1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
|
// Copyright (C) 2003 Dolphin Project.
// This program is free software: you can redistribute it and/or modify
// it under the terms of the GNU General Public License as published by
// the Free Software Foundation, version 2.0.
// This program 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 General Public License 2.0 for more details.
// A copy of the GPL 2.0 should have been included with the program.
// If not, see http://www.gnu.org/licenses/
// Official SVN repository and contact information can be found at
// http://code.google.com/p/dolphin-emu/
#include "OCLTextureDecoder.h"
#include "OpenCL.h"
#include <fcntl.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <math.h>
#include <sys/types.h>
#include <sys/stat.h>
struct sDecoders
{
const char name[256]; // kernel name
cl_kernel kernel; // compute kernel
};
const char *Kernel = " \n\
kernel void DecodeI4(global uchar *dst, \n\
const global uchar *src, int width) \n\
{ \n\
int x = get_global_id(0) * 8, y = get_global_id(1) * 8; \n\
int srcOffset = x + y * width / 8; \n\
for (int iy = 0; iy < 8; iy++) \n\
{ \n\
uchar4 val = vload4(srcOffset, src); \n\
uchar8 res; \n\
res.even = (val >> 4) & 0x0F; \n\
res.odd = val & 0x0F; \n\
res |= res << 4; \n\
vstore8(res, 0, dst + ((y + iy)*width + x)); \n\
srcOffset++; \n\
} \n\
} \n\
\n\
kernel void DecodeI8(global uchar *dst, \n\
const global uchar *src, int width) \n\
{ \n\
int x = get_global_id(0) * 8, y = get_global_id(1) * 4; \n\
int srcOffset = ((x * 4) + (y * width)) / 8; \n\
for (int iy = 0; iy < 4; iy++) \n\
{ \n\
vstore8(vload8(srcOffset, src), \n\
0, dst + ((y + iy)*width + x)); \n\
srcOffset++; \n\
} \n\
} \n\
\n\
kernel void DecodeIA8(global uchar *dst, \n\
const global uchar *src, int width) \n\
{ \n\
int x = get_global_id(0) * 4, y = get_global_id(1) * 4; \n\
int srcOffset = ((x * 4) + (y * width)) / 4; \n\
for (int iy = 0; iy < 4; iy++) \n\
{ \n\
uchar8 val = vload8(srcOffset, src); \n\
uchar8 res; \n\
res.odd = val.even; \n\
res.even = val.odd; \n\
vstore8(res, 0, dst + ((y + iy)*width + x) * 2); \n\
srcOffset++; \n\
} \n\
} \n\
\n\
kernel void DecodeIA4(global uchar *dst, \n\
const global uchar *src, int width) \n\
{ \n\
int x = get_global_id(0) * 8, y = get_global_id(1) * 4; \n\
int srcOffset = ((x * 4) + (y * width)) / 8; \n\
for (int iy = 0; iy < 4; iy++) \n\
{ \n\
uchar8 val = vload8(srcOffset, src); \n\
uchar16 res; \n\
res.odd = (val >> 4) & 0x0F; \n\
res.even = val & 0x0F; \n\
res |= res << 4; \n\
vstore16(res, 0, dst + ((y + iy)*width + x) * 2); \n\
srcOffset++; \n\
} \n\
} \n\
\n\
kernel void DecodeRGBA8(global uchar *dst, \n\
const global uchar *src, int width) \n\
{ \n\
int x = get_global_id(0) * 4, y = get_global_id(1) * 4; \n\
int srcOffset = (x * 2) + (y * width) / 2; \n\
for (int iy = 0; iy < 4; iy++) \n\
{ \n\
uchar8 ar = vload8(srcOffset, src); \n\
uchar8 gb = vload8(srcOffset + 4, src); \n\
uchar16 res; \n\
res.even.even = gb.odd; \n\
res.even.odd = ar.odd; \n\
res.odd.even = gb.even; \n\
