Attachment #42071 for bug #67947

Lines 57-63 Link Here

(-)work.orig/ffmpeg-0.4.9-pre1/libavcodec/i386/dsputil_mmx.c (-1 / +1 lines)
57	"pcmpeqd %%" #regd ", %%" #regd " \n\t"\	57	"pcmpeqd %%" #regd ", %%" #regd " \n\t"\
58	"paddb %%" #regd ", %%" #regd " \n\t" ::)	58	"paddb %%" #regd ", %%" #regd " \n\t" ::)
59		59
60	#ifndef PIC	60	#ifndef __PIC__
61	#define MOVQ_BONE(regd) __asm __volatile ("movq %0, %%" #regd " \n\t" ::"m"(mm_bone))	61	#define MOVQ_BONE(regd) __asm __volatile ("movq %0, %%" #regd " \n\t" ::"m"(mm_bone))
62	#define MOVQ_WTWO(regd) __asm __volatile ("movq %0, %%" #regd " \n\t" ::"m"(mm_wtwo))	62	#define MOVQ_WTWO(regd) __asm __volatile ("movq %0, %%" #regd " \n\t" ::"m"(mm_wtwo))
63	#else	63	#else

Lines 96-102 Link Here

(-)work.orig/ffmpeg-0.4.9-pre1/libavcodec/i386/dsputil_mmx_rnd.h (-2 / +2 lines)
96	"addl %5, %3 \n\t"	96	"addl %5, %3 \n\t"
97	"subl $4, %0 \n\t"	97	"subl $4, %0 \n\t"
98	"jnz 1b \n\t"	98	"jnz 1b \n\t"
99	#ifdef PIC //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cant be used	99	#ifdef __PIC__ //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cant be used
100	:"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst)	100	:"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst)
101	#else	101	#else
102	:"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst)	102	:"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst)
Lines 191-197 Link Here
191	"addl $32, %2 \n\t"	191	"addl $32, %2 \n\t"
192	"subl $2, %0 \n\t"	192	"subl $2, %0 \n\t"
193	"jnz 1b \n\t"	193	"jnz 1b \n\t"
194	#ifdef PIC //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cant be used	194	#ifdef __PIC__ //Note "+bm" and "+mb" are buggy too (with gcc 3.2.2 at least) and cant be used
195	:"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst)	195	:"+m"(h), "+a"(src1), "+c"(src2), "+d"(dst)
196	#else	196	#else
197	:"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst)	197	:"+b"(h), "+a"(src1), "+c"(src2), "+d"(dst)

