| Quellcodebibliothek Statistik Leitseite products/Sources/formale Sprachen/Java/Openjdk/src/hotspot/cpu/x86/ (Sun/Oracle ©) Datei vom 29.11.2022 mit Größe 138 kB |
|
SSL stubGenerator_x86_64.cpp Sprache: C |
|||||||||
|
/*
* Copyright (c) 2003, 2022, Oracle and/or its affiliates. All rights reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it * under the terms of the GNU General Public License version 2 only, as * published by the Free Software Foundation. * * This code is distributed in the hope that it will be useful, but WITHOUT * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License * version 2 for more details (a copy is included in the LICENSE file that * accompanied this code). * * You should have received a copy of the GNU General Public License version * 2 along with this work; if not, write to the Free Software Foundation, * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. * * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA * or visit www.oracle.com if you need additional information or have any * questions. * */ #include "precompiled.hpp" #include "asm/macroAssembler.hpp" #include "classfile/vmIntrinsics.hpp" #include "compiler/oopMap.hpp" #include "gc/shared/barrierSet.hpp" #include "gc/shared/barrierSetAssembler.hpp" #include "gc/shared/barrierSetNMethod.hpp" #include "gc/shared/gc_globals.hpp" #include "memory/universe.hpp" #include "prims/jvmtiExport.hpp" #include "runtime/arguments.hpp" #include "runtime/javaThread.hpp" #include "runtime/sharedRuntime.hpp" #include "runtime/stubRoutines.hpp" #include "stubGenerator_x86_64.hpp" #ifdef COMPILER2 #include "opto/runtime.hpp" #include "opto/c2_globals.hpp" #endif #if INCLUDE_JVMCI #include "jvmci/jvmci_globals.hpp" #endif #if INCLUDE_ZGC #include "gc/z/zThreadLocalData.hpp" #endif #if INCLUDE_JFR #include "jfr/support/jfrIntrinsics.hpp" #endif // For a more detailed description of the stub routine structure // see the comment in stubRoutines.hpp #define __ _masm-> #define TIMES_OOP (UseCompressedOops ? Address::times_4 : Address::times_8) #ifdef PRODUCT #define BLOCK_COMMENT(str) /* nothing */ #else #define BLOCK_COMMENT(str) __ block_comment(str) #endif // PRODUCT #define BIND(label) bind(label); BLOCK_COMMENT(#label ":") // // Linux Arguments: // c_rarg0: call wrapper address address // c_rarg1: result address // c_rarg2: result type BasicType // c_rarg3: method Method* // c_rarg4: (interpreter) entry point address // c_rarg5: parameters intptr_t* // 16(rbp): parameter size (in words) int // 24(rbp): thread Thread* // // [ return_from_Java ] <--- rsp // [ argument word n ] // ... // -12 [ argument word 1 ] // -11 [ saved r15 ] <--- rsp_after_call // -10 [ saved r14 ] // -9 [ saved r13 ] // -8 [ saved r12 ] // -7 [ saved rbx ] // -6 [ call wrapper ] // -5 [ result ] // -4 [ result type ] // -3 [ method ] // -2 [ entry point ] // -1 [ parameters ] // 0 [ saved rbp ] <--- rbp // 1 [ return address ] // 2 [ parameter size ] // 3 [ thread ] // // Windows Arguments: // c_rarg0: call wrapper address address // c_rarg1: result address // c_rarg2: result type BasicType // c_rarg3: method Method* // 48(rbp): (interpreter) entry point address // 56(rbp): parameters intptr_t* // 64(rbp): parameter size (in words) int // 72(rbp): thread Thread* // // [ return_from_Java ] <--- rsp // [ argument word n ] // ... // -60 [ argument word 1 ] // -59 [ saved xmm31 ] <--- rsp after_call // [ saved xmm16-xmm30 ] (EVEX enabled, else the space is blank) // -27 [ saved xmm15 ] // [ saved xmm7-xmm14 ] // -9 [ saved xmm6 ] (each xmm register takes 2 slots) // -7 [ saved r15 ] // -6 [ saved r14 ] // -5 [ saved r13 ] // -4 [ saved r12 ] // -3 [ saved rdi ] // -2 [ saved rsi ] // -1 [ saved rbx ] // 0 [ saved rbp ] <--- rbp // 1 [ return address ] // 2 [ call wrapper ] // 3 [ result ] // 4 [ result type ] // 5 [ method ] // 6 [ entry point ] // 7 [ parameters ] // 8 [ parameter size ] // 9 [ thread ] // // Windows reserves the callers stack space for arguments 1-4. // We spill c_rarg0-c_rarg3 to this space. // Call stub stack layout word offsets from rbp #ifdef _WIN64 enum call_stub_layout { xmm_save_first = 6, // save from xmm6 xmm_save_last = 31, // to xmm31 xmm_save_base = -9, rsp_after_call_off = xmm_save_base - 2 * (xmm_save_last - xmm_save_first), // -27 r15_off = -7, r14_off = -6, r13_off = -5, r12_off = -4, rdi_off = -3, rsi_off = -2, rbx_off = -1, rbp_off = 0, retaddr_off = 1, call_wrapper_off = 2, result_off = 3, result_type_off = 4, method_off = 5, entry_point_off = 6, parameters_off = 7, parameter_size_off = 8, thread_off = 9 }; static Address xmm_save(int reg) { assert(reg >= xmm_save_first && reg <= xmm_save_last, "XMM register number out of range"); return Address(rbp, (xmm_save_base - (reg - xmm_save_first) * 2) * wordSize); } #else // !_WIN64 enum call_stub_layout { rsp_after_call_off = -12, mxcsr_off = rsp_after_call_off, r15_off = -11, r14_off = -10, r13_off = -9, r12_off = -8, rbx_off = -7, call_wrapper_off = -6, result_off = -5, result_type_off = -4, method_off = -3, entry_point_off = -2, parameters_off = -1, rbp_off = 0, retaddr_off = 1, parameter_size_off = 2, thread_off = 3 }; #endif // _WIN64 address StubGenerator::generate_call_stub(address& return_address) { assert((int)frame::entry_frame_after_call_words == -(int)rsp_after_call_off + 1 && (int)frame::entry_frame_call_wrapper_offset == (int)call_wrapper_off, "adjust this code"); StubCodeMark mark(this, "StubRoutines", "call_stub"); address start = __ pc(); // same as in generate_catch_exception()! const Address rsp_after_call(rbp, rsp_after_call_off * wordSize); const Address call_wrapper (rbp, call_wrapper_off * wordSize); const Address result (rbp, result_off * wordSize); const Address result_type (rbp, result_type_off * wordSize); const Address method (rbp, method_off * wordSize); const Address entry_point (rbp, entry_point_off * wordSize); const Address parameters (rbp, parameters_off * wordSize); const Address parameter_size(rbp, parameter_size_off * wordSize); // same as in generate_catch_exception()! const Address thread (rbp, thread_off * wordSize); const Address r15_save(rbp, r15_off * wordSize); const Address r14_save(rbp, r14_off * wordSize); const Address r13_save(rbp, r13_off * wordSize); const Address r12_save(rbp, r12_off * wordSize); const Address rbx_save(rbp, rbx_off * wordSize); // stub code __ enter(); __ subptr(rsp, -rsp_after_call_off * wordSize); // save register parameters #ifndef _WIN64 __ movptr(parameters, c_rarg5); // parameters __ movptr(entry_point, c_rarg4); // entry_point #endif __ movptr(method, c_rarg3); // method __ movl(result_type, c_rarg2); // result type __ movptr(result, c_rarg1); // result __ movptr(call_wrapper, c_rarg0); // call wrapper // save regs belonging to calling function __ movptr(rbx_save, rbx); __ movptr(r12_save, r12); __ movptr(r13_save, r13); __ movptr(r14_save, r14); __ movptr(r15_save, r15); #ifdef _WIN64 int last_reg = 15; if (UseAVX > 2) { last_reg = 31; } if (VM_Version::supports_evex()) { for (int i = xmm_save_first; i <= last_reg; i++) { __ vextractf32x4(xmm_save(i), as_XMMRegister(i), 0); } } else { for (int i = xmm_save_first; i <= last_reg; i++) { __ movdqu(xmm_save(i), as_XMMRegister(i)); } } const Address rdi_save(rbp, rdi_off * wordSize); const Address rsi_save(rbp, rsi_off * wordSize); __ movptr(rsi_save, rsi); __ movptr(rdi_save, rdi); #else const Address mxcsr_save(rbp, mxcsr_off * wordSize); { Label skip_ldmx; __ stmxcsr(mxcsr_save); __ movl(rax, mxcsr_save); __ andl(rax, 0xFFC0); // Mask out any pending exceptions (only check control and mask bits) ExternalAddress mxcsr_std(StubRoutines::x86::addr_mxcsr_std()); __ cmp32(rax, mxcsr_std, rscratch1); __ jcc(Assembler::equal, skip_ldmx); __ ldmxcsr(mxcsr_std, rscratch1); __ bind(skip_ldmx); } #endif // Load up thread register __ movptr(r15_thread, thread); __ reinit_heapbase(); #ifdef ASSERT // make sure we have no pending exceptions { Label L; __ cmpptr(Address(r15_thread, Thread::pending_exception_offset()), NULL_WORD); __ jcc(Assembler::equal, L); __ stop("StubRoutines::call_stub: entered with pending exception"); __ bind(L); } #endif // pass parameters if any BLOCK_COMMENT("pass parameters if any"); Label parameters_done; __ movl(c_rarg3, parameter_size); __ testl(c_rarg3, c_rarg3); __ jcc(Assembler::zero, parameters_done); Label loop; __ movptr(c_rarg2, parameters); // parameter pointer __ movl(c_rarg1, c_rarg3); // parameter counter is in c_rarg1 __ BIND(loop); __ movptr(rax, Address(c_rarg2, 0));// get parameter __ addptr(c_rarg2, wordSize); // advance to next parameter __ decrementl(c_rarg1); // decrement counter __ push(rax); // pass parameter __ jcc(Assembler::notZero, loop); // call Java function __ BIND(parameters_done); __ movptr(rbx, method); // get Method* __ movptr(c_rarg1, entry_point); // get entry_point __ mov(r13, rsp); // set sender sp BLOCK_COMMENT("call Java function"); __ call(c_rarg1); BLOCK_COMMENT("call_stub_return_address:"); return_address = __ pc(); // store result depending on type (everything that is not // T_OBJECT, T_LONG, T_FLOAT or T_DOUBLE is treated as T_INT) __ movptr(c_rarg0, result); Label is_long, is_float, is_double, exit; __ movl(c_rarg1, result_type); __ cmpl(c_rarg1, T_OBJECT); __ jcc(Assembler::equal, is_long); __ cmpl(c_rarg1, T_LONG); __ jcc(Assembler::equal, is_long); __ cmpl(c_rarg1, T_FLOAT); __ jcc(Assembler::equal, is_float); __ cmpl(c_rarg1, T_DOUBLE); __ jcc(Assembler::equal, is_double); // handle T_INT case __ movl(Address(c_rarg0, 0), rax); __ BIND(exit); // pop parameters __ lea(rsp, rsp_after_call); #ifdef ASSERT // verify that threads correspond { Label L1, L2, L3; __ cmpptr(r15_thread, thread); __ jcc(Assembler::equal, L1); __ stop("StubRoutines::call_stub: r15_thread is corrupted"); __ bind(L1); __ get_thread(rbx); __ cmpptr(r15_thread, thread); __ jcc(Assembler::equal, L2); __ stop("StubRoutines::call_stub: r15_thread is modified by call"); __ bind(L2); __ cmpptr(r15_thread, rbx); __ jcc(Assembler::equal, L3); __ stop("StubRoutines::call_stub: threads must correspond"); __ bind(L3); } #endif __ pop_cont_fastpath(); // restore regs belonging to calling function #ifdef _WIN64 // emit the restores for xmm regs if (VM_Version::supports_evex()) { for (int i = xmm_save_first; i <= last_reg; i++) { __ vinsertf32x4(as_XMMRegister(i), as_XMMRegister(i), xmm_save(i), 0); } } else { for (int i = xmm_save_first; i <= last_reg; i++) { __ movdqu(as_XMMRegister(i), xmm_save(i)); } } #endif __ movptr(r15, r15_save); __ movptr(r14, r14_save); __ movptr(r13, r13_save); __ movptr(r12, r12_save); __ movptr(rbx, rbx_save); #ifdef _WIN64 __ movptr(rdi, rdi_save); __ movptr(rsi, rsi_save); #else __ ldmxcsr(mxcsr_save); #endif // restore rsp __ addptr(rsp, -rsp_after_call_off * wordSize); // return __ vzeroupper(); __ pop(rbp); __ ret(0); // handle return types different from T_INT __ BIND(is_long); __ movq(Address(c_rarg0, 0), rax); __ jmp(exit); __ BIND(is_float); __ movflt(Address(c_rarg0, 0), xmm0); __ jmp(exit); __ BIND(is_double); __ movdbl(Address(c_rarg0, 0), xmm0); __ jmp(exit); return start; } // Return point for a Java call if there's an exception thrown in // Java code. The exception is caught and transformed into a // pending exception stored in JavaThread that can be tested from // within the VM. // // Note: Usually the parameters are removed by the callee. In case // of an exception crossing an activation frame boundary, that is // not the case if the callee is compiled code => need to setup the // rsp. // // rax: exception oop address StubGenerator::generate_catch_exception() { StubCodeMark mark(this, "StubRoutines", "catch_exception"); address start = __ pc(); // same as in generate_call_stub(): const Address rsp_after_call(rbp, rsp_after_call_off * wordSize); const Address thread (rbp, thread_off * wordSize); #ifdef ASSERT // verify that threads correspond { Label L1, L2, L3; __ cmpptr(r15_thread, thread); __ jcc(Assembler::equal, L1); __ stop("StubRoutines::catch_exception: r15_thread is corrupted"); __ bind(L1); __ get_thread(rbx); __ cmpptr(r15_thread, thread); __ jcc(Assembler::equal, L2); __ stop("StubRoutines::catch_exception: r15_thread is modified by call"); __ bind(L2); __ cmpptr(r15_thread, rbx); __ jcc(Assembler::equal, L3); __ stop("StubRoutines::catch_exception: threads must correspond"); __ bind(L3); } #endif // set pending exception __ verify_oop(rax); __ movptr(Address(r15_thread, Thread::pending_exception_offset()), rax); __ lea(rscratch1, ExternalAddress((address)__FILE__)); __ movptr(Address(r15_thread, Thread::exception_file_offset()), rscratch1); __ movl(Address(r15_thread, Thread::exception_line_offset()), (int) __LINE__); // complete return to VM assert(StubRoutines::_call_stub_return_address != NULL, "_call_stub_return_address must have been generated before"); __ jump(RuntimeAddress(StubRoutines::_call_stub_return_address)); return start; } // Continuation point for runtime calls returning with a pending // exception. The pending exception check happened in the runtime // or native call stub. The pending exception in Thread is // converted into a Java-level exception. // // Contract with Java-level exception handlers: // rax: exception // rdx: throwing pc // // NOTE: At entry of this stub, exception-pc must be on stack !! address StubGenerator::generate_forward_exception() { StubCodeMark mark(this, "StubRoutines", "forward exception"); address start = __ pc(); // Upon entry, the sp points to the return address returning into // Java (interpreted or compiled) code; i.e., the return address // becomes the throwing pc. // // Arguments pushed before the runtime call are still on the stack // but the exception handler will reset the stack pointer -> // ignore them. A potential result in registers can be ignored as // well. #ifdef ASSERT // make sure this code is only executed if there is a pending exception { Label L; __ cmpptr(Address(r15_thread, Thread::pending_exception_offset()), NULL_WORD); __ jcc(Assembler::notEqual, L); __ stop("StubRoutines::forward exception: no pending exception (1)"); __ bind(L); } #endif // compute exception handler into rbx __ movptr(c_rarg0, Address(rsp, 0)); BLOCK_COMMENT("call exception_handler_for_return_address"); __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::exception_handler_for_return_address), r15_thread, c_rarg0); __ mov(rbx, rax); // setup rax & rdx, remove return address & clear pending exception __ pop(rdx); __ movptr(rax, Address(r15_thread, Thread::pending_exception_offset())); __ movptr(Address(r15_thread, Thread::pending_exception_offset()), NULL_WORD); #ifdef ASSERT // make sure exception is set { Label L; __ testptr(rax, rax); __ jcc(Assembler::notEqual, L); __ stop("StubRoutines::forward exception: no pending exception (2)"); __ bind(L); } #endif // continue at exception