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

src/os_cpu/solaris_x86/vm/solaris_x86_32.s@1e9a9d2e6509 (annotated)

src/os_cpu/solaris_x86/vm/solaris_x86_32.s

Thu, 21 Oct 2010 11:55:10 -0700

author

never

date

Thu, 21 Oct 2010 11:55:10 -0700

changeset 2262

1e9a9d2e6509

parent 1958

d93949c5bdcc

child 2384

0a8e0d4345b3

permissions

-rw-r--r--

6970683: improvements to hs_err output
Reviewed-by: kvn, jrose, dholmes, coleenp

duke@435	1	//
trims@1907	2	// Copyright (c) 2004, 2007, Oracle and/or its affiliates. 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	//
trims@1907	19	// Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
trims@1907	20	// or visit www.oracle.com if you need additional information or have any
trims@1907	21	// questions.
duke@435	22	//
duke@435	23
duke@435	24	.globl fixcw
duke@435	25	.globl sse_check
duke@435	26	.globl sse_unavailable
duke@435	27	.globl gs_load
duke@435	28	.globl gs_thread
duke@435	29	.globl _Atomic_cmpxchg_long_gcc
duke@435	30
duke@435	31	// NOTE WELL! The _Copy functions are called directly
duke@435	32	// from server-compiler-generated code via CallLeafNoFP,
duke@435	33	// which means that they *must* either not use floating
duke@435	34	// point or use it in the same manner as does the server
duke@435	35	// compiler.
duke@435	36
duke@435	37	.globl _Copy_conjoint_bytes
duke@435	38	.globl _Copy_arrayof_conjoint_bytes
duke@435	39	.globl _Copy_conjoint_jshorts_atomic
duke@435	40	.globl _Copy_arrayof_conjoint_jshorts
duke@435	41	.globl _Copy_conjoint_jints_atomic
duke@435	42	.globl _Copy_arrayof_conjoint_jints
duke@435	43	.globl _Copy_conjoint_jlongs_atomic
duke@435	44	.globl _mmx_Copy_arrayof_conjoint_jshorts
duke@435	45
duke@435	46	.section .text,"ax"
duke@435	47
duke@435	48	/ Support for void os::Solaris::init_thread_fpu_state() in os_solaris_i486.cpp
duke@435	49	/ Set fpu to 53 bit precision. This happens too early to use a stub.
duke@435	50	.align 16
duke@435	51	fixcw:
duke@435	52	pushl $0x27f
duke@435	53	fldcw 0(%esp)
duke@435	54	popl %eax
duke@435	55	ret
duke@435	56
duke@435	57	.align 16
duke@435	58	.globl SafeFetch32
duke@435	59	.globl SafeFetchN
duke@435	60	.globl Fetch32PFI, Fetch32Resume
duke@435	61	SafeFetch32:
duke@435	62	SafeFetchN:
duke@435	63	movl 0x8(%esp), %eax
duke@435	64	movl 0x4(%esp), %ecx
duke@435	65	Fetch32PFI:
duke@435	66	movl (%ecx), %eax
duke@435	67	Fetch32Resume:
duke@435	68	ret
duke@435	69
duke@435	70
duke@435	71	.align 16
duke@435	72	.globl SpinPause
duke@435	73	SpinPause:
duke@435	74	rep
duke@435	75	nop
duke@435	76	movl $1, %eax
duke@435	77	ret
duke@435	78
duke@435	79
duke@435	80	/ Test SSE availability, used by os_solaris_i486.cpp
duke@435	81	.align 16
duke@435	82	sse_check:
duke@435	83	/ Fault if SSE not available
duke@435	84	xorps %xmm0,%xmm0
duke@435	85	/ No fault
duke@435	86	movl $1,%eax
duke@435	87	ret
duke@435	88	/ Signal handler continues here if SSE is not available
duke@435	89	sse_unavailable:
duke@435	90	xorl %eax,%eax
