Mercurial > libavcodec.hg
annotate x86/idct_sse2_xvid.c @ 11642:52bc4e605761 libavcodec
vertical align in libx264.c
author | darkshikari |
---|---|
date | Mon, 19 Apr 2010 21:20:20 +0000 |
parents | 34a65026fa06 |
children | 7dd2a45249a9 |
rev | line source |
---|---|
8430 | 1 /* |
2 * XVID MPEG-4 VIDEO CODEC | |
3 * - SSE2 inverse discrete cosine transform - | |
4 * | |
5 * Copyright(C) 2003 Pascal Massimino <skal@planet-d.net> | |
6 * | |
7 * Conversion to gcc syntax with modifications | |
8 * by Alexander Strange <astrange@ithinksw.com> | |
9 * | |
10 * Originally from dct/x86_asm/fdct_sse2_skal.asm in Xvid. | |
11 * | |
12 * This file is part of FFmpeg. | |
13 * | |
14 * Vertical pass is an implementation of the scheme: | |
15 * Loeffler C., Ligtenberg A., and Moschytz C.S.: | |
16 * Practical Fast 1D DCT Algorithm with Eleven Multiplications, | |
17 * Proc. ICASSP 1989, 988-991. | |
18 * | |
19 * Horizontal pass is a double 4x4 vector/matrix multiplication, | |
20 * (see also Intel's Application Note 922: | |
21 * http://developer.intel.com/vtune/cbts/strmsimd/922down.htm | |
22 * Copyright (C) 1999 Intel Corporation) | |
23 * | |
24 * More details at http://skal.planet-d.net/coding/dct.html | |
25 * | |
26 * FFmpeg is free software; you can redistribute it and/or | |
27 * modify it under the terms of the GNU Lesser General Public | |
28 * License as published by the Free Software Foundation; either | |
29 * version 2.1 of the License, or (at your option) any later version. | |
30 * | |
31 * FFmpeg is distributed in the hope that it will be useful, | |
32 * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
33 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
34 * Lesser General Public License for more details. | |
35 * | |
36 * You should have received a copy of the GNU Lesser General Public License | |
37 * along with FFmpeg; if not, write to the Free Software Foundation, | |
38 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | |
39 */ | |
40 | |
41 #include "libavcodec/dsputil.h" | |
42 #include "idct_xvid.h" | |
10114
8b9fc0c8f1cc
Move declarations of some mmx functions to dsputil_mmx.h
mru
parents:
8718
diff
changeset
|
43 #include "dsputil_mmx.h" |
8430 | 44 |
45 /*! | |
8718
e9d9d946f213
Use full internal pathname in doxygen @file directives.
diego
parents:
8590
diff
changeset
|
46 * @file libavcodec/x86/idct_sse2_xvid.c |