res.odd.odd = ar.even; \n\
vstore16(res, 0, dst + ((y + iy)*width + x) * 4); \n\
srcOffset++; \n\
} \n\
} \n\
\n\
kernel void DecodeRGB565(global ushort *dst, \n\
const global ushort *src, int width) \n\
{ \n\
int x = get_global_id(0) * 4, y = get_global_id(1) * 4; \n\
int srcOffset = x + (y * width) / 4; \n\
for (int iy = 0; iy < 4; iy++) \n\
{ \n\
ushort4 val = vload4(srcOffset, src); \n\
val = (val >> 8) | (val << 8); \n\
vstore4(val, 0, dst + ((y + iy)*width + x)); \n\
srcOffset++; \n\
} \n\
} \n\
\n\
kernel void DecodeRGB5A3(global uchar *dst, \n\
const global uchar *src, int width) \n\
{ \n\
int x = get_global_id(0) * 4, y = get_global_id(1) * 4; \n\
int srcOffset = x + (y * width) / 4; \n\
for (int iy = 0; iy < 4; iy++) \n\
{ \n\
ushort8 val = convert_ushort8(vload8(srcOffset, src));\n\
ushort4 vs = val.odd | (val.even << 8); \n\
uchar16 resNoAlpha; \n\
resNoAlpha.odd.odd = (uchar4)(0xFF); \n\
resNoAlpha.even.odd = convert_uchar4(vs >> 7) & 0xF8; \n\
resNoAlpha.odd.even = convert_uchar4(vs >> 2) & 0xF8; \n\
resNoAlpha.even.even = convert_uchar4(vs << 3) & 0xF8;\n\
// Better but cause color bleeding \n\
//resNoAlpha |= resNoAlpha >> 5; \n\
uchar16 resAlpha; \n\
resAlpha.even.odd = convert_uchar4(vs >> 8) & 0x0F; \n\
resAlpha.odd.even = convert_uchar4(vs >> 4) & 0x0F; \n\
resAlpha.even.even = convert_uchar4(vs) & 0x0F; \n\
resAlpha |= resNoAlpha << 4; \n\
resAlpha.odd.odd = convert_uchar4(vs >> 7) & 0xE0; \n\
resAlpha.odd.odd |= (resAlpha.odd.odd >> 3) \n\
| (resAlpha.odd.odd >> 6); \n\
uchar16 choice = (uchar16)((uchar4)(vs.s0 >> 8), \n\
(uchar4)(vs.s1 >> 8), \n\
(uchar4)(vs.s2 >> 8), \n\
(uchar4)(vs.s3 >> 8)); \n\
uchar16 res; \n\
res = select(resAlpha, resNoAlpha, choice); \n\
vstore16(res, 0, dst + ((y + iy)*width + x) * 4); \n\
srcOffset++; \n\
} \n\
} \n\
\n\
kernel void DecodeDXT(global ulong *dst, \n\
const global ulong *src, int width) \n\
{ // TODO: PLEASE NOTE THAT THIS CODE DOES NOT WORK \n\
int x = get_global_id(0) * 8, y = get_global_id(1) * 8; \n\
int srcOffset = ((x * 4) + (y * width)) / 8; \n\
for (int iy = 0; iy < 4; iy++) \n\
{ \n\
vstore8(vload8(srcOffset, src), \n\
0, dst + ((y + iy)*width + x)); \n\
srcOffset++; \n\
} \n\
} \n\
";
cl_program g_program;
// NULL terminated set of kernels
sDecoders Decoders[] = {
{"DecodeI4", NULL},
{"DecodeI8", NULL},
{"DecodeIA4", NULL},
{"DecodeIA8", NULL},
{"DecodeRGBA8", NULL},
{"DecodeRGB565", NULL},
{"DecodeRGB5A3", NULL},
{"DecodeDXT", NULL},
{"", NULL},
};
bool g_Inited = false;
cl_mem g_clsrc, g_cldst; // texture buffer memory objects
void TexDecoder_OpenCL_Initialize() {
#if defined(HAVE_OPENCL) && HAVE_OPENCL
if(!g_Inited)
{
if(!OpenCL::Initialize())
return;
g_program = OpenCL::CompileProgram(Kernel);
int i = 0;
while(strlen(Decoders[i].name) > 0) {
Decoders[i].kernel = OpenCL::CompileKernel(g_program, Decoders[i].name);
i++;
}
// Allocating maximal Wii texture size in advance, so that we don't have to allocate/deallocate per texture
g_clsrc = clCreateBuffer(OpenCL::GetContext(), CL_MEM_READ_ONLY , 1024 * 1024 * sizeof(u32), NULL, NULL);
