Skip to main content

100 posts tagged with "java"

View All Tags

java unsafe

· 2 min read

背景​

java的unsafe包是有很多底层的api暴露出来,举例,java的netty就大量使用这个api

例子​

下面是java的unsafe包里面的allocateMemory方法.

public class UnsafeDemo {

    public static void main(String[] args) {
        var unsafe = getUnsafe();
        var memory = unsafe.allocateMemory(100);
        System.out.println(memory);
    }

    private static Unsafe getUnsafe() {
        try {
            Field field = Unsafe.class.getDeclaredField("theUnsafe");
            field.setAccessible(true);
            return (Unsafe) field.get(null);
        } catch (Exception e) {
            return null;
        }
    }
}

这里memory 返回的是一个地址

实际上是调用Unsafe_AllocateMemory0

UNSAFE_ENTRY(jlong, Unsafe_AllocateMemory0(JNIEnv *env, jobject unsafe, jlong size)) {
  size_t sz = (size_t)size;

  assert(is_aligned(sz, HeapWordSize), "sz not aligned");

  void* x = os::malloc(sz, mtOther);

  return addr_to_java(x);
} UNSAFE_END

最后调用的是glibc 的malloc

void* os::malloc(size_t size, MEMFLAGS memflags, const NativeCallStack& stack) {

  // Special handling for NMT preinit phase before arguments are parsed
  void* rc = NULL;
  if (NMTPreInit::handle_malloc(&rc, size)) {
    // No need to fill with 0 because DumpSharedSpaces doesn't use these
    // early allocations.
    return rc;
  }

  DEBUG_ONLY(check_crash_protection());

  // On malloc(0), implementations of malloc(3) have the choice to return either
  // NULL or a unique non-NULL pointer. To unify libc behavior across our platforms
  // we chose the latter.
  size = MAX2((size_t)1, size);

  // For the test flag -XX:MallocMaxTestWords
  if (has_reached_max_malloc_test_peak(size)) {
    return NULL;
  }

  const size_t outer_size = size + MemTracker::overhead_per_malloc();

  // Check for overflow.
  if (outer_size < size) {
    return NULL;
  }

  ALLOW_C_FUNCTION(::malloc, void* const outer_ptr = ::malloc(outer_size);)      <--  malloc 分配内存
  if (outer_ptr == NULL) {
    return NULL;
  }

  void* const inner_ptr = MemTracker::record_malloc((address)outer_ptr, size, memflags, stack);

  if (DumpSharedSpaces) {
    // Need to deterministically fill all the alignment gaps in C++ structures.
    ::memset(inner_ptr, 0, size);
  } else {
    DEBUG_ONLY(::memset(inner_ptr, uninitBlockPad, size);)
  }
  DEBUG_ONLY(break_if_ptr_caught(inner_ptr);)
  return inner_ptr;
}

相关阅读​

each jvm bytecode implement in x86 with asm

· 7 min read

背景​

想要了解jvm的bytecode 的汇编实现 ，目标平台是x86

汇编格式​

同样一个汇编语句：将1赋值给rax

汇编有两种表达方式

desc/描述intelAT&T
将1写入rax寄存器mov eax,1movl $1,%eax
将rab+3 的地址的值写入raxmov eax,[ebx+3]movl 3(%ebx),%eax

stack frame​

在x86 64 位的模式下 rbcp 是用r13 , 描述的是下一个指令，i = instruction
r14则存了本地变量指针

// Global Register Names
static const Register rbcp     = LP64_ONLY(r13) NOT_LP64(rsi);
static const Register rlocals  = LP64_ONLY(r14) NOT_LP64(rdi);

这里LP64_ONLY()和NOT_LP64()是通过宏_LP64来确定的

__LP64__
_LP64
These macros are defined, with value 1, if (and only if) the compilation is for a target where long int and pointer both use 64-bits and int uses 32-bit.

amd64 下面的寄存器​

java的stack frame

寄存器含义、描述
r14存了本地变量的基地址
r13指向下一个执行的bytecode

类似c的堆栈，java 的栈如下：

stack

相关阅读

frame 用下面的结构描述​

主要包括：

  • _sp :指向栈
  • _pc ： 指向指令
jdk/src/hotspot/share/runtime/frame.hpp
class frame {
 private:
  // Instance variables:
  intptr_t* _sp; // stack pointer (from Thread::last_Java_sp)  , java 的stack 指针
  address   _pc; // program counter (the next instruction after the call)   下一个指令的指针

  CodeBlob* _cb; // CodeBlob that "owns" pc
  enum deopt_state {
    not_deoptimized,
    is_deoptimized,
    unknown
  };

  deopt_state _deopt_state;

...

};

bytecode​

enum TosState {         // describes the tos cache contents
  btos = 0,             // byte, bool tos cached
  ztos = 1,             // byte, bool tos cached
  ctos = 2,             // char tos cached
  stos = 3,             // short tos cached
  itos = 4,             // int tos cached
  ltos = 5,             // long tos cached
  ftos = 6,             // float tos cached
  dtos = 7,             // double tos cached
  atos = 8,             // object cached
  vtos = 9,             // tos not cached
  number_of_states,
  ilgl                  // illegal state: should not occur
};

iload​

bytecodeenumasm
iload21
$65 = (address) 0x7fffe1012693 "A\017\266]\002\203\373\025\017\204J"
(gdb) x/20i 0x7fffe1012693
   0x7fffe1012693:	movzbl 0x2(%r13),%ebx
   0x7fffe1012698:	cmp    $0x15,%ebx            <---  下一个bytecode
   0x7fffe101269b:	je     0x7fffe10126eb        <--   跳转到 done
   0x7fffe10126a1:	cmp    $0xe0,%ebx             <-- 判断下一个是否是_fast_iload
   0x7fffe10126a7:	mov    $0xe1,%ecx              <------ 下一个是_fast_iload 则重写成fast_iload2
   0x7fffe10126ac:	je     0x7fffe10126bd           <--------  跳转到rewrite label
   0x7fffe10126ae:	cmp    $0x34,%ebx
   0x7fffe10126b1:	mov    $0xe2,%ecx
   0x7fffe10126b6:	je     0x7fffe10126bd
   0x7fffe10126b8:	mov    $0xe0,%ecx
   0x7fffe10126bd:	movzbl 0x0(%r13),%ebx
   0x7fffe10126c2:	cmp    $0x15,%ebx
   0x7fffe10126c5:	je     0x7fffe10126e7
   0x7fffe10126cb:	cmp    %ecx,%ebx
   0x7fffe10126cd:	je     0x7fffe10126e7
   0x7fffe10126d3:	movabs $0x7ffff74ef9d7,%rdi
   0x7fffe10126dd:	and    $0xfffffffffffffff0,%rsp
   0x7fffe10126e1:	call   0x7ffff694f3c0 <_ZN14MacroAssembler7debug64EPclPl>
   0x7fffe10126e6:	hlt    
   0x7fffe10126e7:	mov    %cl,0x0(%r13)

源码分析

void TemplateTable::iload_internal(RewriteControl rc) {
  transition(vtos, itos);
  if (RewriteFrequentPairs && rc == may_rewrite) {
    Label rewrite, done;
    Register bc = r4;

    // get next bytecode
    __ load_unsigned_byte(r1, at_bcp(Bytecodes::length_for(Bytecodes::_iload)));     

    // if _iload, wait to rewrite to iload2.  We only want to rewrite the
    // last two iloads in a pair.  Comparing against fast_iload means that
    // the next bytecode is neither an iload or a caload, and therefore
    // an iload pair.
    __ cmpw(r1, Bytecodes::_iload);         <--- 下一个bytecode
    __ br(Assembler::EQ, done);             <---- 跳转到done

    // if _fast_iload rewrite to _fast_iload2
    __ cmpw(r1, Bytecodes::_fast_iload);       <-- 判断下一个是否是_fast_iload
    __ movw(bc, Bytecodes::_fast_iload2);    <------ 下一个是_fast_iload 则重写成fast_iload2
    __ br(Assembler::EQ, rewrite);           <--------  跳转到rewrite label  

    // if _caload rewrite to _fast_icaload
    __ cmpw(r1, Bytecodes::_caload);
    __ movw(bc, Bytecodes::_fast_icaload);
    __ br(Assembler::EQ, rewrite);

    // else rewrite to _fast_iload
    __ movw(bc, Bytecodes::_fast_iload);

    // rewrite
    // bc: new bytecode
    __ bind(rewrite);
    patch_bytecode(Bytecodes::_iload, bc, r1, false);
    __ bind(done);

  }

  // do iload, get the local value into tos
  locals_index(r1);
  __ ldr(r0, iaddress(r1));

}

aconst_null​

bytecodedescenum
aconst_nullpush a null reference onto the stack0x01
void TemplateTable::aconst_null() {
  transition(vtos, atos);
  __ xorl(rax, rax);   // rax 就是栈顶
}

istore​

bytecodedescenum
istoreStore int into local variable54, // 0x36

可以通过这个bytecode 了解怎么访问本地变量

void TemplateTable::istore() {
  transition(itos, vtos);   // 这里只是一个断言assert , 断言之前的状态是itos , 之后的状态是vtos , 实际上是由def来定义的
  locals_index(rbx); // 将偏移 也就是index 写入rbx
  __ movl(iaddress(rbx), rax); //iaddress 就是 rlocal + rbx 也就是获取最后的跳转地址 ,然后将rax写入偏移地址
}