8430 | 47 * @brief SSE2 idct compatible with xvidmmx |
48 */ | |
49 | |
50 #define X8(x) x,x,x,x,x,x,x,x | |
51 | |
52 #define ROW_SHIFT 11 | |
53 #define COL_SHIFT 6 | |
54 | |
10961
34a65026fa06
Move array specifiers outside DECLARE_ALIGNED() invocations
mru
parents:
10114
diff
changeset
|
55 DECLARE_ASM_CONST(16, int16_t, tan1)[] = {X8(13036)}; // tan( pi/16) |
34a65026fa06
Move array specifiers outside DECLARE_ALIGNED() invocations
mru
parents:
10114
diff
changeset
|
56 DECLARE_ASM_CONST(16, int16_t, tan2)[] = {X8(27146)}; // tan(2pi/16) = sqrt(2)-1 |
34a65026fa06
Move array specifiers outside DECLARE_ALIGNED() invocations
mru
parents:
10114
diff
changeset
|
57 DECLARE_ASM_CONST(16, int16_t, tan3)[] = {X8(43790)}; // tan(3pi/16)-1 |
34a65026fa06
Move array specifiers outside DECLARE_ALIGNED() invocations
mru
parents:
10114
diff
changeset
|
58 DECLARE_ASM_CONST(16, int16_t, sqrt2)[]= {X8(23170)}; // 0.5/sqrt(2) |
34a65026fa06
Move array specifiers outside DECLARE_ALIGNED() invocations
mru
parents:
10114
diff
changeset
|
59 DECLARE_ASM_CONST(8, uint8_t, m127)[] = {X8(127)}; |
8430 | 60 |
10961
34a65026fa06
Move array specifiers outside DECLARE_ALIGNED() invocations
mru
parents:
10114
diff
changeset
|
61 DECLARE_ASM_CONST(16, int16_t, iTab1)[] = { |
8430 | 62 0x4000, 0x539f, 0xc000, 0xac61, 0x4000, 0xdd5d, 0x4000, 0xdd5d, |
63 0x4000, 0x22a3, 0x4000, 0x22a3, 0xc000, 0x539f, 0x4000, 0xac61, | |
64 0x3249, 0x11a8, 0x4b42, 0xee58, 0x11a8, 0x4b42, 0x11a8, 0xcdb7, | |
65 0x58c5, 0x4b42, 0xa73b, 0xcdb7, 0x3249, 0xa73b, 0x4b42, 0xa73b | |
66 }; | |
67 | |
10961
34a65026fa06
Move array specifiers outside DECLARE_ALIGNED() invocations
mru
parents:
10114
diff
changeset
|
68 DECLARE_ASM_CONST(16, int16_t, iTab2)[] = { |
8430 | 69 0x58c5, 0x73fc, 0xa73b, 0x8c04, 0x58c5, 0xcff5, 0x58c5, 0xcff5, |
70 0x58c5, 0x300b, 0x58c5, 0x300b, 0xa73b, 0x73fc, 0x58c5, 0x8c04, | |
71 0x45bf, 0x187e, 0x6862, 0xe782, 0x187e, 0x6862, 0x187e, 0xba41, | |
72 0x7b21, 0x6862, 0x84df, 0xba41, 0x45bf, 0x84df, 0x6862, 0x84df | |
73 }; | |
74 | |
10961
34a65026fa06
Move array specifiers outside DECLARE_ALIGNED() invocations
mru
parents:
10114
diff
changeset
|
75 DECLARE_ASM_CONST(16, int16_t, iTab3)[] = { |
8430 | 76 0x539f, 0x6d41, 0xac61, 0x92bf, 0x539f, 0xd2bf, 0x539f, 0xd2bf, |
77 0x539f, 0x2d41, 0x539f, 0x2d41, 0xac61, 0x6d41, 0x539f, 0x92bf, | |
