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

src/os_cpu/solaris_x86/vm/solaris_x86_64.s@1e9a9d2e6509 (annotated)

src/os_cpu/solaris_x86/vm/solaris_x86_64.s

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

author

never

date

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

changeset 2262

1e9a9d2e6509

parent 1907

c18cbe5936b8

child 5400

980532a806a5

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, 2005, 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 fs_load
duke@435	25	.globl fs_thread
duke@435	26
duke@435	27	// NOTE WELL! The _Copy functions are called directly
duke@435	28	// from server-compiler-generated code via CallLeafNoFP,
duke@435	29	// which means that they *must* either not use floating
duke@435	30	// point or use it in the same manner as does the server
duke@435	31	// compiler.
duke@435	32
duke@435	33	.globl _Copy_arrayof_conjoint_bytes
duke@435	34	.globl _Copy_conjoint_jshorts_atomic
duke@435	35	.globl _Copy_arrayof_conjoint_jshorts
duke@435	36	.globl _Copy_conjoint_jints_atomic
duke@435	37	.globl _Copy_arrayof_conjoint_jints
duke@435	38	.globl _Copy_conjoint_jlongs_atomic
duke@435	39	.globl _Copy_arrayof_conjoint_jlongs
duke@435	40
duke@435	41	.section .text,"ax"
duke@435	42
duke@435	43	/ Fast thread accessors, used by threadLS_solaris_amd64.cpp
duke@435	44	.align 16
duke@435	45	fs_load:
duke@435	46	movq %fs:(%rdi),%rax
duke@435	47	ret
duke@435	48
duke@435	49	.align 16
duke@435	50	fs_thread:
duke@435	51	movq %fs:0x0,%rax
duke@435	52	ret
duke@435	53
duke@435	54	.globl SafeFetch32, Fetch32PFI, Fetch32Resume
duke@435	55	.align 16
duke@435	56	// Prototype: int SafeFetch32 (int * Adr, int ErrValue)
duke@435	57	SafeFetch32:
duke@435	58	movl %esi, %eax
duke@435	59	Fetch32PFI:
duke@435	60	movl (%rdi), %eax
duke@435	61	Fetch32Resume:
duke@435	62	ret
duke@435	63
duke@435	64	.globl SafeFetchN, FetchNPFI, FetchNResume
duke@435	65	.align 16
duke@435	66	// Prototype: intptr_t SafeFetchN (intptr_t * Adr, intptr_t ErrValue)
duke@435	67	SafeFetchN:
duke@435	68	movq %rsi, %rax
duke@435	69	FetchNPFI:
duke@435	70	movq (%rdi), %rax
duke@435	71	FetchNResume:
duke@435	72	ret
duke@435	73
duke@435	74	.globl SpinPause
duke@435	75	.align 16
duke@435	76	SpinPause:
duke@435	77	rep
duke@435	78	nop
duke@435	79	movq $1, %rax
duke@435	80	ret
duke@435	81
duke@435	82
duke@435	83	/ Support for void Copy::arrayof_conjoint_bytes(void* from,
duke@435	84	/ void* to,
duke@435	85	/ size_t count)
duke@435	86	/ rdi - from
duke@435	87	/ rsi - to
duke@435	88	/ rdx - count, treated as ssize_t
duke@435	89	/
duke@435	90	.align 16
duke@435	91	_Copy_arrayof_conjoint_bytes:
duke@435	92	movq %rdx,%r8 / byte count
duke@435	93	shrq $3,%rdx / qword count
