[1] = (dst32[1] & mask) | pval[1];
+					dst32[2] = (dst32[2] & mask) | pval[2];
+					dst32[3] = (dst32[3] & mask) | pval[3];
+				}
+
+				src += next_line;
+				dst += next_line;
+			}
+		} else {
+			u32 *src32, *dst32;
+			u32 pval[4], v, v1, mask;
+			int i, j, w, f;
+
+			src = (u8 *)info->screen_base + (sy - 1) * next_line + ((sx + width + 8) & ~15) / (8 / BPL);
+			dst = (u8 *)info->screen_base + (dy - 1) * next_line + ((dx + width + 8) & ~15) / (8 / BPL);
+
+			mask = 0xff00ff;
+			f = 0;
+			w = width;
+			if ((dx + width) & 15)
+				f = 1;
+			if (sx & 15) {
+				f |= 2;
+				w += 8;
+			}
+			w >>= 4;
+			for (i = height; i; i--) {
+				src32 = (u32 *)src;
+				dst32 = (u32 *)dst;
+
+				if (f & 1) {
+					pval[0] = dst32[-1] & mask;
+					pval[1] = dst32[-2] & mask;
+					pval[2] = dst32[-3] & mask;
+					pval[3] = dst32[-4] & mask;
+				} else {
+					pval[0] = (*--src32 >> 8) & mask;
+					pval[1] = (*--src32 >> 8) & mask;
+					pval[2] = (*--src32 >> 8) & mask;
+					pval[3] = (*--src32 >> 8) & mask;
+				}
+
+				for (j = w; j > 0; j--) {
+					v = *--src32;
+					v1 = v & mask;
+					*--dst32 = pval[0] | (v1 << 8);
+					pval[0] = (v ^ v1) >> 8;
+					v = *--src32;
+					v1 = v & mask;
+					*--dst32 = pval[1] | (v1 << 8);
+					pval[1] = (v ^ v1) >> 8;
+					v = *--src32;
+					v1 = v & mask;
+					*--dst32 = pval[2] | (v1 << 8);
+					pval[2] = (v ^ v1) >> 8;
+					v = *--src32;
+					v1 = v & mask;
+					*--dst32 = pval[3] | (v1 << 8);
+					pval[3] = (v ^ v1) >> 8;
+				}
+
+				if (!(f & 2)) {
+					dst32[-1] = (dst32[-1] & mask) | pval[0];
+					dst32[-2] = (dst32[-2] & mask) | pval[1];
+					dst32[-3] = (dst32[-3] & mask) | pval[2];
+					dst32[-4] = (dst32[-4] & mask) | pval[3];
+				}
+
+				src -= next_line;
+				dst -= next_line;
+			}
+		}
+	}
+}
+
+void atafb_iplan2p8_fillrect(struct fb_info *info, u_long next_line, u32 color,
+                             int sy, int sx, int height, int width)
+{
+	u32 *dest;
+	int rows, i;
+	u32 cval[4];
+
+	dest = (u32 *)(info->screen_base + sy * next_line + (sx & ~15) / (8 / BPL));
+	if (sx & 15) {
+		u8 *dest8 = (u8 *)dest + 1;
+
+		expand8_col2mask(color, cval);
+
+		for (i = height; i; i--) {
+			fill8_col(dest8, cval);
+			dest8 += next_line;
+		}
+		dest += BPL / 2;
+		width -= 8;
+	}
+
+	expand16_col2mask(color, cval);
+	rows = width >> 4;
+	if (rows) {
+		u32 *d = dest;
+		u32 off = next_line - rows * BPL * 2;
+		for (i = height; i; i--) {
+			d = fill16_col(d, rows, cval);
+			d = (u32 *)((long)d + off);
+		}
+		dest += rows * BPL / 2;
+		width &= 15;
+	}
+
+	if (width) {
+		u8 *dest8 = (u8 *)dest;
+
+		expand8_col2mask(color, cval);
+
+		for (i = height; i; i--) {
+			fill8_col(dest8, cval);
+			dest8 += next_line;
+		}
+	}
+}
+
+void atafb_iplan2p8_linefill(struct fb_info *info, u_long next_line,