handler (return address removed) // rax: exception // rbx: exception handler // rdx: throwing pc __ verify_oop(rax); __ jmp(rbx); return start; } // Support for intptr_t OrderAccess::fence() // // Arguments : // // Result: address StubGenerator::generate_orderaccess_fence() { StubCodeMark mark(this, "StubRoutines", "orderaccess_fence"); address start = __ pc(); __ membar(Assembler::StoreLoad); __ ret(0); return start; } // Support for intptr_t get_previous_sp() // // This routine is used to find the previous stack pointer for the // caller. address StubGenerator::generate_get_previous_sp() { StubCodeMark mark(this, "StubRoutines", "get_previous_sp"); address start = __ pc(); __ movptr(rax, rsp); __ addptr(rax, 8); // return address is at the top of the stack. __ ret(0); return start; } //------------------------------------------------------------------------------ // Support for void verify_mxcsr() // // This routine is used with -Xcheck:jni to verify that native // JNI code does not return to Java code without restoring the // MXCSR register to our expected state. address StubGenerator::generate_verify_mxcsr() { StubCodeMark mark(this, "StubRoutines", "verify_mxcsr"); address start = __ pc(); const Address mxcsr_save(rsp, 0); if (CheckJNICalls) { Label ok_ret; ExternalAddress mxcsr_std(StubRoutines::x86::addr_mxcsr_std()); __ push(rax); __ subptr(rsp, wordSize); // allocate a temp location __ stmxcsr(mxcsr_save); __ movl(rax, mxcsr_save); __ andl(rax, 0xFFC0); // Mask out any pending exceptions (only check control and mask bits) __ cmp32(rax, mxcsr_std, rscratch1); __ jcc(Assembler::equal, ok_ret); __ warn("MXCSR changed by native JNI code, use -XX:+RestoreMXCSROnJNICall"); __ ldmxcsr(mxcsr_std, rscratch1); __ bind(ok_ret); __ addptr(rsp, wordSize); __ pop(rax); } __ ret(0); return start; } address StubGenerator::generate_f2i_fixup() { StubCodeMark mark(this, "StubRoutines", "f2i_fixup"); Address inout(rsp, 5 * wordSize); // return address + 4 saves address start = __ pc(); Label L; __ push(rax); __ push(c_rarg3); __ push(c_rarg2); __ push(c_rarg1); __ movl(rax, 0x7f800000); __ xorl(c_rarg3, c_rarg3); __ movl(c_rarg2, inout); __ movl(c_rarg1, c_rarg2); __ andl(c_rarg1, 0x7fffffff); __ cmpl(rax, c_rarg1); // NaN? -> 0 __ jcc(Assembler::negative, L); __ testl(c_rarg2, c_rarg2); // signed ? min_jint : max_jint __ movl(c_rarg3, 0x80000000); __ movl(rax, 0x7fffffff); __ cmovl(Assembler::positive, c_rarg3, rax); __ bind(L); __ movptr(inout, c_rarg3); __ pop(c_rarg1); __ pop(c_rarg2); __ pop(c_rarg3); __ pop(rax); __ ret(0); return start; } address StubGenerator::generate_f2l_fixup() { StubCodeMark mark(this, "StubRoutines", "f2l_fixup"); Address inout(rsp, 5 * wordSize); // return address + 4 saves address start = __ pc(); Label L; __ push(rax); __ push(c_rarg3); __ push(c_rarg2); __ push(c_rarg1); __ movl(rax, 0x7f800000); __ xorl(c_rarg3, c_rarg3); __ movl(c_rarg2, inout); __ movl(c_rarg1, c_rarg2); __ andl(c_rarg1, 0x7fffffff); __ cmpl(rax, c_rarg1); // NaN? -> 0 __ jcc(Assembler::negative, L); __ testl(c_rarg2, c_rarg2); // signed ? min_jlong : max_jlong __ mov64(c_rarg3, 0x8000000000000000); __ mov64(rax, 0x7fffffffffffffff); __ cmov(Assembler::positive, c_rarg3, rax); __ bind(L); __ movptr(inout, c_rarg3); __ pop(c_rarg1); __ pop(c_rarg2); __ pop(c_rarg3); __ pop(rax); __ ret(0); return start; } address StubGenerator::generate_d2i_fixup() { StubCodeMark mark(this, "StubRoutines", "d2i_fixup"); Address inout(rsp, 6 * wordSize); // return address + 5 saves address start = __ pc(); Label L; __ push(rax); __ push(c_rarg3); __ push(c_rarg2); __ push(c_rarg1); __ push(c_rarg0); __ movl(rax, 0x7ff00000); __ movq(c_rarg2, inout); __ movl(c_rarg3, c_rarg2); __ mov(c_rarg1, c_rarg2); __ mov(c_rarg0, c_rarg2); __ negl(c_rarg3); __ shrptr(c_rarg1, 0x20); __ orl(c_rarg3, c_rarg2); __ andl(c_rarg1, 0x7fffffff); __ xorl(c_rarg2, c_rarg2); __ shrl(c_rarg3, 0x1f); __ orl(c_rarg1, c_rarg3); __ cmpl(rax, c_rarg1); __ jcc(Assembler::negative, L); // NaN -> 0 __ testptr(c_rarg0, c_rarg0); // signed ? min_jint : max_jint __ movl(c_rarg2, 0x80000000); __ movl(rax, 0x7fffffff); __ cmov(Assembler::positive, c_rarg2, rax); __ bind(L); __ movptr(inout, c_rarg2); __ pop(c_rarg0); __ pop(c_rarg1); __ pop(c_rarg2); __ pop(c_rarg3); __ pop(rax); __ ret(0); return start; } address StubGenerator::generate_d2l_fixup() { StubCodeMark mark(this, "StubRoutines", "d2l_fixup"); Address inout(rsp, 6 * wordSize); // return address + 5 saves address start = __ pc(); Label L; __ push(rax); __ push(c_rarg3); __ push(c_rarg2); __ push(c_rarg1); __ push(c_rarg0); __ movl(rax, 0x7ff00000); __ movq(c_rarg2, inout); __ movl(c_rarg3, c_rarg2); __ mov(c_rarg1, c_rarg2); __ mov(c_rarg0, c_rarg2); __ negl(c_rarg3); __ shrptr(c_rarg1, 0x20); __ orl(c_rarg3, c_rarg2); __ andl(c_rarg1, 0x7fffffff); __ xorl(c_rarg2, c_rarg2); __ shrl(c_rarg3, 0x1f); __ orl(c_rarg1, c_rarg3); __ cmpl(rax, c_rarg1); __ jcc(Assembler::negative, L); // NaN -> 0 __ testq(c_rarg0, c_rarg0); // signed ? min_jlong : max_jlong __ mov64(c_rarg2, 0x8000000000000000); __ mov64(rax, 0x7fffffffffffffff); __ cmovq(Assembler::positive, c_rarg2, rax); __ bind(L); __ movq(inout, c_rarg2); __ pop(c_rarg0); __ pop(c_rarg1); __ pop(c_rarg2); __ pop(c_rarg3); __ pop(rax); __ ret(0); return start; } address StubGenerator::generate_count_leading_zeros_lut(const char *stub_name) { __ align64(); StubCodeMark mark(this, "StubRoutines", stub_name); address start = __ pc(); __ emit_data64(0x0101010102020304, relocInfo::none); __ emit_data64(0x0000000000000000, relocInfo::none); __ emit_data64(0x0101010102020304, relocInfo::none); __ emit_data64(0x0000000000000000, relocInfo::none); __ emit_data64(0x0101010102020304, relocInfo::none); __ emit_data64(0x0000000000000000, relocInfo::none); __ emit_data64(0x0101010102020304, relocInfo::none); __ emit_data64(0x0000000000000000, relocInfo::none); return start; } address StubGenerator::generate_popcount_avx_lut(const char *stub_name) { __ align64(); StubCodeMark mark(this, "StubRoutines", stub_name); address start = __ pc(); __ emit_data64(0x0302020102010100, relocInfo::none); __ emit_data64(0x0403030203020201, relocInfo::none); __ emit_data64(0x0302020102010100, relocInfo::none); __ emit_data64(0x0403030203020201, relocInfo::none); __ emit_data64(0x0302020102010100, relocInfo::none); __ emit_data64(0x0403030203020201, relocInfo::none); __ emit_data64(0x0302020102010100, relocInfo::none); __ emit_data64(0x0403030203020201, relocInfo::none); return start; } address StubGenerator::generate_iota_indices(const char *stub_name) { __ align(CodeEntryAlignment); StubCodeMark mark(this, "StubRoutines", stub_name); address start = __ pc(); // B __ emit_data64(0x0706050403020100, relocInfo::none); __ emit_data64(0x0F0E0D0C0B0A0908, relocInfo::none); __ emit_data64(0x1716151413121110, relocInfo::none); __ emit_data64(0x1F1E1D1C1B1A1918, relocInfo::none); __ emit_data64(0x2726252423222120, relocInfo::none); __ emit_data64(0x2F2E2D2C2B2A2928, relocInfo::none); __ emit_data64(0x3736353433323130, relocInfo::none); __ emit_data64(0x3F3E3D3C3B3A3938, relocInfo::none); // W __ emit_data64(0x0003000200010000, relocInfo::none); __ emit_data64(0x0007000600050004, relocInfo::none); __ emit_data64(0x000B000A00090008, relocInfo::none); __ emit_data64(0x000F000E000D000C, relocInfo::none); __ emit_data64(0x0013001200110010, relocInfo::none); __ emit_data64(0x0017001600150014, relocInfo::none); __ emit_data64(0x001B001A00190018, relocInfo::none); __ emit_data64(0x001F001E001D001C, relocInfo::none); // D __ emit_data64(0x0000000100000000, relocInfo::none); __ emit_data64(0x0000000300000002, relocInfo::none); __ emit_data64(0x0000000500000004, relocInfo::none); __ emit_data64(0x0000000700000006, relocInfo::none); __ emit_data64(0x0000000900000008, relocInfo::none); __ emit_data64(0x0000000B0000000A, relocInfo::none); __ emit_data64(0x0000000D0000000C, relocInfo::none); __ emit_data64(0x0000000F0000000E, relocInfo::none); // Q __ emit_data64(0x0000000000000000, relocInfo::none); __ emit_data64(0x0000000000000001, relocInfo::none); __ emit_data64(0x0000000000000002, relocInfo::none); __ emit_data64(0x0000000000000003, relocInfo::none); __ emit_data64(0x0000000000000004, relocInfo::none); __ emit_data64(0x0000000000000005, relocInfo::none); __ emit_data64(0x0000000000000006, relocInfo::none); __ emit_data64(0x0000000000000007, relocInfo::none); // D - FP __ emit_data64(0x3F80000000000000, relocInfo::none); // 0.0f, 1.0f __ emit_data64(0x4040000040000000, relocInfo::none); // 2.0f, 3.0f __ emit_data64(0x40A0000040800000, relocInfo::none); // 4.0f, 5.0f __ emit_data64(0x40E0000040C00000, relocInfo::none); // 6.0f, 7.0f __ emit_data64(0x4110000041000000, relocInfo::none); // 8.0f, 9.0f __ emit_data64(0x4130000041200000, relocInfo::none); // 10.0f, 11.0f __ emit_data64(0x4150000041400000, relocInfo::none); // 12.0f, 13.0f __ emit_data64(0x4170000041600000, relocInfo::none); // 14.0f, 15.0f // Q - FP __ emit_data64(0x0000000000000000, relocInfo::none); // 0.0d __ emit_data64(0x3FF0000000000000, relocInfo::none); // 1.0d __ emit_data64(0x4000000000000000, relocInfo::none); // 2.0d __ emit_data64(0x4008000000000000, relocInfo::none); // 3.0d __ emit_data64(0x4010000000000000, relocInfo::none); // 4.0d __ emit_data64(0x4014000000000000, relocInfo::none); // 5.0d __ emit_data64(0x4018000000000000, relocInfo::none); // 6.0d __ emit_data64(0x401c000000000000, relocInfo::none); // 7.0d return start; } address StubGenerator::generate_vector_reverse_bit_lut(const char *stub_name) { __ align(CodeEntryAlignment); StubCodeMark mark(this, "StubRoutines", stub_name); address start = __ pc(); __ emit_data64(0x0E060A020C040800, relocInfo::none); __ emit_data64(0x0F070B030D050901, relocInfo::none); __ emit_data64(0x0E060A020C040800, relocInfo::none); __ emit_data64(0x0F070B030D050901, relocInfo::none); __ emit_data64(0x0E060A020C040800, relocInfo::none); __ emit_data64(0x0F070B030D050901, relocInfo::none); __ emit_data64(0x0E060A020C040800, relocInfo::none); __ emit_data64(0x0F070B030D050901, relocInfo::none); return start; } address StubGenerator::generate_vector_reverse_byte_perm_mask_long(const char *stub __ align(CodeEntryAlignment); StubCodeMark mark(this, "StubRoutines", stub_name); address start = __ pc(); __ emit_data64(0x0001020304050607, relocInfo::none); __ emit_data64(0x08090A0B0C0D0E0F, relocInfo::none); __ emit_data64(0x0001020304050607, relocInfo::none); __ emit_data64(0x08090A0B0C0D0E0F, relocInfo::none); __ emit_data64(0x0001020304050607, relocInfo::none); __ emit_data64(0x08090A0B0C0D0E0F, relocInfo::none); __ emit_data64(0x0001020304050607, relocInfo::none); __ emit_data64(0x08090A0B0C0D0E0F, relocInfo::none); return start; } address StubGenerator::generate_vector_reverse_byte_perm_mask_int(const char *stub_ __ align(CodeEntryAlignment); StubCodeMark mark(this, "StubRoutines", stub_name); address start = __ pc(); __ emit_data64(0x0405060700010203, relocInfo::none); __ emit_data64(0x0C0D0E0F08090A0B, relocInfo::none); __ emit_data64(0x0405060700010203, relocInfo::none); __ emit_data64(0x0C0D0E0F08090A0B, relocInfo::none); __ emit_data64(0x0405060700010203, relocInfo::none); __ emit_data64(0x0C0D0E0F08090A0B, relocInfo::none); __ emit_data64(0x0405060700010203, relocInfo::none); __ emit_data64(0x0C0D0E0F08090A0B, relocInfo::none); return start; } address StubGenerator::generate_vector_reverse_byte_perm_mask_short(const char *stu __ align(CodeEntryAlignment); StubCodeMark mark(this, "StubRoutines", stub_name); address start = __ pc(); __ emit_data64(0x0607040502030001, relocInfo::none); __ emit_data64(0x0E0F0C0D0A0B0809, relocInfo::none); __ emit_data64(0x0607040502030001, relocInfo::none); __ emit_data64(0x0E0F0C0D0A0B0809, relocInfo::none); __ emit_data64(0x0607040502030001, relocInfo::none); __ emit_data64(0x0E0F0C0D0A0B0809, relocInfo::none); __ emit_data64(0x0607040502030001, relocInfo::none); __ emit_data64(0x0E0F0C0D0A0B0809, relocInfo::none); return start; } address StubGenerator::generate_vector_byte_shuffle_mask(const char *stub_name) { __ align(CodeEntryAlignment); StubCodeMark mark(this, "StubRoutines", stub_name); address start = __ pc(); __ emit_data64(0x7070707070707070, relocInfo::none); __ emit_data64(0x7070707070707070, relocInfo::none); __ emit_data64(0xF0F0F0F0F0F0F0F0, relocInfo::none); __ emit_data64(0xF0F0F0F0F0F0F0F0, relocInfo::none); return start; } address StubGenerator::generate_fp_mask(const char *stub_name, int64_t mask) { __ align(CodeEntryAlignment); StubCodeMark mark(this, "StubRoutines", stub_name); address start = __ pc(); __ emit_data64( mask, relocInfo::none ); __ emit_data64( mask, relocInfo::none ); return start; } address StubGenerator::generate_vector_mask(const char *stub_name, int64_t mask) { __ align(CodeEntryAlignment); StubCodeMark mark(this, "StubRoutines", stub_name); address start = __ pc(); __ emit_data64(mask, relocInfo::none); __ emit_data64(mask, relocInfo::none); __ emit_data64(mask, relocInfo::none); __ emit_data64(mask, relocInfo::none); __ emit_data64(mask, relocInfo::none); __ emit_data64(mask, relocInfo::none); __ emit_data64(mask, relocInfo::none); __ emit_data64(mask, relocInfo::none); return start; } address StubGenerator::generate_vector_byte_perm_mask(const char *stub_name) { __ align(CodeEntryAlignment); StubCodeMark mark(this, "StubRoutines", stub_name); address start = __ pc(); __ emit_data64(0x0000000000000001, relocInfo::none); __ emit_data64(0x0000000000000003, relocInfo::none); __ emit_data64(0x0000000000000005, relocInfo::none); __ emit_data64(0x0000000000000007, relocInfo::none); __ emit_data64(0x0000000000000000, relocInfo::none); __ emit_data64(0x0000000000000002, relocInfo::none); __ emit_data64(0x0000000000000004, relocInfo::none); __ emit_data64(0x0000000000000006, relocInfo::none); return start; } address StubGenerator::generate_vector_fp_mask(const char *stub_name, int64_t mask) { __ align(CodeEntryAlignment); StubCodeMark mark(this, "StubRoutines", stub_name); address start = __ pc(); __ emit_data64(mask, relocInfo::none); __ emit_data64(mask, relocInfo::none); __ emit_data64(mask, relocInfo::none); __ emit_data64(mask, relocInfo::none); __ emit_data64(mask, relocInfo::none); __ emit_data64(mask, relocInfo::none); __ emit_data64(mask, relocInfo::none); __ emit_data64(mask, relocInfo::none); return start; } address StubGenerator::generate_vector_custom_i32(const char *stub_name, Assembler:: int32_t val0, int32_t val1, int32_t val2, int32_t val3, int32_t val4, int32_t val5, int32_t val6, int32_t val7, int32_t val8, int32_t val9, int32_t val10, int32_t val11, int32_t val12, int32_t val13, int32_t val14, int32_t val15) { __ align(CodeEntryAlignment); StubCodeMark mark(this, "StubRoutines", stub_name); address start = __ pc(); assert(len != Assembler::AVX_NoVec, "vector len must be