Lines 94-102 Link Here

(-)work.orig/ffmpeg-0.4.9-pre1/libavcodec/i386/mpegvideo_mmx_template.c (-2 / +45 lines)
94	SPREADW(%%mm3)	94	SPREADW(%%mm3)
95	"pxor %%mm7, %%mm7 \n\t" // 0	95	"pxor %%mm7, %%mm7 \n\t" // 0
96	"pxor %%mm4, %%mm4 \n\t" // 0	96	"pxor %%mm4, %%mm4 \n\t" // 0
97	"movq (%2), %%mm5 \n\t" // qmat[0]	97	#ifdef __PIC__
		98	#else
		99	"movq (%0), %%mm5 \n\t" // qmat[0]
		100	#endif
98	"pxor %%mm6, %%mm6 \n\t"	101	"pxor %%mm6, %%mm6 \n\t"
99	"psubw (%3), %%mm6 \n\t" // -bias[0]	102	#ifdef __PIC__
		103	#else
		104	"psubw (%1), %%mm6 \n\t" // -bias[0]
		105	#endif
		106	#ifdef __PIC__
		107	: "+a" (last_non_zero_p1)
		108	: "r" (qmat), "r" (bias)
		109	);
		110	asm volatile(
		111	#endif
100	"movl $-128, %%eax \n\t"	112	"movl $-128, %%eax \n\t"
101	".balign 16 \n\t"	113	".balign 16 \n\t"
102	"1: \n\t"	114	"1: \n\t"
Lines 110-118 Link Here
110	"por %%mm0, %%mm4 \n\t"	122	"por %%mm0, %%mm4 \n\t"
111	"pxor %%mm1, %%mm0 \n\t"	123	"pxor %%mm1, %%mm0 \n\t"
112	"psubw %%mm1, %%mm0 \n\t" // out=((ABS(block[i])qmat[0] - bias[0]qmat[0])>>16)*sign(block[i])	124	"psubw %%mm1, %%mm0 \n\t" // out=((ABS(block[i])qmat[0] - bias[0]qmat[0])>>16)*sign(block[i])
		125	#ifdef __PIC__
		126	"movq %%mm0, (%3, %%eax) \n\t"
		127	#else
113	"movq %%mm0, (%5, %%eax) \n\t"	128	"movq %%mm0, (%5, %%eax) \n\t"
		129	#endif
114	"pcmpeqw %%mm7, %%mm0 \n\t" // out==0 ? 0xFF : 0x00	130	"pcmpeqw %%mm7, %%mm0 \n\t" // out==0 ? 0xFF : 0x00
		131	#ifdef __PIC__
		132	"movq (%2, %%eax), %%mm1 \n\t"
		133	#else
115	"movq (%4, %%eax), %%mm1 \n\t"	134	"movq (%4, %%eax), %%mm1 \n\t"
		135	#endif
116	"movq %%mm7, (%1, %%eax) \n\t" // 0	136	"movq %%mm7, (%1, %%eax) \n\t" // 0
117	"pandn %%mm1, %%mm0 \n\t"	137	"pandn %%mm1, %%mm0 \n\t"
118	PMAXW(%%mm0, %%mm3)	138	PMAXW(%%mm0, %%mm3)
Lines 126-136 Link Here
126	PMAXW(%%mm0, %%mm3)	146	PMAXW(%%mm0, %%mm3)
127	"movd %%mm3, %%eax \n\t"	147	"movd %%mm3, %%eax \n\t"
128	"movzbl %%al, %%eax \n\t" // last_non_zero_p1	148	"movzbl %%al, %%eax \n\t" // last_non_zero_p1
		149	#ifdef __PIC__
		150	: "+a" (last_non_zero_p1)
		151	: "r" (block+64),
		152	"r" (inv_zigzag_direct16+64), "r" (temp_block+64)
		153	#else
129	: "+a" (last_non_zero_p1)	154	: "+a" (last_non_zero_p1)
130	: "r" (block+64), "r" (qmat), "r" (bias),	155	: "r" (block+64), "r" (qmat), "r" (bias),
131	"r" (inv_zigzag_direct16+64), "r" (temp_block+64)	156	"r" (inv_zigzag_direct16+64), "r" (temp_block+64)
		157	#endif
132	);	158	);
