Tue, 14 Oct 2008 15:10:26 -0700
6532536: Optimize arraycopy stubs for Intel cpus
Summary: Use SSE2 movdqu in arraycopy stubs on newest Intel's cpus
Reviewed-by: rasbold
duke@435 | 1 | /* |
duke@435 | 2 | * Copyright 1998-2006 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 | |
duke@435 | 26 | #include "incls/_precompiled.incl" |
duke@435 | 27 | #include "incls/_runtime_x86_32.cpp.incl" |
duke@435 | 28 | |
duke@435 | 29 | #define __ masm-> |
duke@435 | 30 | |
duke@435 | 31 | ExceptionBlob* OptoRuntime::_exception_blob; |
duke@435 | 32 | |
duke@435 | 33 | //------------------------------generate_exception_blob--------------------------- |
duke@435 | 34 | // creates exception blob at the end |
duke@435 | 35 | // Using exception blob, this code is jumped from a compiled method. |
duke@435 | 36 | // |
duke@435 | 37 | // Given an exception pc at a call we call into the runtime for the |
duke@435 | 38 | // handler in this method. This handler might merely restore state |
duke@435 | 39 | // (i.e. callee save registers) unwind the frame and jump to the |
duke@435 | 40 | // exception handler for the nmethod if there is no Java level handler |
duke@435 | 41 | // for the nmethod. |
duke@435 | 42 | // |
duke@435 | 43 | // This code is entered with a jmp. |
duke@435 | 44 | // |
duke@435 | 45 | // Arguments: |
duke@435 | 46 | // rax,: exception oop |
duke@435 | 47 | // rdx: exception pc |
duke@435 | 48 | // |
duke@435 | 49 | // Results: |
duke@435 | 50 | // rax,: exception oop |
duke@435 | 51 | // rdx: exception pc in caller or ??? |
duke@435 | 52 | // destination: exception handler of caller |
duke@435 | 53 | // |
duke@435 | 54 | // Note: the exception pc MUST be at a call (precise debug information) |
duke@435 | 55 | // Only register rax, rdx, rcx are not callee saved. |
duke@435 | 56 | // |
duke@435 | 57 | |
duke@435 | 58 | void OptoRuntime::generate_exception_blob() { |
duke@435 | 59 | |
duke@435 | 60 | // Capture info about frame layout |
duke@435 | 61 | enum layout { |
duke@435 | 62 | thread_off, // last_java_sp |
duke@435 | 63 | // The frame sender code expects that rbp will be in the "natural" place and |
duke@435 | 64 | // will override any oopMap setting for it. We must therefore force the layout |
duke@435 | 65 | // so that it agrees with the frame sender code. |
duke@435 | 66 | rbp_off, |
duke@435 | 67 | return_off, // slot for return address |
duke@435 | 68 | framesize |
duke@435 | 69 | }; |
duke@435 | 70 | |
duke@435 | 71 | // allocate space for the code |
duke@435 | 72 | ResourceMark rm; |
duke@435 | 73 | // setup code generation tools |
duke@435 | 74 | CodeBuffer buffer("exception_blob", 512, 512); |
duke@435 | 75 | MacroAssembler* masm = new MacroAssembler(&buffer); |
duke@435 | 76 | |
duke@435 | 77 | OopMapSet *oop_maps = new OopMapSet(); |
duke@435 | 78 | |
duke@435 | 79 | address start = __ pc(); |
duke@435 | 80 | |
never@739 | 81 | __ push(rdx); |
never@739 | 82 | __ subptr(rsp, return_off * wordSize); // Prolog! |
duke@435 | 83 | |
duke@435 | 84 | // rbp, location is implicitly known |
never@739 | 85 | __ movptr(Address(rsp,rbp_off *wordSize), rbp); |
duke@435 | 86 | |
duke@435 | 87 | // Store exception in Thread object. We cannot pass any arguments to the |
