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

src/os_cpu/linux_x86/vm/linux_x86_32.s@f03d0a26bf83 (annotated)

src/os_cpu/linux_x86/vm/linux_x86_32.s

Thu, 22 Apr 2010 13:23:15 -0700

author

jcoomes

date

Thu, 22 Apr 2010 13:23:15 -0700

changeset 1845

f03d0a26bf83

parent 435

a61af66fc99e

child 1907

c18cbe5936b8

permissions

-rw-r--r--

6888954: argument formatting for assert() and friends
Reviewed-by: kvn, twisti, apetrusenko, never, dcubed

duke@435	1	#
duke@435	2	# Copyright 2004-2007 Sun Microsystems, Inc. All Rights Reserved.
duke@435	3	# DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
duke@435	4	#
duke@435	5	# This code is free software; you can redistribute it and/or modify it
duke@435	6	# under the terms of the GNU General Public License version 2 only, as
duke@435	7	# published by the Free Software Foundation.
duke@435	8	#
duke@435	9	# This code is distributed in the hope that it will be useful, but WITHOUT
duke@435	10	# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
duke@435	11	# FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
duke@435	12	# version 2 for more details (a copy is included in the LICENSE file that
duke@435	13	# accompanied this code).
duke@435	14	#
duke@435	15	# You should have received a copy of the GNU General Public License version
duke@435	16	# 2 along with this work; if not, write to the Free Software Foundation,
duke@435	17	# Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
duke@435	18	#
duke@435	19	# Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
duke@435	20	# CA 95054 USA or visit www.sun.com if you need additional information or
duke@435	21	# have any questions.
duke@435	22	#
duke@435	23
duke@435	24
duke@435	25	# NOTE WELL! The _Copy functions are called directly
duke@435	26	# from server-compiler-generated code via CallLeafNoFP,
duke@435	27	# which means that they *must* either not use floating
duke@435	28	# point or use it in the same manner as does the server
duke@435	29	# compiler.
duke@435	30
duke@435	31	.globl _Copy_conjoint_bytes
duke@435	32	.globl _Copy_arrayof_conjoint_bytes
duke@435	33	.globl _Copy_conjoint_jshorts_atomic
duke@435	34	.globl _Copy_arrayof_conjoint_jshorts
duke@435	35	.globl _Copy_conjoint_jints_atomic
duke@435	36	.globl _Copy_arrayof_conjoint_jints
duke@435	37	.globl _Copy_conjoint_jlongs_atomic
duke@435	38	.globl _mmx_Copy_arrayof_conjoint_jshorts
duke@435	39
duke@435	40	.globl _Atomic_cmpxchg_long
duke@435	41
duke@435	42	.text
duke@435	43
duke@435	44	.globl SafeFetch32, Fetch32PFI, Fetch32Resume
duke@435	45	.globl SafeFetchN
duke@435	46	## TODO: avoid exposing Fetch32PFI and Fetch32Resume.