+			     int dy, int dx, u32 width,
+			     const u8 *data, u32 bgcolor, u32 fgcolor)
+{
+	u32 *dest;
+	const u16 *data16;
+	int rows;
+	u32 fgm[4], bgm[4], m;
+
+	dest = (u32 *)(info->screen_base + dy * next_line + (dx & ~15) / (8 / BPL));
+	if (dx & 15) {
+		fill8_2col((u8 *)dest + 1, fgcolor, bgcolor, *data++);
+		dest += BPL / 2;
+		width -= 8;
+	}
+
+	if (width >= 16) {
+		data16 = (const u16 *)data;
+		expand16_2col2mask(fgcolor, bgcolor, fgm, bgm);
+
+		for (rows = width / 16; rows; rows--) {
+			u16 d = *data16++;
+			m = d | ((u32)d << 16);
+			*dest++ = (m & fgm[0]) ^ bgm[0];
+			*dest++ = (m & fgm[1]) ^ bgm[1];
+			*dest++ = (m & fgm[2]) ^ bgm[2];
+			*dest++ = (m & fgm[3]) ^ bgm[3];
+		}
+
+		data = (const u8 *)data16;
+		width &= 15;
+	}
+
+	if (width)
+		fill8_2col((u8 *)dest, fgcolor, bgcolor, *data);
+}
+
+#ifdef MODULE
+MODULE_LICENSE("GPL");
+
+int init_module(void)
+{
+	return 0;
+}
+
+void cleanup_module(void)
+{
+}
+#endif /* MODULE */
+
+
+    /*
+     *  Visible symbols for modules
+     */
+
+EXPORT_SYMBOL(atafb_iplan2p8_copyarea);
+EXPORT_SYMBOL(atafb_iplan2p8_fillrect);
+EXPORT_SYMBOL(atafb_iplan2p8_linefill);

drivers/video/atafb_mfb.c

+/*
+ *  linux/drivers/video/mfb.c -- Low level frame buffer operations for
+ *				 monochrome
+ *
+ *	Created 5 Apr 1997 by Geert Uytterhoeven
+ *
+ *  This file is subject to the terms and conditions of the GNU General Public
+ *  License.  See the file COPYING in the main directory of this archive for
+ *  more details.
+ */
+
+#include <linux/module.h>
+#include <linux/string.h>
+#include <linux/fb.h>
+
+#include "atafb.h"
+#include "atafb_utils.h"
+
+
+    /*
+     *  Monochrome
+     */
+
+void atafb_mfb_copyarea(struct fb_info *info, u_long next_line,
+			int sy, int sx, int dy, int dx,
+			int height, int width)
+{
+	u8 *src, *dest;
+	u_int rows;
+
+	if (sx == 0 && dx == 0 && width == next_line) {
+		src = (u8 *)info->screen_base + sy * (width >> 3);
+		dest = (u8 *)info->screen_base + dy * (width >> 3);
+		fb_memmove(dest, src, height * (width >> 3));
+	} else if (dy <= sy) {
+		src = (u8 *)info->screen_base + sy * next_line + (sx >> 3);
+		dest = (u8 *)info->screen_base + dy * next_line + (dx >> 3);
+		for (rows = height; rows--;) {
+			fb_memmove(dest, src, width >> 3);
+			src += next_line;
+			dest += next_line;
+		}
+	} else {
+		src = (u8 *)info->screen_base + (sy + height - 1) * next_line + (sx >> 3);
+		dest = (u8 *)info->screen_base + (dy + height - 1) * next_line + (dx >> 3);
+		for (rows = height; rows--;) {
+			fb_memmove(dest, src, width >> 3);
+			src -= next_line;
+			dest -= next_line;
+		}
+	}
+}
+
+void atafb_mfb_fillrect(struct fb_info *info, u_long next_line, u32 color,
+			int sy, int sx, int height, int width)
+{
+	u8 *dest;
+	u_int rows;
+
+	dest = (u8 *)info->screen_base + sy * next_line + (sx >> 3);