133	// note the asm is split cuz gcc doesnt like that many operands ...	159	// note the asm is split cuz gcc doesnt like that many operands ...
		160	// note even more splitting up above when __PIC__ - we're one operand down
		161	// on X86 with __PIC__ compared to normal.
134	asm volatile(	162	asm volatile(
135	"movd %1, %%mm1 \n\t" // max_qcoeff	163	"movd %1, %%mm1 \n\t" // max_qcoeff
136	SPREADW(%%mm1)	164	SPREADW(%%mm1)
Lines 142-147 Link Here
142	);	170	);
143	}else{ // FMT_H263	171	}else{ // FMT_H263
144	asm volatile(	172	asm volatile(
		173	#ifdef __PIC__
		174	"pushl %%ebx \n\t"
		175	"movl %5, %%ebx \n\t"
		176	#else
		177	#endif
145	"movd %%eax, %%mm3 \n\t" // last_non_zero_p1	178	"movd %%eax, %%mm3 \n\t" // last_non_zero_p1
146	SPREADW(%%mm3)	179	SPREADW(%%mm3)
147	"pxor %%mm7, %%mm7 \n\t" // 0	180	"pxor %%mm7, %%mm7 \n\t" // 0
Lines 161-167 Link Here
161	"por %%mm0, %%mm4 \n\t"	194	"por %%mm0, %%mm4 \n\t"
162	"pxor %%mm1, %%mm0 \n\t"	195	"pxor %%mm1, %%mm0 \n\t"
163	"psubw %%mm1, %%mm0 \n\t" // out=((ABS(block[i])qmat[0] - bias[0]qmat[0])>>16)*sign(block[i])	196	"psubw %%mm1, %%mm0 \n\t" // out=((ABS(block[i])qmat[0] - bias[0]qmat[0])>>16)*sign(block[i])
		197	#ifdef __PIC__
		198	"movq %%mm0, (%%ebx, %%eax) \n\t"
		199	#else
164	"movq %%mm0, (%5, %%eax) \n\t"	200	"movq %%mm0, (%5, %%eax) \n\t"
		201	#endif
165	"pcmpeqw %%mm7, %%mm0 \n\t" // out==0 ? 0xFF : 0x00	202	"pcmpeqw %%mm7, %%mm0 \n\t" // out==0 ? 0xFF : 0x00
166	"movq (%4, %%eax), %%mm1 \n\t"	203	"movq (%4, %%eax), %%mm1 \n\t"
167	"movq %%mm7, (%1, %%eax) \n\t" // 0	204	"movq %%mm7, (%1, %%eax) \n\t" // 0
Lines 177-185 Link Here
177	PMAXW(%%mm0, %%mm3)	214	PMAXW(%%mm0, %%mm3)
178	"movd %%mm3, %%eax \n\t"	215	"movd %%mm3, %%eax \n\t"
179	"movzbl %%al, %%eax \n\t" // last_non_zero_p1	216	"movzbl %%al, %%eax \n\t" // last_non_zero_p1
		217	#ifdef __PIC__
		218	: "+a" (last_non_zero_p1)
		219	: "r" (block+64), "r" (qmat+64), "r" (bias+64),
		220	"r" (inv_zigzag_direct16+64), "m" (temp_block+64)
		221	#else
180	: "+a" (last_non_zero_p1)	222	: "+a" (last_non_zero_p1)
181	: "r" (block+64), "r" (qmat+64), "r" (bias+64),	223	: "r" (block+64), "r" (qmat+64), "r" (bias+64),
182	"r" (inv_zigzag_direct16+64), "r" (temp_block+64)	224	"r" (inv_zigzag_direct16+64), "r" (temp_block+64)
		225	#endif
183	);	226	);
184	// note the asm is split cuz gcc doesnt like that many operands ...	227	// note the asm is split cuz gcc doesnt like that many operands ...
185	asm volatile(	228	asm volatile(