duke@435	47	## Instead, the signal handler would call a new SafeFetchTriage(FaultingEIP)
duke@435	48	## routine to vet the address. If the address is the faulting LD then
duke@435	49	## SafeFetchTriage() would return the resume-at EIP, otherwise null.
duke@435	50	.type SafeFetch32,@function
duke@435	51	.p2align 4,,15
duke@435	52	SafeFetch32:
duke@435	53	SafeFetchN:
duke@435	54	movl 0x8(%esp), %eax
duke@435	55	movl 0x4(%esp), %ecx
duke@435	56	Fetch32PFI:
duke@435	57	movl (%ecx), %eax
duke@435	58	Fetch32Resume:
duke@435	59	ret
duke@435	60
duke@435	61
duke@435	62	.globl SpinPause
duke@435	63	.type SpinPause,@function
duke@435	64	.p2align 4,,15
duke@435	65	SpinPause:
duke@435	66	rep
duke@435	67	nop
duke@435	68	movl $1, %eax
duke@435	69	ret
duke@435	70
duke@435	71	# Support for void Copy::conjoint_bytes(void* from,
duke@435	72	# void* to,
duke@435	73	# size_t count)
duke@435	74	.p2align 4,,15
duke@435	75	.type _Copy_conjoint_bytes,@function
duke@435	76	_Copy_conjoint_bytes:
duke@435	77	pushl %esi
duke@435	78	movl 4+12(%esp),%ecx # count
duke@435	79	pushl %edi
duke@435	80	movl 8+ 4(%esp),%esi # from
duke@435	81	movl 8+ 8(%esp),%edi # to
duke@435	82	cmpl %esi,%edi
duke@435	83	leal -1(%esi,%ecx),%eax # from + count - 1
duke@435	84	jbe cb_CopyRight
duke@435	85	cmpl %eax,%edi
duke@435	86	jbe cb_CopyLeft
duke@435	87	# copy from low to high
duke@435	88	cb_CopyRight:
duke@435	89	cmpl $3,%ecx
duke@435	90	jbe 5f # <= 3 bytes
duke@435	91	# align source address at dword address boundary
duke@435	92	movl %ecx,%eax # original count
duke@435	93	movl $4,%ecx
duke@435	94	subl %esi,%ecx
duke@435	95	andl $3,%ecx # prefix byte count
duke@435	96	jz 1f # no prefix
duke@435	97	subl %ecx,%eax # byte count less prefix
duke@435	98	# copy prefix
duke@435	99	subl %esi,%edi
duke@435	100	0: movb (%esi),%dl
duke@435	101	movb %dl,(%edi,%esi,1)
duke@435	102	addl $1,%esi
duke@435	103	subl $1,%ecx
duke@435	104	jnz 0b
duke@435	105	addl %esi,%edi
duke@435	106	1: movl %eax,%ecx # byte count less prefix
duke@435	107	shrl $2,%ecx # dword count
duke@435	108	jz 4f # no dwords to move
duke@435	109	cmpl $32,%ecx
duke@435	110	jbe 2f # <= 32 dwords
duke@435	111	# copy aligned dwords
duke@435	112	rep; smovl
duke@435	113	jmp 4f
duke@435	114	# copy aligned dwords
duke@435	115	2: subl %esi,%edi
duke@435	116	.p2align 4,,15
duke@435	117	3: movl (%esi),%edx
duke@435	118	movl %edx,(%edi,%esi,1)
duke@435	119	addl $4,%esi
duke@435	120	subl $1,%ecx
duke@435	121	jnz 3b
duke@435	122	addl %esi,%edi
duke@435	123	4: movl %eax,%ecx # byte count less prefix
duke@435	124	andl $3,%ecx # suffix byte count
duke@435	125	jz 7f # no suffix
duke@435	126	# copy suffix
duke@435	127	5: xorl %eax,%eax
duke@435	128	6: movb (%esi,%eax,1),%dl
