Tue, 31 Mar 2009 14:07:08 -0700
6761600: Use sse 4.2 in intrinsics
Summary: Use SSE 4.2 in intrinsics for String.{compareTo/equals/indexOf} and Arrays.equals.
Reviewed-by: kvn, never, jrose
1 /*
2 * Copyright 1997-2008 Sun Microsystems, Inc. All Rights Reserved.
3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
4 *
5 * This code is free software; you can redistribute it and/or modify it
6 * under the terms of the GNU General Public License version 2 only, as
7 * published by the Free Software Foundation.
8 *
9 * This code is distributed in the hope that it will be useful, but WITHOUT
10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
12 * version 2 for more details (a copy is included in the LICENSE file that
13 * accompanied this code).
14 *
15 * You should have received a copy of the GNU General Public License version
16 * 2 along with this work; if not, write to the Free Software Foundation,
17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
18 *
19 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
20 * CA 95054 USA or visit www.sun.com if you need additional information or
21 * have any questions.
22 *
23 */
25 #include "incls/_precompiled.incl"
26 #include "incls/_vtableStubs_x86_32.cpp.incl"
28 // machine-dependent part of VtableStubs: create VtableStub of correct size and
29 // initialize its code
31 #define __ masm->
33 #ifndef PRODUCT
34 extern "C" void bad_compiled_vtable_index(JavaThread* thread, oop receiver, int index);
35 #endif
37 // These stubs are used by the compiler only.
38 // Argument registers, which must be preserved:
39 // rcx - receiver (always first argument)
40 // rdx - second argument (if any)
41 // Other registers that might be usable:
42 // rax - inline cache register (is interface for itable stub)
43 // rbx - method (used when calling out to interpreter)
44 // Available now, but may become callee-save at some point:
45 // rsi, rdi
46 // Note that rax and rdx are also used for return values.
47 //
48 VtableStub* VtableStubs::create_vtable_stub(int vtable_index) {
49 const int i486_code_length = VtableStub::pd_code_size_limit(true);
50 VtableStub* s = new(i486_code_length) VtableStub(true, vtable_index);
51 ResourceMark rm;
52 CodeBuffer cb(s->entry_point(), i486_code_length);
53 MacroAssembler* masm = new MacroAssembler(&cb);
55 #ifndef PRODUCT
57 if (CountCompiledCalls) {
58 __ incrementl(ExternalAddress((address) SharedRuntime::nof_megamorphic_calls_addr()));
59 }
60 #endif /* PRODUCT */
62 // get receiver (need to skip return address on top of stack)
63 assert(VtableStub::receiver_location() == rcx->as_VMReg(), "receiver expected in rcx");
65 // get receiver klass
66 address npe_addr = __ pc();
67 __ movptr(rax, Address(rcx, oopDesc::klass_offset_in_bytes()));
68 // compute entry offset (in words)
69 int entry_offset = instanceKlass::vtable_start_offset() + vtable_index*vtableEntry::size();
70 #ifndef PRODUCT
71 if (DebugVtables) {
72 Label L;
73 // check offset vs vtable length
74 __ cmpl(Address(rax, instanceKlass::vtable_length_offset()*wordSize), vtable_index*vtableEntry::size());
75 __ jcc(Assembler::greater, L);
76 __ movl(rbx, vtable_index);
77 __ call_VM(noreg, CAST_FROM_FN_PTR(address, bad_compiled_vtable_index), rcx, rbx);
78 __ bind(L);
79 }
80 #endif // PRODUCT
82 const Register method = rbx;
84 // load methodOop and target address
85 __ movptr(method, Address(rax, entry_offset*wordSize + vtableEntry::method_offset_in_bytes()));
86 if (DebugVtables) {
87 Label L;
88 __ cmpptr(method, (int32_t)NULL_WORD);
89 __ jcc(Assembler::equal, L);
90 __ cmpptr(Address(method, methodOopDesc::from_compiled_offset()), (int32_t)NULL_WORD);
91 __ jcc(Assembler::notZero, L);
92 __ stop("Vtable entry is NULL");
93 __ bind(L);
94 }
96 // rax,: receiver klass
97 // method (rbx): methodOop
98 // rcx: receiver
99 address ame_addr = __ pc();
100 __ jmp( Address(method, methodOopDesc::from_compiled_offset()));
102 masm->flush();
104 if (PrintMiscellaneous && (WizardMode || Verbose)) {
105 tty->print_cr("vtable #%d at "PTR_FORMAT"[%d] left over: %d",
106 vtable_index, s->entry_point(),
107 (int)(s->code_end() - s->entry_point()),
108 (int)(s->code_end() - __ pc()));
109 }
110 guarantee(__ pc() <= s->code_end(), "overflowed buffer");
112 s->set_exception_points(npe_addr, ame_addr);
113 return s;
114 }
117 VtableStub* VtableStubs::create_itable_stub(int itable_index) {
118 // Note well: pd_code_size_limit is the absolute minimum we can get away with. If you
119 // add code here, bump the code stub size returned by pd_code_size_limit!
