Attachment #89932 for bug #121871




/* public domain, do what you want */

#ifdef __PIC__
# define MUNG(sym)                 sym##@GOTOFF(%ebp)
# define MUNG_ARR(sym, args...)    sym##@GOTOFF(%ebp,##args)
#else
# define MUNG(sym)                 sym
# define MUNG_ARR(sym, args...)    sym(,##args)
#endif

#ifdef __PIC__
# undef __i686 /* gcc define gets in our way */
	.section .gnu.linkonce.t.__i686.get_pc_thunk.bp,"ax",@progbits
.globl __i686.get_pc_thunk.bp
	.hidden  __i686.get_pc_thunk.bp
	.type    __i686.get_pc_thunk.bp,@function
__i686.get_pc_thunk.bp:
	movl (%esp), %ebp
	ret
# define LOAD_PIC_REG_BP() \
	call __i686.get_pc_thunk.bp ; \
	addl $_GLOBAL_OFFSET_TABLE_, %ebp

	.section .gnu.linkonce.t.__i686.get_pc_thunk.si,"ax",@progbits
.globl __i686.get_pc_thunk.si
	.hidden  __i686.get_pc_thunk.si
	.type    __i686.get_pc_thunk.si,@function
__i686.get_pc_thunk.si:
	movl (%esp), %esi
	ret
# define LOAD_PIC_REG_SI() \
	call __i686.get_pc_thunk.si ; \
	addl $_GLOBAL_OFFSET_TABLE_, %esi
#else
# define LOAD_PIC_REG_BP()
# define LOAD_PIC_REG_SI()
#endif

Lines 53-69 scratch2: .quad 0 Link Here

(-)libdv-0.104-old/libdv/dct_block_mmx.S (-42 / +56 lines)
53	scratch3: .quad 0	53	scratch3: .quad 0
54	scratch4: .quad 0	54	scratch4: .quad 0
55		55
		56	#include "asm_common.S"
		57
56	.text	58	.text
57		59
58	.align 8	60	.align 8
59	.global _dv_dct_88_block_mmx	61	.global _dv_dct_88_block_mmx
		62	.hidden _dv_dct_88_block_mmx
		63	.type _dv_dct_88_block_mmx,@function
60	_dv_dct_88_block_mmx:	64	_dv_dct_88_block_mmx:
61		65
62	pushl %ebp	66	pushl %ebp
63	movl %esp, %ebp
64	pushl %esi	67	pushl %esi
65		68
66	movl 8(%ebp), %esi # source	69	LOAD_PIC_REG_BP()
		70
		71	movl 12(%esp), %esi # source
67		72
68	# column 0	73	# column 0
69	movq 16*0(%esi), %mm0 # v0	74	movq 16*0(%esi), %mm0 # v0
Lines 86-107 _dv_dct_88_block_mmx: Link Here
86		91
87	movq 16*3(%esi), %mm5 # v3	92	movq 16*3(%esi), %mm5 # v3
88	movq 16*4(%esi), %mm7 # v4	93	movq 16*4(%esi), %mm7 # v4
89	movq %mm7, scratch1 # scratch1: v4 ;	94	movq %mm7, MUNG(scratch1) # scratch1: v4 ;
90	movq %mm5, %mm7 # duplicate v3	95	movq %mm5, %mm7 # duplicate v3
91	paddw scratch1, %mm5 # v03: v3+v4	96	paddw MUNG(scratch1), %mm5 # v03: v3+v4
92	psubw scratch1, %mm7 # v04: v3-v4	97	psubw MUNG(scratch1), %mm7 # v04: v3-v4
93	movq %mm5, scratch2 # scratch2: v03	98	movq %mm5, MUNG(scratch2) # scratch2: v03
94	movq %mm0, %mm5 # mm5: v00	99	movq %mm0, %mm5 # mm5: v00
95		100
96	paddw scratch2, %mm0 # v10: v00+v03	101	paddw MUNG(scratch2), %mm0 # v10: v00+v03
97	psubw scratch2, %mm5 # v13: v00-v03	102	psubw MUNG(scratch2), %mm5 # v13: v00-v03
98	movq %mm3, scratch3 # scratch3: v02	103	movq %mm3, MUNG(scratch3) # scratch3: v02
99	movq %mm1, %mm3 # duplicate v01	104	movq %mm1, %mm3 # duplicate v01
100		105
101	paddw scratch3, %mm1 # v11: v01+v02	106	paddw MUNG(scratch3), %mm1 # v11: v01+v02
102	psubw scratch3, %mm3 # v12: v01-v02	107	psubw MUNG(scratch3), %mm3 # v12: v01-v02
103		108
104	movq %mm6, scratch4 # scratch4: v05	109	movq %mm6, MUNG(scratch4) # scratch4: v05
105	movq %mm0, %mm6 # duplicate v10	110	movq %mm0, %mm6 # duplicate v10
106		111
107	paddw %mm1, %mm0 # v10+v11	112	paddw %mm1, %mm0 # v10+v11
Lines 111-120 _dv_dct_88_block_mmx: Link Here
111	movq %mm6, 16*4(%esi) # out4: v10-v11	116	movq %mm6, 16*4(%esi) # out4: v10-v11
112		117
113	movq %mm4, %mm0 # mm0: v06	118	movq %mm4, %mm0 # mm0: v06
114	paddw scratch4, %mm4 # v15: v05+v06	119	paddw MUNG(scratch4), %mm4 # v15: v05+v06
115	paddw %mm2, %mm0 # v16: v07+v06	120	paddw %mm2, %mm0 # v16: v07+v06
116		121
117	pmulhw WA3, %mm4 # v35~: WA3*v15	122	pmulhw MUNG(WA3), %mm4 # v35~: WA3*v15
118	psllw $1, %mm4 # v35: compensate the coeefient scale	123	psllw $1, %mm4 # v35: compensate the coeefient scale
119		124
120	movq %mm4, %mm6 # duplicate v35	125	movq %mm4, %mm6 # duplicate v35
Lines 123-129 _dv_dct_88_block_mmx: Link Here
123		128
124	paddw %mm5, %mm3 # v22: v12+v13	129	paddw %mm5, %mm3 # v22: v12+v13
125		130
126	pmulhw WA1, %mm3 # v32~: WA1*v22	131	pmulhw MUNG(WA1), %mm3 # v32~: WA1*v22
127	psllw $16-NSHIFT, %mm3 # v32: compensate the coeefient scale	132	psllw $16-NSHIFT, %mm3 # v32: compensate the coeefient scale
128	movq %mm5, %mm6 # duplicate v13	133	movq %mm5, %mm6 # duplicate v13
129		134
Lines 134-146 _dv_dct_88_block_mmx: Link Here
134	movq %mm6, 16*6(%esi) # out6: v13-v32	139	movq %mm6, 16*6(%esi) # out6: v13-v32
135		140
136		141
137	paddw scratch4, %mm7 # v14n: v04+v05	142	paddw MUNG(scratch4), %mm7 # v14n: v04+v05
138	movq %mm0, %mm5 # duplicate v16	143	movq %mm0, %mm5 # duplicate v16
139		144
140	psubw %mm7, %mm0 # va1: v16-v14n	145	psubw %mm7, %mm0 # va1: v16-v14n
141	pmulhw WA5, %mm0 # va0~: va1*WA5	146	pmulhw MUNG(WA5), %mm0 # va0~: va1*WA5
142	pmulhw WA4, %mm5 # v36~~: v16*WA4	147	pmulhw MUNG(WA4), %mm5 # v36~~: v16*WA4
143	pmulhw WA2, %mm7 # v34~~: v14n*WA2	148	pmulhw MUNG(WA2), %mm7 # v34~~: v14n*WA2
144	psllw $16-WA4_SHIFT, %mm5 # v36: compensate the coeefient scale	149	psllw $16-WA4_SHIFT, %mm5 # v36: compensate the coeefient scale
145	psllw $16-NSHIFT, %mm7 # v34: compensate the coeefient scale	150	psllw $16-NSHIFT, %mm7 # v34: compensate the coeefient scale
146		151
Lines 188-209 _dv_dct_88_block_mmx: Link Here
188		193
189	movq 16*3(%esi), %mm5 # v3	194	movq 16*3(%esi), %mm5 # v3
190	movq 16*4(%esi), %mm7 # v4	195	movq 16*4(%esi), %mm7 # v4
191	movq %mm7, scratch1 # scratch1: v4 ;	196	movq %mm7, MUNG(scratch1) # scratch1: v4 ;
192	movq %mm5, %mm7 # duplicate v3	197	movq %mm5, %mm7 # duplicate v3
193	paddw scratch1, %mm5 # v03: v3+v4	198	paddw MUNG(scratch1), %mm5 # v03: v3+v4
194	psubw scratch1, %mm7 # v04: v3-v4	199	psubw MUNG(scratch1), %mm7 # v04: v3-v4
195	movq %mm5, scratch2 # scratch2: v03	200	movq %mm5, MUNG(scratch2) # scratch2: v03
196	movq %mm0, %mm5 # mm5: v00	201	movq %mm0, %mm5 # mm5: v00
197		202
198	paddw scratch2, %mm0 # v10: v00+v03	203	paddw MUNG(scratch2), %mm0 # v10: v00+v03
199	psubw scratch2, %mm5 # v13: v00-v03	204	psubw MUNG(scratch2), %mm5 # v13: v00-v03
200	movq %mm3, scratch3 # scratc3: v02	205	movq %mm3, MUNG(scratch3) # scratc3: v02
201	movq %mm1, %mm3 # duplicate v01	206	movq %mm1, %mm3 # duplicate v01
202		207
203	paddw scratch3, %mm1 # v11: v01+v02	208	paddw MUNG(scratch3), %mm1 # v11: v01+v02
204	psubw scratch3, %mm3 # v12: v01-v02	209	psubw MUNG(scratch3), %mm3 # v12: v01-v02
205		210
206	movq %mm6, scratch4 # scratc4: v05	211	movq %mm6, MUNG(scratch4) # scratc4: v05
207	movq %mm0, %mm6 # duplicate v10	212	movq %mm0, %mm6 # duplicate v10
208		213
209	paddw %mm1, %mm0 # v10+v11	214	paddw %mm1, %mm0 # v10+v11
Lines 213-222 _dv_dct_88_block_mmx: Link Here
213	movq %mm6, 16*4(%esi) # out4: v10-v11	218	movq %mm6, 16*4(%esi) # out4: v10-v11
214		219
215	movq %mm4, %mm0 # mm0: v06	220	movq %mm4, %mm0 # mm0: v06
216	paddw scratch4, %mm4 # v15: v05+v06	221	paddw MUNG(scratch4), %mm4 # v15: v05+v06
217	paddw %mm2, %mm0 # v16: v07+v06	222	paddw %mm2, %mm0 # v16: v07+v06
218		223
219	pmulhw WA3, %mm4 # v35~: WA3*v15	224	pmulhw MUNG(WA3), %mm4 # v35~: WA3*v15
220	psllw $16-NSHIFT, %mm4 # v35: compensate the coeefient scale	225	psllw $16-NSHIFT, %mm4 # v35: compensate the coeefient scale
221		226
222	movq %mm4, %mm6 # duplicate v35	227	movq %mm4, %mm6 # duplicate v35
Lines 225-231 _dv_dct_88_block_mmx: Link Here
225		230
226	paddw %mm5, %mm3 # v22: v12+v13	231	paddw %mm5, %mm3 # v22: v12+v13
227		232
228	pmulhw WA1, %mm3 # v32~: WA3*v15	233	pmulhw MUNG(WA1), %mm3 # v32~: WA3*v15
229	psllw $16-NSHIFT, %mm3 # v32: compensate the coeefient scale	234	psllw $16-NSHIFT, %mm3 # v32: compensate the coeefient scale
230	movq %mm5, %mm6 # duplicate v13	235	movq %mm5, %mm6 # duplicate v13
231		236
Lines 235-247 _dv_dct_88_block_mmx: Link Here
235	movq %mm5, 16*2(%esi) # out2: v13+v32	240	movq %mm5, 16*2(%esi) # out2: v13+v32
236	movq %mm6, 16*6(%esi) # out6: v13-v32	241	movq %mm6, 16*6(%esi) # out6: v13-v32
237		242
238	paddw scratch4, %mm7 # v14n: v04+v05	243	paddw MUNG(scratch4), %mm7 # v14n: v04+v05
239	movq %mm0, %mm5 # duplicate v16	244	movq %mm0, %mm5 # duplicate v16
240		245
241	psubw %mm7, %mm0 # va1: v16-v14n	246	psubw %mm7, %mm0 # va1: v16-v14n
242	pmulhw WA2, %mm7 # v34~~: v14n*WA2	247	pmulhw MUNG(WA2), %mm7 # v34~~: v14n*WA2
243	pmulhw WA5, %mm0 # va0~: va1*WA5	248	pmulhw MUNG(WA5), %mm0 # va0~: va1*WA5
244	pmulhw WA4, %mm5 # v36~~: v16*WA4	249	pmulhw MUNG(WA4), %mm5 # v36~~: v16*WA4
245	psllw $16-NSHIFT, %mm7	250	psllw $16-NSHIFT, %mm7
246	psllw $16-WA4_SHIFT, %mm5 # v36: compensate the coeffient	251	psllw $16-WA4_SHIFT, %mm5 # v36: compensate the coeffient
247	# scale note that WA4 is shifted 1 bit less than the others	252	# scale note that WA4 is shifted 1 bit less than the others
Lines 272-277 _dv_dct_88_block_mmx: Link Here
272		277
273	.align 8	278	.align 8
274	.global _dv_dct_block_mmx_postscale_88	279	.global _dv_dct_block_mmx_postscale_88
		280	.hidden _dv_dct_block_mmx_postscale_88
		281	.type _dv_dct_block_mmx_postscale_88,@function
275	_dv_dct_block_mmx_postscale_88:	282	_dv_dct_block_mmx_postscale_88:
276		283
277	pushl %ebp	284	pushl %ebp
Lines 748-761 _dv_dct_block_mmx_postscale_88: Link Here
748		755
749	.align 8	756	.align 8
750	.global _dv_dct_248_block_mmx	757	.global _dv_dct_248_block_mmx
		758	.hidden _dv_dct_248_block_mmx
		759	.type _dv_dct_248_block_mmx,@function
751	_dv_dct_248_block_mmx:	760	_dv_dct_248_block_mmx:
752		761
753	pushl %ebp	762	pushl %ebp
754	movl %esp, %ebp
755	pushl %esi	763	pushl %esi
756	pushl %edi	764	pushl %edi
757		765
758	movl 8(%ebp), %esi # source	766	LOAD_PIC_REG_BP()
		767
		768	movl 16(%esp), %esi # source
759		769
760	# column 0	770	# column 0
761		771
Lines 779-785 _dv_dct_248_block_mmx: Link Here
779	paddw %mm1, %mm0 # v20: v10+v11	789	paddw %mm1, %mm0 # v20: v10+v11
780	psubw %mm1, %mm3 # v21: v10-v11	790	psubw %mm1, %mm3 # v21: v10-v11
781		791
782	pmulhw WA1, %mm5 # v32~: WA1*v22	792	pmulhw MUNG(WA1), %mm5 # v32~: WA1*v22
783	movq %mm4, %mm2	793	movq %mm4, %mm2
784	psllw $16-NSHIFT, %mm5 # v32: compensate the coeffient scale	794	psllw $16-NSHIFT, %mm5 # v32: compensate the coeffient scale
785		795
Lines 818-824 _dv_dct_248_block_mmx: Link Here
818	paddw %mm1, %mm0 # v20: v10+v11	828	paddw %mm1, %mm0 # v20: v10+v11
819	psubw %mm1, %mm3 # v21: v10-v11	829	psubw %mm1, %mm3 # v21: v10-v11
820		830
821	pmulhw WA1, %mm5 # v32~: WA1*v22	831	pmulhw MUNG(WA1), %mm5 # v32~: WA1*v22
822	movq %mm4, %mm2	832	movq %mm4, %mm2
823	psllw $16-NSHIFT, %mm5 # v32: compensate the coeffient scale	833	psllw $16-NSHIFT, %mm5 # v32: compensate the coeffient scale
824		834
Lines 855-861 _dv_dct_248_block_mmx: Link Here
855	paddw %mm1, %mm0 # v20: v10+v11	865	paddw %mm1, %mm0 # v20: v10+v11
856	psubw %mm1, %mm3 # v21: v10-v11	866	psubw %mm1, %mm3 # v21: v10-v11
857		867
858	pmulhw WA1, %mm5 # v32~: WA1*v22	868	pmulhw MUNG(WA1), %mm5 # v32~: WA1*v22
859	movq %mm4, %mm2	869	movq %mm4, %mm2
860	psllw $16-NSHIFT, %mm5 # v32: compensate the coeffient scale	870	psllw $16-NSHIFT, %mm5 # v32: compensate the coeffient scale
861		871
Lines 892-898 _dv_dct_248_block_mmx: Link Here
892	paddw %mm1, %mm0 # v20: v10+v11	902	paddw %mm1, %mm0 # v20: v10+v11
893	psubw %mm1, %mm3 # v21: v10-v11	903	psubw %mm1, %mm3 # v21: v10-v11
894		904
895	pmulhw WA1, %mm5 # v32~: WA1*v22	905	pmulhw MUNG(WA1), %mm5 # v32~: WA1*v22
896	movq %mm4, %mm2	906	movq %mm4, %mm2
897	psllw $16-NSHIFT, %mm5 # v32: compensate the coeffient scale	907	psllw $16-NSHIFT, %mm5 # v32: compensate the coeffient scale
898		908
Lines 912-917 _dv_dct_248_block_mmx: Link Here
912		922
913	.align 8	923	.align 8
914	.global _dv_dct_248_block_mmx_post_sum	924	.global _dv_dct_248_block_mmx_post_sum
		925	.hidden _dv_dct_248_block_mmx_post_sum
		926	.type _dv_dct_248_block_mmx_post_sum,@function
915	_dv_dct_248_block_mmx_post_sum:	927	_dv_dct_248_block_mmx_post_sum:
916		928
917	pushl %ebp	929	pushl %ebp
Lines 992-997 _dv_dct_248_block_mmx_post_sum: Link Here
992		1004
993	.align 8	1005	.align 8
994	.global _dv_dct_block_mmx_postscale_248	1006	.global _dv_dct_block_mmx_postscale_248
		1007	.hidden _dv_dct_block_mmx_postscale_248
		1008	.type _dv_dct_block_mmx_postscale_248,@function
995	_dv_dct_block_mmx_postscale_248:	1009	_dv_dct_block_mmx_postscale_248:
996		1010
997	pushl %ebp	1011	pushl %ebp