+
+	if (sx == 0 && width == next_line) {
+		if (color)
+			fb_memset255(dest, height * (width >> 3));
+		else
+			fb_memclear(dest, height * (width >> 3));
+	} else {
+		for (rows = height; rows--; dest += next_line) {
+			if (color)
+				fb_memset255(dest, width >> 3);
+			else
+				fb_memclear_small(dest, width >> 3);
+		}
+	}
+}
+
+void atafb_mfb_linefill(struct fb_info *info, u_long next_line,
+			int dy, int dx, u32 width,
+			const u8 *data, u32 bgcolor, u32 fgcolor)
+{
+	u8 *dest;
+	u_int rows;
+
+	dest = (u8 *)info->screen_base + dy * next_line + (dx >> 3);
+
+	for (rows = width / 8; rows--; /* check margins */ ) {
+		// use fast_memmove or fb_memmove
+		*dest++ = *data++;
+	}
+}
+
+#ifdef MODULE
+MODULE_LICENSE("GPL");
+
+int init_module(void)
+{
+	return 0;
+}
+
+void cleanup_module(void)
+{
+}
+#endif /* MODULE */
+
+
+    /*
+     *  Visible symbols for modules
+     */
+
+EXPORT_SYMBOL(atafb_mfb_copyarea);
+EXPORT_SYMBOL(atafb_mfb_fillrect);
+EXPORT_SYMBOL(atafb_mfb_linefill);

drivers/video/atafb_utils.h

+#ifndef _VIDEO_ATAFB_UTILS_H
+#define _VIDEO_ATAFB_UTILS_H
+
+/* ================================================================= */
+/*                      Utility Assembler Functions                  */
+/* ================================================================= */
+
+/* ====================================================================== */
+
+/* Those of a delicate disposition might like to skip the next couple of
+ * pages.
+ *
+ * These functions are drop in replacements for memmove and
+ * memset(_, 0, _). However their five instances add at least a kilobyte
+ * to the object file. You have been warned.
+ *
+ * Not a great fan of assembler for the sake of it, but I think
+ * that these routines are at least 10 times faster than their C
+ * equivalents for large blits, and that's important to the lowest level of
+ * a graphics driver. Question is whether some scheme with the blitter
+ * would be faster. I suspect not for simple text system - not much
+ * asynchrony.
+ *
+ * Code is very simple, just gruesome expansion. Basic strategy is to
+ * increase data moved/cleared at each step to 16 bytes to reduce
+ * instruction per data move overhead. movem might be faster still
+ * For more than 15 bytes, we try to align the write direction on a
+ * longword boundary to get maximum speed. This is even more gruesome.
+ * Unaligned read/write used requires 68020+ - think this is a problem?
+ *
+ * Sorry!
+ */
+
+
+/* ++roman: I've optimized Robert's original versions in some minor
+ * aspects, e.g. moveq instead of movel, let gcc choose the registers,
+ * use movem in some places...
+ * For other modes than 1 plane, lots of more such assembler functions
+ * were needed (e.g. the ones using movep or expanding color values).
