前段时间看到说Lua 5.4用了一种新的通用随机数算法，替换掉本来内部使用的CRT的随机数引擎。我看了一下大致的实现，CPU和空间复杂度任然保持了一个较低的水平，并且循环节和说是随机性都还不错。我们游戏项目中原本对大量随机数场景的随机数算法使用的是基于线性同余的TAUS88，但是使用过程中发现这个算法分布上还是有一些不是很理想，所以就想把这个新的科研成果也用进我们项目中试试看效果。
新写的 xoshiro / xoroshiro 随机数引擎的特点如下:
- CPU和内存消耗大幅低于梅森旋转（MT19937，C++11 STL自带算法）
- 循环节和随机性优于TAUS88（基于线性同余）
- Lua 5.4内置随机数算法，种子初始化流程同Lua 5.4
- Website: http://xoshiro.di.unimi.it/
- Paper: http://vigna.di.unimi.it/ftp/papers/ScrambledLinear.pdf
- 实现的算法引擎支持 xoshiro256**、xoshiro256+、xoroshiro128**、xoroshiro128+。xoroshiro64、xoroshiro512和xoroshiro1024的模式不太一样没有接入，实际上前面四个就足够几乎所有的应用场景了。
以下是简化版代码，实际上分三个文件:
- 上半部是xoshiro / xoroshiro算法实现
- 中部是随机数功能封装，用于提供多引擎下统一的随机数接口。剥离里项目中的梅森旋转和TAUS88引擎。
- 下半部是简化的单元测试，单元测试数据是采集自原作者的纯C代码的运行结果。
完整的代码和单元测试见 atframework 的utils项目: http://github.com/atframework/atframe_utils
/**
* @file random_xor_combine_core.h
* @brief 伪随机数生成器 - xoshiro算法核心
* @see http://xoshiro.di.unimi.it
* @note 2018年的新型的全功能型伪随机数算法，也是lua 5.4内建的伪随机数算法
*
* Licensed under the MIT licenses.
* @version 1.0
* @author OWenT
* @date 2018年09月30日
*
* @history
*
*/
#ifndef UTIL_RANDOM_XOSHIRO_CORE_H
#define UTIL_RANDOM_XOSHIRO_CORE_H
// #pragma once
#include <stdint.h>
#include <cstddef>
namespace util {
namespace random {
namespace core {
/**
* @breif just for
* xoshiro256**
* xoshiro256+
* xoroshiro128**
* xoroshiro128+
* @note not support for xoroshiro64** 、xoroshiro64*、xoroshiro512** 、xoroshiro512*、xoroshiro1024** 、xoroshiro1024*
*/
template <typename UIntType, bool is_plus, int iidx, int n1, int n2>
class xoshinro_engine {
public:
typedef UIntType result_type;
typedef result_type seed_type[4];
private:
/// \endcond
seed_type xoshinro_seed_;
static inline result_type rotl(const result_type x, int k) {
return (x << k) | (x >> ((sizeof(result_type) * 8) - static_cast<result_type>(k)));
}
template <typename, bool>
struct next_init;
template <typename T>
struct next_init<T, true> {
static inline result_type call(seed_type &s) { return s[0] + s[3]; }
};
template <typename T>
struct next_init<T, false> {
static inline result_type call(seed_type &s) { return rotl(s[iidx] * 5, 7) * 9; }
};
protected:
result_type next() {
const result_type ret = next_init<UIntType, is_plus>::call(xoshinro_seed_);
const result_type t = xoshinro_seed_[1] << n1;
xoshinro_seed_[2] ^= xoshinro_seed_[0];
xoshinro_seed_[3] ^= xoshinro_seed_[1];
xoshinro_seed_[1] ^= xoshinro_seed_[2];
xoshinro_seed_[0] ^= xoshinro_seed_[3];
xoshinro_seed_[2] ^= t;
xoshinro_seed_[3] = rotl(xoshinro_seed_[3], n2);
return ret;
}
void jump(const seed_type &JUMP) {
result_type s0 = 0;
result_type s1 = 0;
result_type s2 = 0;
result_type s3 = 0;
for (size_t i = 0; i < sizeof(JUMP) / sizeof(JUMP[0]); i++) {
for (size_t b = 0; b < sizeof(result_type) * 8; b++) {
if (JUMP[i] & result_type(1) << b) {
