• file
• revisions
• annotate
• diff
• comparison
• raw

media/libvpx/vp8/encoder/x86/encodeopt.asm@ac0c01689b40 (annotated)

media/libvpx/vp8/encoder/x86/encodeopt.asm

Thu, 15 Jan 2015 15:59:08 +0100

author

Michael Schloh von Bennewitz <michael@schloh.com>

date

Thu, 15 Jan 2015 15:59:08 +0100

branch

TOR_BUG_9701

changeset 10

ac0c01689b40

permissions

-rw-r--r--

Implement a real Private Browsing Mode condition by changing the API/ABI;
This solves Tor bug #9701, complying with disk avoidance documented in
https://www.torproject.org/projects/torbrowser/design/#disk-avoidance.

michael@0	1	;
michael@0	2	; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
michael@0	3	;
michael@0	4	; Use of this source code is governed by a BSD-style license
michael@0	5	; that can be found in the LICENSE file in the root of the source
michael@0	6	; tree. An additional intellectual property rights grant can be found
michael@0	7	; in the file PATENTS. All contributing project authors may
michael@0	8	; be found in the AUTHORS file in the root of the source tree.
michael@0	9	;
michael@0	10
michael@0	11
michael@0	12	%include "vpx_ports/x86_abi_support.asm"
michael@0	13
michael@0	14	;int vp8_block_error_xmm(short *coeff_ptr, short *dcoef_ptr)
michael@0	15	global sym(vp8_block_error_xmm) PRIVATE
michael@0	16	sym(vp8_block_error_xmm):
michael@0	17	push rbp
michael@0	18	mov rbp, rsp
michael@0	19	SHADOW_ARGS_TO_STACK 2
michael@0	20	push rsi
michael@0	21	push rdi
michael@0	22	; end prologue
michael@0	23
michael@0	24	mov rsi, arg(0) ;coeff_ptr
michael@0	25	mov rdi, arg(1) ;dcoef_ptr
michael@0	26
michael@0	27	movdqa xmm0, [rsi]
michael@0	28	movdqa xmm1, [rdi]
michael@0	29
michael@0	30	movdqa xmm2, [rsi+16]
michael@0	31	movdqa xmm3, [rdi+16]
michael@0	32
michael@0	33	psubw xmm0, xmm1
michael@0	34	psubw xmm2, xmm3
michael@0	35
michael@0	36	pmaddwd xmm0, xmm0
michael@0	37	pmaddwd xmm2, xmm2
michael@0	38
michael@0	39	paddd xmm0, xmm2
michael@0	40
michael@0	41	pxor xmm5, xmm5
michael@0	42	movdqa xmm1, xmm0
michael@0	43
michael@0	44	punpckldq xmm0, xmm5
michael@0	45	punpckhdq xmm1, xmm5
michael@0	46
michael@0	47	paddd xmm0, xmm1
michael@0	48	movdqa xmm1, xmm0
michael@0	49
michael@0	50	psrldq xmm0, 8
michael@0	51	paddd xmm0, xmm1
michael@0	52
michael@0	53	movq rax, xmm0
michael@0	54
michael@0	55	pop rdi
michael@0	56	pop rsi
michael@0	57	; begin epilog
michael@0	58	UNSHADOW_ARGS
michael@0	59	pop rbp
michael@0	60	ret
michael@0	61
michael@0	62	;int vp8_block_error_mmx(short *coeff_ptr, short *dcoef_ptr)
michael@0	63	global sym(vp8_block_error_mmx) PRIVATE
michael@0	64	sym(vp8_block_error_mmx):
michael@0	65	push rbp
michael@0	66	mov rbp, rsp
michael@0	67	SHADOW_ARGS_TO_STACK 2
michael@0	68	push rsi
