git.wh0rd.org / patches.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
initial import
[patches.git] / libdv-0.104-pic-fix.patch
1 diff -urp libdv-0.104-old/libdv/dct_block_mmx.S libdv-0.104/libdv/dct_block_mmx.S
2 --- libdv-0.104-old/libdv/dct_block_mmx.S 2005-10-23 19:40:58.000000000 +0200
3 +++ libdv-0.104/libdv/dct_block_mmx.S 2005-10-24 00:11:39.000000000 +0200
4 @@ -53,6 +53,17 @@ scratch2: .quad 0
5 scratch3: .quad 0
6 scratch4: .quad 0
7
8 +#ifdef __PIC__
9 +# undef __i686 /* gcc define gets in our way */
10 + .section .gnu.linkonce.t.__i686.get_pc_thunk.bp,"ax",@progbits
11 +.globl __i686.get_pc_thunk.bp
12 + .hidden __i686.get_pc_thunk.bp
13 + .type __i686.get_pc_thunk.bp,@function
14 +__i686.get_pc_thunk.bp:
15 + movl (%esp), %ebp
16 + ret
17 +#endif
18 +
19 .text
20
21 .align 8
22 @@ -60,10 +71,14 @@ scratch4: .quad 0
23 _dv_dct_88_block_mmx:
24
25 pushl %ebp
26 - movl %esp, %ebp
27 pushl %esi
28
29 - movl 8(%ebp), %esi # source
30 +#ifdef __PIC__
31 + call __i686.get_pc_thunk.bp
32 + addl $_GLOBAL_OFFSET_TABLE_, %ebp
33 +#endif
34 +
35 + movl 12(%ebp), %esi # source
36
37 # column 0
38 movq 16*0(%esi), %mm0 # v0
39 @@ -86,22 +101,45 @@ _dv_dct_88_block_mmx:
40
41 movq 16*3(%esi), %mm5 # v3
42 movq 16*4(%esi), %mm7 # v4
43 +#ifdef __PIC__
44 + movq %mm7, scratch1@GOTOFF(%ebp) # scratch1: v4 ;
45 +#else
46 movq %mm7, scratch1 # scratch1: v4 ;
47 +#endif
48 movq %mm5, %mm7 # duplicate v3
49 +#ifdef __PIC__
50 + paddw scratch1@GOTOFF(%ebp), %mm5 # v03: v3+v4
51 + psubw scratch1@GOTOFF(%ebp), %mm7 # v04: v3-v4
52 + movq %mm5, scratch2@GOTOFF(%ebp) # scratch2: v03
53 +#else
54 paddw scratch1, %mm5 # v03: v3+v4
55 psubw scratch1, %mm7 # v04: v3-v4
56 movq %mm5, scratch2 # scratch2: v03
57 +#endif
58 movq %mm0, %mm5 # mm5: v00
59
60 +#ifdef __PIC__
61 + paddw scratch2@GOTOFF(%ebp), %mm0 # v10: v00+v03
62 + psubw scratch2@GOTOFF(%ebp), %mm5 # v13: v00-v03
63 + movq %mm3, scratch3@GOTOFF(%ebp) # scratch3: v02
64 +#else
65 paddw scratch2, %mm0 # v10: v00+v03
66 psubw scratch2, %mm5 # v13: v00-v03
67 movq %mm3, scratch3 # scratch3: v02
68 +#endif
69 movq %mm1, %mm3 # duplicate v01
70
71 +#ifdef __PIC__
72 + paddw scratch3@GOTOFF(%ebp), %mm1 # v11: v01+v02
73 + psubw scratch3@GOTOFF(%ebp), %mm3 # v12: v01-v02
74 +
75 + movq %mm6, scratch4@GOTOFF(%ebp) # scratch4: v05
76 +#else
77 paddw scratch3, %mm1 # v11: v01+v02
78 psubw scratch3, %mm3 # v12: v01-v02
79
80 movq %mm6, scratch4 # scratch4: v05
81 +#endif
82 movq %mm0, %mm6 # duplicate v10
83
84 paddw %mm1, %mm0 # v10+v11
85 @@ -111,10 +149,18 @@ _dv_dct_88_block_mmx:
86 movq %mm6, 16*4(%esi) # out4: v10-v11
87
88 movq %mm4, %mm0 # mm0: v06
89 +#ifdef __PIC__
90 + paddw scratch4@GOTOFF(%ebp), %mm4 # v15: v05+v06
91 +#else
92 paddw scratch4, %mm4 # v15: v05+v06
93 +#endif
94 paddw %mm2, %mm0 # v16: v07+v06
95
96 +#ifdef __PIC__
97 + pmulhw WA3@GOTOFF(%ebp), %mm4 # v35~: WA3*v15
98 +#else
99 pmulhw WA3, %mm4 # v35~: WA3*v15
100 +#endif
101 psllw $1, %mm4 # v35: compensate the coeefient scale
102
103 movq %mm4, %mm6 # duplicate v35
104 @@ -123,7 +169,11 @@ _dv_dct_88_block_mmx:
105
106 paddw %mm5, %mm3 # v22: v12+v13
107
108 +#ifdef __PIC__
109 + pmulhw WA1@GOTOFF(%ebp), %mm3 # v32~: WA1*v22
110 +#else
111 pmulhw WA1, %mm3 # v32~: WA1*v22
112 +#endif
113 psllw $16-NSHIFT, %mm3 # v32: compensate the coeefient scale
114 movq %mm5, %mm6 # duplicate v13
115
116 @@ -134,13 +184,23 @@ _dv_dct_88_block_mmx:
117 movq %mm6, 16*6(%esi) # out6: v13-v32
118
119
120 +#ifdef __PIC__
121 + paddw scratch4@GOTOFF(%ebp), %mm7 # v14n: v04+v05
122 +#else
123 paddw scratch4, %mm7 # v14n: v04+v05
124 +#endif
125 movq %mm0, %mm5 # duplicate v16
126
127 psubw %mm7, %mm0 # va1: v16-v14n
128 +#ifdef __PIC__
129 + pmulhw WA5@GOTOFF(%ebp), %mm0 # va0~: va1*WA5
130 + pmulhw WA4@GOTOFF(%ebp), %mm5 # v36~~: v16*WA4
131 + pmulhw WA2@GOTOFF(%ebp), %mm7 # v34~~: v14n*WA2
132 +#else
133 pmulhw WA5, %mm0 # va0~: va1*WA5
134 pmulhw WA4, %mm5 # v36~~: v16*WA4
135 pmulhw WA2, %mm7 # v34~~: v14n*WA2
136 +#endif
137 psllw $16-WA4_SHIFT, %mm5 # v36: compensate the coeefient scale
138 psllw $16-NSHIFT, %mm7 # v34: compensate the coeefient scale
139
140 @@ -188,22 +248,45 @@ _dv_dct_88_block_mmx:
141
142 movq 16*3(%esi), %mm5 # v3
143 movq 16*4(%esi), %mm7 # v4
144 +#ifdef __PIC__
145 + movq %mm7, scratch1@GOTOFF(%ebp) # scratch1: v4 ;
146 +#else
147 movq %mm7, scratch1 # scratch1: v4 ;
148 +#endif
149 movq %mm5, %mm7 # duplicate v3
150 +#ifdef __PIC__
151 + paddw scratch1@GOTOFF(%ebp), %mm5 # v03: v3+v4
152 + psubw scratch1@GOTOFF(%ebp), %mm7 # v04: v3-v4
153 + movq %mm5, scratch2@GOTOFF(%ebp) # scratch2: v03
154 +#else
155 paddw scratch1, %mm5 # v03: v3+v4
156 psubw scratch1, %mm7 # v04: v3-v4
157 movq %mm5, scratch2 # scratch2: v03
158 +#endif
159 movq %mm0, %mm5 # mm5: v00
160
161 +#ifdef __PIC__
162 + paddw scratch2@GOTOFF(%ebp), %mm0 # v10: v00+v03
163 + psubw scratch2@GOTOFF(%ebp), %mm5 # v13: v00-v03
164 + movq %mm3, scratch3@GOTOFF(%ebp) # scratc3: v02
165 +#else
166 paddw scratch2, %mm0 # v10: v00+v03
167 psubw scratch2, %mm5 # v13: v00-v03
168 movq %mm3, scratch3 # scratc3: v02
169 +#endif
170 movq %mm1, %mm3 # duplicate v01
171
172 +#ifdef __PIC__
173 + paddw scratch3@GOTOFF(%ebp), %mm1 # v11: v01+v02
174 + psubw scratch3@GOTOFF(%ebp), %mm3 # v12: v01-v02
175 +
176 + movq %mm6, scratch4@GOTOFF(%ebp) # scratc4: v05
177 +#else
178 paddw scratch3, %mm1 # v11: v01+v02
179 psubw scratch3, %mm3 # v12: v01-v02
180
181 movq %mm6, scratch4 # scratc4: v05
182 +#endif
183 movq %mm0, %mm6 # duplicate v10
184
185 paddw %mm1, %mm0 # v10+v11
186 @@ -213,10 +296,18 @@ _dv_dct_88_block_mmx:
187 movq %mm6, 16*4(%esi) # out4: v10-v11
188
189 movq %mm4, %mm0 # mm0: v06
190 +#ifdef __PIC__
191 + paddw scratch4@GOTOFF(%ebp), %mm4 # v15: v05+v06
192 +#else
193 paddw scratch4, %mm4 # v15: v05+v06
194 +#endif
195 paddw %mm2, %mm0 # v16: v07+v06
196
197 +#ifdef __PIC__
198 + pmulhw WA3@GOTOFF(%ebp), %mm4 # v35~: WA3*v15
199 +#else
200 pmulhw WA3, %mm4 # v35~: WA3*v15
201 +#endif
202 psllw $16-NSHIFT, %mm4 # v35: compensate the coeefient scale
203
204 movq %mm4, %mm6 # duplicate v35
205 @@ -225,7 +316,11 @@ _dv_dct_88_block_mmx:
206
207 paddw %mm5, %mm3 # v22: v12+v13
208
209 +#ifdef __PIC__
210 + pmulhw WA1@GOTOFF(%ebp), %mm3 # v32~: WA3*v15
211 +#else
212 pmulhw WA1, %mm3 # v32~: WA3*v15
213 +#endif
214 psllw $16-NSHIFT, %mm3 # v32: compensate the coeefient scale
215 movq %mm5, %mm6 # duplicate v13
216
217 @@ -235,13 +330,23 @@ _dv_dct_88_block_mmx:
218 movq %mm5, 16*2(%esi) # out2: v13+v32
219 movq %mm6, 16*6(%esi) # out6: v13-v32
220
221 +#ifdef __PIC__
222 + paddw scratch4@GOTOFF(%ebp), %mm7 # v14n: v04+v05
223 +#else
224 paddw scratch4, %mm7 # v14n: v04+v05
225 +#endif
226 movq %mm0, %mm5 # duplicate v16
227
228 psubw %mm7, %mm0 # va1: v16-v14n
229 +#ifdef __PIC__
230 + pmulhw WA2@GOTOFF(%ebp), %mm7 # v34~~: v14n*WA2
231 + pmulhw WA5@GOTOFF(%ebp), %mm0 # va0~: va1*WA5
232 + pmulhw WA4@GOTOFF(%ebp), %mm5 # v36~~: v16*WA4
233 +#else
234 pmulhw WA2, %mm7 # v34~~: v14n*WA2
235 pmulhw WA5, %mm0 # va0~: va1*WA5
236 pmulhw WA4, %mm5 # v36~~: v16*WA4
237 +#endif
238 psllw $16-NSHIFT, %mm7
