Valgrind是一个GPL的软件，用于Linux（For x86, amd64 and ppc32）程序的内存调试和代码剖析。你可以在它的环境中运行你的程序来监视内存的使用情况，比如C 语言中的malloc和free或者 C++中的new和 delete。使用Valgrind的工具包，你可以自动的检测许多内存管理和线程的bug，避免花费太多的时间在bug寻找上，使得你的程序更加稳固。
Valgrind的主要功能
Valgrind工具包包含多个工具，如Memcheck,Cachegrind,Helgrind, Callgrind，Massif。下面分别介绍个工具的作用：
Memcheck 工具主要检查下面的程序错误：
1.使用未初始化的内存 (Use of uninitialised memory)
2.使用已经释放了的内存 (Reading/writing memory after it has been free’d)
3.使用超过 malloc分配的内存空间(Reading/writing off the end of malloc’d blocks)
4.对堆栈的非法访问 (Reading/writing inappropriate areas on the stack)
5.申请的空间是否有释放 (Memory leaks – where pointers to malloc’d blocks are lost forever)
6.malloc/free/new/delete申请和释放内存的匹配(Mismatched use of malloc/new/new [] vs free/delete/delete [])
7.src和dst的重叠(Overlapping src and dst pointers in memcpy() and related functions)
Callgrind
Callgrind收集程序运行时的一些数据，函数调用关系等信息，还可以有选择地进行cache 模拟。在运行结束时，它会把分析数据写入一个文件。callgrind_annotate可以把这个文件的内容转化成可读的形式。
Cachegrind
它模拟 CPU中的一级缓存I1,D1和L2二级缓存，能够精确地指出程序中 cache的丢失和命中。如果需要，它还能够为我们提供cache丢失次数，内存引用次数，以及每行代码，每个函数，每个模块，整个程序产生的指令数。这对优化程序有很大的帮助。
Helgrind
它主要用来检查多线程程序中出现的竞争问题。Helgrind 寻找内存中被多个线程访问，而又没有一贯加锁的区域，这些区域往往是线程之间失去同步的地方，而且会导致难以发掘的错误。Helgrind实现了名为” Eraser” 的竞争检测算法，并做了进一步改进，减少了报告错误的次数。
Massif
堆栈分析器，它能测量程序在堆栈中使用了多少内存，告诉我们堆块，堆管理块和栈的大小。Massif能帮助我们减少内存的使用，在带有虚拟内存的现代系统中，它还能够加速我们程序的运行，减少程序停留在交换区中的几率。
Ubuntu 安装valgrind：
#> sudo apt-get install valgrind
valgrind 使用：
#include <stdio.h> #include <stdlib.h> #include <string.h> int main() { char *ptr = (char *)malloc(1024); ptr[1024] = 0; // memcpy(ptr+1, ptr, 100); char ch = ptr[1024]; free(ptr); free(ptr); char *ptr1; *ptr1 = ‘a‘; char *ptr2 = (char *)malloc(1024); return 0; }
#> valgrind --tool=memcheck --leak-check=yes ./valgTest
注意在编译程序的时候加上-g选项，打印错误信息的时候会给出行号。上面的执行结果给出了六个错误。
下面看一个多线程竞争的情况：
#include <pthread.h> #include <stdio.h> int a = 0; void* child_fn (void* arg) { a++; return NULL; } int main () { pthread_t child; pthread_create(&child,NULL, child_fn, NULL); a++; pthread_join(child,NULL); return 0; }
zhuyp@ubuntu:~/valgrind$ valgrind --tool=helgrind ./helTest
==5639== Helgrind, a thread error detector
==5639== Copyright (C) 2007-2013, and GNU GPL‘d, by OpenWorks LLP et al.