Lines 3167-3176 Link Here

(-)work.orig/ffmpeg-0.4.9-pre1/libavcodec/libpostproc/postprocess_template.c (-1 / +44 lines)
3167	{	3167	{
3168	#ifdef HAVE_MMX	3168	#ifdef HAVE_MMX
3169	asm volatile(	3169	asm volatile(
		3170	#ifdef __PIC__
		3171	"pushl %%ebx \n\t"
		3172	"movl %5, %%ebx \n\t"
		3173	#endif
3170	"movq (%%eax), %%mm2 \n\t" // packedYOffset	3174	"movq (%%eax), %%mm2 \n\t" // packedYOffset
3171	"movq 8(%%eax), %%mm3 \n\t" // packedYScale	3175	"movq 8(%%eax), %%mm3 \n\t" // packedYScale
3172	"leal (%2,%4), %%eax \n\t"	3176	"leal (%2,%4), %%eax \n\t"
		3177	#ifdef __PIC__
		3178	"leal (%3,%%ebx), %%edx \n\t"
		3179	#else
3173	"leal (%3,%5), %%edx \n\t"	3180	"leal (%3,%5), %%edx \n\t"
		3181	#endif
3174	"pxor %%mm4, %%mm4 \n\t"	3182	"pxor %%mm4, %%mm4 \n\t"
3175	#ifdef HAVE_MMX2	3183	#ifdef HAVE_MMX2
3176	#define SCALED_CPY(src1, src2, dst1, dst2) \	3184	#define SCALED_CPY(src1, src2, dst1, dst2) \
Lines 3223-3235 Link Here
3223	"movq %%mm1, " #dst2 " \n\t"\	3231	"movq %%mm1, " #dst2 " \n\t"\
3224		3232
3225	#endif //!HAVE_MMX2	3233	#endif //!HAVE_MMX2
3226		3234	#ifdef __PIC__
		3235	SCALED_CPY((%2) , (%2, %4) , (%3) , (%3, %%ebx))
		3236	SCALED_CPY((%2, %4, 2), (%%eax, %4, 2), (%3, %%ebx, 2), (%%edx, %%ebx, 2))
		3237	SCALED_CPY((%2, %4, 4), (%%eax, %4, 4), (%3, %%ebx, 4), (%%edx, %%ebx, 4))
		3238	"leal (%%eax,%4,4), %%eax \n\t"
		3239	"leal (%%edx,%%ebx,4), %%edx \n\t"
		3240	SCALED_CPY((%%eax, %4), (%%eax, %4, 2), (%%edx, %%ebx), (%%edx, %%ebx, 2))
		3241	"popl %%ebx \n\t"
		3242	#else
3227	SCALED_CPY((%2) , (%2, %4) , (%3) , (%3, %5))	3243	SCALED_CPY((%2) , (%2, %4) , (%3) , (%3, %5))
3228	SCALED_CPY((%2, %4, 2), (%%eax, %4, 2), (%3, %5, 2), (%%edx, %5, 2))	3244	SCALED_CPY((%2, %4, 2), (%%eax, %4, 2), (%3, %5, 2), (%%edx, %5, 2))
3229	SCALED_CPY((%2, %4, 4), (%%eax, %4, 4), (%3, %5, 4), (%%edx, %5, 4))	3245	SCALED_CPY((%2, %4, 4), (%%eax, %4, 4), (%3, %5, 4), (%%edx, %5, 4))
3230	"leal (%%eax,%4,4), %%eax \n\t"	3246	"leal (%%eax,%4,4), %%eax \n\t"
3231	"leal (%%edx,%5,4), %%edx \n\t"	3247	"leal (%%edx,%5,4), %%edx \n\t"
3232	SCALED_CPY((%%eax, %4), (%%eax, %4, 2), (%%edx, %5), (%%edx, %5, 2))	3248	SCALED_CPY((%%eax, %4), (%%eax, %4, 2), (%%edx, %5), (%%edx, %5, 2))
		3249	#endif
3233		3250
3234		3251
3235	: "=&a" (packedOffsetAndScale)	3252	: "=&a" (packedOffsetAndScale)
Lines 3237-3243 Link Here
3237	"r"(src),	3254	"r"(src),
3238	"r"(dst),	3255	"r"(dst),
3239	"r" (srcStride),	3256	"r" (srcStride),
		3257	#ifdef __PIC__
		3258	"m" (dstStride)
		3259	#else
3240	"r" (dstStride)	3260	"r" (dstStride)
		3261	#endif
3241	: "%edx"	3262	: "%edx"
3242	);	3263	);
3243	#else	3264	#else
Lines 3250-3257 Link Here
3250	{	3271	{
3251	#ifdef HAVE_MMX	3272	#ifdef HAVE_MMX
3252	asm volatile(	3273	asm volatile(
		3274	#ifdef __PIC__