specified"); __ emit_data(val0, relocInfo::none, 0); __ emit_data(val1, relocInfo::none, 0); __ emit_data(val2, relocInfo::none, 0); __ emit_data(val3, relocInfo::none, 0); if (len >= Assembler::AVX_256bit) { __ emit_data(val4, relocInfo::none, 0); __ emit_data(val5, relocInfo::none, 0); __ emit_data(val6, relocInfo::none, 0); __ emit_data(val7, relocInfo::none, 0); if (len >= Assembler::AVX_512bit) { __ emit_data(val8, relocInfo::none, 0); __ emit_data(val9, relocInfo::none, 0); __ emit_data(val10, relocInfo::none, 0); __ emit_data(val11, relocInfo::none, 0); __ emit_data(val12, relocInfo::none, 0); __ emit_data(val13, relocInfo::none, 0); __ emit_data(val14, relocInfo::none, 0); __ emit_data(val15, relocInfo::none, 0); } } return start; } // Non-destructive plausibility checks for oops // // Arguments: // all args on stack! // // Stack after saving c_rarg3: // [tos + 0]: saved c_rarg3 // [tos + 1]: saved c_rarg2 // [tos + 2]: saved r12 (several TemplateTable methods use it) // [tos + 3]: saved flags // [tos + 4]: return address // * [tos + 5]: error message (char*) // * [tos + 6]: object to verify (oop) // * [tos + 7]: saved rax - saved by caller and bashed // * [tos + 8]: saved r10 (rscratch1) - saved by caller // * = popped on exit address StubGenerator::generate_verify_oop() { StubCodeMark mark(this, "StubRoutines", "verify_oop"); address start = __ pc(); Label exit, error; __ pushf(); __ incrementl(ExternalAddress((address) StubRoutines::verify_oop_count_addr()), rsc __ push(r12); // save c_rarg2 and c_rarg3 __ push(c_rarg2); __ push(c_rarg3); enum { // After previous pushes. oop_to_verify = 6 * wordSize, saved_rax = 7 * wordSize, saved_r10 = 8 * wordSize, // Before the call to MacroAssembler::debug(), see below. return_addr = 16 * wordSize, error_msg = 17 * wordSize }; // get object __ movptr(rax, Address(rsp, oop_to_verify)); // make sure object is 'reasonable' __ testptr(rax, rax); __ jcc(Assembler::zero, exit); // if obj is NULL it is OK #if INCLUDE_ZGC if (UseZGC) { // Check if metadata bits indicate a bad oop __ testptr(rax, Address(r15_thread, ZThreadLocalData::address_bad_mask_offset())); __ jcc(Assembler::notZero, error); } #endif // Check if the oop is in the right area of memory __ movptr(c_rarg2, rax); __ movptr(c_rarg3, (intptr_t) Universe::verify_oop_mask()); __ andptr(c_rarg2, c_rarg3); __ movptr(c_rarg3, (intptr_t) Universe::verify_oop_bits()); __ cmpptr(c_rarg2, c_rarg3); __ jcc(Assembler::notZero, error); // make sure klass is 'reasonable', which is not zero. __ load_klass(rax, rax, rscratch1); // get klass __ testptr(rax, rax); __ jcc(Assembler::zero, error); // if klass is NULL it is broken // return if everything seems ok __ bind(exit); __ movptr(rax, Address(rsp, saved_rax)); // get saved rax back __ movptr(rscratch1, Address(rsp, saved_r10)); // get saved r10 back __ pop(c_rarg3); // restore c_rarg3 __ pop(c_rarg2); // restore c_rarg2 __ pop(r12); // restore r12 __ popf(); // restore flags __ ret(4 * wordSize); // pop caller saved stuff // handle errors __ bind(error); __ movptr(rax, Address(rsp, saved_rax)); // get saved rax back __ movptr(rscratch1, Address(rsp, saved_r10)); // get saved r10 back __ pop(c_rarg3); // get saved c_rarg3 back __ pop(c_rarg2); // get saved c_rarg2 back __ pop(r12); // get saved r12 back __ popf(); // get saved flags off stack -- // will be ignored __ pusha(); // push registers // (rip is already // already pushed) // debug(char* msg, int64_t pc, int64_t regs[]) // We've popped the registers we'd saved (c_rarg3, c_rarg2 and flags), and // pushed all the registers, so now the stack looks like: // [tos + 0] 16 saved registers // [tos + 16] return address // * [tos + 17] error message (char*) // * [tos + 18] object to verify (oop) // * [tos + 19] saved rax - saved by caller and bashed // * [tos + 20] saved r10 (rscratch1) - saved by caller // * = popped on exit __ movptr(c_rarg0, Address(rsp, error_msg)); // pass address of error message __ movptr(c_rarg1, Address(rsp, return_addr)); // pass return address __ movq(c_rarg2, rsp); // pass address of regs on stack __ mov(r12, rsp); // remember rsp __ subptr(rsp, frame::arg_reg_save_area_bytes); // windows __ andptr(rsp, -16); // align stack as required by ABI BLOCK_COMMENT("call MacroAssembler::debug"); __ call(RuntimeAddress(CAST_FROM_FN_PTR(address, MacroAssembler::debug64))); __ hlt(); return start; } // Shuffle first three arg regs on Windows into Linux/Solaris locations. // // Outputs: // rdi - rcx // rsi - rdx // rdx - r8 // rcx - r9 // // Registers r9 and r10 are used to save rdi and rsi on Windows, which latter // are non-volatile. r9 and r10 should not be used by the caller. // void StubGenerator::setup_arg_regs(int nargs) { const Register saved_rdi = r9; const Register saved_rsi = r10; assert(nargs == 3 || nargs == 4, "else fix"); #ifdef _WIN64 assert(c_rarg0 == rcx && c_rarg1 == rdx && c_rarg2 == r8 && c_rarg3 == r9, "unexpected argument registers"); if (nargs >= 4) __ mov(rax, r9); // r9 is also saved_rdi __ movptr(saved_rdi, rdi); __ movptr(saved_rsi, rsi); __ mov(rdi, rcx); // c_rarg0 __ mov(rsi, rdx); // c_rarg1 __ mov(rdx, r8); // c_rarg2 if (nargs >= 4) __ mov(rcx, rax); // c_rarg3 (via rax) #else assert(c_rarg0 == rdi && c_rarg1 == rsi && c_rarg2 == rdx && c_rarg3 == rcx, "unexpected argument registers"); #endif DEBUG_ONLY(_regs_in_thread = false;) } void StubGenerator::restore_arg_regs() { assert(!_regs_in_thread, "wrong call to restore_arg_regs"); const Register saved_rdi = r9; const Register saved_rsi = r10; #ifdef _WIN64 __ movptr(rdi, saved_rdi); __ movptr(rsi, saved_rsi); #endif } // This is used in places where r10 is a scratch register, and can // be adapted if r9 is needed also. void StubGenerator::setup_arg_regs_using_thread() { const Register saved_r15 = r9; #ifdef _WIN64 __ mov(saved_r15, r15); // r15 is callee saved and needs to be restored __ get_thread(r15_thread); assert(c_rarg0 == rcx && c_rarg1 == rdx && c_rarg2 == r8 && c_rarg3 == r9, "unexpected argument registers"); __ movptr(Address(r15_thread, in_bytes(JavaThread::windows_saved_rdi_offset())), rd __ movptr(Address(r15_thread, in_bytes(JavaThread::windows_saved_rsi_offset())), rs __ mov(rdi, rcx); // c_rarg0 __ mov(rsi, rdx); // c_rarg1 __ mov(rdx, r8); // c_rarg2 #else assert(c_rarg0 == rdi && c_rarg1 == rsi && c_rarg2 == rdx && c_rarg3 == rcx, "unexpected argument registers"); #endif DEBUG_ONLY(_regs_in_thread = true;) } void StubGenerator::restore_arg_regs_using_thread() { assert(_regs_in_thread, "wrong call to restore_arg_regs"); const Register saved_r15 = r9; #ifdef _WIN64 __ get_thread(r15_thread); __ movptr(rsi, Address(r15_thread, in_bytes(JavaThread::windows_saved_rsi_offset()) __ movptr(rdi, Address(r15_thread, in_bytes(JavaThread::windows_saved_rdi_offset()) __ mov(r15, saved_r15); // r15 is callee saved and needs to be restored #endif } void StubGenerator::setup_argument_regs(BasicType type) { if (type == T_BYTE || type == T_SHORT) { setup_arg_regs(); // from => rdi, to => rsi, count => rdx // r9 and r10 may be used to save non-volatile registers } else { setup_arg_regs_using_thread(); // from => rdi, to => rsi, count => rdx // r9 is used to save r15_thread } } void StubGenerator::restore_argument_regs(BasicType type) { if (type == T_BYTE || type == T_SHORT) { restore_arg_regs(); } else { restore_arg_regs_using_thread(); } } address StubGenerator::generate_data_cache_writeback() { const