==5639== Using Valgrind-3.10.0.SVN and LibVEX; rerun with -h for copyright info
==5639== Command: ./helTest
==5639==
==5639== ---Thread-Announcement------------------------------------------
==5639==
==5639== Thread #1 is the program‘s root thread
==5639==
==5639== ---Thread-Announcement------------------------------------------
==5639==
==5639== Thread #2 was created
==5639== at 0x51562CE: clone (clone.S:74)
==5639== by 0x4E44199: do_clone.constprop.3 (createthread.c:75)
==5639== by 0x4E458BA: pthread_create@@GLIBC_2.2.5 (createthread.c:245)
==5639== by 0x4C30C90: ??? (in /usr/lib/valgrind/vgpreload_helgrind-amd64-linux.so)
==5639== by 0x40068D: main (helTest.c:15)
==5639==
==5639== ----------------------------------------------------------------
==5639==
==5639== Possible data race during read of size 4 at 0x60104C by thread #1
==5639== Locks held: none
==5639== at 0x40068E: main (helTest.c:16)
==5639==
==5639== This conflicts with a previous write of size 4 by thread #2
==5639== Locks held: none
==5639== at 0x40065E: child_fn (helTest.c:8)
==5639== by 0x4C30E26: ??? (in /usr/lib/valgrind/vgpreload_helgrind-amd64-linux.so)
==5639== by 0x4E45181: start_thread (pthread_create.c:312)
==5639== by 0x515630C: clone (clone.S:111)
==5639==
==5639== ----------------------------------------------------------------
==5639==
==5639== Possible data race during write of size 4 at 0x60104C by thread #1
==5639== Locks held: none
==5639== at 0x400697: main (helTest.c:16)
==5639==
==5639== This conflicts with a previous write of size 4 by thread #2
==5639== Locks held: none
==5639== at 0x40065E: child_fn (helTest.c:8)
==5639== by 0x4C30E26: ??? (in /usr/lib/valgrind/vgpreload_helgrind-amd64-linux.so)
==5639== by 0x4E45181: start_thread (pthread_create.c:312)
==5639== by 0x515630C: clone (clone.S:111)
==5639==
==5639==
==5639== For counts of detected and suppressed errors, rerun with: -v
==5639== Use --history-level=approx or =none to gain increased speed, at
==5639== the cost of reduced accuracy of conflicting-access information
==5639== ERROR SUMMARY: 2 errors from 2 contexts (suppressed: 0 from 0)
前面写过一个使用coredump调试死锁的方法，这里使用valgrind 的 helgrind 工具也可以检查出死锁问题。
#include <iostream> #include <thread> #include <mutex> #include <chrono> using namespace std; mutex m1,m2; void func_2() { m2.lock(); cout<< "about to dead_lock"<<endl; m1.lock(); } void func_1() { m1.lock(); chrono::milliseconds dura( 1000 );// delay to trigger dead_lock this_thread::sleep_for( dura ); m2.lock(); } int main() { thread t1(func_1); thread t2(func_2); t1.join(); t2.join(); return 0; }
zhuyp@ubuntu:~/valgrind$ g++ -Wall dead_lock_demo.cpp -o dead_lock_demo -g -std=c++11 -lpthread
zhuyp@ubuntu:~/valgrind$ valgrind --tool=helgrind ./dead_lock_demo
==5646== Helgrind, a thread error detector
==5646== Copyright (C) 2007-2013, and GNU GPL‘d, by OpenWorks LLP et al.