Lines 57-62 scratch4: .quad 0 Link Here

(-)libdv-0.104-old/libdv/dct_block_mmx_x86_64.S (+4 lines)
57		57
58	.align 8	58	.align 8
59	.global _dv_dct_88_block_mmx_x86_64	59	.global _dv_dct_88_block_mmx_x86_64
		60	.hidden _dv_dct_88_block_mmx_x86_64
		61	.type _dv_dct_88_block_mmx_x86_64,@function
60	_dv_dct_88_block_mmx_x86_64:	62	_dv_dct_88_block_mmx_x86_64:
61		63
62	/* void _dv_dct_88_block_mmx_x86_64(int16_t* block); */	64	/* void _dv_dct_88_block_mmx_x86_64(int16_t* block); */
Lines 269-274 _dv_dct_88_block_mmx_x86_64: Link Here
269		271
270	.align 8	272	.align 8
271	.global _dv_dct_block_mmx_x86_64_postscale_88	273	.global _dv_dct_block_mmx_x86_64_postscale_88
		274	.hidden _dv_dct_block_mmx_x86_64_postscale_88
		275	.type _dv_dct_block_mmx_x86_64_postscale_88,@function
272	_dv_dct_block_mmx_x86_64_postscale_88:	276	_dv_dct_block_mmx_x86_64_postscale_88:
273		277
274	/* void _dv_dct_block_mmx_x86_64_postscale_88(int16_t* block, int16_t* postscale_matrix); */	278	/* void _dv_dct_block_mmx_x86_64_postscale_88(int16_t* block, int16_t* postscale_matrix); */

Lines 205-210 dv_reconfigure(int clamp_luma, int clamp Link Here

(-)libdv-0.104-old/libdv/dv.c (-2 / +5 lines)
205	} /* dv_reconfigure */	205	} /* dv_reconfigure */
206		206
207		207
		208	extern uint8_t dv_quant_offset[4];
		209	extern uint8_t dv_quant_shifts[22][4];
		210
208	static inline void	211	static inline void
209	dv_decode_macroblock(dv_decoder_t dv, dv_macroblock_t mb, unsigned int quality) {	212	dv_decode_macroblock(dv_decoder_t dv, dv_macroblock_t mb, unsigned int quality) {
210	int i;	213	int i;
Lines 218-224 dv_decode_macroblock(dv_decoder_t *dv, d Link Here
218	dv_idct_248 (co248, mb->b[i].coeffs);	221	dv_idct_248 (co248, mb->b[i].coeffs);
219	} else {	222	} else {
220	#if ARCH_X86	223	#if ARCH_X86
221	_dv_quant_88_inverse_x86(mb->b[i].coeffs,mb->qno,mb->b[i].class_no);	224	_dv_quant_88_inverse_x86(mb->b[i].coeffs,mb->qno,mb->b[i].class_no,dv_quant_offset,dv_quant_shifts);
222	_dv_idct_88(mb->b[i].coeffs);	225	_dv_idct_88(mb->b[i].coeffs);
223	#elif ARCH_X86_64	226	#elif ARCH_X86_64
224	_dv_quant_88_inverse_x86_64(mb->b[i].coeffs,mb->qno,mb->b[i].class_no);	227	_dv_quant_88_inverse_x86_64(mb->b[i].coeffs,mb->qno,mb->b[i].class_no);
Lines 250-256 dv_decode_video_segment(dv_decoder_t *dv Link Here
250	dv_idct_248 (co248, mb->b[b].coeffs);	253	dv_idct_248 (co248, mb->b[b].coeffs);
251	} else {	254	} else {
252	#if ARCH_X86	255	#if ARCH_X86
253	_dv_quant_88_inverse_x86(bl->coeffs,mb->qno,bl->class_no);	256	_dv_quant_88_inverse_x86(bl->coeffs,mb->qno,bl->class_no,dv_quant_offset,dv_quant_shifts);
254	_dv_weight_88_inverse(bl->coeffs);	257	_dv_weight_88_inverse(bl->coeffs);
255	_dv_idct_88(bl->coeffs);	258	_dv_idct_88(bl->coeffs);
256	#elif ARCH_X86_64	259	#elif ARCH_X86_64




}

extern unsigned long _dv_vlc_encode_block_mmx(dv_coeff_t* coeffs,
					  dv_vlc_entry_t ** out,
					  dv_vlc_entry_t * lookup);

extern unsigned long _dv_vlc_encode_block_mmx_x86_64(dv_coeff_t* coeffs,
					  dv_vlc_entry_t ** out);

#elif ARCH_X86
	int num_bits;

	num_bits = _dv_vlc_encode_block_mmx(coeffs, &o, vlc_encode_lookup);
	emms();
#else
	int num_bits;

	return num_bits;
}

extern unsigned long _dv_vlc_num_bits_block_x86(dv_coeff_t* coeffs, unsigned char* lookup);
extern unsigned long _dv_vlc_num_bits_block_x86_64(dv_coeff_t* coeffs);

extern unsigned long _dv_vlc_num_bits_block(dv_coeff_t* coeffs)

#elif ARCH_X86_64
	return _dv_vlc_num_bits_block_x86_64(coeffs);
#else
	return _dv_vlc_num_bits_block_x86(coeffs, vlc_num_bits_lookup);
#endif
}


Lines 23-35 Link Here

(-)libdv-0.104-old/libdv/encode_x86.S (-9 / +23 lines)
23	* The libdv homepage is http://libdv.sourceforge.net/.	23	* The libdv homepage is http://libdv.sourceforge.net/.
24	*/	24	*/
25		25
26	.data
27	ALLONE: .word 1,1,1,1
28	VLCADDMASK: .byte 255,0,0,0,255,0,0,0
29
30	.text	26	.text
31		27
32	.global _dv_vlc_encode_block_mmx	28	.global _dv_vlc_encode_block_mmx
		29	.hidden _dv_vlc_encode_block_mmx
		30	.type _dv_vlc_encode_block_mmx,@function
33	_dv_vlc_encode_block_mmx:	31	_dv_vlc_encode_block_mmx:
34	pushl %ebx	32	pushl %ebx
35	pushl %esi	33	pushl %esi
Lines 45-55 _dv_vlc_encode_block_mmx: Link Here
45		43
46	movl $63, %ecx	44	movl $63, %ecx
47		45
48	movl vlc_encode_lookup, %esi	46	movl 4+4*4+8(%esp), %esi # vlc_encode_lookup
49		47
50	pxor %mm0, %mm0	48	pxor %mm0, %mm0
51	pxor %mm2, %mm2	49	pxor %mm2, %mm2
52	movq VLCADDMASK, %mm1	50	pushl $0x000000FF # these four lines
		51	pushl $0x000000FF # load VLCADDMASK
		52	movq (%esp), %mm1 # into %mm1 off the stack
		53	addl $8, %esp # --> no TEXTRELs
53	xorl %ebp, %ebp	54	xorl %ebp, %ebp
54	subl $8, %edx	55	subl $8, %edx
55	vlc_encode_block_mmx_loop:	56	vlc_encode_block_mmx_loop:
Lines 106-111 vlc_encode_block_out: Link Here
106	ret	107	ret
107		108
108	.global _dv_vlc_num_bits_block_x86	109	.global _dv_vlc_num_bits_block_x86
		110	.hidden _dv_vlc_num_bits_block_x86
		111	.type _dv_vlc_num_bits_block_x86,@function
109	_dv_vlc_num_bits_block_x86:	112	_dv_vlc_num_bits_block_x86:
110	pushl %ebx	113	pushl %ebx
111	pushl %esi	114	pushl %esi
Lines 121-127 _dv_vlc_num_bits_block_x86: Link Here
121	addl $2, %edi	124	addl $2, %edi
122		125
123	movl $63, %ecx	126	movl $63, %ecx
124	movl vlc_num_bits_lookup, %esi	127	movl 4+4*4+4(%esp), %esi # vlc_num_bits_lookup
125		128
126	vlc_num_bits_block_x86_loop:	129	vlc_num_bits_block_x86_loop:
127	movw (%edi), %ax	130	movw (%edi), %ax
Lines 164-169 vlc_num_bits_block_out: Link Here
164	ret	167	ret
165		168
166	.global _dv_vlc_encode_block_pass_1_x86	169	.global _dv_vlc_encode_block_pass_1_x86
		170	.hidden _dv_vlc_encode_block_pass_1_x86
		171	.type _dv_vlc_encode_block_pass_1_x86,@function
167	_dv_vlc_encode_block_pass_1_x86:	172	_dv_vlc_encode_block_pass_1_x86:
168	pushl %ebx	173	pushl %ebx
169	pushl %esi	174	pushl %esi
Lines 240-245 vlc_encode_block_pass1_x86_out: Link Here
240	ret	245	ret
241		246
242	.global _dv_classify_mmx	247	.global _dv_classify_mmx
		248	.hidden _dv_classify_mmx
		249	.type _dv_classify_mmx,@function
243	_dv_classify_mmx:	250	_dv_classify_mmx:
244		251
245	pushl %ebp	252	pushl %ebp
Lines 345-350 _dv_classify_mmx: Link Here
345	don't know why... */	352	don't know why... */
346		353
347	.global _dv_reorder_block_mmx	354	.global _dv_reorder_block_mmx
		355	.hidden _dv_reorder_block_mmx
		356	.type _dv_reorder_block_mmx,@function
348	_dv_reorder_block_mmx:	357	_dv_reorder_block_mmx:
349		358
350	pushl %ebp	359	pushl %ebp
Lines 460-465 reorder_loop: Link Here
460	ret	469	ret
461		470
462	.global _dv_need_dct_248_mmx_rows	471	.global _dv_need_dct_248_mmx_rows
		472	.hidden _dv_need_dct_248_mmx_rows
		473	.type _dv_need_dct_248_mmx_rows,@function
463	_dv_need_dct_248_mmx_rows:	474	_dv_need_dct_248_mmx_rows:
464		475
465	pushl %ebp	476	pushl %ebp
Lines 579-586 _dv_need_dct_248_mmx_rows: Link Here
579	paddw %mm5, %mm1	590	paddw %mm5, %mm1
580		591
581	paddw %mm1, %mm0	592	paddw %mm1, %mm0
582		593
583	pmaddwd ALLONE, %mm0	594	pushl $0x00010001 # these four lines
		595	pushl $0x00010001 # load ALLONE
		596	pmaddwd (%esp), %mm0 # into %mm0 off the stack
		597	addl $8, %esp # --> no TEXTRELs
584	movq %mm0, %mm1	598	movq %mm0, %mm1
585	psrlq $32, %mm1	599	psrlq $32, %mm1
586	paddd %mm1, %mm0	600	paddd %mm1, %mm0

Lines 30-35 VLCADDMASK: .byte 255,0,0,0,255,0,0,0 Link Here

(-)libdv-0.104-old/libdv/encode_x86_64.S (+12 lines)
30	.text	30	.text
31		31
32	.global _dv_vlc_encode_block_mmx_x86_64	32	.global _dv_vlc_encode_block_mmx_x86_64
		33	.hidden _dv_vlc_encode_block_mmx_x86_64
		34	.type _dv_vlc_encode_block_mmx_x86_64,@function
33	_dv_vlc_encode_block_mmx_x86_64:	35	_dv_vlc_encode_block_mmx_x86_64:
34		36
35	/* extern unsigned long _dv_vlc_encode_block_mmx(dv_coeff_t* coeffs,	37	/* extern unsigned long _dv_vlc_encode_block_mmx(dv_coeff_t* coeffs,
Lines 113-118 vlc_encode_block_out: Link Here
113	ret	115	ret
114		116
115	.global _dv_vlc_num_bits_block_x86_64	117	.global _dv_vlc_num_bits_block_x86_64
		118	.hidden _dv_vlc_num_bits_block_x86_64
		119	.type _dv_vlc_num_bits_block_x86_64,@function
116	_dv_vlc_num_bits_block_x86_64:	120	_dv_vlc_num_bits_block_x86_64:
117		121
118	/* extern unsigned long _dv_vlc_num_bits_block_x86_64(dv_coeff_t* coeffs); */	122	/* extern unsigned long _dv_vlc_num_bits_block_x86_64(dv_coeff_t* coeffs); */
Lines 173-178 vlc_num_bits_block_out: Link Here
173	ret	177	ret
174		178
175	.global _dv_vlc_encode_block_pass_1_x86_64	179	.global _dv_vlc_encode_block_pass_1_x86_64
		180	.hidden _dv_vlc_encode_block_pass_1_x86_64
		181	.type _dv_vlc_encode_block_pass_1_x86_64,@function
176	_dv_vlc_encode_block_pass_1_x86_64:	182	_dv_vlc_encode_block_pass_1_x86_64:
177		183
178	/*	184	/*
Lines 251-256 vlc_encode_block_pass1_x86_out: Link Here
251	ret	257	ret
252		258
253	.global _dv_classify_mmx_x86_64	259	.global _dv_classify_mmx_x86_64
		260	.hidden _dv_classify_mmx_x86_64
		261	.type _dv_classify_mmx_x86_64,@function
254	_dv_classify_mmx_x86_64:	262	_dv_classify_mmx_x86_64:
255		263
256	/* extern int _dv_classify_mmx_x86_64(dv_coeff_t * a, rdi	264	/* extern int _dv_classify_mmx_x86_64(dv_coeff_t * a, rdi
Lines 355-360 _dv_classify_mmx_x86_64: Link Here
355	don't know why... */	363	don't know why... */
356		364
357	.global _dv_reorder_block_mmx_x86_64	365	.global _dv_reorder_block_mmx_x86_64
		366	.hidden _dv_reorder_block_mmx_x86_64
		367	.type _dv_reorder_block_mmx_x86_64,@function
358	_dv_reorder_block_mmx_x86_64:	368	_dv_reorder_block_mmx_x86_64:
359		369
360	/extern int _dv_reorder_block_mmx_x86_64(dv_coeff_t a, rdi	370	/extern int _dv_reorder_block_mmx_x86_64(dv_coeff_t a, rdi
Lines 469-474 reorder_loop: Link Here
469	ret	479	ret
470		480
471	.global _dv_need_dct_248_mmx_x86_64_rows	481	.global _dv_need_dct_248_mmx_x86_64_rows
		482	.hidden _dv_need_dct_248_mmx_x86_64_rows
		483	.type _dv_need_dct_248_mmx_x86_64_rows,@function
472	_dv_need_dct_248_mmx_x86_64_rows:	484	_dv_need_dct_248_mmx_x86_64_rows:
473		485
474	/* extern int _dv_need_dct_248_mmx_x86_64_rows(dv_coeff_t * bl); rdi */	486	/* extern int _dv_need_dct_248_mmx_x86_64_rows(dv_coeff_t * bl); rdi */