michael@0	69	push rdi
michael@0	70	; end prolog
michael@0	71
michael@0	72
michael@0	73	mov rsi, arg(0) ;coeff_ptr
michael@0	74	pxor mm7, mm7
michael@0	75
michael@0	76	mov rdi, arg(1) ;dcoef_ptr
michael@0	77	movq mm3, [rsi]
michael@0	78
michael@0	79	movq mm4, [rdi]
michael@0	80	movq mm5, [rsi+8]
michael@0	81
michael@0	82	movq mm6, [rdi+8]
michael@0	83	pxor mm1, mm1 ; from movd mm1, dc ; dc =0
michael@0	84
michael@0	85	movq mm2, mm7
michael@0	86	psubw mm5, mm6
michael@0	87
michael@0	88	por mm1, mm2
michael@0	89	pmaddwd mm5, mm5
michael@0	90
michael@0	91	pcmpeqw mm1, mm7
michael@0	92	psubw mm3, mm4
michael@0	93
michael@0	94	pand mm1, mm3
michael@0	95	pmaddwd mm1, mm1
michael@0	96
michael@0	97	paddd mm1, mm5
michael@0	98	movq mm3, [rsi+16]
michael@0	99
michael@0	100	movq mm4, [rdi+16]
michael@0	101	movq mm5, [rsi+24]
michael@0	102
michael@0	103	movq mm6, [rdi+24]
michael@0	104	psubw mm5, mm6
michael@0	105
michael@0	106	pmaddwd mm5, mm5
michael@0	107	psubw mm3, mm4
michael@0	108
michael@0	109	pmaddwd mm3, mm3
michael@0	110	paddd mm3, mm5
michael@0	111
michael@0	112	paddd mm1, mm3
michael@0	113	movq mm0, mm1
michael@0	114
michael@0	115	psrlq mm1, 32
michael@0	116	paddd mm0, mm1
michael@0	117
michael@0	118	movq rax, mm0
michael@0	119
michael@0	120	pop rdi
michael@0	121	pop rsi
michael@0	122	; begin epilog
michael@0	123	UNSHADOW_ARGS
michael@0	124	pop rbp
michael@0	125	ret
michael@0	126
michael@0	127
michael@0	128	;int vp8_mbblock_error_mmx_impl(short *coeff_ptr, short *dcoef_ptr, int dc);
michael@0	129	global sym(vp8_mbblock_error_mmx_impl) PRIVATE
michael@0	130	sym(vp8_mbblock_error_mmx_impl):
michael@0	131	push rbp
michael@0	132	mov rbp, rsp
michael@0	133	SHADOW_ARGS_TO_STACK 3
michael@0	134	push rsi
michael@0	135	push rdi
michael@0	136	; end prolog
michael@0	137
michael@0	138
michael@0	139	mov rsi, arg(0) ;coeff_ptr
michael@0	140	pxor mm7, mm7
michael@0	141
michael@0	142	mov rdi, arg(1) ;dcoef_ptr
michael@0	143	pxor mm2, mm2
michael@0	144
michael@0	145	movd mm1, dword ptr arg(2) ;dc
michael@0	146	por mm1, mm2
michael@0	147
michael@0	148	pcmpeqw mm1, mm7
michael@0	149	mov rcx, 16
michael@0	150
michael@0	151	.mberror_loop_mmx:
michael@0	152	movq mm3, [rsi]
michael@0	153	movq mm4, [rdi]
michael@0	154
michael@0	155	movq mm5, [rsi+8]
michael@0	156	movq mm6, [rdi+8]
michael@0	157
michael@0	158
michael@0	159	psubw mm5, mm6
michael@0	160	pmaddwd mm5, mm5
michael@0	161
michael@0	162	psubw mm3, mm4
michael@0	163	pand mm3, mm1
michael@0	164
michael@0	165	pmaddwd mm3, mm3
michael@0	166	paddd mm2, mm5
michael@0	167
michael@0	168	paddd mm2, mm3
michael@0	169	movq mm3, [rsi+16]
michael@0	170
michael@0	171	movq mm4, [rdi+16]
michael@0	172	movq mm5, [rsi+24]
michael@0	173
michael@0	174	movq mm6, [rdi+24]