duke@435	129	movb %dl,(%edi,%eax,1)
duke@435	130	addl $1,%eax
duke@435	131	subl $1,%ecx
duke@435	132	jnz 6b
duke@435	133	7: popl %edi
duke@435	134	popl %esi
duke@435	135	ret
duke@435	136	# copy from high to low
duke@435	137	cb_CopyLeft:
duke@435	138	std
duke@435	139	leal -4(%edi,%ecx),%edi # to + count - 4
duke@435	140	movl %eax,%esi # from + count - 1
duke@435	141	movl %ecx,%eax
duke@435	142	subl $3,%esi # from + count - 4
duke@435	143	cmpl $3,%ecx
duke@435	144	jbe 5f # <= 3 bytes
duke@435	145	1: shrl $2,%ecx # dword count
duke@435	146	jz 4f # no dwords to move
duke@435	147	cmpl $32,%ecx
duke@435	148	ja 3f # > 32 dwords
duke@435	149	# copy dwords, aligned or not
duke@435	150	subl %esi,%edi
duke@435	151	.p2align 4,,15
duke@435	152	2: movl (%esi),%edx
duke@435	153	movl %edx,(%edi,%esi,1)
duke@435	154	subl $4,%esi
duke@435	155	subl $1,%ecx
duke@435	156	jnz 2b
duke@435	157	addl %esi,%edi
duke@435	158	jmp 4f
duke@435	159	# copy dwords, aligned or not
duke@435	160	3: rep; smovl
duke@435	161	4: movl %eax,%ecx # byte count
duke@435	162	andl $3,%ecx # suffix byte count
duke@435	163	jz 7f # no suffix
duke@435	164	# copy suffix
duke@435	165	5: subl %esi,%edi
duke@435	166	addl $3,%esi
duke@435	167	6: movb (%esi),%dl
duke@435	168	movb %dl,(%edi,%esi,1)
duke@435	169	subl $1,%esi
duke@435	170	subl $1,%ecx
duke@435	171	jnz 6b
duke@435	172	7: cld
duke@435	173	popl %edi
duke@435	174	popl %esi
duke@435	175	ret
duke@435	176
duke@435	177	# Support for void Copy::arrayof_conjoint_bytes(void* from,
duke@435	178	# void* to,
duke@435	179	# size_t count)
duke@435	180	#
duke@435	181	# Same as _Copy_conjoint_bytes, except no source alignment check.
duke@435	182	.p2align 4,,15
duke@435	183	.type _Copy_arrayof_conjoint_bytes,@function
duke@435	184	_Copy_arrayof_conjoint_bytes:
duke@435	185	pushl %esi
duke@435	186	movl 4+12(%esp),%ecx # count
duke@435	187	pushl %edi
duke@435	188	movl 8+ 4(%esp),%esi # from
duke@435	189	movl 8+ 8(%esp),%edi # to
duke@435	190	cmpl %esi,%edi
duke@435	191	leal -1(%esi,%ecx),%eax # from + count - 1
duke@435	192	jbe acb_CopyRight
duke@435	193	cmpl %eax,%edi
duke@435	194	jbe acb_CopyLeft
duke@435	195	# copy from low to high
duke@435	196	acb_CopyRight:
duke@435	197	cmpl $3,%ecx
duke@435	198	jbe 5f
duke@435	199	1: movl %ecx,%eax
duke@435	200	shrl $2,%ecx
duke@435	201	jz 4f
duke@435	202	cmpl $32,%ecx
duke@435	203	ja 3f
duke@435	204	# copy aligned dwords
duke@435	205	subl %esi,%edi
duke@435	206	.p2align 4,,15
duke@435	207	2: movl (%esi),%edx
duke@435	208	movl %edx,(%edi,%esi,1)
duke@435	209	addl $4,%esi
duke@435	210	subl $1,%ecx
duke@435	211	jnz 2b
duke@435	212	addl %esi,%edi
duke@435	213	jmp 4f
duke@435	214	# copy aligned dwords