239 psllw $16-WA4_SHIFT, %mm5 # v36: compensate the coeffient
240 # scale note that WA4 is shifted 1 bit less than the others
241 @@ -751,11 +856,15 @@ _dv_dct_block_mmx_postscale_88:
242 _dv_dct_248_block_mmx:
243
244 pushl %ebp
245 - movl %esp, %ebp
246 pushl %esi
247 pushl %edi
248
249 - movl 8(%ebp), %esi # source
250 +#ifdef __PIC__
251 + call __i686.get_pc_thunk.bp
252 + addl $_GLOBAL_OFFSET_TABLE_, %ebp
253 +#endif
254 +
255 + movl 16(%ebp), %esi # source
256
257 # column 0
258
259 @@ -779,7 +888,11 @@ _dv_dct_248_block_mmx:
260 paddw %mm1, %mm0 # v20: v10+v11
261 psubw %mm1, %mm3 # v21: v10-v11
262
263 +#ifdef __PIC__
264 + pmulhw WA1@GOTOFF(%ebp), %mm5 # v32~: WA1*v22
265 +#else
266 pmulhw WA1, %mm5 # v32~: WA1*v22
267 +#endif
268 movq %mm4, %mm2
269 psllw $16-NSHIFT, %mm5 # v32: compensate the coeffient scale
270
271 @@ -818,7 +931,11 @@ _dv_dct_248_block_mmx:
272 paddw %mm1, %mm0 # v20: v10+v11
273 psubw %mm1, %mm3 # v21: v10-v11
274
275 +#ifdef __PIC__
276 + pmulhw WA1@GOTOFF(%ebp), %mm5 # v32~: WA1*v22
277 +#else
278 pmulhw WA1, %mm5 # v32~: WA1*v22
279 +#endif
280 movq %mm4, %mm2
281 psllw $16-NSHIFT, %mm5 # v32: compensate the coeffient scale
282
283 @@ -855,7 +972,11 @@ _dv_dct_248_block_mmx:
284 paddw %mm1, %mm0 # v20: v10+v11
285 psubw %mm1, %mm3 # v21: v10-v11
286
287 +#ifdef __PIC__
288 + pmulhw WA1@GOTOFF(%ebp), %mm5 # v32~: WA1*v22
289 +#else
290 pmulhw WA1, %mm5 # v32~: WA1*v22
291 +#endif
292 movq %mm4, %mm2
293 psllw $16-NSHIFT, %mm5 # v32: compensate the coeffient scale
294
295 @@ -892,7 +1013,11 @@ _dv_dct_248_block_mmx:
296 paddw %mm1, %mm0 # v20: v10+v11
297 psubw %mm1, %mm3 # v21: v10-v11
298
299 +#ifdef __PIC__
300 + pmulhw WA1@GOTOFF(%ebp), %mm5 # v32~: WA1*v22
301 +#else
302 pmulhw WA1, %mm5 # v32~: WA1*v22
303 +#endif
304 movq %mm4, %mm2
305 psllw $16-NSHIFT, %mm5 # v32: compensate the coeffient scale
306
307 diff -urp libdv-0.104-old/libdv/dv.c libdv-0.104/libdv/dv.c
308 --- libdv-0.104-old/libdv/dv.c 2004-10-20 05:49:24.000000000 +0200
309 +++ libdv-0.104/libdv/dv.c 2005-10-24 00:59:57.000000000 +0200
310 @@ -205,6 +205,9 @@ dv_reconfigure(int clamp_luma, int clamp
311 } /* dv_reconfigure */
312
313
314 +extern uint8_t dv_quant_offset[4];
315 +extern uint8_t dv_quant_shifts[22][4];
316 +
317 static inline void
318 dv_decode_macroblock(dv_decoder_t *dv, dv_macroblock_t *mb, unsigned int quality) {
319 int i;
320 @@ -218,7 +221,7 @@ dv_decode_macroblock(dv_decoder_t *dv, d
321 dv_idct_248 (co248, mb->b[i].coeffs);
322 } else {
323 #if ARCH_X86
324 - _dv_quant_88_inverse_x86(mb->b[i].coeffs,mb->qno,mb->b[i].class_no);
325 + _dv_quant_88_inverse_x86(mb->b[i].coeffs,mb->qno,mb->b[i].class_no,dv_quant_offset,dv_quant_shifts);
326 _dv_idct_88(mb->b[i].coeffs);
327 #elif ARCH_X86_64
328 _dv_quant_88_inverse_x86_64(mb->b[i].coeffs,mb->qno,mb->b[i].class_no);
329 @@ -250,7 +253,7 @@ dv_decode_video_segment(dv_decoder_t *dv
330 dv_idct_248 (co248, mb->b[b].coeffs);
331 } else {
332 #if ARCH_X86
333 - _dv_quant_88_inverse_x86(bl->coeffs,mb->qno,bl->class_no);
334 + _dv_quant_88_inverse_x86(bl->coeffs,mb->qno,bl->class_no,dv_quant_offset,dv_quant_shifts);
335 _dv_weight_88_inverse(bl->coeffs);
336 _dv_idct_88(bl->coeffs);
337 #elif ARCH_X86_64
338 diff -urp libdv-0.104-old/libdv/encode.c libdv-0.104/libdv/encode.c
339 --- libdv-0.104-old/libdv/encode.c 2004-11-17 04:36:30.000000000 +0100
340 +++ libdv-0.104/libdv/encode.c 2005-10-24 01:17:41.000000000 +0200
341 @@ -521,7 +521,8 @@ static void reorder_block(dv_block_t *bl
342 }
343
344 extern unsigned long _dv_vlc_encode_block_mmx(dv_coeff_t* coeffs,
345 - dv_vlc_entry_t ** out);
346 + dv_vlc_entry_t ** out,
347 + dv_vlc_entry_t * lookup);
348
349 extern unsigned long _dv_vlc_encode_block_mmx_x86_64(dv_coeff_t* coeffs,
350 dv_vlc_entry_t ** out);
351 @@ -558,7 +559,7 @@ static unsigned long vlc_encode_block(dv
352 #elif ARCH_X86
353 int num_bits;
354
355 - num_bits = _dv_vlc_encode_block_mmx(coeffs, &o);
356 + num_bits = _dv_vlc_encode_block_mmx(coeffs, &o, vlc_encode_lookup);
357 emms();
358 #else
359 int num_bits;
360 @@ -574,7 +575,7 @@ static unsigned long vlc_encode_block(dv
361 return num_bits;
362 }
363
364 -extern unsigned long _dv_vlc_num_bits_block_x86(dv_coeff_t* coeffs);
365 +extern unsigned long _dv_vlc_num_bits_block_x86(dv_coeff_t* coeffs, unsigned char* lookup);
366 extern unsigned long _dv_vlc_num_bits_block_x86_64(dv_coeff_t* coeffs);
367
368 extern unsigned long _dv_vlc_num_bits_block(dv_coeff_t* coeffs)
369 @@ -600,7 +601,7 @@ extern unsigned long _dv_vlc_num_bits_bl
370 #elif ARCH_X86_64
371 return _dv_vlc_num_bits_block_x86_64(coeffs);
372 #else
373 - return _dv_vlc_num_bits_block_x86(coeffs);
374 + return _dv_vlc_num_bits_block_x86(coeffs, vlc_num_bits_lookup);
375 #endif
376 }
377
378 diff -urp libdv-0.104-old/libdv/encode_x86.S libdv-0.104/libdv/encode_x86.S
379 --- libdv-0.104-old/libdv/encode_x86.S 2005-10-23 19:40:58.000000000 +0200
380 +++ libdv-0.104/libdv/encode_x86.S 2005-10-24 01:18:32.000000000 +0200
381 @@ -23,10 +23,6 @@
382 * The libdv homepage is http://libdv.sourceforge.net/.
383 */
384
385 -.data
386 -ALLONE: .word 1,1,1,1
387 -VLCADDMASK: .byte 255,0,0,0,255,0,0,0
388 -
389 .text
390
391 .global _dv_vlc_encode_block_mmx
392 @@ -45,11 +41,14 @@ _dv_vlc_encode_block_mmx:
393
394 movl $63, %ecx
395
396 - movl vlc_encode_lookup, %esi
397 + movl 4+4*4+8(%esp), %esi # vlc_encode_lookup
398
399 pxor %mm0, %mm0
400 pxor %mm2, %mm2
401 - movq VLCADDMASK, %mm1
402 + pushl $0x000000FF
403 + pushl $0x000000FF
404 + movq (%esp), %mm1
405 + addl $8, %esp
406 xorl %ebp, %ebp
407 subl $8, %edx
408 vlc_encode_block_mmx_loop:
409 @@ -121,7 +120,7 @@ _dv_vlc_num_bits_block_x86:
410 addl $2, %edi
411
412 movl $63, %ecx
413 - movl vlc_num_bits_lookup, %esi
414 + movl 4+4*4+4(%esp), %esi # vlc_num_bits_lookup
415
416 vlc_num_bits_block_x86_loop:
417 movw (%edi), %ax
418 @@ -579,8 +578,11 @@ _dv_need_dct_248_mmx_rows:
419 paddw %mm5, %mm1
420
421 paddw %mm1, %mm0
422 -
423 - pmaddwd ALLONE, %mm0
424 +
425 + pushl $0x00010001
426 + pushl $0x00010001
427 + pmaddwd (%esp), %mm0
428 + addl $8, %esp
429 movq %mm0, %mm1
430 psrlq $32, %mm1
431 paddd %mm1, %mm0
432 diff -urp libdv-0.104-old/libdv/idct_block_mmx.S libdv-0.104/libdv/idct_block_mmx.S
433 --- libdv-0.104-old/libdv/idct_block_mmx.S 2005-10-23 19:40:58.000000000 +0200
434 +++ libdv-0.104/libdv/idct_block_mmx.S 2005-10-24 01:12:12.000000000 +0200
435 @@ -8,16 +8,37 @@
436
437
438
439 +#ifdef __PIC__
440 +# undef __i686 /* gcc define gets in our way */
441 + .section .gnu.linkonce.t.__i686.get_pc_thunk.bp,"ax",@progbits
442 +.globl __i686.get_pc_thunk.bp
443 + .hidden __i686.get_pc_thunk.bp
444 + .type __i686.get_pc_thunk.bp,@function
445 +__i686.get_pc_thunk.bp:
446 + movl (%esp), %ebp
447 + ret
448 +#endif
449 +
450 .text
451 +
452 .align 4
453 .globl _dv_idct_block_mmx
454 .type _dv_idct_block_mmx,@function
455 _dv_idct_block_mmx:
456 pushl %ebp
457 - movl %esp,%ebp
458 pushl %esi
459 +
460 +#ifdef __PIC__
461 + call __i686.get_pc_thunk.bp
462 + addl $_GLOBAL_OFFSET_TABLE_, %ebp
463 +#endif
464 +
465 +#ifdef __PIC__
466 + leal preSC@GOTOFF(%ebp), %ecx
467 +#else
468 leal preSC, %ecx