这里iaddress(rbx) 其实是rlocals+rbx 的偏移,也就是相对于本地变量的偏移

static inline Address iaddress(Register r) {
  return Address(rlocals, r, Address::times_ptr);
}

iaddress 的源码在这里: src\hotspot\cpu\x86\assembler_x86.hpp 调用顺序是iaddress -> Address

static inline Address iaddress(Register r) {
  return Address(rlocals, r, Address::times_ptr);
}
  Address(Register base, Register index, ScaleFactor scale, int disp = 0)
    : _base (base),
      _index(index),
      _xmmindex(xnoreg),
      _scale(scale),
      _disp (disp),
      _isxmmindex(false) {
    assert(!index->is_valid() == (scale == Address::no_scale),
           "inconsistent address");
  }

def istore展开​

前面不是看到transition(itos, vtos); , 这个transition只是一个类似测试时候的断言,真正是在def 处理的

  def(Bytecodes::_istore              , ubcp|____|clvm|____, itos, vtos, istore              ,  _           );

下面我们看看def展开,会慢慢展开成

void TemplateTable::def(Bytecodes::Code code, int flags, TosState in, TosState out, void (*gen)(int arg), int arg) {
  ...
  Template* t = is_wide ? template_for_wide(code) : template_for(code);
  // setup entry
  t->initialize(flags, in, out, gen, arg);
  assert(t->bytecode() == code, "just checkin'");
}

这里的 in 和out 会在TemplateInterpreterGenerator::generate_and_dispatch的时候使用


//------------------------------------------------------------------------------------------------------------------------

void TemplateInterpreterGenerator::generate_and_dispatch(Template* t, TosState tos_out) {
#ifndef PRODUCT
  // debugging code
  if (CountBytecodes || TraceBytecodes || StopInterpreterAt > 0) count_bytecode();
  if (PrintBytecodeHistogram)                                    histogram_bytecode(t);
  if (PrintBytecodePairHistogram)                                histogram_bytecode_pair(t);
  if (TraceBytecodes)                                            trace_bytecode(t);
  if (StopInterpreterAt > 0)                                     stop_interpreter_at();
  __ verify_FPU(1, t->tos_in());
#endif // !PRODUCT
  int step = 0;
  if (!t->does_dispatch()) {
    step = t->is_wide() ? Bytecodes::wide_length_for(t->bytecode()) : Bytecodes::length_for(t->bytecode());
    if (tos_out == ilgl) tos_out = t->tos_out();
    // compute bytecode size
    assert(step > 0, "just checkin'");
    // setup stuff for dispatching next bytecode
    if (ProfileInterpreter && VerifyDataPointer
        && MethodData::bytecode_has_profile(t->bytecode())) {
      __ verify_method_data_pointer();
    }
    __ dispatch_prolog(tos_out, step);
  }
  // generate template
  t->generate(_masm);
  // advance
  if (t->does_dispatch()) {
#ifdef ASSERT
    // make sure execution doesn't go beyond this point if code is broken
    __ should_not_reach_here();
#endif // ASSERT
  } else {
    // dispatch to next bytecode
    __ dispatch_epilog(tos_out, step);
  }
}

reference​

utf8 and utf16 and encoding and java

· One min read

背景​

java 的字符串会设计很多编码相关的问题,全部整理一下

知识点​

Code Unit​

code unit 描述的是一个编码的的最小单位(注意一个Unicode 平面对应的字符可能由多个code unit 组成)

编码unit code
utf-81字节
utf-162字节

java 的char​

java 的char 是2个字节,类型的范围是 0 到 2^16 - 1.

clickhouse mybatis batch insert cpu raise up

· 2 min read

背景​

背景

使用jdbc clickhouse,批量写入,发现cpu升高非常多,升到了90%多

排查及原因​

相关环境​

jdk: jdk11 clickhouse 使用的sql驱动:

               <groupId>com.clickhouse</groupId>
                <artifactId>clickhouse-jdbc</artifactId>
                <version>0.3.2-patch11</version>

原因: 使用mybatis plus 拼写sql,批量写入2000条 sql 用mybatis 的xml 拼写 类似:

insert into table values (row1_field1 , row1_field2 ),(row1_field1 , row1_field2) .... 这里是用mybatis 的xml foreache 2000 次

线上warning日志:

Please consider to use one and only one values expression, for example: use 'values(?)' instead of 'values(?),(?)'.

由于jdbc 的parser 比较慢,需要将perpare语句改成以下形式:

insert into table values ( ?, ? )  ## 只有一次

采用的是

// Note: "insert into table values(?,?)" is treated as "insert into mytable"
try (PreparedStatement ps = conn.prepareStatement("insert into table values(?,?)")) {
    ps.setString(1, "test"); // id
    ps.setObject(2, LocalDateTime.now()); // timestamp
    ps.addBatch(); // append parameters to the query
    ...
    ps.executeBatch(); // issue the composed query: insert into mytable values(...)(...)...(...)
}

结果​

cpu 从80%降低到30%以内 优化前:

优化前

优化后: 优化后

相关阅读​

java 常用命令

· One min read
命令描述
java -XshowSettings:all -version获取所有配置

Garbage-First Garbage Collection 简单概况

· 8 min read

Garbage-First Garbage Collection​

Garbage-First简称g1算法，是java目前可选的一个gc算法。

目标： 在一定的软实时性条件下，保证整体的吞吐

算法构成：

  • 堆等大小： 整个内存堆被划分为相同大小的块。

The Garbage-First collector achieves these goals via sev- eral techniques. The heap is partitioned into a set of equal- sized heap regions, much like the train cars of the Mature- Object Space collector of Hudson and Moss [22]. However, whereas the remembered sets of the Mature-Object Space collector are unidirectional, recording pointers from older regions to younger but not vice versa, Garbage-First remem- bered sets record pointers from all regions (with some excep- tions, described in sections 2.4 and 4.6). Recording all ref- erences allows an arbitrary set of heap regions to be chosen for collection. A concurrent thread processes log records cre- ated by special mutator write barriers to keep remembered sets up-to-date, allowing shorter collections.

源码分析​

  • Young gc
void G1YoungCollector::collect() {
  // Do timing/tracing/statistics/pre- and post-logging/verification work not
  // directly related to the collection. They should not be accounted for in
  // collection work timing.

  // The G1YoungGCTraceTime message depends on collector state, so must come after
  // determining collector state.
  G1YoungGCTraceTime tm(this, _gc_cause);

  // JFR
  G1YoungGCJFRTracerMark jtm(gc_timer_stw(), gc_tracer_stw(), _gc_cause);
  // JStat/MXBeans
  G1MonitoringScope ms(monitoring_support(),
                       false /* full_gc */,
                       collector_state()->in_mixed_phase() /* all_memory_pools_affected */);
  // Create the heap printer before internal pause timing to have
  // heap information printed as last part of detailed GC log.
  G1HeapPrinterMark hpm(_g1h);
  // Young GC internal pause timing
  G1YoungGCNotifyPauseMark npm(this);

  // Verification may use the workers, so they must be set up before.
  // Individual parallel phases may override this.
  set_young_collection_default_active_worker_threads();

  // Wait for root region scan here to make sure that it is done before any
  // use of the STW workers to maximize cpu use (i.e. all cores are available
  // just to do that).
  wait_for_root_region_scanning();

  G1YoungGCVerifierMark vm(this);
  {
    // Actual collection work starts and is executed (only) in this scope.

    // Young GC internal collection timing. The elapsed time recorded in the
    // policy for the collection deliberately elides verification (and some
    // other trivial setup above).
    policy()->record_young_collection_start();

    calculate_collection_set(jtm.evacuation_info(), _target_pause_time_ms);

    G1RedirtyCardsQueueSet rdcqs(G1BarrierSet::dirty_card_queue_set().allocator());
    G1PreservedMarksSet preserved_marks_set(workers()->active_workers());
    G1ParScanThreadStateSet per_thread_states(_g1h,
                                              &rdcqs,
                                              &preserved_marks_set,
                                              workers()->active_workers(),
                                              collection_set()->young_region_length(),
                                              collection_set()->optional_region_length(),
                                              &_evac_failure_regions);

    pre_evacuate_collection_set(jtm.evacuation_info(), &per_thread_states);

    bool may_do_optional_evacuation = collection_set()->optional_region_length() != 0;
    // Actually do the work...
    evacuate_initial_collection_set(&per_thread_states, may_do_optional_evacuation);

    if (may_do_optional_evacuation) {
      evacuate_optional_collection_set(&per_thread_states);
    }
    post_evacuate_collection_set(jtm.evacuation_info(), &per_thread_states);

    // Refine the type of a concurrent mark operation now that we did the
    // evacuation, eventually aborting it.
    _concurrent_operation_is_full_mark = policy()->concurrent_operation_is_full_mark("Revise IHOP");

    // Need to report the collection pause now since record_collection_pause_end()
    // modifies it to the next state.
    jtm.report_pause_type(collector_state()->young_gc_pause_type(_concurrent_operation_is_full_mark));

    policy()->record_young_collection_end(_concurrent_operation_is_full_mark, evacuation_failed());
  }
  TASKQUEUE_STATS_ONLY(print_taskqueue_stats());
  TASKQUEUE_STATS_ONLY(reset_taskqueue_stats());
}