duke@435	215	3: rep; smovl
duke@435	216	4: movl %eax,%ecx
duke@435	217	andl $3,%ecx
duke@435	218	jz 7f
duke@435	219	# copy suffix
duke@435	220	5: xorl %eax,%eax
duke@435	221	6: movb (%esi,%eax,1),%dl
duke@435	222	movb %dl,(%edi,%eax,1)
duke@435	223	addl $1,%eax
duke@435	224	subl $1,%ecx
duke@435	225	jnz 6b
duke@435	226	7: popl %edi
duke@435	227	popl %esi
duke@435	228	ret
duke@435	229	acb_CopyLeft:
duke@435	230	std
duke@435	231	leal -4(%edi,%ecx),%edi # to + count - 4
duke@435	232	movl %eax,%esi # from + count - 1
duke@435	233	movl %ecx,%eax
duke@435	234	subl $3,%esi # from + count - 4
duke@435	235	cmpl $3,%ecx
duke@435	236	jbe 5f
duke@435	237	1: shrl $2,%ecx
duke@435	238	jz 4f
duke@435	239	cmpl $32,%ecx
duke@435	240	jbe 2f # <= 32 dwords
duke@435	241	rep; smovl
duke@435	242	jmp 4f
duke@435	243	.=.+8
duke@435	244	2: subl %esi,%edi
duke@435	245	.p2align 4,,15
duke@435	246	3: movl (%esi),%edx
duke@435	247	movl %edx,(%edi,%esi,1)
duke@435	248	subl $4,%esi
duke@435	249	subl $1,%ecx
duke@435	250	jnz 3b
duke@435	251	addl %esi,%edi
duke@435	252	4: movl %eax,%ecx
duke@435	253	andl $3,%ecx
duke@435	254	jz 7f
duke@435	255	5: subl %esi,%edi
duke@435	256	addl $3,%esi
duke@435	257	6: movb (%esi),%dl
duke@435	258	movb %dl,(%edi,%esi,1)
duke@435	259	subl $1,%esi
duke@435	260	subl $1,%ecx
duke@435	261	jnz 6b
duke@435	262	7: cld
duke@435	263	popl %edi
duke@435	264	popl %esi
duke@435	265	ret
duke@435	266
duke@435	267	# Support for void Copy::conjoint_jshorts_atomic(void* from,
duke@435	268	# void* to,
duke@435	269	# size_t count)
duke@435	270	.p2align 4,,15
duke@435	271	.type _Copy_conjoint_jshorts_atomic,@function
duke@435	272	_Copy_conjoint_jshorts_atomic:
duke@435	273	pushl %esi
duke@435	274	movl 4+12(%esp),%ecx # count
duke@435	275	pushl %edi
duke@435	276	movl 8+ 4(%esp),%esi # from
duke@435	277	movl 8+ 8(%esp),%edi # to
duke@435	278	cmpl %esi,%edi
duke@435	279	leal -2(%esi,%ecx,2),%eax # from + count*2 - 2
duke@435	280	jbe cs_CopyRight
duke@435	281	cmpl %eax,%edi
duke@435	282	jbe cs_CopyLeft
duke@435	283	# copy from low to high
duke@435	284	cs_CopyRight:
duke@435	285	# align source address at dword address boundary
duke@435	286	movl %esi,%eax # original from
duke@435	287	andl $3,%eax # either 0 or 2
duke@435	288	jz 1f # no prefix
duke@435	289	# copy prefix
duke@435	290	movw (%esi),%dx
duke@435	291	movw %dx,(%edi)
duke@435	292	addl %eax,%esi # %eax == 2
duke@435	293	addl %eax,%edi
duke@435	294	subl $1,%ecx
duke@435	295	1: movl %ecx,%eax # word count less prefix
duke@435	296	sarl %ecx # dword count
duke@435	297	jz 4f # no dwords to move
duke@435	298	cmpl $32,%ecx
duke@435	299	jbe 2f # <= 32 dwords
duke@435	300	# copy aligned dwords
duke@435	301	rep; smovl
