1/*
2 *  Generic 1-bit or 8-bit source to 1-32 bit destination expansion
3 *  for frame buffer located in system RAM with packed pixels of any depth.
4 *
5 *  Based almost entirely on cfbimgblt.c
6 *
7 *      Copyright (C)  April 2007 Antonino Daplas <adaplas@pol.net>
8 *
9 *  This file is subject to the terms and conditions of the GNU General Public
10 *  License.  See the file COPYING in the main directory of this archive for
11 *  more details.
12 */
13#include <linux/module.h>
14#include <linux/string.h>
15#include <linux/fb.h>
16#include <asm/types.h>
17
18#define DEBUG
19
20#ifdef DEBUG
21#define DPRINTK(fmt, args...) printk(KERN_DEBUG "%s: " fmt,__func__,## args)
22#else
23#define DPRINTK(fmt, args...)
24#endif
25
26static const u32 cfb_tab8_be[] = {
27    0x00000000,0x000000ff,0x0000ff00,0x0000ffff,
28    0x00ff0000,0x00ff00ff,0x00ffff00,0x00ffffff,
29    0xff000000,0xff0000ff,0xff00ff00,0xff00ffff,
30    0xffff0000,0xffff00ff,0xffffff00,0xffffffff
31};
32
33static const u32 cfb_tab8_le[] = {
34    0x00000000,0xff000000,0x00ff0000,0xffff0000,
35    0x0000ff00,0xff00ff00,0x00ffff00,0xffffff00,
36    0x000000ff,0xff0000ff,0x00ff00ff,0xffff00ff,
37    0x0000ffff,0xff00ffff,0x00ffffff,0xffffffff
38};
39
40static const u32 cfb_tab16_be[] = {
41    0x00000000, 0x0000ffff, 0xffff0000, 0xffffffff
42};
43
44static const u32 cfb_tab16_le[] = {
45    0x00000000, 0xffff0000, 0x0000ffff, 0xffffffff
46};
47
48static const u32 cfb_tab32[] = {
49	0x00000000, 0xffffffff
50};
51
52static void color_imageblit(const struct fb_image *image, struct fb_info *p,
53			    void *dst1, u32 start_index, u32 pitch_index)
54{
55	/* Draw the penguin */
56	u32 *dst, *dst2;
57	u32 color = 0, val, shift;
58	int i, n, bpp = p->var.bits_per_pixel;
59	u32 null_bits = 32 - bpp;
60	u32 *palette = (u32 *) p->pseudo_palette;
61	const u8 *src = image->data;
62
63	dst2 = dst1;
64	for (i = image->height; i--; ) {
65		n = image->width;
66		dst = dst1;
67		shift = 0;
68		val = 0;
69
70		if (start_index) {
71			u32 start_mask = ~(FB_SHIFT_HIGH(p, ~(u32)0,
72							 start_index));
73			val = *dst & start_mask;
74			shift = start_index;
75		}
76		while (n--) {
77			if (p->fix.visual == FB_VISUAL_TRUECOLOR ||
78			    p->fix.visual == FB_VISUAL_DIRECTCOLOR )
79				color = palette[*src];
80			else
81				color = *src;
82			color <<= FB_LEFT_POS(p, bpp);
83			val |= FB_SHIFT_HIGH(p, color, shift);
84			if (shift >= null_bits) {
85				*dst++ = val;
86
87				val = (shift == null_bits) ? 0 :
88					FB_SHIFT_LOW(p, color, 32 - shift);
89			}
90			shift += bpp;
91			shift &= (32 - 1);
92			src++;
93		}
94		if (shift) {
95			u32 end_mask = FB_SHIFT_HIGH(p, ~(u32)0, shift);
96
97			*dst &= end_mask;
98			*dst |= val;
99		}
100		dst1 += p->fix.line_length;
101		if (pitch_index) {
102			dst2 += p->fix.line_length;
103			dst1 = (u8 *)((long)dst2 & ~(sizeof(u32) - 1));
104
105			start_index += pitch_index;
106			start_index &= 32 - 1;
107		}
108	}
109}
110
111static void slow_imageblit(const struct fb_image *image, struct fb_info *p,
112				  void *dst1, u32 fgcolor, u32 bgcolor,