真正复制的代码

evacuate_live_objects(


MAYBE_INLINE_EVACUATION
void G1ParScanThreadState::dispatch_task(ScannerTask task) {
  verify_task(task);
  if (task.is_narrow_oop_ptr()) {
    do_oop_evac(task.to_narrow_oop_ptr());
  } else if (task.is_oop_ptr()) {   //oop 复制
    do_oop_evac(task.to_oop_ptr());
  } else {
    do_partial_array(task.to_partial_array_task());
  }
}

堆栈：

Thread 23 "GC Thread#4" hit Breakpoint 1, G1ParScanThreadState::do_copy_to_survivor_space (this=0x7fff7c000d90, region_attr=..., old=0x716809d28, old_mark=...) at /home/dai/jdk/src/hotspot/share/gc/g1/g1ParScanThreadState.cpp:443
443	  assert(region_attr.is_in_cset(),
(gdb) bt
#0  G1ParScanThreadState::do_copy_to_survivor_space (this=0x7fff7c000d90, region_attr=..., old=0x716809d28, old_mark=...) at /home/dai/jdk/src/hotspot/share/gc/g1/g1ParScanThreadState.cpp:443
#1  0x00007ffff64ab3f6 in G1ParScanThreadState::copy_to_survivor_space (this=0x7fff7c000d90, region_attr=..., old=0x716809d28, old_mark=...)
    at /home/dai/jdk/src/hotspot/share/gc/g1/g1ParScanThreadState.cpp:555
#2  0x00007ffff64de15e in G1ParCopyClosure<(G1Barrier)0, false>::do_oop_work<oopDesc*> (this=0x7fff7c001478, p=0x7ffff02e1cc8) at /home/dai/jdk/src/hotspot/share/gc/g1/g1OopClosures.inline.hpp:240
#3  0x00007ffff64dcbc3 in G1ParCopyClosure<(G1Barrier)0, false>::do_oop (this=0x7fff7c001478, p=0x7ffff02e1cc8) at /home/dai/jdk/src/hotspot/share/gc/g1/g1OopClosures.hpp:167
#4  0x00007ffff6546dd8 in chunk_oops_do (f=0x7fff7c001478, chunk=0x7ffff02e1cb0, chunk_top=0x7ffff02e1cd0 "\350\034\200\026\a") at /home/dai/jdk/src/hotspot/share/runtime/handles.cpp:100
#5  0x00007ffff6546e23 in HandleArea::oops_do (this=0x7ffff02e1c30, f=0x7fff7c001478) at /home/dai/jdk/src/hotspot/share/runtime/handles.cpp:108
#6  0x00007ffff6d85dd4 in Thread::oops_do_no_frames (this=0x7ffff02e1160, f=0x7fff7c001478, cf=0x7fff7c001520) at /home/dai/jdk/src/hotspot/share/runtime/thread.cpp:550
#7  0x00007ffff6d8a513 in JavaThread::oops_do_no_frames (this=0x7ffff02e1160, f=0x7fff7c001478, cf=0x7fff7c001520) at /home/dai/jdk/src/hotspot/share/runtime/thread.cpp:1968
#8  0x00007ffff6d85e28 in Thread::oops_do (this=0x7ffff02e1160, f=0x7fff7c001478, cf=0x7fff7c001520) at /home/dai/jdk/src/hotspot/share/runtime/thread.cpp:580
#9  0x00007ffff6d91359 in ParallelOopsDoThreadClosure::do_thread (this=0x7fff87dfaa00, t=0x7ffff02e1160) at /home/dai/jdk/src/hotspot/share/runtime/thread.cpp:3620
#10 0x00007ffff6d8c40b in Threads::possibly_parallel_threads_do (is_par=true, tc=0x7fff87dfaa00) at /home/dai/jdk/src/hotspot/share/runtime/thread.cpp:2545
#11 0x00007ffff6d8eac8 in Threads::possibly_parallel_oops_do (is_par=true, f=0x7fff7c001478, cf=0x7fff7c001520) at /home/dai/jdk/src/hotspot/share/runtime/thread.cpp:3626
#12 0x00007ffff64dea53 in G1RootProcessor::process_java_roots (this=0x7fffc9723df0, closures=0x7fff7c001470, phase_times=0x7fffb8001380, worker_id=1)
    at /home/dai/jdk/src/hotspot/share/gc/g1/g1RootProcessor.cpp:183
#13 0x00007ffff64de78e in G1RootProcessor::evacuate_roots (this=0x7fffc9723df0, pss=0x7fff7c000d90, worker_id=1) at /home/dai/jdk/src/hotspot/share/gc/g1/g1RootProcessor.cpp:60
#14 0x00007ffff64f06b8 in G1EvacuateRegionsTask::scan_roots (this=0x7fffc9723f50, pss=0x7fff7c000d90, worker_id=1) at /home/dai/jdk/src/hotspot/share/gc/g1/g1YoungCollector.cpp:706
#15 0x00007ffff64f0632 in G1EvacuateRegionsBaseTask::work (this=0x7fffc9723f50, worker_id=1) at /home/dai/jdk/src/hotspot/share/gc/g1/g1YoungCollector.cpp:693
#16 0x00007ffff6e8bb7c in WorkerTaskDispatcher::worker_run_task (this=0x7ffff00a4c88) at /home/dai/jdk/src/hotspot/share/gc/shared/workerThread.cpp:67
#17 0x00007ffff6e8c074 in WorkerThread::run (this=0x7fffb800df30) at /home/dai/jdk/src/hotspot/share/gc/shared/workerThread.cpp:159
#18 0x00007ffff6d8557f in Thread::call_run (this=0x7fffb800df30) at /home/dai/jdk/src/hotspot/share/runtime/thread.cpp:358
#19 0x00007ffff6acc1e7 in thread_native_entry (thread=0x7fffb800df30) at /home/dai/jdk/src/hotspot/os/linux/os_linux.cpp:705
#20 0x00007ffff7c94ac3 in start_thread (arg=<optimized out>) at ./nptl/pthread_create.c:442
#21 0x00007ffff7d26a40 in clone3 () at ../sysdeps/unix/sysv/linux/x86_64/clone3.S:81

内存复制：

static void pd_disjoint_words(const HeapWord* from, HeapWord* to, size_t count) {
#ifdef AMD64
  switch (count) {
  case 8:  to[7] = from[7];
  case 7:  to[6] = from[6];
  case 6:  to[5] = from[5];
  case 5:  to[4] = from[4];
  case 4:  to[3] = from[3];
  case 3:  to[2] = from[2];
  case 2:  to[1] = from[1];
  case 1:  to[0] = from[0];
  case 0:  break;
  default:
    (void)memcpy(to, from, count * HeapWordSize);
    break;
  }
#else
  // Includes a zero-count check.
  intx temp;
  __asm__ volatile("        testl   %6,%6       ;"
                   "        jz      3f          ;"
                   "        cmpl    $32,%6      ;"
                   "        ja      2f          ;"
                   "        subl    %4,%1       ;"
                   "1:      movl    (%4),%3     ;"
                   "        movl    %7,(%5,%4,1);"
                   "        addl    $4,%0       ;"
                   "        subl    $1,%2        ;"
                   "        jnz     1b          ;"
                   "        jmp     3f          ;"
                   "2:      rep;    smovl       ;"
                   "3:      nop                  "
                   : "=S" (from), "=D" (to), "=c" (count), "=r" (temp)
                   : "0"  (from), "1"  (to), "2"  (count), "3"  (temp)
                   : "memory", "cc");
#endif // AMD64
}

并发标记: D:\jdk\src\hotspot\share\gc\g1\g1ConcurrentMark.inline.hpp

inline bool G1ConcurrentMark::mark_in_bitmap(uint const worker_id, oop const obj) {
  HeapRegion* const hr = _g1h->heap_region_containing(obj);

  if (hr->obj_allocated_since_marking_start(obj)) {
    return false;
  }

  // Some callers may have stale objects to mark above TAMS after humongous reclaim.
  // Can't assert that this is a valid object at this point, since it might be in the process of being copied by another thread.
  assert(!hr->is_continues_humongous(), "Should not try to mark object " PTR_FORMAT " in Humongous continues region %u above TAMS " PTR_FORMAT, p2i(obj), hr->hrm_index(), p2i(hr->top_at_mark_start()));

  bool success = _mark_bitmap.par_mark(obj);
  if (success) {
    add_to_liveness(worker_id, obj, obj->size());
  }
  return success;
}

分配内存：

HeapWord*
G1CollectedHeap::mem_allocate(size_t word_size,
                              bool*  gc_overhead_limit_was_exceeded) {
  assert_heap_not_locked_and_not_at_safepoint();

  if (is_humongous(word_size)) {
    return attempt_allocation_humongous(word_size);
  }
  size_t dummy = 0;
  return attempt_allocation(word_size, word_size, &dummy);
}
  product(uint, GCCardSizeInBytes, 512,                                     \
          "Card table entry size (in bytes) for card based collectors")     \
          range(128, NOT_LP64(512) LP64_ONLY(1024))                         \
          constraint(GCCardSizeInBytesConstraintFunc,AtParse)


cartable 为什么是512

  _card_size = GCCardSizeInBytes;
  _card_shift = log2i_exact(_card_size);
  _card_size_in_words = _card_size / sizeof(HeapWord);
template <DecoratorSet decorators, typename BarrierSetT>
template <typename T>
inline void G1BarrierSet::AccessBarrier<decorators, BarrierSetT>::
oop_store_not_in_heap(T* addr, oop new_value) {
  // Apply SATB barriers for all non-heap references, to allow
  // concurrent scanning of such references.
  G1BarrierSet *bs = barrier_set_cast<G1BarrierSet>(BarrierSet::barrier_set());
  bs->write_ref_field_pre<decorators>(addr);
  Raw::oop_store(addr, new_value);
}