duke@435	94	cmpq %rdi,%rsi
duke@435	95	leaq -1(%rdi,%r8,1),%rax / from + bcount*1 - 1
duke@435	96	jbe acb_CopyRight
duke@435	97	cmpq %rax,%rsi
duke@435	98	jbe acb_CopyLeft
duke@435	99	acb_CopyRight:
duke@435	100	leaq -8(%rdi,%rdx,8),%rax / from + qcount*8 - 8
duke@435	101	leaq -8(%rsi,%rdx,8),%rcx / to + qcount*8 - 8
duke@435	102	negq %rdx
duke@435	103	jmp 7f
duke@435	104	.align 16
duke@435	105	1: movq 8(%rax,%rdx,8),%rsi
duke@435	106	movq %rsi,8(%rcx,%rdx,8)
duke@435	107	addq $1,%rdx
duke@435	108	jnz 1b
duke@435	109	2: testq $4,%r8 / check for trailing dword
duke@435	110	jz 3f
duke@435	111	movl 8(%rax),%esi / copy trailing dword
duke@435	112	movl %esi,8(%rcx)
duke@435	113	addq $4,%rax
duke@435	114	addq $4,%rcx / original %rsi is trashed, so we
duke@435	115	/ can't use it as a base register
duke@435	116	3: testq $2,%r8 / check for trailing word
duke@435	117	jz 4f
duke@435	118	movw 8(%rax),%si / copy trailing word
duke@435	119	movw %si,8(%rcx)
duke@435	120	addq $2,%rcx
duke@435	121	4: testq $1,%r8 / check for trailing byte
duke@435	122	jz 5f
duke@435	123	movb -1(%rdi,%r8,1),%al / copy trailing byte
duke@435	124	movb %al,8(%rcx)
duke@435	125	5: ret
duke@435	126	.align 16
duke@435	127	6: movq -24(%rax,%rdx,8),%rsi
duke@435	128	movq %rsi,-24(%rcx,%rdx,8)
duke@435	129	movq -16(%rax,%rdx,8),%rsi
duke@435	130	movq %rsi,-16(%rcx,%rdx,8)
duke@435	131	movq -8(%rax,%rdx,8),%rsi
duke@435	132	movq %rsi,-8(%rcx,%rdx,8)
duke@435	133	movq (%rax,%rdx,8),%rsi
duke@435	134	movq %rsi,(%rcx,%rdx,8)
duke@435	135	7: addq $4,%rdx
duke@435	136	jle 6b
duke@435	137	subq $4,%rdx
duke@435	138	jl 1b
duke@435	139	jmp 2b
duke@435	140	acb_CopyLeft:
duke@435	141	testq $1,%r8 / check for trailing byte
duke@435	142	jz 1f
duke@435	143	movb -1(%rdi,%r8,1),%cl / copy trailing byte
duke@435	144	movb %cl,-1(%rsi,%r8,1)
duke@435	145	subq $1,%r8 / adjust for possible trailing word
duke@435	146	1: testq $2,%r8 / check for trailing word
duke@435	147	jz 2f
duke@435	148	movw -2(%rdi,%r8,1),%cx / copy trailing word
duke@435	149	movw %cx,-2(%rsi,%r8,1)
duke@435	150	2: testq $4,%r8 / check for trailing dword
duke@435	151	jz 5f
duke@435	152	movl (%rdi,%rdx,8),%ecx / copy trailing dword
duke@435	153	movl %ecx,(%rsi,%rdx,8)
duke@435	154	jmp 5f
duke@435	155	.align 16
duke@435	156	3: movq -8(%rdi,%rdx,8),%rcx
duke@435	157	movq %rcx,-8(%rsi,%rdx,8)
duke@435	158	subq $1,%rdx
duke@435	159	jnz 3b
duke@435	160	ret
duke@435	161	.align 16
duke@435	162	4: movq 24(%rdi,%rdx,8),%rcx
duke@435	163	movq %rcx,24(%rsi,%rdx,8)
duke@435	164	movq 16(%rdi,%rdx,8),%rcx
duke@435	165	movq %rcx,16(%rsi,%rdx,8)
duke@435	166	movq 8(%rdi,%rdx,8),%rcx
duke@435	167	movq %rcx,8(%rsi,%rdx,8)
duke@435	168	movq (%rdi,%rdx,8),%rcx