+ */
+
+/* ++andreas: more optimizations:
+   subl #65536,d0 replaced by clrw d0; subql #1,d0 for dbcc
+   addal is faster than addaw
+   movep is rather expensive compared to ordinary move's
+   some functions rewritten in C for clarity, no speed loss */
+
+static inline void *fb_memclear_small(void *s, size_t count)
+{
+	if (!count)
+		return 0;
+
+	asm volatile ("\n"
+		"	lsr.l	#1,%1 ; jcc 1f ; move.b %2,-(%0)\n"
+		"1:	lsr.l	#1,%1 ; jcc 1f ; move.w %2,-(%0)\n"
+		"1:	lsr.l	#1,%1 ; jcc 1f ; move.l %2,-(%0)\n"
+		"1:	lsr.l	#1,%1 ; jcc 1f ; move.l %2,-(%0) ; move.l %2,-(%0)\n"
+		"1:"
+		: "=a" (s), "=d" (count)
+		: "d" (0), "0" ((char *)s + count), "1" (count));
+	asm volatile ("\n"
+		"	subq.l  #1,%1\n"
+		"	jcs	3f\n"
+		"	move.l	%2,%%d4; move.l %2,%%d5; move.l %2,%%d6\n"
+		"2:	movem.l	%2/%%d4/%%d5/%%d6,-(%0)\n"
+		"	dbra	%1,2b\n"
+		"3:"
+		: "=a" (s), "=d" (count)
+		: "d" (0), "0" (s), "1" (count)
+		: "d4", "d5", "d6"
+		);
+
+	return 0;
+}
+
+
+static inline void *fb_memclear(void *s, size_t count)
+{
+	if (!count)
+		return 0;
+
+	if (count < 16) {
+		asm volatile ("\n"
+			"	lsr.l	#1,%1 ; jcc 1f ; clr.b (%0)+\n"
+			"1:	lsr.l	#1,%1 ; jcc 1f ; clr.w (%0)+\n"
+			"1:	lsr.l	#1,%1 ; jcc 1f ; clr.l (%0)+\n"
+			"1:	lsr.l	#1,%1 ; jcc 1f ; clr.l (%0)+ ; clr.l (%0)+\n"
+			"1:"
+			: "=a" (s), "=d" (count)
+			: "0" (s), "1" (count));
+	} else {
+		long tmp;
+		asm volatile ("\n"
+			"	move.l	%1,%2\n"
+			"	lsr.l	#1,%2 ; jcc 1f ; clr.b (%0)+ ; subq.w #1,%1\n"
+			"	lsr.l	#1,%2 ; jcs 2f\n"  /* %0 increased=>bit 2 switched*/
+			"	clr.w	(%0)+  ; subq.w  #2,%1 ; jra 2f\n"
+			"1:	lsr.l	#1,%2 ; jcc 2f\n"
+			"	clr.w	(%0)+  ; subq.w  #2,%1\n"
+			"2:	move.w	%1,%2; lsr.l #2,%1 ; jeq 6f\n"
+			"	lsr.l	#1,%1 ; jcc 3f ; clr.l (%0)+\n"
+			"3:	lsr.l	#1,%1 ; jcc 4f ; clr.l (%0)+ ; clr.l (%0)+\n"
+			"4:	subq.l	#1,%1 ; jcs 6f\n"
+			"5:	clr.l	(%0)+; clr.l (%0)+ ; clr.l (%0)+ ; clr.l (%0)+\n"
+			"	dbra	%1,5b ; clr.w %1; subq.l #1,%1; jcc 5b\n"
+			"6:	move.w	%2,%1; btst #1,%1 ; jeq 7f ; clr.w (%0)+\n"
+			"7:	btst	#0,%1 ; jeq 8f ; clr.b (%0)+\n"
+			"8:"
+			: "=a" (s), "=d" (count), "=d" (tmp)
+			: "0" (s), "1" (count));
+	}
+
+	return 0;
+}
+
+
+static inline void *fb_memset255(void *s, size_t count)