Lines 8-23 Link Here

(-)libdv-0.104-old/libdv/idct_block_mmx.S (-32 / +38 lines)
8		8
9		9
10		10
		11	#include "asm_common.S"
		12
11	.text	13	.text
		14
12	.align 4	15	.align 4
13	.globl _dv_idct_block_mmx	16	.globl _dv_idct_block_mmx
		17	.hidden _dv_idct_block_mmx
14	.type _dv_idct_block_mmx,@function	18	.type _dv_idct_block_mmx,@function
15	_dv_idct_block_mmx:	19	_dv_idct_block_mmx:
16	pushl %ebp	20	pushl %ebp
17	movl %esp,%ebp
18	pushl %esi	21	pushl %esi
19	leal preSC, %ecx	22
20	movl 8(%ebp),%esi /* source matrix */	23	LOAD_PIC_REG_BP()
		24
		25	leal MUNG(preSC), %ecx
		26	movl 12(%esp),%esi /* source matrix */
21		27
22	/*	28	/*
23	* column 0: even part	29	* column 0: even part
Lines 35-41 _dv_idct_block_mmx: Link Here
35	movq %mm1, %mm2 /* added 11/1/96 */	41	movq %mm1, %mm2 /* added 11/1/96 */
36	pmulhw 88(%esi),%mm5 / V8 */	42	pmulhw 88(%esi),%mm5 / V8 */
37	psubsw %mm0, %mm1 /* V16 */	43	psubsw %mm0, %mm1 /* V16 */
38	pmulhw x5a825a825a825a82, %mm1 /* 23170 ->V18 */	44	pmulhw MUNG(x5a825a825a825a82), %mm1 /* 23170 ->V18 */
39	paddsw %mm0, %mm2 /* V17 */	45	paddsw %mm0, %mm2 /* V17 */
40	movq %mm2, %mm0 /* duplicate V17 */	46	movq %mm2, %mm0 /* duplicate V17 */
41	psraw $1, %mm2 /* t75=t82 */	47	psraw $1, %mm2 /* t75=t82 */
Lines 76-82 _dv_idct_block_mmx: Link Here
76	paddsw %mm0, %mm3 /* V29 ; free mm0 */	82	paddsw %mm0, %mm3 /* V29 ; free mm0 */
77	movq %mm7, %mm1 /* duplicate V26 */	83	movq %mm7, %mm1 /* duplicate V26 */
78	psraw $1, %mm3 /* t91=t94 */	84	psraw $1, %mm3 /* t91=t94 */
79	pmulhw x539f539f539f539f,%mm7 /* V33 */	85	pmulhw MUNG(x539f539f539f539f),%mm7 /* V33 */
80	psraw $1, %mm1 /* t96 */	86	psraw $1, %mm1 /* t96 */
81	movq %mm5, %mm0 /* duplicate V2 */	87	movq %mm5, %mm0 /* duplicate V2 */
82	psraw $2, %mm4 /* t85=t87 */	88	psraw $2, %mm4 /* t85=t87 */
Lines 84-98 _dv_idct_block_mmx: Link Here
84	psubsw %mm4, %mm0 /* V28 ; free mm4 */	90	psubsw %mm4, %mm0 /* V28 ; free mm4 */
85	movq %mm0, %mm2 /* duplicate V28 */	91	movq %mm0, %mm2 /* duplicate V28 */
86	psraw $1, %mm5 /* t90=t93 */	92	psraw $1, %mm5 /* t90=t93 */
87	pmulhw x4546454645464546,%mm0 /* V35 */	93	pmulhw MUNG(x4546454645464546),%mm0 /* V35 */
88	psraw $1, %mm2 /* t97 */	94	psraw $1, %mm2 /* t97 */
89	movq %mm5, %mm4 /* duplicate t90=t93 */	95	movq %mm5, %mm4 /* duplicate t90=t93 */
90	psubsw %mm2, %mm1 /* V32 ; free mm2 */	96	psubsw %mm2, %mm1 /* V32 ; free mm2 */
91	pmulhw x61f861f861f861f8,%mm1 /* V36 */	97	pmulhw MUNG(x61f861f861f861f8),%mm1 /* V36 */
92	psllw $1, %mm7 /* t107 */	98	psllw $1, %mm7 /* t107 */
93	paddsw %mm3, %mm5 /* V31 */	99	paddsw %mm3, %mm5 /* V31 */
94	psubsw %mm3, %mm4 /* V30 ; free mm3 */	100	psubsw %mm3, %mm4 /* V30 ; free mm3 */
95	pmulhw x5a825a825a825a82,%mm4 /* V34 */	101	pmulhw MUNG(x5a825a825a825a82),%mm4 /* V34 */
96	nop	102	nop
97	psubsw %mm1, %mm0 /* V38 */	103	psubsw %mm1, %mm0 /* V38 */
98	psubsw %mm7, %mm1 /* V37 ; free mm7 */	104	psubsw %mm7, %mm1 /* V37 ; free mm7 */
Lines 159-165 _dv_idct_block_mmx: Link Here
159	psubsw %mm7, %mm1 /* V50 */	165	psubsw %mm7, %mm1 /* V50 */
160	pmulhw 89(%esi), %mm5 / V9 */	166	pmulhw 89(%esi), %mm5 / V9 */
161	paddsw %mm7, %mm2 /* V51 */	167	paddsw %mm7, %mm2 /* V51 */
162	pmulhw x5a825a825a825a82, %mm1 /* 23170 ->V52 */	168	pmulhw MUNG(x5a825a825a825a82), %mm1 /* 23170 ->V52 */
163	movq %mm2, %mm6 /* duplicate V51 */	169	movq %mm2, %mm6 /* duplicate V51 */
164	psraw $1, %mm2 /* t138=t144 */	170	psraw $1, %mm2 /* t138=t144 */
165	movq %mm3, %mm4 /* duplicate V1 */	171	movq %mm3, %mm4 /* duplicate V1 */
Lines 200-210 _dv_idct_block_mmx: Link Here
200	* even more by doing the correction step in a later stage when the number	206	* even more by doing the correction step in a later stage when the number
201	* is actually multiplied by 16	207	* is actually multiplied by 16
202	*/	208	*/
203	paddw x0005000200010001, %mm4	209	paddw MUNG(x0005000200010001), %mm4
204	psubsw %mm6, %mm3 /* V60 ; free mm6 */	210	psubsw %mm6, %mm3 /* V60 ; free mm6 */
205	psraw $1, %mm0 /* t154=t156 */	211	psraw $1, %mm0 /* t154=t156 */
206	movq %mm3, %mm1 /* duplicate V60 */	212	movq %mm3, %mm1 /* duplicate V60 */
207	pmulhw x539f539f539f539f, %mm1 /* V67 */	213	pmulhw MUNG(x539f539f539f539f), %mm1 /* V67 */
208	movq %mm5, %mm6 /* duplicate V3 */	214	movq %mm5, %mm6 /* duplicate V3 */
209	psraw $2, %mm4 /* t148=t150 */	215	psraw $2, %mm4 /* t148=t150 */
210	paddsw %mm4, %mm5 /* V61 */	216	paddsw %mm4, %mm5 /* V61 */
Lines 213-225 _dv_idct_block_mmx: Link Here
213	psllw $1, %mm1 /* t169 */	219	psllw $1, %mm1 /* t169 */
214	paddsw %mm0, %mm5 /* V65 -> result */	220	paddsw %mm0, %mm5 /* V65 -> result */
215	psubsw %mm0, %mm4 /* V64 ; free mm0 */	221	psubsw %mm0, %mm4 /* V64 ; free mm0 */
216	pmulhw x5a825a825a825a82, %mm4 /* V68 */	222	pmulhw MUNG(x5a825a825a825a82), %mm4 /* V68 */
217	psraw $1, %mm3 /* t158 */	223	psraw $1, %mm3 /* t158 */
218	psubsw %mm6, %mm3 /* V66 */	224	psubsw %mm6, %mm3 /* V66 */
219	movq %mm5, %mm2 /* duplicate V65 */	225	movq %mm5, %mm2 /* duplicate V65 */
220	pmulhw x61f861f861f861f8, %mm3 /* V70 */	226	pmulhw MUNG(x61f861f861f861f8), %mm3 /* V70 */
221	psllw $1, %mm6 /* t165 */	227	psllw $1, %mm6 /* t165 */
222	pmulhw x4546454645464546, %mm6 /* V69 */	228	pmulhw MUNG(x4546454645464546), %mm6 /* V69 */
223	psraw $1, %mm2 /* t172 */	229	psraw $1, %mm2 /* t172 */
224	/* moved from next block */	230	/* moved from next block */
225	movq 85(%esi), %mm0 / V56 */	231	movq 85(%esi), %mm0 / V56 */
Lines 344-350 _dv_idct_block_mmx: Link Here
344	* movq 8*13(%esi), %mm4 tmt13	350	* movq 8*13(%esi), %mm4 tmt13
345	*/	351	*/
346	psubsw %mm4, %mm3 /* V134 */	352	psubsw %mm4, %mm3 /* V134 */
347	pmulhw x5a825a825a825a82, %mm3 /* 23170 ->V136 */	353	pmulhw MUNG(x5a825a825a825a82), %mm3 /* 23170 ->V136 */
348	movq 89(%esi), %mm6 / tmt9 */	354	movq 89(%esi), %mm6 / tmt9 */
349	paddsw %mm4, %mm5 /* V135 ; mm4 free */	355	paddsw %mm4, %mm5 /* V135 ; mm4 free */
350	movq %mm0, %mm4 /* duplicate tmt1 */	356	movq %mm0, %mm4 /* duplicate tmt1 */
Lines 373-389 _dv_idct_block_mmx: Link Here
373	psubsw %mm7, %mm0 /* V144 */	379	psubsw %mm7, %mm0 /* V144 */
374	movq %mm0, %mm3 /* duplicate V144 */	380	movq %mm0, %mm3 /* duplicate V144 */
375	paddsw %mm7, %mm2 /* V147 ; free mm7 */	381	paddsw %mm7, %mm2 /* V147 ; free mm7 */
376	pmulhw x539f539f539f539f, %mm0 /* 21407-> V151 */	382	pmulhw MUNG(x539f539f539f539f), %mm0 /* 21407-> V151 */
377	movq %mm1, %mm7 /* duplicate tmt3 */	383	movq %mm1, %mm7 /* duplicate tmt3 */
378	paddsw %mm5, %mm7 /* V145 */	384	paddsw %mm5, %mm7 /* V145 */
379	psubsw %mm5, %mm1 /* V146 ; free mm5 */	385	psubsw %mm5, %mm1 /* V146 ; free mm5 */
380	psubsw %mm1, %mm3 /* V150 */	386	psubsw %mm1, %mm3 /* V150 */
381	movq %mm7, %mm5 /* duplicate V145 */	387	movq %mm7, %mm5 /* duplicate V145 */
382	pmulhw x4546454645464546, %mm1 /* 17734-> V153 */	388	pmulhw MUNG(x4546454645464546), %mm1 /* 17734-> V153 */
383	psubsw %mm2, %mm5 /* V148 */	389	psubsw %mm2, %mm5 /* V148 */
384	pmulhw x61f861f861f861f8, %mm3 /* 25080-> V154 */	390	pmulhw MUNG(x61f861f861f861f8), %mm3 /* 25080-> V154 */
385	psllw $2, %mm0 /* t311 */	391	psllw $2, %mm0 /* t311 */
386	pmulhw x5a825a825a825a82, %mm5 /* 23170-> V152 */	392	pmulhw MUNG(x5a825a825a825a82), %mm5 /* 23170-> V152 */
387	paddsw %mm2, %mm7 /* V149 ; free mm2 */	393	paddsw %mm2, %mm7 /* V149 ; free mm2 */
388	psllw $1, %mm1 /* t313 */	394	psllw $1, %mm1 /* t313 */
389	nop /* without the nop - freeze here for one clock */	395	nop /* without the nop - freeze here for one clock */
Lines 409-415 _dv_idct_block_mmx: Link Here
409	paddsw %mm3, %mm6 /* V164 ; free mm3 */	415	paddsw %mm3, %mm6 /* V164 ; free mm3 */
410	movq %mm4, %mm3 /* duplicate V142 */	416	movq %mm4, %mm3 /* duplicate V142 */
411	psubsw %mm5, %mm4 /* V165 ; free mm5 */	417	psubsw %mm5, %mm4 /* V165 ; free mm5 */
412	movq %mm2, scratch7 /* out7 */	418	movq %mm2, MUNG(scratch7) /* out7 */
413	psraw $4, %mm6	419	psraw $4, %mm6
414	psraw $4, %mm4	420	psraw $4, %mm4
415	paddsw %mm5, %mm3 /* V162 */	421	paddsw %mm5, %mm3 /* V162 */
Lines 420-430 _dv_idct_block_mmx: Link Here
420	*/	426	*/
421	movq %mm6, 89(%esi) / out9 */	427	movq %mm6, 89(%esi) / out9 */
422	paddsw %mm1, %mm0 /* V161 */	428	paddsw %mm1, %mm0 /* V161 */
423	movq %mm3, scratch5 /* out5 */	429	movq %mm3, MUNG(scratch5) /* out5 */
424	psubsw %mm1, %mm5 /* V166 ; free mm1 */	430	psubsw %mm1, %mm5 /* V166 ; free mm1 */
425	movq %mm4, 811(%esi) / out11 */	431	movq %mm4, 811(%esi) / out11 */
426	psraw $4, %mm5	432	psraw $4, %mm5
427	movq %mm0, scratch3 /* out3 */	433	movq %mm0, MUNG(scratch3) /* out3 */
428	movq %mm2, %mm4 /* duplicate V140 */	434	movq %mm2, %mm4 /* duplicate V140 */
429	movq %mm5, 813(%esi) / out13 */	435	movq %mm5, 813(%esi) / out13 */
430	paddsw %mm7, %mm2 /* V160 */	436	paddsw %mm7, %mm2 /* V160 */
Lines 434-440 _dv_idct_block_mmx: Link Here
434	/* moved from the next block */	440	/* moved from the next block */
435	movq 8*3(%esi), %mm7	441	movq 8*3(%esi), %mm7
436	psraw $4, %mm4	442	psraw $4, %mm4
437	movq %mm2, scratch1 /* out1 */	443	movq %mm2, MUNG(scratch1) /* out1 */
438	/* moved from the next block */	444	/* moved from the next block */
439	movq %mm0, %mm1	445	movq %mm0, %mm1
440	movq %mm4, 815(%esi) / out15 */	446	movq %mm4, 815(%esi) / out15 */
Lines 491-505 _dv_idct_block_mmx: Link Here
491	paddsw %mm4, %mm3 /* V113 ; free mm4 */	497	paddsw %mm4, %mm3 /* V113 ; free mm4 */
492	movq %mm0, %mm4 /* duplicate V110 */	498	movq %mm0, %mm4 /* duplicate V110 */
493	paddsw %mm1, %mm2 /* V111 */	499	paddsw %mm1, %mm2 /* V111 */
494	pmulhw x539f539f539f539f, %mm0 /* 21407-> V117 */	500	pmulhw MUNG(x539f539f539f539f), %mm0 /* 21407-> V117 */
495	psubsw %mm1, %mm5 /* V112 ; free mm1 */	501	psubsw %mm1, %mm5 /* V112 ; free mm1 */
496	psubsw %mm5, %mm4 /* V116 */	502	psubsw %mm5, %mm4 /* V116 */
497	movq %mm2, %mm1 /* duplicate V111 */	503	movq %mm2, %mm1 /* duplicate V111 */
498	pmulhw x4546454645464546, %mm5 /* 17734-> V119 */	504	pmulhw MUNG(x4546454645464546), %mm5 /* 17734-> V119 */
499	psubsw %mm3, %mm2 /* V114 */	505	psubsw %mm3, %mm2 /* V114 */
500	pmulhw x61f861f861f861f8, %mm4 /* 25080-> V120 */	506	pmulhw MUNG(x61f861f861f861f8), %mm4 /* 25080-> V120 */
501	paddsw %mm3, %mm1 /* V115 ; free mm3 */	507	paddsw %mm3, %mm1 /* V115 ; free mm3 */
502	pmulhw x5a825a825a825a82, %mm2 /* 23170-> V118 */	508	pmulhw MUNG(x5a825a825a825a82), %mm2 /* 23170-> V118 */
503	psllw $2, %mm0 /* t266 */	509	psllw $2, %mm0 /* t266 */
504	movq %mm1, (%esi) /* save V115 */	510	movq %mm1, (%esi) /* save V115 */
505	psllw $1, %mm5 /* t268 */	511	psllw $1, %mm5 /* t268 */
Lines 517-523 _dv_idct_block_mmx: Link Here
517	movq %mm6, %mm3 /* duplicate tmt4 */	523	movq %mm6, %mm3 /* duplicate tmt4 */
518	psubsw %mm0, %mm6 /* V100 */	524	psubsw %mm0, %mm6 /* V100 */
519	paddsw %mm0, %mm3 /* V101 ; free mm0 */	525	paddsw %mm0, %mm3 /* V101 ; free mm0 */
520	pmulhw x5a825a825a825a82, %mm6 /* 23170 ->V102 */	526	pmulhw MUNG(x5a825a825a825a82), %mm6 /* 23170 ->V102 */
521	movq %mm7, %mm5 /* duplicate tmt0 */	527	movq %mm7, %mm5 /* duplicate tmt0 */
522	movq 88(%esi), %mm1 / tmt8 */	528	movq 88(%esi), %mm1 / tmt8 */
523	paddsw %mm1, %mm7 /* V103 */	529	paddsw %mm1, %mm7 /* V103 */
Lines 551-560 _dv_idct_block_mmx: Link Here
551	movq 82(%esi), %mm3 / V123 */	557	movq 82(%esi), %mm3 / V123 */
552	paddsw %mm4, %mm7 /* out0 */	558	paddsw %mm4, %mm7 /* out0 */
553	/* moved up from next block */	559	/* moved up from next block */
554	movq scratch3, %mm0	560	movq MUNG(scratch3), %mm0
555	psraw $4, %mm7	561	psraw $4, %mm7
556	/* moved up from next block */	562	/* moved up from next block */
557	movq scratch5, %mm6	563	movq MUNG(scratch5), %mm6
558	psubsw %mm4, %mm1 /* out14 ; free mm4 */	564	psubsw %mm4, %mm1 /* out14 ; free mm4 */
559	paddsw %mm3, %mm5 /* out2 */	565	paddsw %mm3, %mm5 /* out2 */
560	psraw $4, %mm1	566	psraw $4, %mm1
Lines 565-571 _dv_idct_block_mmx: Link Here
565	movq %mm5, 82(%esi) / out2 ; free mm5 */	571	movq %mm5, 82(%esi) / out2 ; free mm5 */
566	psraw $4, %mm2	572	psraw $4, %mm2
567	/* moved up to the prev block */	573	/* moved up to the prev block */
568	movq scratch7, %mm4	574	movq MUNG(scratch7), %mm4
569	/* moved up to the prev block */	575	/* moved up to the prev block */
570	psraw $4, %mm0	576	psraw $4, %mm0
571	movq %mm2, 812(%esi) / out12 ; free mm2 */	577	movq %mm2, 812(%esi) / out12 ; free mm2 */
Lines 579-585 _dv_idct_block_mmx: Link Here
579	* psraw $4, %mm0	585	* psraw $4, %mm0
580	* psraw $4, %mm6	586	* psraw $4, %mm6
581	*/	587	*/
582	movq scratch1, %mm1	588	movq MUNG(scratch1), %mm1
583	psraw $4, %mm4	589	psraw $4, %mm4
584	movq %mm0, 83(%esi) / out3 */	590	movq %mm0, 83(%esi) / out3 */
585	psraw $4, %mm1	591	psraw $4, %mm1

Lines 17-22 Link Here

(-)libdv-0.104-old/libdv/idct_block_mmx_x86_64.S (+1 lines)
17	.text	17	.text
18	.align 4	18	.align 4
19	.globl _dv_idct_block_mmx_x86_64	19	.globl _dv_idct_block_mmx_x86_64
		20	.hidden _dv_idct_block_mmx_x86_64
20	.type _dv_idct_block_mmx_x86_64,@function	21	.type _dv_idct_block_mmx_x86_64,@function
21	_dv_idct_block_mmx_x86_64:	22	_dv_idct_block_mmx_x86_64:
22	/* void _dv_idct_88(dv_coeff_t block) /	23	/* void _dv_idct_88(dv_coeff_t block) /

Lines 477-482 dv_parse_ac_coeffs(dv_videosegment_t *se Link Here

(-)libdv-0.104-old/libdv/parse.c (+7 lines)
477	exit(0);	477	exit(0);
478	#endif	478	#endif
479	} /* dv_parse_ac_coeffs */	479	} /* dv_parse_ac_coeffs */
		480	#if defined __GNUC__ && __ELF__
		481	# define dv_strong_hidden_alias(name, aliasname) \
		482	extern __typeof (name) aliasname __attribute__ ((alias (#name), visibility ("hidden")))
		483	dv_strong_hidden_alias(dv_parse_ac_coeffs, asm_dv_parse_ac_coeffs);
		484	#else
		485	int asm_dv_parse_ac_coeffs(dv_videosegment_t *seg) { return dv_parse_ac_coeffs(seg); }
		486	#endif
480		487
481	/* ---------------------------------------------------------------------------	488	/* ---------------------------------------------------------------------------
482	*/	489	*/