78 0x41b3, 0x1712, 0x6254, 0xe8ee, 0x1712, 0x6254, 0x1712, 0xbe4d, | |
79 0x73fc, 0x6254, 0x8c04, 0xbe4d, 0x41b3, 0x8c04, 0x6254, 0x8c04 | |
80 }; | |
81 | |
10961
34a65026fa06
Move array specifiers outside DECLARE_ALIGNED() invocations
mru
parents:
10114
diff
changeset
|
82 DECLARE_ASM_CONST(16, int16_t, iTab4)[] = { |
8430 | 83 0x4b42, 0x6254, 0xb4be, 0x9dac, 0x4b42, 0xd746, 0x4b42, 0xd746, |
84 0x4b42, 0x28ba, 0x4b42, 0x28ba, 0xb4be, 0x6254, 0x4b42, 0x9dac, | |
85 0x3b21, 0x14c3, 0x587e, 0xeb3d, 0x14c3, 0x587e, 0x14c3, 0xc4df, | |
86 0x6862, 0x587e, 0x979e, 0xc4df, 0x3b21, 0x979e, 0x587e, 0x979e | |
87 }; | |
88 | |
10961
34a65026fa06
Move array specifiers outside DECLARE_ALIGNED() invocations
mru
parents:
10114
diff
changeset
|
89 DECLARE_ASM_CONST(16, int32_t, walkenIdctRounders)[] = { |
8430 | 90 65536, 65536, 65536, 65536, |
91 3597, 3597, 3597, 3597, | |
92 2260, 2260, 2260, 2260, | |
93 1203, 1203, 1203, 1203, | |
94 120, 120, 120, 120, | |
95 512, 512, 512, 512 | |
96 }; | |
97 | |
98 // Temporary storage before the column pass | |
99 #define ROW1 "%%xmm6" | |
100 #define ROW3 "%%xmm4" | |
101 #define ROW5 "%%xmm5" | |
102 #define ROW7 "%%xmm7" | |
103 | |
104 #define CLEAR_ODD(r) "pxor "r","r" \n\t" | |
105 #define PUT_ODD(dst) "pshufhw $0x1B, %%xmm2, "dst" \n\t" | |
106 | |
8590 | 107 #if ARCH_X86_64 |
8430 | 108 |
109 # define ROW0 "%%xmm8" | |
110 # define REG0 ROW0 | |
111 # define ROW2 "%%xmm9" | |
112 # define REG2 ROW2 | |
113 # define ROW4 "%%xmm10" | |
114 # define REG4 ROW4 | |
115 # define ROW6 "%%xmm11" | |
116 # define REG6 ROW6 | |
117 # define CLEAR_EVEN(r) CLEAR_ODD(r) | |
118 # define PUT_EVEN(dst) PUT_ODD(dst) | |
119 # define XMMS "%%xmm12" | |
120 # define MOV_32_ONLY "#" | |
121 # define SREG2 REG2 | |
122 # define TAN3 "%%xmm13" | |
123 # define TAN1 "%%xmm14" | |
124 | |
125 #else | |
126 | |
127 # define ROW0 "(%0)" | |
128 # define REG0 "%%xmm4" | |
129 # define ROW2 "2*16(%0)" | |
130 # define REG2 "%%xmm4" | |
131 # define ROW4 "4*16(%0)" | |
132 # define REG4 "%%xmm6" | |
133 # define ROW6 "6*16(%0)" | |
134 # define REG6 "%%xmm6" | |
135 # define CLEAR_EVEN(r) | |
136 # define PUT_EVEN(dst) \ | |
137 "pshufhw $0x1B, %%xmm2, %%xmm2 \n\t" \ | |
138 "movdqa %%xmm2, "dst" \n\t" | |
139 # define XMMS "%%xmm2" | |
140 # define MOV_32_ONLY "movdqa " | |
141 # define SREG2 "%%xmm7" | |
142 # define TAN3 "%%xmm0" | |