+{
+	if (!count)
+		return 0;
+
+	asm volatile ("\n"
+		"	lsr.l	#1,%1 ; jcc 1f ; move.b %2,-(%0)\n"
+		"1:	lsr.l	#1,%1 ; jcc 1f ; move.w %2,-(%0)\n"
+		"1:	lsr.l	#1,%1 ; jcc 1f ; move.l %2,-(%0)\n"
+		"1:	lsr.l	#1,%1 ; jcc 1f ; move.l %2,-(%0) ; move.l %2,-(%0)\n"
+		"1:"
+		: "=a" (s), "=d" (count)
+		: "d" (-1), "0" ((char *)s+count), "1" (count));
+	asm volatile ("\n"
+		"	subq.l	#1,%1 ; jcs 3f\n"
+		"	move.l	%2,%%d4; move.l %2,%%d5; move.l %2,%%d6\n"
+		"2:	movem.l	%2/%%d4/%%d5/%%d6,-(%0)\n"
+		"	dbra	%1,2b\n"
+		"3:"
+		: "=a" (s), "=d" (count)
+		: "d" (-1), "0" (s), "1" (count)
+		: "d4", "d5", "d6");
+
+	return 0;
+}
+
+
+static inline void *fb_memmove(void *d, const void *s, size_t count)
+{
+	if (d < s) {
+		if (count < 16) {
+			asm volatile ("\n"
+				"	lsr.l	#1,%2 ; jcc 1f ; move.b (%1)+,(%0)+\n"
+				"1:	lsr.l	#1,%2 ; jcc 1f ; move.w (%1)+,(%0)+\n"
+				"1:	lsr.l	#1,%2 ; jcc 1f ; move.l (%1)+,(%0)+\n"
+				"1:	lsr.l	#1,%2 ; jcc 1f ; move.l (%1)+,(%0)+ ; move.l (%1)+,(%0)+\n"
+				"1:"
+				: "=a" (d), "=a" (s), "=d" (count)
+				: "0" (d), "1" (s), "2" (count));
+		} else {
+			long tmp;
+			asm volatile ("\n"
+				"	move.l	%0,%3\n"
+				"	lsr.l	#1,%3 ; jcc 1f ; move.b (%1)+,(%0)+ ; subqw #1,%2\n"
+				"	lsr.l	#1,%3 ; jcs 2f\n"  /* %0 increased=>bit 2 switched*/
+				"	move.w	(%1)+,(%0)+  ; subqw  #2,%2 ; jra 2f\n"
+				"1:	lsr.l   #1,%3 ; jcc 2f\n"
+				"	move.w	(%1)+,(%0)+  ; subqw  #2,%2\n"
+				"2:	move.w	%2,%-; lsr.l #2,%2 ; jeq 6f\n"
+				"	lsr.l	#1,%2 ; jcc 3f ; move.l (%1)+,(%0)+\n"
+				"3:	lsr.l	#1,%2 ; jcc 4f ; move.l (%1)+,(%0)+ ; move.l (%1)+,(%0)+\n"
+				"4:	subq.l	#1,%2 ; jcs 6f\n"
+				"5:	move.l	(%1)+,(%0)+; move.l (%1)+,(%0)+\n"
+				"	move.l	(%1)+,(%0)+; move.l (%1)+,(%0)+\n"
+				"	dbra	%2,5b ; clr.w %2; subq.l #1,%2; jcc 5b\n"
+				"6:	move.w	%+,%2; btst #1,%2 ; jeq 7f ; move.w (%1)+,(%0)+\n"
+				"7:	btst	#0,%2 ; jeq 8f ; move.b (%1)+,(%0)+\n"
+				"8:"
+				: "=a" (d), "=a" (s), "=d" (count), "=d" (tmp)
+				: "0" (d), "1" (s), "2" (count));
+		}
+	} else {
+		if (count < 16) {
+			asm volatile ("\n"
+				"	lsr.l	#1,%2 ; jcc 1f ; move.b -(%1),-(%0)\n"
+				"1:	lsr.l	#1,%2 ; jcc 1f ; move.w -(%1),-(%0)\n"
+				"1:	lsr.l	#1,%2 ; jcc 1f ; move.l -(%1),-(%0)\n"
