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

media/libvpx/vp8/common/x86/sad_sse4.asm@ac0c01689b40 (annotated)

media/libvpx/vp8/common/x86/sad_sse4.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	%macro PROCESS_16X2X8 1
michael@0	15	%if %1
michael@0	16	movdqa xmm0, XMMWORD PTR [rsi]
michael@0	17	movq xmm1, MMWORD PTR [rdi]
michael@0	18	movq xmm3, MMWORD PTR [rdi+8]
michael@0	19	movq xmm2, MMWORD PTR [rdi+16]
michael@0	20	punpcklqdq xmm1, xmm3
michael@0	21	punpcklqdq xmm3, xmm2
michael@0	22
michael@0	23	movdqa xmm2, xmm1
michael@0	24	mpsadbw xmm1, xmm0, 0x0
michael@0	25	mpsadbw xmm2, xmm0, 0x5
michael@0	26
michael@0	27	psrldq xmm0, 8
michael@0	28
michael@0	29	movdqa xmm4, xmm3
michael@0	30	mpsadbw xmm3, xmm0, 0x0
michael@0	31	mpsadbw xmm4, xmm0, 0x5
michael@0	32
michael@0	33	paddw xmm1, xmm2
michael@0	34	paddw xmm1, xmm3
michael@0	35	paddw xmm1, xmm4
michael@0	36	%else
michael@0	37	movdqa xmm0, XMMWORD PTR [rsi]
michael@0	38	movq xmm5, MMWORD PTR [rdi]
michael@0	39	movq xmm3, MMWORD PTR [rdi+8]
michael@0	40	movq xmm2, MMWORD PTR [rdi+16]
michael@0	41	punpcklqdq xmm5, xmm3
michael@0	42	punpcklqdq xmm3, xmm2
michael@0	43
michael@0	44	movdqa xmm2, xmm5
michael@0	45	mpsadbw xmm5, xmm0, 0x0
michael@0	46	mpsadbw xmm2, xmm0, 0x5
michael@0	47
michael@0	48	psrldq xmm0, 8
michael@0	49
michael@0	50	movdqa xmm4, xmm3
michael@0	51	mpsadbw xmm3, xmm0, 0x0
michael@0	52	mpsadbw xmm4, xmm0, 0x5
michael@0	53
michael@0	54	paddw xmm5, xmm2
michael@0	55	paddw xmm5, xmm3
michael@0	56	paddw xmm5, xmm4
michael@0	57
michael@0	58	paddw xmm1, xmm5
michael@0	59	%endif
michael@0	60	movdqa xmm0, XMMWORD PTR [rsi + rax]
michael@0	61	movq xmm5, MMWORD PTR [rdi+ rdx]
michael@0	62	movq xmm3, MMWORD PTR [rdi+ rdx+8]
michael@0	63	movq xmm2, MMWORD PTR [rdi+ rdx+16]
michael@0	64	punpcklqdq xmm5, xmm3
michael@0	65	punpcklqdq xmm3, xmm2
michael@0	66
michael@0	67	lea rsi, [rsi+rax*2]
michael@0	68	lea rdi, [rdi+rdx*2]
michael@0	69
michael@0	70	movdqa xmm2, xmm5
michael@0	71	mpsadbw xmm5, xmm0, 0x0
michael@0	72	mpsadbw xmm2, xmm0, 0x5
michael@0	73
michael@0	74	psrldq xmm0, 8
michael@0	75	movdqa xmm4, xmm3
michael@0	76	mpsadbw xmm3, xmm0, 0x0
michael@0	77	mpsadbw xmm4, xmm0, 0x5
michael@0	78
michael@0	79	paddw xmm5, xmm2
michael@0	80	paddw xmm5, xmm3
michael@0	81	paddw xmm5, xmm4
michael@0	82
michael@0	83	paddw xmm1, xmm5
michael@0	84	%endmacro
michael@0	85
michael@0	86	%macro PROCESS_8X2X8 1
michael@0	87	%if %1
michael@0	88	movq xmm0, MMWORD PTR [rsi]
michael@0	89	movq xmm1, MMWORD PTR [rdi]
michael@0	90	movq xmm3, MMWORD PTR [rdi+8]
michael@0	91	punpcklqdq xmm1, xmm3
michael@0	92
michael@0	93	movdqa xmm2, xmm1