duke@435	169	movq %rcx,(%rsi,%rdx,8)
duke@435	170	5: subq $4,%rdx
duke@435	171	jge 4b
duke@435	172	addq $4,%rdx
duke@435	173	jg 3b
duke@435	174	ret
duke@435	175
duke@435	176	/ Support for void Copy::arrayof_conjoint_jshorts(void* from,
duke@435	177	/ void* to,
duke@435	178	/ size_t count)
duke@435	179	/ Equivalent to
duke@435	180	/ conjoint_jshorts_atomic
duke@435	181	/
duke@435	182	/ If 'from' and/or 'to' are aligned on 4- or 2-byte boundaries, we
duke@435	183	/ let the hardware handle it. The tow or four words within dwords
duke@435	184	/ or qwords that span cache line boundaries will still be loaded
duke@435	185	/ and stored atomically.
duke@435	186	/
duke@435	187	/ rdi - from
duke@435	188	/ rsi - to
duke@435	189	/ rdx - count, treated as ssize_t
duke@435	190	/
duke@435	191	.align 16
duke@435	192	_Copy_arrayof_conjoint_jshorts:
duke@435	193	_Copy_conjoint_jshorts_atomic:
duke@435	194	movq %rdx,%r8 / word count
duke@435	195	shrq $2,%rdx / qword count
duke@435	196	cmpq %rdi,%rsi
duke@435	197	leaq -2(%rdi,%r8,2),%rax / from + wcount*2 - 2
duke@435	198	jbe acs_CopyRight
duke@435	199	cmpq %rax,%rsi
duke@435	200	jbe acs_CopyLeft
duke@435	201	acs_CopyRight:
duke@435	202	leaq -8(%rdi,%rdx,8),%rax / from + qcount*8 - 8
duke@435	203	leaq -8(%rsi,%rdx,8),%rcx / to + qcount*8 - 8
duke@435	204	negq %rdx
duke@435	205	jmp 6f
duke@435	206	1: movq 8(%rax,%rdx,8),%rsi
duke@435	207	movq %rsi,8(%rcx,%rdx,8)
duke@435	208	addq $1,%rdx
duke@435	209	jnz 1b
duke@435	210	2: testq $2,%r8 / check for trailing dword
duke@435	211	jz 3f
duke@435	212	movl 8(%rax),%esi / copy trailing dword
duke@435	213	movl %esi,8(%rcx)
duke@435	214	addq $4,%rcx / original %rsi is trashed, so we
duke@435	215	/ can't use it as a base register
duke@435	216	3: testq $1,%r8 / check for trailing word
duke@435	217	jz 4f
duke@435	218	movw -2(%rdi,%r8,2),%si / copy trailing word
duke@435	219	movw %si,8(%rcx)
duke@435	220	4: ret
duke@435	221	.align 16
duke@435	222	5: movq -24(%rax,%rdx,8),%rsi
duke@435	223	movq %rsi,-24(%rcx,%rdx,8)
duke@435	224	movq -16(%rax,%rdx,8),%rsi
duke@435	225	movq %rsi,-16(%rcx,%rdx,8)
duke@435	226	movq -8(%rax,%rdx,8),%rsi
duke@435	227	movq %rsi,-8(%rcx,%rdx,8)
duke@435	228	movq (%rax,%rdx,8),%rsi
duke@435	229	movq %rsi,(%rcx,%rdx,8)
duke@435	230	6: addq $4,%rdx
duke@435	231	jle 5b
duke@435	232	subq $4,%rdx
duke@435	233	jl 1b
duke@435	234	jmp 2b
duke@435	235	acs_CopyLeft:
duke@435	236	testq $1,%r8 / check for trailing word
duke@435	237	jz 1f
duke@435	238	movw -2(%rdi,%r8,2),%cx / copy trailing word
duke@435	239	movw %cx,-2(%rsi,%r8,2)
duke@435	240	1: testq $2,%r8 / check for trailing dword
duke@435	241	jz 4f
duke@435	242	movl (%rdi,%rdx,8),%ecx / copy trailing dword