duke@435	302	jmp 4f
duke@435	303	# copy aligned dwords
duke@435	304	2: subl %esi,%edi
duke@435	305	.p2align 4,,15
duke@435	306	3: movl (%esi),%edx
duke@435	307	movl %edx,(%edi,%esi,1)
duke@435	308	addl $4,%esi
duke@435	309	subl $1,%ecx
duke@435	310	jnz 3b
duke@435	311	addl %esi,%edi
duke@435	312	4: andl $1,%eax # suffix count
duke@435	313	jz 5f # no suffix
duke@435	314	# copy suffix
duke@435	315	movw (%esi),%dx
duke@435	316	movw %dx,(%edi)
duke@435	317	5: popl %edi
duke@435	318	popl %esi
duke@435	319	ret
duke@435	320	# copy from high to low
duke@435	321	cs_CopyLeft:
duke@435	322	std
duke@435	323	leal -4(%edi,%ecx,2),%edi # to + count*2 - 4
duke@435	324	movl %eax,%esi # from + count*2 - 2
duke@435	325	movl %ecx,%eax
duke@435	326	subl $2,%esi # from + count*2 - 4
duke@435	327	1: sarl %ecx # dword count
duke@435	328	jz 4f # no dwords to move
duke@435	329	cmpl $32,%ecx
duke@435	330	ja 3f # > 32 dwords
duke@435	331	subl %esi,%edi
duke@435	332	.p2align 4,,15
duke@435	333	2: movl (%esi),%edx
duke@435	334	movl %edx,(%edi,%esi,1)
duke@435	335	subl $4,%esi
duke@435	336	subl $1,%ecx
duke@435	337	jnz 2b
duke@435	338	addl %esi,%edi
duke@435	339	jmp 4f
duke@435	340	3: rep; smovl
duke@435	341	4: andl $1,%eax # suffix count
duke@435	342	jz 5f # no suffix
duke@435	343	# copy suffix
duke@435	344	addl $2,%esi
duke@435	345	addl $2,%edi
duke@435	346	movw (%esi),%dx
duke@435	347	movw %dx,(%edi)
duke@435	348	5: cld
duke@435	349	popl %edi
duke@435	350	popl %esi
duke@435	351	ret
duke@435	352
duke@435	353	# Support for void Copy::arrayof_conjoint_jshorts(void* from,
duke@435	354	# void* to,
duke@435	355	# size_t count)
duke@435	356	.p2align 4,,15
duke@435	357	.type _Copy_arrayof_conjoint_jshorts,@function
duke@435	358	_Copy_arrayof_conjoint_jshorts:
duke@435	359	pushl %esi
duke@435	360	movl 4+12(%esp),%ecx # count
duke@435	361	pushl %edi
duke@435	362	movl 8+ 4(%esp),%esi # from
duke@435	363	movl 8+ 8(%esp),%edi # to
duke@435	364	cmpl %esi,%edi
duke@435	365	leal -2(%esi,%ecx,2),%eax # from + count*2 - 2
duke@435	366	jbe acs_CopyRight
duke@435	367	cmpl %eax,%edi
duke@435	368	jbe acs_CopyLeft
duke@435	369	acs_CopyRight:
duke@435	370	movl %ecx,%eax # word count
duke@435	371	sarl %ecx # dword count
duke@435	372	jz 4f # no dwords to move
duke@435	373	cmpl $32,%ecx
duke@435	374	jbe 2f # <= 32 dwords
duke@435	375	# copy aligned dwords
duke@435	376	rep; smovl
duke@435	377	jmp 4f
duke@435	378	# copy aligned dwords
duke@435	379	.=.+5
duke@435	380	2: subl %esi,%edi
duke@435	381	.p2align 4,,15
duke@435	382	3: movl (%esi),%edx
duke@435	383	movl %edx,(%edi,%esi,1)
duke@435	384	addl $4,%esi
duke@435	385	subl $1,%ecx
duke@435	386	jnz 3b
duke@435	387	addl %esi,%edi