michael@0	94	mpsadbw xmm1, xmm0, 0x0
michael@0	95	mpsadbw xmm2, xmm0, 0x5
michael@0	96	paddw xmm1, xmm2
michael@0	97	%else
michael@0	98	movq xmm0, MMWORD PTR [rsi]
michael@0	99	movq xmm5, MMWORD PTR [rdi]
michael@0	100	movq xmm3, MMWORD PTR [rdi+8]
michael@0	101	punpcklqdq xmm5, xmm3
michael@0	102
michael@0	103	movdqa xmm2, xmm5
michael@0	104	mpsadbw xmm5, xmm0, 0x0
michael@0	105	mpsadbw xmm2, xmm0, 0x5
michael@0	106	paddw xmm5, xmm2
michael@0	107
michael@0	108	paddw xmm1, xmm5
michael@0	109	%endif
michael@0	110	movq xmm0, MMWORD PTR [rsi + rax]
michael@0	111	movq xmm5, MMWORD PTR [rdi+ rdx]
michael@0	112	movq xmm3, MMWORD PTR [rdi+ rdx+8]
michael@0	113	punpcklqdq xmm5, xmm3
michael@0	114
michael@0	115	lea rsi, [rsi+rax*2]
michael@0	116	lea rdi, [rdi+rdx*2]
michael@0	117
michael@0	118	movdqa xmm2, xmm5
michael@0	119	mpsadbw xmm5, xmm0, 0x0
michael@0	120	mpsadbw xmm2, xmm0, 0x5
michael@0	121	paddw xmm5, xmm2
michael@0	122
michael@0	123	paddw xmm1, xmm5
michael@0	124	%endmacro
michael@0	125
michael@0	126	%macro PROCESS_4X2X8 1
michael@0	127	%if %1
michael@0	128	movd xmm0, [rsi]
michael@0	129	movq xmm1, MMWORD PTR [rdi]
michael@0	130	movq xmm3, MMWORD PTR [rdi+8]
michael@0	131	punpcklqdq xmm1, xmm3
michael@0	132
michael@0	133	mpsadbw xmm1, xmm0, 0x0
michael@0	134	%else
michael@0	135	movd xmm0, [rsi]
michael@0	136	movq xmm5, MMWORD PTR [rdi]
michael@0	137	movq xmm3, MMWORD PTR [rdi+8]
michael@0	138	punpcklqdq xmm5, xmm3
michael@0	139
michael@0	140	mpsadbw xmm5, xmm0, 0x0
michael@0	141
michael@0	142	paddw xmm1, xmm5
michael@0	143	%endif
michael@0	144	movd xmm0, [rsi + rax]
michael@0	145	movq xmm5, MMWORD PTR [rdi+ rdx]
michael@0	146	movq xmm3, MMWORD PTR [rdi+ rdx+8]
michael@0	147	punpcklqdq xmm5, xmm3
michael@0	148
michael@0	149	lea rsi, [rsi+rax*2]
michael@0	150	lea rdi, [rdi+rdx*2]
michael@0	151
michael@0	152	mpsadbw xmm5, xmm0, 0x0
michael@0	153
michael@0	154	paddw xmm1, xmm5
michael@0	155	%endmacro
michael@0	156
michael@0	157
michael@0	158	;void vp8_sad16x16x8_sse4(
michael@0	159	; const unsigned char *src_ptr,
michael@0	160	; int src_stride,
michael@0	161	; const unsigned char *ref_ptr,
michael@0	162	; int ref_stride,
michael@0	163	; unsigned short *sad_array);
michael@0	164	global sym(vp8_sad16x16x8_sse4) PRIVATE
michael@0	165	sym(vp8_sad16x16x8_sse4):
michael@0	166	push rbp
michael@0	167	mov rbp, rsp
michael@0	168	SHADOW_ARGS_TO_STACK 5
michael@0	169	push rsi
michael@0	170	push rdi
michael@0	171	; end prolog
michael@0	172
michael@0	173	mov rsi, arg(0) ;src_ptr
michael@0	174	mov rdi, arg(2) ;ref_ptr
michael@0	175
michael@0	176	movsxd rax, dword ptr arg(1) ;src_stride