113				  u32 start_index, u32 pitch_index)
114{
115	u32 shift, color = 0, bpp = p->var.bits_per_pixel;
116	u32 *dst, *dst2;
117	u32 val, pitch = p->fix.line_length;
118	u32 null_bits = 32 - bpp;
119	u32 spitch = (image->width+7)/8;
120	const u8 *src = image->data, *s;
121	u32 i, j, l;
122
123	dst2 = dst1;
124	fgcolor <<= FB_LEFT_POS(p, bpp);
125	bgcolor <<= FB_LEFT_POS(p, bpp);
126
127	for (i = image->height; i--; ) {
128		shift = val = 0;
129		l = 8;
130		j = image->width;
131		dst = dst1;
132		s = src;
133
134		/* write leading bits */
135		if (start_index) {
136			u32 start_mask = ~(FB_SHIFT_HIGH(p, ~(u32)0,
137							 start_index));
138			val = *dst & start_mask;
139			shift = start_index;
140		}
141
142		while (j--) {
143			l--;
144			color = (*s & (1 << l)) ? fgcolor : bgcolor;
145			val |= FB_SHIFT_HIGH(p, color, shift);
146
147			/* Did the bitshift spill bits to the next long? */
148			if (shift >= null_bits) {
149				*dst++ = val;
150				val = (shift == null_bits) ? 0 :
151					FB_SHIFT_LOW(p, color, 32 - shift);
152			}
153			shift += bpp;
154			shift &= (32 - 1);
155			if (!l) { l = 8; s++; }
156		}
157
158		/* write trailing bits */
159 		if (shift) {
160			u32 end_mask = FB_SHIFT_HIGH(p, ~(u32)0, shift);
161
162			*dst &= end_mask;
163			*dst |= val;
164		}
165
166		dst1 += pitch;
167		src += spitch;
168		if (pitch_index) {
169			dst2 += pitch;
170			dst1 = (u8 *)((long)dst2 & ~(sizeof(u32) - 1));
171			start_index += pitch_index;
172			start_index &= 32 - 1;
173		}
174
175	}
176}
177
178/*
179 * fast_imageblit - optimized monochrome color expansion
180 *
181 * Only if:  bits_per_pixel == 8, 16, or 32
182 *           image->width is divisible by pixel/dword (ppw);
183 *           fix->line_legth is divisible by 4;
184 *           beginning and end of a scanline is dword aligned
185 */
186static void fast_imageblit(const struct fb_image *image, struct fb_info *p,
187				  void *dst1, u32 fgcolor, u32 bgcolor)
188{
189	u32 fgx = fgcolor, bgx = bgcolor, bpp = p->var.bits_per_pixel;
190	u32 ppw = 32/bpp, spitch = (image->width + 7)/8;
191	u32 bit_mask, eorx, shift;
192	const u8 *s = image->data, *src;
193	u32 *dst;
194	const u32 *tab;
195	size_t tablen;
196	u32 colortab[16];
197	int i, j, k;
198
199	switch (bpp) {
200	case 8:
201		tab = fb_be_math(p) ? cfb_tab8_be : cfb_tab8_le;
202		tablen = 16;
203		break;
204	case 16:
205		tab = fb_be_math(p) ? cfb_tab16_be : cfb_tab16_le;
206		tablen = 4;
207		break;
208	case 32:
209		tab = cfb_tab32;
210		tablen = 2;
211		break;
212	default:
213		return;
214	}
215
216	for (i = ppw-1; i--; ) {
217		fgx <<= bpp;
218		bgx <<= bpp;
219		fgx |= fgcolor;
220		bgx |= bgcolor;
221	}
222
223	bit_mask = (1 << ppw) - 1;
224	eorx = fgx ^ bgx;
225	k = image->width/ppw;
226
227	for (i = 0; i < tablen; ++i)
228		colortab[i] = (tab[i] & eorx) ^ bgx;
229
230	for (i = image->height; i--; ) {
231		dst = dst1;
232		shift = 8;
233		src = s;
234
235		/*
236		 * Manually unroll the per-line copying loop for better
237		 * performance. This works until we processed the last
238		 * completely filled source byte (inclusive).