+				"1:	lsr.l	#1,%2 ; jcc 1f ; move.l -(%1),-(%0) ; move.l -(%1),-(%0)\n"
+				"1:"
+				: "=a" (d), "=a" (s), "=d" (count)
+				: "0" ((char *) d + count), "1" ((char *) s + count), "2" (count));
+		} else {
+			long tmp;
+
+			asm volatile ("\n"
+				"	move.l	%0,%3\n"
+				"	lsr.l	#1,%3 ; jcc 1f ; move.b -(%1),-(%0) ; subqw #1,%2\n"
+				"	lsr.l	#1,%3 ; jcs 2f\n"  /* %0 increased=>bit 2 switched*/
+				"	move.w	-(%1),-(%0) ; subqw  #2,%2 ; jra 2f\n"
+				"1:	lsr.l	#1,%3 ; jcc 2f\n"
+				"	move.w	-(%1),-(%0) ; subqw  #2,%2\n"
+				"2:	move.w	%2,%-; lsr.l #2,%2 ; jeq 6f\n"
+				"	lsr.l	#1,%2 ; jcc 3f ; move.l -(%1),-(%0)\n"
+				"3:	lsr.l	#1,%2 ; jcc 4f ; move.l -(%1),-(%0) ; move.l -(%1),-(%0)\n"
+				"4:	subq.l	#1,%2 ; jcs 6f\n"
+				"5:	move.l	-(%1),-(%0); move.l -(%1),-(%0)\n"
+				"	move.l	-(%1),-(%0); move.l -(%1),-(%0)\n"
+				"	dbra	%2,5b ; clr.w %2; subq.l #1,%2; jcc 5b\n"
+				"6:	move.w	%+,%2; btst #1,%2 ; jeq 7f ; move.w -(%1),-(%0)\n"
+				"7:	btst	#0,%2 ; jeq 8f ; move.b -(%1),-(%0)\n"
+				"8:"
+				: "=a" (d), "=a" (s), "=d" (count), "=d" (tmp)
+				: "0" ((char *) d + count), "1" ((char *) s + count), "2" (count));
+		}
+	}
+
+	return 0;
+}
+
+
+/* ++andreas: Simple and fast version of memmove, assumes size is
+   divisible by 16, suitable for moving the whole screen bitplane */
+static inline void fast_memmove(char *dst, const char *src, size_t size)
+{
+	if (!size)
+		return;
+	if (dst < src)
+		asm volatile ("\n"
+			"1:	movem.l	(%0)+,%%d0/%%d1/%%a0/%%a1\n"
+			"	movem.l	%%d0/%%d1/%%a0/%%a1,%1@\n"
+			"	addq.l	#8,%1; addq.l #8,%1\n"
+			"	dbra	%2,1b\n"
+			"	clr.w	%2; subq.l #1,%2\n"
+			"	jcc	1b"
+			: "=a" (src), "=a" (dst), "=d" (size)
+			: "0" (src), "1" (dst), "2" (size / 16 - 1)
+			: "d0", "d1", "a0", "a1", "memory");
+	else
+		asm volatile ("\n"
+			"1:	subq.l	#8,%0; subq.l #8,%0\n"
+			"	movem.l	%0@,%%d0/%%d1/%%a0/%%a1\n"
+			"	movem.l	%%d0/%%d1/%%a0/%%a1,-(%1)\n"
+			"	dbra	%2,1b\n"
+			"	clr.w	%2; subq.l #1,%2\n"
+			"	jcc 1b"
+			: "=a" (src), "=a" (dst), "=d" (size)
+			: "0" (src + size), "1" (dst + size), "2" (size / 16 - 1)
+			: "d0", "d1", "a0", "a1", "memory");
+}
+
+#ifdef BPL
+
+/*
+ * This expands a up to 8 bit color into two longs
+ * for movel operations.