s0 ^= xoshinro_seed_[0];
s1 ^= xoshinro_seed_[1];
s2 ^= xoshinro_seed_[2];
s3 ^= xoshinro_seed_[3];
}
next();
}
}
xoshinro_seed_[0] = s0;
xoshinro_seed_[1] = s1;
xoshinro_seed_[2] = s2;
xoshinro_seed_[3] = s3;
}
public:
xoshinro_engine() {
xoshinro_seed_[0] = 0;
xoshinro_seed_[1] = 0;
xoshinro_seed_[2] = 0;
xoshinro_seed_[3] = 0;
}
xoshinro_engine(result_type s) {
xoshinro_seed_[0] = 0;
xoshinro_seed_[1] = 0;
xoshinro_seed_[2] = 0;
xoshinro_seed_[3] = 0;
init_seed(s);
}
void init_seed(result_type s) {
xoshinro_seed_[0] = s;
xoshinro_seed_[1] = 0xff;
xoshinro_seed_[2] = 0;
xoshinro_seed_[3] = 0;
// just like in lua 5.4
for (int i = 0; i < 16; ++i) {
next();
}
}
template <typename It>
void init_seed(It &first, It last) {
It begin = first;
for (int i = 0; i < 4; ++i) {
if (begin != last) {
xoshinro_seed_[i] = *begin;
++begin;
} else {
xoshinro_seed_[i] = 0;
}
}
// just like in lua 5.4
for (int i = 0; i < 16; ++i) {
next();
}
}
result_type random() { return next(); }
result_type operator()() { return random(); }
inline const seed_type &get_seed() const { return xoshinro_seed_; }
};
template <bool is_plus>
class xoshinro_engine_128 : public xoshinro_engine<uint32_t, is_plus, 0, 9, 11> {
public:
typedef xoshinro_engine<uint32_t, is_plus, 0, 9, 11> base_type;
typedef typename base_type::result_type result_type;
typedef typename base_type::seed_type seed_type;
public:
xoshinro_engine_128() {}
xoshinro_engine_128(result_type s) : base_type(s) {}
using base_type::jump;
/**
* @brief just like call next() for 2^64 times
*/
void jump() {
static const result_type jump_params[4] = {0x8764000b, 0xf542d2d3, 0x6fa035c3, 0x77f2db5b};
jump(jump_params);
}
};
template <bool is_plus>
class xoshinro_engine_256 : public xoshinro_engine<uint64_t, is_plus, 1, 17, 45> {
public:
typedef xoshinro_engine<uint64_t, is_plus, 1, 17, 45> base_type;
typedef typename base_type::result_type result_type;
typedef typename base_type::seed_type seed_type;
public:
xoshinro_engine_256() {}
xoshinro_engine_256(result_type s) : base_type(s) {}
using base_type::jump;
/**
* @brief just like call next() for 2^128 times
*/
void jump() {
static const result_type jump_params[4] = {0x180ec6d33cfd0aba, 0xd5a61266f0c9392c, 0xa9582618e03fc9aa,
0x39abdc4529b1661c};
jump(jump_params);
}
/**
* @brief just like call next() for 2^192 times
*/
void long_jump() {
static const result_type jump_params[4] = {0x76e15d3efefdcbbf, 0xc5004e441c522fb3, 0x77710069854ee241,
0x39109bb02acbe635};
jump(jump_params);
}
};
} // namespace core
} // namespace random
} // namespace util
#endif /* UTIL_RANDOM_XOSHIRO_CORE_H */
// ============================ 随机数引擎封装 ================================
/**
* @file random_generator.h
* @brief 伪随机数生成器
*
* Licensed under the MIT licenses.