写入barrier逻辑​

写入barrier逻辑

void G1BarrierSetAssembler::oop_store_at(MacroAssembler* masm, DecoratorSet decorators, BasicType type,
                                         Address dst, Register val, Register tmp1, Register tmp2, Register tmp3) {
  bool in_heap = (decorators & IN_HEAP) != 0;
  bool as_normal = (decorators & AS_NORMAL) != 0;

  bool needs_pre_barrier = as_normal;
  bool needs_post_barrier = val != noreg && in_heap;

  Register rthread = LP64_ONLY(r15_thread) NOT_LP64(rcx);
  // flatten object address if needed
  // We do it regardless of precise because we need the registers
  if (dst.index() == noreg && dst.disp() == 0) {
    if (dst.base() != tmp1) {
      __ movptr(tmp1, dst.base());
    }
  } else {
    __ lea(tmp1, dst);
  }

#ifndef _LP64
  InterpreterMacroAssembler *imasm = static_cast<InterpreterMacroAssembler*>(masm);
#endif

  NOT_LP64(__ get_thread(rcx));
  NOT_LP64(imasm->save_bcp());

  if (needs_pre_barrier) {
    g1_write_barrier_pre(masm /*masm*/,
                         tmp1 /* obj */,
                         tmp2 /* pre_val */,
                         rthread /* thread */,
                         tmp3  /* tmp */,
                         val != noreg /* tosca_live */,
                         false /* expand_call */);
  }
  if (val == noreg) {
    BarrierSetAssembler::store_at(masm, decorators, type, Address(tmp1, 0), val, noreg, noreg, noreg);
  } else {
    Register new_val = val;
    if (needs_post_barrier) {
      // G1 barrier needs uncompressed oop for region cross check.
      if (UseCompressedOops) {
        new_val = tmp2;
        __ movptr(new_val, val);
      }
    }
    BarrierSetAssembler::store_at(masm, decorators, type, Address(tmp1, 0), val, noreg, noreg, noreg);
    if (needs_post_barrier) {
      g1_write_barrier_post(masm /*masm*/,
                            tmp1 /* store_adr */,
                            new_val /* new_val */,
                            rthread /* thread */,
                            tmp3 /* tmp */,
                            tmp2 /* tmp2 */);
    }
  }
  NOT_LP64(imasm->restore_bcp());
}

hread 2 "java" hit Breakpoint 1, 0x00007ffff6fa66d8 in G1BarrierSetAssembler::oop_store_at(MacroAssembler*, unsigned long, BasicType, Address, Register, Register, Register, Register) ()
   from /usr/lib/jvm/java-21-openjdk-amd64/lib/server/libjvm.so
(gdb) bt
#0  0x00007ffff6fa66d8 in G1BarrierSetAssembler::oop_store_at(MacroAssembler*, unsigned long, BasicType, Address, Register, Register, Register, Register) ()
   from /usr/lib/jvm/java-21-openjdk-amd64/lib/server/libjvm.so
#1  0x00007ffff7482012 in ModRefBarrierSetAssembler::store_at(MacroAssembler*, unsigned long, BasicType, Address, Register, Register, Register, Register) ()
   from /usr/lib/jvm/java-21-openjdk-amd64/lib/server/libjvm.so
#2  0x00007ffff740992a in MacroAssembler::store_heap_oop(Address, Register, Register, Register, Register, unsigned long) ()
   from /usr/lib/jvm/java-21-openjdk-amd64/lib/server/libjvm.so
#3  0x00007ffff77506ec in TemplateTable::aastore() ()
   from /usr/lib/jvm/java-21-openjdk-amd64/lib/server/libjvm.so
#4  0x00007ffff7746b47 in Template::generate(InterpreterMacroAssembler*) ()
   from /usr/lib/jvm/java-21-openjdk-amd64/lib/server/libjvm.so
#5  0x00007ffff7736a0c in TemplateInterpreterGenerator::generate_and_dispatch(Template*, TosState) ()
   from /usr/lib/jvm/java-21-openjdk-amd64/lib/server/libjvm.so
#6  0x00007ffff77427f6 in TemplateInterpreterGenerator::set_vtos_entry_points(Template*, unsigned char*&, unsigned char*&, unsigned char*&, unsigned char*&, unsigned char*&, unsigned char*&, unsigned char*&, unsigned char*&, unsigned char*&) () from /usr/lib/jvm/java-21-openjdk-amd64/lib/server/libjvm.so
#7  0x00007ffff77370d7 in TemplateInterpreterGenerator::set_entry_points(Bytecodes::Code) () from /usr/lib/jvm/java-21-openjdk-amd64/lib/server/libjvm.so
#8  0x00007ffff773855b in TemplateInterpreterGenerator::generate_all() ()
--Type <RET> for more, q to quit, c to continue without paging--
   from /usr/lib/jvm/java-21-openjdk-amd64/lib/server/libjvm.so
#9  0x00007ffff7735d45 in TemplateInterpreter::initialize_code() () from /usr/lib/jvm/java-21-openjdk-amd64/lib/server/libjvm.so
#10 0x00007ffff70f0007 in interpreter_init_code() () from /usr/lib/jvm/java-21-openjdk-amd64/lib/server/libjvm.so
#11 0x00007ffff70cc117 in init_globals2() () from /usr/lib/jvm/java-21-openjdk-amd64/lib/server/libjvm.so
#12 0x00007ffff7777839 in Threads::create_vm(JavaVMInitArgs*, bool*) () from /usr/lib/jvm/java-21-openjdk-amd64/lib/server/libjvm.so
#13 0x00007ffff719c702 in JNI_CreateJavaVM () from /usr/lib/jvm/java-21-openjdk-amd64/lib/server/libjvm.so
#14 0x00007ffff7f81123 in JavaMain () from /usr/lib/jvm/java-21-openjdk-amd64/bin/../lib/libjli.so
#15 0x00007ffff7f84f0d in ThreadJavaMain () from /usr/lib/jvm/java-21-openjdk-amd64/bin/../lib/libjli.so
#16 0x00007ffff7e233ec in start_thread (arg=<optimized out>) at ./nptl/pthread_create.c:444
#17 0x00007ffff7ea3a4c in clone3 () at ../sysdeps/unix/sysv/linux/x86_64/clone3.S:81

相关阅读​

java 一次gc排查

· 2 min read

背景​

测试环境有个old generation 非常占用内存.所以需要排查

过程​

问题出现​

同事压测的时候,发现old generation 占用内存非常大 例子是下图的样子: java_gc_generation

相关配置:

  • jdk 11
  • 最大堆内存2g
  • gc算法用默认的g1

排查问题​

dump 文件​

使用jamp命令将java的内存dump出来

jmap -dump:format=b,,live,file=<file-path> <pid>

mat工具分析​

然后去下载mat工具下载地址 安装之后打开,点击Dominator Tree: Dominator Tree

就可以看到对象以及大小

Dominator Tree detail

可以看到有一个900m大小的HashSet , 这是一个去重的set , 每次都会往这里塞入设备id ,当循环结束,会自动不会再被引用,然后会被gc回收.

结论​

这不是内存泄漏,而是一个有1kw 数据的的大set ,也就是一个大对象

我们找运维调大了最大的堆内存,问题解决

后续优化: 后续我们打算不使用HashSet 塞字符串去重,而是用布隆过滤器去重

java 对象大小

· 15 min read

背景​

排查一个gc问题的时候想到需要了解java的对象大小

举例这样创建一个1099 的对象会占用多少个字节呢?

那如果是一个特定的对象NumClass

public class NumClass{
    public int num1 ;
    public int num2 ; 
}

那样创建NunClass[] 又占用多少字节呢?

## 这样 的array占用多少字节呢? 
Object[] array = new Object[10245] ; 
## 这样 的array占用多少字节呢? 
Object[] array = new NumClass[10245] ;

结论是两者占用的字节是一样的

object 数组例子​

先上最简单的new Object的代码，在这个例子里面： 创建了一个长度是10245的对象数组

package com;
public class Hello{
        public static volatile Object[] arr ;
        public static void main(String [] argv){
                arr = new Object[10245];    // 创建一个对象数组 ， 数组也是一个对象 ， 那么这个对象有多大呢？ 
                arr[1] = "hihi";

        }
}

实际上是在64位机器上，没有开启指针压缩的情况下是： 8+ 4 + 10245*8个字节长度

oop 是指针oopDesc* 的别名

typedef class oopDesc*                    oop;
typedef class   instanceOopDesc*            instanceOop;
typedef class   arrayOopDesc*               arrayOop;
typedef class     objArrayOopDesc*            objArrayOop;
typedef class     typeArrayOopDesc*           typeArrayOop;
+-----------++----------++-------++------+           
|           ||          ||       ||      |           
|           ||          ||       ||      |           
|  MarkWord || length   ||  oop  ||  oop |                  
+-----------++----------++-------++------+ repeat 10254 次

jol​

核心函数:

jol\jol-core\src\main\java\org\openjdk\jol\layouters\CurrentLayouter.java

    @Override
    public ClassLayout layout(ClassData data) {
        VirtualMachine vm = VM.current();

        if (data.isArray()) {
            // special case of arrays
            int base = vm.arrayBaseOffset(data.arrayComponentType());
            int scale = vm.arrayIndexScale(data.arrayComponentType());

            long instanceSize = MathUtil.align(base + data.arrayLength() * scale, vm.objectAlignment());