Register src = c_rarg0; // source address __ align(CodeEntryAlignment); StubCodeMark mark(this, "StubRoutines", "_data_cache_writeback"); address start = __ pc(); __ enter(); __ cache_wb(Address(src, 0)); __ leave(); __ ret(0); return start; } address StubGenerator::generate_data_cache_writeback_sync() { const Register is_pre = c_rarg0; // pre or post sync __ align(CodeEntryAlignment); StubCodeMark mark(this, "StubRoutines", "_data_cache_writeback_sync"); // pre wbsync is a no-op // post wbsync translates to an sfence Label skip; address start = __ pc(); __ enter(); __ cmpl(is_pre, 0); __ jcc(Assembler::notEqual, skip); __ cache_wbsync(false); __ bind(skip); __ leave(); __ ret(0); return start; } // ofs and limit are use for multi-block byte array. // int com.sun.security.provider.MD5.implCompress(byte[] b, int ofs) address StubGenerator::generate_md5_implCompress(bool multi_block, const char *name) { __ align(CodeEntryAlignment); StubCodeMark mark(this, "StubRoutines", name); address start = __ pc(); const Register buf_param = r15; const Address state_param(rsp, 0 * wordSize); const Address ofs_param (rsp, 1 * wordSize ); const Address limit_param(rsp, 1 * wordSize + 4); __ enter(); __ push(rbx); __ push(rdi); __ push(rsi); __ push(r15); __ subptr(rsp, 2 * wordSize); __ movptr(buf_param, c_rarg0); __ movptr(state_param, c_rarg1); if (multi_block) { __ movl(ofs_param, c_rarg2); __ movl(limit_param, c_rarg3); } __ fast_md5(buf_param, state_param, ofs_param, limit_param, multi_block); __ addptr(rsp, 2 * wordSize); __ pop(r15); __ pop(rsi); __ pop(rdi); __ pop(rbx); __ leave(); __ ret(0); return start; } address StubGenerator::generate_upper_word_mask() { __ align64(); StubCodeMark mark(this, "StubRoutines", "upper_word_mask"); address start = __ pc(); __ emit_data64(0x0000000000000000, relocInfo::none); __ emit_data64(0xFFFFFFFF00000000, relocInfo::none); return start; } address StubGenerator::generate_shuffle_byte_flip_mask() { __ align64(); StubCodeMark mark(this, "StubRoutines", "shuffle_byte_flip_mask"); address start = __ pc(); __ emit_data64(0x08090a0b0c0d0e0f, relocInfo::none); __ emit_data64(0x0001020304050607, relocInfo::none); return start; } // ofs and limit are use for multi-block byte array. // int com.sun.security.provider.DigestBase.implCompressMultiBlock(byte[] b, int ofs, address StubGenerator::generate_sha1_implCompress(bool multi_block, const char *name) __ align(CodeEntryAlignment); StubCodeMark mark(this, "StubRoutines", name); address start = __ pc(); Register buf = c_rarg0; Register state = c_rarg1; Register ofs = c_rarg2; Register limit = c_rarg3; const XMMRegister abcd = xmm0; const XMMRegister e0 = xmm1; const XMMRegister e1 = xmm2; const XMMRegister msg0 = xmm3; const XMMRegister msg1 = xmm4; const XMMRegister msg2 = xmm5; const XMMRegister msg3 = xmm6; const XMMRegister shuf_mask = xmm7; __ enter(); __ subptr(rsp, 4 * wordSize); __ fast_sha1(abcd, e0, e1, msg0, msg1, msg2, msg3, shuf_mask, buf, state, ofs, limit, rsp, multi_block); __ addptr(rsp, 4 * wordSize); __ leave(); __ ret(0); return start; } address StubGenerator::generate_pshuffle_byte_flip_mask() { __ align64(); StubCodeMark mark(this, "StubRoutines", "pshuffle_byte_flip_mask"); address start = __ pc(); __ emit_data64(0x0405060700010203, relocInfo::none); __ emit_data64(0x0c0d0e0f08090a0b, relocInfo::none); if (VM_Version::supports_avx2()) { __ emit_data64(0x0405060700010203, relocInfo::none); // second copy __ emit_data64(0x0c0d0e0f08090a0b, relocInfo::none); // _SHUF_00BA __ emit_data64(0x0b0a090803020100, relocInfo::none); __ emit_data64(0xFFFFFFFFFFFFFFFF, relocInfo::none); __ emit_data64(0x0b0a090803020100, relocInfo::none); __ emit_data64(0xFFFFFFFFFFFFFFFF, relocInfo::none); // _SHUF_DC00 __ emit_data64(0xFFFFFFFFFFFFFFFF, relocInfo::none); __ emit_data64(0x0b0a090803020100, relocInfo::none); __ emit_data64(0xFFFFFFFFFFFFFFFF, relocInfo::none); __ emit_data64(0x0b0a090803020100, relocInfo::none); } return start; } //Mask for byte-swapping a couple of qwords in an XMM register using (v)pshufb. address StubGenerator::generate_pshuffle_byte_flip_mask_sha512() { __ align32(); StubCodeMark mark(this, "StubRoutines", "pshuffle_byte_flip_mask_sha512"); address start = __ pc(); if (VM_Version::supports_avx2()) { __ emit_data64(0x0001020304050607, relocInfo::none); // PSHUFFLE_BYTE_FLIP_MASK __ emit_data64(0x08090a0b0c0d0e0f, relocInfo::none); __ emit_data64(0x1011121314151617, relocInfo::none); __ emit_data64(0x18191a1b1c1d1e1f, relocInfo::none); __ emit_data64(0x0000000000000000, relocInfo::none); //MASK_YMM_LO __ emit_data64(0x0000000000000000, relocInfo::none); __ emit_data64(0xFFFFFFFFFFFFFFFF, relocInfo::none); __ emit_data64(0xFFFFFFFFFFFFFFFF, relocInfo::none); } return start; } // ofs and limit are use for multi-block byte array. // int com.sun.security.provider.DigestBase.implCompressMultiBlock(byte[] b, int ofs, address StubGenerator::generate_sha256_implCompress(bool multi_block, const char *nam assert(VM_Version::supports_sha() || VM_Version::supports_avx2(), ""); __ align(CodeEntryAlignment); StubCodeMark mark(this, "StubRoutines", name); address start = __ pc(); Register buf = c_rarg0; Register state = c_rarg1; Register ofs = c_rarg2; Register limit = c_rarg3; const XMMRegister msg = xmm0; const XMMRegister state0 = xmm1; const XMMRegister state1 = xmm2; const XMMRegister msgtmp0 = xmm3; const XMMRegister msgtmp1 = xmm4; const XMMRegister msgtmp2 = xmm5; const XMMRegister msgtmp3 = xmm6; const XMMRegister msgtmp4 = xmm7; const XMMRegister shuf_mask = xmm8; __ enter(); __ subptr(rsp, 4 * wordSize); if (VM_Version::supports_sha()) { __ fast_sha256(msg, state0, state1, msgtmp0, msgtmp1, msgtmp2, msgtmp3, msgtmp4, buf, state, ofs, limit, rsp, multi_block, shuf_mask); } else if (VM_Version::supports_avx2()) { __ sha256_AVX2(msg, state0, state1, msgtmp0, msgtmp1, msgtmp2, msgtmp3, msgtmp4, buf, state, ofs, limit, rsp, multi_block, shuf_mask); } __ addptr(rsp, 4 * wordSize); __ vzeroupper(); __ leave(); __ ret(0); return start; } address StubGenerator::generate_sha512_implCompress(bool multi_block, const char *nam assert(VM_Version::supports_avx2(), ""); assert(VM_Version::supports_bmi2(), ""); __ align(CodeEntryAlignment); StubCodeMark mark(this, "StubRoutines", name); address start = __ pc(); Register buf = c_rarg0; Register state = c_rarg1; Register ofs = c_rarg2; Register limit = c_rarg3; const XMMRegister msg = xmm0; const XMMRegister state0 = xmm1; const XMMRegister state1 = xmm2; const XMMRegister msgtmp0 = xmm3; const XMMRegister msgtmp1 = xmm4; const XMMRegister msgtmp2 = xmm5; const XMMRegister msgtmp3 = xmm6; const XMMRegister msgtmp4 = xmm7; const XMMRegister shuf_mask = xmm8; __ enter(); __ sha512_AVX2(msg, state0, state1, msgtmp0, msgtmp1, msgtmp2, msgtmp3, msgtmp4, buf, state, ofs, limit, rsp, multi_block, shuf_mask); __ vzeroupper(); __ leave(); __ ret(0); return start; } address StubGenerator::base64_shuffle_addr() { __ align64(); StubCodeMark mark(this, "StubRoutines", "shuffle_base64"); address start = __ pc(); assert(((unsigned long long)start & 0x3f) == 0, "Alignment problem (0x%08llx)", (unsigned long long)start); __ emit_data64(0x0405030401020001, relocInfo::none); __ emit_data64(0x0a0b090a07080607, relocInfo::none); __ emit_data64(0x10110f100d0e0c0d, relocInfo::none); __ emit_data64(0x1617151613141213, relocInfo::none); __ emit_data64(0x1c1d1b1c191a1819, relocInfo::none); __ emit_data64(0x222321221f201e1f, relocInfo::none); __ emit_data64(0x2829272825262425, relocInfo::none); __ emit_data64(0x2e2f2d2e2b2c2a2b, relocInfo::none); return start; } address StubGenerator::base64_avx2_shuffle_addr() { __ align32(); StubCodeMark mark(this, "StubRoutines", "avx2_shuffle_base64"); address start = __ pc(); __ emit_data64(0x0809070805060405, relocInfo::none); __ emit_data64(0x0e0f0d0e0b0c0a0b, relocInfo::none); __ emit_data64(0x0405030401020001, relocInfo::none); __ emit_data64(0x0a0b090a07080607, relocInfo::none); return start; } address StubGenerator::base64_avx2_input_mask_addr() { __ align32(); StubCodeMark mark(this, "StubRoutines", "avx2_input_mask_base64"); address start = __ pc(); __ emit_data64(0x8000000000000000, relocInfo::none); __ emit_data64(0x8000000080000000, relocInfo::none); __ emit_data64(0x8000000080000000, relocInfo::none); __ emit_data64(0x8000000080000000, relocInfo::none); return start; } address StubGenerator::base64_avx2_lut_addr() { __ align32(); StubCodeMark mark(this, "StubRoutines", "avx2_lut_base64"); address start = __ pc(); __ emit_data64(0xfcfcfcfcfcfc4741, relocInfo::none); __ emit_data64(0x0000f0edfcfcfcfc, relocInfo::none); __ emit_data64(0xfcfcfcfcfcfc4741, relocInfo::none); __ emit_data64(0x0000f0edfcfcfcfc, relocInfo::none); // URL LUT __ emit_data64(0xfcfcfcfcfcfc4741, relocInfo::none); __ emit_data64(0x000020effcfcfcfc, relocInfo::none); __ emit_data64(0xfcfcfcfcfcfc4741, relocInfo::none); __ emit_data64(0x000020effcfcfcfc, relocInfo::none); return start; } address StubGenerator::base64_encoding_table_addr() { __ align64(); StubCodeMark mark(this, "StubRoutines", "encoding_table_base64"); address start = __ pc(); assert(((unsigned long long)start & 0x3f) == 0, "Alignment problem (0x%08llx)", (uns __ emit_data64(0x4847464544434241, relocInfo::none); __ emit_data64(0x504f4e4d4c4b4a49, relocInfo::none); __ emit_data64(0x5857565554535251, relocInfo::none); __ emit_data64(0x6665646362615a59, relocInfo::none); __ emit_data64(0x6e6d6c6b6a696867, relocInfo::none); __ emit_data64(0x767574737271706f, relocInfo::none); __ emit_data64(0x333231307a797877, relocInfo::none); __ emit_data64(0x2f2b393837363534, relocInfo::none); // URL table __ emit_data64(0x4847464544434241, relocInfo::none); __ emit_data64(0x504f4e4d4c4b4a49, relocInfo::none); __ emit_data64(0x5857565554535251, relocInfo::none); __ emit_data64(0x6665646362615a59, relocInfo::none); __ emit_data64(0x6e6d6c6b6a696867, relocInfo::none); __ emit_data64(0x767574737271706f, relocInfo::none); __ emit_data64(0x333231307a797877, relocInfo::none); __ emit_data64(0x5f2d393837363534, relocInfo::none); return start; } // Code for generating Base64 encoding. // Intrinsic function prototype in Base64.java: // private void encodeBlock(byte[] src, int sp, int sl, byte[] dst, int dp, // boolean isURL) { address StubGenerator::generate_base64_encodeBlock() { __ align(CodeEntryAlignment); StubCodeMark mark(this, "StubRoutines", "implEncode"); address start = __ pc(); __ enter(); // Save callee-saved registers before using them __ push(r12); __ push(r13); __ push(r14); __ push(r15); // arguments const Register source = c_rarg0; // Source Array const Register start_offset = c_rarg1; // start offset const Register end_offset = c_rarg2; // end offset const Register dest = c_rarg3; // destination array #ifndef _WIN64 const Register dp = c_rarg4; // Position for writing to dest array const Register isURL = c_rarg5; // Base64 or URL character set #else const Address dp_mem(rbp, 6 * wordSize); // length is on stack on Win64 const Address isURL_mem(rbp, 7 * wordSize); const Register isURL = r10; // pick the volatile windows register const Register dp = r12; __ movl(dp, dp_mem); __ movl(isURL, isURL_mem); #endif const Register length = r14; const Register encode_table = r13; Label L_process3, L_exit, L_processdata, L_vbmiLoop, L_not512, L_32byteLoop; // calculate length from offsets __ movl(length, end_offset); __ subl(length, start_offset); __ cmpl(length, 0); __ jcc(Assembler::lessEqual, L_exit); // Code for 512-bit VBMI encoding. Encodes 48 input bytes into 64 // output bytes. We read 64 input bytes and ignore the last 16, so be // sure not to read past the end of the input buffer. if (VM_Version::supports_avx512_vbmi()) { __ cmpl(length, 64); // Do not overrun input buffer. __ jcc(Assembler::below, L_not512); __ shll(isURL, 6); // index into decode table based on isURL __ lea(encode_table, ExternalAddress(StubRoutines::x86::base64_encoding_table_addr __ addptr(encode_table, isURL); __ shrl(isURL, 6); // restore isURL __ mov64(rax, 0x3036242a1016040aull); // Shifts __ evmovdquq(xmm3, ExternalAddress(StubRoutines::x86::base64_shuffle_addr()), Assem __ evmovdquq(xmm2, Address(encode_table, 0), Assembler::AVX_512bit); __ evpbroadcastq(xmm1, rax, Assembler::AVX_512bit); __ align32(); __ BIND(L_vbmiLoop); __ vpermb(xmm0, xmm3, Address(source, start_offset), Assembler::AVX_512bit); __ subl(length, 48); // Put the input bytes into the proper lanes for writing, then // encode them. __ evpmultishiftqb(xmm0, xmm1, xmm0, Assembler::AVX_512bit); __ vpermb(xmm0, xmm0, xmm2, Assembler::AVX_512bit); // Write to destination __ evmovdquq(Address(dest, dp), xmm0, Assembler::AVX_512bit); __ addptr(dest, 64); __ addptr(source, 48); __ cmpl(length, 64); __ jcc(Assembler::aboveEqual, L_vbmiLoop); __ vzeroupper(); } __ BIND(L_not512); if (VM_Version::supports_avx2() && VM_Version::supports_avx512vlbw()) { /* ** This AVX2 encoder is based off the paper at: ** https://dl.acm.org/doi/10.1145/3132709 ** ** We use AVX2 SIMD instructions to encode 24 bytes into 32 ** output bytes. ** */ // Lengths under 32 bytes are done with scalar routine __ cmpl(length, 31); __ jcc(Assembler::belowEqual, L_process3); // Set up supporting constant table data __ vmovdqu(xmm9, ExternalAddress(StubRoutines::x86::base64_avx2_shuffle_addr()), ra // 6-bit mask for 2nd and 4th (and multiples) 6-bit values __ movl(rax, 0x0fc0fc00); __ vmovdqu(xmm1, ExternalAddress(StubRoutines::x86::base64_avx2_input_mask_addr()) __ evpbroadcastd(xmm8, rax, Assembler::AVX_256bit); // Multiplication constant for "shifting" right by 6 and 10 // bits __ movl(rax, 0x04000040); __ subl(length, 24); __ evpbroadcastd(xmm7, rax, Assembler::AVX_256bit); // For the first load, we mask off reading of the first 4 // bytes into the register. This is so we can get 4 3-byte // chunks into each lane of the register, avoiding having to // handle end conditions. We then shuffle these bytes into a // specific order so that manipulation is easier. // // The initial read loads the XMM register like this: // // Lower 128-bit lane: // +----+----+----+----+----+----+----+----+----+----+----+----+----+----+----+-- // | XX | XX | XX | XX | A0 | A1 | A2 | B0 | B1 | B2 | C0 | C1 // | C2 | D0 | D1 | D2 | // +----+----+----+----+----+----+----+----+----+----+----+----+----+----+----+-- // // Upper 128-bit lane: // +----+----+----+----+----+----+----+----+----+----+----+----+----+----+----+-- // | E0 | E1 | E2 | F0 | F1 | F2 | G0 | G1 | G2 | H0 | H1 | H2 // | XX | XX | XX | XX | // +----+----+----+----+----+----+----+----+----+----+----+----+----+----+----+-- // // Where A0 is the first input byte, B0 is the fourth, etc. // The alphabetical significance denotes the 3 bytes to be // consumed and encoded into 4 bytes. // // We then shuffle the register so each 32-bit word contains // the