143 # define TAN1 "%%xmm2" | |
144 | |
145 #endif | |
146 | |
147 #define ROUND(x) "paddd "MANGLE(x) | |
148 | |
149 #define JZ(reg, to) \ | |
150 "testl "reg","reg" \n\t" \ | |
151 "jz "to" \n\t" | |
152 | |
153 #define JNZ(reg, to) \ | |
154 "testl "reg","reg" \n\t" \ | |
155 "jnz "to" \n\t" | |
156 | |
157 #define TEST_ONE_ROW(src, reg, clear) \ | |
158 clear \ | |
159 "movq "src", %%mm1 \n\t" \ | |
160 "por 8+"src", %%mm1 \n\t" \ | |
161 "paddusb %%mm0, %%mm1 \n\t" \ | |
162 "pmovmskb %%mm1, "reg" \n\t" | |
163 | |
164 #define TEST_TWO_ROWS(row1, row2, reg1, reg2, clear1, clear2) \ | |
165 clear1 \ | |
166 clear2 \ | |
167 "movq "row1", %%mm1 \n\t" \ | |
168 "por 8+"row1", %%mm1 \n\t" \ | |
169 "movq "row2", %%mm2 \n\t" \ | |
170 "por 8+"row2", %%mm2 \n\t" \ | |
171 "paddusb %%mm0, %%mm1 \n\t" \ | |
172 "paddusb %%mm0, %%mm2 \n\t" \ | |
173 "pmovmskb %%mm1, "reg1" \n\t" \ | |
174 "pmovmskb %%mm2, "reg2" \n\t" | |
175 | |
176 ///IDCT pass on rows. | |
177 #define iMTX_MULT(src, table, rounder, put) \ | |
178 "movdqa "src", %%xmm3 \n\t" \ | |
179 "movdqa %%xmm3, %%xmm0 \n\t" \ | |
180 "pshufd $0x11, %%xmm3, %%xmm1 \n\t" /* 4602 */ \ | |
181 "punpcklqdq %%xmm0, %%xmm0 \n\t" /* 0246 */ \ | |
182 "pmaddwd "table", %%xmm0 \n\t" \ | |
183 "pmaddwd 16+"table", %%xmm1 \n\t" \ | |
184 "pshufd $0xBB, %%xmm3, %%xmm2 \n\t" /* 5713 */ \ | |
185 "punpckhqdq %%xmm3, %%xmm3 \n\t" /* 1357 */ \ | |
186 "pmaddwd 32+"table", %%xmm2 \n\t" \ | |
187 "pmaddwd 48+"table", %%xmm3 \n\t" \ | |
188 "paddd %%xmm1, %%xmm0 \n\t" \ | |
189 "paddd %%xmm3, %%xmm2 \n\t" \ | |
190 rounder", %%xmm0 \n\t" \ | |
191 "movdqa %%xmm2, %%xmm3 \n\t" \ | |
192 "paddd %%xmm0, %%xmm2 \n\t" \ | |
193 "psubd %%xmm3, %%xmm0 \n\t" \ | |
194 "psrad $11, %%xmm2 \n\t" \ | |
195 "psrad $11, %%xmm0 \n\t" \ | |
196 "packssdw %%xmm0, %%xmm2 \n\t" \ | |
197 put \ | |
198 "1: \n\t" | |
199 | |
200 #define iLLM_HEAD \ | |
201 "movdqa "MANGLE(tan3)", "TAN3" \n\t" \ | |
202 "movdqa "MANGLE(tan1)", "TAN1" \n\t" \ | |
203 | |
204 ///IDCT pass on columns. | |
205 #define iLLM_PASS(dct) \ | |
206 "movdqa "TAN3", %%xmm1 \n\t" \ | |
207 "movdqa "TAN1", %%xmm3 \n\t" \ | |
208 "pmulhw %%xmm4, "TAN3" \n\t" \ | |
209 "pmulhw %%xmm5, %%xmm1 \n\t" \ | |
210 "paddsw %%xmm4, "TAN3" \n\t" \ | |
211 "paddsw %%xmm5, %%xmm1 \n\t" \ | |
212 "psubsw %%xmm5, "TAN3" \n\t" \ | |
213 "paddsw %%xmm4, %%xmm1 \n\t" \ | |