michael@0	177	movsxd rdx, dword ptr arg(3) ;ref_stride
michael@0	178
michael@0	179	PROCESS_16X2X8 1
michael@0	180	PROCESS_16X2X8 0
michael@0	181	PROCESS_16X2X8 0
michael@0	182	PROCESS_16X2X8 0
michael@0	183	PROCESS_16X2X8 0
michael@0	184	PROCESS_16X2X8 0
michael@0	185	PROCESS_16X2X8 0
michael@0	186	PROCESS_16X2X8 0
michael@0	187
michael@0	188	mov rdi, arg(4) ;Results
michael@0	189	movdqa XMMWORD PTR [rdi], xmm1
michael@0	190
michael@0	191	; begin epilog
michael@0	192	pop rdi
michael@0	193	pop rsi
michael@0	194	UNSHADOW_ARGS
michael@0	195	pop rbp
michael@0	196	ret
michael@0	197
michael@0	198
michael@0	199	;void vp8_sad16x8x8_sse4(
michael@0	200	; const unsigned char *src_ptr,
michael@0	201	; int src_stride,
michael@0	202	; const unsigned char *ref_ptr,
michael@0	203	; int ref_stride,
michael@0	204	; unsigned short *sad_array
michael@0	205	;);
michael@0	206	global sym(vp8_sad16x8x8_sse4) PRIVATE
michael@0	207	sym(vp8_sad16x8x8_sse4):
michael@0	208	push rbp
michael@0	209	mov rbp, rsp
michael@0	210	SHADOW_ARGS_TO_STACK 5
michael@0	211	push rsi
michael@0	212	push rdi
michael@0	213	; end prolog
michael@0	214
michael@0	215	mov rsi, arg(0) ;src_ptr
michael@0	216	mov rdi, arg(2) ;ref_ptr
michael@0	217
michael@0	218	movsxd rax, dword ptr arg(1) ;src_stride
michael@0	219	movsxd rdx, dword ptr arg(3) ;ref_stride
michael@0	220
michael@0	221	PROCESS_16X2X8 1
michael@0	222	PROCESS_16X2X8 0
michael@0	223	PROCESS_16X2X8 0
michael@0	224	PROCESS_16X2X8 0
michael@0	225
michael@0	226	mov rdi, arg(4) ;Results
michael@0	227	movdqa XMMWORD PTR [rdi], xmm1
michael@0	228
michael@0	229	; begin epilog
michael@0	230	pop rdi
michael@0	231	pop rsi
michael@0	232	UNSHADOW_ARGS
michael@0	233	pop rbp
michael@0	234	ret
michael@0	235
michael@0	236
michael@0	237	;void vp8_sad8x8x8_sse4(
michael@0	238	; const unsigned char *src_ptr,
michael@0	239	; int src_stride,
michael@0	240	; const unsigned char *ref_ptr,
michael@0	241	; int ref_stride,
michael@0	242	; unsigned short *sad_array
michael@0	243	;);
michael@0	244	global sym(vp8_sad8x8x8_sse4) PRIVATE
michael@0	245	sym(vp8_sad8x8x8_sse4):
michael@0	246	push rbp
michael@0	247	mov rbp, rsp
michael@0	248	SHADOW_ARGS_TO_STACK 5
michael@0	249	push rsi
michael@0	250	push rdi
michael@0	251	; end prolog
michael@0	252
michael@0	253	mov rsi, arg(0) ;src_ptr
michael@0	254	mov rdi, arg(2) ;ref_ptr
michael@0	255
michael@0	256	movsxd rax, dword ptr arg(1) ;src_stride
michael@0	257	movsxd rdx, dword ptr arg(3) ;ref_stride
michael@0	258
michael@0	259	PROCESS_8X2X8 1
michael@0	260	PROCESS_8X2X8 0
michael@0	261	PROCESS_8X2X8 0
michael@0	262	PROCESS_8X2X8 0
michael@0	263