duke@435	91	ret
duke@435	92
duke@435	93	/ Fast thread accessors, used by threadLS_solaris_i486.cpp
duke@435	94	.align 16
duke@435	95	gs_load:
duke@435	96	movl 4(%esp),%ecx
duke@435	97	movl %gs:(%ecx),%eax
duke@435	98	ret
duke@435	99
duke@435	100	.align 16
duke@435	101	gs_thread:
duke@435	102	movl %gs:0x0,%eax
duke@435	103	ret
duke@435	104
duke@435	105	/ Support for void Copy::conjoint_bytes(void* from,
duke@435	106	/ void* to,
duke@435	107	/ size_t count)
duke@435	108	.align 16
duke@435	109	_Copy_conjoint_bytes:
duke@435	110	pushl %esi
duke@435	111	movl 4+12(%esp),%ecx / count
duke@435	112	pushl %edi
duke@435	113	movl 8+ 4(%esp),%esi / from
duke@435	114	movl 8+ 8(%esp),%edi / to
duke@435	115	cmpl %esi,%edi
duke@435	116	leal -1(%esi,%ecx),%eax / from + count - 1
duke@435	117	jbe cb_CopyRight
duke@435	118	cmpl %eax,%edi
duke@435	119	jbe cb_CopyLeft
duke@435	120	/ copy from low to high
duke@435	121	cb_CopyRight:
duke@435	122	cmpl $3,%ecx
duke@435	123	jbe 5f / <= 3 bytes
duke@435	124	/ align source address at dword address boundary
duke@435	125	movl %ecx,%eax / original count
duke@435	126	movl $4,%ecx
duke@435	127	subl %esi,%ecx
duke@435	128	andl $3,%ecx / prefix byte count
duke@435	129	jz 1f / no prefix
duke@435	130	subl %ecx,%eax / byte count less prefix
duke@435	131	/ copy prefix
duke@435	132	subl %esi,%edi
duke@435	133	0: movb (%esi),%dl
duke@435	134	movb %dl,(%edi,%esi,1)
duke@435	135	addl $1,%esi
duke@435	136	subl $1,%ecx
duke@435	137	jnz 0b
duke@435	138	addl %esi,%edi
duke@435	139	1: movl %eax,%ecx / byte count less prefix
duke@435	140	shrl $2,%ecx / dword count
duke@435	141	jz 4f / no dwords to move
duke@435	142	cmpl $32,%ecx
duke@435	143	jbe 2f / <= 32 dwords
duke@435	144	/ copy aligned dwords
duke@435	145	rep; smovl
duke@435	146	jmp 4f
duke@435	147	/ copy aligned dwords
duke@435	148	2: subl %esi,%edi
duke@435	149	.align 16
duke@435	150	3: movl (%esi),%edx
duke@435	151	movl %edx,(%edi,%esi,1)
duke@435	152	addl $4,%esi
duke@435	153	subl $1,%ecx
duke@435	154	jnz 3b
duke@435	155	addl %esi,%edi
duke@435	156	4: movl %eax,%ecx / byte count less prefix
kvn@1958	157	5: andl $3,%ecx / suffix byte count
duke@435	158	jz 7f / no suffix
duke@435	159	/ copy suffix
kvn@1958	160	xorl %eax,%eax
duke@435	161	6: movb (%esi,%eax,1),%dl
duke@435	162	movb %dl,(%edi,%eax,1)
duke@435	163	addl $1,%eax
duke@435	164	subl $1,%ecx
duke@435	165	jnz 6b
duke@435	166	7: popl %edi
duke@435	167	popl %esi
duke@435	168	ret
duke@435	169	/ copy from high to low
duke@435	170	cb_CopyLeft:
duke@435	171	std
duke@435	172	leal -4(%edi,%ecx),%edi / to + count - 4
duke@435	173	movl %eax,%esi / from + count - 1
duke@435	174	movl %ecx,%eax
duke@435	175	subl $3,%esi / from + count - 4
duke@435	176	cmpl $3,%ecx
duke@435	177	jbe 5f / <= 3 bytes
duke@435	178	1: shrl $2,%ecx / dword count
duke@435	179	jz 4f / no dwords to move
duke@435	180	cmpl $32,%ecx
duke@435	181	ja 3f / > 32 dwords
duke@435	182	/ copy dwords, aligned or not
duke@435	183	subl %esi,%edi