469 - movl 8(%ebp),%esi /* source matrix */
470 +#endif
471 + movl 12(%esp),%esi /* source matrix */
472
473 /*
474 * column 0: even part
475 @@ -35,7 +56,11 @@ _dv_idct_block_mmx:
476 movq %mm1, %mm2 /* added 11/1/96 */
477 pmulhw 8*8(%esi),%mm5 /* V8 */
478 psubsw %mm0, %mm1 /* V16 */
479 +#ifdef __PIC__
480 + pmulhw x5a825a825a825a82@GOTOFF(%ebp), %mm1 /* 23170 ->V18 */
481 +#else
482 pmulhw x5a825a825a825a82, %mm1 /* 23170 ->V18 */
483 +#endif
484 paddsw %mm0, %mm2 /* V17 */
485 movq %mm2, %mm0 /* duplicate V17 */
486 psraw $1, %mm2 /* t75=t82 */
487 @@ -76,7 +101,11 @@ _dv_idct_block_mmx:
488 paddsw %mm0, %mm3 /* V29 ; free mm0 */
489 movq %mm7, %mm1 /* duplicate V26 */
490 psraw $1, %mm3 /* t91=t94 */
491 +#ifdef __PIC__
492 + pmulhw x539f539f539f539f@GOTOFF(%ebp),%mm7 /* V33 */
493 +#else
494 pmulhw x539f539f539f539f,%mm7 /* V33 */
495 +#endif
496 psraw $1, %mm1 /* t96 */
497 movq %mm5, %mm0 /* duplicate V2 */
498 psraw $2, %mm4 /* t85=t87 */
499 @@ -84,15 +113,27 @@ _dv_idct_block_mmx:
500 psubsw %mm4, %mm0 /* V28 ; free mm4 */
501 movq %mm0, %mm2 /* duplicate V28 */
502 psraw $1, %mm5 /* t90=t93 */
503 +#ifdef __PIC__
504 + pmulhw x4546454645464546@GOTOFF(%ebp),%mm0 /* V35 */
505 +#else
506 pmulhw x4546454645464546,%mm0 /* V35 */
507 +#endif
508 psraw $1, %mm2 /* t97 */
509 movq %mm5, %mm4 /* duplicate t90=t93 */
510 psubsw %mm2, %mm1 /* V32 ; free mm2 */
511 +#ifdef __PIC__
512 + pmulhw x61f861f861f861f8@GOTOFF(%ebp),%mm1 /* V36 */
513 +#else
514 pmulhw x61f861f861f861f8,%mm1 /* V36 */
515 +#endif
516 psllw $1, %mm7 /* t107 */
517 paddsw %mm3, %mm5 /* V31 */
518 psubsw %mm3, %mm4 /* V30 ; free mm3 */
519 +#ifdef __PIC__
520 + pmulhw x5a825a825a825a82@GOTOFF(%ebp),%mm4 /* V34 */
521 +#else
522 pmulhw x5a825a825a825a82,%mm4 /* V34 */
523 +#endif
524 nop
525 psubsw %mm1, %mm0 /* V38 */
526 psubsw %mm7, %mm1 /* V37 ; free mm7 */
527 @@ -159,7 +200,11 @@ _dv_idct_block_mmx:
528 psubsw %mm7, %mm1 /* V50 */
529 pmulhw 8*9(%esi), %mm5 /* V9 */
530 paddsw %mm7, %mm2 /* V51 */
531 +#ifdef __PIC__
532 + pmulhw x5a825a825a825a82@GOTOFF(%ebp), %mm1 /* 23170 ->V52 */
533 +#else
534 pmulhw x5a825a825a825a82, %mm1 /* 23170 ->V52 */
535 +#endif
536 movq %mm2, %mm6 /* duplicate V51 */
537 psraw $1, %mm2 /* t138=t144 */
538 movq %mm3, %mm4 /* duplicate V1 */
539 @@ -200,11 +245,19 @@ _dv_idct_block_mmx:
540 * even more by doing the correction step in a later stage when the number
541 * is actually multiplied by 16
542 */
543 +#ifdef __PIC__
544 + paddw x0005000200010001@GOTOFF(%ebp), %mm4
545 +#else
546 paddw x0005000200010001, %mm4
547 +#endif
548 psubsw %mm6, %mm3 /* V60 ; free mm6 */
549 psraw $1, %mm0 /* t154=t156 */
550 movq %mm3, %mm1 /* duplicate V60 */
551 +#ifdef __PIC__
552 + pmulhw x539f539f539f539f@GOTOFF(%ebp), %mm1 /* V67 */
553 +#else
554 pmulhw x539f539f539f539f, %mm1 /* V67 */
555 +#endif
556 movq %mm5, %mm6 /* duplicate V3 */
557 psraw $2, %mm4 /* t148=t150 */
558 paddsw %mm4, %mm5 /* V61 */
559 @@ -213,13 +266,25 @@ _dv_idct_block_mmx:
560 psllw $1, %mm1 /* t169 */
561 paddsw %mm0, %mm5 /* V65 -> result */
562 psubsw %mm0, %mm4 /* V64 ; free mm0 */
563 +#ifdef __PIC__
564 + pmulhw x5a825a825a825a82@GOTOFF(%ebp), %mm4 /* V68 */
565 +#else
566 pmulhw x5a825a825a825a82, %mm4 /* V68 */
567 +#endif
568 psraw $1, %mm3 /* t158 */
569 psubsw %mm6, %mm3 /* V66 */
570 movq %mm5, %mm2 /* duplicate V65 */
571 +#ifdef __PIC__
572 + pmulhw x61f861f861f861f8@GOTOFF(%ebp), %mm3 /* V70 */
573 +#else
574 pmulhw x61f861f861f861f8, %mm3 /* V70 */
575 +#endif
576 psllw $1, %mm6 /* t165 */
577 +#ifdef __PIC__
578 + pmulhw x4546454645464546@GOTOFF(%ebp), %mm6 /* V69 */
579 +#else
580 pmulhw x4546454645464546, %mm6 /* V69 */
581 +#endif
582 psraw $1, %mm2 /* t172 */
583 /* moved from next block */
584 movq 8*5(%esi), %mm0 /* V56 */
585 @@ -344,7 +409,11 @@ _dv_idct_block_mmx:
586 * movq 8*13(%esi), %mm4 tmt13
587 */
588 psubsw %mm4, %mm3 /* V134 */
589 +#ifdef __PIC__
590 + pmulhw x5a825a825a825a82@GOTOFF(%ebp), %mm3 /* 23170 ->V136 */
591 +#else
592 pmulhw x5a825a825a825a82, %mm3 /* 23170 ->V136 */
593 +#endif
594 movq 8*9(%esi), %mm6 /* tmt9 */
595 paddsw %mm4, %mm5 /* V135 ; mm4 free */
596 movq %mm0, %mm4 /* duplicate tmt1 */
597 @@ -373,17 +442,33 @@ _dv_idct_block_mmx:
598 psubsw %mm7, %mm0 /* V144 */
599 movq %mm0, %mm3 /* duplicate V144 */
600 paddsw %mm7, %mm2 /* V147 ; free mm7 */
601 +#ifdef __PIC__
602 + pmulhw x539f539f539f539f@GOTOFF(%ebp), %mm0 /* 21407-> V151 */
603 +#else
604 pmulhw x539f539f539f539f, %mm0 /* 21407-> V151 */
605 +#endif
606 movq %mm1, %mm7 /* duplicate tmt3 */
607 paddsw %mm5, %mm7 /* V145 */
608 psubsw %mm5, %mm1 /* V146 ; free mm5 */
609 psubsw %mm1, %mm3 /* V150 */
610 movq %mm7, %mm5 /* duplicate V145 */
611 +#ifdef __PIC__
612 + pmulhw x4546454645464546@GOTOFF(%ebp), %mm1 /* 17734-> V153 */
613 +#else
614 pmulhw x4546454645464546, %mm1 /* 17734-> V153 */
615 +#endif
616 psubsw %mm2, %mm5 /* V148 */
617 +#ifdef __PIC__
618 + pmulhw x61f861f861f861f8@GOTOFF(%ebp), %mm3 /* 25080-> V154 */
619 +#else
620 pmulhw x61f861f861f861f8, %mm3 /* 25080-> V154 */
621 +#endif
622 psllw $2, %mm0 /* t311 */
623 +#ifdef __PIC__
624 + pmulhw x5a825a825a825a82@GOTOFF(%ebp), %mm5 /* 23170-> V152 */
625 +#else
626 pmulhw x5a825a825a825a82, %mm5 /* 23170-> V152 */
627 +#endif
628 paddsw %mm2, %mm7 /* V149 ; free mm2 */
629 psllw $1, %mm1 /* t313 */
630 nop /* without the nop - freeze here for one clock */
631 @@ -409,7 +494,11 @@ _dv_idct_block_mmx:
632 paddsw %mm3, %mm6 /* V164 ; free mm3 */
633 movq %mm4, %mm3 /* duplicate V142 */
634 psubsw %mm5, %mm4 /* V165 ; free mm5 */
635 +#ifdef __PIC__
636 + movq %mm2, scratch7@GOTOFF(%ebp) /* out7 */
637 +#else
638 movq %mm2, scratch7 /* out7 */
639 +#endif
640 psraw $4, %mm6
641 psraw $4, %mm4
642 paddsw %mm5, %mm3 /* V162 */
643 @@ -420,11 +509,19 @@ _dv_idct_block_mmx:
644 */
645 movq %mm6, 8*9(%esi) /* out9 */
646 paddsw %mm1, %mm0 /* V161 */
647 +#ifdef __PIC__
648 + movq %mm3, scratch5@GOTOFF(%ebp) /* out5 */
649 +#else
650 movq %mm3, scratch5 /* out5 */
651 +#endif
652 psubsw %mm1, %mm5 /* V166 ; free mm1 */
653 movq %mm4, 8*11(%esi) /* out11 */
654 psraw $4, %mm5
655 +#ifdef __PIC__
656 + movq %mm0, scratch3@GOTOFF(%ebp) /* out3 */
657 +#else
658 movq %mm0, scratch3 /* out3 */
659 +#endif
660 movq %mm2, %mm4 /* duplicate V140 */
661 movq %mm5, 8*13(%esi) /* out13 */
662 paddsw %mm7, %mm2 /* V160 */
663 @@ -434,7 +531,11 @@ _dv_idct_block_mmx:
664 /* moved from the next block */
665 movq 8*3(%esi), %mm7
666 psraw $4, %mm4
667 +#ifdef __PIC__
668 + movq %mm2, scratch1@GOTOFF(%ebp) /* out1 */
669 +#else
670 movq %mm2, scratch1 /* out1 */
671 +#endif
672 /* moved from the next block */
673 movq %mm0, %mm1
674 movq %mm4, 8*15(%esi) /* out15 */
675 @@ -491,15 +592,31 @@ _dv_idct_block_mmx:
676 paddsw %mm4, %mm3 /* V113 ; free mm4 */
677 movq %mm0, %mm4 /* duplicate V110 */
678 paddsw %mm1, %mm2 /* V111 */
679 +#ifdef __PIC__
680 + pmulhw x539f539f539f539f@GOTOFF(%ebp), %mm0 /* 21407-> V117 */
681 +#else
682 pmulhw x539f539f539f539f, %mm0 /* 21407-> V117 */
683 +#endif
684 psubsw %mm1, %mm5 /* V112 ; free mm1 */
685 psubsw %mm5, %mm4 /* V116 */