g_cldst = clCreateBuffer(OpenCL::GetContext(), CL_MEM_WRITE_ONLY, 1024 * 1024 * sizeof(u32), NULL, NULL);
g_Inited = true;
}
#endif
}
void TexDecoder_OpenCL_Shutdown() {
#if defined(HAVE_OPENCL) && HAVE_OPENCL
if(g_clsrc)
clReleaseMemObject(g_clsrc);
if(g_cldst)
clReleaseMemObject(g_cldst);
#endif
}
PC_TexFormat TexDecoder_Decode_OpenCL(u8 *dst, const u8 *src, int width, int height, int texformat, int tlutaddr, int tlutfmt)
{
#if defined(HAVE_OPENCL) && HAVE_OPENCL
cl_int err;
cl_kernel kernelToRun = Decoders[0].kernel;
float sizeOfDst = sizeof(u8), sizeOfSrc = sizeof(u8), xSkip, ySkip;
PC_TexFormat formatResult;
switch(texformat)
{
case GX_TF_I4:
kernelToRun = Decoders[0].kernel;
sizeOfSrc = sizeof(u8) / 2.0f;
sizeOfDst = sizeof(u8);
xSkip = 8;
ySkip = 8;
formatResult = PC_TEX_FMT_I4_AS_I8;
break;
case GX_TF_I8:
kernelToRun = Decoders[1].kernel;
sizeOfSrc = sizeOfDst = sizeof(u8);
xSkip = 8;
ySkip = 4;
formatResult = PC_TEX_FMT_I8;
break;
case GX_TF_IA4:
// Maybe a cleaner way is needed
kernelToRun = Decoders[2].kernel;
sizeOfSrc = sizeof(u8);
sizeOfDst = sizeof(u16);
xSkip = 8;
ySkip = 4;
formatResult = PC_TEX_FMT_IA4_AS_IA8;
break;
case GX_TF_IA8:
kernelToRun = Decoders[3].kernel;
sizeOfSrc = sizeOfDst = sizeof(u16);
xSkip = 4;
ySkip = 4;
formatResult = PC_TEX_FMT_IA8;
break;
case GX_TF_RGBA8:
kernelToRun = Decoders[4].kernel;
sizeOfSrc = sizeOfDst = sizeof(u32);
xSkip = 4;
ySkip = 4;
formatResult = PC_TEX_FMT_BGRA32;
break;
case GX_TF_RGB565:
kernelToRun = Decoders[5].kernel;
sizeOfSrc = sizeOfDst = sizeof(u16);
xSkip = 4;
ySkip = 4;
formatResult = PC_TEX_FMT_RGB565;
break;
case GX_TF_RGB5A3:
kernelToRun = Decoders[6].kernel;
sizeOfSrc = sizeof(u16);
sizeOfDst = sizeof(u32);
xSkip = 4;
ySkip = 4;
formatResult = PC_TEX_FMT_BGRA32;
break;
case GX_TF_CMPR:
return PC_TEX_FMT_NONE; // <-- TODO: Fix CMPR
kernelToRun = Decoders[7].kernel;
sizeOfSrc = sizeOfDst = sizeof(u32);
xSkip = 8;
ySkip = 8;
formatResult = PC_TEX_FMT_BGRA32;
break;
default:
return PC_TEX_FMT_NONE;
}
clEnqueueWriteBuffer(OpenCL::GetCommandQueue(), g_clsrc, CL_TRUE, 0, (size_t)(width * height * sizeOfSrc), src, 0, NULL, NULL);
clSetKernelArg(kernelToRun, 0, sizeof(cl_mem), &g_cldst);
clSetKernelArg(kernelToRun, 1, sizeof(cl_mem), &g_clsrc);
clSetKernelArg(kernelToRun, 2, sizeof(cl_int), &width);
size_t global[] = { (size_t)(width / xSkip), (size_t)(height / ySkip) };
// No work-groups for now
/*
size_t local;
err = clGetKernelWorkGroupInfo(kernelToRun, OpenCL::device_id, CL_KERNEL_WORK_GROUP_SIZE, sizeof(local), &local, NULL);
if(err)
PanicAlert("Error obtaining work-group information");
*/
err = clEnqueueNDRangeKernel(OpenCL::GetCommandQueue(), kernelToRun, 2, NULL, global, NULL, 0, NULL, NULL);
if(err)
PanicAlert("Error queueing kernel");
clFinish(OpenCL::GetCommandQueue());
clEnqueueReadBuffer(OpenCL::GetCommandQueue(), g_cldst, CL_TRUE, 0, (size_t)(width * height * sizeOfDst), dst, 0, NULL, NULL);
return formatResult;
#else
return PC_TEX_FMT_NONE;
#endif
/* switch (texformat)
{
case GX_TF_C4:
if (tlutfmt == 2)