* @version 1.0
* @author OWenT
* @date 2018年09月30日
*
* @history
*
*/
#ifndef UTIL_RANDOM_GENERATOR_H
#define UTIL_RANDOM_GENERATOR_H
// #pragma once
#include <limits>
#include <numeric>
#include <ctime>
// #include "random_mt_core.h"
// #include "random_xor_combine_core.h"
// #include "random_xoshiro_core.h"
namespace util {
namespace random {
/**
* 随机数包装类，用以提供高级功能
*/
template <typename CoreType>
class random_manager_wrapper {
public:
typedef CoreType core_type;
typedef typename core_type::result_type result_type;
private:
core_type core_;
public:
random_manager_wrapper() {}
random_manager_wrapper(result_type rd_seed) : core_(rd_seed) {}
inline core_type & get_core() { return core_; }
inline const core_type &get_core() const { return core_; }
/**
* 初始化随机数种子
* @param [in] rd_seed 随机数种子
*/
void init_seed(result_type rd_seed) { core_.init_seed(rd_seed); }
/**
* 初始化随机数种子
* @note 取值范围为 [first, last)
* @param [in] first 随机数种子散列值起始位置
* @param [in] last 随机数种子散列值结束位置
*/
template <typename It>
void init_seed(It &first, It last) {
core_.init_seed(first, last);
}
/**
* 产生一个随机数
* @return 产生的随机数
*/
result_type random() { return core_(); }
/**
* 产生一个随机数
* @return 产生的随机数
*/
result_type operator()() { return random(); }
/**
* 产生一个随机数
* @param [in] lowest 下限
* @param [in] highest 上限(必须大于lowest)
* @note 取值范围 [lowest, highest)
* @return 产生的随机数
*/
template <typename ResaultType>
ResaultType random_between(ResaultType lowest, ResaultType highest) {
if (highest <= lowest) {
return lowest;
}
result_type res = (*this)();
return static_cast<ResaultType>(res % static_cast<result_type>(highest - lowest)) + lowest;
}
};
// ============== 随机数生成器 - xoshiro 算法(比梅森旋转算法消耗更少的内存，但是循环节更小，随机性比taus好) ==============
// @see http://xoshiro.di.unimi.it
// 循环节： 2^128 − 1
typedef random_manager_wrapper<core::xoshinro_engine_128<false> > xoroshiro128_starstar;
// 循环节： 2^128 − 1，少一次旋转，更快一点点
typedef random_manager_wrapper<core::xoshinro_engine_128<true> > xoroshiro128_plus;
// 循环节： 2^256 − 1
typedef random_manager_wrapper<core::xoshinro_engine_256<false> > xoshiro256_starstar;
// 循环节： 2^256 − 1，少一次旋转，更快一点点
typedef random_manager_wrapper<core::xoshinro_engine_256<true> > xoshiro256_plus;
} // namespace random
} // namespace util
#endif /* _UTIL_RANDOM_GENERATOR_H_ */
// ================================= 简易单元测试 =================================
#include <cstring>
#include <iostream>
#include <list>
#ifdef max
#undef max
#endif
// #include "random/random_generator.h"
// ===================== 这里是单元测试组件的简化版 =====================
#define test_case_func_name(test_name, case_name) test_func_test_##test_name##_case_##case_name
#define test_case_obj_name(test_name, case_name) test_obj_test_##test_name##_case_##case_name##_obj
#define CASE_TEST(test_name, case_name) \