686 movq %mm2, %mm1 /* duplicate V111 */
687 +#ifdef __PIC__
688 + pmulhw x4546454645464546@GOTOFF(%ebp), %mm5 /* 17734-> V119 */
689 +#else
690 pmulhw x4546454645464546, %mm5 /* 17734-> V119 */
691 +#endif
692 psubsw %mm3, %mm2 /* V114 */
693 +#ifdef __PIC__
694 + pmulhw x61f861f861f861f8@GOTOFF(%ebp), %mm4 /* 25080-> V120 */
695 +#else
696 pmulhw x61f861f861f861f8, %mm4 /* 25080-> V120 */
697 +#endif
698 paddsw %mm3, %mm1 /* V115 ; free mm3 */
699 +#ifdef __PIC__
700 + pmulhw x5a825a825a825a82@GOTOFF(%ebp), %mm2 /* 23170-> V118 */
701 +#else
702 pmulhw x5a825a825a825a82, %mm2 /* 23170-> V118 */
703 +#endif
704 psllw $2, %mm0 /* t266 */
705 movq %mm1, (%esi) /* save V115 */
706 psllw $1, %mm5 /* t268 */
707 @@ -517,7 +634,11 @@ _dv_idct_block_mmx:
708 movq %mm6, %mm3 /* duplicate tmt4 */
709 psubsw %mm0, %mm6 /* V100 */
710 paddsw %mm0, %mm3 /* V101 ; free mm0 */
711 +#ifdef __PIC__
712 + pmulhw x5a825a825a825a82@GOTOFF(%ebp), %mm6 /* 23170 ->V102 */
713 +#else
714 pmulhw x5a825a825a825a82, %mm6 /* 23170 ->V102 */
715 +#endif
716 movq %mm7, %mm5 /* duplicate tmt0 */
717 movq 8*8(%esi), %mm1 /* tmt8 */
718 paddsw %mm1, %mm7 /* V103 */
719 @@ -551,10 +672,18 @@ _dv_idct_block_mmx:
720 movq 8*2(%esi), %mm3 /* V123 */
721 paddsw %mm4, %mm7 /* out0 */
722 /* moved up from next block */
723 +#ifdef __PIC__
724 + movq scratch3@GOTOFF(%ebp), %mm0
725 +#else
726 movq scratch3, %mm0
727 +#endif
728 psraw $4, %mm7
729 /* moved up from next block */
730 +#ifdef __PIC__
731 + movq scratch5@GOTOFF(%ebp), %mm6
732 +#else
733 movq scratch5, %mm6
734 +#endif
735 psubsw %mm4, %mm1 /* out14 ; free mm4 */
736 paddsw %mm3, %mm5 /* out2 */
737 psraw $4, %mm1
738 @@ -565,7 +694,11 @@ _dv_idct_block_mmx:
739 movq %mm5, 8*2(%esi) /* out2 ; free mm5 */
740 psraw $4, %mm2
741 /* moved up to the prev block */
742 +#ifdef __PIC__
743 + movq scratch7@GOTOFF(%ebp), %mm4
744 +#else
745 movq scratch7, %mm4
746 +#endif
747 /* moved up to the prev block */
748 psraw $4, %mm0
749 movq %mm2, 8*12(%esi) /* out12 ; free mm2 */
750 @@ -579,7 +712,11 @@ _dv_idct_block_mmx:
751 * psraw $4, %mm0
752 * psraw $4, %mm6
753 */
754 +#ifdef __PIC__
755 + movq scratch1@GOTOFF(%ebp), %mm1
756 +#else
757 movq scratch1, %mm1
758 +#endif
759 psraw $4, %mm4
760 movq %mm0, 8*3(%esi) /* out3 */
761 psraw $4, %mm1
762 diff -urp libdv-0.104-old/libdv/quant.c libdv-0.104/libdv/quant.c
763 --- libdv-0.104-old/libdv/quant.c 2004-10-20 05:49:24.000000000 +0200
764 +++ libdv-0.104/libdv/quant.c 2005-10-24 01:06:24.000000000 +0200
765 @@ -144,7 +144,7 @@ uint8_t dv_quant_offset[4] = { 6,3,0,1
766 uint32_t dv_quant_248_mul_tab [2] [22] [64];
767 uint32_t dv_quant_88_mul_tab [2] [22] [64];
768
769 -extern void _dv_quant_x86(dv_coeff_t *block,int qno,int klass);
770 +extern void _dv_quant_x86(dv_coeff_t *block,int qno,int klass,uint8_t dv_quant_offset[],uint8_t dv_quant_shifts[][]);
771 extern void _dv_quant_x86_64(dv_coeff_t *block,int qno,int klass);
772 static void quant_248_inverse_std(dv_coeff_t *block,int qno,int klass,dv_248_coeff_t *co);
773 static void quant_248_inverse_mmx(dv_coeff_t *block,int qno,int klass,dv_248_coeff_t *co);
774 @@ -210,7 +210,7 @@ void _dv_quant(dv_coeff_t *block,int qno
775 _dv_quant_x86_64(block, qno, klass);
776 emms();
777 #else
778 - _dv_quant_x86(block, qno, klass);
779 + _dv_quant_x86(block, qno, klass, dv_quant_offset, dv_quant_shifts);
780 emms();
781 #endif
782 }
783 diff -urp libdv-0.104-old/libdv/quant.h libdv-0.104/libdv/quant.h
784 --- libdv-0.104-old/libdv/quant.h 2004-10-20 05:49:24.000000000 +0200
785 +++ libdv-0.104/libdv/quant.h 2005-10-24 00:57:43.000000000 +0200
786 @@ -27,7 +27,7 @@ extern void _dv_quant(dv_coeff_t *block,
787 extern void _dv_quant_88_inverse(dv_coeff_t *block,int qno,int klass);
788 extern void (*_dv_quant_248_inverse) (dv_coeff_t *block,int qno,int klass,
789 dv_248_coeff_t *co);
790 -extern void _dv_quant_88_inverse_x86(dv_coeff_t *block,int qno,int klass);
791 +extern void _dv_quant_88_inverse_x86(dv_coeff_t *block,int qno,int klass, uint8_t offset[], uint8_t shifts[][]);
792 extern void _dv_quant_88_inverse_x86_64(dv_coeff_t *block,int qno,int klass);
793 extern void dv_quant_init (void);
794 #ifdef __cplusplus
795 diff -urp libdv-0.104-old/libdv/quant_x86.S libdv-0.104/libdv/quant_x86.S
796 --- libdv-0.104-old/libdv/quant_x86.S 2005-10-23 19:40:58.000000000 +0200
797 +++ libdv-0.104/libdv/quant_x86.S 2005-10-24 01:10:21.000000000 +0200
798 @@ -71,10 +71,13 @@ _dv_quant_88_inverse_x86:
799
800 /* pq = dv_quant_shifts[qno + dv_quant_offset[class]]; */
801 movl ARGn(1),%eax /* qno */
802 + movl ARGn(3),%ebx /* dv_quant_offset */
803 + addl ARGn(2),%ebx /* class */
804 + movzbl (%ebx),%ecx
805 movl ARGn(2),%ebx /* class */
806 - movzbl dv_quant_offset(%ebx),%ecx
807 addl %ecx,%eax
808 - leal dv_quant_shifts(,%eax,4),%edx /* edx is pq */
809 + movl ARGn(4),%edx /* dv_quant_shifts */
810 + leal (%edx,%eax,4),%edx /* edx is pq */
811
812 /* extra = (class == 3); */
813 /* 0 1 2 3 */
814 @@ -212,11 +215,13 @@ _dv_quant_x86:
815
816 /* pq = dv_quant_shifts[qno + dv_quant_offset[class]]; */
817 movl ARGn(1),%eax /* qno */
818 + movl ARGn(3),%ebx /* offset */
819 + addl ARGn(2),%ebx /* class */
820 + movzbl (%ebx),%ecx
821 movl ARGn(2),%ebx /* class */
822 -
823 - movzbl dv_quant_offset(%ebx),%ecx
824 + movl ARGn(4),%edx /* shifts */
825 addl %ecx,%eax
826 - leal dv_quant_shifts(,%eax,4),%edx /* edx is pq */
827 + leal (%edx,%eax,4),%edx /* edx is pq */
828
829 /* extra = (class == 3); */
830 /* 0 1 2 3 */
831 diff -urp libdv-0.104-old/libdv/rgbtoyuv.S libdv-0.104/libdv/rgbtoyuv.S
832 --- libdv-0.104-old/libdv/rgbtoyuv.S 2005-10-23 19:40:58.000000000 +0200
833 +++ libdv-0.104/libdv/rgbtoyuv.S 2005-10-24 00:46:34.000000000 +0200
834 @@ -110,20 +110,30 @@ VR0GR: .long 0,0
835 VBG0B: .long 0,0
836
837 #endif
838 -
839 +
840 +#ifdef __PIC__
841 +# undef __i686 /* gcc define gets in our way */
842 + .section .gnu.linkonce.t.__i686.get_pc_thunk.bp,"ax",@progbits
843 +.globl __i686.get_pc_thunk.bp
844 + .hidden __i686.get_pc_thunk.bp
845 + .type __i686.get_pc_thunk.bp,@function
846 +__i686.get_pc_thunk.bp:
847 + movl (%esp), %ebp
848 + ret
849 +#endif
850 +
851 .text
852
853 -#define _inPtr 8
854 -#define _rows 12
855 -#define _columns 16
856 -#define _outyPtr 20
857 -#define _outuPtr 24
858 -#define _outvPtr 28
859 +#define _inPtr 24+8
860 +#define _rows 24+12
861 +#define _columns 24+16
862 +#define _outyPtr 24+20
863 +#define _outuPtr 24+24
864 +#define _outvPtr 24+28
865
866 _dv_rgbtoycb_mmx:
867
868 pushl %ebp
869 - movl %esp, %ebp
870 pushl %eax
871 pushl %ebx
872 pushl %ecx
873 @@ -131,46 +141,103 @@ _dv_rgbtoycb_mmx:
874 pushl %esi
875 pushl %edi
876
877 +#ifdef __PIC__
878 + call __i686.get_pc_thunk.bp
879 + addl $_GLOBAL_OFFSET_TABLE_, %ebp
880 +#endif
881 +
882 +#ifdef __PIC__
883 + leal ZEROSX@GOTOFF(%ebp), %eax #This section gets around a bug
884 +#else
885 leal ZEROSX, %eax #This section gets around a bug
886 +#endif
887 movq (%eax), %mm0 #unlikely to persist
888 +#ifdef __PIC__
889 + movq %mm0, ZEROS@GOTOFF(%ebp)
890 + leal OFFSETDX@GOTOFF(%ebp), %eax
891 +#else
892 movq %mm0, ZEROS
893 leal OFFSETDX, %eax
894 +#endif