            SortedSet<FieldLayout> result = new TreeSet<>();
            result.add(new FieldLayout(FieldData.create(data.arrayClass(), "<elements>", data.arrayComponentType()), base, scale * data.arrayLength()));
            return ClassLayout.create(data, result, CURRENT, instanceSize, false);
        }

        Collection<FieldData> fields = data.fields();

        SortedSet<FieldLayout> result = new TreeSet<>();
        for (FieldData f : fields) {
            result.add(new FieldLayout(f, vm.fieldOffset(f.refField()), vm.sizeOfField(f.typeClass())));
        }

        long instanceSize;
        if (result.isEmpty()) {
            instanceSize = vm.objectHeaderSize();
        } else {
            FieldLayout f = result.last();
            instanceSize = f.offset() + f.size();
            // TODO: This calculation is incorrect if there is a trailing @Contended field, or the instance is @Contended
        }
        instanceSize = MathUtil.align(instanceSize, vm.objectAlignment());
        return ClassLayout.create(data, result, CURRENT, instanceSize, true);
    }
  static int length_offset_in_bytes() {
    return UseCompressedClassPointers ? klass_gap_offset_in_bytes() :
                               sizeof(arrayOopDesc);
  }

创建一个数组

(gdb) bt
#0  arrayOopDesc::length_offset_in_bytes () at /var/jdk/src/hotspot/share/oops/arrayOop.hpp:83
#1  0x00007f93c2fdc06e in arrayOopDesc::length_addr_impl (obj_ptr=0x7162010c0) at /var/jdk/src/hotspot/share/oops/arrayOop.hpp:67
#2  0x00007f93c2fdc0ba in arrayOopDesc::length (this=0x7162010c0) at /var/jdk/src/hotspot/share/oops/arrayOop.hpp:114
#3  0x00007f93c3239521 in arrayOopDesc::is_within_bounds (this=0x7162010c0, index=10) at /var/jdk/src/hotspot/share/oops/arrayOop.hpp:110
#4  0x00007f93c3239690 in typeArrayOopDesc::byte_at_addr (this=0x7162010c0, which=10) at /var/jdk/src/hotspot/share/oops/typeArrayOop.inline.hpp:48
#5  0x00007f93c323972d in typeArrayOopDesc::byte_at_put (this=0x7162010c0, which=10, contents=112 'p') at /var/jdk/src/hotspot/share/oops/typeArrayOop.inline.hpp:96
#6  0x00007f93c389e259 in java_lang_String::create_from_unicode (unicode=0x7f93bc037c88, length=33, __the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/classfile/javaClasses.cpp:300
#7  0x00007f93c3f7b91d in StringTable::do_intern (string_or_null_h=..., name=0x7f93bc037c88, len=33, hash=2694772907, __the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/classfile/stringTable.cpp:347
#8  0x00007f93c3f7b87e in StringTable::intern (string_or_null_h=..., name=0x7f93bc037c88, len=33, __the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/classfile/stringTable.cpp:336
#9  0x00007f93c3f7b5b0 in StringTable::intern (symbol=0x7f93c0018d38, __the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/classfile/stringTable.cpp:296
#10 0x00007f93c35740b5 in ConstantPool::uncached_string_at (this=0x7f938c02e0a0, which=250, __the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/oops/constantPool.cpp:1171
#11 0x00007f93c36a19cd in fieldDescriptor::string_initial_value (this=0x7f93c2bd3a68, __the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/runtime/fieldDescriptor.cpp:103
#12 0x00007f93c389ff2d in initialize_static_string_field (fd=0x7f93c2bd3a68, mirror=..., __the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/classfile/javaClasses.cpp:809
#13 0x00007f93c38a0444 in initialize_static_field (fd=0x7f93c2bd3a68, mirror=..., __the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/classfile/javaClasses.cpp:866
#14 0x00007f93c3869f51 in InstanceKlass::do_local_static_fields (this=0x8000431a0, f=0x7f93c38a0365 <initialize_static_field(fieldDescriptor*, Handle, JavaThread*)>, mirror=..., __the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/oops/instanceKlass.cpp:1672
#15 0x00007f93c38a08d5 in java_lang_Class::initialize_mirror_fields (k=0x8000431a0, mirror=..., protection_domain=..., classData=..., __the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/classfile/javaClasses.cpp:930
#16 0x00007f93c38a10d1 in java_lang_Class::create_mirror (k=0x8000431a0, class_loader=..., module=..., protection_domain=..., classData=..., __the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/classfile/javaClasses.cpp:1035
#17 0x00007f93c34c6144 in ClassFileParser::fill_instance_klass (this=0x7f93c2bd3dd0, ik=0x8000431a0, changed_by_loadhook=false, cl_inst_info=..., __the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/classfile/classFileParser.cpp:5426
#18 0x00007f93c34c532e in ClassFileParser::create_instance_klass (this=0x7f93c2bd3dd0, changed_by_loadhook=false, cl_inst_info=..., __the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/classfile/classFileParser.cpp:5255
#19 0x00007f93c3b554d1 in KlassFactory::create_from_stream (stream=0x7f93bc036fb0, name=0x7f93c00001f8, loader_data=0x7f93bc121170, cl_info=..., __the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/classfile/klassFactory.cpp:202
#20 0x00007f93c34d86e9 in ClassLoader::load_class (name=0x7f93c00001f8, search_append_only=false, __the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/classfile/classLoader.cpp:1231
#21 0x00007f93c4000806 in SystemDictionary::load_instance_class_impl (class_name=0x7f93c00001f8, class_loader=..., __the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/classfile/systemDictionary.cpp:1289
#22 0x00007f93c4000ba3 in SystemDictionary::load_instance_class (name_hash=2036240099, name=0x7f93c00001f8, class_loader=..., __the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/classfile/systemDictionary.cpp:1354
#23 0x00007f93c3ffeca9 in SystemDictionary::resolve_instance_class_or_null (name=0x7f93c00001f8, class_loader=..., protection_domain=..., __the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/classfile/systemDictionary.cpp:723
#24 0x00007f93c3ffd82e in SystemDictionary::resolve_instance_class_or_null_helper (class_name=0x7f93c00001f8, class_loader=..., protection_domain=..., __the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/classfile/systemDictionary.cpp:294
#25 0x00007f93c3ffd6d4 in SystemDictionary::resolve_or_null (class_name=0x7f93c00001f8, class_loader=..., protection_domain=..., __the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/classfile/systemDictionary.cpp:277
#26 0x00007f93c3ffd617 in SystemDictionary::resolve_or_fail (class_name=0x7f93c00001f8, class_loader=..., protection_domain=..., throw_error=true, __the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/classfile/systemDictionary.cpp:263
#27 0x00007f93c32b8d98 in SystemDictionary::resolve_or_fail (class_name=0x7f93c00001f8, throw_error=true, __the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/classfile/systemDictionary.hpp:100
#28 0x00007f93c40dca98 in vmClasses::resolve (id=vmClassID::Throwable_klass_knum, __the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/classfile/vmClasses.cpp:99
#29 0x00007f93c40dcb96 in vmClasses::resolve_until (limit_id=vmClassID::SoftReference_klass_knum, start_id=@0x7f93c2bd48f0: vmClassID::Cloneable_klass_knum, __the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/classfile/vmClasses.cpp:108
#30 0x00007f93c40dd59a in vmClasses::resolve_through (last_id=vmClassID::Reference_klass_knum, start_id=@0x7f93c2bd48f0: vmClassID::Cloneable_klass_knum, __the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/classfile/vmClasses.hpp:64
#31 0x00007f93c40dce23 in vmClasses::resolve_all (__the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/classfile/vmClasses.cpp:168
#32 0x00007f93c4001ab2 in SystemDictionary::initialize (__the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/classfile/systemDictionary.cpp:1655
#33 0x00007f93c40812fb in Universe::genesis (__the_thread__=0x7f93bc028fa0) at /var/jdk/src/hotspot/share/memory/universe.cpp:335
#34 0x00007f93c408378f in universe2_init () at /var/jdk/src/hotspot/share/memory/universe.cpp:937
#35 0x00007f93c3863d8a in init_globals () at /var/jdk/src/hotspot/share/runtime/init.cpp:132
#36 0x00007f93c404a1ca in Threads::create_vm (args=0x7f93c2bd4d50, canTryAgain=0x7f93c2bd4c5b) at /var/jdk/src/hotspot/share/runtime/thread.cpp:2843
#37 0x00007f93c396f43b in JNI_CreateJavaVM_inner (vm=0x7f93c2bd4da8, penv=0x7f93c2bd4db0, args=0x7f93c2bd4d50) at /var/jdk/src/hotspot/share/prims/jni.cpp:3613
#38 0x00007f93c396f787 in JNI_CreateJavaVM (vm=0x7f93c2bd4da8, penv=0x7f93c2bd4db0, args=0x7f93c2bd4d50) at /var/jdk/src/hotspot/share/prims/jni.cpp:3701
#39 0x00007f93c50e6a6a in InitializeJVM (pvm=0x7f93c2bd4da8, penv=0x7f93c2bd4db0, ifn=0x7f93c2bd4e00) at /var/jdk/src/java.base/share/native/libjli/java.c:1459
#40 0x00007f93c50e35ec in JavaMain (_args=0x7ffedd44e1a0) at /var/jdk/src/java.base/share/native/libjli/java.c:411
#41 0x00007f93c50ea5ec in ThreadJavaMain (args=0x7ffedd44e1a0) at /var/jdk/src/java.base/unix/native/libjli/java_md.c:651
#42 0x00007f93c4f45b43 in start_thread (arg=<optimized out>) at ./nptl/pthread_create.c:442
#43 0x00007f93c4fd6bb4 in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:100

