diff --git a/cpp/Makefile b/cpp/Makefile
index 7cc66b2c..4a3677e9 100644
--- a/cpp/Makefile
+++ b/cpp/Makefile
@@ -1,11 +1,11 @@
 
-CXXFLAGS = -I.. -I. -Wall -g
-#CXXFLAGS = -I.. -I. -Wall -g -O4
-LDFLAGS = -L.
+#CXXFLAGS = -I.. -I. -Wall -g
+CXXFLAGS = -I.. -I. -Wall -g -O4
+LDFLAGS = -L. $(CXXFLAGS)
 
 NEED_PREPROCESS = zone.hpp
 
-all: test
+all: test bench
 
 %.hpp: %.hpp.erb
 	erb $< > $@
@@ -13,10 +13,13 @@ all: test
 test: $(NEED_PREPROCESS) unpack.o unpack_inline.o object.o zone.o test.o object.hpp unpack.hpp pack.hpp
 	$(CXX) $(LDFLAGS) unpack.o unpack_inline.o zone.o object.o test.o -o $@
 
+bench: $(NEED_PREPROCESS) unpack.o unpack_inline.o object.o zone.o bench.o object.hpp unpack.hpp pack.hpp
+	$(CXX) $(LDFLAGS) unpack.o unpack_inline.o zone.o object.o bench.o -o $@
+
 .PHONY: clean
 clean:
 	$(RM) unpack.o unpack_inline.o object.o zone.o
-	$(RM) test.o
-	$(RM) test
+	$(RM) test.o test
+	$(RM) bench.o bench
 	$(RM) $(NEED_PREPROCESS)
 
diff --git a/cpp/bench.cpp b/cpp/bench.cpp
new file mode 100644
index 00000000..1a9f6f92
--- /dev/null
+++ b/cpp/bench.cpp
@@ -0,0 +1,181 @@
+#include <msgpack/unpack.hpp>
+#include <msgpack/pack.hpp>
+#include <string.h>
+#include <sys/time.h>
+#include <iostream>
+#include <stdexcept>
+#include <string>
+
+static const unsigned int TASK_INT_NUM = 1<<24;
+static const unsigned int TASK_STR_LEN = 1<<15;
+//static const unsigned int TASK_INT_NUM = 1<<23;
+//static const unsigned int TASK_STR_LEN = 1<<14;
+static const void* TASK_STR_PTR;
+
+
+class simple_timer {
+public:
+	void reset() { gettimeofday(&m_timeval, NULL); }
+	void show_stat(size_t bufsz)
+	{
+		struct timeval endtime;
+		gettimeofday(&endtime, NULL);
+		double sec = (endtime.tv_sec - m_timeval.tv_sec)
+			+ (double)(endtime.tv_usec - m_timeval.tv_usec) / 1000 / 1000;
+		std::cout << sec << " sec" << std::endl;
+		std::cout << (double(bufsz)/1024/1024) << " MB" << std::endl;
+		std::cout << (bufsz/sec/1024/1024*8) << " Mbps" << std::endl;
+	}
+private:
+	timeval m_timeval;
+};
+
+
+class simple_buffer {
+public:
+	static const size_t DEFAULT_INITIAL_SIZE = 32*1024;//512*1024*1024*2;
+
+	simple_buffer(size_t initial_size = DEFAULT_INITIAL_SIZE) :
+		m_storage((char*)malloc(initial_size)),
+		m_allocated(initial_size),
+		m_used(0)
+	{
+		if(!m_storage) { throw std::bad_alloc(); }
+	}
+
+	~simple_buffer()
+	{
+		free(m_storage);
+	}
+
+public:
+	inline simple_buffer& append(const char* buf, size_t len)
+	{
+		if(m_allocated - m_used < len) {
+			expand_buffer(len);
+		}
+		memcpy(m_storage + m_used, buf, len);
+		m_used += len;
+		return *this;
+	}
+
+	void clear()
+	{
+		m_used = 0;
+	}
+
+private:
+	void expand_buffer(size_t req)
+	{
+		size_t nsize = m_allocated * 2;
+		size_t at_least = m_used + req;
+		while(nsize < at_least) { nsize *= 2; }
+		char* tmp = (char*)realloc(m_storage, nsize);
+		if(!tmp) { throw std::bad_alloc(); }
+		m_storage = tmp;
+		m_allocated = nsize;
+	}
+
+public:
+	size_t size() const { return m_used; }
+	const char* data() const { return m_storage; }
+
+private:
+	char* m_storage;
+	size_t m_allocated;
+	size_t m_used;
+};
+
+
+void bench_msgpack_int()
+{
+	simple_buffer buf;
+	simple_timer timer;
+
+	std::cout << "----" << std::endl;
+	std::cout << "pack integer" << std::endl;
+
+	timer.reset();
+	{
+		msgpack::packer<simple_buffer> pk(buf);
+		pk.pack_array(TASK_INT_NUM);
+		for(unsigned int i=0; i < TASK_INT_NUM; ++i) {
+			pk.pack_unsigned_int(i);
+		}
+	}
+	timer.show_stat(buf.size());
+
+	std::cout << "----" << std::endl;
+	std::cout << "unpack integer" << std::endl;
+
+	msgpack::zone z;
+	msgpack::object obj;
+
+	timer.reset();
+	{
+		obj = msgpack::unpack(buf.data(), buf.size(), z);
+	}
+	timer.show_stat(buf.size());
+
+	std::cout << "----" << std::endl;
+	std::cout << "dynamic pack integer" << std::endl;
+
+	buf.clear();
+
+	timer.reset();
+	msgpack::pack(buf, obj);
+	timer.show_stat(buf.size());
+}
+
+void bench_msgpack_str()
+{
+	simple_buffer buf;
+	simple_timer timer;
+
+	std::cout << "----" << std::endl;
+	std::cout << "pack string" << std::endl;
+
+	timer.reset();
+	{
+		msgpack::packer<simple_buffer> pk(buf);
+		pk.pack_array(TASK_STR_LEN);
+		for(unsigned int i=0; i < TASK_STR_LEN; ++i) {
+			pk.pack_raw(TASK_STR_PTR, i);
+		}
+	}
+	timer.show_stat(buf.size());
+
+	std::cout << "----" << std::endl;
+	std::cout << "unpack string" << std::endl;
+
+	msgpack::zone z;
+	msgpack::object obj;
+
+	timer.reset();
+	{
+		obj = msgpack::unpack(buf.data(), buf.size(), z);
+	}
+	timer.show_stat(buf.size());
+
+	std::cout << "----" << std::endl;
+	std::cout << "dynamic pack string" << std::endl;
+
+	buf.clear();
+
+	timer.reset();
+	msgpack::pack(buf, obj);
+	timer.show_stat(buf.size());
+}
+
+int main(void)
+{
+	void* str = malloc(TASK_STR_LEN);
+	memset(str, 'a', TASK_STR_LEN);
+	TASK_STR_PTR = str;
+
+	bench_msgpack_int();
+	bench_msgpack_str();
+
+	return 0;
+}
+
diff --git a/cpp/unpack.cpp b/cpp/unpack.cpp
index d4845938..807ded9f 100644
--- a/cpp/unpack.cpp
+++ b/cpp/unpack.cpp
@@ -85,8 +85,8 @@ void unpacker::expand_buffer(size_t len)
 		void* tmp = malloc(next_size);
 		if(!tmp) { throw std::bad_alloc(); }
 		memcpy(tmp, m_buffer, m_used);
-
 		free(m_buffer);
+
 		m_buffer = tmp;
 		m_free = next_size - m_used;