Mercurial > hg > icedtea8-forest > hotspot
view src/cpu/aarch64/vm/interpreter_aarch64.cpp @ 10905:f57189b7648d
8257192: Integrate AArch64 JIT port into 8u
7009641: Don't fail VM when CodeCache is full
8073108: [AArch64] Use x86 and SPARC CPU instructions for GHASH acceleration
8130309: Need to bailout cleanly if creation of stubs fails when codecache is out of space (AArch64 changes)
8131779: AARCH64: add Montgomery multiply intrinsic
8132875: AArch64: Fix error introduced into AArch64 CodeCache by commit for 8130309
8135018: AARCH64: Missing memory barriers for CMS collector
8145320: Create unsafe_arraycopy and generic_arraycopy for AArch64
8148328: aarch64: redundant lsr instructions in stub code.
8148783: aarch64: SEGV running SpecJBB2013
8148948: aarch64: generate_copy_longs calls align() incorrectly
8149080: AArch64: Recognise disjoint array copy in stub code
8149365: aarch64: memory copy does not prefetch on backwards copy
8149907: aarch64: use load/store pair instructions in call_stub
8150038: aarch64: make use of CBZ and CBNZ when comparing narrow pointer with zero
8150045: arraycopy causes segfaults in SATB during garbage collection
8150082: aarch64: optimise small array copy
8150229: aarch64: pipeline class for several instructions is not set correctly
8150313: aarch64: optimise array copy using SIMD instructions
8150394: aarch64: add support for 8.1 LSE CAS instructions
8150652: Remove unused code in AArch64 back end
8151340: aarch64: prefetch the destination word for write prior to ldxr/stxr loops.
8151502: optimize pd_disjoint_words and pd_conjoint_words
8151775: aarch64: add support for 8.1 LSE atomic operations
8152537: aarch64: Make use of CBZ and CBNZ when comparing unsigned values with zero.
8152840: aarch64: improve _unsafe_arraycopy stub routine
8153172: aarch64: hotspot crashes after the 8.1 LSE patch is merged
8153713: aarch64: improve short array clearing using store pair
8153797: aarch64: Add Arrays.fill stub code
8154413: AArch64: Better byte behaviour
8154537: AArch64: some integer rotate instructions are never emitted
8154739: AArch64: TemplateTable::fast_xaccess loads in wrong mode
8155015: Aarch64: bad assert in spill generation code
8155100: AArch64: Relax alignment requirement for byte_map_base
8155612: Aarch64: vector nodes need to support misaligned offset
8155617: aarch64: ClearArray does not use DC ZVA
8155627: Enable SA on AArch64
8155653: TestVectorUnalignedOffset.java not pushed with 8155612
8156731: aarch64: java/util/Arrays/Correct.java fails due to _generic_arraycopy stub routine
8157841: aarch64: prefetch ignores cache line size
8157906: aarch64: some more integer rotate instructions are never emitted
8158913: aarch64: SEGV running Spark terasort
8159052: aarch64: optimise unaligned copies in pd_disjoint_words and pd_conjoint_words
8159063: aarch64: optimise unaligned array copy long
8160748: [AArch64] Inconsistent types for ideal_reg
8161072: AArch64: jtreg compiler/uncommontrap/TestDeoptOOM failure
8161190: AArch64: Fix overflow in immediate cmp instruction
8164113: AArch64: follow-up the fix for 8161598
8165673: AArch64: Fix JNI floating point argument handling
8167200: AArch64: Broken stack pointer adjustment in interpreter
8167421: AArch64: in one core system, fatal error: Illegal threadstate encountered
8167595: AArch64: SEGV in stub code cipherBlockChaining_decryptAESCrypt
8168699: Validate special case invocations [AArch64 support]
8168888: Port 8160591: Improve internal array handling to AArch64.