michael@0	264	mov rdi, arg(4) ;Results
michael@0	265	movdqa XMMWORD PTR [rdi], xmm1
michael@0	266
michael@0	267	; begin epilog
michael@0	268	pop rdi
michael@0	269	pop rsi
michael@0	270	UNSHADOW_ARGS
michael@0	271	pop rbp
michael@0	272	ret
michael@0	273
michael@0	274
michael@0	275	;void vp8_sad8x16x8_sse4(
michael@0	276	; const unsigned char *src_ptr,
michael@0	277	; int src_stride,
michael@0	278	; const unsigned char *ref_ptr,
michael@0	279	; int ref_stride,
michael@0	280	; unsigned short *sad_array
michael@0	281	;);
michael@0	282	global sym(vp8_sad8x16x8_sse4) PRIVATE
michael@0	283	sym(vp8_sad8x16x8_sse4):
michael@0	284	push rbp
michael@0	285	mov rbp, rsp
michael@0	286	SHADOW_ARGS_TO_STACK 5
michael@0	287	push rsi
michael@0	288	push rdi
michael@0	289	; end prolog
michael@0	290
michael@0	291	mov rsi, arg(0) ;src_ptr
michael@0	292	mov rdi, arg(2) ;ref_ptr
michael@0	293
michael@0	294	movsxd rax, dword ptr arg(1) ;src_stride
michael@0	295	movsxd rdx, dword ptr arg(3) ;ref_stride
michael@0	296
michael@0	297	PROCESS_8X2X8 1
michael@0	298	PROCESS_8X2X8 0
michael@0	299	PROCESS_8X2X8 0
michael@0	300	PROCESS_8X2X8 0
michael@0	301	PROCESS_8X2X8 0
michael@0	302	PROCESS_8X2X8 0
michael@0	303	PROCESS_8X2X8 0
michael@0	304	PROCESS_8X2X8 0
michael@0	305	mov rdi, arg(4) ;Results
michael@0	306	movdqa XMMWORD PTR [rdi], xmm1
michael@0	307
michael@0	308	; begin epilog
michael@0	309	pop rdi
michael@0	310	pop rsi
michael@0	311	UNSHADOW_ARGS
michael@0	312	pop rbp
michael@0	313	ret
michael@0	314
michael@0	315
michael@0	316	;void vp8_sad4x4x8_c(
michael@0	317	; const unsigned char *src_ptr,
michael@0	318	; int src_stride,
michael@0	319	; const unsigned char *ref_ptr,
michael@0	320	; int ref_stride,
michael@0	321	; unsigned short *sad_array
michael@0	322	;);
michael@0	323	global sym(vp8_sad4x4x8_sse4) PRIVATE
michael@0	324	sym(vp8_sad4x4x8_sse4):
michael@0	325	push rbp
michael@0	326	mov rbp, rsp
michael@0	327	SHADOW_ARGS_TO_STACK 5
michael@0	328	push rsi
michael@0	329	push rdi
michael@0	330	; end prolog
michael@0	331
michael@0	332	mov rsi, arg(0) ;src_ptr
michael@0	333	mov rdi, arg(2) ;ref_ptr
michael@0	334
michael@0	335	movsxd rax, dword ptr arg(1) ;src_stride
michael@0	336	movsxd rdx, dword ptr arg(3) ;ref_stride
michael@0	337
michael@0	338	PROCESS_4X2X8 1
michael@0	339	PROCESS_4X2X8 0
michael@0	340
michael@0	341	mov rdi, arg(4) ;Results
michael@0	342	movdqa XMMWORD PTR [rdi], xmm1
michael@0	343
michael@0	344	; begin epilog
michael@0	345	pop rdi
michael@0	346	pop rsi
michael@0	347	UNSHADOW_ARGS
michael@0	348	pop rbp
michael@0	349	ret
michael@0	350
michael@0	351
michael@0	352
michael@0	353