895 movq (%eax), %mm0
896 +#ifdef __PIC__
897 + movq %mm0, OFFSETD@GOTOFF(%ebp)
898 + leal OFFSETWX@GOTOFF(%ebp), %eax
899 +#else
900 movq %mm0, OFFSETD
901 leal OFFSETWX, %eax
902 +#endif
903 movq (%eax), %mm0
904 +#ifdef __PIC__
905 + movq %mm0, OFFSETW@GOTOFF(%ebp)
906 + leal OFFSETBX@GOTOFF(%ebp), %eax
907 +#else
908 movq %mm0, OFFSETW
909 leal OFFSETBX, %eax
910 +#endif
911 movq (%eax), %mm0
912 +#ifdef __PIC__
913 + movq %mm0, OFFSETB@GOTOFF(%ebp)
914 + leal YR0GRX@GOTOFF(%ebp), %eax
915 +#else
916 movq %mm0, OFFSETB
917 leal YR0GRX, %eax
918 +#endif
919 movq (%eax), %mm0
920 +#ifdef __PIC__
921 + movq %mm0, YR0GR@GOTOFF(%ebp)
922 + leal YBG0BX@GOTOFF(%ebp), %eax
923 +#else
924 movq %mm0, YR0GR
925 leal YBG0BX, %eax
926 +#endif
927 movq (%eax), %mm0
928 +#ifdef __PIC__
929 + movq %mm0, YBG0B@GOTOFF(%ebp)
930 + leal UR0GRX@GOTOFF(%ebp), %eax
931 +#else
932 movq %mm0, YBG0B
933 leal UR0GRX, %eax
934 +#endif
935 movq (%eax), %mm0
936 +#ifdef __PIC__
937 + movq %mm0, UR0GR@GOTOFF(%ebp)
938 + leal UBG0BX@GOTOFF(%ebp), %eax
939 +#else
940 movq %mm0, UR0GR
941 leal UBG0BX, %eax
942 +#endif
943 movq (%eax), %mm0
944 +#ifdef __PIC__
945 + movq %mm0, UBG0B@GOTOFF(%ebp)
946 + leal VR0GRX@GOTOFF(%ebp), %eax
947 +#else
948 movq %mm0, UBG0B
949 leal VR0GRX, %eax
950 +#endif
951 movq (%eax), %mm0
952 +#ifdef __PIC__
953 + movq %mm0, VR0GR@GOTOFF(%ebp)
954 + leal VBG0BX@GOTOFF(%ebp), %eax
955 +#else
956 movq %mm0, VR0GR
957 leal VBG0BX, %eax
958 +#endif
959 movq (%eax), %mm0
960 +#ifdef __PIC__
961 + movq %mm0, VBG0B@GOTOFF(%ebp)
962 +#else
963 movq %mm0, VBG0B
964 -
965 - movl _rows(%ebp), %eax
966 - movl _columns(%ebp), %ebx
967 +#endif
968 + movl _rows(%esp), %eax
969 + movl _columns(%esp), %ebx
970 mull %ebx #number pixels
971 shrl $3, %eax #number of loops
972 movl %eax, %edi #loop counter in edi
973 - movl _inPtr(%ebp), %eax
974 - movl _outyPtr(%ebp), %ebx
975 - movl _outuPtr(%ebp), %ecx
976 - movl _outvPtr(%ebp), %edx
977 + movl _inPtr(%esp), %eax
978 + movl _outyPtr(%esp), %ebx
979 + movl _outuPtr(%esp), %ecx
980 + movl _outvPtr(%esp), %edx
981 rgbtoycb_mmx_loop:
982 movq (%eax), %mm1 #load G2R2B1G1R1B0G0R0
983 pxor %mm6, %mm6 #0 -> mm6
984 @@ -184,29 +251,57 @@ rgbtoycb_mmx_loop:
985 punpcklbw %mm6, %mm1 #B1G1R1B0 -> mm1
986 movq %mm0, %mm2 #R1B0G0R0 -> mm2
987
988 +#ifdef __PIC__
989 + pmaddwd YR0GR@GOTOFF(%ebp), %mm0 #yrR1,ygG0+yrR0 -> mm0
990 +#else
991 pmaddwd YR0GR, %mm0 #yrR1,ygG0+yrR0 -> mm0
992 +#endif
993 movq %mm1, %mm3 #B1G1R1B0 -> mm3
994
995 +#ifdef __PIC__
996 + pmaddwd YBG0B@GOTOFF(%ebp), %mm1 #ybB1+ygG1,ybB0 -> mm1
997 +#else
998 pmaddwd YBG0B, %mm1 #ybB1+ygG1,ybB0 -> mm1
999 +#endif
1000 movq %mm2, %mm4 #R1B0G0R0 -> mm4
1001
1002 +#ifdef __PIC__
1003 + pmaddwd UR0GR@GOTOFF(%ebp), %mm2 #urR1,ugG0+urR0 -> mm2
1004 +#else
1005 pmaddwd UR0GR, %mm2 #urR1,ugG0+urR0 -> mm2
1006 +#endif
1007 movq %mm3, %mm5 #B1G1R1B0 -> mm5
1008
1009 +#ifdef __PIC__
1010 + pmaddwd UBG0B@GOTOFF(%ebp), %mm3 #ubB1+ugG1,ubB0 -> mm3
1011 +#else
1012 pmaddwd UBG0B, %mm3 #ubB1+ugG1,ubB0 -> mm3
1013 +#endif
1014 punpckhbw %mm6, %mm7 # 00G2R2 -> mm7
1015
1016 +#ifdef __PIC__
1017 + pmaddwd VR0GR@GOTOFF(%ebp), %mm4 #vrR1,vgG0+vrR0 -> mm4
1018 +#else
1019 pmaddwd VR0GR, %mm4 #vrR1,vgG0+vrR0 -> mm4
1020 +#endif
1021 paddd %mm1, %mm0 #Y1Y0 -> mm0
1022
1023 +#ifdef __PIC__
1024 + pmaddwd VBG0B@GOTOFF(%ebp), %mm5 #vbB1+vgG1,vbB0 -> mm5
1025 +#else
1026 pmaddwd VBG0B, %mm5 #vbB1+vgG1,vbB0 -> mm5
1027 +#endif
1028
1029 movq 8(%eax), %mm1 #R5B4G4R4B3G3R3B2 -> mm1
1030 paddd %mm3, %mm2 #U1U0 -> mm2
1031
1032 movq %mm1, %mm6 #R5B4G4R4B3G3R3B2 -> mm6
1033
1034 +#ifdef __PIC__
1035 + punpcklbw ZEROS@GOTOFF(%ebp), %mm1 #B3G3R3B2 -> mm1
1036 +#else
1037 punpcklbw ZEROS, %mm1 #B3G3R3B2 -> mm1
1038 +#endif
1039 paddd %mm5, %mm4 #V1V0 -> mm4
1040
1041 movq %mm1, %mm5 #B3G3R3B2 -> mm5
1042 @@ -214,29 +309,61 @@ rgbtoycb_mmx_loop:
1043
1044 paddd %mm7, %mm1 #R3B200+00G2R2=R3B2G2R2->mm1
1045
1046 +#ifdef __PIC__
1047 + punpckhbw ZEROS@GOTOFF(%ebp), %mm6 #R5B4G4R3 -> mm6
1048 +#else
1049 punpckhbw ZEROS, %mm6 #R5B4G4R3 -> mm6
1050 +#endif
1051 movq %mm1, %mm3 #R3B2G2R2 -> mm3
1052
1053 +#ifdef __PIC__
1054 + pmaddwd YR0GR@GOTOFF(%ebp), %mm1 #yrR3,ygG2+yrR2 -> mm1
1055 +#else
1056 pmaddwd YR0GR, %mm1 #yrR3,ygG2+yrR2 -> mm1
1057 +#endif
1058 movq %mm5, %mm7 #B3G3R3B2 -> mm7
1059
1060 +#ifdef __PIC__
1061 + pmaddwd YBG0B@GOTOFF(%ebp), %mm5 #ybB3+ygG3,ybB2 -> mm5
1062 +#else
1063 pmaddwd YBG0B, %mm5 #ybB3+ygG3,ybB2 -> mm5
1064 +#endif
1065 psrad $FIXPSHIFT, %mm0 #32-bit scaled Y1Y0 -> mm0
1066
1067 +#ifdef __PIC__
1068 + movq %mm6, TEMP0@GOTOFF(%ebp) #R5B4G4R4 -> TEMP0
1069 +#else
1070 movq %mm6, TEMP0 #R5B4G4R4 -> TEMP0
1071 +#endif
1072 movq %mm3, %mm6 #R3B2G2R2 -> mm6
1073 +#ifdef __PIC__
1074 + pmaddwd UR0GR@GOTOFF(%ebp), %mm6 #urR3,ugG2+urR2 -> mm6
1075 +#else
1076 pmaddwd UR0GR, %mm6 #urR3,ugG2+urR2 -> mm6
1077 +#endif
1078 psrad $FIXPSHIFT, %mm2 #32-bit scaled U1U0 -> mm2
1079
1080 paddd %mm5, %mm1 #Y3Y2 -> mm1
1081 movq %mm7, %mm5 #B3G3R3B2 -> mm5
1082 +#ifdef __PIC__
1083 + pmaddwd UBG0B@GOTOFF(%ebp), %mm7 #ubB3+ugG3,ubB2
1084 +#else
1085 pmaddwd UBG0B, %mm7 #ubB3+ugG3,ubB2
1086 +#endif
1087 psrad $FIXPSHIFT, %mm1 #32-bit scaled Y3Y2 -> mm1
1088
1089 +#ifdef __PIC__
1090 + pmaddwd VR0GR@GOTOFF(%ebp), %mm3 #vrR3,vgG2+vgR2
1091 +#else
1092 pmaddwd VR0GR, %mm3 #vrR3,vgG2+vgR2
1093 +#endif
1094 packssdw %mm1, %mm0 #Y3Y2Y1Y0 -> mm0
1095
1096 +#ifdef __PIC__
1097 + pmaddwd VBG0B@GOTOFF(%ebp), %mm5 #vbB3+vgG3,vbB2 -> mm5
1098 +#else
1099 pmaddwd VBG0B, %mm5 #vbB3+vgG3,vbB2 -> mm5
1100 +#endif
1101 psrad $FIXPSHIFT, %mm4 #32-bit scaled V1V0 -> mm4
1102
1103 movq 16(%eax), %mm1 #B7G7R7B6G6R6B5G5 -> mm7
1104 @@ -251,58 +378,114 @@ rgbtoycb_mmx_loop:
1105 movq %mm7, %mm5 #R7B6G6R6B5G500 -> mm5
1106 psrad $FIXPSHIFT, %mm3 #32-bit scaled V3V2 -> mm3
1107
1108 +#ifdef __PIC__
1109 + paddw OFFSETY@GOTOFF(%ebp), %mm0
1110 +#else
1111 paddw OFFSETY, %mm0
1112 +#endif
1113 movq %mm0, (%ebx) #store Y3Y2Y1Y0
1114 packssdw %mm6, %mm2 #32-bit scaled U3U2U1U0 -> mm2
1115
1116 +#ifdef __PIC__
1117 + movq TEMP0@GOTOFF(%ebp), %mm0 #R5B4G4R4 -> mm0
1118 +#else
1119 movq TEMP0, %mm0 #R5B4G4R4 -> mm0
1120 +#endif
1121 addl $8, %ebx
1122 -
1123 +
1124 +#ifdef __PIC__
1125 + punpcklbw ZEROS@GOTOFF(%ebp), %mm7 #B5G500 -> mm7
1126 +#else
1127 punpcklbw ZEROS, %mm7 #B5G500 -> mm7
1128 +#endif
1129 movq %mm0, %mm6 #R5B4G4R4 -> mm6
1130
1131 +#ifdef __PIC__
1132 + movq %mm2, TEMPU@GOTOFF(%ebp) #32-bit scaled U3U2U1U0 -> TEMPU
1133 +#else
1134 movq %mm2, TEMPU #32-bit scaled U3U2U1U0 -> TEMPU
1135 +#endif
1136 psrlq $32, %mm0 #00R5B4 -> mm0
1137
1138 paddw %mm0, %mm7 #B5G5R5B4 -> mm7
1139 movq %mm6, %mm2 #B5B4G4R4 -> mm2
1140
1141 +#ifdef __PIC__
1142 + pmaddwd YR0GR@GOTOFF(%ebp), %mm2 #yrR5,ygG4+yrR4 -> mm2
1143 +#else
1144 pmaddwd YR0GR, %mm2 #yrR5,ygG4+yrR4 -> mm2
1145 +#endif
1146 movq %mm7, %mm0 #B5G5R5B4 -> mm0
1147
1148 +#ifdef __PIC__
1149 + pmaddwd YBG0B@GOTOFF(%ebp), %mm7 #ybB5+ygG5,ybB4 -> mm7
1150 +#else
1151 pmaddwd YBG0B, %mm7 #ybB5+ygG5,ybB4 -> mm7
1152 +#endif
1153 packssdw %mm3, %mm4 #32-bit scaled V3V2V1V0 -> mm4
1154
1155 addl $24, %eax #increment RGB count