duke@435	184	.align 16
duke@435	185	2: movl (%esi),%edx
duke@435	186	movl %edx,(%edi,%esi,1)
duke@435	187	subl $4,%esi
duke@435	188	subl $1,%ecx
duke@435	189	jnz 2b
duke@435	190	addl %esi,%edi
duke@435	191	jmp 4f
duke@435	192	/ copy dwords, aligned or not
duke@435	193	3: rep; smovl
duke@435	194	4: movl %eax,%ecx / byte count
kvn@1958	195	5: andl $3,%ecx / suffix byte count
duke@435	196	jz 7f / no suffix
duke@435	197	/ copy suffix
kvn@1958	198	subl %esi,%edi
duke@435	199	addl $3,%esi
duke@435	200	6: movb (%esi),%dl
duke@435	201	movb %dl,(%edi,%esi,1)
duke@435	202	subl $1,%esi
duke@435	203	subl $1,%ecx
duke@435	204	jnz 6b
duke@435	205	7: cld
duke@435	206	popl %edi
duke@435	207	popl %esi
duke@435	208	ret
duke@435	209
duke@435	210	/ Support for void Copy::arrayof_conjoint_bytes(void* from,
duke@435	211	/ void* to,
duke@435	212	/ size_t count)
duke@435	213	/
duke@435	214	/ Same as _Copy_conjoint_bytes, except no source alignment check.
duke@435	215	.align 16
duke@435	216	_Copy_arrayof_conjoint_bytes:
duke@435	217	pushl %esi
duke@435	218	movl 4+12(%esp),%ecx / count
duke@435	219	pushl %edi
duke@435	220	movl 8+ 4(%esp),%esi / from
duke@435	221	movl 8+ 8(%esp),%edi / to
duke@435	222	cmpl %esi,%edi
duke@435	223	leal -1(%esi,%ecx),%eax / from + count - 1
duke@435	224	jbe acb_CopyRight
duke@435	225	cmpl %eax,%edi
duke@435	226	jbe acb_CopyLeft
duke@435	227	/ copy from low to high
duke@435	228	acb_CopyRight:
duke@435	229	cmpl $3,%ecx
duke@435	230	jbe 5f
duke@435	231	1: movl %ecx,%eax
duke@435	232	shrl $2,%ecx
duke@435	233	jz 4f
duke@435	234	cmpl $32,%ecx
duke@435	235	ja 3f
duke@435	236	/ copy aligned dwords
duke@435	237	subl %esi,%edi
duke@435	238	.align 16
duke@435	239	2: movl (%esi),%edx
duke@435	240	movl %edx,(%edi,%esi,1)
duke@435	241	addl $4,%esi
duke@435	242	subl $1,%ecx
duke@435	243	jnz 2b
duke@435	244	addl %esi,%edi
duke@435	245	jmp 4f
duke@435	246	/ copy aligned dwords
duke@435	247	3: rep; smovl
duke@435	248	4: movl %eax,%ecx
kvn@1958	249	5: andl $3,%ecx
duke@435	250	jz 7f
duke@435	251	/ copy suffix
kvn@1958	252	xorl %eax,%eax
duke@435	253	6: movb (%esi,%eax,1),%dl
duke@435	254	movb %dl,(%edi,%eax,1)
duke@435	255	addl $1,%eax
duke@435	256	subl $1,%ecx
duke@435	257	jnz 6b
duke@435	258	7: popl %edi
duke@435	259	popl %esi
duke@435	260	ret
duke@435	261	acb_CopyLeft:
duke@435	262	std
duke@435	263	leal -4(%edi,%ecx),%edi / to + count - 4
duke@435	264	movl %eax,%esi / from + count - 1
duke@435	265	movl %ecx,%eax
duke@435	266	subl $3,%esi / from + count - 4
duke@435	267	cmpl $3,%ecx
duke@435	268	jbe 5f
duke@435	269	1: shrl $2,%ecx
duke@435	270	jz 4f
duke@435	271	cmpl $32,%ecx
duke@435	272	jbe 2f / <= 32 dwords
duke@435	273	rep; smovl
duke@435	274	jmp 4f
duke@435	275	.=.+8
duke@435	276	2: subl %esi,%edi
duke@435	277	.align 16
duke@435	278	3: movl (%esi),%edx
duke@435	279	movl %edx,(%edi,%esi,1)
duke@435	280	subl $4,%esi
duke@435	281	subl $1,%ecx
duke@435	282	jnz 3b
duke@435	283	addl %esi,%edi
duke@435	284	4: movl %eax,%ecx