{
// Special decoding is required for TLUT format 5A3
for (int y = 0; y < height; y += 8)
for (int x = 0; x < width; x += 8)
for (int iy = 0; iy < 8; iy++, src += 4)
decodebytesC4_5A3_To_BGRA32((u32*)dst + (y + iy) * width + x, src, tlutaddr);
}
else
{
for (int y = 0; y < height; y += 8)
for (int x = 0; x < width; x += 8)
for (int iy = 0; iy < 8; iy++, src += 4)
decodebytesC4_To_Raw16((u16*)dst + (y + iy) * width + x, src, tlutaddr);
}
return GetPCFormatFromTLUTFormat(tlutfmt);
case GX_TF_C8:
if (tlutfmt == 2)
{
// Special decoding is required for TLUT format 5A3
for (int y = 0; y < height; y += 4)
for (int x = 0; x < width; x += 8)
for (int iy = 0; iy < 4; iy++, src += 8)
decodebytesC8_5A3_To_BGRA32((u32*)dst + (y + iy) * width + x, src, tlutaddr);
}
else
{
for (int y = 0; y < height; y += 4)
for (int x = 0; x < width; x += 8)
for (int iy = 0; iy < 4; iy++, src += 8)
decodebytesC8_To_Raw16((u16*)dst + (y + iy) * width + x, src, tlutaddr);
}
return GetPCFormatFromTLUTFormat(tlutfmt);
case GX_TF_C14X2:
if (tlutfmt == 2)
{
// Special decoding is required for TLUT format 5A3
for (int y = 0; y < height; y += 4)
for (int x = 0; x < width; x += 4)
for (int iy = 0; iy < 4; iy++, src += 8)
decodebytesC14X2_5A3_To_BGRA32((u32*)dst + (y + iy) * width + x, (u16*)src, tlutaddr);
}
else
{
for (int y = 0; y < height; y += 4)
for (int x = 0; x < width; x += 4)
for (int iy = 0; iy < 4; iy++, src += 8)
decodebytesC14X2_To_Raw16((u16*)dst + (y + iy) * width + x, (u16*)src, tlutaddr);
}
return GetPCFormatFromTLUTFormat(tlutfmt);
case GX_TF_RGB565:
{
for (int y = 0; y < height; y += 4)
for (int x = 0; x < width; x += 4)
for (int iy = 0; iy < 4; iy++, src += 8)
{
u16 *ptr = (u16 *)dst + (y + iy) * width + x;
u16 *s = (u16 *)src;
for(int j = 0; j < 4; j++)
*ptr++ = Common::swap16(*s++);
}
}
return PC_TEX_FMT_RGB565;
case GX_TF_RGB5A3:
{
for (int y = 0; y < height; y += 4)
for (int x = 0; x < width; x += 4)
for (int iy = 0; iy < 4; iy++, src += 8)
//decodebytesRGB5A3((u32*)dst+(y+iy)*width+x, (u16*)src, 4);
decodebytesRGB5A3((u32*)dst+(y+iy)*width+x, (u16*)src);
}
return PC_TEX_FMT_BGRA32;
case GX_TF_RGBA8: // speed critical
{
for (int y = 0; y < height; y += 4)
for (int x = 0; x < width; x += 4)
{
for (int iy = 0; iy < 4; iy++)
decodebytesARGB8_4((u32*)dst + (y+iy)*width + x, (u16*)src + 4 * iy, (u16*)src + 4 * iy + 16);
src += 64;
}
}
return PC_TEX_FMT_BGRA32;
case GX_TF_CMPR: // speed critical
// The metroid games use this format almost exclusively.
{
for (int y = 0; y < height; y += 8)
{
for (int x = 0; x < width; x += 8)
{
decodeDXTBlock((u32*)dst + y * width + x, (DXTBlock*)src, width);
src += sizeof(DXTBlock);
decodeDXTBlock((u32*)dst + y * width + x + 4, (DXTBlock*)src, width);
src += sizeof(DXTBlock);
decodeDXTBlock((u32*)dst + (y + 4) * width + x, (DXTBlock*)src, width);
src += sizeof(DXTBlock);
decodeDXTBlock((u32*)dst + (y + 4) * width + x + 4, (DXTBlock*)src, width);
src += sizeof(DXTBlock);
}
}
return PC_TEX_FMT_BGRA32;
}
}
*/
// The "copy" texture formats, too?
return PC_TEX_FMT_NONE;
}
|