1156
1157 +#ifdef __PIC__
1158 + movq %mm4, TEMPV@GOTOFF(%ebp) #(V3V2V1V0)/256 -> mm4
1159 +#else
1160 movq %mm4, TEMPV #(V3V2V1V0)/256 -> mm4
1161 +#endif
1162 movq %mm6, %mm4 #B5B4G4R4 -> mm4
1163
1164 +#ifdef __PIC__
1165 + pmaddwd UR0GR@GOTOFF(%ebp), %mm6 #urR5,ugG4+urR4
1166 +#else
1167 pmaddwd UR0GR, %mm6 #urR5,ugG4+urR4
1168 +#endif
1169 movq %mm0, %mm3 #B5G5R5B4 -> mm0
1170
1171 +#ifdef __PIC__
1172 + pmaddwd UBG0B@GOTOFF(%ebp), %mm0 #ubB5+ugG5,ubB4
1173 +#else
1174 pmaddwd UBG0B, %mm0 #ubB5+ugG5,ubB4
1175 +#endif
1176 paddd %mm7, %mm2 #Y5Y4 -> mm2
1177
1178 +#ifdef __PIC__
1179 + pmaddwd VR0GR@GOTOFF(%ebp), %mm4 #vrR5,vgG4+vrR4 -> mm4
1180 +#else
1181 pmaddwd VR0GR, %mm4 #vrR5,vgG4+vrR4 -> mm4
1182 +#endif
1183 pxor %mm7, %mm7 #0 -> mm7
1184
1185 +#ifdef __PIC__
1186 + pmaddwd VBG0B@GOTOFF(%ebp), %mm3 #vbB5+vgG5,vbB4 -> mm3
1187 +#else
1188 pmaddwd VBG0B, %mm3 #vbB5+vgG5,vbB4 -> mm3
1189 +#endif
1190 punpckhbw %mm7, %mm1 #B7G7R7B6 -> mm1
1191
1192 paddd %mm6, %mm0 #U5U4 -> mm0
1193 movq %mm1, %mm6 #B7G7R7B6 -> mm6
1194
1195 +#ifdef __PIC__
1196 + pmaddwd YBG0B@GOTOFF(%ebp), %mm6 #ybB7+ygG7,ybB6 -> mm6
1197 +#else
1198 pmaddwd YBG0B, %mm6 #ybB7+ygG7,ybB6 -> mm6
1199 +#endif
1200 punpckhbw %mm7, %mm5 #R7B6G6R6 -> mm5
1201
1202 movq %mm5, %mm7 #R7B6G6R6 -> mm7
1203 paddd %mm4, %mm3 #V5V4 -> mm3
1204
1205 +#ifdef __PIC__
1206 + pmaddwd YR0GR@GOTOFF(%ebp), %mm5 #yrR7,ygG6+yrR6 -> mm5
1207 +#else
1208 pmaddwd YR0GR, %mm5 #yrR7,ygG6+yrR6 -> mm5
1209 +#endif
1210 movq %mm1, %mm4 #B7G7R7B6 -> mm4
1211
1212 +#ifdef __PIC__
1213 + pmaddwd UBG0B@GOTOFF(%ebp), %mm4 #ubB7+ugG7,ubB6 -> mm4
1214 +#else
1215 pmaddwd UBG0B, %mm4 #ubB7+ugG7,ubB6 -> mm4
1216 +#endif
1217 psrad $FIXPSHIFT, %mm0 #32-bit scaled U5U4 -> mm0
1218
1219 psrad $FIXPSHIFT, %mm2 #32-bit scaled Y5Y4 -> mm2
1220 @@ -310,25 +493,49 @@ rgbtoycb_mmx_loop:
1221 paddd %mm5, %mm6 #Y7Y6 -> mm6
1222 movq %mm7, %mm5 #R7B6G6R6 -> mm5
1223
1224 +#ifdef __PIC__
1225 + pmaddwd UR0GR@GOTOFF(%ebp), %mm7 #urR7,ugG6+ugR6 -> mm7
1226 +#else
1227 pmaddwd UR0GR, %mm7 #urR7,ugG6+ugR6 -> mm7
1228 +#endif
1229 psrad $FIXPSHIFT, %mm3 #32-bit scaled V5V4 -> mm3
1230
1231 +#ifdef __PIC__
1232 + pmaddwd VBG0B@GOTOFF(%ebp), %mm1 #vbB7+vgG7,vbB6 -> mm1
1233 +#else
1234 pmaddwd VBG0B, %mm1 #vbB7+vgG7,vbB6 -> mm1
1235 +#endif
1236 psrad $FIXPSHIFT, %mm6 #32-bit scaled Y7Y6 -> mm6
1237
1238 packssdw %mm6, %mm2 #Y7Y6Y5Y4 -> mm2
1239
1240 +#ifdef __PIC__
1241 + pmaddwd VR0GR@GOTOFF(%ebp), %mm5 #vrR7,vgG6+vrR6 -> mm5
1242 +#else
1243 pmaddwd VR0GR, %mm5 #vrR7,vgG6+vrR6 -> mm5
1244 +#endif
1245 paddd %mm4, %mm7 #U7U6 -> mm7
1246
1247 psrad $FIXPSHIFT, %mm7 #32-bit scaled U7U6 -> mm7
1248 +#ifdef __PIC__
1249 + paddw OFFSETY@GOTOFF(%ebp), %mm2
1250 +#else
1251 paddw OFFSETY, %mm2
1252 +#endif
1253 movq %mm2, (%ebx) #store Y7Y6Y5Y4
1254
1255 +#ifdef __PIC__
1256 + movq ALLONE@GOTOFF(%ebp), %mm6
1257 +#else
1258 movq ALLONE, %mm6
1259 +#endif
1260 packssdw %mm7, %mm0 #32-bit scaled U7U6U5U4 -> mm0
1261
1262 +#ifdef __PIC__
1263 + movq TEMPU@GOTOFF(%ebp), %mm4 #32-bit scaled U3U2U1U0 -> mm4
1264 +#else
1265 movq TEMPU, %mm4 #32-bit scaled U3U2U1U0 -> mm4
1266 +#endif
1267 pmaddwd %mm6, %mm0 #U7U6U5U4 averaged -> (U7U6)(U5U4)=UU3 UU2->mm0
1268
1269 pmaddwd %mm6, %mm4 #U3U2U1U0 averaged -> (U3U2)(U1U0)=UU1 UU0->mm4
1270 @@ -338,8 +545,12 @@ rgbtoycb_mmx_loop:
1271
1272 psrad $FIXPSHIFT, %mm1 #32-bit scaled V7V6 -> mm1
1273 psraw $1, %mm4 #divide UU3 UU2 UU1 UU0 by 2 -> mm4
1274 -
1275 +
1276 +#ifdef __PIC__
1277 + movq TEMPV@GOTOFF(%ebp), %mm5 #32-bit scaled V3V2V1V0 -> mm5
1278 +#else
1279 movq TEMPV, %mm5 #32-bit scaled V3V2V1V0 -> mm5
1280 +#endif
1281
1282 movq %mm4, (%ecx) # store U
1283
1284 @@ -425,14 +636,22 @@ _dv_ppm_copy_y_block_mmx:
1285 _dv_pgm_copy_y_block_mmx:
1286
1287 pushl %ebp
1288 - movl %esp, %ebp
1289 pushl %esi
1290 pushl %edi
1291 -
1292 - movl 8(%ebp), %edi # dest
1293 - movl 12(%ebp), %esi # src
1294
1295 +#ifdef __PIC__
1296 + call __i686.get_pc_thunk.bp
1297 + addl $_GLOBAL_OFFSET_TABLE_, %ebp
1298 +#endif
1299 +
1300 + movl 16(%esp), %edi # dest
1301 + movl 20(%esp), %esi # src
1302 +
1303 +#ifdef __PIC__
1304 + movq OFFSETY@GOTOFF(%ebp), %mm7
1305 +#else
1306 movq OFFSETY, %mm7
1307 +#endif
1308 pxor %mm6, %mm6
1309
1310 movq (%esi), %mm0
1311 @@ -567,14 +786,22 @@ _dv_pgm_copy_y_block_mmx:
1312 _dv_video_copy_y_block_mmx:
1313
1314 pushl %ebp
1315 - movl %esp, %ebp
1316 pushl %esi
1317 pushl %edi
1318 -
1319 - movl 8(%ebp), %edi # dest
1320 - movl 12(%ebp), %esi # src
1321
1322 +#ifdef __PIC__
1323 + call __i686.get_pc_thunk.bp
1324 + addl $_GLOBAL_OFFSET_TABLE_, %ebp
1325 +#endif
1326 +
1327 + movl 16(%esp), %edi # dest
1328 + movl 20(%esp), %esi # src
1329 +
1330 +#ifdef __PIC__
1331 + movq OFFSETBX@GOTOFF(%ebp), %mm7
1332 +#else
1333 movq OFFSETBX, %mm7
1334 +#endif
1335 pxor %mm6, %mm6
1336
1337 movq (%esi), %mm0
1338 @@ -855,16 +1082,23 @@ _dv_ppm_copy_pal_c_block_mmx:
1339 _dv_pgm_copy_pal_c_block_mmx:
1340
1341 pushl %ebp
1342 - movl %esp, %ebp
1343 pushl %esi
1344 pushl %edi
1345 pushl %ebx
1346 -
1347 - movl 8(%ebp), %edi # dest
1348 - movl 12(%ebp), %esi # src
1349
1350 +#ifdef __PIC__
1351 + call __i686.get_pc_thunk.bp
1352 + addl $_GLOBAL_OFFSET_TABLE_, %ebp
1353 +#endif
1354 +
1355 + movl 20(%esp), %edi # dest
1356 + movl 24(%esp), %esi # src
1357
1358 +#ifdef __PIC__
1359 + movq OFFSETBX@GOTOFF(%ebp), %mm7
1360 +#else
1361 movq OFFSETBX, %mm7
1362 +#endif
1363 pxor %mm6, %mm6
1364
1365
1366 @@ -1003,15 +1237,23 @@ _dv_pgm_copy_pal_c_block_mmx:
1367 _dv_video_copy_pal_c_block_mmx:
1368
1369 pushl %ebp
1370 - movl %esp, %ebp
1371 pushl %esi
1372 pushl %edi
1373 pushl %ebx
1374 -
1375 - movl 8(%ebp), %edi # dest
1376 - movl 12(%ebp), %esi # src
1377
1378 +#ifdef __PIC__
1379 + call __i686.get_pc_thunk.bp
1380 + addl $_GLOBAL_OFFSET_TABLE_, %ebp
1381 +#endif
1382 +
1383 + movl 20(%esp), %edi # dest
1384 + movl 24(%esp), %esi # src
1385 +
1386 +#ifdef __PIC__
1387 + movq OFFSETBX@GOTOFF(%ebp), %mm7
1388 +#else
1389 movq OFFSETBX, %mm7
1390 +#endif
1391 paddw %mm7, %mm7
1392 pxor %mm6, %mm6
1393
1394 @@ -1098,18 +1340,25 @@ video_copy_pal_c_block_mmx_loop:
1395 _dv_ppm_copy_ntsc_c_block_mmx:
1396
1397 pushl %ebp
1398 - movl %esp, %ebp
1399 pushl %esi
1400 pushl %edi
1401 pushl %ebx
1402 -
1403 - movl 8(%ebp), %edi # dest
1404 - movl 12(%ebp), %esi # src
1405 +
1406 +#ifdef __PIC__
1407 + call __i686.get_pc_thunk.bp
1408 + addl $_GLOBAL_OFFSET_TABLE_, %ebp
1409 +#endif
1410 +
1411 + movl 20(%esp), %edi # dest
1412 + movl 24(%esp), %esi # src
1413
1414 movl $4, %ebx
1415
1416 +#ifdef __PIC__
1417 + movq ALLONE@GOTOFF(%ebp), %mm6
1418 +#else
1419 movq ALLONE, %mm6
1420 -
1421 +#endif
1422 ppm_copy_ntsc_c_block_mmx_loop:
1423
1424 movq (%esi), %mm0
1425 @@ -1171,14 +1420,22 @@ ppm_copy_ntsc_c_block_mmx_loop:
1426 _dv_pgm_copy_ntsc_c_block_mmx:
1427
1428 pushl %ebp
1429 - movl %esp, %ebp
1430 pushl %esi
1431 pushl %edi
1432 -
1433 - movl 8(%ebp), %edi # dest
1434 - movl 12(%ebp), %esi # src
1435
1436 +#ifdef __PIC__
1437 + call __i686.get_pc_thunk.bp
1438 + addl $_GLOBAL_OFFSET_TABLE_, %ebp
1439 +#endif
1440 +
1441 + movl 16(%esp), %edi # dest