duke@435 | 88 | // handle_exception call, since we do not want to make any assumption |
duke@435 | 89 | // about the size of the frame where the exception happened in. |
duke@435 | 90 | __ get_thread(rcx); |
never@739 | 91 | __ movptr(Address(rcx, JavaThread::exception_oop_offset()), rax); |
never@739 | 92 | __ movptr(Address(rcx, JavaThread::exception_pc_offset()), rdx); |
duke@435 | 93 | |
duke@435 | 94 | // This call does all the hard work. It checks if an exception handler |
duke@435 | 95 | // exists in the method. |
duke@435 | 96 | // If so, it returns the handler address. |
duke@435 | 97 | // If not, it prepares for stack-unwinding, restoring the callee-save |
duke@435 | 98 | // registers of the frame being removed. |
duke@435 | 99 | // |
never@739 | 100 | __ movptr(Address(rsp, thread_off * wordSize), rcx); // Thread is first argument |
duke@435 | 101 | __ set_last_Java_frame(rcx, noreg, noreg, NULL); |
duke@435 | 102 | |
duke@435 | 103 | __ call(RuntimeAddress(CAST_FROM_FN_PTR(address, OptoRuntime::handle_exception_C))); |
duke@435 | 104 | |
duke@435 | 105 | // No registers to map, rbp is known implicitly |
duke@435 | 106 | oop_maps->add_gc_map( __ pc() - start, new OopMap( framesize, 0 )); |
duke@435 | 107 | __ get_thread(rcx); |
duke@435 | 108 | __ reset_last_Java_frame(rcx, false, false); |
duke@435 | 109 | |
duke@435 | 110 | // Restore callee-saved registers |
never@739 | 111 | __ movptr(rbp, Address(rsp, rbp_off * wordSize)); |
duke@435 | 112 | |
never@739 | 113 | __ addptr(rsp, return_off * wordSize); // Epilog! |
never@739 | 114 | __ pop(rdx); // Exception pc |
duke@435 | 115 | |
duke@435 | 116 | |
duke@435 | 117 | // rax,: exception handler for given <exception oop/exception pc> |
duke@435 | 118 | |
duke@435 | 119 | // We have a handler in rax, (could be deopt blob) |
duke@435 | 120 | // rdx - throwing pc, deopt blob will need it. |
duke@435 | 121 | |
never@739 | 122 | __ push(rax); |
duke@435 | 123 | |
duke@435 | 124 | // rcx contains handler address |
duke@435 | 125 | |
duke@435 | 126 | __ get_thread(rcx); // TLS |
duke@435 | 127 | // Get the exception |
never@739 | 128 | __ movptr(rax, Address(rcx, JavaThread::exception_oop_offset())); |
duke@435 | 129 | // Get the exception pc in case we are deoptimized |
never@739 | 130 | __ movptr(rdx, Address(rcx, JavaThread::exception_pc_offset())); |
duke@435 | 131 | #ifdef ASSERT |
never@739 | 132 | __ movptr(Address(rcx, JavaThread::exception_handler_pc_offset()), (int32_t)NULL_WORD); |
never@739 | 133 | __ movptr(Address(rcx, JavaThread::exception_pc_offset()), (int32_t)NULL_WORD); |
duke@435 | 134 | #endif |
duke@435 | 135 | // Clear the exception oop so GC no longer processes it as a root. |
never@739 | 136 | __ movptr(Address(rcx, JavaThread::exception_oop_offset()), (int32_t)NULL_WORD); |
duke@435 | 137 | |
never@739 | 138 | __ pop(rcx); |
duke@435 | 139 | |
duke@435 | 140 | // rax,: exception oop |
duke@435 | 141 | // rcx: exception handler |
duke@435 | 142 | // rdx: exception pc |
duke@435 | 143 | __ jmp (rcx); |
duke@435 | 144 | |
duke@435 | 145 | // ------------- |
duke@435 | 146 | // make sure all code is generated |
duke@435 | 147 | masm->flush(); |
duke@435 | 148 | |
duke@435 | 149 | _exception_blob = ExceptionBlob::create(&buffer, oop_maps, framesize); |
duke@435 | 150 | } |