kvn@1958	285	5: andl $3,%ecx
duke@435	286	jz 7f
kvn@1958	287	subl %esi,%edi
duke@435	288	addl $3,%esi
duke@435	289	6: movb (%esi),%dl
duke@435	290	movb %dl,(%edi,%esi,1)
duke@435	291	subl $1,%esi
duke@435	292	subl $1,%ecx
duke@435	293	jnz 6b
duke@435	294	7: cld
duke@435	295	popl %edi
duke@435	296	popl %esi
duke@435	297	ret
duke@435	298
duke@435	299	/ Support for void Copy::conjoint_jshorts_atomic(void* from,
duke@435	300	/ void* to,
duke@435	301	/ size_t count)
duke@435	302	.align 16
duke@435	303	_Copy_conjoint_jshorts_atomic:
duke@435	304	pushl %esi
duke@435	305	movl 4+12(%esp),%ecx / count
duke@435	306	pushl %edi
duke@435	307	movl 8+ 4(%esp),%esi / from
duke@435	308	movl 8+ 8(%esp),%edi / to
duke@435	309	cmpl %esi,%edi
duke@435	310	leal -2(%esi,%ecx,2),%eax / from + count*2 - 2
duke@435	311	jbe cs_CopyRight
duke@435	312	cmpl %eax,%edi
duke@435	313	jbe cs_CopyLeft
duke@435	314	/ copy from low to high
duke@435	315	cs_CopyRight:
duke@435	316	/ align source address at dword address boundary
duke@435	317	movl %esi,%eax / original from
duke@435	318	andl $3,%eax / either 0 or 2
duke@435	319	jz 1f / no prefix
duke@435	320	/ copy prefix
kvn@1958	321	subl $1,%ecx
kvn@1958	322	jl 5f / zero count
duke@435	323	movw (%esi),%dx
duke@435	324	movw %dx,(%edi)
duke@435	325	addl %eax,%esi / %eax == 2
duke@435	326	addl %eax,%edi
duke@435	327	1: movl %ecx,%eax / word count less prefix
duke@435	328	sarl %ecx / dword count
duke@435	329	jz 4f / no dwords to move
duke@435	330	cmpl $32,%ecx
duke@435	331	jbe 2f / <= 32 dwords
duke@435	332	/ copy aligned dwords
duke@435	333	rep; smovl
duke@435	334	jmp 4f
duke@435	335	/ copy aligned dwords
duke@435	336	2: subl %esi,%edi
duke@435	337	.align 16
duke@435	338	3: movl (%esi),%edx
duke@435	339	movl %edx,(%edi,%esi,1)
duke@435	340	addl $4,%esi
duke@435	341	subl $1,%ecx
duke@435	342	jnz 3b
duke@435	343	addl %esi,%edi
duke@435	344	4: andl $1,%eax / suffix count
duke@435	345	jz 5f / no suffix
duke@435	346	/ copy suffix
duke@435	347	movw (%esi),%dx
duke@435	348	movw %dx,(%edi)
duke@435	349	5: popl %edi
duke@435	350	popl %esi
duke@435	351	ret
duke@435	352	/ copy from high to low
duke@435	353	cs_CopyLeft:
duke@435	354	std
duke@435	355	leal -4(%edi,%ecx,2),%edi / to + count*2 - 4
duke@435	356	movl %eax,%esi / from + count*2 - 2
duke@435	357	movl %ecx,%eax
duke@435	358	subl $2,%esi / from + count*2 - 4
duke@435	359	1: sarl %ecx / dword count
duke@435	360	jz 4f / no dwords to move
duke@435	361	cmpl $32,%ecx
duke@435	362	ja 3f / > 32 dwords
duke@435	363	subl %esi,%edi
duke@435	364	.align 16
duke@435	365	2: movl (%esi),%edx
duke@435	366	movl %edx,(%edi,%esi,1)
duke@435	367	subl $4,%esi
duke@435	368	subl $1,%ecx
duke@435	369	jnz 2b
duke@435	370	addl %esi,%edi
duke@435	371	jmp 4f
duke@435	372	3: rep; smovl
duke@435	373	4: andl $1,%eax / suffix count
duke@435	374	jz 5f / no suffix
duke@435	375	/ copy suffix
duke@435	376	addl $2,%esi
duke@435	377	addl $2,%edi
duke@435	378	movw (%esi),%dx
duke@435	379	movw %dx,(%edi)