1442 + movl 20(%esp), %esi # src
1443 +
1444 +#ifdef __PIC__
1445 + movq OFFSETBX@GOTOFF(%ebp), %mm7
1446 +#else
1447 movq OFFSETBX, %mm7
1448 +#endif
1449 paddw %mm7, %mm7
1450 pxor %mm6, %mm6
1451
1452 @@ -1328,15 +1585,23 @@ _dv_pgm_copy_ntsc_c_block_mmx:
1453 _dv_video_copy_ntsc_c_block_mmx:
1454
1455 pushl %ebp
1456 - movl %esp, %ebp
1457 pushl %esi
1458 pushl %edi
1459 pushl %ebx
1460 -
1461 - movl 8(%ebp), %edi # dest
1462 - movl 12(%ebp), %esi # src
1463
1464 +#ifdef __PIC__
1465 + call __i686.get_pc_thunk.bp
1466 + addl $_GLOBAL_OFFSET_TABLE_, %ebp
1467 +#endif
1468 +
1469 + movl 20(%esp), %edi # dest
1470 + movl 24(%esp), %esi # src
1471 +
1472 +#ifdef __PIC__
1473 + movq OFFSETBX@GOTOFF(%ebp), %mm7
1474 +#else
1475 movq OFFSETBX, %mm7
1476 +#endif
1477 paddw %mm7, %mm7
1478 pxor %mm6, %mm6
1479
1480 diff -urp libdv-0.104-old/libdv/vlc_x86.S libdv-0.104/libdv/vlc_x86.S
1481 --- libdv-0.104-old/libdv/vlc_x86.S 2005-10-23 19:40:58.000000000 +0200
1482 +++ libdv-0.104/libdv/vlc_x86.S 2005-10-25 01:47:14.000000000 +0200
1483 @@ -1,29 +1,76 @@
1484 #include "asmoff.h"
1485 .text
1486 +
1487 +#ifdef __PIC__
1488 +# undef __i686 /* gcc define gets in our way */
1489 + .section .gnu.linkonce.t.__i686.get_pc_thunk.bp,"ax",@progbits
1490 +.globl __i686.get_pc_thunk.bp
1491 + .hidden __i686.get_pc_thunk.bp
1492 + .type __i686.get_pc_thunk.bp,@function
1493 +__i686.get_pc_thunk.bp:
1494 + movl (%esp), %ebp
1495 + ret
1496 +
1497 + .section .gnu.linkonce.t.__i686.get_pc_thunk.si,"ax",@progbits
1498 +.globl __i686.get_pc_thunk.si
1499 + .hidden __i686.get_pc_thunk.si
1500 + .type __i686.get_pc_thunk.si,@function
1501 +__i686.get_pc_thunk.si:
1502 + movl (%esp), %esi
1503 + ret
1504 +#endif
1505 +
1506 .align 4
1507 .globl dv_decode_vlc
1508 .type dv_decode_vlc,@function
1509 dv_decode_vlc:
1510 pushl %ebx
1511 + pushl %ebp
1512 +
1513 +#ifdef __PIC__
1514 + call __i686.get_pc_thunk.bp
1515 + addl $_GLOBAL_OFFSET_TABLE_, %ebp
1516 +#endif
1517
1518 - /* Args are at 8(%esp). */
1519 - movl 8(%esp),%eax /* %eax is bits */
1520 - movl 12(%esp),%ebx /* %ebx is maxbits */
1521 + /* Args are at 12(%esp). */
1522 + movl 12(%esp),%eax /* %eax is bits */
1523 + movl 16(%esp),%ebx /* %ebx is maxbits */
1524 andl $0x3f,%ebx /* limit index range STL*/
1525
1526 +#ifdef __ELF__
1527 + movl dv_vlc_class_index_mask@GOTOFF(%ebp,%ebx,4),%edx
1528 +#else
1529 movl dv_vlc_class_index_mask(,%ebx,4),%edx
1530 +#endif
1531 andl %eax,%edx
1532 +#ifdef __ELF__
1533 + movl dv_vlc_class_index_rshift@GOTOFF(%ebp,%ebx,4),%ecx
1534 +#else
1535 movl dv_vlc_class_index_rshift(,%ebx,4),%ecx
1536 +#endif
1537 sarl %cl,%edx
1538 +#ifdef __ELF__
1539 + movl dv_vlc_classes@GOTOFF(%ebp,%ebx,4),%ecx
1540 +#else
1541 movl dv_vlc_classes(,%ebx,4),%ecx
1542 +#endif
1543 movsbl (%ecx,%edx,1),%edx /* %edx is class */
1544 -
1545 +
1546 +#ifdef __ELF__
1547 + movl dv_vlc_index_mask@GOTOFF(%ebp,%edx,4),%ebx
1548 + movl dv_vlc_index_rshift@GOTOFF(%ebp,%edx,4),%ecx
1549 +#else
1550 movl dv_vlc_index_mask(,%edx,4),%ebx
1551 movl dv_vlc_index_rshift(,%edx,4),%ecx
1552 +#endif
1553 andl %eax,%ebx
1554 sarl %cl,%ebx
1555
1556 +#ifdef __ELF__
1557 + movl dv_vlc_lookups@GOTOFF(%ebp,%edx,4),%edx
1558 +#else
1559 movl dv_vlc_lookups(,%edx,4),%edx
1560 +#endif
1561 movl (%edx,%ebx,4),%edx
1562
1563 /* Now %edx holds result, like this:
1564 @@ -42,7 +89,11 @@ dv_decode_vlc:
1565 movl %edx,%ecx
1566 sarl $8,%ecx
1567 andl $0xff,%ecx
1568 +#ifdef __ELF__
1569 + movl sign_mask@GOTOFF(%ebp,%ecx,4),%ebx
1570 +#else
1571 movl sign_mask(,%ecx,4),%ebx
1572 +#endif
1573 andl %ebx,%eax
1574 negl %eax
1575 sarl $31,%eax
1576 @@ -63,14 +114,14 @@ dv_decode_vlc:
1577 *result = broken;
1578 Note that the 'broken' pattern is all ones (i.e. 0xffffffff)
1579 */
1580 - movl 12(%esp),%ebx /* %ebx is maxbits */
1581 + movl 16(%esp),%ebx /* %ebx is maxbits */
1582 subl %ecx,%ebx
1583 sbbl %ebx,%ebx
1584 orl %ebx,%edx
1585
1586 - movl 16(%esp),%eax
1587 + movl 20(%esp),%eax
1588 movl %edx,(%eax)
1589 -
1590 + popl %ebp
1591 popl %ebx
1592 ret
1593
1594 @@ -80,21 +131,38 @@ dv_decode_vlc:
1595 .type __dv_decode_vlc,@function
1596 __dv_decode_vlc:
1597 pushl %ebx
1598 + pushl %ebp
1599 +
1600 +#ifdef __PIC__
1601 + call __i686.get_pc_thunk.bp
1602 + addl $_GLOBAL_OFFSET_TABLE_, %ebp
1603 +#endif
1604
1605 - /* Args are at 8(%esp). */
1606 - movl 8(%esp),%eax /* %eax is bits */
1607 + /* Args are at 12(%esp). */
1608 + movl 12(%esp),%eax /* %eax is bits */
1609
1610 movl %eax,%edx /* %edx is class */
1611 andl $0xfe00,%edx
1612 sarl $9,%edx
1613 +#ifdef __PIC__
1614 + movsbl dv_vlc_class_lookup5@GOTOFF(%ebp,%edx),%edx
1615 +
1616 + movl dv_vlc_index_mask@GOTOFF(%ebp,%edx,4),%ebx
1617 + movl dv_vlc_index_rshift@GOTOFF(%ebp,%edx,4),%ecx
1618 +#else
1619 movsbl dv_vlc_class_lookup5(%edx),%edx
1620 -
1621 +
1622 movl dv_vlc_index_mask(,%edx,4),%ebx
1623 movl dv_vlc_index_rshift(,%edx,4),%ecx
1624 +#endif
1625 andl %eax,%ebx
1626 sarl %cl,%ebx
1627
1628 +#ifdef __PIC__
1629 + movl dv_vlc_lookups@GOTOFF(%ebp,%edx,4),%edx
1630 +#else
1631 movl dv_vlc_lookups(,%edx,4),%edx
1632 +#endif
1633 movl (%edx,%ebx,4),%edx
1634
1635 /* Now %edx holds result, like this:
1636 @@ -112,7 +180,11 @@ __dv_decode_vlc:
1637 movl %edx,%ecx
1638 sarl $8,%ecx
1639 andl $0xff,%ecx
1640 +#ifdef __PIC__
1641 + movl sign_mask@GOTOFF(%ebp,%ecx,4),%ecx
1642 +#else
1643 movl sign_mask(,%ecx,4),%ecx
1644 +#endif
1645 andl %ecx,%eax
1646 negl %eax
1647 sarl $31,%eax
1648 @@ -127,9 +199,9 @@ __dv_decode_vlc:
1649 xorl %eax,%edx
1650 subl %eax,%edx
1651
1652 - movl 12(%esp),%eax
1653 + movl 16(%esp),%eax
1654 movl %edx,(%eax)
1655 -
1656 + popl %ebp
1657 popl %ebx
1658 ret
1659
1660 @@ -147,6 +219,11 @@ dv_parse_ac_coeffs_pass0:
1661 pushl %esi
1662 pushl %ebp
1663
1664 +#ifdef __PIC__
1665 + call __i686.get_pc_thunk.si
1666 + addl $_GLOBAL_OFFSET_TABLE_, %esi
1667 +#endif
1668 +
1669 #define ARGn(N) (20+(4*(N)))(%esp)
1670
1671 /*
1672 @@ -159,8 +236,10 @@ dv_parse_ac_coeffs_pass0:
1673 ebp bl
1674 */
1675 movl ARGn(2),%ebp
1676 +#ifndef __PIC__
1677 movl ARGn(0),%esi
1678 movl bitstream_t_buf(%esi),%esi
1679 +#endif
1680 movl dv_block_t_offset(%ebp),%edi
1681 movl dv_block_t_reorder(%ebp),%ebx
1682
1683 @@ -170,7 +249,11 @@ dv_parse_ac_coeffs_pass0:
1684
1685 movq dv_block_t_coeffs(%ebp),%mm1
1686 pxor %mm0,%mm0
1687 +#ifdef __PIC__
1688 + pand const_f_0_0_0@GOTOFF(%esi),%mm1
1689 +#else
1690 pand const_f_0_0_0,%mm1
1691 +#endif
1692 movq %mm1,dv_block_t_coeffs(%ebp)
1693 movq %mm0,(dv_block_t_coeffs + 8)(%ebp)
1694 movq %mm0,(dv_block_t_coeffs + 16)(%ebp)
1695 @@ -191,9 +274,17 @@ dv_parse_ac_coeffs_pass0:
1696 readloop:
1697 movl %edi,%ecx
1698 shrl $3,%ecx
1699 +#ifdef __PIC__
1700 + pushl %esi
1701 + movl ARGn(1),%esi
1702 + movl bitstream_t_buf(%esi),%esi
1703 +#endif
1704 movzbl (%esi,%ecx,1),%eax
1705 movzbl 1(%esi,%ecx,1),%edx
1706 movzbl 2(%esi,%ecx,1),%ecx
1707 +#ifdef __PIC__
1708 + popl %esi
1709 +#endif
1710 shll $16,%eax
1711 shll $8,%edx
1712 orl %ecx,%eax
1713 @@ -217,7 +308,11 @@ readloop:
1714
1715 /* Attempt to use the shortcut first. If it hits, then
1716 this vlc term has been decoded. */
1717 +#ifdef __PIC__