duke@435	243	movl %ecx,(%rsi,%rdx,8)
duke@435	244	jmp 4f
duke@435	245	2: movq -8(%rdi,%rdx,8),%rcx
duke@435	246	movq %rcx,-8(%rsi,%rdx,8)
duke@435	247	subq $1,%rdx
duke@435	248	jnz 2b
duke@435	249	ret
duke@435	250	.align 16
duke@435	251	3: movq 24(%rdi,%rdx,8),%rcx
duke@435	252	movq %rcx,24(%rsi,%rdx,8)
duke@435	253	movq 16(%rdi,%rdx,8),%rcx
duke@435	254	movq %rcx,16(%rsi,%rdx,8)
duke@435	255	movq 8(%rdi,%rdx,8),%rcx
duke@435	256	movq %rcx,8(%rsi,%rdx,8)
duke@435	257	movq (%rdi,%rdx,8),%rcx
duke@435	258	movq %rcx,(%rsi,%rdx,8)
duke@435	259	4: subq $4,%rdx
duke@435	260	jge 3b
duke@435	261	addq $4,%rdx
duke@435	262	jg 2b
duke@435	263	ret
duke@435	264
duke@435	265	/ Support for void Copy::arrayof_conjoint_jints(jint* from,
duke@435	266	/ jint* to,
duke@435	267	/ size_t count)
duke@435	268	/ Equivalent to
duke@435	269	/ conjoint_jints_atomic
duke@435	270	/
duke@435	271	/ If 'from' and/or 'to' are aligned on 4-byte boundaries, we let
duke@435	272	/ the hardware handle it. The two dwords within qwords that span
duke@435	273	/ cache line boundaries will still be loaded and stored atomically.
duke@435	274	/
duke@435	275	/ rdi - from
duke@435	276	/ rsi - to
duke@435	277	/ rdx - count, treated as ssize_t
duke@435	278	/
duke@435	279	.align 16
duke@435	280	_Copy_arrayof_conjoint_jints:
duke@435	281	_Copy_conjoint_jints_atomic:
duke@435	282	movq %rdx,%r8 / dword count
duke@435	283	shrq %rdx / qword count
duke@435	284	cmpq %rdi,%rsi
duke@435	285	leaq -4(%rdi,%r8,4),%rax / from + dcount*4 - 4
duke@435	286	jbe aci_CopyRight
duke@435	287	cmpq %rax,%rsi
duke@435	288	jbe aci_CopyLeft
duke@435	289	aci_CopyRight:
duke@435	290	leaq -8(%rdi,%rdx,8),%rax / from + qcount*8 - 8
duke@435	291	leaq -8(%rsi,%rdx,8),%rcx / to + qcount*8 - 8
duke@435	292	negq %rdx
duke@435	293	jmp 5f
duke@435	294	.align 16
duke@435	295	1: movq 8(%rax,%rdx,8),%rsi
duke@435	296	movq %rsi,8(%rcx,%rdx,8)
duke@435	297	addq $1,%rdx
duke@435	298	jnz 1b
duke@435	299	2: testq $1,%r8 / check for trailing dword
duke@435	300	jz 3f
duke@435	301	movl 8(%rax),%esi / copy trailing dword
duke@435	302	movl %esi,8(%rcx)
duke@435	303	3: ret
duke@435	304	.align 16
duke@435	305	4: movq -24(%rax,%rdx,8),%rsi
duke@435	306	movq %rsi,-24(%rcx,%rdx,8)
duke@435	307	movq -16(%rax,%rdx,8),%rsi
duke@435	308	movq %rsi,-16(%rcx,%rdx,8)
duke@435	309	movq -8(%rax,%rdx,8),%rsi
duke@435	310	movq %rsi,-8(%rcx,%rdx,8)
duke@435	311	movq (%rax,%rdx,8),%rsi
duke@435	312	movq %rsi,(%rcx,%rdx,8)
duke@435	313	5: addq $4,%rdx
duke@435	314	jle 4b
duke@435	315	subq $4,%rdx
duke@435	316	jl 1b
duke@435	317	jmp 2b
duke@435	318	aci_CopyLeft:
duke@435	319	testq $1,%r8 / check for trailing dword
duke@435	320	jz 3f
duke@435	321	movl -4(%rdi,%r8,4),%ecx / copy trailing dword
