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
|
/*
* Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg 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
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "../swscale_internal.h"
#include "libavutil/x86/asm.h"
#include "libavutil/x86/cpu.h"
#define RET 0xC3 // near return opcode for x86
#define PREFETCH "prefetchnta"
#if HAVE_INLINE_ASM
av_cold int ff_init_hscaler_mmxext(int dstW, int xInc, uint8_t *filterCode,
int16_t *filter, int32_t *filterPos,
int numSplits)
{
uint8_t *fragmentA;
x86_reg imm8OfPShufW1A;
x86_reg imm8OfPShufW2A;
x86_reg fragmentLengthA;
uint8_t *fragmentB;
x86_reg imm8OfPShufW1B;
x86_reg imm8OfPShufW2B;
x86_reg fragmentLengthB;
int fragmentPos;
int xpos, i;
// create an optimized horizontal scaling routine
/* This scaler is made of runtime-generated MMXEXT code using specially tuned
* pshufw instructions. For every four output pixels, if four input pixels
* are enough for the fast bilinear scaling, then a chunk of fragmentB is
* used. If five input pixels are needed, then a chunk of fragmentA is used.
*/
// code fragment
__asm__ volatile (
"jmp 9f \n\t"
// Begin
"0: \n\t"
"movq (%%"FF_REG_d", %%"FF_REG_a"), %%mm3 \n\t"
"movd (%%"FF_REG_c", %%"FF_REG_S"), %%mm0 \n\t"
"movd 1(%%"FF_REG_c", %%"FF_REG_S"), %%mm1 \n\t"
"punpcklbw %%mm7, %%mm1 \n\t"
"punpcklbw %%mm7, %%mm0 \n\t"
"pshufw $0xFF, %%mm1, %%mm1 \n\t"
"1: \n\t"
"pshufw $0xFF, %%mm0, %%mm0 \n\t"
"2: \n\t"
"psubw %%mm1, %%mm0 \n\t"
"movl 8(%%"FF_REG_b", %%"FF_REG_a"), %%esi \n\t"
"pmullw %%mm3, %%mm0 \n\t"
"psllw $7, %%mm1 \n\t"
"paddw %%mm1, %%mm0 \n\t"
"movq %%mm0, (%%"FF_REG_D", %%"FF_REG_a") \n\t"
"add $8, %%"FF_REG_a" \n\t"
// End
"9: \n\t"
"lea " LOCAL_MANGLE(0b) ", %0 \n\t"
"lea " LOCAL_MANGLE(1b) ", %1 \n\t"
"lea " LOCAL_MANGLE(2b) ", %2 \n\t"
"dec %1 \n\t"
"dec %2 \n\t"
"sub %0, %1 \n\t"
"sub %0, %2 \n\t"
"lea " LOCAL_MANGLE(9b) ", %3 \n\t"
"sub %0, %3 \n\t"
: "=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),
"=r" (fragmentLengthA)
);
__asm__ volatile (
"jmp 9f \n\t"
// Begin
"0: \n\t"
"movq (%%"FF_REG_d", %%"FF_REG_a"), %%mm3 \n\t"
"movd (%%"FF_REG_c", %%"FF_REG_S"), %%mm0 \n\t"
"punpcklbw %%mm7, %%mm0 \n\t"
"pshufw $0xFF, %%mm0, %%mm1 \n\t"
"1: \n\t"
"pshufw $0xFF, %%mm0, %%mm0 \n\t"
"2: \n\t"
"psubw %%mm1, %%mm0 \n\t"
"movl 8(%%"FF_REG_b", %%"FF_REG_a"), %%esi \n\t"
"pmullw %%mm3, %%mm0 \n\t"
"psllw $7, %%mm1 \n\t"
"paddw %%mm1, %%mm0 \n\t"
"movq %%mm0, (%%"FF_REG_D", %%"FF_REG_a") \n\t"
"add $8, %%"FF_REG_a" \n\t"
// End
"9: \n\t"
"lea " LOCAL_MANGLE(0b) ", %0 \n\t"
"lea " LOCAL_MANGLE(1b) ", %1 \n\t"
"lea " LOCAL_MANGLE(2b) ", %2 \n\t"
"dec %1 \n\t"
"dec %2 \n\t"
"sub %0, %1 \n\t"
"sub %0, %2 \n\t"
"lea " LOCAL_MANGLE(9b) ", %3 \n\t"
"sub %0, %3 \n\t"
: "=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),