		3275	"pushl %%ebx \n\t"
		3276	"movl %3, %%ebx \n\t"
		3277	#endif
3253	"leal (%0,%2), %%eax \n\t"	3278	"leal (%0,%2), %%eax \n\t"
		3279	#ifdef __PIC__
		3280	"leal (%1,%%ebx), %%edx \n\t"
		3281	#else
3254	"leal (%1,%3), %%edx \n\t"	3282	"leal (%1,%3), %%edx \n\t"
		3283	#endif
3255		3284
3256	#define SIMPLE_CPY(src1, src2, dst1, dst2) \	3285	#define SIMPLE_CPY(src1, src2, dst1, dst2) \
3257	"movq " #src1 ", %%mm0 \n\t"\	3286	"movq " #src1 ", %%mm0 \n\t"\
Lines 3259-3275 Link Here
3259	"movq %%mm0, " #dst1 " \n\t"\	3288	"movq %%mm0, " #dst1 " \n\t"\
3260	"movq %%mm1, " #dst2 " \n\t"\	3289	"movq %%mm1, " #dst2 " \n\t"\
3261		3290
		3291	#ifdef __PIC__
		3292	SIMPLE_CPY((%0) , (%0, %2) , (%1) , (%1, %%ebx))
		3293	SIMPLE_CPY((%0, %2, 2), (%%eax, %2, 2), (%1, %%ebx, 2), (%%edx, %%ebx, 2))
		3294	SIMPLE_CPY((%0, %2, 4), (%%eax, %2, 4), (%1, %%ebx, 4), (%%edx, %%ebx, 4))
		3295	"leal (%%eax,%2,4), %%eax \n\t"
		3296	"leal (%%edx,%%ebx,4), %%edx \n\t"
		3297	SIMPLE_CPY((%%eax, %2), (%%eax, %2, 2), (%%edx, %%ebx), (%%edx, %%ebx, 2))
		3298	"popl %%ebx \n\t"
		3299	#else
3262	SIMPLE_CPY((%0) , (%0, %2) , (%1) , (%1, %3))	3300	SIMPLE_CPY((%0) , (%0, %2) , (%1) , (%1, %3))
3263	SIMPLE_CPY((%0, %2, 2), (%%eax, %2, 2), (%1, %3, 2), (%%edx, %3, 2))	3301	SIMPLE_CPY((%0, %2, 2), (%%eax, %2, 2), (%1, %3, 2), (%%edx, %3, 2))
3264	SIMPLE_CPY((%0, %2, 4), (%%eax, %2, 4), (%1, %3, 4), (%%edx, %3, 4))	3302	SIMPLE_CPY((%0, %2, 4), (%%eax, %2, 4), (%1, %3, 4), (%%edx, %3, 4))
3265	"leal (%%eax,%2,4), %%eax \n\t"	3303	"leal (%%eax,%2,4), %%eax \n\t"
3266	"leal (%%edx,%3,4), %%edx \n\t"	3304	"leal (%%edx,%3,4), %%edx \n\t"
3267	SIMPLE_CPY((%%eax, %2), (%%eax, %2, 2), (%%edx, %3), (%%edx, %3, 2))	3305	SIMPLE_CPY((%%eax, %2), (%%eax, %2, 2), (%%edx, %3), (%%edx, %3, 2))
		3306	#endif
3268		3307
3269	: : "r" (src),	3308	: : "r" (src),
3270	"r" (dst),	3309	"r" (dst),
3271	"r" (srcStride),	3310	"r" (srcStride),
		3311	#ifdef __PIC__
		3312	"m" (dstStride)
		3313	#else
3272	"r" (dstStride)	3314	"r" (dstStride)
		3315	#endif
3273	: "%eax", "%edx"	3316	: "%eax", "%edx"
3274	);	3317	);
3275	#else	3318	#else

Lines 716-722 Link Here

(-)work.orig/ffmpeg-0.4.9-pre1/libavcodec/msmpeg4.c (-1 / +1 lines)
716	necessitate to modify mpegvideo.c. The problem comes from the	716	necessitate to modify mpegvideo.c. The problem comes from the
717	fact they decided to store the quantized DC (which would lead	717	fact they decided to store the quantized DC (which would lead
718	to problems if Q could vary !) */	718	to problems if Q could vary !) */
719	#if defined ARCH_X86 && !defined PIC	719	#if defined ARCH_X86 && !defined __PIC__
720	asm volatile(	720	asm volatile(
721	"movl %3, %%eax \n\t"	721	"movl %3, %%eax \n\t"
722	"shrl $1, %%eax \n\t"	722	"shrl $1, %%eax \n\t"

Return to bug 67947