duke@435	380	5: cld
duke@435	381	popl %edi
duke@435	382	popl %esi
duke@435	383	ret
duke@435	384
duke@435	385	/ Support for void Copy::arrayof_conjoint_jshorts(void* from,
duke@435	386	/ void* to,
duke@435	387	/ size_t count)
duke@435	388	.align 16
duke@435	389	_Copy_arrayof_conjoint_jshorts:
duke@435	390	pushl %esi
duke@435	391	movl 4+12(%esp),%ecx / count
duke@435	392	pushl %edi
duke@435	393	movl 8+ 4(%esp),%esi / from
duke@435	394	movl 8+ 8(%esp),%edi / to
duke@435	395	cmpl %esi,%edi
duke@435	396	leal -2(%esi,%ecx,2),%eax / from + count*2 - 2
duke@435	397	jbe acs_CopyRight
duke@435	398	cmpl %eax,%edi
duke@435	399	jbe acs_CopyLeft
duke@435	400	acs_CopyRight:
duke@435	401	movl %ecx,%eax / word count
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	jbe 2f / <= 32 dwords
duke@435	406	/ copy aligned dwords
duke@435	407	rep; smovl
duke@435	408	jmp 4f
duke@435	409	/ copy aligned dwords
duke@435	410	.=.+5
duke@435	411	2: subl %esi,%edi
duke@435	412	.align 16
duke@435	413	3: movl (%esi),%edx
duke@435	414	movl %edx,(%edi,%esi,1)
duke@435	415	addl $4,%esi
duke@435	416	subl $1,%ecx
duke@435	417	jnz 3b
duke@435	418	addl %esi,%edi
duke@435	419	4: andl $1,%eax / suffix count
duke@435	420	jz 5f / no suffix
duke@435	421	/ copy suffix
duke@435	422	movw (%esi),%dx
duke@435	423	movw %dx,(%edi)
duke@435	424	5: popl %edi
duke@435	425	popl %esi
duke@435	426	ret
duke@435	427	acs_CopyLeft:
duke@435	428	std
duke@435	429	leal -4(%edi,%ecx,2),%edi / to + count*2 - 4
duke@435	430	movl %eax,%esi / from + count*2 - 2
duke@435	431	movl %ecx,%eax
duke@435	432	subl $2,%esi / from + count*2 - 4
duke@435	433	sarl %ecx / dword count
duke@435	434	jz 4f / no dwords to move
duke@435	435	cmpl $32,%ecx
duke@435	436	ja 3f / > 32 dwords
duke@435	437	subl %esi,%edi
duke@435	438	.align 16
duke@435	439	2: movl (%esi),%edx
duke@435	440	movl %edx,(%edi,%esi,1)
duke@435	441	subl $4,%esi
duke@435	442	subl $1,%ecx
duke@435	443	jnz 2b
duke@435	444	addl %esi,%edi
duke@435	445	jmp 4f
duke@435	446	3: rep; smovl
duke@435	447	4: andl $1,%eax / suffix count
duke@435	448	jz 5f / no suffix
duke@435	449	/ copy suffix
duke@435	450	addl $2,%esi
duke@435	451	addl $2,%edi
duke@435	452	movw (%esi),%dx
duke@435	453	movw %dx,(%edi)
duke@435	454	5: cld
duke@435	455	popl %edi
duke@435	456	popl %esi
duke@435	457	ret
duke@435	458
duke@435	459	/ Support for void Copy::conjoint_jints_atomic(void* from,
duke@435	460	/ void* to,
duke@435	461	/ size_t count)
duke@435	462	/ Equivalent to
duke@435	463	/ arrayof_conjoint_jints
duke@435	464	.align 16
duke@435	465	_Copy_conjoint_jints_atomic:
duke@435	466	_Copy_arrayof_conjoint_jints:
duke@435	467	pushl %esi
duke@435	468	movl 4+12(%esp),%ecx / count
duke@435	469	pushl %edi
duke@435	470	movl 8+ 4(%esp),%esi / from
duke@435	471	movl 8+ 8(%esp),%edi / to
duke@435	472	cmpl %esi,%edi
duke@435	473	leal -4(%esi,%ecx,4),%eax / from + count*4 - 4
duke@435	474	jbe ci_CopyRight
duke@435	475	cmpl %eax,%edi
duke@435	476	jbe ci_CopyLeft
duke@435	477	ci_CopyRight:
duke@435	478	cmpl $32,%ecx
duke@435	479	jbe 2f / <= 32 dwords
duke@435	480	rep; smovl
duke@435	481	popl %edi
duke@435	482	popl %esi
duke@435	483	ret
duke@435	484	.=.+10
duke@435	485	2: subl %esi,%edi
kvn@1958	486	jmp 4f
duke@435	487	.align 16
duke@435	488	3: movl (%esi),%edx
duke@435	489	movl %edx,(%edi,%esi,1)
duke@435	490	addl $4,%esi
kvn@1958	491	4: subl $1,%ecx
kvn@1958	492	jge 3b
duke@435	493	popl %edi
duke@435	494	popl %esi
duke@435	495	ret
duke@435	496	ci_CopyLeft:
duke@435	497	std
duke@435	498	leal -4(%edi,%ecx,4),%edi / to + count*4 - 4
duke@435	499	cmpl $32,%ecx
kvn@1958	500	ja 4f / > 32 dwords
duke@435	501	subl %eax,%edi / eax == from + count*4 - 4
kvn@1958	502	jmp 3f
duke@435	503	.align 16
duke@435	504	2: movl (%eax),%edx
duke@435	505	movl %edx,(%edi,%eax,1)
duke@435	506	subl $4,%eax
kvn@1958	507	3: subl $1,%ecx
kvn@1958	508	jge 2b
duke@435	509	cld
duke@435	510	popl %edi
duke@435	511	popl %esi
duke@435	512	ret
kvn@1958	513	4: movl %eax,%esi / from + count*4 - 4
duke@435	514	rep; smovl
duke@435	515	cld
duke@435	516	popl %edi
duke@435	517	popl %esi
duke@435	518	ret
duke@435	519
duke@435	520	/ Support for void Copy::conjoint_jlongs_atomic(jlong* from,
duke@435	521	/ jlong* to,
duke@435	522	/ size_t count)
duke@435	523	/
duke@435	524	/ 32-bit
duke@435	525	/
duke@435	526	/ count treated as signed
duke@435	527	/
duke@435	528	/ if (from > to) {
duke@435	529	/ while (--count >= 0) {
duke@435	530	/ *to++ = *from++;
duke@435	531	/ }
duke@435	532	/ } else {
duke@435	533	/ while (--count >= 0) {
duke@435	534	/ to[count] = from[count];
duke@435	535	/ }
duke@435	536	/ }
duke@435	537	.align 16
duke@435	538	_Copy_conjoint_jlongs_atomic:
duke@435	539	movl 4+8(%esp),%ecx / count
duke@435	540	movl 4+0(%esp),%eax / from
duke@435	541	movl 4+4(%esp),%edx / to
duke@435	542	cmpl %eax,%edx
duke@435	543	jae cla_CopyLeft
duke@435	544	cla_CopyRight:
duke@435	545	subl %eax,%edx
duke@435	546	jmp 2f
duke@435	547	.align 16
duke@435	548	1: fildll (%eax)
duke@435	549	fistpll (%edx,%eax,1)
duke@435	550	addl $8,%eax
duke@435	551	2: subl $1,%ecx
duke@435	552	jge 1b
duke@435	553	ret
duke@435	554	.align 16
duke@435	555	3: fildll (%eax,%ecx,8)
duke@435	556	fistpll (%edx,%ecx,8)
duke@435	557	cla_CopyLeft:
duke@435	558	subl $1,%ecx
duke@435	559	jge 3b
duke@435	560	ret
duke@435	561
duke@435	562	/ Support for void Copy::arrayof_conjoint_jshorts(void* from,
duke@435	563	/ void* to,
duke@435	564	/ size_t count)
duke@435	565	.align 16
duke@435	566	_mmx_Copy_arrayof_conjoint_jshorts:
duke@435	567	pushl %esi
duke@435	568	movl 4+12(%esp),%ecx
duke@435	569	pushl %edi
duke@435	570	movl 8+ 4(%esp),%esi
duke@435	571	movl 8+ 8(%esp),%edi
duke@435	572	cmpl %esi,%edi
duke@435	573	leal -2(%esi,%ecx,2),%eax
duke@435	574	jbe mmx_acs_CopyRight
duke@435	575	cmpl %eax,%edi
duke@435	576	jbe mmx_acs_CopyLeft
duke@435	577	mmx_acs_CopyRight:
duke@435	578	movl %ecx,%eax
duke@435	579	sarl %ecx
duke@435	580	je 5f
duke@435	581	cmpl $33,%ecx
duke@435	582	jae 3f