+ */
+static const u32 four2long[] = {
+	0x00000000, 0x000000ff, 0x0000ff00, 0x0000ffff,
+	0x00ff0000, 0x00ff00ff, 0x00ffff00, 0x00ffffff,
+	0xff000000, 0xff0000ff, 0xff00ff00, 0xff00ffff,
+	0xffff0000, 0xffff00ff, 0xffffff00, 0xffffffff,
+};
+
+static inline void expand8_col2mask(u8 c, u32 m[])
+{
+	m[0] = four2long[c & 15];
+#if BPL > 4
+	m[1] = four2long[c >> 4];
+#endif
+}
+
+static inline void expand8_2col2mask(u8 fg, u8 bg, u32 fgm[], u32 bgm[])
+{
+	fgm[0] = four2long[fg & 15] ^ (bgm[0] = four2long[bg & 15]);
+#if BPL > 4
+	fgm[1] = four2long[fg >> 4] ^ (bgm[1] = four2long[bg >> 4]);
+#endif
+}
+
+/*
+ * set an 8bit value to a color
+ */
+static inline void fill8_col(u8 *dst, u32 m[])
+{
+	u32 tmp = m[0];
+	dst[0] = tmp;
+	dst[2] = (tmp >>= 8);
+#if BPL > 2
+	dst[4] = (tmp >>= 8);
+	dst[6] = tmp >> 8;
+#endif
+#if BPL > 4
+	tmp = m[1];
+	dst[8] = tmp;
+	dst[10] = (tmp >>= 8);
+	dst[12] = (tmp >>= 8);
+	dst[14] = tmp >> 8;
+#endif
+}
+
+/*
+ * set an 8bit value according to foreground/background color
+ */
+static inline void fill8_2col(u8 *dst, u8 fg, u8 bg, u32 mask)
+{
+	u32 fgm[2], bgm[2], tmp;
+
+	expand8_2col2mask(fg, bg, fgm, bgm);
+
+	mask |= mask << 8;
+#if BPL > 2
+	mask |= mask << 16;
+#endif
+	tmp = (mask & fgm[0]) ^ bgm[0];
+	dst[0] = tmp;
+	dst[2] = (tmp >>= 8);
+#if BPL > 2
+	dst[4] = (tmp >>= 8);
+	dst[6] = tmp >> 8;
+#endif
+#if BPL > 4
+	tmp = (mask & fgm[1]) ^ bgm[1];
+	dst[8] = tmp;
+	dst[10] = (tmp >>= 8);
+	dst[12] = (tmp >>= 8);
+	dst[14] = tmp >> 8;
+#endif
+}
+
+static const u32 two2word[] = {
+	0x00000000, 0xffff0000, 0x0000ffff, 0xffffffff
+};
+
+static inline void expand16_col2mask(u8 c, u32 m[])
+{
+	m[0] = two2word[c & 3];
+#if BPL > 2
+	m[1] = two2word[(c >> 2) & 3];
+#endif
+#if BPL > 4
+	m[2] = two2word[(c >> 4) & 3];
+	m[3] = two2word[c >> 6];
+#endif
+}
+
+static inline void expand16_2col2mask(u8 fg, u8 bg, u32 fgm[], u32 bgm[])
+{
+	bgm[0] = two2word[bg & 3];
+	fgm[0] = two2word[fg & 3] ^ bgm[0];
+#if BPL > 2
+	bgm[1] = two2word[(bg >> 2) & 3];
+	fgm[1] = two2word[(fg >> 2) & 3] ^ bgm[1];
+#endif
+#if BPL > 4
+	bgm[2] = two2word[(bg >> 4) & 3];
+	fgm[2] = two2word[(fg >> 4) & 3] ^ bgm[2];
+	bgm[3] = two2word[bg >> 6];
+	fgm[3] = two2word[fg >> 6] ^ bgm[3];
+#endif
+}
+
+static inline u32 *fill16_col(u32 *dst, int rows, u32 m[])
+{
+	while (rows) {
+		*dst++ = m[0];
+#if BPL > 2
+		*dst++ = m[1];
+#endif
+#if BPL > 4
+		*dst++ = m[2];
+		*dst++ = m[3];
+#endif
+		rows--;
+	}
+	return dst;
+}
+
+static inline void memmove32_col(void *dst, void *src, u32 mask, u32 h, u32 bytes)
+{
+	u32 *s, *d, v;
+
+        s = src;
+        d = dst;
+        do {
+                v = (*s++ & mask) | (*d  & ~mask);
+                *d++ = v;
+#if BPL > 2
+                v = (*s++ & mask) | (*d  & ~mask);
+                *d++ = v;
+#endif
+#if BPL > 4
+                v = (*s++ & mask) | (*d  & ~mask);
+                *d++ = v;
+                v = (*s++ & mask) | (*d  & ~mask);
+                *d++ = v;
+#endif
+                d = (u32 *)((u8 *)d + bytes);
+                s = (u32 *)((u8 *)s + bytes);
+        } while (--h);
+}
+
+#endif
+
+#endif /* _VIDEO_ATAFB_UTILS_H */