1 /* ScummVM - Graphic Adventure Engine
2 *
3 * ScummVM is the legal property of its developers, whose names
4 * are too numerous to list here. Please refer to the COPYRIGHT
5 * file distributed with this source distribution.
6 *
7 * This program is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public License
9 * as published by the Free Software Foundation; either version 2
10 * of the License, or (at your option) any later version.
11 *
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
20 *
21 */
22
23 #include "graphics/scaler/intern.h"
24 #include "graphics/scaler/scalebit.h"
25 #include "common/util.h"
26 #include "common/system.h"
27 #include "common/textconsole.h"
28
29 int gBitFormat = 565;
30
31 #ifdef USE_HQ_SCALERS
32 // RGB-to-YUV lookup table
33 extern "C" {
34
35 #ifdef USE_NASM
36 // NOTE: if your compiler uses different mangled names, add another
37 // condition here
38
39 #if !defined(_WIN32) && !defined(MACOSX) && !defined(__OS2__)
40 #define RGBtoYUV _RGBtoYUV
41 #define hqx_highbits _hqx_highbits
42 #define hqx_lowbits _hqx_lowbits
43 #define hqx_low2bits _hqx_low2bits
44 #define hqx_low3bits _hqx_low3bits
45 #define hqx_greenMask _hqx_greenMask
46 #define hqx_redBlueMask _hqx_redBlueMask
47 #define hqx_green_redBlue_Mask _hqx_green_redBlue_Mask
48 #endif
49
50 uint32 hqx_highbits = 0xF7DEF7DE;
51 uint32 hqx_lowbits = 0x0821;
52 uint32 hqx_low2bits = 0x0C63;
53 uint32 hqx_low3bits = 0x1CE7;
54 uint32 hqx_greenMask = 0;
55 uint32 hqx_redBlueMask = 0;
56 uint32 hqx_green_redBlue_Mask = 0;
57
58 #endif
59
60 /**
61 * 16bit RGB to YUV conversion table. This table is setup by InitLUT().
62 * Used by the hq scaler family.
63 *
64 * FIXME/TODO: The RGBtoYUV table sucks up 256 KB. This is bad.
65 * In addition we never free it...
66 *
67 * Note: a memory lookup table is *not* necessarily faster than computing
68 * these things on the fly, because of its size. The table together with
69 * the code, plus the input/output GFX data, may not fit in the cache on some
70 * systems, so main memory has to be accessed, which is about the worst thing
71 * that can happen to code which tries to be fast...
72 *
73 * So we should think about ways to get this smaller / removed. Maybe we can
74 * use the same technique employed by our MPEG code to reduce the size of the
75 * lookup table at the cost of some additional computations?
76 *
77 * Of course, the above is largely a conjecture, and the actual speed
78 * differences are likely to vary a lot between different architectures and
79 * CPUs.
80 */
81 uint32 *RGBtoYUV = 0;
82 }
83
InitLUT(Graphics::PixelFormat format)84 void InitLUT(Graphics::PixelFormat format) {
85 uint8 r, g, b;
86 int Y, u, v;
87
88 assert(format.bytesPerPixel == 2);
89
90 // Allocate the YUV/LUT buffers on the fly if needed.
91 if (RGBtoYUV == 0)
92 RGBtoYUV = (uint32 *)malloc(65536 * sizeof(uint32));
93
94 if (!RGBtoYUV)
95 error("[InitLUT] Cannot allocate memory for YUV/LUT buffers");
96
97 for (int color = 0; color < 65536; ++color) {
98 format.colorToRGB(color, r, g, b);
99 Y = (r + g + b) >> 2;
100 u = 128 + ((r - b) >> 2);
101 v = 128 + ((-r + 2 * g - b) >> 3);
102 RGBtoYUV[color] = (Y << 16) | (u << 8) | v;
103 }
104
105 #ifdef USE_NASM
106 hqx_lowbits = (1 << format.rShift) | (1 << format.gShift) | (1 << format.bShift),
107 hqx_low2bits = (3 << format.rShift) | (3 << format.gShift) | (3 << format.bShift),
108 hqx_low3bits = (7 << format.rShift) | (7 << format.gShift) | (7 << format.bShift),
109
110 hqx_highbits = format.RGBToColor(255, 255, 255) ^ hqx_lowbits;
111
112 // FIXME: The following code only does the right thing
113 // if the color order is RGB or BGR, i.e., green is in the middle.