Lines 144-150 uint8_t dv_quant_offset[4] = { 6,3,0,1 Link Here

(-)libdv-0.104-old/libdv/quant.c (-2 / +2 lines)
144	uint32_t dv_quant_248_mul_tab [2] [22] [64];	144	uint32_t dv_quant_248_mul_tab [2] [22] [64];
145	uint32_t dv_quant_88_mul_tab [2] [22] [64];	145	uint32_t dv_quant_88_mul_tab [2] [22] [64];
146		146
147	extern void _dv_quant_x86(dv_coeff_t *block,int qno,int klass);	147	extern void _dv_quant_x86(dv_coeff_t block,int qno,int klass,uint8_t dv_quant_offset,uint8_t *dv_quant_shifts);
148	extern void _dv_quant_x86_64(dv_coeff_t *block,int qno,int klass);	148	extern void _dv_quant_x86_64(dv_coeff_t *block,int qno,int klass);
149	static void quant_248_inverse_std(dv_coeff_t block,int qno,int klass,dv_248_coeff_t co);	149	static void quant_248_inverse_std(dv_coeff_t block,int qno,int klass,dv_248_coeff_t co);
150	static void quant_248_inverse_mmx(dv_coeff_t block,int qno,int klass,dv_248_coeff_t co);	150	static void quant_248_inverse_mmx(dv_coeff_t block,int qno,int klass,dv_248_coeff_t co);
Lines 210-216 void _dv_quant(dv_coeff_t *block,int qno Link Here
210	_dv_quant_x86_64(block, qno, klass);	210	_dv_quant_x86_64(block, qno, klass);
211	emms();	211	emms();
212	#else	212	#else
213	_dv_quant_x86(block, qno, klass);	213	_dv_quant_x86(block, qno, klass, dv_quant_offset, dv_quant_shifts);
214	emms();	214	emms();
215	#endif	215	#endif
216	}	216	}

Lines 27-33 extern void _dv_quant(dv_coeff_t *block, Link Here

(-)libdv-0.104-old/libdv/quant.h (-1 / +1 lines)
27	extern void _dv_quant_88_inverse(dv_coeff_t *block,int qno,int klass);	27	extern void _dv_quant_88_inverse(dv_coeff_t *block,int qno,int klass);
28	extern void (_dv_quant_248_inverse) (dv_coeff_t block,int qno,int klass,	28	extern void (_dv_quant_248_inverse) (dv_coeff_t block,int qno,int klass,
29	dv_248_coeff_t *co);	29	dv_248_coeff_t *co);
30	extern void _dv_quant_88_inverse_x86(dv_coeff_t *block,int qno,int klass);	30	extern void _dv_quant_88_inverse_x86(dv_coeff_t block,int qno,int klass, uint8_t offset, uint8_t *shifts);
31	extern void _dv_quant_88_inverse_x86_64(dv_coeff_t *block,int qno,int klass);	31	extern void _dv_quant_88_inverse_x86_64(dv_coeff_t *block,int qno,int klass);
32	extern void dv_quant_init (void);	32	extern void dv_quant_init (void);
33	#ifdef __cplusplus	33	#ifdef __cplusplus

Lines 55-60 void _dv_quant_88_inverse(dv_coeff_t *bl Link Here

(-)libdv-0.104-old/libdv/quant_x86.S (-5 / +14 lines)
55	.text	55	.text
56	.align 4	56	.align 4
57	.globl _dv_quant_88_inverse_x86	57	.globl _dv_quant_88_inverse_x86
		58	.hidden _dv_quant_88_inverse_x86
		59	.type _dv_quant_88_inverse_x86,@function
58	_dv_quant_88_inverse_x86:	60	_dv_quant_88_inverse_x86:
59	pushl %ebx	61	pushl %ebx
60	pushl %esi	62	pushl %esi
Lines 71-80 _dv_quant_88_inverse_x86: Link Here
71		73
72	/* pq = dv_quant_shifts[qno + dv_quant_offset[class]]; */	74	/* pq = dv_quant_shifts[qno + dv_quant_offset[class]]; */
73	movl ARGn(1),%eax /* qno */	75	movl ARGn(1),%eax /* qno */
		76	movl ARGn(3),%ebx /* dv_quant_offset */
		77	addl ARGn(2),%ebx /* class */
		78	movzbl (%ebx),%ecx
74	movl ARGn(2),%ebx /* class */	79	movl ARGn(2),%ebx /* class */
75	movzbl dv_quant_offset(%ebx),%ecx
76	addl %ecx,%eax	80	addl %ecx,%eax
77	leal dv_quant_shifts(,%eax,4),%edx /* edx is pq */	81	movl ARGn(4),%edx /* dv_quant_shifts */
		82	leal (%edx,%eax,4),%edx /* edx is pq */
78		83
79	/* extra = (class == 3); */	84	/* extra = (class == 3); */
80	/* 0 1 2 3 */	85	/* 0 1 2 3 */
Lines 193-198 _dv_quant_88_inverse_x86: Link Here
193		198
194	.align 4	199	.align 4
195	.globl _dv_quant_x86	200	.globl _dv_quant_x86
		201	.hidden _dv_quant_x86
		202	.type _dv_quant_x86,@function
196	_dv_quant_x86:	203	_dv_quant_x86:
197	pushl %ebx	204	pushl %ebx
198	pushl %ecx	205	pushl %ecx
Lines 212-222 _dv_quant_x86: Link Here
212		219
213	/* pq = dv_quant_shifts[qno + dv_quant_offset[class]]; */	220	/* pq = dv_quant_shifts[qno + dv_quant_offset[class]]; */
214	movl ARGn(1),%eax /* qno */	221	movl ARGn(1),%eax /* qno */
		222	movl ARGn(3),%ebx /* offset */
		223	addl ARGn(2),%ebx /* class */
		224	movzbl (%ebx),%ecx
215	movl ARGn(2),%ebx /* class */	225	movl ARGn(2),%ebx /* class */
216		226	movl ARGn(4),%edx /* shifts */
217	movzbl dv_quant_offset(%ebx),%ecx
218	addl %ecx,%eax	227	addl %ecx,%eax
219	leal dv_quant_shifts(,%eax,4),%edx /* edx is pq */	228	leal (%edx,%eax,4),%edx /* edx is pq */
220		229
221	/* extra = (class == 3); */	230	/* extra = (class == 3); */
222	/* 0 1 2 3 */	231	/* 0 1 2 3 */

Lines 55-60 void _dv_quant_88_inverse(dv_coeff_t *bl Link Here

(-)libdv-0.104-old/libdv/quant_x86_64.S (+4 lines)
55	.text	55	.text
56	.align 4	56	.align 4
57	.globl _dv_quant_88_inverse_x86_64	57	.globl _dv_quant_88_inverse_x86_64
		58	.hidden _dv_quant_88_inverse_x86_64
		59	.type _dv_quant_88_inverse_x86_64,@function
58	_dv_quant_88_inverse_x86_64:	60	_dv_quant_88_inverse_x86_64:
59		61
60	/* Args are at block=rdi, qno=rsi, class=rdx */	62	/* Args are at block=rdi, qno=rsi, class=rdx */
Lines 195-200 _dv_quant_88_inverse_x86_64: Link Here
195		197
196	.align 4	198	.align 4
197	.globl _dv_quant_x86_64	199	.globl _dv_quant_x86_64
		200	.hidden _dv_quant_x86_64
		201	.type _dv_quant_x86_64,@function
198	_dv_quant_x86_64:	202	_dv_quant_x86_64:
199		203
200	/* Args are at block=rdi, qno=rsi, class=rdx */	204	/* Args are at block=rdi, qno=rsi, class=rdx */




#define DV_WIDTH_SHORT_HALF 720
#define DV_WIDTH_BYTE_HALF  360	
		
.global _dv_rgbtoycb_mmx
# .global yuvtoycb_mmx

.data

.align 8

VBG0B:  .long   0,0
	
#endif	

#include "asm_common.S"

.text

#define _inPtr     24+8
#define _rows      24+12
#define _columns   24+16
#define _outyPtr   24+20
#define _outuPtr   24+24
#define _outvPtr   24+28

.global _dv_rgbtoycb_mmx
.hidden _dv_rgbtoycb_mmx
.type _dv_rgbtoycb_mmx,@function
_dv_rgbtoycb_mmx:

	pushl   %ebp
	movl    %esp, %ebp
	pushl   %eax
	pushl   %ebx
	pushl   %ecx

	pushl   %esi
	pushl   %edi

	LOAD_PIC_REG_BP()

	leal    MUNG(ZEROSX), %eax    #This section gets around a bug
	movq    (%eax), %mm0    #unlikely to persist
	movq    %mm0, MUNG(ZEROS)
	leal    MUNG(OFFSETDX), %eax
	movq    (%eax), %mm0
	movq    %mm0, MUNG(OFFSETD)
	leal    MUNG(OFFSETWX), %eax
	movq    (%eax), %mm0
	movq    %mm0, MUNG(OFFSETW)
	leal    MUNG(OFFSETBX), %eax
	movq    (%eax), %mm0
	movq    %mm0, MUNG(OFFSETB)
	leal    MUNG(YR0GRX), %eax
	movq    (%eax), %mm0
	movq    %mm0, MUNG(YR0GR)
	leal    MUNG(YBG0BX), %eax
	movq    (%eax), %mm0
	movq    %mm0, MUNG(YBG0B)
	leal    MUNG(UR0GRX), %eax
	movq    (%eax), %mm0
	movq    %mm0, MUNG(UR0GR)
	leal    MUNG(UBG0BX), %eax
	movq    (%eax), %mm0
	movq    %mm0, MUNG(UBG0B)
	leal    MUNG(VR0GRX), %eax
	movq    (%eax), %mm0
	movq    %mm0, MUNG(VR0GR)
	leal    MUNG(VBG0BX), %eax
	movq    (%eax), %mm0
	movq    %mm0, MUNG(VBG0B)
	movl    _rows(%esp), %eax
	movl    _columns(%esp), %ebx
	movl    _columns(%ebp), %ebx
	mull    %ebx            #number pixels
	shrl    $3, %eax        #number of loops
	movl    %eax, %edi      #loop counter in edi
	movl    _inPtr(%esp), %eax
	movl    _outyPtr(%esp), %ebx
	movl    _outuPtr(%esp), %ecx
	movl    _outvPtr(%esp), %edx
rgbtoycb_mmx_loop: 
	movq    (%eax), %mm1    #load G2R2B1G1R1B0G0R0
	pxor    %mm6, %mm6      #0 -> mm6

	punpcklbw %mm6, %mm1     #B1G1R1B0 -> mm1
	movq    %mm0, %mm2      #R1B0G0R0 -> mm2

	pmaddwd MUNG(YR0GR), %mm0     #yrR1,ygG0+yrR0 -> mm0
	movq    %mm1, %mm3      #B1G1R1B0 -> mm3

	pmaddwd MUNG(YBG0B), %mm1     #ybB1+ygG1,ybB0 -> mm1
	movq    %mm2, %mm4      #R1B0G0R0 -> mm4

	pmaddwd MUNG(UR0GR), %mm2     #urR1,ugG0+urR0 -> mm2
	movq    %mm3, %mm5      #B1G1R1B0 -> mm5

	pmaddwd MUNG(UBG0B), %mm3     #ubB1+ugG1,ubB0 -> mm3
	punpckhbw       %mm6, %mm7 #    00G2R2 -> mm7

	pmaddwd MUNG(VR0GR), %mm4     #vrR1,vgG0+vrR0 -> mm4
	paddd   %mm1, %mm0      #Y1Y0 -> mm0

	pmaddwd MUNG(VBG0B), %mm5     #vbB1+vgG1,vbB0 -> mm5

	movq    8(%eax), %mm1   #R5B4G4R4B3G3R3B2 -> mm1
	paddd   %mm3, %mm2      #U1U0 -> mm2

	movq    %mm1, %mm6      #R5B4G4R4B3G3R3B2 -> mm6

	punpcklbw       MUNG(ZEROS), %mm1     #B3G3R3B2 -> mm1
	paddd   %mm5, %mm4      #V1V0 -> mm4

	movq    %mm1, %mm5      #B3G3R3B2 -> mm5


	paddd   %mm7, %mm1      #R3B200+00G2R2=R3B2G2R2->mm1

	punpckhbw       MUNG(ZEROS), %mm6     #R5B4G4R3 -> mm6
	movq    %mm1, %mm3      #R3B2G2R2 -> mm3

	pmaddwd MUNG(YR0GR), %mm1     #yrR3,ygG2+yrR2 -> mm1
	movq    %mm5, %mm7      #B3G3R3B2 -> mm7

	pmaddwd MUNG(YBG0B), %mm5     #ybB3+ygG3,ybB2 -> mm5
	psrad   $FIXPSHIFT, %mm0       #32-bit scaled Y1Y0 -> mm0

	movq    %mm6, MUNG(TEMP0)     #R5B4G4R4 -> TEMP0
	movq    %mm3, %mm6      #R3B2G2R2 -> mm6
	pmaddwd MUNG(UR0GR), %mm6     #urR3,ugG2+urR2 -> mm6
	psrad   $FIXPSHIFT, %mm2       #32-bit scaled U1U0 -> mm2

	paddd   %mm5, %mm1      #Y3Y2 -> mm1
	movq    %mm7, %mm5      #B3G3R3B2 -> mm5
	pmaddwd MUNG(UBG0B), %mm7     #ubB3+ugG3,ubB2
	psrad   $FIXPSHIFT, %mm1 #32-bit scaled Y3Y2 -> mm1

	pmaddwd MUNG(VR0GR), %mm3     #vrR3,vgG2+vgR2
	packssdw        %mm1, %mm0      #Y3Y2Y1Y0 -> mm0

	pmaddwd MUNG(VBG0B), %mm5     #vbB3+vgG3,vbB2 -> mm5
	psrad   $FIXPSHIFT, %mm4       #32-bit scaled V1V0 -> mm4

	movq    16(%eax), %mm1  #B7G7R7B6G6R6B5G5 -> mm7

	movq    %mm7, %mm5      #R7B6G6R6B5G500 -> mm5
	psrad   $FIXPSHIFT, %mm3       #32-bit scaled V3V2 -> mm3

	paddw	MUNG(OFFSETY), %mm0
	movq    %mm0, (%ebx)     #store Y3Y2Y1Y0 
	packssdw %mm6, %mm2      #32-bit scaled U3U2U1U0 -> mm2

	movq    MUNG(TEMP0), %mm0     #R5B4G4R4 -> mm0
	addl	$8, %ebx

	punpcklbw       MUNG(ZEROS), %mm7     #B5G500 -> mm7
	movq    %mm0, %mm6      #R5B4G4R4 -> mm6

	movq    %mm2, MUNG(TEMPU)     #32-bit scaled U3U2U1U0 -> TEMPU
	psrlq   $32, %mm0       #00R5B4 -> mm0

	paddw   %mm0, %mm7      #B5G5R5B4 -> mm7
	movq    %mm6, %mm2      #B5B4G4R4 -> mm2

	pmaddwd MUNG(YR0GR), %mm2     #yrR5,ygG4+yrR4 -> mm2
	movq    %mm7, %mm0      #B5G5R5B4 -> mm0

	pmaddwd MUNG(YBG0B), %mm7     #ybB5+ygG5,ybB4 -> mm7
	packssdw        %mm3, %mm4      #32-bit scaled V3V2V1V0 -> mm4

	addl    $24, %eax       #increment RGB count

	movq    %mm4, MUNG(TEMPV)     #(V3V2V1V0)/256 -> mm4
	movq    %mm6, %mm4      #B5B4G4R4 -> mm4

	pmaddwd MUNG(UR0GR), %mm6     #urR5,ugG4+urR4
	movq    %mm0, %mm3      #B5G5R5B4 -> mm0

	pmaddwd MUNG(UBG0B), %mm0     #ubB5+ugG5,ubB4
	paddd   %mm7, %mm2      #Y5Y4 -> mm2

	pmaddwd         MUNG(VR0GR), %mm4     #vrR5,vgG4+vrR4 -> mm4
	pxor    %mm7, %mm7      #0 -> mm7

	pmaddwd MUNG(VBG0B), %mm3     #vbB5+vgG5,vbB4 -> mm3
	punpckhbw       %mm7, %mm1      #B7G7R7B6 -> mm1

	paddd   %mm6, %mm0      #U5U4 -> mm0
	movq    %mm1, %mm6      #B7G7R7B6 -> mm6

	pmaddwd MUNG(YBG0B), %mm6     #ybB7+ygG7,ybB6 -> mm6
	punpckhbw       %mm7, %mm5      #R7B6G6R6 -> mm5

	movq    %mm5, %mm7      #R7B6G6R6 -> mm7
	paddd   %mm4, %mm3      #V5V4 -> mm3

	pmaddwd MUNG(YR0GR), %mm5     #yrR7,ygG6+yrR6 -> mm5
	movq    %mm1, %mm4      #B7G7R7B6 -> mm4

	pmaddwd MUNG(UBG0B), %mm4     #ubB7+ugG7,ubB6 -> mm4
	psrad   $FIXPSHIFT, %mm0       #32-bit scaled U5U4 -> mm0

	psrad   $FIXPSHIFT, %mm2       #32-bit scaled Y5Y4 -> mm2

	paddd   %mm5, %mm6      #Y7Y6 -> mm6
	movq    %mm7, %mm5      #R7B6G6R6 -> mm5

	pmaddwd MUNG(UR0GR), %mm7     #urR7,ugG6+ugR6 -> mm7
	psrad   $FIXPSHIFT, %mm3       #32-bit scaled V5V4 -> mm3

	pmaddwd MUNG(VBG0B), %mm1     #vbB7+vgG7,vbB6 -> mm1
	psrad   $FIXPSHIFT, %mm6 #32-bit scaled Y7Y6 -> mm6

	packssdw %mm6, %mm2     #Y7Y6Y5Y4 -> mm2

	pmaddwd MUNG(VR0GR), %mm5     #vrR7,vgG6+vrR6 -> mm5
	paddd   %mm4, %mm7      #U7U6 -> mm7    

	psrad   $FIXPSHIFT, %mm7       #32-bit scaled U7U6 -> mm7
	paddw	MUNG(OFFSETY), %mm2
	movq	%mm2, (%ebx)    #store Y7Y6Y5Y4 

	movq	MUNG(ALLONE), %mm6
	packssdw %mm7, %mm0     #32-bit scaled U7U6U5U4 -> mm0

	movq    MUNG(TEMPU), %mm4     #32-bit scaled U3U2U1U0 -> mm4
	pmaddwd	%mm6, %mm0      #U7U6U5U4 averaged -> (U7U6)(U5U4)=UU3 UU2->mm0
	
	pmaddwd	%mm6, %mm4      #U3U2U1U0 averaged -> (U3U2)(U1U0)=UU1 UU0->mm4


	psrad   $FIXPSHIFT, %mm1       #32-bit scaled V7V6 -> mm1
	psraw	$1, %mm4 	#divide UU3 UU2 UU1 UU0 by 2 -> mm4

	movq    MUNG(TEMPV), %mm5     #32-bit scaled V3V2V1V0 -> mm5

	movq	%mm4, (%ecx)    # store U	


	ret     

.global _dv_ppm_copy_y_block_mmx
.hidden _dv_ppm_copy_y_block_mmx
.type	_dv_ppm_copy_y_block_mmx,@function
_dv_ppm_copy_y_block_mmx:

	pushl   %ebp

	ret

.global _dv_pgm_copy_y_block_mmx
.hidden _dv_pgm_copy_y_block_mmx
.type	_dv_ppm_copy_y_block_mmx,@function
_dv_pgm_copy_y_block_mmx:

	pushl   %ebp
	movl    %esp, %ebp
	pushl   %esi
	pushl	%edi
	
	movl    8(%ebp), %edi          # dest
	movl    12(%ebp), %esi         # src

	LOAD_PIC_REG_BP()

	movl    16(%esp), %edi          # dest
	movl    20(%esp), %esi         # src

	movq	MUNG(OFFSETY), %mm7
	pxor	%mm6, %mm6
	
	movq	(%esi), %mm0

	ret

.global _dv_video_copy_y_block_mmx
.hidden _dv_video_copy_y_block_mmx
.type	_dv_video_copy_y_block_mmx,@function
_dv_video_copy_y_block_mmx:

	pushl   %ebp
	movl    %esp, %ebp
	pushl   %esi
	pushl	%edi
	
	movl    8(%ebp), %edi          # dest
	movl    12(%ebp), %esi         # src

	LOAD_PIC_REG_BP()

	movl    16(%esp), %edi          # dest
	movl    20(%esp), %esi         # src

	movq	MUNG(OFFSETBX), %mm7
	pxor	%mm6, %mm6
	
	movq	(%esi), %mm0

	
		