"=r" (fragmentLengthB)
);
xpos = 0; // lumXInc/2 - 0x8000; // difference between pixel centers
fragmentPos = 0;
for (i = 0; i < dstW / numSplits; i++) {
int xx = xpos >> 16;
if ((i & 3) == 0) {
int a = 0;
int b = ((xpos + xInc) >> 16) - xx;
int c = ((xpos + xInc * 2) >> 16) - xx;
int d = ((xpos + xInc * 3) >> 16) - xx;
int inc = (d + 1 < 4);
uint8_t *fragment = inc ? fragmentB : fragmentA;
x86_reg imm8OfPShufW1 = inc ? imm8OfPShufW1B : imm8OfPShufW1A;
x86_reg imm8OfPShufW2 = inc ? imm8OfPShufW2B : imm8OfPShufW2A;
x86_reg fragmentLength = inc ? fragmentLengthB : fragmentLengthA;
int maxShift = 3 - (d + inc);
int shift = 0;
if (filterCode) {
filter[i] = ((xpos & 0xFFFF) ^ 0xFFFF) >> 9;
filter[i + 1] = (((xpos + xInc) & 0xFFFF) ^ 0xFFFF) >> 9;
filter[i + 2] = (((xpos + xInc * 2) & 0xFFFF) ^ 0xFFFF) >> 9;
filter[i + 3] = (((xpos + xInc * 3) & 0xFFFF) ^ 0xFFFF) >> 9;
filterPos[i / 2] = xx;
memcpy(filterCode + fragmentPos, fragment, fragmentLength);
filterCode[fragmentPos + imm8OfPShufW1] = (a + inc) |
((b + inc) << 2) |
((c + inc) << 4) |
((d + inc) << 6);
filterCode[fragmentPos + imm8OfPShufW2] = a | (b << 2) |
(c << 4) |
(d << 6);
if (i + 4 - inc >= dstW)
shift = maxShift; // avoid overread
else if ((filterPos[i / 2] & 3) <= maxShift)
shift = filterPos[i / 2] & 3; // align
if (shift && i >= shift) {
filterCode[fragmentPos + imm8OfPShufW1] += 0x55 * shift;
filterCode[fragmentPos + imm8OfPShufW2] += 0x55 * shift;
filterPos[i / 2] -= shift;
}
}
fragmentPos += fragmentLength;
if (filterCode)
filterCode[fragmentPos] = RET;
}
xpos += xInc;
}
if (filterCode)
filterPos[((i / 2) + 1) & (~1)] = xpos >> 16; // needed to jump to the next part
return fragmentPos + 1;
}
void ff_hyscale_fast_mmxext(SwsContext *c, int16_t *dst,
int dstWidth, const uint8_t *src,
int srcW, int xInc)
{
int32_t *filterPos = c->hLumFilterPos;
int16_t *filter = c->hLumFilter;
void *mmxextFilterCode = c->lumMmxextFilterCode;
int i;
#if ARCH_X86_64
uint64_t retsave;
#else
#if defined(PIC)
uint64_t ebxsave;
#endif
#endif
__asm__ volatile(
#if ARCH_X86_64
"mov -8(%%rsp), %%"FF_REG_a" \n\t"
"mov %%"FF_REG_a", %5 \n\t" // retsave
#else
#if defined(PIC)
"mov %%"FF_REG_b", %5 \n\t" // ebxsave
#endif
#endif
"pxor %%mm7, %%mm7 \n\t"
"mov %0, %%"FF_REG_c" \n\t"
"mov %1, %%"FF_REG_D" \n\t"
"mov %2, %%"FF_REG_d" \n\t"
"mov %3, %%"FF_REG_b" \n\t"
"xor %%"FF_REG_a", %%"FF_REG_a" \n\t" // i
PREFETCH" (%%"FF_REG_c") \n\t"
PREFETCH" 32(%%"FF_REG_c") \n\t"
PREFETCH" 64(%%"FF_REG_c") \n\t"
#if ARCH_X86_64
#define CALL_MMXEXT_FILTER_CODE \
"movl (%%"FF_REG_b"), %%esi \n\t"\
"call *%4 \n\t"\
"movl (%%"FF_REG_b", %%"FF_REG_a"), %%esi \n\t"\
"add %%"FF_REG_S", %%"FF_REG_c" \n\t"\
"add %%"FF_REG_a", %%"FF_REG_D" \n\t"\
"xor %%"FF_REG_a", %%"FF_REG_a" \n\t"\
#else
#define CALL_MMXEXT_FILTER_CODE \