114 hqx_greenMask = format.RGBToColor( 0, 255, 0);
115 hqx_redBlueMask = format.RGBToColor(255, 0, 255);
116
117 hqx_green_redBlue_Mask = (hqx_greenMask << 16) | hqx_redBlueMask;
118 #endif
119 }
120 #endif
121
122
123 /** Lookup table for the DotMatrix scaler. */
124 uint16 g_dotmatrix[16] = {0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0};
125
126 /** Init the scaler subsystem. */
InitScalers(uint32 BitFormat)127 void InitScalers(uint32 BitFormat) {
128 gBitFormat = BitFormat;
129
130 // FIXME: The pixelformat should be param to this function, not the bitformat.
131 // Until then, determine the pixelformat in other ways. Unfortunately,
132 // calling OSystem::getOverlayFormat() here might not be safe on all ports.
133 Graphics::PixelFormat format;
134 if (gBitFormat == 555) {
135 format = Graphics::createPixelFormat<555>();
136 } else if (gBitFormat == 565) {
137 format = Graphics::createPixelFormat<565>();
138 } else {
139 assert(g_system);
140 format = g_system->getOverlayFormat();
141 }
142
143 #ifdef USE_HQ_SCALERS
144 InitLUT(format);
145 #endif
146
147 // Build dotmatrix lookup table for the DotMatrix scaler.
148 g_dotmatrix[0] = g_dotmatrix[10] = format.RGBToColor( 0, 63, 0);
149 g_dotmatrix[1] = g_dotmatrix[11] = format.RGBToColor( 0, 0, 63);
150 g_dotmatrix[2] = g_dotmatrix[ 8] = format.RGBToColor(63, 0, 0);
151 g_dotmatrix[4] = g_dotmatrix[ 6] =
152 g_dotmatrix[12] = g_dotmatrix[14] = format.RGBToColor(63, 63, 63);
153 }
154
DestroyScalers()155 void DestroyScalers() {
156 #ifdef USE_HQ_SCALERS
157 free(RGBtoYUV);
158 RGBtoYUV = 0;
159 #endif
160 }
161
162
163 /**
164 * Trivial 'scaler' - in fact it doesn't do any scaling but just copies the
165 * source to the destination.
166 */
Normal1x(const uint8 * srcPtr,uint32 srcPitch,uint8 * dstPtr,uint32 dstPitch,int width,int height)167 void Normal1x(const uint8 *srcPtr, uint32 srcPitch, uint8 *dstPtr, uint32 dstPitch,
168 int width, int height) {
169 // Spot the case when it can all be done in 1 hit
170 if ((srcPitch == sizeof(uint16) * (uint)width) && (dstPitch == sizeof(uint16) * (uint)width)) {
171 memcpy(dstPtr, srcPtr, sizeof(uint16) * width * height);
172 return;
173 }
174 while (height--) {
175 memcpy(dstPtr, srcPtr, sizeof(uint16) * width);
176 srcPtr += srcPitch;
177 dstPtr += dstPitch;
178 }
179 }
180
181 #ifdef USE_SCALERS
182
183
184 #ifdef USE_ARM_SCALER_ASM
185 extern "C" void Normal2xARM(const uint8 *srcPtr,
186 uint32 srcPitch,
187 uint8 *dstPtr,
188 uint32 dstPitch,
189 int width,
190 int height);
191
Normal2x(const uint8 * srcPtr,uint32 srcPitch,uint8 * dstPtr,uint32 dstPitch,int width,int height)192 void Normal2x(const uint8 *srcPtr,
193 uint32 srcPitch,
194 uint8 *dstPtr,
195 uint32 dstPitch,
196 int width,
197 int height) {
198 Normal2xARM(srcPtr, srcPitch, dstPtr, dstPitch, width, height);
199 }
200
201 #else
202 /**
203 * Trivial nearest-neighbor 2x scaler.
204 */
Normal2x(const uint8 * srcPtr,uint32 srcPitch,uint8 * dstPtr,uint32 dstPitch,int width,int height)205 void Normal2x(const uint8 *srcPtr, uint32 srcPitch, uint8 *dstPtr, uint32 dstPitch,
206 int width, int height) {
207 uint8 *r;
208
209 assert(IS_ALIGNED(dstPtr, 4));
210 while (height--) {
211 r = dstPtr;
212 for (int i = 0; i < width; ++i, r += 4) {
213 uint32 color = *(((const uint16 *)srcPtr) + i);
214
215 color |= color << 16;
216
217 *(uint32 *)(r) = color;
218 *(uint32 *)(r + dstPitch) = color;
219 }
220 srcPtr += srcPitch;
221 dstPtr += dstPitch << 1;
222 }
223 }
224 #endif
225
226 /**
227 * Trivial nearest-neighbor 3x scaler.