==5646== Using Valgrind-3.10.0.SVN and LibVEX; rerun with -h for copyright info
==5646== Command: ./dead_lock_demo
==5646==
about to dead_lock // 执行到这里，Ctrl + C 结束进程，则打印出如下信息
^C==5646== ---Thread-Announcement------------------------------------------
==5646==
==5646== Thread #2 was created
==5646== at 0x56702CE: clone (clone.S:74)
==5646== by 0x4E44199: do_clone.constprop.3 (createthread.c:75)
==5646== by 0x4E458BA: pthread_create@@GLIBC_2.2.5 (createthread.c:245)
==5646== by 0x4C30C90: ??? (in /usr/lib/valgrind/vgpreload_helgrind-amd64-linux.so)
==5646== by 0x510CE3E: std::thread::_M_start_thread(std::shared_ptr<std::thread::_Impl_base>) (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.19)
==5646== by 0x401789: std::thread::thread<void (&)()>(void (&)()) (thread:135)
==5646== by 0x40112D: main (dead_lock_demo.cpp:33)
==5646==
==5646== ----------------------------------------------------------------
==5646==
==5646== Thread #2: Exiting thread still holds 1 lock
==5646== at 0x4E4BF2C: __lll_lock_wait (lowlevellock.S:135)
==5646== by 0x4E47656: _L_lock_909 (pthread_mutex_lock.c:151)
==5646== by 0x4E4747E: pthread_mutex_lock (pthread_mutex_lock.c:79)
==5646== by 0x4C32072: pthread_mutex_lock (in /usr/lib/valgrind/vgpreload_helgrind-amd64-linux.so)
==5646== by 0x40101B: __gthread_mutex_lock(pthread_mutex_t*) (gthr-default.h:748)
==5646== by 0x401421: std::mutex::lock() (mutex:134)
==5646== by 0x401111: func_1() (dead_lock_demo.cpp:25)
==5646== by 0x40278C: void std::_Bind_simple<void (*())()>::_M_invoke<>(std::_Index_tuple<>) (functional:1732)
==5646== by 0x4026E6: std::_Bind_simple<void (*())()>::operator()() (functional:1720)
==5646== by 0x40267F: std::thread::_Impl<std::_Bind_simple<void (*())()> >::_M_run() (thread:115)
==5646== by 0x510CBEF: ??? (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.19)
==5646== by 0x4C30E26: ??? (in /usr/lib/valgrind/vgpreload_helgrind-amd64-linux.so)
==5646==
==5646== ---Thread-Announcement------------------------------------------
==5646==
==5646== Thread #3 was created
==5646== at 0x56702CE: clone (clone.S:74)
==5646== by 0x4E44199: do_clone.constprop.3 (createthread.c:75)
==5646== by 0x4E458BA: pthread_create@@GLIBC_2.2.5 (createthread.c:245)
==5646== by 0x4C30C90: ??? (in /usr/lib/valgrind/vgpreload_helgrind-amd64-linux.so)
==5646== by 0x510CE3E: std::thread::_M_start_thread(std::shared_ptr<std::thread::_Impl_base>) (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.19)
==5646== by 0x401789: std::thread::thread<void (&)()>(void (&)()) (thread:135)
==5646== by 0x40113E: main (dead_lock_demo.cpp:35)
==5646==
==5646== ----------------------------------------------------------------
==5646==
==5646== Thread #3: Exiting thread still holds 1 lock
==5646== at 0x4E4BF2C: __lll_lock_wait (lowlevellock.S:135)
==5646== by 0x4E47656: _L_lock_909 (pthread_mutex_lock.c:151)
==5646== by 0x4E4747E: pthread_mutex_lock (pthread_mutex_lock.c:79)
==5646== by 0x4C32072: pthread_mutex_lock (in /usr/lib/valgrind/vgpreload_helgrind-amd64-linux.so)
==5646== by 0x40101B: __gthread_mutex_lock(pthread_mutex_t*) (gthr-default.h:748)
==5646== by 0x401421: std::mutex::lock() (mutex:134)
==5646== by 0x4010DF: func_2() (dead_lock_demo.cpp:14)
==5646== by 0x40278C: void std::_Bind_simple<void (*())()>::_M_invoke<>(std::_Index_tuple<>) (functional:1732)
==5646== by 0x4026E6: std::_Bind_simple<void (*())()>::operator()() (functional:1720)
==5646== by 0x40267F: std::thread::_Impl<std::_Bind_simple<void (*())()> >::_M_run() (thread:115)
==5646== by 0x510CBEF: ??? (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.19)
==5646== by 0x4C30E26: ??? (in /usr/lib/valgrind/vgpreload_helgrind-amd64-linux.so)
==5646==
==5646==
==5646== For counts of detected and suppressed errors, rerun with: -v
==5646== Use --history-level=approx or =none to gain increased speed, at
==5646== the cost of reduced accuracy of conflicting-access information
==5646== ERROR SUMMARY: 2 errors from 2 contexts (suppressed: 1 from 1)
吐槽一下，刚刚在Linux操作并编辑该博文，各种不适应啊，还是换到Windows下来编辑要方便些。。。
ubuntu 上使用valgrind,布布扣,bubuko.com
原文：http://www.cnblogs.com/zhuyp1015/p/3901191.html