src/cpu/x86/vm/vtableStubs_x86_64.cpp

Tue, 31 Mar 2009 14:07:08 -0700

author
cfang
date
Tue, 31 Mar 2009 14:07:08 -0700
changeset 1116
fbde8ec322d0
parent 1058
9adddb8c0fc8
child 1144
1d037ecd7960
permissions
-rw-r--r--

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 2003-2008 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_64.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,
    35                                           oop receiver,
    36                                           int index);
    37 #endif
    39 VtableStub* VtableStubs::create_vtable_stub(int vtable_index) {
    40   const int amd64_code_length = VtableStub::pd_code_size_limit(true);
    41   VtableStub* s = new(amd64_code_length) VtableStub(true, vtable_index);
    42   ResourceMark rm;
    43   CodeBuffer cb(s->entry_point(), amd64_code_length);
    44   MacroAssembler* masm = new MacroAssembler(&cb);
    46 #ifndef PRODUCT
    47   if (CountCompiledCalls) {
    48     __ incrementl(ExternalAddress((address) SharedRuntime::nof_megamorphic_calls_addr()));
    49   }
    50 #endif
    52   // get receiver (need to skip return address on top of stack)
    53   assert(VtableStub::receiver_location() == j_rarg0->as_VMReg(), "receiver expected in j_rarg0");
    55   // Free registers (non-args) are rax, rbx
    57   // get receiver klass
    58   address npe_addr = __ pc();
    59   __ load_klass(rax, j_rarg0);
    61   // compute entry offset (in words)
    62   int entry_offset =
    63     instanceKlass::vtable_start_offset() + vtable_index * vtableEntry::size();
    65 #ifndef PRODUCT
    66   if (DebugVtables) {
    67     Label L;
    68     // check offset vs vtable length
    69     __ cmpl(Address(rax, instanceKlass::vtable_length_offset() * wordSize),
    70             vtable_index * vtableEntry::size());
    71     __ jcc(Assembler::greater, L);
    72     __ movl(rbx, vtable_index);
    73     __ call_VM(noreg,
    74                CAST_FROM_FN_PTR(address, bad_compiled_vtable_index), j_rarg0, rbx);
    75     __ bind(L);
    76   }
    77 #endif // PRODUCT
    79   // load methodOop and target address
    80   const Register method = rbx;
    82   __ movptr(method, Address(rax,
    83                             entry_offset * wordSize +
    84                             vtableEntry::method_offset_in_bytes()));
    85   if (DebugVtables) {
    86     Label L;
    87     __ cmpptr(method, (int32_t)NULL_WORD);
    88     __ jcc(Assembler::equal, L);
    89     __ cmpptr(Address(method, methodOopDesc::from_compiled_offset()), (int32_t)NULL_WORD);
    90     __ jcc(Assembler::notZero, L);
    91     __ stop("Vtable entry is NULL");
    92     __ bind(L);
    93   }
    94   // rax: receiver klass
    95   // rbx: methodOop
    96   // rcx: receiver
    97   address ame_addr = __ pc();
    98   __ jmp( Address(rbx, methodOopDesc::from_compiled_offset()));
   100   __ flush();
   102   if (PrintMiscellaneous && (WizardMode || Verbose)) {
   103     tty->print_cr("vtable #%d at "PTR_FORMAT"[%d] left over: %d",
   104                   vtable_index, s->entry_point(),
   105                   (int)(s->code_end() - s->entry_point()),
   106                   (int)(s->code_end() - __ pc()));
   107   }
   108   guarantee(__ pc() <= s->code_end(), "overflowed buffer");
   110   s->set_exception_points(npe_addr, ame_addr);
   111   return s;
   112 }
   115 VtableStub* VtableStubs::create_itable_stub(int itable_index) {
   116   // Note well: pd_code_size_limit is the absolute minimum we can get
   117   // away with.  If you add code here, bump the code stub size
   118   // returned by pd_code_size_limit!