duke@435	583	1: subl %esi,%edi
duke@435	584	.align 16
duke@435	585	2: movl (%esi),%edx
duke@435	586	movl %edx,(%edi,%esi,1)
duke@435	587	addl $4,%esi
duke@435	588	subl $1,%ecx
duke@435	589	jnz 2b
duke@435	590	addl %esi,%edi
duke@435	591	jmp 5f
duke@435	592	3: smovl / align to 8 bytes, we know we are 4 byte aligned to start
duke@435	593	subl $1,%ecx
duke@435	594	4: .align 16
duke@435	595	movq 0(%esi),%mm0
duke@435	596	addl $64,%edi
duke@435	597	movq 8(%esi),%mm1
duke@435	598	subl $16,%ecx
duke@435	599	movq 16(%esi),%mm2
duke@435	600	movq %mm0,-64(%edi)
duke@435	601	movq 24(%esi),%mm0
duke@435	602	movq %mm1,-56(%edi)
duke@435	603	movq 32(%esi),%mm1
duke@435	604	movq %mm2,-48(%edi)
duke@435	605	movq 40(%esi),%mm2
duke@435	606	movq %mm0,-40(%edi)
duke@435	607	movq 48(%esi),%mm0
duke@435	608	movq %mm1,-32(%edi)
duke@435	609	movq 56(%esi),%mm1
duke@435	610	movq %mm2,-24(%edi)
duke@435	611	movq %mm0,-16(%edi)
duke@435	612	addl $64,%esi
duke@435	613	movq %mm1,-8(%edi)
duke@435	614	cmpl $16,%ecx
duke@435	615	jge 4b
duke@435	616	emms
duke@435	617	testl %ecx,%ecx
duke@435	618	ja 1b
duke@435	619	5: andl $1,%eax
duke@435	620	je 7f
duke@435	621	6: movw (%esi),%dx
duke@435	622	movw %dx,(%edi)
duke@435	623	7: popl %edi
duke@435	624	popl %esi
duke@435	625	ret
duke@435	626	mmx_acs_CopyLeft:
duke@435	627	std
duke@435	628	leal -4(%edi,%ecx,2),%edi
duke@435	629	movl %eax,%esi
duke@435	630	movl %ecx,%eax
duke@435	631	subl $2,%esi
duke@435	632	sarl %ecx
duke@435	633	je 4f
duke@435	634	cmpl $32,%ecx
duke@435	635	ja 3f
duke@435	636	subl %esi,%edi
duke@435	637	.align 16
duke@435	638	2: movl (%esi),%edx
duke@435	639	movl %edx,(%edi,%esi,1)
duke@435	640	subl $4,%esi
duke@435	641	subl $1,%ecx
duke@435	642	jnz 2b
duke@435	643	addl %esi,%edi
duke@435	644	jmp 4f
duke@435	645	3: rep; smovl
duke@435	646	4: andl $1,%eax
duke@435	647	je 6f
duke@435	648	addl $2,%esi
duke@435	649	addl $2,%edi
duke@435	650	5: movw (%esi),%dx
duke@435	651	movw %dx,(%edi)
duke@435	652	6: cld
duke@435	653	popl %edi
duke@435	654	popl %esi
duke@435	655	ret
duke@435	656
duke@435	657
duke@435	658	/ Support for jlong Atomic::cmpxchg(jlong exchange_value,
duke@435	659	/ volatile jlong* dest,
duke@435	660	/ jlong compare_value,
duke@435	661	/ bool is_MP)
duke@435	662	/ Used only for Solaris/gcc builds
duke@435	663	.align 16
duke@435	664	_Atomic_cmpxchg_long_gcc:
duke@435	665	/ 8(%esp) : return PC
duke@435	666	pushl %ebx / 4(%esp) : old %ebx
duke@435	667	pushl %edi / 0(%esp) : old %edi
duke@435	668	movl 12(%esp), %ebx / 12(%esp) : exchange_value (low)
duke@435	669	movl 16(%esp), %ecx / 16(%esp) : exchange_value (high)
duke@435	670	movl 24(%esp), %eax / 24(%esp) : compare_value (low)
duke@435	671	movl 28(%esp), %edx / 28(%esp) : compare_value (high)
duke@435	672	movl 20(%esp), %edi / 20(%esp) : dest
duke@435	673	cmpl $0, 32(%esp) / 32(%esp) : is_MP
duke@435	674	je 1f
duke@435	675	lock
duke@435	676	1: cmpxchg8b (%edi)
duke@435	677	popl %edi
duke@435	678	popl %ebx
duke@435	679	ret