8170100: AArch64: Crash in C1-compiled code accessing References
8170188: jtreg test compiler/types/TestMeetIncompatibleInterfaceArrays.java causes JVM crash
8170873: PPC64/aarch64: Poor StrictMath performance due to non-optimized compilation
8171537: aarch64: compiler/c1/Test6849574.java generates guarantee failure in C1
8172881: AArch64: assertion failure: the int pressure is incorrect
8173472: AArch64: C1 comparisons with null only use 32-bit instructions
8176100: [AArch64] [REDO][REDO] G1 Needs pre barrier on dereference of weak JNI handles
8177661: Correct ad rule output register types from iRegX to iRegXNoSp
8179954: AArch64: C1 and C2 volatile accesses are not sequentially consistent
8182581: aarch64: fix for crash caused by earlyret of compiled method
8183925: [AArch64] Decouple crash protection from watcher thread
8186325: AArch64: jtreg test hotspot/test/gc/g1/TestJNIWeakG1/TestJNIWeakG1.java SEGV
8187224: aarch64: some inconsistency between aarch64_ad.m4 and aarch64.ad
8189170: [AArch64] Add option to disable stack overflow checking in primordial thread for use with JNI_CreateJavaJVM
8193133: Assertion failure because 0xDEADDEAD can be in-heap
8195685: AArch64 port of 8174962: Better interface invocations
8195859: AArch64: vtableStubs gtest fails after 8174962
8196136: AArch64: Correct register use in patch for JDK-8194686
8196221: AArch64: Mistake in committed patch for JDK-8195859
8199712: [AArch64] Flight Recorder
8203481: Incorrect constraint for unextended_sp in frame:safe_for_sender
8203699: java/lang/invoke/SpecialInterfaceCall fails with SIGILL on aarch64
8205421: AARCH64: StubCodeMark should be placed after alignment
8206163: AArch64: incorrect code generation for StoreCM
8207345: Trampoline generation code reads from uninitialized memory
8207838: AArch64: Float registers incorrectly restored in JNI call
8209413: AArch64: NPE in clhsdb jstack command
8209414: [AArch64] method handle invocation does not respect JVMTI interp_only mode
8209415: Fix JVMTI test failure HS202
8209420: Track membars for volatile accesses so they can be properly optimized
8209835: Aarch64: elide barriers on all volatile operations
8210425: [AArch64] sharedRuntimeTrig/sharedRuntimeTrans compiled without optimization
8211064: [AArch64] Interpreter and c1 don't correctly handle jboolean results in native calls
8211233: MemBarNode::trailing_membar() and MemBarNode::leading_membar() need to handle dying subgraphs better
8213134: AArch64: vector shift failed with MaxVectorSize=8
8213419: [AArch64] C2 may hang in MulLNode::Ideal()/MulINode::Ideal() with gcc 8.2.1
8214857: "bad trailing membar" assert failure at memnode.cpp:3220
8215951: AArch64: jtreg test vmTestbase/nsk/jvmti/PopFrame/popframe005 segfaults
8215961: jdk/jfr/event/os/TestCPUInformation.java fails on AArch64
8216350: AArch64: monitor unlock fast path not called
8216989: CardTableBarrierSetAssembler::gen_write_ref_array_post_barrier() does not check for zero length on AARCH64
8217368: AArch64: C2 recursive stack locking optimisation not triggered
8218185: aarch64: missing LoadStore barrier in TemplateTable::putfield_or_static
8219011: Implement MacroAssembler::warn method on AArch64
8219635: aarch64: missing LoadStore barrier in TemplateTable::fast_storefield
8221220: AArch64: Add StoreStore membar explicitly for Volatile Writes in TemplateTable
8221658: aarch64: add necessary predicate for ubfx patterns
8224671: AArch64: mauve System.arraycopy test failure
8224828: aarch64: rflags is not correct after safepoint poll
8224851: AArch64: fix warnings and errors with Clang and GCC 8.3
8224880: AArch64: java/javac error with AllocatePrefetchDistance
8228400: Remove built-in AArch64 simulator
8228406: Superfluous change in chaitin.hpp
8228593: Revert explicit JDK 7 support additions
8228716: Revert InstanceKlass::print_on debug additions
8228718: Revert incorrect backport of JDK-8129757 to 8-aarch64
8228725: AArch64: Purge method call format support
8228747: Revert "unused" attribute from test_arraycopy_func
8228767: Revert ResourceMark additions
8228770: Revert development hsdis changes
8229123: Revert build fixes for aarch64/zero
8229124: Revert disassembler.cpp changes
8229145: Revert TemplateTable::bytecode() visibility change