.global _dv_ppm_copy_pal_c_block_mmx
.hidden _dv_ppm_copy_pal_c_block_mmx
.type	_dv_ppm_copy_pal_c_block_mmx,@function
_dv_ppm_copy_pal_c_block_mmx:
				
	pushl   %ebp

	ret

.global _dv_pgm_copy_pal_c_block_mmx
.hidden _dv_ppm_copy_pal_c_block_mmx
.type	_dv_pgm_copy_pal_c_block_mmx,@function
_dv_pgm_copy_pal_c_block_mmx:
				
	pushl   %ebp
	movl    %esp, %ebp
	pushl   %esi
	pushl	%edi
	pushl	%ebx
	
	movl    8(%ebp), %edi          # dest
	movl    12(%ebp), %esi         # src

	LOAD_PIC_REG_BP()

	movl    20(%esp), %edi          # dest
	movl    24(%esp), %esi         # src

	movq	MUNG(OFFSETBX), %mm7
	pxor	%mm6, %mm6

	

	ret

.global _dv_video_copy_pal_c_block_mmx
.hidden _dv_video_copy_pal_c_block_mmx
.type	_dv_video_copy_pal_c_block_mmx,@function
_dv_video_copy_pal_c_block_mmx:
				
	pushl   %ebp
	movl    %esp, %ebp
	pushl   %esi
	pushl	%edi
	pushl	%ebx
	
	movl    8(%ebp), %edi          # dest
	movl    12(%ebp), %esi         # src

	LOAD_PIC_REG_BP()

	movl    20(%esp), %edi          # dest
	movl    24(%esp), %esi         # src

	movq	MUNG(OFFSETBX), %mm7
	paddw	%mm7, %mm7
	pxor	%mm6, %mm6


	ret
	
.global _dv_ppm_copy_ntsc_c_block_mmx
.hidden _dv_ppm_copy_ntsc_c_block_mmx
.type	_dv_ppm_copy_ntsc_c_block_mmx,@function
_dv_ppm_copy_ntsc_c_block_mmx:
				
	pushl   %ebp
	movl    %esp, %ebp
	pushl   %esi
	pushl	%edi
	pushl	%ebx

	LOAD_PIC_REG_BP()

	movl    20(%esp), %edi          # dest
	movl    24(%esp), %esi         # src

	movl	$4, %ebx	

	movq	MUNG(ALLONE), %mm6
	
ppm_copy_ntsc_c_block_mmx_loop:	
	
	movq	(%esi), %mm0

	ret

.global _dv_pgm_copy_ntsc_c_block_mmx
.hidden _dv_pgm_copy_ntsc_c_block_mmx
.type	_dv_pgm_copy_ntsc_c_block_mmx,@function
_dv_pgm_copy_ntsc_c_block_mmx:
				
	pushl   %ebp
	movl    %esp, %ebp
	pushl   %esi
	pushl	%edi
	
	movl    8(%ebp), %edi          # dest
	movl    12(%ebp), %esi         # src

	LOAD_PIC_REG_BP()

	movl    16(%esp), %edi          # dest
	movl    20(%esp), %esi         # src

	movq	MUNG(OFFSETBX), %mm7
	paddw	%mm7, %mm7
	pxor	%mm6, %mm6


	ret

.global _dv_video_copy_ntsc_c_block_mmx
.hidden _dv_video_copy_ntsc_c_block_mmx
.type	_dv_video_copy_ntsc_c_block_mmx,@function
_dv_video_copy_ntsc_c_block_mmx:
				
	pushl   %ebp
	movl    %esp, %ebp
	pushl   %esi
	pushl	%edi
	pushl	%ebx
	
	movl    8(%ebp), %edi          # dest
	movl    12(%ebp), %esi         # src

	LOAD_PIC_REG_BP()

	movl    20(%esp), %edi          # dest
	movl    24(%esp), %esi         # src

	movq	MUNG(OFFSETBX), %mm7
	paddw	%mm7, %mm7
	pxor	%mm6, %mm6


Lines 41-49 Link Here

(-)libdv-0.104-old/libdv/rgbtoyuv_x86_64.S (-3 lines)
41	#define DV_WIDTH_SHORT_HALF 720	41	#define DV_WIDTH_SHORT_HALF 720
42	#define DV_WIDTH_BYTE_HALF 360	42	#define DV_WIDTH_BYTE_HALF 360
43		43
44	.global _dv_rgbtoycb_mmx_x86_64
45	# .global yuvtoycb_mmx_x86_64
46
47	.data	44	.data
48		45
49	.align 8	46	.align 8

Lines 1-5 Link Here

(-)libdv-0.104-old/libdv/transpose_x86.S (+2 lines)
1	.text	1	.text
2	.global _dv_transpose_mmx	2	.global _dv_transpose_mmx
		3	.hidden _dv_transpose_mmx
		4	.type _dv_transpose_mmx,@function
3		5
4	_dv_transpose_mmx:	6	_dv_transpose_mmx:
5	pushl %ebp	7	pushl %ebp

Lines 1-5 Link Here

(-)libdv-0.104-old/libdv/transpose_x86_64.S (+2 lines)
1	.text	1	.text
2	.global _dv_transpose_mmx_x86_64	2	.global _dv_transpose_mmx_x86_64
		3	.hidden _dv_transpose_mmx_x86_64
		4	.type _dv_transpose_mmx_x86_64,@function
3		5
4	_dv_transpose_mmx_x86_64:	6	_dv_transpose_mmx_x86_64:
5		7