duke@435	322	movl %ecx,-4(%rsi,%r8,4)
duke@435	323	jmp 3f
duke@435	324	1: movq -8(%rdi,%rdx,8),%rcx
duke@435	325	movq %rcx,-8(%rsi,%rdx,8)
duke@435	326	subq $1,%rdx
duke@435	327	jnz 1b
duke@435	328	ret
duke@435	329	.align 16
duke@435	330	2: movq 24(%rdi,%rdx,8),%rcx
duke@435	331	movq %rcx,24(%rsi,%rdx,8)
duke@435	332	movq 16(%rdi,%rdx,8),%rcx
duke@435	333	movq %rcx,16(%rsi,%rdx,8)
duke@435	334	movq 8(%rdi,%rdx,8),%rcx
duke@435	335	movq %rcx,8(%rsi,%rdx,8)
duke@435	336	movq (%rdi,%rdx,8),%rcx
duke@435	337	movq %rcx,(%rsi,%rdx,8)
duke@435	338	3: subq $4,%rdx
duke@435	339	jge 2b
duke@435	340	addq $4,%rdx
duke@435	341	jg 1b
duke@435	342	ret
duke@435	343
duke@435	344	/ Support for void Copy::arrayof_conjoint_jlongs(jlong* from,
duke@435	345	/ jlong* to,
duke@435	346	/ size_t count)
duke@435	347	/ Equivalent to
duke@435	348	/ conjoint_jlongs_atomic
duke@435	349	/ arrayof_conjoint_oops
duke@435	350	/ conjoint_oops_atomic
duke@435	351	/
duke@435	352	/ rdi - from
duke@435	353	/ rsi - to
duke@435	354	/ rdx - count, treated as ssize_t
duke@435	355	/
duke@435	356	.align 16
duke@435	357	_Copy_arrayof_conjoint_jlongs:
duke@435	358	_Copy_conjoint_jlongs_atomic:
duke@435	359	cmpq %rdi,%rsi
duke@435	360	leaq -8(%rdi,%rdx,8),%rax / from + count*8 - 8
duke@435	361	jbe acl_CopyRight
duke@435	362	cmpq %rax,%rsi
duke@435	363	jbe acl_CopyLeft
duke@435	364	acl_CopyRight:
duke@435	365	leaq -8(%rsi,%rdx,8),%rcx / to + count*8 - 8
duke@435	366	negq %rdx
duke@435	367	jmp 3f
duke@435	368	1: movq 8(%rax,%rdx,8),%rsi
duke@435	369	movq %rsi,8(%rcx,%rdx,8)
duke@435	370	addq $1,%rdx
duke@435	371	jnz 1b
duke@435	372	ret
duke@435	373	.align 16
duke@435	374	2: movq -24(%rax,%rdx,8),%rsi
duke@435	375	movq %rsi,-24(%rcx,%rdx,8)
duke@435	376	movq -16(%rax,%rdx,8),%rsi
duke@435	377	movq %rsi,-16(%rcx,%rdx,8)
duke@435	378	movq -8(%rax,%rdx,8),%rsi
duke@435	379	movq %rsi,-8(%rcx,%rdx,8)
duke@435	380	movq (%rax,%rdx,8),%rsi
duke@435	381	movq %rsi,(%rcx,%rdx,8)
duke@435	382	3: addq $4,%rdx
duke@435	383	jle 2b
duke@435	384	subq $4,%rdx
duke@435	385	jl 1b
duke@435	386	ret
duke@435	387	4: movq -8(%rdi,%rdx,8),%rcx
duke@435	388	movq %rcx,-8(%rsi,%rdx,8)
duke@435	389	subq $1,%rdx
duke@435	390	jnz 4b
duke@435	391	ret
duke@435	392	.align 16
duke@435	393	5: movq 24(%rdi,%rdx,8),%rcx
duke@435	394	movq %rcx,24(%rsi,%rdx,8)
duke@435	395	movq 16(%rdi,%rdx,8),%rcx
duke@435	396	movq %rcx,16(%rsi,%rdx,8)
duke@435	397	movq 8(%rdi,%rdx,8),%rcx
duke@435	398	movq %rcx,8(%rsi,%rdx,8)
duke@435	399	movq (%rdi,%rdx,8),%rcx
duke@435	400	movq %rcx,(%rsi,%rdx,8)
duke@435	401	acl_CopyLeft:
duke@435	402	subq $4,%rdx
duke@435	403	jge 5b
duke@435	404	addq $4,%rdx
duke@435	405	jg 4b
duke@435	406	ret