duke@435	388	4: andl $1,%eax # suffix count
duke@435	389	jz 5f # no suffix
duke@435	390	# copy suffix
duke@435	391	movw (%esi),%dx
duke@435	392	movw %dx,(%edi)
duke@435	393	5: popl %edi
duke@435	394	popl %esi
duke@435	395	ret
duke@435	396	acs_CopyLeft:
duke@435	397	std
duke@435	398	leal -4(%edi,%ecx,2),%edi # to + count*2 - 4
duke@435	399	movl %eax,%esi # from + count*2 - 2
duke@435	400	movl %ecx,%eax
duke@435	401	subl $2,%esi # from + count*2 - 4
duke@435	402	sarl %ecx # dword count
duke@435	403	jz 4f # no dwords to move
duke@435	404	cmpl $32,%ecx
duke@435	405	ja 3f # > 32 dwords
duke@435	406	subl %esi,%edi
duke@435	407	.p2align 4,,15
duke@435	408	2: movl (%esi),%edx
duke@435	409	movl %edx,(%edi,%esi,1)
duke@435	410	subl $4,%esi
duke@435	411	subl $1,%ecx
duke@435	412	jnz 2b
duke@435	413	addl %esi,%edi
duke@435	414	jmp 4f
duke@435	415	3: rep; smovl
duke@435	416	4: andl $1,%eax # suffix count
duke@435	417	jz 5f # no suffix
duke@435	418	# copy suffix
duke@435	419	addl $2,%esi
duke@435	420	addl $2,%edi
duke@435	421	movw (%esi),%dx
duke@435	422	movw %dx,(%edi)
duke@435	423	5: cld
duke@435	424	popl %edi
duke@435	425	popl %esi
duke@435	426	ret
duke@435	427
duke@435	428	# Support for void Copy::conjoint_jints_atomic(void* from,
duke@435	429	# void* to,
duke@435	430	# size_t count)
duke@435	431	# Equivalent to
duke@435	432	# arrayof_conjoint_jints
duke@435	433	.p2align 4,,15
duke@435	434	.type _Copy_conjoint_jints_atomic,@function
duke@435	435	.type _Copy_arrayof_conjoint_jints,@function
duke@435	436	_Copy_conjoint_jints_atomic:
duke@435	437	_Copy_arrayof_conjoint_jints:
duke@435	438	pushl %esi
duke@435	439	movl 4+12(%esp),%ecx # count
duke@435	440	pushl %edi
duke@435	441	movl 8+ 4(%esp),%esi # from
duke@435	442	movl 8+ 8(%esp),%edi # to
duke@435	443	cmpl %esi,%edi
duke@435	444	leal -4(%esi,%ecx,4),%eax # from + count*4 - 4
duke@435	445	jbe ci_CopyRight
duke@435	446	cmpl %eax,%edi
duke@435	447	jbe ci_CopyLeft
duke@435	448	ci_CopyRight:
duke@435	449	cmpl $32,%ecx
duke@435	450	jbe 2f # <= 32 dwords
duke@435	451	rep; smovl
duke@435	452	popl %edi
duke@435	453	popl %esi
duke@435	454	ret
duke@435	455	.=.+10
duke@435	456	2: subl %esi,%edi
duke@435	457	.p2align 4,,15
duke@435	458	3: movl (%esi),%edx
duke@435	459	movl %edx,(%edi,%esi,1)
duke@435	460	addl $4,%esi
duke@435	461	subl $1,%ecx
duke@435	462	jnz 3b
duke@435	463	popl %edi
duke@435	464	popl %esi
duke@435	465	ret
duke@435	466	ci_CopyLeft:
duke@435	467	std
duke@435	468	leal -4(%edi,%ecx,4),%edi # to + count*4 - 4
duke@435	469	cmpl $32,%ecx
duke@435	470	ja 3f # > 32 dwords
duke@435	471	subl %eax,%edi # eax == from + count*4 - 4
duke@435	472	.p2align 4,,15
duke@435	473	2: movl (%eax),%edx