分配数组大小：

  static int array_size(int length) {
    const uint OopsPerHeapWord = HeapWordSize/heapOopSize;
    assert(OopsPerHeapWord >= 1 && (HeapWordSize % heapOopSize == 0),
           "Else the following (new) computation would be in error");
    uint res = ((uint)length + OopsPerHeapWord - 1)/OopsPerHeapWord;
#ifdef ASSERT
    // The old code is left in for sanity-checking; it'll
    // go away pretty soon. XXX
    // Without UseCompressedOops, this is simply:
    // oop->length() * HeapWordsPerOop;
    // With narrowOops, HeapWordsPerOop is 1/2 or equal 0 as an integer.
    // The oop elements are aligned up to wordSize
    const uint HeapWordsPerOop = heapOopSize/HeapWordSize;
    uint old_res;
    if (HeapWordsPerOop > 0) {
      old_res = length * HeapWordsPerOop;
    } else {
      old_res = align_up((uint)length, OopsPerHeapWord)/OopsPerHeapWord;
    }
    assert(res == old_res, "Inconsistency between old and new.");
#endif  // ASSERT
    return res;
  }
Thread 2 "java" hit Breakpoint 5, jni_invoke_static (env=0x7fca48029310, result=0x7fca4c534bf0, receiver=0x0, call_type=JNI_STATIC, method_id=0x7fca48542d50, args=0x7fca4c534c60, __the_thread__=0x7fca48029030) at /var/jdk/src/hotspot/share/prims/jni.cpp:881
881       args->push_arguments_on(&java_args);
(gdb) p method._value->print()
{method}
 - this oop:          0x00007fca14411240
 - method holder:     'com/Hello'
 - constants:         0x00007fca14411020 constant pool [20] {0x00007fca14411020} for 'com/Hello' cache=0x00007fca14411348
 - access:            0x9  public static 
 - name:              'main'
 - signature:         '([Ljava/lang/String;)V'
 - max stack:         2
 - max locals:        1
 - size of params:    1
 - method size:       13
 - vtable index:      -2
 - i2i entry:         0x00007fca3900dc00
 - adapters:          AHE@0x00007fca4812b8d0: 0xb i2c: 0x00007fca39114d60 c2i: 0x00007fca39114e1a c2iUV: 0x00007fca39114de4 c2iNCI: 0x00007fca39114e57
 - compiled entry     0x00007fca39114e1a
 - code size:         10
 - code start:        0x00007fca14411230
 - code end (excl):   0x00007fca1441123a
 - checked ex length: 0
 - linenumber start:  0x00007fca1441123a
 - localvar length:   0
$17 = void
(gdb) bt
#0  jni_invoke_static (env=0x7fca48029310, result=0x7fca4c534bf0, receiver=0x0, call_type=JNI_STATIC, method_id=0x7fca48542d50, args=0x7fca4c534c60, __the_thread__=0x7fca48029030) at /var/jdk/src/hotspot/share/prims/jni.cpp:881
#1  0x00007fca4d2c141c in jni_CallStaticVoidMethod (env=0x7fca48029310, cls=0x7fca4802b368, methodID=0x7fca48542d50) at /var/jdk/src/hotspot/share/prims/jni.cpp:1710
#2  0x00007fca4ea4415e in JavaMain (_args=0x7fff1ad56b60) at /var/jdk/src/java.base/share/native/libjli/java.c:545
#3  0x00007fca4ea4a5ec in ThreadJavaMain (args=0x7fff1ad56b60) at /var/jdk/src/java.base/unix/native/libjli/java_md.c:651
#4  0x00007fca4e8a5b43 in start_thread (arg=<optimized out>) at ./nptl/pthread_create.c:442
#5  0x00007fca4e936bb4 in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:100
Thread 2 "java" hit Breakpoint 5, jni_invoke_static (env=0x7f8674029310, result=0x7f86791b0bf0, receiver=0x0, call_type=JNI_STATIC, method_id=0x7f8674652560, args=0x7f86791b0c60, __the_thread__=0x7f8674029030) at /var/jdk/src/hotspot/share/prims/jni.cpp:881
881       args->push_arguments_on(&java_args);
(gdb) p method._value->print()
{method}
 - this oop:          0x00007f8644411240
 - method holder:     'com/Hello'
 - constants:         0x00007f8644411020 constant pool [20] {0x00007f8644411020} for 'com/Hello' cache=0x00007f8644411348
 - access:            0x9  public static 
 - name:              'main'
 - signature:         '([Ljava/lang/String;)V'
 - max stack:         2
 - max locals:        1
 - size of params:    1
 - method size:       13
 - vtable index:      -2
 - i2i entry:         0x00007f866500dc00
 - adapters:          AHE@0x00007f867412b8d0: 0xb i2c: 0x00007f8665114d60 c2i: 0x00007f8665114e1a c2iUV: 0x00007f8665114de4 c2iNCI: 0x00007f8665114e57
 - compiled entry     0x00007f8665114e1a
 - code size:         10
 - code start:        0x00007f8644411230
 - code end (excl):   0x00007f864441123a
 - checked ex length: 0
 - linenumber start:  0x00007f864441123a
 - localvar length:   0
$23 = void
(gdb) enable 1
(gdb) c
Continuing.

Thread 2 "java" hit Breakpoint 1, oopFactory::new_objArray (klass=0x800040f80, length=2019, __the_thread__=0x7f8674029030) at /var/jdk/src/hotspot/share/memory/oopFactory.cpp:118
118       assert(klass->is_klass(), "must be instance class");
(gdb) bt
#0  oopFactory::new_objArray (klass=0x800040f80, length=2019, __the_thread__=0x7f8674029030) at /var/jdk/src/hotspot/share/memory/oopFactory.cpp:118
#1  0x00007f8679e68a5b in InterpreterRuntime::anewarray (current=0x7f8674029030, pool=0x7f8644411020, index=2, size=2019) at /var/jdk/src/hotspot/share/interpreter/interpreterRuntime.cpp:254
#2  0x00007f8665024083 in ?? ()
#3  0x00007f867b4520a0 in TemplateInterpreter::_active_table () from /var/jdk/build/linux-x86_64-server-slowdebug/jdk/lib/server/libjvm.so
#4  0x00007f8665024002 in ?? ()
#5  0x00007f86791b07b0 in ?? ()
#6  0x00007f8644411233 in ?? ()
#7  0x00007f86791b0808 in ?? ()
#8  0x00007f8644411348 in ?? ()
#9  0x0000000000000000 in ?? ()
Bottom (innermost) frame selected; you cannot go down.
(gdb) p _do_zero 
$35 = true
(gdb) n
413       if (_do_zero) {
(gdb) n
414         mem_clear(mem);
(gdb) n
416       arrayOopDesc::set_length(mem, _length);
(gdb) l
411       // concurrent GC.
412       assert(_length >= 0, "length should be non-negative");
413       if (_do_zero) {
414         mem_clear(mem);
415       }
416       arrayOopDesc::set_length(mem, _length);
417       return finish(mem);
418     }
419
420     oop ClassAllocator::initialize(HeapWord* mem) const {
(gdb) up
#1  0x00007f867a27af4b in MemAllocator::allocate (this=0x7f86791b0650) at /var/jdk/src/hotspot/share/gc/shared/memAllocator.cpp:365
365           obj = initialize(mem);
(gdb) down
#0  ObjArrayAllocator::initialize (this=0x7f86791b0650, mem=0x715e73dd0) at /var/jdk/src/hotspot/share/gc/shared/memAllocator.cpp:416
416       arrayOopDesc::set_length(mem, _length);
(gdb) s
arrayOopDesc::set_length (mem=0x715e73dd0, length=2019) at /var/jdk/src/hotspot/share/oops/arrayOop.hpp:122
122         *length_addr_impl(mem) = length;
(gdb) s
arrayOopDesc::length_addr_impl (obj_ptr=0x715e73dd0) at /var/jdk/src/hotspot/share/oops/arrayOop.hpp:66
66          char* ptr = static_cast<char*>(obj_ptr);
(gdb) l
61          return (int)hs;
62        }
63
64        // Returns the address of the length "field".  See length_offset_in_bytes().
65        static int* length_addr_impl(void* obj_ptr) {
66          char* ptr = static_cast<char*>(obj_ptr);
67          return reinterpret_cast<int*>(ptr + length_offset_in_bytes());
68        }
69
70        // Check whether an element of a typeArrayOop with the given type must be
(gdb) n
67          return reinterpret_cast<int*>(ptr + length_offset_in_bytes());
(gdb) s
arrayOopDesc::length_offset_in_bytes () at /var/jdk/src/hotspot/share/oops/arrayOop.hpp:83
83                                     sizeof(arrayOopDesc);
(gdb) l
78        // The _length field is not declared in C++.  It is allocated after the
79        // declared nonstatic fields in arrayOopDesc if not compressed, otherwise
80        // it occupies the second half of the _klass field in oopDesc.
81        static int length_offset_in_bytes() {
82          return UseCompressedClassPointers ? klass_gap_offset_in_bytes() :
83                                     sizeof(arrayOopDesc);
84        }
85
86        // Returns the offset of the first element.
87        static int base_offset_in_bytes(BasicType type) {
(gdb) p Use
Display all 161 possibilities? (y or n)
(gdb) p UseCompressedClassPointers   
$36 = true
(gdb) s
82          return UseCompressedClassPointers ? klass_gap_offset_in_bytes() :
(gdb) s
oopDesc::klass_gap_offset_in_bytes () at /var/jdk/src/hotspot/share/oops/oop.hpp:307
307         assert(has_klass_gap(), "only applicable to compressed klass pointers");
(gdb) l
302
303       // for code generation
304       static int mark_offset_in_bytes()      { return offset_of(oopDesc, _mark); }
305       static int klass_offset_in_bytes()     { return offset_of(oopDesc, _metadata._klass); }
306       static int klass_gap_offset_in_bytes() {
307         assert(has_klass_gap(), "only applicable to compressed klass pointers");
308         return klass_offset_in_bytes() + sizeof(narrowKlass);
309       }
310
311       // for error reporting
(gdb) n
oopDesc::klass_gap_offset_in_bytes () at /var/jdk/src/hotspot/share/oops/oop.hpp:308
308         return klass_offset_in_bytes() + sizeof(narrowKlass);
(gdb) n
309       }
(gdb) n
arrayOopDesc::length_offset_in_bytes () at /var/jdk/src/hotspot/share/oops/arrayOop.hpp:83
83                                     sizeof(arrayOopDesc);
(gdb) n
84        }
(gdb) n
arrayOopDesc::length_addr_impl (obj_ptr=0x715e73dd0) at /var/jdk/src/hotspot/share/oops/arrayOop.hpp:68
68        }
(gdb) n
arrayOopDesc::set_length (mem=0x715e73dd0, length=2019) at /var/jdk/src/hotspot/share/oops/arrayOop.hpp:122
122         *length_addr_impl(mem) = length;
(gdb) n              
123       }
(gdb) p  (int *)mem@20
Only values in memory can be extended with '@'.
(gdb) p  *(int *)mem@20
$37 = {-1163019586, -1163019586, -1163019586, 2019, 0 <repeats 16 times>}