214 "pmulhw %%xmm7, %%xmm3 \n\t" \ | |
215 "pmulhw %%xmm6, "TAN1" \n\t" \ | |
216 "paddsw %%xmm6, %%xmm3 \n\t" \ | |
217 "psubsw %%xmm7, "TAN1" \n\t" \ | |
218 "movdqa %%xmm3, %%xmm7 \n\t" \ | |
219 "movdqa "TAN1", %%xmm6 \n\t" \ | |
220 "psubsw %%xmm1, %%xmm3 \n\t" \ | |
221 "psubsw "TAN3", "TAN1" \n\t" \ | |
222 "paddsw %%xmm7, %%xmm1 \n\t" \ | |
223 "paddsw %%xmm6, "TAN3" \n\t" \ | |
224 "movdqa %%xmm3, %%xmm6 \n\t" \ | |
225 "psubsw "TAN3", %%xmm3 \n\t" \ | |
226 "paddsw %%xmm6, "TAN3" \n\t" \ | |
227 "movdqa "MANGLE(sqrt2)", %%xmm4 \n\t" \ | |
228 "pmulhw %%xmm4, %%xmm3 \n\t" \ | |
229 "pmulhw %%xmm4, "TAN3" \n\t" \ | |
230 "paddsw "TAN3", "TAN3" \n\t" \ | |
231 "paddsw %%xmm3, %%xmm3 \n\t" \ | |
232 "movdqa "MANGLE(tan2)", %%xmm7 \n\t" \ | |
233 MOV_32_ONLY ROW2", "REG2" \n\t" \ | |
234 MOV_32_ONLY ROW6", "REG6" \n\t" \ | |
235 "movdqa %%xmm7, %%xmm5 \n\t" \ | |
236 "pmulhw "REG6", %%xmm7 \n\t" \ | |
237 "pmulhw "REG2", %%xmm5 \n\t" \ | |
238 "paddsw "REG2", %%xmm7 \n\t" \ | |
239 "psubsw "REG6", %%xmm5 \n\t" \ | |
240 MOV_32_ONLY ROW0", "REG0" \n\t" \ | |
241 MOV_32_ONLY ROW4", "REG4" \n\t" \ | |
242 MOV_32_ONLY" "TAN1", (%0) \n\t" \ | |
243 "movdqa "REG0", "XMMS" \n\t" \ | |
244 "psubsw "REG4", "REG0" \n\t" \ | |
245 "paddsw "XMMS", "REG4" \n\t" \ | |
246 "movdqa "REG4", "XMMS" \n\t" \ | |
247 "psubsw %%xmm7, "REG4" \n\t" \ | |
248 "paddsw "XMMS", %%xmm7 \n\t" \ | |
249 "movdqa "REG0", "XMMS" \n\t" \ | |
250 "psubsw %%xmm5, "REG0" \n\t" \ | |
251 "paddsw "XMMS", %%xmm5 \n\t" \ | |
252 "movdqa %%xmm5, "XMMS" \n\t" \ | |
253 "psubsw "TAN3", %%xmm5 \n\t" \ | |
254 "paddsw "XMMS", "TAN3" \n\t" \ | |
255 "movdqa "REG0", "XMMS" \n\t" \ | |
256 "psubsw %%xmm3, "REG0" \n\t" \ | |
257 "paddsw "XMMS", %%xmm3 \n\t" \ | |
258 MOV_32_ONLY" (%0), "TAN1" \n\t" \ | |
259 "psraw $6, %%xmm5 \n\t" \ | |
260 "psraw $6, "REG0" \n\t" \ | |
261 "psraw $6, "TAN3" \n\t" \ | |
262 "psraw $6, %%xmm3 \n\t" \ | |
263 "movdqa "TAN3", 1*16("dct") \n\t" \ | |
264 "movdqa %%xmm3, 2*16("dct") \n\t" \ | |
265 "movdqa "REG0", 5*16("dct") \n\t" \ | |
266 "movdqa %%xmm5, 6*16("dct") \n\t" \ | |
267 "movdqa %%xmm7, %%xmm0 \n\t" \ | |
268 "movdqa "REG4", %%xmm4 \n\t" \ | |
269 "psubsw %%xmm1, %%xmm7 \n\t" \ | |
270 "psubsw "TAN1", "REG4" \n\t" \ | |
271 "paddsw %%xmm0, %%xmm1 \n\t" \ | |
272 "paddsw %%xmm4, "TAN1" \n\t" \ | |
273 "psraw $6, %%xmm1 \n\t" \ | |