sequence: // A1 A0 A2 A1, B1, B0, B2, B1, etc. // Each of these byte sequences are then manipulated into 4 // 6-bit values ready for encoding. // // If we focus on one set of 3-byte chunks, changing the // nomenclature such that A0 => a, A1 => b, and A2 => c, we // shuffle such that each 24-bit chunk contains: // // b7 b6 b5 b4 b3 b2 b1 b0 | a7 a6 a5 a4 a3 a2 a1 a0 | c7 c6 // c5 c4 c3 c2 c1 c0 | b7 b6 b5 b4 b3 b2 b1 b0 // Explain this step. // b3 b2 b1 b0 c5 c4 c3 c2 | c1 c0 d5 d4 d3 d2 d1 d0 | a5 a4 // a3 a2 a1 a0 b5 b4 | b3 b2 b1 b0 c5 c4 c3 c2 // // W first and off all but bits 4-9 and 16-21 (c5..c0 and // a5..a0) and shift them using a vector multiplication // operation (vpmulhuw) which effectively shifts c right by 6 // bits and a right by 10 bits. We similarly mask bits 10-15 // (d5..d0) and 22-27 (b5..b0) and shift them left by 8 and 4 // bits respectively. This is done using vpmullw. We end up // with 4 6-bit values, thus splitting the 3 input bytes, // ready for encoding: // 0 0 d5..d0 0 0 c5..c0 0 0 b5..b0 0 0 a5..a0 // // For translation, we recognize that there are 5 distinct // ranges of legal Base64 characters as below: // // +-------------+-------------+------------+ // | 6-bit value | ASCII range | offset | // +-------------+-------------+------------+ // | 0..25 | A..Z | 65 | // | 26..51 | a..z | 71 | // | 52..61 | 0..9 | -4 | // | 62 | + or - | -19 or -17 | // | 63 | / or _ | -16 or 32 | // +-------------+-------------+------------+ // // We note that vpshufb does a parallel lookup in a // destination register using the lower 4 bits of bytes from a // source register. If we use a saturated subtraction and // subtract 51 from each 6-bit value, bytes from [0,51] // saturate to 0, and [52,63] map to a range of [1,12]. We // distinguish the [0,25] and [26,51] ranges by assigning a // value of 13 for all 6-bit values less than 26. We end up // with: // // +-------------+-------------+------------+ // | 6-bit value | Reduced | offset | // +-------------+-------------+------------+ // | 0..25 | 13 | 65 | // | 26..51 | 0 | 71 | // | 52..61 | 0..9 | -4 | // | 62 | 11 | -19 or -17 | // | 63 | 12 | -16 or 32 | // +-------------+-------------+------------+ // // We then use a final vpshufb to add the appropriate offset, // translating the bytes. // // Load input bytes - only 28 bytes. Mask the first load to // not load into the full register. __ vpmaskmovd(xmm1, xmm1, Address(source, start_offset, Address::times_1, -4), Assemble // Move 3-byte chunks of input (12 bytes) into 16 bytes, // ordering by: // 1, 0, 2, 1; 4, 3, 5, 4; etc. This groups 6-bit chunks // for easy masking __ vpshufb(xmm1, xmm1, xmm9, Assembler::AVX_256bit); __ addl(start_offset, 24); // Load masking register for first and third (and multiples) // 6-bit values. __ movl(rax, 0x003f03f0); __ evpbroadcastd(xmm6, rax, Assembler::AVX_256bit); // Multiplication constant for "shifting" left by 4 and 8 bits __ movl(rax, 0x01000010); __ evpbroadcastd(xmm5, rax, Assembler::AVX_256bit); // Isolate 6-bit chunks of interest __ vpand(xmm0, xmm8, xmm1, Assembler::AVX_256bit); // Load constants for encoding __ movl(rax, 0x19191919); __ evpbroadcastd(xmm3, rax, Assembler::AVX_256bit); __ movl(rax, 0x33333333); __ evpbroadcastd(xmm4, rax, Assembler::AVX_256bit); // Shift output bytes 0 and 2 into proper lanes __ vpmulhuw(xmm2, xmm0, xmm7, Assembler::AVX_256bit); // Mask and shift output bytes 1 and 3 into proper lanes and // combine __ vpand(xmm0, xmm6, xmm1, Assembler::AVX_256bit); __ vpmullw(xmm0, xmm5, xmm0, Assembler::AVX_256bit); __ vpor(xmm0, xmm0, xmm2, Assembler::AVX_256bit); // Find out which are 0..25. This indicates which input // values fall in the range of 'A'-'Z', which require an // additional offset (see comments above) __ vpcmpgtb(xmm2, xmm0, xmm3, Assembler::AVX_256bit); __ vpsubusb(xmm1, xmm0, xmm4, Assembler::AVX_256bit); __ vpsubb(xmm1, xmm1, xmm2, Assembler::AVX_256bit); // Load the proper lookup table __ lea(r11, ExternalAddress(StubRoutines::x86::base64_avx2_lut_addr())); __ movl(r15, isURL); __ shll(r15, 5); __ vmovdqu(xmm2, Address(r11, r15)); // Shuffle the offsets based on the range calculation done // above. This allows us to add the correct offset to the // 6-bit value corresponding to the range documented above. __ vpshufb(xmm1, xmm2, xmm1, Assembler::AVX_256bit); __ vpaddb(xmm0, xmm1, xmm0, Assembler::AVX_256bit); // Store the encoded bytes __ vmovdqu(Address(dest, dp), xmm0); __ addl(dp, 32); __ cmpl(length, 31); __ jcc(Assembler::belowEqual, L_process3); __ align32(); __ BIND(L_32byteLoop); // Get next 32 bytes __ vmovdqu(xmm1, Address(source, start_offset, Address::times_1, -4)); __ subl(length, 24); __ addl(start_offset, 24); // This logic is identical to the above, with only constant // register loads removed. Shuffle the input, mask off 6-bit // chunks, shift them into place, then add the offset to // encode. __ vpshufb(xmm1, xmm1, xmm9, Assembler::AVX_256bit); __ vpand(xmm0, xmm8, xmm1, Assembler::AVX_256bit); __ vpmulhuw(xmm10, xmm0, xmm7, Assembler::AVX_256bit); __ vpand(xmm0, xmm6, xmm1, Assembler::AVX_256bit); __ vpmullw(xmm0, xmm5, xmm0, Assembler::AVX_256bit); __ vpor(xmm0, xmm0, xmm10, Assembler::AVX_256bit); __ vpcmpgtb(xmm10, xmm0, xmm3, Assembler::AVX_256bit); __ vpsubusb(xmm1, xmm0, xmm4, Assembler::AVX_256bit); __ vpsubb(xmm1, xmm1, xmm10, Assembler::AVX_256bit); __ vpshufb(xmm1, xmm2, xmm1, Assembler::AVX_256bit); __ vpaddb(xmm0, xmm1, xmm0, Assembler::AVX_256bit); // Store the encoded bytes __ vmovdqu(Address(dest, dp), xmm0); __ addl(dp, 32); __ cmpl(length, 31); __ jcc(Assembler::above, L_32byteLoop); __ BIND(L_process3); __ vzeroupper(); } else { __ BIND(L_process3); } __ cmpl(length, 3); __ jcc(Assembler::below, L_exit); // Load the encoding table based on isURL __ lea(r11, ExternalAddress(StubRoutines::x86::base64_encoding_table_addr())); __ movl(r15, isURL); __ shll(r15, 6); __ addptr(r11, r15); __ BIND(L_processdata); // Load 3 bytes __ load_unsigned_byte(r15, Address(source, start_offset)); __ load_unsigned_byte(r10, Address(source, start_offset, Address::times_1, 1)); __ load_unsigned_byte(r13, Address(source, start_offset, Address::times_1, 2)); // Build a 32-bit word with bytes 1, 2, 0, 1 __ movl(rax, r10); __ shll(r10, 24); __ orl(rax, r10); __ subl(length, 3); __ shll(r15, 8); __ shll(r13, 16); __ orl(rax, r15); __ addl(start_offset, 3); __ orl(rax, r13); // At this point, rax contains | byte1 | byte2 | byte0 | byte1 // r13 has byte2 << 16 - need low-order 6 bits to translate. // This translated byte is the fourth output byte. __ shrl(r13, 16); __ andl(r13, 0x3f); // The high-order 6 bits of r15 (byte0) is translated. // The translated byte is the first output byte. __ shrl(r15, 10); __ load_unsigned_byte(r13, Address(r11, r13)); __ load_unsigned_byte(r15, Address(r11, r15)); __ movb(Address(dest, dp, Address::times_1, 3), r13); // Extract high-order 4 bits of byte1 and low-order 2 bits of byte0. // This translated byte is the second output byte. __ shrl(rax, 4); __ movl(r10, rax); __ andl(rax, 0x3f); --> -------------------- --> maximum size reached --> -------------------- ¤ Die Informationen auf dieser Webseite wurden nach bestem Wissen sorgfältig zusammengestellt. Es wird jedoch weder Vollständigkeit, noch Richtigkeit, noch Qualität der bereit gestellten Informationen zugesichert.0.24Bemerkung: (vorverarbeitet) ¤*© Formatika GbR, Deutschland |
|
2026-03-28