duke@435	474	movl %edx,(%edi,%eax,1)
duke@435	475	subl $4,%eax
duke@435	476	subl $1,%ecx
duke@435	477	jnz 2b
duke@435	478	cld
duke@435	479	popl %edi
duke@435	480	popl %esi
duke@435	481	ret
duke@435	482	3: movl %eax,%esi # from + count*4 - 4
duke@435	483	rep; smovl
duke@435	484	cld
duke@435	485	popl %edi
duke@435	486	popl %esi
duke@435	487	ret
duke@435	488
duke@435	489	# Support for void Copy::conjoint_jlongs_atomic(jlong* from,
duke@435	490	# jlong* to,
duke@435	491	# size_t count)
duke@435	492	#
duke@435	493	# 32-bit
duke@435	494	#
duke@435	495	# count treated as signed
duke@435	496	#
duke@435	497	# if (from > to) {
duke@435	498	# while (--count >= 0) {
duke@435	499	# *to++ = *from++;
duke@435	500	# }
duke@435	501	# } else {
duke@435	502	# while (--count >= 0) {
duke@435	503	# to[count] = from[count];
duke@435	504	# }
duke@435	505	# }
duke@435	506	.p2align 4,,15
duke@435	507	.type _Copy_conjoint_jlongs_atomic,@function
duke@435	508	_Copy_conjoint_jlongs_atomic:
duke@435	509	movl 4+8(%esp),%ecx # count
duke@435	510	movl 4+0(%esp),%eax # from
duke@435	511	movl 4+4(%esp),%edx # to
duke@435	512	cmpl %eax,%edx
duke@435	513	jae cla_CopyLeft
duke@435	514	cla_CopyRight:
duke@435	515	subl %eax,%edx
duke@435	516	jmp 2f
duke@435	517	.p2align 4,,15
duke@435	518	1: fildll (%eax)
duke@435	519	fistpll (%edx,%eax,1)
duke@435	520	addl $8,%eax
duke@435	521	2: subl $1,%ecx
duke@435	522	jge 1b
duke@435	523	ret
duke@435	524	.p2align 4,,15
duke@435	525	3: fildll (%eax,%ecx,8)
duke@435	526	fistpll (%edx,%ecx,8)
duke@435	527	cla_CopyLeft:
duke@435	528	subl $1,%ecx
duke@435	529	jge 3b
duke@435	530	ret
duke@435	531
duke@435	532	# Support for void Copy::arrayof_conjoint_jshorts(void* from,
duke@435	533	# void* to,
duke@435	534	# size_t count)
duke@435	535	.p2align 4,,15
duke@435	536	.type _mmx_Copy_arrayof_conjoint_jshorts,@function
duke@435	537	_mmx_Copy_arrayof_conjoint_jshorts:
duke@435	538	pushl %esi
duke@435	539	movl 4+12(%esp),%ecx
duke@435	540	pushl %edi
duke@435	541	movl 8+ 4(%esp),%esi
duke@435	542	movl 8+ 8(%esp),%edi
duke@435	543	cmpl %esi,%edi
duke@435	544	leal -2(%esi,%ecx,2),%eax
duke@435	545	jbe mmx_acs_CopyRight
duke@435	546	cmpl %eax,%edi
duke@435	547	jbe mmx_acs_CopyLeft
duke@435	548	mmx_acs_CopyRight:
duke@435	549	movl %ecx,%eax
duke@435	550	sarl %ecx
duke@435	551	je 5f
duke@435	552	cmpl $33,%ecx
duke@435	553	jae 3f
duke@435	554	1: subl %esi,%edi
duke@435	555	.p2align 4,,15
duke@435	556	2: movl (%esi),%edx
duke@435	557	movl %edx,(%edi,%esi,1)
duke@435	558	addl $4,%esi
duke@435	559	subl $1,%ecx
duke@435	560	jnz 2b
duke@435	561	addl %esi,%edi
duke@435	562	jmp 5f
duke@435	563	3: smovl # align to 8 bytes, we know we are 4 byte aligned to start