274 "psraw $6, %%xmm7 \n\t" \ | |
275 "psraw $6, "TAN1" \n\t" \ | |
276 "psraw $6, "REG4" \n\t" \ | |
277 "movdqa %%xmm1, ("dct") \n\t" \ | |
278 "movdqa "TAN1", 3*16("dct") \n\t" \ | |
279 "movdqa "REG4", 4*16("dct") \n\t" \ | |
280 "movdqa %%xmm7, 7*16("dct") \n\t" | |
281 | |
282 ///IDCT pass on columns, assuming rows 4-7 are zero. | |
283 #define iLLM_PASS_SPARSE(dct) \ | |
284 "pmulhw %%xmm4, "TAN3" \n\t" \ | |
285 "paddsw %%xmm4, "TAN3" \n\t" \ | |
286 "movdqa %%xmm6, %%xmm3 \n\t" \ | |
287 "pmulhw %%xmm6, "TAN1" \n\t" \ | |
288 "movdqa %%xmm4, %%xmm1 \n\t" \ | |
289 "psubsw %%xmm1, %%xmm3 \n\t" \ | |
290 "paddsw %%xmm6, %%xmm1 \n\t" \ | |
291 "movdqa "TAN1", %%xmm6 \n\t" \ | |
292 "psubsw "TAN3", "TAN1" \n\t" \ | |
293 "paddsw %%xmm6, "TAN3" \n\t" \ | |
294 "movdqa %%xmm3, %%xmm6 \n\t" \ | |
295 "psubsw "TAN3", %%xmm3 \n\t" \ | |
296 "paddsw %%xmm6, "TAN3" \n\t" \ | |
297 "movdqa "MANGLE(sqrt2)", %%xmm4 \n\t" \ | |
298 "pmulhw %%xmm4, %%xmm3 \n\t" \ | |
299 "pmulhw %%xmm4, "TAN3" \n\t" \ | |
300 "paddsw "TAN3", "TAN3" \n\t" \ | |
301 "paddsw %%xmm3, %%xmm3 \n\t" \ | |
302 "movdqa "MANGLE(tan2)", %%xmm5 \n\t" \ | |
303 MOV_32_ONLY ROW2", "SREG2" \n\t" \ | |
304 "pmulhw "SREG2", %%xmm5 \n\t" \ | |
305 MOV_32_ONLY ROW0", "REG0" \n\t" \ | |
306 "movdqa "REG0", %%xmm6 \n\t" \ | |
307 "psubsw "SREG2", %%xmm6 \n\t" \ | |
308 "paddsw "REG0", "SREG2" \n\t" \ | |
309 MOV_32_ONLY" "TAN1", (%0) \n\t" \ | |
310 "movdqa "REG0", "XMMS" \n\t" \ | |
311 "psubsw %%xmm5, "REG0" \n\t" \ | |
312 "paddsw "XMMS", %%xmm5 \n\t" \ | |
313 "movdqa %%xmm5, "XMMS" \n\t" \ | |
314 "psubsw "TAN3", %%xmm5 \n\t" \ | |
315 "paddsw "XMMS", "TAN3" \n\t" \ | |
316 "movdqa "REG0", "XMMS" \n\t" \ | |
317 "psubsw %%xmm3, "REG0" \n\t" \ | |
318 "paddsw "XMMS", %%xmm3 \n\t" \ | |
319 MOV_32_ONLY" (%0), "TAN1" \n\t" \ | |
320 "psraw $6, %%xmm5 \n\t" \ | |
321 "psraw $6, "REG0" \n\t" \ | |
322 "psraw $6, "TAN3" \n\t" \ | |
323 "psraw $6, %%xmm3 \n\t" \ | |
324 "movdqa "TAN3", 1*16("dct") \n\t" \ | |
325 "movdqa %%xmm3, 2*16("dct") \n\t" \ | |
326 "movdqa "REG0", 5*16("dct") \n\t" \ | |
327 "movdqa %%xmm5, 6*16("dct") \n\t" \ | |
328 "movdqa "SREG2", %%xmm0 \n\t" \ | |
329 "movdqa %%xmm6, %%xmm4 \n\t" \ | |
330 "psubsw %%xmm1, "SREG2" \n\t" \ | |
331 "psubsw "TAN1", %%xmm6 \n\t" \ | |
332 "paddsw %%xmm0, %%xmm1 \n\t" \ | |
333 "paddsw %%xmm4, "TAN1" \n\t" \ | |