Lines 1-29 Link Here

(-)libdv-0.104-old/libdv/vlc_x86.S (-26 / +159 lines)
1	#include "asmoff.h"	1	#include "asmoff.h"
		2	#include "asm_common.S"
		3
2	.text	4	.text
3	.align 4	5	.align 4
4	.globl dv_decode_vlc	6	.globl dv_decode_vlc
		7	.globl asm_dv_decode_vlc
		8	.hidden asm_dv_decode_vlc
		9	asm_dv_decode_vlc = dv_decode_vlc
		10
5	.type dv_decode_vlc,@function	11	.type dv_decode_vlc,@function
6	dv_decode_vlc:	12	dv_decode_vlc:
7	pushl %ebx	13	pushl %ebx
		14	pushl %ebp
		15
		16	LOAD_PIC_REG_BP()
8		17
9	/* Args are at 8(%esp). */	18	/* Args are at 12(%esp). */
10	movl 8(%esp),%eax /* %eax is bits */	19	movl 12(%esp),%eax /* %eax is bits */
11	movl 12(%esp),%ebx /* %ebx is maxbits */	20	movl 16(%esp),%ebx /* %ebx is maxbits */
12	andl $0x3f,%ebx /* limit index range STL*/	21	andl $0x3f,%ebx /* limit index range STL*/
13		22
14	movl dv_vlc_class_index_mask(,%ebx,4),%edx	23	movl MUNG_ARR(dv_vlc_class_index_mask,%ebx,4),%edx
15	andl %eax,%edx	24	andl %eax,%edx
16	movl dv_vlc_class_index_rshift(,%ebx,4),%ecx	25	movl MUNG_ARR(dv_vlc_class_index_rshift,%ebx,4),%ecx
17	sarl %cl,%edx	26	sarl %cl,%edx
18	movl dv_vlc_classes(,%ebx,4),%ecx	27	movl MUNG_ARR(dv_vlc_classes,%ebx,4),%ecx
19	movsbl (%ecx,%edx,1),%edx /* %edx is class */	28	movsbl (%ecx,%edx,1),%edx /* %edx is class */
20		29
21	movl dv_vlc_index_mask(,%edx,4),%ebx	30	movl MUNG_ARR(dv_vlc_index_mask,%edx,4),%ebx
22	movl dv_vlc_index_rshift(,%edx,4),%ecx	31	movl MUNG_ARR(dv_vlc_index_rshift,%edx,4),%ecx
23	andl %eax,%ebx	32	andl %eax,%ebx
24	sarl %cl,%ebx	33	sarl %cl,%ebx
25		34
26	movl dv_vlc_lookups(,%edx,4),%edx	35	movl MUNG_ARR(dv_vlc_lookups,%edx,4),%edx
27	movl (%edx,%ebx,4),%edx	36	movl (%edx,%ebx,4),%edx
28		37
29	/* Now %edx holds result, like this:	38	/* Now %edx holds result, like this:
Lines 42-48 dv_decode_vlc: Link Here
42	movl %edx,%ecx	51	movl %edx,%ecx
43	sarl $8,%ecx	52	sarl $8,%ecx
44	andl $0xff,%ecx	53	andl $0xff,%ecx
45	movl sign_mask(,%ecx,4),%ebx	54	movl MUNG_ARR(sign_mask,%ecx,4),%ebx
46	andl %ebx,%eax	55	andl %ebx,%eax
47	negl %eax	56	negl %eax
48	sarl $31,%eax	57	sarl $31,%eax
Lines 63-76 dv_decode_vlc: Link Here
63	*result = broken;	72	*result = broken;
64	Note that the 'broken' pattern is all ones (i.e. 0xffffffff)	73	Note that the 'broken' pattern is all ones (i.e. 0xffffffff)
65	*/	74	*/
66	movl 12(%esp),%ebx /* %ebx is maxbits */	75	movl 16(%esp),%ebx /* %ebx is maxbits */
67	subl %ecx,%ebx	76	subl %ecx,%ebx
68	sbbl %ebx,%ebx	77	sbbl %ebx,%ebx
69	orl %ebx,%edx	78	orl %ebx,%edx
70		79
71	movl 16(%esp),%eax	80	movl 20(%esp),%eax
72	movl %edx,(%eax)	81	movl %edx,(%eax)
73		82	popl %ebp
74	popl %ebx	83	popl %ebx
75	ret	84	ret
76		85
Lines 80-100 dv_decode_vlc: Link Here
80	.type __dv_decode_vlc,@function	89	.type __dv_decode_vlc,@function
81	__dv_decode_vlc:	90	__dv_decode_vlc:
82	pushl %ebx	91	pushl %ebx
		92	pushl %ebp
		93
		94	LOAD_PIC_REG_BP()
83		95
84	/* Args are at 8(%esp). */	96	/* Args are at 12(%esp). */
85	movl 8(%esp),%eax /* %eax is bits */	97	movl 12(%esp),%eax /* %eax is bits */
86		98
87	movl %eax,%edx /* %edx is class */	99	movl %eax,%edx /* %edx is class */
88	andl $0xfe00,%edx	100	andl $0xfe00,%edx
89	sarl $9,%edx	101	sarl $9,%edx
		102	#ifdef __PIC__
		103	movsbl dv_vlc_class_lookup5@GOTOFF(%ebp,%edx),%edx
		104	#else
90	movsbl dv_vlc_class_lookup5(%edx),%edx	105	movsbl dv_vlc_class_lookup5(%edx),%edx
91		106	#endif
92	movl dv_vlc_index_mask(,%edx,4),%ebx	107
93	movl dv_vlc_index_rshift(,%edx,4),%ecx	108	movl MUNG_ARR(dv_vlc_index_mask,%edx,4),%ebx
		109	movl MUNG_ARR(dv_vlc_index_rshift,%edx,4),%ecx
94	andl %eax,%ebx	110	andl %eax,%ebx
95	sarl %cl,%ebx	111	sarl %cl,%ebx
96		112
97	movl dv_vlc_lookups(,%edx,4),%edx	113	movl MUNG_ARR(dv_vlc_lookups,%edx,4),%edx
98	movl (%edx,%ebx,4),%edx	114	movl (%edx,%ebx,4),%edx
99		115
100	/* Now %edx holds result, like this:	116	/* Now %edx holds result, like this:
Lines 112-118 __dv_decode_vlc: Link Here
112	movl %edx,%ecx	128	movl %edx,%ecx
113	sarl $8,%ecx	129	sarl $8,%ecx
114	andl $0xff,%ecx	130	andl $0xff,%ecx
115	movl sign_mask(,%ecx,4),%ecx	131	movl MUNG_ARR(sign_mask,%ecx,4),%ecx
116	andl %ecx,%eax	132	andl %ecx,%eax
117	negl %eax	133	negl %eax
118	sarl $31,%eax	134	sarl $31,%eax
Lines 127-135 __dv_decode_vlc: Link Here
127	xorl %eax,%edx	143	xorl %eax,%edx
128	subl %eax,%edx	144	subl %eax,%edx
129		145
130	movl 12(%esp),%eax	146	movl 16(%esp),%eax
131	movl %edx,(%eax)	147	movl %edx,(%eax)
132		148	popl %ebp
133	popl %ebx	149	popl %ebx
134	ret	150	ret
135		151
Lines 140-152 void dv_parse_ac_coeffs_pass0(bitstream_ Link Here
140	*/	156	*/
141	.text	157	.text
142	.align 4	158	.align 4
		159	.globl asm_dv_parse_ac_coeffs_pass0
		160	.hidden asm_dv_parse_ac_coeffs_pass0
		161	asm_dv_parse_ac_coeffs_pass0 = dv_parse_ac_coeffs_pass0
		162
143	.globl dv_parse_ac_coeffs_pass0	163	.globl dv_parse_ac_coeffs_pass0
		164	.type dv_parse_ac_coeffs_pass0,@function
144	dv_parse_ac_coeffs_pass0:	165	dv_parse_ac_coeffs_pass0:
145	pushl %ebx	166	pushl %ebx
146	pushl %edi	167	pushl %edi
147	pushl %esi	168	pushl %esi
148	pushl %ebp	169	pushl %ebp
149		170
		171	LOAD_PIC_REG_SI()
		172
150	#define ARGn(N) (20+(4*(N)))(%esp)	173	#define ARGn(N) (20+(4*(N)))(%esp)
151		174
152	/*	175	/*
Lines 159-166 dv_parse_ac_coeffs_pass0: Link Here
159	ebp bl	182	ebp bl
160	*/	183	*/
161	movl ARGn(2),%ebp	184	movl ARGn(2),%ebp
		185	#ifndef __PIC__
162	movl ARGn(0),%esi	186	movl ARGn(0),%esi
163	movl bitstream_t_buf(%esi),%esi	187	movl bitstream_t_buf(%esi),%esi
		188	#endif
164	movl dv_block_t_offset(%ebp),%edi	189	movl dv_block_t_offset(%ebp),%edi
165	movl dv_block_t_reorder(%ebp),%ebx	190	movl dv_block_t_reorder(%ebp),%ebx
166		191
Lines 170-176 dv_parse_ac_coeffs_pass0: Link Here
170		195
171	movq dv_block_t_coeffs(%ebp),%mm1	196	movq dv_block_t_coeffs(%ebp),%mm1
172	pxor %mm0,%mm0	197	pxor %mm0,%mm0
		198	#ifdef __PIC__
		199	pand const_f_0_0_0@GOTOFF(%esi),%mm1
		200	#else
173	pand const_f_0_0_0,%mm1	201	pand const_f_0_0_0,%mm1
		202	#endif
174	movq %mm1,dv_block_t_coeffs(%ebp)	203	movq %mm1,dv_block_t_coeffs(%ebp)
175	movq %mm0,(dv_block_t_coeffs + 8)(%ebp)	204	movq %mm0,(dv_block_t_coeffs + 8)(%ebp)
176	movq %mm0,(dv_block_t_coeffs + 16)(%ebp)	205	movq %mm0,(dv_block_t_coeffs + 16)(%ebp)
Lines 191-199 dv_parse_ac_coeffs_pass0: Link Here
191	readloop:	220	readloop:
192	movl %edi,%ecx	221	movl %edi,%ecx
193	shrl $3,%ecx	222	shrl $3,%ecx
		223	#ifdef __PIC__
		224	movl ARGn(0),%eax
		225	addl bitstream_t_buf(%eax),%ecx
		226	movzbl (%ecx),%eax
		227	movzbl 1(%ecx),%edx
		228	movzbl 2(%ecx),%ecx
		229	#else
194	movzbl (%esi,%ecx,1),%eax	230	movzbl (%esi,%ecx,1),%eax
195	movzbl 1(%esi,%ecx,1),%edx	231	movzbl 1(%esi,%ecx,1),%edx
196	movzbl 2(%esi,%ecx,1),%ecx	232	movzbl 2(%esi,%ecx,1),%ecx
		233	#endif
197	shll $16,%eax	234	shll $16,%eax
198	shll $8,%edx	235	shll $8,%edx
199	orl %ecx,%eax	236	orl %ecx,%eax
Lines 217-223 readloop: Link Here
217		254
218	/* Attempt to use the shortcut first. If it hits, then	255	/* Attempt to use the shortcut first. If it hits, then
219	this vlc term has been decoded. */	256	this vlc term has been decoded. */
		257	#ifdef __PIC__
		258	movl dv_vlc_class1_shortcut@GOTOFF(%esi,%ecx,4),%edx
		259	#else
220	movl dv_vlc_class1_shortcut(,%ecx,4),%edx	260	movl dv_vlc_class1_shortcut(,%ecx,4),%edx
		261	#endif
221	test $0x80,%edx	262	test $0x80,%edx
222	je done_decode	263	je done_decode
223		264
Lines 228-239 readloop: Link Here
228	movl %ebx,dv_block_t_reorder(%ebp)	269	movl %ebx,dv_block_t_reorder(%ebp)
229		270
230	/* %eax is bits */	271	/* %eax is bits */
231		272	#ifdef __PIC__
		273	movsbl dv_vlc_class_lookup5@GOTOFF(%esi,%ecx),%ecx
		274
		275	movl dv_vlc_index_mask@GOTOFF(%esi,%ecx,4),%ebx
		276	movl dv_vlc_lookups@GOTOFF(%esi,%ecx,4),%edx
		277	movl dv_vlc_index_rshift@GOTOFF(%esi,%ecx,4),%ecx
		278	#else
232	movsbl dv_vlc_class_lookup5(%ecx),%ecx	279	movsbl dv_vlc_class_lookup5(%ecx),%ecx
233		280
234	movl dv_vlc_index_mask(,%ecx,4),%ebx	281	movl dv_vlc_index_mask(,%ecx,4),%ebx
235	movl dv_vlc_lookups(,%ecx,4),%edx	282	movl dv_vlc_lookups(,%ecx,4),%edx
236	movl dv_vlc_index_rshift(,%ecx,4),%ecx	283	movl dv_vlc_index_rshift(,%ecx,4),%ecx
		284	#endif
237	andl %eax,%ebx	285	andl %eax,%ebx
238	sarl %cl,%ebx	286	sarl %cl,%ebx
239		287
Lines 256-262 readloop: Link Here
256	movl %edx,%ecx	304	movl %edx,%ecx
257	sarl $8,%ecx	305	sarl $8,%ecx
258	andl $0xff,%ecx	306	andl $0xff,%ecx
		307	#ifdef __PIC__
		308	movl sign_mask@GOTOFF(%esi,%ecx,4),%ecx
		309	#else
259	movl sign_mask(,%ecx,4),%ecx	310	movl sign_mask(,%ecx,4),%ecx
		311	#endif
260	andl %ecx,%eax	312	andl %ecx,%eax
261	negl %eax	313	negl %eax
262	sarl $31,%eax	314	sarl $31,%eax
Lines 326-335 alldone: Link Here
326		378
327	slowpath:	379	slowpath:
328	/* slow path: use dv_decode_vlc */;	380	/* slow path: use dv_decode_vlc */;
		381	#ifdef __PIC__
		382	pushl %esi
		383	leal vlc@GOTOFF(%esi),%esi
		384	xchgl %esi,(%esp) /* last parameter is &vlc */
		385	#else
329	pushl $vlc /* last parameter is &vlc */	386	pushl $vlc /* last parameter is &vlc */
		387	#endif
330	pushl %edx /* bits_left */	388	pushl %edx /* bits_left */
331	pushl %eax /* bits */	389	pushl %eax /* bits */
332	call dv_decode_vlc	390	call asm_dv_decode_vlc
333	addl $12,%esp	391	addl $12,%esp
334	test $0x80,%edx /* If (vlc.run < 0) break */	392	test $0x80,%edx /* If (vlc.run < 0) break */
335	jne escape	393	jne escape
Lines 359-370 show16: Link Here
359	gint dv_parse_video_segment(dv_videosegment_t *seg, guint quality) {	417	gint dv_parse_video_segment(dv_videosegment_t *seg, guint quality) {
360	*/	418	*/
361	.globl dv_parse_video_segment	419	.globl dv_parse_video_segment
		420	.type dv_parse_video_segment,@function
362	dv_parse_video_segment:	421	dv_parse_video_segment:
363	pushl %ebx	422	pushl %ebx
364	pushl %edi	423	pushl %edi
365	pushl %esi	424	pushl %esi
366	pushl %ebp	425	pushl %ebp
367		426
		427	LOAD_PIC_REG_SI()
		428
368	#define ARGn(N) (20+(4*(N)))(%esp)	429	#define ARGn(N) (20+(4*(N)))(%esp)
369		430
370	movl ARGn(1),%eax /* quality */	431	movl ARGn(1),%eax /* quality */
Lines 373-379 dv_parse_video_segment: Link Here
373	jz its_mono	434	jz its_mono
374	movl $6,%ebx	435	movl $6,%ebx
375	its_mono:	436	its_mono:
		437	#ifdef __PIC__
		438	movl %ebx,n_blocks@GOTOFF(%esi)
		439	#else
376	movl %ebx,n_blocks	440	movl %ebx,n_blocks
		441	#endif
377		442
378	/*	443	/*
379	* ebx seg/b	444	* ebx seg/b
Lines 384-398 its_mono: Link Here
384	* ebp bl	449	* ebp bl
385	*/	450	*/
386	movl ARGn(0),%ebx	451	movl ARGn(0),%ebx
		452	#ifndef __PIC__
387	movl dv_videosegment_t_bs(%ebx),%esi	453	movl dv_videosegment_t_bs(%ebx),%esi
388	movl bitstream_t_buf(%esi),%esi	454	movl bitstream_t_buf(%esi),%esi
		455	#endif
389	leal dv_videosegment_t_mb(%ebx),%edi	456	leal dv_videosegment_t_mb(%ebx),%edi
390		457
391	movl $0,%eax	458	movl $0,%eax
392	movl $0,%ecx	459	movl $0,%ecx
393	macloop:	460	macloop:
		461	#ifdef __PIC__
		462	movl %eax,m@GOTOFF(%esi)
		463	movl %ecx,mb_start@GOTOFF(%esi)
		464	#else
394	movl %eax,m	465	movl %eax,m
395	movl %ecx,mb_start	466	movl %ecx,mb_start
		467	#endif
396		468
397	movl ARGn(0),%ebx	469	movl ARGn(0),%ebx
398		470
Lines 400-406 macloop: Link Here
400	/* mb->qno = bitstream_get(bs,4); */	472	/* mb->qno = bitstream_get(bs,4); */
401	movl %ecx,%edx	473	movl %ecx,%edx
402	shr $3,%edx	474	shr $3,%edx
		475	#ifdef __PIC__
		476	movl dv_videosegment_t_bs(%ebx),%ecx
		477	movl bitstream_t_buf(%ecx),%ecx
		478	movzbl 3(%ecx,%edx,1),%edx
		479	#else
403	movzbl 3(%esi,%edx,1),%edx	480	movzbl 3(%esi,%edx,1),%edx
		481	#endif
404	andl $0xf,%edx	482	andl $0xf,%edx
405	movl %edx,dv_macroblock_t_qno(%edi)	483	movl %edx,dv_macroblock_t_qno(%edi)
406		484
Lines 411-417 macloop: Link Here
411	movl %edx,dv_macroblock_t_eob_count(%edi)	489	movl %edx,dv_macroblock_t_eob_count(%edi)
412		490
413	/* mb->i = (seg->i + dv_super_map_vertical[m]) % (seg->isPAL?12:10); */	491	/* mb->i = (seg->i + dv_super_map_vertical[m]) % (seg->isPAL?12:10); */
		492	#ifdef __PIC__
		493	movl dv_super_map_vertical@GOTOFF(%esi,%eax,4),%edx
		494	#else
414	movl dv_super_map_vertical(,%eax,4),%edx	495	movl dv_super_map_vertical(,%eax,4),%edx
		496	#endif
415	movl dv_videosegment_t_i(%ebx),%ecx	497	movl dv_videosegment_t_i(%ebx),%ecx
416	addl %ecx,%edx	498	addl %ecx,%edx
417		499
Lines 422-432 skarly: Link Here
422	andl $1,%ecx	504	andl $1,%ecx
423	shll $5,%ecx /* ecx = (isPAL ? 32 : 0) */	505	shll $5,%ecx /* ecx = (isPAL ? 32 : 0) */
424		506
		507	#ifdef __PIC__
		508	leal mod_10@GOTOFF(%esi),%edx
		509	movzbl (%edx,%ecx,1),%edx /* uses mod_12 for PAL */
		510	#else
425	movzbl mod_10(%edx,%ecx,1),%edx /* uses mod_12 for PAL */	511	movzbl mod_10(%edx,%ecx,1),%edx /* uses mod_12 for PAL */
		512	#endif
426	movl %edx,dv_macroblock_t_i(%edi)	513	movl %edx,dv_macroblock_t_i(%edi)
427		514
428	/* mb->j = dv_super_map_horizontal[m]; */	515	/* mb->j = dv_super_map_horizontal[m]; */
		516	#ifdef __PIC__
		517	movl dv_super_map_horizontal@GOTOFF(%esi,%eax,4),%edx
		518	#else
429	movl dv_super_map_horizontal(,%eax,4),%edx	519	movl dv_super_map_horizontal(,%eax,4),%edx
		520	#endif
430	movl %edx,dv_macroblock_t_j(%edi)	521	movl %edx,dv_macroblock_t_j(%edi)
431		522
432	/* mb->k = seg->k; */	523	/* mb->k = seg->k; */
Lines 445-456 blkloop: Link Here
445	+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+	536	+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+---+
446	*/	537	*/
447	/* dc = bitstream_get(bs,9); */	538	/* dc = bitstream_get(bs,9); */
		539	#ifdef __PIC__
		540	movl mb_start@GOTOFF(%esi),%ecx
		541	#else
448	movl mb_start,%ecx	542	movl mb_start,%ecx
		543	#endif
449	shr $3,%ecx	544	shr $3,%ecx
		545	#ifdef __PIC__
		546	movzbl blk_start@GOTOFF(%esi,%ebx),%edx
		547	#else
450	movzbl blk_start(%ebx),%edx	548	movzbl blk_start(%ebx),%edx
		549	#endif
451	addl %ecx,%edx	550	addl %ecx,%edx
		551	#ifdef __PIC__
		552	movl ARGn(0),%ecx
		553	movl dv_videosegment_t_bs(%ecx),%ecx
		554	movl bitstream_t_buf(%ecx),%ecx
		555	movzbl (%ecx,%edx,1),%eax /* hi byte */
		556	movzbl 1(%ecx,%edx,1),%ecx /* lo byte */
		557	#else
452	movzbl (%esi,%edx,1),%eax /* hi byte */	558	movzbl (%esi,%edx,1),%eax /* hi byte */
453	movzbl 1(%esi,%edx,1),%ecx /* lo byte */	559	movzbl 1(%esi,%edx,1),%ecx /* lo byte */
		560	#endif
454	shll $8,%eax	561	shll $8,%eax
455	orl %ecx,%eax	562	orl %ecx,%eax
456		563
Lines 477-483 blkloop: Link Here
477		584
478	/* bl->reorder = &dv_reorder[bl->dct_mode][1]; */	585	/* bl->reorder = &dv_reorder[bl->dct_mode][1]; */
479	shll $6,%eax	586	shll $6,%eax
		587	#ifdef __PIC__
		588	leal dv_reorder@GOTOFF+1(%esi),%eax
		589	#else
480	addl $(dv_reorder+1),%eax	590	addl $(dv_reorder+1),%eax
		591	#endif
481	movl %eax,dv_block_t_reorder(%ebp)	592	movl %eax,dv_block_t_reorder(%ebp)
482		593
483	/* bl->reorder_sentinel = bl->reorder + 63; */	594	/* bl->reorder_sentinel = bl->reorder + 63; */
Lines 485-497 blkloop: Link Here
485	movl %eax,dv_block_t_reorder_sentinel(%ebp)	596	movl %eax,dv_block_t_reorder_sentinel(%ebp)
486		597
487	/* bl->offset= mb_start + dv_parse_bit_start[b]; */	598	/* bl->offset= mb_start + dv_parse_bit_start[b]; */
		599	#ifdef __PIC__
		600	movl mb_start@GOTOFF(%esi),%ecx
		601	movl dv_parse_bit_start@GOTOFF(%esi,%ebx,4),%eax
		602	#else
488	movl mb_start,%ecx	603	movl mb_start,%ecx
489	movl dv_parse_bit_start(,%ebx,4),%eax	604	movl dv_parse_bit_start(,%ebx,4),%eax
		605	#endif
490	addl %ecx,%eax	606	addl %ecx,%eax
491	movl %eax,dv_block_t_offset(%ebp)	607	movl %eax,dv_block_t_offset(%ebp)
492		608
493	/* bl->end= mb_start + dv_parse_bit_end[b]; */	609	/* bl->end= mb_start + dv_parse_bit_end[b]; */
		610	#ifdef __PIC__
		611	movl dv_parse_bit_end@GOTOFF(%esi,%ebx,4),%eax
		612	#else
494	movl dv_parse_bit_end(,%ebx,4),%eax	613	movl dv_parse_bit_end(,%ebx,4),%eax
		614	#endif
495	addl %ecx,%eax	615	addl %ecx,%eax
496	movl %eax,dv_block_t_end(%ebp)	616	movl %eax,dv_block_t_end(%ebp)
497		617
Lines 503-509 blkloop: Link Here
503	/* no AC pass. Just zero out the remaining coeffs */	623	/* no AC pass. Just zero out the remaining coeffs */
504	movq dv_block_t_coeffs(%ebp),%mm1	624	movq dv_block_t_coeffs(%ebp),%mm1
505	pxor %mm0,%mm0	625	pxor %mm0,%mm0
		626	#ifdef __PIC__
		627	pand const_f_0_0_0@GOTOFF(%esi),%mm1
		628	#else
506	pand const_f_0_0_0,%mm1	629	pand const_f_0_0_0,%mm1
		630	#endif
507	movq %mm1,dv_block_t_coeffs(%ebp)	631	movq %mm1,dv_block_t_coeffs(%ebp)
508	movq %mm0,(dv_block_t_coeffs + 8)(%ebp)	632	movq %mm0,(dv_block_t_coeffs + 8)(%ebp)
509	movq %mm0,(dv_block_t_coeffs + 16)(%ebp)	633	movq %mm0,(dv_block_t_coeffs + 16)(%ebp)
Lines 528-545 do_ac_pass: Link Here
528	pushl %ebp	652	pushl %ebp
529	pushl %edi	653	pushl %edi
530	pushl %eax	654	pushl %eax
531	call dv_parse_ac_coeffs_pass0	655	call asm_dv_parse_ac_coeffs_pass0
532	addl $12,%esp	656	addl $12,%esp
533	done_ac:	657	done_ac:
534		658
		659	#ifdef __PIC__
		660	movl n_blocks@GOTOFF(%esi),%eax
		661	#else
535	movl n_blocks,%eax	662	movl n_blocks,%eax
		663	#endif
536	addl $dv_block_t_size,%ebp	664	addl $dv_block_t_size,%ebp
537	incl %ebx	665	incl %ebx
538	cmpl %eax,%ebx	666	cmpl %eax,%ebx
539	jnz blkloop	667	jnz blkloop
540		668
		669	#ifdef __PIC__
		670	movl m@GOTOFF(%esi),%eax
		671	movl mb_start@GOTOFF(%esi),%ecx
		672	#else
541	movl m,%eax	673	movl m,%eax
542	movl mb_start,%ecx	674	movl mb_start,%ecx
		675	#endif
543	addl $(8 * 80),%ecx	676	addl $(8 * 80),%ecx
544	addl $dv_macroblock_t_size,%edi	677	addl $dv_macroblock_t_size,%edi
545	incl %eax	678	incl %eax
Lines 557-563 done_ac: Link Here
557		690
558	andl $DV_QUALITY_AC_MASK,%eax	691	andl $DV_QUALITY_AC_MASK,%eax
559	cmpl $DV_QUALITY_AC_2,%eax	692	cmpl $DV_QUALITY_AC_2,%eax
560	jz dv_parse_ac_coeffs	693	jz asm_dv_parse_ac_coeffs
561	movl $0,%eax	694	movl $0,%eax
562	ret	695	ret
563		696

Lines 169-175 void dv_parse_ac_coeffs_pass0(bitstream_ Link Here

(-)libdv-0.104-old/libdv/vlc_x86_64.S (-1 / +3 lines)
169	.text	169	.text
170	.align 4	170	.align 4
171	.globl dv_parse_ac_coeffs_pass0	171	.globl dv_parse_ac_coeffs_pass0
172		172	.type dv_parse_ac_coeffs_pass0,@function
		173
173	dv_parse_ac_coeffs_pass0:	174	dv_parse_ac_coeffs_pass0:
174		175
175	/* Args are at rdi=bs, rsi=mb, rdx=bl */	176	/* Args are at rdi=bs, rsi=mb, rdx=bl */
Lines 422-427 show16: /* not u Link Here
422	gint dv_parse_video_segment(dv_videosegment_t *seg, guint quality) {	423	gint dv_parse_video_segment(dv_videosegment_t *seg, guint quality) {
423	*/	424	*/
424	.globl dv_parse_video_segment	425	.globl dv_parse_video_segment
		426	.type dv_parse_video_segment,@function
425	dv_parse_video_segment:	427	dv_parse_video_segment:
426		428
427	/* Args are at rdi=seg, rsi=quality */	429	/* Args are at rdi=seg, rsi=quality */

Return to bug 121871

Line 0 Link Here

(-)libdv-0.104-old/libdv/asm_common.S (+37 lines)
	1	/* public domain, do what you want */
	2
	3	#ifdef __PIC__
	4	# define MUNG(sym) sym##@GOTOFF(%ebp)
	5	# define MUNG_ARR(sym, args...) sym##@GOTOFF(%ebp,##args)
	6	#else
	7	# define MUNG(sym) sym
	8	# define MUNG_ARR(sym, args...) sym(,##args)
	9	#endif
	10
	11	#ifdef __PIC__
	12	# undef __i686 /* gcc define gets in our way */
	13	.section .gnu.linkonce.t.__i686.get_pc_thunk.bp,"ax",@progbits
	14	.globl __i686.get_pc_thunk.bp
	15	.hidden __i686.get_pc_thunk.bp
	16	.type __i686.get_pc_thunk.bp,@function
	17	__i686.get_pc_thunk.bp:
	18	movl (%esp), %ebp
	19	ret
	20	# define LOAD_PIC_REG_BP() \
	21	call __i686.get_pc_thunk.bp ; \
	22	addl $_GLOBAL_OFFSET_TABLE_, %ebp
	23
	24	.section .gnu.linkonce.t.__i686.get_pc_thunk.si,"ax",@progbits
	25	.globl __i686.get_pc_thunk.si
	26	.hidden __i686.get_pc_thunk.si
	27	.type __i686.get_pc_thunk.si,@function
	28	__i686.get_pc_thunk.si:
	29	movl (%esp), %esi
	30	ret
	31	# define LOAD_PIC_REG_SI() \
	32	call __i686.get_pc_thunk.si ; \
	33	addl $_GLOBAL_OFFSET_TABLE_, %esi
	34	#else
	35	# define LOAD_PIC_REG_BP()
	36	# define LOAD_PIC_REG_SI()
	37	#endif