239		 */
240		switch (ppw) {
241		case 4: /* 8 bpp */
242			for (j = k; j >= 2; j -= 2, ++src) {
243				*dst++ = colortab[(*src >> 4) & bit_mask];
244				*dst++ = colortab[(*src >> 0) & bit_mask];
245			}
246			break;
247		case 2: /* 16 bpp */
248			for (j = k; j >= 4; j -= 4, ++src) {
249				*dst++ = colortab[(*src >> 6) & bit_mask];
250				*dst++ = colortab[(*src >> 4) & bit_mask];
251				*dst++ = colortab[(*src >> 2) & bit_mask];
252				*dst++ = colortab[(*src >> 0) & bit_mask];
253			}
254			break;
255		case 1: /* 32 bpp */
256			for (j = k; j >= 8; j -= 8, ++src) {
257				*dst++ = colortab[(*src >> 7) & bit_mask];
258				*dst++ = colortab[(*src >> 6) & bit_mask];
259				*dst++ = colortab[(*src >> 5) & bit_mask];
260				*dst++ = colortab[(*src >> 4) & bit_mask];
261				*dst++ = colortab[(*src >> 3) & bit_mask];
262				*dst++ = colortab[(*src >> 2) & bit_mask];
263				*dst++ = colortab[(*src >> 1) & bit_mask];
264				*dst++ = colortab[(*src >> 0) & bit_mask];
265			}
266			break;
267		}
268
269		/*
270		 * For image widths that are not a multiple of 8, there
271		 * are trailing pixels left on the current line. Print
272		 * them as well.
273		 */
274		for (; j--; ) {
275			shift -= ppw;
276			*dst++ = colortab[(*src >> shift) & bit_mask];
277			if (!shift) {
278				shift = 8;
279				++src;
280			}
281		}
282
283		dst1 += p->fix.line_length;
284		s += spitch;
285	}
286}
287
288void sys_imageblit(struct fb_info *p, const struct fb_image *image)
289{
290	u32 fgcolor, bgcolor, start_index, bitstart, pitch_index = 0;
291	u32 bpl = sizeof(u32), bpp = p->var.bits_per_pixel;
292	u32 width = image->width;
293	u32 dx = image->dx, dy = image->dy;
294	void *dst1;
295
296	if (p->state != FBINFO_STATE_RUNNING)
297		return;
298
299	if (!(p->flags & FBINFO_VIRTFB))
300		fb_warn_once(p, "Framebuffer is not in virtual address space.");
301
302	bitstart = (dy * p->fix.line_length * 8) + (dx * bpp);
303	start_index = bitstart & (32 - 1);
304	pitch_index = (p->fix.line_length & (bpl - 1)) * 8;
305
306	bitstart /= 8;
307	bitstart &= ~(bpl - 1);
308	dst1 = (void __force *)p->screen_base + bitstart;
309
310	if (p->fbops->fb_sync)
311		p->fbops->fb_sync(p);
312
313	if (image->depth == 1) {
314		if (p->fix.visual == FB_VISUAL_TRUECOLOR ||
315		    p->fix.visual == FB_VISUAL_DIRECTCOLOR) {
316			fgcolor = ((u32*)(p->pseudo_palette))[image->fg_color];
317			bgcolor = ((u32*)(p->pseudo_palette))[image->bg_color];
318		} else {
319			fgcolor = image->fg_color;
320			bgcolor = image->bg_color;
321		}
322
323		if (32 % bpp == 0 && !start_index && !pitch_index &&
324		    ((width & (32/bpp-1)) == 0) &&
325		    bpp >= 8 && bpp <= 32)
326			fast_imageblit(image, p, dst1, fgcolor, bgcolor);
327		else
328			slow_imageblit(image, p, dst1, fgcolor, bgcolor,
329					start_index, pitch_index);
330	} else
331		color_imageblit(image, p, dst1, start_index, pitch_index);
332}
333
334EXPORT_SYMBOL(sys_imageblit);
335
336MODULE_AUTHOR("Antonino Daplas <adaplas@pol.net>");
337MODULE_DESCRIPTION("1-bit/8-bit to 1-32 bit color expansion (sys-to-sys)");
338MODULE_LICENSE("GPL");
339
340