michael@0	175	psubw mm5, mm6
michael@0	176
michael@0	177	pmaddwd mm5, mm5
michael@0	178	psubw mm3, mm4
michael@0	179
michael@0	180	pmaddwd mm3, mm3
michael@0	181	paddd mm2, mm5
michael@0	182
michael@0	183	paddd mm2, mm3
michael@0	184	add rsi, 32
michael@0	185
michael@0	186	add rdi, 32
michael@0	187	sub rcx, 1
michael@0	188
michael@0	189	jnz .mberror_loop_mmx
michael@0	190
michael@0	191	movq mm0, mm2
michael@0	192	psrlq mm2, 32
michael@0	193
michael@0	194	paddd mm0, mm2
michael@0	195	movq rax, mm0
michael@0	196
michael@0	197	pop rdi
michael@0	198	pop rsi
michael@0	199	; begin epilog
michael@0	200	UNSHADOW_ARGS
michael@0	201	pop rbp
michael@0	202	ret
michael@0	203
michael@0	204
michael@0	205	;int vp8_mbblock_error_xmm_impl(short *coeff_ptr, short *dcoef_ptr, int dc);
michael@0	206	global sym(vp8_mbblock_error_xmm_impl) PRIVATE
michael@0	207	sym(vp8_mbblock_error_xmm_impl):
michael@0	208	push rbp
michael@0	209	mov rbp, rsp
michael@0	210	SHADOW_ARGS_TO_STACK 3
michael@0	211	SAVE_XMM 6
michael@0	212	push rsi
michael@0	213	push rdi
michael@0	214	; end prolog
michael@0	215
michael@0	216
michael@0	217	mov rsi, arg(0) ;coeff_ptr
michael@0	218	pxor xmm6, xmm6
michael@0	219
michael@0	220	mov rdi, arg(1) ;dcoef_ptr
michael@0	221	pxor xmm4, xmm4
michael@0	222
michael@0	223	movd xmm5, dword ptr arg(2) ;dc
michael@0	224	por xmm5, xmm4
michael@0	225
michael@0	226	pcmpeqw xmm5, xmm6
michael@0	227	mov rcx, 16
michael@0	228
michael@0	229	.mberror_loop:
michael@0	230	movdqa xmm0, [rsi]
michael@0	231	movdqa xmm1, [rdi]
michael@0	232
michael@0	233	movdqa xmm2, [rsi+16]
michael@0	234	movdqa xmm3, [rdi+16]
michael@0	235
michael@0	236
michael@0	237	psubw xmm2, xmm3
michael@0	238	pmaddwd xmm2, xmm2
michael@0	239
michael@0	240	psubw xmm0, xmm1
michael@0	241	pand xmm0, xmm5
michael@0	242
michael@0	243	pmaddwd xmm0, xmm0
michael@0	244	add rsi, 32
michael@0	245
michael@0	246	add rdi, 32
michael@0	247
michael@0	248	sub rcx, 1
michael@0	249	paddd xmm4, xmm2
michael@0	250
michael@0	251	paddd xmm4, xmm0
michael@0	252	jnz .mberror_loop
michael@0	253
michael@0	254	movdqa xmm0, xmm4
michael@0	255	punpckldq xmm0, xmm6
michael@0	256
michael@0	257	punpckhdq xmm4, xmm6
michael@0	258	paddd xmm0, xmm4
michael@0	259
michael@0	260	movdqa xmm1, xmm0
michael@0	261	psrldq xmm0, 8
michael@0	262
michael@0	263	paddd xmm0, xmm1
michael@0	264	movq rax, xmm0
michael@0	265
michael@0	266	pop rdi
michael@0	267	pop rsi
michael@0	268	; begin epilog
michael@0	269	RESTORE_XMM
michael@0	270	UNSHADOW_ARGS
michael@0	271	pop rbp
michael@0	272	ret
michael@0	273
michael@0	274
michael@0	275	;int vp8_mbuverror_mmx_impl(short *s_ptr, short *d_ptr);
michael@0	276	global sym(vp8_mbuverror_mmx_impl) PRIVATE
michael@0	277	sym(vp8_mbuverror_mmx_impl):