static void test_case_func_name(test_name, case_name) (); \
static test_case_base test_case_obj_name(test_name, case_name) (#test_name, #case_name, test_case_func_name(test_name, case_name)); \
void test_case_func_name(test_name, case_name) ()
struct test_case_base;
static std::list<test_case_base*> g_all_tests;
static test_case_base* g_current_test;
static size_t g_success_count = 0;
static size_t g_failed_count = 0;
struct test_case_base {
typedef void (*fn_t)();
test_case_base(const char* test_name, const char* case_name, fn_t fn): test_name_(test_name), case_name_(case_name), fn_(fn) {
g_all_tests.push_back(this);
}
const char* test_name_;
const char* case_name_;
fn_t fn_;
};
#define CASE_EXPECT_EQ(L, R) \
if ((L) == (R)) { \
++ g_success_count; \
} else {\
++ g_failed_count; \
std::cerr<< "Expect Failed: "<< #L<< " == " << #R<< " @"<< __FILE__<< ":"<< __LINE__<< std::endl; \
}
#define CASE_EXPECT_NE(L, R) \
if ((L) != (R)) { \
++ g_success_count; \
} else {\
++ g_failed_count; \
std::cerr<< "Expect Failed: "<< #L<< " != " << #R<< " @"<< __FILE__<< ":"<< __LINE__<< std::endl; \
}
#define CASE_EXPECT_TRUE(EXPR) \
if ((EXPR)) { \
++ g_success_count; \
} else {\
++ g_failed_count; \
std::cerr<< "Expect True Failed: "<< #EXPR<< " @"<< __FILE__<< ":"<< __LINE__<< std::endl; \
}
// ---------------------- 这里是单元测试组件的简化版 ----------------------
CASE_TEST(random_test, xoroshiro128_starstar) {
util::random::xoroshiro128_starstar gen1;
gen1.init_seed(0x12345678);
uint32_t checked1[] = {0x6baaab3a, 0x9385fdbb, 0x99ae85fe, 0x88e2358c, 0x6592b0ce,
0x20eadc84, 0x6a8d80b7, 0xa676abb7, 0x726ef19a, 0xd4e33183};
for (size_t i = 0; i < sizeof(checked1) / sizeof(checked1[0]); ++i) {
CASE_EXPECT_EQ(gen1.random(), checked1[i]);
}
gen1.get_core().jump();
uint32_t checked2[] = {0xb40ef1a3, 0xeaf5f926, 0xdaedd531, 0xd75a0342, 0xf5cd493d,
0x4081f259, 0x92a08a09, 0x8e7856ff, 0x3eecdb2, 0x34cd820c};
for (size_t i = 0; i < sizeof(checked2) / sizeof(checked2[0]); ++i) {
CASE_EXPECT_EQ(gen1.random(), checked2[i]);
}
uint32_t arr_seeds[] = {0x01, 0x02, 0x03, 0x04};
uint32_t *ptr_seeds = &arr_seeds[0];
gen1.init_seed(ptr_seeds, ptr_seeds + 4);
uint32_t checked3[] = {0x483919e4, 0xa90f4ce8, 0xe55d1499, 0x17c2185e, 0xdb11c3f7,
0x63557f98, 0xc7ad6874, 0x4a4a45ca, 0x74dec471, 0xf7e82f2a};
for (size_t i = 0; i < sizeof(checked3) / sizeof(checked3[0]); ++i) {
CASE_EXPECT_EQ(gen1.random(), checked3[i]);
}
gen1.get_core().jump();
uint32_t checked4[] = {0xf30187c0, 0xe533ff68, 0x628920df, 0xa998b9e7, 0xcac83081,
0xe23e175b, 0x8aef7f7a, 0x43e56ba2, 0x59dc108b, 0x35b5536c};
for (size_t i = 0; i < sizeof(checked4) / sizeof(checked4[0]); ++i) {
CASE_EXPECT_EQ(gen1.random(), checked4[i]);
}
}
CASE_TEST(random_test, xoroshiro128_plus) {