1718 + movl dv_vlc_class1_shortcut@GOTOFF(%esi,%ecx,4),%edx
1719 +#else
1720 movl dv_vlc_class1_shortcut(,%ecx,4),%edx
1721 +#endif
1722 test $0x80,%edx
1723 je done_decode
1724
1725 @@ -228,12 +323,19 @@ readloop:
1726 movl %ebx,dv_block_t_reorder(%ebp)
1727
1728 /* %eax is bits */
1729 -
1730 +#ifdef __PIC__
1731 + movsbl dv_vlc_class_lookup5@GOTOFF(%esi,%ecx),%ecx
1732 +
1733 + movl dv_vlc_index_mask@GOTOFF(%esi,%ecx,4),%ebx
1734 + movl dv_vlc_lookups@GOTOFF(%esi,%ecx,4),%edx
1735 + movl dv_vlc_index_rshift@GOTOFF(%esi,%ecx,4),%ecx
1736 +#else
1737 movsbl dv_vlc_class_lookup5(%ecx),%ecx
1738
1739 movl dv_vlc_index_mask(,%ecx,4),%ebx
1740 movl dv_vlc_lookups(,%ecx,4),%edx
1741 movl dv_vlc_index_rshift(,%ecx,4),%ecx
1742 +#endif
1743 andl %eax,%ebx
1744 sarl %cl,%ebx
1745
1746 @@ -256,7 +358,11 @@ readloop:
1747 movl %edx,%ecx
1748 sarl $8,%ecx
1749 andl $0xff,%ecx
1750 +#ifdef __PIC__
1751 + movl sign_mask@GOTOFF(%esi,%ecx,4),%ecx
1752 +#else
1753 movl sign_mask(,%ecx,4),%ecx
1754 +#endif
1755 andl %ecx,%eax
1756 negl %eax
1757 sarl $31,%eax
1758 @@ -326,10 +432,20 @@ alldone:
1759
1760 slowpath:
1761 /* slow path: use dv_decode_vlc */;
1762 +#ifdef __PIC__
1763 + pushl %esi
1764 + leal vlc@GOTOFF(%esi),%esi
1765 + xchgl %esi,(%esp) /* last parameter is &vlc */
1766 +#else
1767 pushl $vlc /* last parameter is &vlc */
1768 +#endif
1769 pushl %edx /* bits_left */
1770 pushl %eax /* bits */
1771 +#ifdef __PIC__
1772 + call dv_decode_vlc@PLT
1773 +#else
1774 call dv_decode_vlc
1775 +#endif
1776 addl $12,%esp
1777 test $0x80,%edx /* If (vlc.run < 0) break */
1778 jne escape
1779 @@ -365,6 +481,11 @@ dv_parse_video_segment:
1780 pushl %esi
1781 pushl %ebp
1782
1783 +#ifdef __PIC__
1784 + call __i686.get_pc_thunk.si
1785 + addl $_GLOBAL_OFFSET_TABLE_, %esi
1786 +#endif
1787 +
1788 #define ARGn(N) (20+(4*(N)))(%esp)
1789
1790 movl ARGn(1),%eax /* quality */
1791 @@ -373,7 +494,11 @@ dv_parse_video_segment:
1792 jz its_mono
1793 movl $6,%ebx
1794 its_mono:
1795 +#ifdef __PIC__
1796 + movl %ebx,n_blocks@GOTOFF(%esi)
1797 +#else
1798 movl %ebx,n_blocks
1799 +#endif
1800
1801 /*
1802 * ebx seg/b
1803 @@ -384,15 +509,22 @@ its_mono:
1804 * ebp bl
1805 */
1806 movl ARGn(0),%ebx
1807 +#ifndef __PIC__
1808 movl dv_videosegment_t_bs(%ebx),%esi
1809 movl bitstream_t_buf(%esi),%esi
1810 +#endif
1811 leal dv_videosegment_t_mb(%ebx),%edi
1812
1813 movl $0,%eax
1814 movl $0,%ecx
1815 macloop:
1816 +#ifdef __PIC__
1817 + movl %eax,m@GOTOFF(%esi)
1818 + movl %ecx,mb_start@GOTOFF(%esi)
1819 +#else
1820 movl %eax,m
1821 movl %ecx,mb_start
1822 +#endif
1823
1824 movl ARGn(0),%ebx
1825
1826 @@ -400,7 +532,15 @@ macloop:
1827 /* mb->qno = bitstream_get(bs,4); */
1828 movl %ecx,%edx
1829 shr $3,%edx
1830 +#ifdef __PIC__
1831 + pushl %esi
1832 + movl dv_videosegment_t_bs(%ebx),%esi
1833 + movl bitstream_t_buf(%esi),%esi
1834 +#endif
1835 movzbl 3(%esi,%edx,1),%edx
1836 +#ifdef __PIC__
1837 + popl %esi
1838 +#endif
1839 andl $0xf,%edx
1840 movl %edx,dv_macroblock_t_qno(%edi)
1841
1842 @@ -411,7 +551,11 @@ macloop:
1843 movl %edx,dv_macroblock_t_eob_count(%edi)
1844
1845 /* mb->i = (seg->i + dv_super_map_vertical[m]) % (seg->isPAL?12:10); */
1846 +#ifdef __PIC__
1847 + movl dv_super_map_vertical@GOTOFF(%esi,%eax,4),%edx
1848 +#else
1849 movl dv_super_map_vertical(,%eax,4),%edx
1850 +#endif
1851 movl dv_videosegment_t_i(%ebx),%ecx
1852 addl %ecx,%edx
1853
1854 @@ -422,11 +566,20 @@ skarly:
1855 andl $1,%ecx
1856 shll $5,%ecx /* ecx = (isPAL ? 32 : 0) */
1857
1858 +#ifdef __PIC__
1859 + leal mod_10@GOTOFF(%esi,%edx),%edx
1860 + movzbl (%edx,%ecx,1),%edx /* uses mod_12 for PAL */
1861 +#else
1862 movzbl mod_10(%edx,%ecx,1),%edx /* uses mod_12 for PAL */
1863 +#endif
1864 movl %edx,dv_macroblock_t_i(%edi)
1865
1866 /* mb->j = dv_super_map_horizontal[m]; */
1867 +#ifdef __PIC__
1868 + movl dv_super_map_horizontal@GOTOFF(%esi,%eax,4),%edx
1869 +#else
1870 movl dv_super_map_horizontal(,%eax,4),%edx
1871 +#endif
1872 movl %edx,dv_macroblock_t_j(%edi)
1873
1874 /* mb->k = seg->k; */
1875 @@ -445,12 +598,29 @@ blkloop:
1876 +---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+
1877 */
1878 /* dc = bitstream_get(bs,9); */
1879 +#ifdef __PIC__
1880 + movl mb_start@GOTOFF(%esi),%ecx
1881 +#else
1882 movl mb_start,%ecx
1883 +#endif
1884 shr $3,%ecx
1885 +#ifdef __PIC__
1886 + movzbl blk_start@GOTOFF(%esi,%ebx),%edx
1887 +#else
1888 movzbl blk_start(%ebx),%edx
1889 +#endif
1890 addl %ecx,%edx
1891 +#ifdef __PIC__
1892 + pushl %esi
1893 + movl ARGn(1),%esi
1894 + movl dv_videosegment_t_bs(%esi),%esi
1895 + movl bitstream_t_buf(%esi),%esi
1896 +#endif
1897 movzbl (%esi,%edx,1),%eax /* hi byte */
1898 movzbl 1(%esi,%edx,1),%ecx /* lo byte */
1899 +#ifdef __PIC__
1900 + popl %esi
1901 +#endif
1902 shll $8,%eax
1903 orl %ecx,%eax
1904
1905 @@ -477,7 +647,11 @@ blkloop:
1906
1907 /* bl->reorder = &dv_reorder[bl->dct_mode][1]; */
1908 shll $6,%eax
1909 +#ifdef __PIC__
1910 + addl dv_reorder@GOTOFF+1(%esi),%eax
1911 +#else
1912 addl $(dv_reorder+1),%eax
1913 +#endif
1914 movl %eax,dv_block_t_reorder(%ebp)
1915
1916 /* bl->reorder_sentinel = bl->reorder + 63; */
1917 @@ -485,13 +659,22 @@ blkloop:
1918 movl %eax,dv_block_t_reorder_sentinel(%ebp)
1919
1920 /* bl->offset= mb_start + dv_parse_bit_start[b]; */
1921 +#ifdef __PIC__
1922 + movl mb_start@GOTOFF(%esi),%ecx
1923 + movl dv_parse_bit_start@GOTOFF(%esi,%ebx,4),%eax
1924 +#else
1925 movl mb_start,%ecx
1926 movl dv_parse_bit_start(,%ebx,4),%eax
1927 +#endif
1928 addl %ecx,%eax
1929 movl %eax,dv_block_t_offset(%ebp)
1930
1931 /* bl->end= mb_start + dv_parse_bit_end[b]; */
1932 +#ifdef __PIC__
1933 + movl dv_parse_bit_end@GOTOFF(%esi,%ebx,4),%eax
1934 +#else
1935 movl dv_parse_bit_end(,%ebx,4),%eax
1936 +#endif
1937 addl %ecx,%eax
1938 movl %eax,dv_block_t_end(%ebp)
1939
1940 @@ -503,7 +686,11 @@ blkloop:
1941 /* no AC pass. Just zero out the remaining coeffs */
1942 movq dv_block_t_coeffs(%ebp),%mm1
1943 pxor %mm0,%mm0
1944 +#ifdef __PIC__
1945 + pand const_f_0_0_0@GOTOFF(%esi),%mm1
1946 +#else
1947 pand const_f_0_0_0,%mm1
1948 +#endif
1949 movq %mm1,dv_block_t_coeffs(%ebp)
1950 movq %mm0,(dv_block_t_coeffs + 8)(%ebp)
1951 movq %mm0,(dv_block_t_coeffs + 16)(%ebp)
1952 @@ -528,18 +715,31 @@ do_ac_pass:
1953 pushl %ebp
1954 pushl %edi
1955 pushl %eax
1956 +#ifdef __PIC__
1957 + call dv_parse_ac_coeffs_pass0@PLT
1958 +#else
1959 call dv_parse_ac_coeffs_pass0
1960 +#endif
1961 addl $12,%esp
1962 done_ac:
1963
1964 +#ifdef __PIC__
1965 + movl n_blocks@GOTOFF(%esi),%eax
1966 +#else
1967 movl n_blocks,%eax
1968 +#endif
1969 addl $dv_block_t_size,%ebp
1970 incl %ebx
1971 cmpl %eax,%ebx
1972 jnz blkloop
1973
1974 +#ifdef __PIC__
1975 + movl m@GOTOFF(%esi),%eax
1976 + movl mb_start@GOTOFF(%esi),%ecx
1977 +#else
1978 movl m,%eax
1979 movl mb_start,%ecx
1980 +#endif
1981 addl $(8 * 80),%ecx
1982 addl $dv_macroblock_t_size,%edi
1983 incl %eax
1984 @@ -557,7 +757,11 @@ done_ac:
1985
1986 andl $DV_QUALITY_AC_MASK,%eax
1987 cmpl $DV_QUALITY_AC_2,%eax
1988 +#ifdef __PIC__
1989 + jz dv_parse_ac_coeffs@PLT
1990 +#else
1991 jz dv_parse_ac_coeffs
1992 +#endif
1993 movl $0,%eax
1994 ret
1995
random patches; mostly (all?) dead