byte code 生成入口​

// jdk/src/hotspot/share/interpreter/templateInterpreter.cpp
void DispatchTable::set_entry(int i, EntryPoint& entry) {
  assert(0 <= i && i < length, "index out of bounds");
  assert(number_of_states == 10, "check the code below");
  _table[btos][i] = entry.entry(btos);
  _table[ztos][i] = entry.entry(ztos);
  _table[ctos][i] = entry.entry(ctos);
  _table[stos][i] = entry.entry(stos);
  _table[atos][i] = entry.entry(atos);
  _table[itos][i] = entry.entry(itos);
  _table[ltos][i] = entry.entry(ltos);
  _table[ftos][i] = entry.entry(ftos);
  _table[dtos][i] = entry.entry(dtos);
  _table[vtos][i] = entry.entry(vtos);
}
//src/hotspot/share/interpreter/bytecodes.hpp
   _new                  = 187, // 0xbb
    _newarray             = 188, // 0xbc
    _anewarray            = 189, // 0xbd
    _arraylength          = 190, // 0xbe

堆栈：

Thread 2 "java" hit Breakpoint 13, 0x00007fffe1011b13 in ?? ()
(gdb) x/20i $pc
=> 0x7fffe1011b13:	movzwl 0x1(%r13),%eax
   0x7fffe1011b18:	bswap  %eax
   0x7fffe1011b1a:	sar    $0x10,%eax
   0x7fffe1011b1d:	movzbl 0x3(%r13),%ebx
   0x7fffe1011b22:	add    $0x3,%r13
   0x7fffe1011b26:	movabs $0x7ffff7bca0a0,%r10
   0x7fffe1011b30:	jmp    *(%r10,%rbx,8)
   0x7fffe1011b34:	nop
   0x7fffe1011b35:	nop
   0x7fffe1011b36:	nop
   0x7fffe1011b37:	nop
   0x7fffe1011b38:	int3   
   0x7fffe1011b39:	int3   
   0x7fffe1011b3a:	int3   
   0x7fffe1011b3b:	int3   
   0x7fffe1011b3c:	int3   
   0x7fffe1011b3d:	int3   
   0x7fffe1011b3e:	int3   
   0x7fffe1011b3f:	int3   
   0x7fffe1011b40:	and    %al,(%rax,%rax,1)

0x00007fffe1011b1d in ?? ()
(gdb) info registers 
rax            0x2805              10245
rbx            0x11                17
rcx            0x2                 2
rdx            0x8                 8
rsi            0x555555581230      93824992416304
rdi            0x7ffff0028f70      140737220087664
rbp            0x7ffff59fe7f8      0x7ffff59fe7f8
rsp            0x7ffff59fe7b0      0x7ffff59fe7b0
r8             0x8                 8
r9             0x0                 0
r10            0x7ffff7bcc8a0      140737349732512
r11            0x216               534
r12            0x0                 0
r13            0x7fffb4411230      140736217551408
r14            0x7ffff59fe808      140737314285576
r15            0x7ffff0028f70      140737220087664
rip            0x7fffe1011b1d      0x7fffe1011b1d
eflags         0x216               [ PF AF IF ]
cs             0x33                51
ss             0x2b                43
ds             0x0                 0
es             0x0                 0
fs             0x0                 0
gs             0x0                 0

anewarray 汇编代码：

(gdb) x/20i $pc
=> 0x7fffe102400a:	lea    0x8(%rsp),%rax
   0x7fffe102400f:	mov    %r13,-0x40(%rbp)
   0x7fffe1024013:	cmpq   $0x0,-0x10(%rbp)
   0x7fffe102401b:	je     0x7fffe1024035
   0x7fffe1024021:	movabs $0x7ffff71becc8,%rdi
   0x7fffe102402b:	and    $0xfffffffffffffff0,%rsp
   0x7fffe102402f:	call   0x7ffff694f3c0 <_ZN14MacroAssembler7debug64EPclPl>
   0x7fffe1024034:	hlt    
   0x7fffe1024035:	push   %r10
   0x7fffe1024037:	cmp    0x16ae2ec2(%rip),%r12        # 0x7ffff7b06f00 <_ZN14CompressedOops11_narrow_oopE>
   0x7fffe102403e:	je     0x7fffe1024058
   0x7fffe1024044:	movabs $0x7ffff7311c28,%rdi
   0x7fffe102404e:	and    $0xfffffffffffffff0,%rsp
   0x7fffe1024052:	call   0x7ffff694f3c0 <_ZN14MacroAssembler7debug64EPclPl>
   0x7fffe1024057:	hlt    
   0x7fffe1024058:	pop    %r10
   0x7fffe102405a:	mov    %r15,%rdi
   0x7fffe102405d:	vzeroupper 
   0x7fffe1024060:	mov    %rbp,0x2d0(%r15)
   0x7fffe1024067:	mov    %rax,0x2c0(%r15)
(gdb) x/200i $pc
=> 0x7fffe102400a:	lea    0x8(%rsp),%rax
   0x7fffe102400f:	mov    %r13,-0x40(%rbp)
   0x7fffe1024013:	cmpq   $0x0,-0x10(%rbp)
   0x7fffe102401b:	je     0x7fffe1024035
   0x7fffe1024021:	movabs $0x7ffff71becc8,%rdi
   0x7fffe102402b:	and    $0xfffffffffffffff0,%rsp
   0x7fffe102402f:	call   0x7ffff694f3c0 <_ZN14MacroAssembler7debug64EPclPl>
   0x7fffe1024034:	hlt    
   0x7fffe1024035:	push   %r10
   0x7fffe1024037:	cmp    0x16ae2ec2(%rip),%r12        # 0x7ffff7b06f00 <_ZN14CompressedOops11_narrow_oopE>
   0x7fffe102403e:	je     0x7fffe1024058
   0x7fffe1024044:	movabs $0x7ffff7311c28,%rdi
   0x7fffe102404e:	and    $0xfffffffffffffff0,%rsp
   0x7fffe1024052:	call   0x7ffff694f3c0 <_ZN14MacroAssembler7debug64EPclPl>
   0x7fffe1024057:	hlt    
   0x7fffe1024058:	pop    %r10
   0x7fffe102405a:	mov    %r15,%rdi
   0x7fffe102405d:	vzeroupper 
   0x7fffe1024060:	mov    %rbp,0x2d0(%r15)
   0x7fffe1024067:	mov    %rax,0x2c0(%r15)
   0x7fffe102406e:	test   $0xf,%esp
   0x7fffe1024074:	je     0x7fffe102408c
   0x7fffe102407a:	sub    $0x8,%rsp
   0x7fffe102407e:	call   0x7ffff65cf968 <_ZN18InterpreterRuntime9anewarrayEP10JavaThreadP12ConstantPoolii>
   0x7fffe1024083:	add    $0x8,%rsp
   0x7fffe1024087:	jmp    0x7fffe1024091
   0x7fffe102408c:	call   0x7ffff65cf968 <_ZN18InterpreterRuntime9anewarrayEP10JavaThreadP12ConstantPoolii>
   0x7fffe1024091:	push   %rax
   0x7fffe1024092:	push   %rdi
   0x7fffe1024093:	push   %rsi
   0x7fffe1024094:	push   %rdx
--Type <RET> for more, q to quit, c to continue without paging--
   0x7fffe1024095:	push   %rcx
   0x7fffe1024096:	push   %r8
   0x7fffe1024098:	push   %r9
   0x7fffe102409a:	push   %r10
   0x7fffe102409c:	push   %r11
   0x7fffe102409e:	test   $0xf,%esp
   0x7fffe10240a4:	je     0x7fffe10240bc
   0x7fffe10240aa:	sub    $0x8,%rsp
   0x7fffe10240ae:	call   0x7ffff5d1c04e <_ZN6Thread7currentEv>
   0x7fffe10240b3:	add    $0x8,%rsp
   0x7fffe10240b7:	jmp    0x7fffe10240c1
   0x7fffe10240bc:	call   0x7ffff5d1c04e <_ZN6Thread7currentEv>
   0x7fffe10240c1:	pop    %r11
   0x7fffe10240c3:	pop    %r10
   0x7fffe10240c5:	pop    %r9
   0x7fffe10240c7:	pop    %r8
   0x7fffe10240c9:	pop    %rcx
   0x7fffe10240ca:	pop    %rdx
   0x7fffe10240cb:	pop    %rsi
   0x7fffe10240cc:	pop    %rdi
   0x7fffe10240cd:	cmp    %rax,%r15
   0x7fffe10240d0:	je     0x7fffe10240ea
   0x7fffe10240d6:	movabs $0x7ffff7311da0,%rdi
   0x7fffe10240e0:	and    $0xfffffffffffffff0,%rsp
   0x7fffe10240e4:	call   0x7ffff694f3c0 <_ZN14MacroAssembler7debug64EPclPl>
   0x7fffe10240e9:	hlt    
   0x7fffe10240ea:	pop    %rax
   0x7fffe10240eb:	movq   $0x0,0x2c0(%r15)
   0x7fffe10240f6:	movq   $0x0,0x2d0(%r15)
   0x7fffe1024101:	movq   $0x0,0x2c8(%r15)
   0x7fffe102410c:	vzeroupper 
--Type <RET> for more, q to quit, c to continue without paging--
   0x7fffe102410f:	cmpq   $0x0,0x8(%r15)
   0x7fffe1024117:	je     0x7fffe1024122
   0x7fffe102411d:	jmp    0x7fffe1000c20
   0x7fffe1024122:	mov    0x318(%r15),%rax
   0x7fffe1024129:	movq   $0x0,0x318(%r15)
   0x7fffe1024134:	mov    -0x40(%rbp),%r13
   0x7fffe1024138:	mov    -0x38(%rbp),%r14
   0x7fffe102413c:	ret    
   0x7fffe102413d:	movzbl 0x3(%r13),%ebx
   0x7fffe1024142:	add    $0x3,%r13
   0x7fffe1024146:	movabs $0x7ffff7bcc0a0,%r10
   0x7fffe1024150:	jmp    *(%r10,%rbx,8)
   0x7fffe1024154:	nop
   0x7fffe1024155:	nop