228 */
Normal3x(const uint8 * srcPtr,uint32 srcPitch,uint8 * dstPtr,uint32 dstPitch,int width,int height)229 void Normal3x(const uint8 *srcPtr, uint32 srcPitch, uint8 *dstPtr, uint32 dstPitch,
230 int width, int height) {
231 uint8 *r;
232 const uint32 dstPitch2 = dstPitch * 2;
233 const uint32 dstPitch3 = dstPitch * 3;
234
235 assert(IS_ALIGNED(dstPtr, 2));
236 while (height--) {
237 r = dstPtr;
238 for (int i = 0; i < width; ++i, r += 6) {
239 uint16 color = *(((const uint16 *)srcPtr) + i);
240
241 *(uint16 *)(r + 0) = color;
242 *(uint16 *)(r + 2) = color;
243 *(uint16 *)(r + 4) = color;
244 *(uint16 *)(r + 0 + dstPitch) = color;
245 *(uint16 *)(r + 2 + dstPitch) = color;
246 *(uint16 *)(r + 4 + dstPitch) = color;
247 *(uint16 *)(r + 0 + dstPitch2) = color;
248 *(uint16 *)(r + 2 + dstPitch2) = color;
249 *(uint16 *)(r + 4 + dstPitch2) = color;
250 }
251 srcPtr += srcPitch;
252 dstPtr += dstPitch3;
253 }
254 }
255
256 #define interpolate_1_1 interpolate16_1_1<ColorMask>
257 #define interpolate_1_1_1_1 interpolate16_1_1_1_1<ColorMask>
258
259 /**
260 * Trivial nearest-neighbor 1.5x scaler.
261 */
262 template<typename ColorMask>
Normal1o5xTemplate(const uint8 * srcPtr,uint32 srcPitch,uint8 * dstPtr,uint32 dstPitch,int width,int height)263 void Normal1o5xTemplate(const uint8 *srcPtr, uint32 srcPitch, uint8 *dstPtr, uint32 dstPitch,
264 int width, int height) {
265 uint8 *r;
266 const uint32 dstPitch2 = dstPitch * 2;
267 const uint32 dstPitch3 = dstPitch * 3;
268 const uint32 srcPitch2 = srcPitch * 2;
269
270 assert(IS_ALIGNED(dstPtr, 2));
271 while (height > 0) {
272 r = dstPtr;
273 for (int i = 0; i < width; i += 2, r += 6) {
274 uint16 color0 = *(((const uint16 *)srcPtr) + i);
275 uint16 color1 = *(((const uint16 *)srcPtr) + i + 1);
276 uint16 color2 = *(((const uint16 *)(srcPtr + srcPitch)) + i);
277 uint16 color3 = *(((const uint16 *)(srcPtr + srcPitch)) + i + 1);
278
279 *(uint16 *)(r + 0) = color0;
280 *(uint16 *)(r + 2) = interpolate_1_1(color0, color1);
281 *(uint16 *)(r + 4) = color1;
282 *(uint16 *)(r + 0 + dstPitch) = interpolate_1_1(color0, color2);
283 *(uint16 *)(r + 2 + dstPitch) = interpolate_1_1_1_1(color0, color1, color2, color3);
284 *(uint16 *)(r + 4 + dstPitch) = interpolate_1_1(color1, color3);
285 *(uint16 *)(r + 0 + dstPitch2) = color2;
286 *(uint16 *)(r + 2 + dstPitch2) = interpolate_1_1(color2, color3);
287 *(uint16 *)(r + 4 + dstPitch2) = color3;
288 }
289 srcPtr += srcPitch2;
290 dstPtr += dstPitch3;
291 height -= 2;
292 }
293 }
294
Normal1o5x(const uint8 * srcPtr,uint32 srcPitch,uint8 * dstPtr,uint32 dstPitch,int width,int height)295 void Normal1o5x(const uint8 *srcPtr, uint32 srcPitch, uint8 *dstPtr, uint32 dstPitch, int width, int height) {
296 if (gBitFormat == 565)
297 Normal1o5xTemplate<Graphics::ColorMasks<565> >(srcPtr, srcPitch, dstPtr, dstPitch, width, height);
298 else
299 Normal1o5xTemplate<Graphics::ColorMasks<555> >(srcPtr, srcPitch, dstPtr, dstPitch, width, height);
300 }
301
302 /**
303 * The Scale2x filter, also known as AdvMame2x.
304 * See also http://scale2x.sourceforge.net
305 */
AdvMame2x(const uint8 * srcPtr,uint32 srcPitch,uint8 * dstPtr,uint32 dstPitch,int width,int height)306 void AdvMame2x(const uint8 *srcPtr, uint32 srcPitch, uint8 *dstPtr, uint32 dstPitch,
307 int width, int height) {
308 scale(2, dstPtr, dstPitch, srcPtr - srcPitch, srcPitch, 2, width, height);
309 }
310
311 /**
312 * The Scale3x filter, also known as AdvMame3x.