duke@435	564	subl $1,%ecx
duke@435	565	4: .p2align 4,,15
duke@435	566	movq 0(%esi),%mm0
duke@435	567	addl $64,%edi
duke@435	568	movq 8(%esi),%mm1
duke@435	569	subl $16,%ecx
duke@435	570	movq 16(%esi),%mm2
duke@435	571	movq %mm0,-64(%edi)
duke@435	572	movq 24(%esi),%mm0
duke@435	573	movq %mm1,-56(%edi)
duke@435	574	movq 32(%esi),%mm1
duke@435	575	movq %mm2,-48(%edi)
duke@435	576	movq 40(%esi),%mm2
duke@435	577	movq %mm0,-40(%edi)
duke@435	578	movq 48(%esi),%mm0
duke@435	579	movq %mm1,-32(%edi)
duke@435	580	movq 56(%esi),%mm1
duke@435	581	movq %mm2,-24(%edi)
duke@435	582	movq %mm0,-16(%edi)
duke@435	583	addl $64,%esi
duke@435	584	movq %mm1,-8(%edi)
duke@435	585	cmpl $16,%ecx
duke@435	586	jge 4b
duke@435	587	emms
duke@435	588	testl %ecx,%ecx
duke@435	589	ja 1b
duke@435	590	5: andl $1,%eax
duke@435	591	je 7f
duke@435	592	6: movw (%esi),%dx
duke@435	593	movw %dx,(%edi)
duke@435	594	7: popl %edi
duke@435	595	popl %esi
duke@435	596	ret
duke@435	597	mmx_acs_CopyLeft:
duke@435	598	std
duke@435	599	leal -4(%edi,%ecx,2),%edi
duke@435	600	movl %eax,%esi
duke@435	601	movl %ecx,%eax
duke@435	602	subl $2,%esi
duke@435	603	sarl %ecx
duke@435	604	je 4f
duke@435	605	cmpl $32,%ecx
duke@435	606	ja 3f
duke@435	607	subl %esi,%edi
duke@435	608	.p2align 4,,15
duke@435	609	2: movl (%esi),%edx
duke@435	610	movl %edx,(%edi,%esi,1)
duke@435	611	subl $4,%esi
duke@435	612	subl $1,%ecx
duke@435	613	jnz 2b
duke@435	614	addl %esi,%edi
duke@435	615	jmp 4f
duke@435	616	3: rep; smovl
duke@435	617	4: andl $1,%eax
duke@435	618	je 6f
duke@435	619	addl $2,%esi
duke@435	620	addl $2,%edi
duke@435	621	5: movw (%esi),%dx
duke@435	622	movw %dx,(%edi)
duke@435	623	6: cld
duke@435	624	popl %edi
duke@435	625	popl %esi
duke@435	626	ret
duke@435	627
duke@435	628
duke@435	629	# Support for jlong Atomic::cmpxchg(jlong exchange_value,
duke@435	630	# volatile jlong* dest,
duke@435	631	# jlong compare_value,
duke@435	632	# bool is_MP)
duke@435	633	#
duke@435	634	.p2align 4,,15
duke@435	635	.type _Atomic_cmpxchg_long,@function
duke@435	636	_Atomic_cmpxchg_long:
duke@435	637	# 8(%esp) : return PC
duke@435	638	pushl %ebx # 4(%esp) : old %ebx
duke@435	639	pushl %edi # 0(%esp) : old %edi
duke@435	640	movl 12(%esp), %ebx # 12(%esp) : exchange_value (low)
duke@435	641	movl 16(%esp), %ecx # 16(%esp) : exchange_value (high)
duke@435	642	movl 24(%esp), %eax # 24(%esp) : compare_value (low)
duke@435	643	movl 28(%esp), %edx # 28(%esp) : compare_value (high)
duke@435	644	movl 20(%esp), %edi # 20(%esp) : dest
duke@435	645	cmpl $0, 32(%esp) # 32(%esp) : is_MP
duke@435	646	je 1f
duke@435	647	lock
duke@435	648	1: cmpxchg8b (%edi)
duke@435	649	popl %edi
duke@435	650	popl %ebx
duke@435	651	ret
duke@435	652