"movl (%%"FF_REG_b"), %%esi \n\t"\
"call *%4 \n\t"\
"addl (%%"FF_REG_b", %%"FF_REG_a"), %%"FF_REG_c" \n\t"\
"add %%"FF_REG_a", %%"FF_REG_D" \n\t"\
"xor %%"FF_REG_a", %%"FF_REG_a" \n\t"\
#endif /* ARCH_X86_64 */
CALL_MMXEXT_FILTER_CODE
CALL_MMXEXT_FILTER_CODE
CALL_MMXEXT_FILTER_CODE
CALL_MMXEXT_FILTER_CODE
CALL_MMXEXT_FILTER_CODE
CALL_MMXEXT_FILTER_CODE
CALL_MMXEXT_FILTER_CODE
CALL_MMXEXT_FILTER_CODE
#if ARCH_X86_64
"mov %5, %%"FF_REG_a" \n\t"
"mov %%"FF_REG_a", -8(%%rsp) \n\t"
#else
#if defined(PIC)
"mov %5, %%"FF_REG_b" \n\t"
#endif
#endif
:: "m" (src), "m" (dst), "m" (filter), "m" (filterPos),
"m" (mmxextFilterCode)
#if ARCH_X86_64
,"m"(retsave)
#else
#if defined(PIC)
,"m" (ebxsave)
#endif
#endif
: "%"FF_REG_a, "%"FF_REG_c, "%"FF_REG_d, "%"FF_REG_S, "%"FF_REG_D
#if ARCH_X86_64 || !defined(PIC)
,"%"FF_REG_b
#endif
);
for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--)
dst[i] = src[srcW-1]*128;
}
void ff_hcscale_fast_mmxext(SwsContext *c, int16_t *dst1, int16_t *dst2,
int dstWidth, const uint8_t *src1,
const uint8_t *src2, int srcW, int xInc)
{
int32_t *filterPos = c->hChrFilterPos;
int16_t *filter = c->hChrFilter;
void *mmxextFilterCode = c->chrMmxextFilterCode;
int i;
#if ARCH_X86_64
DECLARE_ALIGNED(8, uint64_t, retsave);
#else
#if defined(PIC)
DECLARE_ALIGNED(8, uint64_t, ebxsave);
#endif
#endif
__asm__ volatile(
#if ARCH_X86_64
"mov -8(%%rsp), %%"FF_REG_a" \n\t"
"mov %%"FF_REG_a", %7 \n\t" // retsave
#else
#if defined(PIC)
"mov %%"FF_REG_b", %7 \n\t" // ebxsave
#endif
#endif
"pxor %%mm7, %%mm7 \n\t"
"mov %0, %%"FF_REG_c" \n\t"
"mov %1, %%"FF_REG_D" \n\t"
"mov %2, %%"FF_REG_d" \n\t"
"mov %3, %%"FF_REG_b" \n\t"
"xor %%"FF_REG_a", %%"FF_REG_a" \n\t" // i
PREFETCH" (%%"FF_REG_c") \n\t"
PREFETCH" 32(%%"FF_REG_c") \n\t"
PREFETCH" 64(%%"FF_REG_c") \n\t"
CALL_MMXEXT_FILTER_CODE
CALL_MMXEXT_FILTER_CODE
CALL_MMXEXT_FILTER_CODE
CALL_MMXEXT_FILTER_CODE
"xor %%"FF_REG_a", %%"FF_REG_a" \n\t" // i
"mov %5, %%"FF_REG_c" \n\t" // src2
"mov %6, %%"FF_REG_D" \n\t" // dst2
PREFETCH" (%%"FF_REG_c") \n\t"
PREFETCH" 32(%%"FF_REG_c") \n\t"
PREFETCH" 64(%%"FF_REG_c") \n\t"
CALL_MMXEXT_FILTER_CODE
CALL_MMXEXT_FILTER_CODE
CALL_MMXEXT_FILTER_CODE
CALL_MMXEXT_FILTER_CODE
#if ARCH_X86_64
"mov %7, %%"FF_REG_a" \n\t"
"mov %%"FF_REG_a", -8(%%rsp) \n\t"
#else
#if defined(PIC)
"mov %7, %%"FF_REG_b" \n\t"
#endif
#endif
:: "m" (src1), "m" (dst1), "m" (filter), "m" (filterPos),
"m" (mmxextFilterCode), "m" (src2), "m"(dst2)
#if ARCH_X86_64
,"m"(retsave)
#else
#if defined(PIC)
,"m" (ebxsave)
#endif
#endif
: "%"FF_REG_a, "%"FF_REG_c, "%"FF_REG_d, "%"FF_REG_S, "%"FF_REG_D
#if ARCH_X86_64 || !defined(PIC)
,"%"FF_REG_b
#endif
);
for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) {
dst1[i] = src1[srcW-1]*128;
dst2[i] = src2[srcW-1]*128;
}
}
#endif //HAVE_INLINE_ASM
|