334 "psraw $6, %%xmm1 \n\t" \ | |
335 "psraw $6, "SREG2" \n\t" \ | |
336 "psraw $6, "TAN1" \n\t" \ | |
337 "psraw $6, %%xmm6 \n\t" \ | |
338 "movdqa %%xmm1, ("dct") \n\t" \ | |
339 "movdqa "TAN1", 3*16("dct") \n\t" \ | |
340 "movdqa %%xmm6, 4*16("dct") \n\t" \ | |
341 "movdqa "SREG2", 7*16("dct") \n\t" | |
342 | |
343 inline void ff_idct_xvid_sse2(short *block) | |
344 { | |
345 __asm__ volatile( | |
346 "movq "MANGLE(m127)", %%mm0 \n\t" | |
347 iMTX_MULT("(%0)", MANGLE(iTab1), ROUND(walkenIdctRounders), PUT_EVEN(ROW0)) | |
348 iMTX_MULT("1*16(%0)", MANGLE(iTab2), ROUND(walkenIdctRounders+1*16), PUT_ODD(ROW1)) | |
349 iMTX_MULT("2*16(%0)", MANGLE(iTab3), ROUND(walkenIdctRounders+2*16), PUT_EVEN(ROW2)) | |
350 | |
351 TEST_TWO_ROWS("3*16(%0)", "4*16(%0)", "%%eax", "%%ecx", CLEAR_ODD(ROW3), CLEAR_EVEN(ROW4)) | |
352 JZ("%%eax", "1f") | |
353 iMTX_MULT("3*16(%0)", MANGLE(iTab4), ROUND(walkenIdctRounders+3*16), PUT_ODD(ROW3)) | |
354 | |
355 TEST_TWO_ROWS("5*16(%0)", "6*16(%0)", "%%eax", "%%edx", CLEAR_ODD(ROW5), CLEAR_EVEN(ROW6)) | |
356 TEST_ONE_ROW("7*16(%0)", "%%esi", CLEAR_ODD(ROW7)) | |
357 iLLM_HEAD | |
358 ASMALIGN(4) | |
359 JNZ("%%ecx", "2f") | |
360 JNZ("%%eax", "3f") | |
361 JNZ("%%edx", "4f") | |
362 JNZ("%%esi", "5f") | |
363 iLLM_PASS_SPARSE("%0") | |
364 "jmp 6f \n\t" | |
365 "2: \n\t" | |
366 iMTX_MULT("4*16(%0)", MANGLE(iTab1), "#", PUT_EVEN(ROW4)) | |
367 "3: \n\t" | |
368 iMTX_MULT("5*16(%0)", MANGLE(iTab4), ROUND(walkenIdctRounders+4*16), PUT_ODD(ROW5)) | |
369 JZ("%%edx", "1f") | |
370 "4: \n\t" | |
371 iMTX_MULT("6*16(%0)", MANGLE(iTab3), ROUND(walkenIdctRounders+5*16), PUT_EVEN(ROW6)) | |
372 JZ("%%esi", "1f") | |
373 "5: \n\t" | |
374 iMTX_MULT("7*16(%0)", MANGLE(iTab2), ROUND(walkenIdctRounders+5*16), PUT_ODD(ROW7)) | |
8590 | 375 #if !ARCH_X86_64 |
8430 | 376 iLLM_HEAD |
377 #endif | |
378 iLLM_PASS("%0") | |
379 "6: \n\t" | |
380 : "+r"(block) | |
381 : | |
382 : "%eax", "%ecx", "%edx", "%esi", "memory"); | |
383 } | |
384 | |
385 void ff_idct_xvid_sse2_put(uint8_t *dest, int line_size, short *block) | |
386 { | |
387 ff_idct_xvid_sse2(block); | |
388 put_pixels_clamped_mmx(block, dest, line_size); | |
389 } | |
390 | |
391 void ff_idct_xvid_sse2_add(uint8_t *dest, int line_size, short *block) | |
392 { | |
393 ff_idct_xvid_sse2(block); | |
394 add_pixels_clamped_mmx(block, dest, line_size); | |
395 } |