内存分配​

用的是jdk11以上版本，我这个jdk是用g1来做gc的，所以看看g1 是怎么分配的

(gdb) bt
#0  HeapRegion::par_allocate_impl (this=0x7ffff00e11b0, min_word_size=256, desired_word_size=63020, actual_size=0x7ffff59fcba8)
    at /home/dai/jdk/src/hotspot/share/gc/g1/heapRegion.inline.hpp:63
#1  0x00007ffff640bdcc in HeapRegion::par_allocate (this=0x7ffff00e11b0, min_word_size=256, desired_word_size=63020, actual_word_size=0x7ffff59fcba8)
    at /home/dai/jdk/src/hotspot/share/gc/g1/heapRegion.inline.hpp:225
#2  0x00007ffff640bfcb in G1AllocRegion::par_allocate (this=0x7ffff0052e10, alloc_region=0x7ffff00e11b0, min_word_size=256, desired_word_size=63020, 
    actual_word_size=0x7ffff59fcba8) at /home/dai/jdk/src/hotspot/share/gc/g1/g1AllocRegion.inline.hpp:63
#3  0x00007ffff640c0c6 in G1AllocRegion::attempt_allocation (this=0x7ffff0052e10, min_word_size=256, desired_word_size=63020, actual_word_size=0x7ffff59fcba8)
    at /home/dai/jdk/src/hotspot/share/gc/g1/g1AllocRegion.inline.hpp:77
#4  0x00007ffff6447142 in G1Allocator::attempt_allocation (this=0x7ffff0052d50, min_word_size=256, desired_word_size=63020, actual_word_size=0x7ffff59fcba8)
    at /home/dai/jdk/src/hotspot/share/gc/g1/g1Allocator.inline.hpp:62
#5  0x00007ffff6447b1d in G1CollectedHeap::attempt_allocation (this=0x7ffff0048bf0, min_word_size=256, desired_word_size=63020, 
    actual_word_size=0x7ffff59fcba8) at /home/dai/jdk/src/hotspot/share/gc/g1/g1CollectedHeap.cpp:709
#6  0x00007ffff64385ea in G1CollectedHeap::allocate_new_tlab (this=0x7ffff0048bf0, min_size=256, requested_size=63020, actual_size=0x7ffff59fcba8)
    at /home/dai/jdk/src/hotspot/share/gc/g1/g1CollectedHeap.cpp:359
#7  0x00007ffff69e1cf6 in MemAllocator::allocate_inside_tlab_slow (this=0x7ffff59fcc00, allocation=...)
    at /home/dai/jdk/src/hotspot/share/gc/shared/memAllocator.cpp:318
#8  0x00007ffff69e1bc2 in MemAllocator::allocate_inside_tlab (this=0x7ffff59fcc00, allocation=...)
    at /home/dai/jdk/src/hotspot/share/gc/shared/memAllocator.cpp:278
#9  0x00007ffff69e1eb9 in MemAllocator::mem_allocate (this=0x7ffff59fcc00, allocation=...) at /home/dai/jdk/src/hotspot/share/gc/shared/memAllocator.cpp:350
#10 0x00007ffff69e1f22 in MemAllocator::allocate (this=0x7ffff59fcc00) at /home/dai/jdk/src/hotspot/share/gc/shared/memAllocator.cpp:363
#11 0x00007ffff6260d84 in CollectedHeap::array_allocate (this=0x7ffff0048bf0, klass=0x8000407c0, size=106, length=825, do_zero=true, 
    __the_thread__=0x7ffff0028f70) at /home/dai/jdk/src/hotspot/share/gc/shared/collectedHeap.inline.hpp:41
#12 0x00007ffff6db9bf2 in TypeArrayKlass::allocate_common (this=0x8000407c0, length=825, do_zero=true, __the_thread__=0x7ffff0028f70)
    at /home/dai/jdk/src/hotspot/share/oops/typeArrayKlass.cpp:93
#13 0x00007ffff62f7428 in TypeArrayKlass::allocate (this=0x8000407c0, length=825, __the_thread__=0x7ffff0028f70)
    at /home/dai/jdk/src/hotspot/share/oops/typeArrayKlass.hpp:68
#14 0x00007ffff6ab4757 in oopFactory::new_typeArray (type=T_BYTE, length=825, __the_thread__=0x7ffff0028f70)
    at /home/dai/jdk/src/hotspot/share/memory/oopFactory.cpp:93
#15 0x00007ffff65cf8e5 in InterpreterRuntime::newarray (current=0x7ffff0028f70, type=T_BYTE, size=825)
    at /home/dai/jdk/src/hotspot/share/interpreter/interpreterRuntime.cpp:247
#16 0x00007fffe1023eb2 in ?? ()
#17 0x00007ffff7bca0a0 in TemplateInterpreter::_active_table () from /home/dai/jdk/build/linux-x86_64-server-slowdebug/jdk/lib/server/libjvm.so
#18 0x00007fffe1023e31 in ?? ()
#19 0x000000062a47ab38 in ?? ()
#20 0x00007ffff59fcd88 in ?? ()
#21 0x00007fffb43a23e6 in ?? ()
#22 0x00007ffff59fcde8 in ?? ()
#23 0x00007fffb43a3520 in ?? ()
#24 0x0000000000000000 in ?? ()

array_copy 反查​

java里面 System.arraycopy 函数就是copy 整个数组到新的数组里面,复制方式是浅拷贝. 最后调用的入口是下面的c++代码.所以可以通过这个函数就可以了解数组的内存布局

void ObjArrayKlass::copy_array(arrayOop s, int src_pos, arrayOop d,
                               int dst_pos, int length, TRAPS) {


  ...
    size_t src_offset = (size_t) objArrayOopDesc::obj_at_offset<oop>(src_pos); <----------------------------- 开始地址
    size_t dst_offset = (size_t) objArrayOopDesc::obj_at_offset<oop>(dst_pos);  <---------------------------- 结束地址
    assert(arrayOopDesc::obj_offset_to_raw<oop>(s, src_offset, NULL) ==
           objArrayOop(s)->obj_at_addr<oop>(src_pos), "sanity");
    assert(arrayOopDesc::obj_offset_to_raw<oop>(d, dst_offset, NULL) ==
           objArrayOop(d)->obj_at_addr<oop>(dst_pos), "sanity");
    do_copy(s, src_offset, d, dst_offset, length, CHECK);
  }
}

所以java的对象数组的内存布局就像下面一样 ， 这里的oop 是一个指针

64位下面是8字节

相关阅读​