Lines 521-527 static void reorder_block(dv_block_t *bl Link Here

(-)libdv-0.104-old/libdv/encode.c (-4 / +5 lines)
521	}	521	}
522		522
523	extern unsigned long _dv_vlc_encode_block_mmx(dv_coeff_t* coeffs,	523	extern unsigned long _dv_vlc_encode_block_mmx(dv_coeff_t* coeffs,
524	dv_vlc_entry_t ** out);	524	dv_vlc_entry_t ** out,
		525	dv_vlc_entry_t * lookup);
525		526
526	extern unsigned long _dv_vlc_encode_block_mmx_x86_64(dv_coeff_t* coeffs,	527	extern unsigned long _dv_vlc_encode_block_mmx_x86_64(dv_coeff_t* coeffs,
527	dv_vlc_entry_t ** out);	528	dv_vlc_entry_t ** out);
Lines 558-564 static unsigned long vlc_encode_block(dv Link Here
558	#elif ARCH_X86	559	#elif ARCH_X86
559	int num_bits;	560	int num_bits;
560		561
561	num_bits = _dv_vlc_encode_block_mmx(coeffs, &o);	562	num_bits = _dv_vlc_encode_block_mmx(coeffs, &o, vlc_encode_lookup);
562	emms();	563	emms();
563	#else	564	#else
564	int num_bits;	565	int num_bits;
Lines 574-580 static unsigned long vlc_encode_block(dv Link Here
574	return num_bits;	575	return num_bits;
575	}	576	}
576		577
577	extern unsigned long _dv_vlc_num_bits_block_x86(dv_coeff_t* coeffs);	578	extern unsigned long _dv_vlc_num_bits_block_x86(dv_coeff_t* coeffs, unsigned char* lookup);
578	extern unsigned long _dv_vlc_num_bits_block_x86_64(dv_coeff_t* coeffs);	579	extern unsigned long _dv_vlc_num_bits_block_x86_64(dv_coeff_t* coeffs);
579		580
580	extern unsigned long _dv_vlc_num_bits_block(dv_coeff_t* coeffs)	581	extern unsigned long _dv_vlc_num_bits_block(dv_coeff_t* coeffs)
Lines 600-606 extern unsigned long _dv_vlc_num_bits_bl Link Here
600	#elif ARCH_X86_64	601	#elif ARCH_X86_64
601	return _dv_vlc_num_bits_block_x86_64(coeffs);	602	return _dv_vlc_num_bits_block_x86_64(coeffs);
602	#else	603	#else
603	return _dv_vlc_num_bits_block_x86(coeffs);	604	return _dv_vlc_num_bits_block_x86(coeffs, vlc_num_bits_lookup);
604	#endif	605	#endif
605	}	606	}
606		607