   119   const int amd64_code_length = VtableStub::pd_code_size_limit(false);
   120   VtableStub* s = new(amd64_code_length) VtableStub(false, itable_index);
   121   ResourceMark rm;
   122   CodeBuffer cb(s->entry_point(), amd64_code_length);
   123   MacroAssembler* masm = new MacroAssembler(&cb);
   125 #ifndef PRODUCT
   126   if (CountCompiledCalls) {
   127     __ incrementl(ExternalAddress((address) SharedRuntime::nof_megamorphic_calls_addr()));
   128   }
   129 #endif
   131   // Entry arguments:
   132   //  rax: Interface
   133   //  j_rarg0: Receiver
   135   // Free registers (non-args) are rax (interface), rbx
   137   // get receiver (need to skip return address on top of stack)
   139   assert(VtableStub::receiver_location() == j_rarg0->as_VMReg(), "receiver expected in j_rarg0");
   140   // get receiver klass (also an implicit null-check)
   141   address npe_addr = __ pc();
   143   // Most registers are in use; we'll use rax, rbx, r10, r11
   144   // (various calling sequences use r[cd]x, r[sd]i, r[89]; stay away from them)
   145   __ load_klass(r10, j_rarg0);
   147   // If we take a trap while this arg is on the stack we will not
   148   // be able to walk the stack properly. This is not an issue except
   149   // when there are mistakes in this assembly code that could generate
   150   // a spurious fault. Ask me how I know...
   152   const Register method = rbx;
   153   Label throw_icce;
   155   // Get methodOop and entrypoint for compiler
   156   __ lookup_interface_method(// inputs: rec. class, interface, itable index
   157                              r10, rax, itable_index,
   158                              // outputs: method, scan temp. reg
   159                              method, r11,
   160                              throw_icce);
   162   // method (rbx): methodOop
   163   // j_rarg0: receiver
   165 #ifdef ASSERT
   166   if (DebugVtables) {
   167     Label L2;
   168     __ cmpptr(method, (int32_t)NULL_WORD);
   169     __ jcc(Assembler::equal, L2);
   170     __ cmpptr(Address(method, methodOopDesc::from_compiled_offset()), (int32_t)NULL_WORD);
   171     __ jcc(Assembler::notZero, L2);
   172     __ stop("compiler entrypoint is null");
   173     __ bind(L2);
   174   }
   175 #endif // ASSERT
   177   // rbx: methodOop
   178   // j_rarg0: receiver
   179   address ame_addr = __ pc();
   180   __ jmp(Address(method, methodOopDesc::from_compiled_offset()));
   182   __ bind(throw_icce);
   183   __ jump(RuntimeAddress(StubRoutines::throw_IncompatibleClassChangeError_entry()));
   185   __ flush();
   187   if (PrintMiscellaneous && (WizardMode || Verbose)) {
   188     tty->print_cr("itable #%d at "PTR_FORMAT"[%d] left over: %d",
   189                   itable_index, s->entry_point(),
   190                   (int)(s->code_end() - s->entry_point()),
   191                   (int)(s->code_end() - __ pc()));
   192   }
   193   guarantee(__ pc() <= s->code_end(), "overflowed buffer");
   195   s->set_exception_points(npe_addr, ame_addr);
   196   return s;
   197 }
   199 int VtableStub::pd_code_size_limit(bool is_vtable_stub) {
   200   if (is_vtable_stub) {
   201     // Vtable stub size
   202     return (DebugVtables ? 512 : 24) + (CountCompiledCalls ? 13 : 0) +
   203            (UseCompressedOops ? 16 : 0);  // 1 leaq can be 3 bytes + 1 long
   204   } else {
   205     // Itable stub size
   206     return (DebugVtables ? 512 : 72) + (CountCompiledCalls ? 13 : 0) +
   207            (UseCompressedOops ? 32 : 0);  // 2 leaqs
   208   }
   209 }
   211 int VtableStub::pd_code_alignment() {
   212   return wordSize;
   213 }

mercurial