313 * See also http://scale2x.sourceforge.net
314 */
AdvMame3x(const uint8 * srcPtr,uint32 srcPitch,uint8 * dstPtr,uint32 dstPitch,int width,int height)315 void AdvMame3x(const uint8 *srcPtr, uint32 srcPitch, uint8 *dstPtr, uint32 dstPitch,
316 int width, int height) {
317 scale(3, dstPtr, dstPitch, srcPtr - srcPitch, srcPitch, 2, width, height);
318 }
319
320 template<typename ColorMask>
TV2xTemplate(const uint8 * srcPtr,uint32 srcPitch,uint8 * dstPtr,uint32 dstPitch,int width,int height)321 void TV2xTemplate(const uint8 *srcPtr, uint32 srcPitch, uint8 *dstPtr, uint32 dstPitch,
322 int width, int height) {
323 const uint32 nextlineSrc = srcPitch / sizeof(uint16);
324 const uint16 *p = (const uint16 *)srcPtr;
325
326 const uint32 nextlineDst = dstPitch / sizeof(uint16);
327 uint16 *q = (uint16 *)dstPtr;
328
329 while (height--) {
330 for (int i = 0, j = 0; i < width; ++i, j += 2) {
331 uint16 p1 = *(p + i);
332 uint32 pi;
333
334 pi = (((p1 & ColorMask::kRedBlueMask) * 7) >> 3) & ColorMask::kRedBlueMask;
335 pi |= (((p1 & ColorMask::kGreenMask) * 7) >> 3) & ColorMask::kGreenMask;
336
337 *(q + j) = p1;
338 *(q + j + 1) = p1;
339 *(q + j + nextlineDst) = (uint16)pi;
340 *(q + j + nextlineDst + 1) = (uint16)pi;
341 }
342 p += nextlineSrc;
343 q += nextlineDst << 1;
344 }
345 }
346
TV2x(const uint8 * srcPtr,uint32 srcPitch,uint8 * dstPtr,uint32 dstPitch,int width,int height)347 void TV2x(const uint8 *srcPtr, uint32 srcPitch, uint8 *dstPtr, uint32 dstPitch, int width, int height) {
348 if (gBitFormat == 565)
349 TV2xTemplate<Graphics::ColorMasks<565> >(srcPtr, srcPitch, dstPtr, dstPitch, width, height);
350 else
351 TV2xTemplate<Graphics::ColorMasks<555> >(srcPtr, srcPitch, dstPtr, dstPitch, width, height);
352 }
353
DOT_16(const uint16 * dotmatrix,uint16 c,int j,int i)354 static inline uint16 DOT_16(const uint16 *dotmatrix, uint16 c, int j, int i) {
355 return c - ((c >> 2) & dotmatrix[((j & 3) << 2) + (i & 3)]);
356 }
357
358
359 // FIXME: This scaler doesn't quite work. Either it needs to know where on the
360 // screen it's drawing, or the dirty rects will have to be adjusted so that
361 // access to the dotmatrix array are made in a consistent way. (Doing that in
362 // a way that also works together with aspect-ratio correction is left as an
363 // exercise for the reader.)
364
DotMatrix(const uint8 * srcPtr,uint32 srcPitch,uint8 * dstPtr,uint32 dstPitch,int width,int height)365 void DotMatrix(const uint8 *srcPtr, uint32 srcPitch, uint8 *dstPtr, uint32 dstPitch,
366 int width, int height) {
367
368 const uint16 *dotmatrix = g_dotmatrix;
369
370 const uint32 nextlineSrc = srcPitch / sizeof(uint16);
371 const uint16 *p = (const uint16 *)srcPtr;
372
373 const uint32 nextlineDst = dstPitch / sizeof(uint16);
374 uint16 *q = (uint16 *)dstPtr;
375
376 for (int j = 0, jj = 0; j < height; ++j, jj += 2) {
377 for (int i = 0, ii = 0; i < width; ++i, ii += 2) {
378 uint16 c = *(p + i);
379 *(q + ii) = DOT_16(dotmatrix, c, jj, ii);
380 *(q + ii + 1) = DOT_16(dotmatrix, c, jj, ii + 1);
381 *(q + ii + nextlineDst) = DOT_16(dotmatrix, c, jj + 1, ii);
382 *(q + ii + nextlineDst + 1) = DOT_16(dotmatrix, c, jj + 1, ii + 1);
383 }
384 p += nextlineSrc;
385 q += nextlineDst << 1;
386 }
387 }
388
389 #endif // #ifdef USE_SCALERS
390