michael@0	278	push rbp
michael@0	279	mov rbp, rsp
michael@0	280	SHADOW_ARGS_TO_STACK 2
michael@0	281	push rsi
michael@0	282	push rdi
michael@0	283	; end prolog
michael@0	284
michael@0	285
michael@0	286	mov rsi, arg(0) ;s_ptr
michael@0	287	mov rdi, arg(1) ;d_ptr
michael@0	288
michael@0	289	mov rcx, 16
michael@0	290	pxor mm7, mm7
michael@0	291
michael@0	292	.mbuverror_loop_mmx:
michael@0	293
michael@0	294	movq mm1, [rsi]
michael@0	295	movq mm2, [rdi]
michael@0	296
michael@0	297	psubw mm1, mm2
michael@0	298	pmaddwd mm1, mm1
michael@0	299
michael@0	300
michael@0	301	movq mm3, [rsi+8]
michael@0	302	movq mm4, [rdi+8]
michael@0	303
michael@0	304	psubw mm3, mm4
michael@0	305	pmaddwd mm3, mm3
michael@0	306
michael@0	307
michael@0	308	paddd mm7, mm1
michael@0	309	paddd mm7, mm3
michael@0	310
michael@0	311
michael@0	312	add rsi, 16
michael@0	313	add rdi, 16
michael@0	314
michael@0	315	dec rcx
michael@0	316	jnz .mbuverror_loop_mmx
michael@0	317
michael@0	318	movq mm0, mm7
michael@0	319	psrlq mm7, 32
michael@0	320
michael@0	321	paddd mm0, mm7
michael@0	322	movq rax, mm0
michael@0	323
michael@0	324	pop rdi
michael@0	325	pop rsi
michael@0	326	; begin epilog
michael@0	327	UNSHADOW_ARGS
michael@0	328	pop rbp
michael@0	329	ret
michael@0	330
michael@0	331
michael@0	332	;int vp8_mbuverror_xmm_impl(short *s_ptr, short *d_ptr);
michael@0	333	global sym(vp8_mbuverror_xmm_impl) PRIVATE
michael@0	334	sym(vp8_mbuverror_xmm_impl):
michael@0	335	push rbp
michael@0	336	mov rbp, rsp
michael@0	337	SHADOW_ARGS_TO_STACK 2
michael@0	338	push rsi
michael@0	339	push rdi
michael@0	340	; end prolog
michael@0	341
michael@0	342
michael@0	343	mov rsi, arg(0) ;s_ptr
michael@0	344	mov rdi, arg(1) ;d_ptr
michael@0	345
michael@0	346	mov rcx, 16
michael@0	347	pxor xmm3, xmm3
michael@0	348
michael@0	349	.mbuverror_loop:
michael@0	350
michael@0	351	movdqa xmm1, [rsi]
michael@0	352	movdqa xmm2, [rdi]
michael@0	353
michael@0	354	psubw xmm1, xmm2
michael@0	355	pmaddwd xmm1, xmm1
michael@0	356
michael@0	357	paddd xmm3, xmm1
michael@0	358
michael@0	359	add rsi, 16
michael@0	360	add rdi, 16
michael@0	361
michael@0	362	dec rcx
michael@0	363	jnz .mbuverror_loop
michael@0	364
michael@0	365	pxor xmm0, xmm0
michael@0	366	movdqa xmm1, xmm3
michael@0	367
michael@0	368	movdqa xmm2, xmm1
michael@0	369	punpckldq xmm1, xmm0
michael@0	370
michael@0	371	punpckhdq xmm2, xmm0
michael@0	372	paddd xmm1, xmm2
michael@0	373
michael@0	374	movdqa xmm2, xmm1
michael@0	375
michael@0	376	psrldq xmm1, 8
michael@0	377	paddd xmm1, xmm2
michael@0	378
michael@0	379	movq rax, xmm1
michael@0	380
michael@0	381	pop rdi
michael@0	382	pop rsi
michael@0	383	; begin epilog
michael@0	384	UNSHADOW_ARGS
michael@0	385	pop rbp
michael@0	386	ret