120 const int i486_code_length = VtableStub::pd_code_size_limit(false);
121 VtableStub* s = new(i486_code_length) VtableStub(false, itable_index);
122 ResourceMark rm;
123 CodeBuffer cb(s->entry_point(), i486_code_length);
124 MacroAssembler* masm = new MacroAssembler(&cb);
126 // Entry arguments:
127 // rax,: Interface
128 // rcx: Receiver
130 #ifndef PRODUCT
131 if (CountCompiledCalls) {
132 __ incrementl(ExternalAddress((address) SharedRuntime::nof_megamorphic_calls_addr()));
133 }
134 #endif /* PRODUCT */
135 // get receiver (need to skip return address on top of stack)
137 assert(VtableStub::receiver_location() == rcx->as_VMReg(), "receiver expected in rcx");
139 // get receiver klass (also an implicit null-check)
140 address npe_addr = __ pc();
141 __ movptr(rsi, Address(rcx, oopDesc::klass_offset_in_bytes()));
143 // Most registers are in use; we'll use rax, rbx, rsi, rdi
144 // (If we need to make rsi, rdi callee-save, do a push/pop here.)
145 const Register method = rbx;
146 Label throw_icce;
148 // Get methodOop and entrypoint for compiler
149 __ lookup_interface_method(// inputs: rec. class, interface, itable index
150 rsi, rax, itable_index,
151 // outputs: method, scan temp. reg
152 method, rdi,
153 throw_icce);
155 // method (rbx): methodOop
156 // rcx: receiver
158 #ifdef ASSERT
159 if (DebugVtables) {
160 Label L1;
161 __ cmpptr(method, (int32_t)NULL_WORD);
162 __ jcc(Assembler::equal, L1);
163 __ cmpptr(Address(method, methodOopDesc::from_compiled_offset()), (int32_t)NULL_WORD);
164 __ jcc(Assembler::notZero, L1);
165 __ stop("methodOop is null");
166 __ bind(L1);
167 }
168 #endif // ASSERT
170 address ame_addr = __ pc();
171 __ jmp(Address(method, methodOopDesc::from_compiled_offset()));
173 __ bind(throw_icce);
174 __ jump(RuntimeAddress(StubRoutines::throw_IncompatibleClassChangeError_entry()));
175 masm->flush();
177 if (PrintMiscellaneous && (WizardMode || Verbose)) {
178 tty->print_cr("itable #%d at "PTR_FORMAT"[%d] left over: %d",
179 itable_index, s->entry_point(),
180 (int)(s->code_end() - s->entry_point()),
181 (int)(s->code_end() - __ pc()));
182 }
183 guarantee(__ pc() <= s->code_end(), "overflowed buffer");
185 s->set_exception_points(npe_addr, ame_addr);
186 return s;
187 }
191 int VtableStub::pd_code_size_limit(bool is_vtable_stub) {
192 if (is_vtable_stub) {
193 // Vtable stub size
194 return (DebugVtables ? 210 : 16) + (CountCompiledCalls ? 6 : 0);
195 } else {
196 // Itable stub size
197 return (DebugVtables ? 256 : 66) + (CountCompiledCalls ? 6 : 0);
198 }
199 }
201 int VtableStub::pd_code_alignment() {
202 return wordSize;
203 }