8233839: aarch64: missing memory barrier in NewObjectArrayStub and NewTypeArrayStub
8237512: AArch64: aarch64TestHook leaks a BufferBlob
8246482: Build failures with +JFR -PCH
8247979: aarch64: missing side effect of killing flags for clearArray_reg_reg
8248219: aarch64: missing memory barrier in fast_storefield and fast_accessfield
Reviewed-by: shade, aph
author | andrew |
---|---|
date | Mon, 01 Feb 2021 03:48:36 +0000 |
parents | |
children | f79e943d15a7 |
line wrap: on
line source
/* * Copyright (c) 2013, Red Hat Inc. * Copyright (c) 2003, 2011, 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 "interpreter/bytecodeHistogram.hpp" #include "interpreter/interpreter.hpp" #include "interpreter/interpreterGenerator.hpp" #include "interpreter/interpreterRuntime.hpp" #include "interpreter/templateTable.hpp" #include "oops/arrayOop.hpp" #include "oops/methodData.hpp" #include "oops/method.hpp" #include "oops/oop.inline.hpp" #include "prims/jvmtiExport.hpp" #include "prims/jvmtiThreadState.hpp" #include "prims/methodHandles.hpp" #include "runtime/arguments.hpp" #include "runtime/deoptimization.hpp" #include "runtime/frame.inline.hpp" #include "runtime/sharedRuntime.hpp" #include "runtime/stubRoutines.hpp" #include "runtime/synchronizer.hpp" #include "runtime/timer.hpp" #include "runtime/vframeArray.hpp" #include "utilities/debug.hpp" #ifdef COMPILER1 #include "c1/c1_Runtime1.hpp" #endif #define __ _masm-> address AbstractInterpreterGenerator::generate_slow_signature_handler() { address entry = __ pc(); __ andr(esp, esp, -16); __ mov(c_rarg3, esp); // rmethod // rlocals // c_rarg3: first stack arg - wordSize // adjust sp __ sub(sp, c_rarg3, 18 * wordSize); __ str(lr, Address(__ pre(sp, -2 * wordSize))); __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::slow_signature_handler), rmethod, rlocals, c_rarg3); // r0: result handler // Stack layout: // rsp: return address <- sp // 1 garbage // 8 integer args (if static first is unused) // 1 float/double identifiers // 8 double args // stack args <- esp // garbage // expression stack bottom // bcp (NULL) // ... // Restore LR __ ldr(lr, Address(__ post(sp, 2 * wordSize))); // Do FP first so we can use c_rarg3 as temp __ ldrw(c_rarg3, Address(sp, 9 * wordSize)); // float/double identifiers for (int i = 0; i < Argument::n_float_register_parameters_c; i++) { const FloatRegister r = as_FloatRegister(i); Label d, done; __ tbnz(c_rarg3, i, d); __ ldrs(r, Address(sp, (10 + i) * wordSize)); __ b(done); __ bind(d); __ ldrd(r, Address(sp, (10 + i) * wordSize)); __ bind(done); } // c_rarg0 contains the result from the call of // InterpreterRuntime::slow_signature_handler so we don't touch it // here. It will be loaded with the JNIEnv* later. __ ldr(c_rarg1, Address(sp, 1 * wordSize)); for (int i = c_rarg2->encoding(); i <= c_rarg7->encoding(); i += 2) { Register rm = as_Register(i), rn = as_Register(i+1); __ ldp(rm, rn, Address(sp, i * wordSize)); } __ add(sp, sp, 18 * wordSize); __ ret(lr); return entry; } // // Various method entries // address InterpreterGenerator::generate_math_entry(AbstractInterpreter::MethodKind kind) { // rmethod: Method* // r13: sender sp // esp: args if (!InlineIntrinsics) return NULL; // Generate a vanilla entry // These don't need a safepoint check because they aren't virtually // callable. We won't enter these intrinsics from compiled code. // If in the future we added an intrinsic which was virtually callable // we'd have to worry about how to safepoint so that this code is used. // mathematical functions inlined by compiler // (interpreter must provide identical implementation // in order to avoid monotonicity bugs when switching // from interpreter to compiler in the middle of some // computation) // // stack: // [ arg ] <-- esp // [ arg ] // retaddr in lr address entry_point = NULL; Register continuation = lr; switch (kind) { case Interpreter::java_lang_math_abs: entry_point = __ pc(); __ ldrd(v0, Address(esp)); __ fabsd(v0, v0); __ mov(sp, r13); // Restore caller's SP break; case Interpreter::java_lang_math_sqrt: entry_point = __ pc(); __ ldrd(v0, Address(esp)); __ fsqrtd(v0, v0); __ mov(sp, r13); break; case Interpreter::java_lang_math_sin : case Interpreter::java_lang_math_cos : case Interpreter::java_lang_math_tan : case Interpreter::java_lang_math_log : case Interpreter::java_lang_math_log10 : case Interpreter::java_lang_math_exp : entry_point = __ pc(); __ ldrd(v0, Address(esp)); __ mov(sp, r13); __ mov(r19, lr); continuation = r19; // The first callee-saved register generate_transcendental_entry(kind, 1); break; case Interpreter::java_lang_math_pow : entry_point = __ pc(); __ mov(r19, lr); continuation = r19; __ ldrd(v0, Address(esp, 2 * Interpreter::stackElementSize)); __ ldrd(v1, Address(esp)); __ mov(sp, r13); generate_transcendental_entry(kind, 2); break; default: ; } if (entry_point) { __ br(continuation); } return entry_point; } // double trigonometrics and transcendentals // static jdouble dsin(jdouble x); // static jdouble dcos(jdouble x); // static jdouble dtan(jdouble x); // static jdouble dlog(jdouble x); // static jdouble dlog10(jdouble x); // static jdouble dexp(jdouble x); // static jdouble dpow(jdouble x, jdouble y); void InterpreterGenerator::generate_transcendental_entry(AbstractInterpreter::MethodKind kind, int fpargs) { address fn; switch (kind) { case Interpreter::java_lang_math_sin : fn = CAST_FROM_FN_PTR(address, SharedRuntime::dsin); break; case Interpreter::java_lang_math_cos : fn = CAST_FROM_FN_PTR(address, SharedRuntime::dcos); break; case Interpreter::java_lang_math_tan : fn = CAST_FROM_FN_PTR(address, SharedRuntime::dtan); break; case Interpreter::java_lang_math_log : fn = CAST_FROM_FN_PTR(address, SharedRuntime::dlog); break; case Interpreter::java_lang_math_log10 : fn = CAST_FROM_FN_PTR(address, SharedRuntime::dlog10); break; case Interpreter::java_lang_math_exp : fn = CAST_FROM_FN_PTR(address, SharedRuntime::dexp); break; case Interpreter::java_lang_math_pow : fn = CAST_FROM_FN_PTR(address, SharedRuntime::dpow); break; default: ShouldNotReachHere(); } __ mov(rscratch1, fn); __ blr(rscratch1); } // Abstract method entry // Attempt to execute abstract method. Throw exception address InterpreterGenerator::generate_abstract_entry(void) { // rmethod: Method* // r13: sender SP address entry_point = __ pc(); // abstract method entry // pop return address, reset last_sp to NULL __ empty_expression_stack(); __ restore_bcp(); // bcp must be correct for exception handler (was destroyed) __ restore_locals(); // make sure locals pointer is correct as well (was destroyed) // throw exception __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::throw_AbstractMethodError)); // the call_VM checks for exception, so we should never return here. __ should_not_reach_here(); return entry_point; } // Empty method, generate a very fast return. address InterpreterGenerator::generate_empty_entry(void) { // rmethod: Method* // r13: sender sp must set sp to this value on return if (!UseFastEmptyMethods) { return NULL; } address entry_point = __ pc(); // If we need a safepoint check, generate full interpreter entry. Label slow_path; { unsigned long offset; assert(SafepointSynchronize::_not_synchronized == 0, "SafepointSynchronize::_not_synchronized"); __ adrp(rscratch2, SafepointSynchronize::address_of_state(), offset); __ ldrw(rscratch2, Address(rscratch2, offset)); __ cbnz(rscratch2, slow_path); } // do nothing for empty methods (do not even increment invocation counter) // Code: _return // _return // return w/o popping parameters __ mov(sp, r13); // Restore caller's SP __ br(lr); __ bind(slow_path); (void) generate_normal_entry(false); return entry_point; } void Deoptimization::unwind_callee_save_values(frame* f, vframeArray* vframe_array) { // This code is sort of the equivalent of C2IAdapter::setup_stack_frame back in // the days we had adapter frames. When we deoptimize a situation where a // compiled caller calls a compiled caller will have registers it expects // to survive the call to the callee. If we deoptimize the callee the only // way we can restore these registers is to have the oldest interpreter // frame that we create restore these values. That is what this routine // will accomplish. // At the moment we have modified c2 to not have any callee save registers // so this problem does not exist and this routine is just a place holder. assert(f->is_interpreted_frame(), "must be interpreted"); }