Lines 41-49 Link Here

(-)libdv-0.104-old/libdv/rgbtoyuv.S (-112 / +137 lines)
41	#define DV_WIDTH_SHORT_HALF 720	41	#define DV_WIDTH_SHORT_HALF 720
42	#define DV_WIDTH_BYTE_HALF 360	42	#define DV_WIDTH_BYTE_HALF 360
43		43
44	.global _dv_rgbtoycb_mmx
45	# .global yuvtoycb_mmx
46
47	.data	44	.data
48		45
49	.align 8	46	.align 8
Lines 110-129 VR0GR: .long 0,0 Link Here
110	VBG0B: .long 0,0	107	VBG0B: .long 0,0
111		108
112	#endif	109	#endif
113		110
		111	#include "asm_common.S"
		112
114	.text	113	.text
115		114
116	#define _inPtr 8	115	#define _inPtr 24+8
117	#define _rows 12	116	#define _rows 24+12
118	#define _columns 16	117	#define _columns 24+16
119	#define _outyPtr 20	118	#define _outyPtr 24+20
120	#define _outuPtr 24	119	#define _outuPtr 24+24
121	#define _outvPtr 28	120	#define _outvPtr 24+28
122		121
		122	.global _dv_rgbtoycb_mmx
		123	.hidden _dv_rgbtoycb_mmx
		124	.type _dv_rgbtoycb_mmx,@function
123	_dv_rgbtoycb_mmx:	125	_dv_rgbtoycb_mmx:
124		126
125	pushl %ebp	127	pushl %ebp
126	movl %esp, %ebp
127	pushl %eax	128	pushl %eax
128	pushl %ebx	129	pushl %ebx
129	pushl %ecx	130	pushl %ecx
Lines 131-176 _dv_rgbtoycb_mmx: Link Here
131	pushl %esi	132	pushl %esi
132	pushl %edi	133	pushl %edi
133		134
134	leal ZEROSX, %eax #This section gets around a bug	135	LOAD_PIC_REG_BP()
		136
		137	leal MUNG(ZEROSX), %eax #This section gets around a bug
135	movq (%eax), %mm0 #unlikely to persist	138	movq (%eax), %mm0 #unlikely to persist
136	movq %mm0, ZEROS	139	movq %mm0, MUNG(ZEROS)
137	leal OFFSETDX, %eax	140	leal MUNG(OFFSETDX), %eax
138	movq (%eax), %mm0	141	movq (%eax), %mm0
139	movq %mm0, OFFSETD	142	movq %mm0, MUNG(OFFSETD)
140	leal OFFSETWX, %eax	143	leal MUNG(OFFSETWX), %eax
141	movq (%eax), %mm0	144	movq (%eax), %mm0
142	movq %mm0, OFFSETW	145	movq %mm0, MUNG(OFFSETW)
143	leal OFFSETBX, %eax	146	leal MUNG(OFFSETBX), %eax
144	movq (%eax), %mm0	147	movq (%eax), %mm0
145	movq %mm0, OFFSETB	148	movq %mm0, MUNG(OFFSETB)
146	leal YR0GRX, %eax	149	leal MUNG(YR0GRX), %eax
147	movq (%eax), %mm0	150	movq (%eax), %mm0
148	movq %mm0, YR0GR	151	movq %mm0, MUNG(YR0GR)
149	leal YBG0BX, %eax	152	leal MUNG(YBG0BX), %eax
150	movq (%eax), %mm0	153	movq (%eax), %mm0
151	movq %mm0, YBG0B	154	movq %mm0, MUNG(YBG0B)
152	leal UR0GRX, %eax	155	leal MUNG(UR0GRX), %eax
153	movq (%eax), %mm0	156	movq (%eax), %mm0
154	movq %mm0, UR0GR	157	movq %mm0, MUNG(UR0GR)
155	leal UBG0BX, %eax	158	leal MUNG(UBG0BX), %eax
156	movq (%eax), %mm0	159	movq (%eax), %mm0
157	movq %mm0, UBG0B	160	movq %mm0, MUNG(UBG0B)
158	leal VR0GRX, %eax	161	leal MUNG(VR0GRX), %eax
159	movq (%eax), %mm0	162	movq (%eax), %mm0
160	movq %mm0, VR0GR	163	movq %mm0, MUNG(VR0GR)
161	leal VBG0BX, %eax	164	leal MUNG(VBG0BX), %eax
162	movq (%eax), %mm0	165	movq (%eax), %mm0
163	movq %mm0, VBG0B	166	movq %mm0, MUNG(VBG0B)
164		167	movl _rows(%esp), %eax
165	movl _rows(%ebp), %eax	168	movl _columns(%esp), %ebx
166	movl _columns(%ebp), %ebx
167	mull %ebx #number pixels	169	mull %ebx #number pixels
168	shrl $3, %eax #number of loops	170	shrl $3, %eax #number of loops
169	movl %eax, %edi #loop counter in edi	171	movl %eax, %edi #loop counter in edi
170	movl _inPtr(%ebp), %eax	172	movl _inPtr(%esp), %eax
171	movl _outyPtr(%ebp), %ebx	173	movl _outyPtr(%esp), %ebx
172	movl _outuPtr(%ebp), %ecx	174	movl _outuPtr(%esp), %ecx
173	movl _outvPtr(%ebp), %edx	175	movl _outvPtr(%esp), %edx
174	rgbtoycb_mmx_loop:	176	rgbtoycb_mmx_loop:
175	movq (%eax), %mm1 #load G2R2B1G1R1B0G0R0	177	movq (%eax), %mm1 #load G2R2B1G1R1B0G0R0
176	pxor %mm6, %mm6 #0 -> mm6	178	pxor %mm6, %mm6 #0 -> mm6
Lines 184-212 rgbtoycb_mmx_loop: Link Here
184	punpcklbw %mm6, %mm1 #B1G1R1B0 -> mm1	186	punpcklbw %mm6, %mm1 #B1G1R1B0 -> mm1
185	movq %mm0, %mm2 #R1B0G0R0 -> mm2	187	movq %mm0, %mm2 #R1B0G0R0 -> mm2
186		188
187	pmaddwd YR0GR, %mm0 #yrR1,ygG0+yrR0 -> mm0	189	pmaddwd MUNG(YR0GR), %mm0 #yrR1,ygG0+yrR0 -> mm0
188	movq %mm1, %mm3 #B1G1R1B0 -> mm3	190	movq %mm1, %mm3 #B1G1R1B0 -> mm3
189		191
190	pmaddwd YBG0B, %mm1 #ybB1+ygG1,ybB0 -> mm1	192	pmaddwd MUNG(YBG0B), %mm1 #ybB1+ygG1,ybB0 -> mm1
191	movq %mm2, %mm4 #R1B0G0R0 -> mm4	193	movq %mm2, %mm4 #R1B0G0R0 -> mm4
192		194
193	pmaddwd UR0GR, %mm2 #urR1,ugG0+urR0 -> mm2	195	pmaddwd MUNG(UR0GR), %mm2 #urR1,ugG0+urR0 -> mm2
194	movq %mm3, %mm5 #B1G1R1B0 -> mm5	196	movq %mm3, %mm5 #B1G1R1B0 -> mm5
195		197
196	pmaddwd UBG0B, %mm3 #ubB1+ugG1,ubB0 -> mm3	198	pmaddwd MUNG(UBG0B), %mm3 #ubB1+ugG1,ubB0 -> mm3
197	punpckhbw %mm6, %mm7 # 00G2R2 -> mm7	199	punpckhbw %mm6, %mm7 # 00G2R2 -> mm7
198		200
199	pmaddwd VR0GR, %mm4 #vrR1,vgG0+vrR0 -> mm4	201	pmaddwd MUNG(VR0GR), %mm4 #vrR1,vgG0+vrR0 -> mm4
200	paddd %mm1, %mm0 #Y1Y0 -> mm0	202	paddd %mm1, %mm0 #Y1Y0 -> mm0
201		203
202	pmaddwd VBG0B, %mm5 #vbB1+vgG1,vbB0 -> mm5	204	pmaddwd MUNG(VBG0B), %mm5 #vbB1+vgG1,vbB0 -> mm5
203		205
204	movq 8(%eax), %mm1 #R5B4G4R4B3G3R3B2 -> mm1	206	movq 8(%eax), %mm1 #R5B4G4R4B3G3R3B2 -> mm1
205	paddd %mm3, %mm2 #U1U0 -> mm2	207	paddd %mm3, %mm2 #U1U0 -> mm2
206		208
207	movq %mm1, %mm6 #R5B4G4R4B3G3R3B2 -> mm6	209	movq %mm1, %mm6 #R5B4G4R4B3G3R3B2 -> mm6
208		210
209	punpcklbw ZEROS, %mm1 #B3G3R3B2 -> mm1	211	punpcklbw MUNG(ZEROS), %mm1 #B3G3R3B2 -> mm1
210	paddd %mm5, %mm4 #V1V0 -> mm4	212	paddd %mm5, %mm4 #V1V0 -> mm4
211		213
212	movq %mm1, %mm5 #B3G3R3B2 -> mm5	214	movq %mm1, %mm5 #B3G3R3B2 -> mm5
Lines 214-242 rgbtoycb_mmx_loop: Link Here
214		216
215	paddd %mm7, %mm1 #R3B200+00G2R2=R3B2G2R2->mm1	217	paddd %mm7, %mm1 #R3B200+00G2R2=R3B2G2R2->mm1
216		218
217	punpckhbw ZEROS, %mm6 #R5B4G4R3 -> mm6	219	punpckhbw MUNG(ZEROS), %mm6 #R5B4G4R3 -> mm6
218	movq %mm1, %mm3 #R3B2G2R2 -> mm3	220	movq %mm1, %mm3 #R3B2G2R2 -> mm3
219		221
220	pmaddwd YR0GR, %mm1 #yrR3,ygG2+yrR2 -> mm1	222	pmaddwd MUNG(YR0GR), %mm1 #yrR3,ygG2+yrR2 -> mm1
221	movq %mm5, %mm7 #B3G3R3B2 -> mm7	223	movq %mm5, %mm7 #B3G3R3B2 -> mm7
222		224
223	pmaddwd YBG0B, %mm5 #ybB3+ygG3,ybB2 -> mm5	225	pmaddwd MUNG(YBG0B), %mm5 #ybB3+ygG3,ybB2 -> mm5
224	psrad $FIXPSHIFT, %mm0 #32-bit scaled Y1Y0 -> mm0	226	psrad $FIXPSHIFT, %mm0 #32-bit scaled Y1Y0 -> mm0
225		227
226	movq %mm6, TEMP0 #R5B4G4R4 -> TEMP0	228	movq %mm6, MUNG(TEMP0) #R5B4G4R4 -> TEMP0
227	movq %mm3, %mm6 #R3B2G2R2 -> mm6	229	movq %mm3, %mm6 #R3B2G2R2 -> mm6
228	pmaddwd UR0GR, %mm6 #urR3,ugG2+urR2 -> mm6	230	pmaddwd MUNG(UR0GR), %mm6 #urR3,ugG2+urR2 -> mm6
229	psrad $FIXPSHIFT, %mm2 #32-bit scaled U1U0 -> mm2	231	psrad $FIXPSHIFT, %mm2 #32-bit scaled U1U0 -> mm2
230		232
231	paddd %mm5, %mm1 #Y3Y2 -> mm1	233	paddd %mm5, %mm1 #Y3Y2 -> mm1
232	movq %mm7, %mm5 #B3G3R3B2 -> mm5	234	movq %mm7, %mm5 #B3G3R3B2 -> mm5
233	pmaddwd UBG0B, %mm7 #ubB3+ugG3,ubB2	235	pmaddwd MUNG(UBG0B), %mm7 #ubB3+ugG3,ubB2
234	psrad $FIXPSHIFT, %mm1 #32-bit scaled Y3Y2 -> mm1	236	psrad $FIXPSHIFT, %mm1 #32-bit scaled Y3Y2 -> mm1
235		237
236	pmaddwd VR0GR, %mm3 #vrR3,vgG2+vgR2	238	pmaddwd MUNG(VR0GR), %mm3 #vrR3,vgG2+vgR2
237	packssdw %mm1, %mm0 #Y3Y2Y1Y0 -> mm0	239	packssdw %mm1, %mm0 #Y3Y2Y1Y0 -> mm0
238		240
239	pmaddwd VBG0B, %mm5 #vbB3+vgG3,vbB2 -> mm5	241	pmaddwd MUNG(VBG0B), %mm5 #vbB3+vgG3,vbB2 -> mm5
240	psrad $FIXPSHIFT, %mm4 #32-bit scaled V1V0 -> mm4	242	psrad $FIXPSHIFT, %mm4 #32-bit scaled V1V0 -> mm4
241		243
242	movq 16(%eax), %mm1 #B7G7R7B6G6R6B5G5 -> mm7	244	movq 16(%eax), %mm1 #B7G7R7B6G6R6B5G5 -> mm7
Lines 251-308 rgbtoycb_mmx_loop: Link Here
251	movq %mm7, %mm5 #R7B6G6R6B5G500 -> mm5	253	movq %mm7, %mm5 #R7B6G6R6B5G500 -> mm5
252	psrad $FIXPSHIFT, %mm3 #32-bit scaled V3V2 -> mm3	254	psrad $FIXPSHIFT, %mm3 #32-bit scaled V3V2 -> mm3
253		255
254	paddw OFFSETY, %mm0	256	paddw MUNG(OFFSETY), %mm0
255	movq %mm0, (%ebx) #store Y3Y2Y1Y0	257	movq %mm0, (%ebx) #store Y3Y2Y1Y0
256	packssdw %mm6, %mm2 #32-bit scaled U3U2U1U0 -> mm2	258	packssdw %mm6, %mm2 #32-bit scaled U3U2U1U0 -> mm2
257		259
258	movq TEMP0, %mm0 #R5B4G4R4 -> mm0	260	movq MUNG(TEMP0), %mm0 #R5B4G4R4 -> mm0
259	addl $8, %ebx	261	addl $8, %ebx
260		262
261	punpcklbw ZEROS, %mm7 #B5G500 -> mm7	263	punpcklbw MUNG(ZEROS), %mm7 #B5G500 -> mm7
262	movq %mm0, %mm6 #R5B4G4R4 -> mm6	264	movq %mm0, %mm6 #R5B4G4R4 -> mm6
263		265
264	movq %mm2, TEMPU #32-bit scaled U3U2U1U0 -> TEMPU	266	movq %mm2, MUNG(TEMPU) #32-bit scaled U3U2U1U0 -> TEMPU
265	psrlq $32, %mm0 #00R5B4 -> mm0	267	psrlq $32, %mm0 #00R5B4 -> mm0
266		268
267	paddw %mm0, %mm7 #B5G5R5B4 -> mm7	269	paddw %mm0, %mm7 #B5G5R5B4 -> mm7
268	movq %mm6, %mm2 #B5B4G4R4 -> mm2	270	movq %mm6, %mm2 #B5B4G4R4 -> mm2
269		271
270	pmaddwd YR0GR, %mm2 #yrR5,ygG4+yrR4 -> mm2	272	pmaddwd MUNG(YR0GR), %mm2 #yrR5,ygG4+yrR4 -> mm2
271	movq %mm7, %mm0 #B5G5R5B4 -> mm0	273	movq %mm7, %mm0 #B5G5R5B4 -> mm0
272		274
273	pmaddwd YBG0B, %mm7 #ybB5+ygG5,ybB4 -> mm7	275	pmaddwd MUNG(YBG0B), %mm7 #ybB5+ygG5,ybB4 -> mm7
274	packssdw %mm3, %mm4 #32-bit scaled V3V2V1V0 -> mm4	276	packssdw %mm3, %mm4 #32-bit scaled V3V2V1V0 -> mm4
275		277
276	addl $24, %eax #increment RGB count	278	addl $24, %eax #increment RGB count
277		279
278	movq %mm4, TEMPV #(V3V2V1V0)/256 -> mm4	280	movq %mm4, MUNG(TEMPV) #(V3V2V1V0)/256 -> mm4
279	movq %mm6, %mm4 #B5B4G4R4 -> mm4	281	movq %mm6, %mm4 #B5B4G4R4 -> mm4
280		282
281	pmaddwd UR0GR, %mm6 #urR5,ugG4+urR4	283	pmaddwd MUNG(UR0GR), %mm6 #urR5,ugG4+urR4
282	movq %mm0, %mm3 #B5G5R5B4 -> mm0	284	movq %mm0, %mm3 #B5G5R5B4 -> mm0
283		285
284	pmaddwd UBG0B, %mm0 #ubB5+ugG5,ubB4	286	pmaddwd MUNG(UBG0B), %mm0 #ubB5+ugG5,ubB4
285	paddd %mm7, %mm2 #Y5Y4 -> mm2	287	paddd %mm7, %mm2 #Y5Y4 -> mm2
286		288
287	pmaddwd VR0GR, %mm4 #vrR5,vgG4+vrR4 -> mm4	289	pmaddwd MUNG(VR0GR), %mm4 #vrR5,vgG4+vrR4 -> mm4
288	pxor %mm7, %mm7 #0 -> mm7	290	pxor %mm7, %mm7 #0 -> mm7
289		291
290	pmaddwd VBG0B, %mm3 #vbB5+vgG5,vbB4 -> mm3	292	pmaddwd MUNG(VBG0B), %mm3 #vbB5+vgG5,vbB4 -> mm3
291	punpckhbw %mm7, %mm1 #B7G7R7B6 -> mm1	293	punpckhbw %mm7, %mm1 #B7G7R7B6 -> mm1
292		294
293	paddd %mm6, %mm0 #U5U4 -> mm0	295	paddd %mm6, %mm0 #U5U4 -> mm0
294	movq %mm1, %mm6 #B7G7R7B6 -> mm6	296	movq %mm1, %mm6 #B7G7R7B6 -> mm6
295		297
296	pmaddwd YBG0B, %mm6 #ybB7+ygG7,ybB6 -> mm6	298	pmaddwd MUNG(YBG0B), %mm6 #ybB7+ygG7,ybB6 -> mm6
297	punpckhbw %mm7, %mm5 #R7B6G6R6 -> mm5	299	punpckhbw %mm7, %mm5 #R7B6G6R6 -> mm5
298		300
299	movq %mm5, %mm7 #R7B6G6R6 -> mm7	301	movq %mm5, %mm7 #R7B6G6R6 -> mm7
300	paddd %mm4, %mm3 #V5V4 -> mm3	302	paddd %mm4, %mm3 #V5V4 -> mm3
301		303
302	pmaddwd YR0GR, %mm5 #yrR7,ygG6+yrR6 -> mm5	304	pmaddwd MUNG(YR0GR), %mm5 #yrR7,ygG6+yrR6 -> mm5
303	movq %mm1, %mm4 #B7G7R7B6 -> mm4	305	movq %mm1, %mm4 #B7G7R7B6 -> mm4
304		306
305	pmaddwd UBG0B, %mm4 #ubB7+ugG7,ubB6 -> mm4	307	pmaddwd MUNG(UBG0B), %mm4 #ubB7+ugG7,ubB6 -> mm4
306	psrad $FIXPSHIFT, %mm0 #32-bit scaled U5U4 -> mm0	308	psrad $FIXPSHIFT, %mm0 #32-bit scaled U5U4 -> mm0
307		309
308	psrad $FIXPSHIFT, %mm2 #32-bit scaled Y5Y4 -> mm2	310	psrad $FIXPSHIFT, %mm2 #32-bit scaled Y5Y4 -> mm2
Lines 310-334 rgbtoycb_mmx_loop: Link Here
310	paddd %mm5, %mm6 #Y7Y6 -> mm6	312	paddd %mm5, %mm6 #Y7Y6 -> mm6
311	movq %mm7, %mm5 #R7B6G6R6 -> mm5	313	movq %mm7, %mm5 #R7B6G6R6 -> mm5
312		314
313	pmaddwd UR0GR, %mm7 #urR7,ugG6+ugR6 -> mm7	315	pmaddwd MUNG(UR0GR), %mm7 #urR7,ugG6+ugR6 -> mm7
314	psrad $FIXPSHIFT, %mm3 #32-bit scaled V5V4 -> mm3	316	psrad $FIXPSHIFT, %mm3 #32-bit scaled V5V4 -> mm3
315		317
316	pmaddwd VBG0B, %mm1 #vbB7+vgG7,vbB6 -> mm1	318	pmaddwd MUNG(VBG0B), %mm1 #vbB7+vgG7,vbB6 -> mm1
317	psrad $FIXPSHIFT, %mm6 #32-bit scaled Y7Y6 -> mm6	319	psrad $FIXPSHIFT, %mm6 #32-bit scaled Y7Y6 -> mm6
318		320
319	packssdw %mm6, %mm2 #Y7Y6Y5Y4 -> mm2	321	packssdw %mm6, %mm2 #Y7Y6Y5Y4 -> mm2
320		322
321	pmaddwd VR0GR, %mm5 #vrR7,vgG6+vrR6 -> mm5	323	pmaddwd MUNG(VR0GR), %mm5 #vrR7,vgG6+vrR6 -> mm5
322	paddd %mm4, %mm7 #U7U6 -> mm7	324	paddd %mm4, %mm7 #U7U6 -> mm7
323		325
324	psrad $FIXPSHIFT, %mm7 #32-bit scaled U7U6 -> mm7	326	psrad $FIXPSHIFT, %mm7 #32-bit scaled U7U6 -> mm7
325	paddw OFFSETY, %mm2	327	paddw MUNG(OFFSETY), %mm2
326	movq %mm2, (%ebx) #store Y7Y6Y5Y4	328	movq %mm2, (%ebx) #store Y7Y6Y5Y4
327		329
328	movq ALLONE, %mm6	330	movq MUNG(ALLONE), %mm6
329	packssdw %mm7, %mm0 #32-bit scaled U7U6U5U4 -> mm0	331	packssdw %mm7, %mm0 #32-bit scaled U7U6U5U4 -> mm0
330		332
331	movq TEMPU, %mm4 #32-bit scaled U3U2U1U0 -> mm4	333	movq MUNG(TEMPU), %mm4 #32-bit scaled U3U2U1U0 -> mm4
332	pmaddwd %mm6, %mm0 #U7U6U5U4 averaged -> (U7U6)(U5U4)=UU3 UU2->mm0	334	pmaddwd %mm6, %mm0 #U7U6U5U4 averaged -> (U7U6)(U5U4)=UU3 UU2->mm0
333		335
334	pmaddwd %mm6, %mm4 #U3U2U1U0 averaged -> (U3U2)(U1U0)=UU1 UU0->mm4	336	pmaddwd %mm6, %mm4 #U3U2U1U0 averaged -> (U3U2)(U1U0)=UU1 UU0->mm4
Lines 338-345 rgbtoycb_mmx_loop: Link Here
338		340
339	psrad $FIXPSHIFT, %mm1 #32-bit scaled V7V6 -> mm1	341	psrad $FIXPSHIFT, %mm1 #32-bit scaled V7V6 -> mm1
340	psraw $1, %mm4 #divide UU3 UU2 UU1 UU0 by 2 -> mm4	342	psraw $1, %mm4 #divide UU3 UU2 UU1 UU0 by 2 -> mm4
341		343
342	movq TEMPV, %mm5 #32-bit scaled V3V2V1V0 -> mm5	344	movq MUNG(TEMPV), %mm5 #32-bit scaled V3V2V1V0 -> mm5
343		345
344	movq %mm4, (%ecx) # store U	346	movq %mm4, (%ecx) # store U
345		347
Lines 372-377 rgbtoycb_mmx_loop: Link Here
372	ret	374	ret
373		375
374	.global _dv_ppm_copy_y_block_mmx	376	.global _dv_ppm_copy_y_block_mmx
		377	.hidden _dv_ppm_copy_y_block_mmx
		378	.type _dv_ppm_copy_y_block_mmx,@function
375	_dv_ppm_copy_y_block_mmx:	379	_dv_ppm_copy_y_block_mmx:
376		380
377	pushl %ebp	381	pushl %ebp
Lines 422-438 _dv_ppm_copy_y_block_mmx: Link Here
422	ret	426	ret
423		427
424	.global _dv_pgm_copy_y_block_mmx	428	.global _dv_pgm_copy_y_block_mmx
		429	.hidden _dv_pgm_copy_y_block_mmx
		430	.type _dv_ppm_copy_y_block_mmx,@function
425	_dv_pgm_copy_y_block_mmx:	431	_dv_pgm_copy_y_block_mmx:
426		432
427	pushl %ebp	433	pushl %ebp
428	movl %esp, %ebp
429	pushl %esi	434	pushl %esi
430	pushl %edi	435	pushl %edi
431
432	movl 8(%ebp), %edi # dest
433	movl 12(%ebp), %esi # src
434		436
435	movq OFFSETY, %mm7	437	LOAD_PIC_REG_BP()
		438
		439	movl 16(%esp), %edi # dest
		440	movl 20(%esp), %esi # src
		441
		442	movq MUNG(OFFSETY), %mm7
436	pxor %mm6, %mm6	443	pxor %mm6, %mm6
437		444
438	movq (%esi), %mm0	445	movq (%esi), %mm0
Lines 564-580 _dv_pgm_copy_y_block_mmx: Link Here
564	ret	571	ret
565		572
566	.global _dv_video_copy_y_block_mmx	573	.global _dv_video_copy_y_block_mmx
		574	.hidden _dv_video_copy_y_block_mmx
		575	.type _dv_video_copy_y_block_mmx,@function
567	_dv_video_copy_y_block_mmx:	576	_dv_video_copy_y_block_mmx:
568		577
569	pushl %ebp	578	pushl %ebp
570	movl %esp, %ebp
571	pushl %esi	579	pushl %esi
572	pushl %edi	580	pushl %edi
573
574	movl 8(%ebp), %edi # dest
575	movl 12(%ebp), %esi # src
576		581
577	movq OFFSETBX, %mm7	582	LOAD_PIC_REG_BP()
		583
		584	movl 16(%esp), %edi # dest
		585	movl 20(%esp), %esi # src
		586
		587	movq MUNG(OFFSETBX), %mm7
578	pxor %mm6, %mm6	588	pxor %mm6, %mm6
579		589
580	movq (%esi), %mm0	590	movq (%esi), %mm0
Lines 709-714 _dv_video_copy_y_block_mmx: Link Here
709		719
710		720
711	.global _dv_ppm_copy_pal_c_block_mmx	721	.global _dv_ppm_copy_pal_c_block_mmx
		722	.hidden _dv_ppm_copy_pal_c_block_mmx
		723	.type _dv_ppm_copy_pal_c_block_mmx,@function
712	_dv_ppm_copy_pal_c_block_mmx:	724	_dv_ppm_copy_pal_c_block_mmx:
713		725
714	pushl %ebp	726	pushl %ebp
Lines 852-870 _dv_ppm_copy_pal_c_block_mmx: Link Here
852	ret	864	ret
853		865
854	.global _dv_pgm_copy_pal_c_block_mmx	866	.global _dv_pgm_copy_pal_c_block_mmx
		867	.hidden _dv_ppm_copy_pal_c_block_mmx
		868	.type _dv_pgm_copy_pal_c_block_mmx,@function
855	_dv_pgm_copy_pal_c_block_mmx:	869	_dv_pgm_copy_pal_c_block_mmx:
856		870
857	pushl %ebp	871	pushl %ebp
858	movl %esp, %ebp
859	pushl %esi	872	pushl %esi
860	pushl %edi	873	pushl %edi
861	pushl %ebx	874	pushl %ebx
862
863	movl 8(%ebp), %edi # dest
864	movl 12(%ebp), %esi # src
865		875
		876	LOAD_PIC_REG_BP()
		877
		878	movl 20(%esp), %edi # dest
		879	movl 24(%esp), %esi # src
866		880
867	movq OFFSETBX, %mm7	881	movq MUNG(OFFSETBX), %mm7
868	pxor %mm6, %mm6	882	pxor %mm6, %mm6
869		883
870		884
Lines 1000-1017 _dv_pgm_copy_pal_c_block_mmx: Link Here
1000	ret	1014	ret
1001		1015
1002	.global _dv_video_copy_pal_c_block_mmx	1016	.global _dv_video_copy_pal_c_block_mmx
		1017	.hidden _dv_video_copy_pal_c_block_mmx
		1018	.type _dv_video_copy_pal_c_block_mmx,@function
1003	_dv_video_copy_pal_c_block_mmx:	1019	_dv_video_copy_pal_c_block_mmx:
1004		1020
1005	pushl %ebp	1021	pushl %ebp
1006	movl %esp, %ebp
1007	pushl %esi	1022	pushl %esi
1008	pushl %edi	1023	pushl %edi
1009	pushl %ebx	1024	pushl %ebx
1010
1011	movl 8(%ebp), %edi # dest
1012	movl 12(%ebp), %esi # src
1013		1025
1014	movq OFFSETBX, %mm7	1026	LOAD_PIC_REG_BP()
		1027
		1028	movl 20(%esp), %edi # dest
		1029	movl 24(%esp), %esi # src
		1030
		1031	movq MUNG(OFFSETBX), %mm7
1015	paddw %mm7, %mm7	1032	paddw %mm7, %mm7
1016	pxor %mm6, %mm6	1033	pxor %mm6, %mm6
1017		1034
Lines 1095-1115 video_copy_pal_c_block_mmx_loop: Link Here
1095	ret	1112	ret
1096		1113
1097	.global _dv_ppm_copy_ntsc_c_block_mmx	1114	.global _dv_ppm_copy_ntsc_c_block_mmx
		1115	.hidden _dv_ppm_copy_ntsc_c_block_mmx
		1116	.type _dv_ppm_copy_ntsc_c_block_mmx,@function
1098	_dv_ppm_copy_ntsc_c_block_mmx:	1117	_dv_ppm_copy_ntsc_c_block_mmx:
1099		1118
1100	pushl %ebp	1119	pushl %ebp
1101	movl %esp, %ebp
1102	pushl %esi	1120	pushl %esi
1103	pushl %edi	1121	pushl %edi
1104	pushl %ebx	1122	pushl %ebx
1105		1123
1106	movl 8(%ebp), %edi # dest	1124	LOAD_PIC_REG_BP()
1107	movl 12(%ebp), %esi # src	1125
		1126	movl 20(%esp), %edi # dest
		1127	movl 24(%esp), %esi # src
1108		1128
1109	movl $4, %ebx	1129	movl $4, %ebx
1110		1130
1111	movq ALLONE, %mm6	1131	movq MUNG(ALLONE), %mm6
1112
1113	ppm_copy_ntsc_c_block_mmx_loop:	1132	ppm_copy_ntsc_c_block_mmx_loop:
1114		1133
1115	movq (%esi), %mm0	1134	movq (%esi), %mm0
Lines 1168-1184 ppm_copy_ntsc_c_block_mmx_loop: Link Here
1168	ret	1187	ret
1169		1188
1170	.global _dv_pgm_copy_ntsc_c_block_mmx	1189	.global _dv_pgm_copy_ntsc_c_block_mmx
		1190	.hidden _dv_pgm_copy_ntsc_c_block_mmx
		1191	.type _dv_pgm_copy_ntsc_c_block_mmx,@function
1171	_dv_pgm_copy_ntsc_c_block_mmx:	1192	_dv_pgm_copy_ntsc_c_block_mmx:
1172		1193
1173	pushl %ebp	1194	pushl %ebp
1174	movl %esp, %ebp
1175	pushl %esi	1195	pushl %esi
1176	pushl %edi	1196	pushl %edi
1177
1178	movl 8(%ebp), %edi # dest
1179	movl 12(%ebp), %esi # src
1180		1197
1181	movq OFFSETBX, %mm7	1198	LOAD_PIC_REG_BP()
		1199
		1200	movl 16(%esp), %edi # dest
		1201	movl 20(%esp), %esi # src
		1202
		1203	movq MUNG(OFFSETBX), %mm7
1182	paddw %mm7, %mm7	1204	paddw %mm7, %mm7
1183	pxor %mm6, %mm6	1205	pxor %mm6, %mm6
1184		1206
Lines 1325-1342 _dv_pgm_copy_ntsc_c_block_mmx: Link Here
1325	ret	1347	ret
1326		1348
1327	.global _dv_video_copy_ntsc_c_block_mmx	1349	.global _dv_video_copy_ntsc_c_block_mmx
		1350	.hidden _dv_video_copy_ntsc_c_block_mmx
		1351	.type _dv_video_copy_ntsc_c_block_mmx,@function
1328	_dv_video_copy_ntsc_c_block_mmx:	1352	_dv_video_copy_ntsc_c_block_mmx:
1329		1353
1330	pushl %ebp	1354	pushl %ebp
1331	movl %esp, %ebp
1332	pushl %esi	1355	pushl %esi
1333	pushl %edi	1356	pushl %edi
1334	pushl %ebx	1357	pushl %ebx
1335
1336	movl 8(%ebp), %edi # dest
1337	movl 12(%ebp), %esi # src
1338		1358
1339	movq OFFSETBX, %mm7	1359	LOAD_PIC_REG_BP()
		1360
		1361	movl 20(%esp), %edi # dest
		1362	movl 24(%esp), %esi # src
		1363
		1364	movq MUNG(OFFSETBX), %mm7
1340	paddw %mm7, %mm7	1365	paddw %mm7, %mm7
1341	pxor %mm6, %mm6	1366	pxor %mm6, %mm6
1342		1367