util::random::xoroshiro128_plus gen1;
gen1.init_seed(0x12345678);
uint32_t checked1[] = {0xf1c950fb, 0x3c9b4647, 0x212ab3fb, 0x9930ad20, 0xe891a595,
0x5f3d9875, 0x264dd138, 0xb23b09e7, 0xb351ccd6, 0xf6342d85};
for (size_t i = 0; i < sizeof(checked1) / sizeof(checked1[0]); ++i) {
CASE_EXPECT_EQ(gen1.random(), checked1[i]);
}
gen1.get_core().jump();
uint32_t checked2[] = {0x65981904, 0x2d406a8b, 0xf722cc27, 0xd93f4e0, 0xc8c324e,
0x47adc6fc, 0x2e331b89, 0x2de59177, 0x68298f16, 0x3c1f786};
for (size_t i = 0; i < sizeof(checked2) / sizeof(checked2[0]); ++i) {
CASE_EXPECT_EQ(gen1.random(), checked2[i]);
}
uint32_t arr_seeds[] = {0x01, 0x02, 0x03, 0x04};
uint32_t *ptr_seeds = &arr_seeds[0];
gen1.init_seed(ptr_seeds, ptr_seeds + 4);
uint32_t checked3[] = {0x6ba5dc24, 0x1f1d11fa, 0xfa4932ff, 0x86a802af, 0x784c234,
0xbabe3a02, 0x77ad7d11, 0x7ebe938, 0x44211069, 0x2a5ba899};
for (size_t i = 0; i < sizeof(checked3) / sizeof(checked3[0]); ++i) {
CASE_EXPECT_EQ(gen1.random(), checked3[i]);
}
gen1.get_core().jump();
uint32_t checked4[] = {0x4956e25b, 0x1637a896, 0xc303f2ef, 0x9f3f6ac7, 0x8048281d,
0xfd181cdb, 0xd50c6ccb, 0xa1e5dc16, 0xe2e6feed, 0xe8901128};
for (size_t i = 0; i < sizeof(checked4) / sizeof(checked4[0]); ++i) {
CASE_EXPECT_EQ(gen1.random(), checked4[i]);
}
}
CASE_TEST(random_test, xoshiro256_starstar) {
util::random::xoshiro256_starstar gen1;
gen1.init_seed(0x12345678);
uint64_t checked1[] = {0x83ea18ca6bf49f0c, 0xca24436623484287, 0xe7965c078396b9dc, 0xb9036864adb81725, 0x5c35ba435d92ab9b,
0x45202e3b4085a2cc, 0xa1e9410f7f528e4e, 0x48eb35365566f0b2, 0x90412e6e9ae131c3, 0xba4a1df61ca0b237};
for (size_t i = 0; i < sizeof(checked1) / sizeof(checked1[0]); ++i) {
CASE_EXPECT_EQ(gen1.random(), checked1[i]);
}
gen1.get_core().jump();
uint64_t checked2[] = {0x555e786541686a07, 0x6b5e064426c7dce4, 0x63a8537c3e2de537, 0x1b1eb7c6ef11c6c, 0x9d6eaac62ce6a4f4,
0x8b2cb439dc19fadd, 0xb749c288597ceac4, 0x92da83e146fdff1d, 0x110d38c051cededc, 0xadab1aacca67bd98};
for (size_t i = 0; i < sizeof(checked2) / sizeof(checked2[0]); ++i) {
CASE_EXPECT_EQ(gen1.random(), checked2[i]);
}
gen1.get_core().long_jump();
uint64_t checked3[] = {0x855835566240e421, 0xef37ac417209f23c, 0x27d0b120fbc1263b, 0x3acf06c9c8bebcfc, 0x8e4c39858e961d77,
0x54ef99ad7133b336, 0x507e39520e592971, 0xf17538afb5568bf2, 0x9b12bf7bc0ad1cb, 0xe30b952ddb0edc22};
for (size_t i = 0; i < sizeof(checked3) / sizeof(checked3[0]); ++i) {
CASE_EXPECT_EQ(gen1.random(), checked3[i]);
}
uint64_t arr_seeds[] = {0x01, 0x02, 0x03, 0x04};
uint64_t *ptr_seeds = &arr_seeds[0];
gen1.init_seed(ptr_seeds, ptr_seeds + 4);
uint64_t checked4[] = {0x167894d082017430, 0x38b95133d3ac9e80, 0x1f5b58670bd33b, 0xa72aa9c1f3c234dc, 0xa369742bfe107109,
0xa10057c6cbaaf5b7, 0xed9d205bdc0671c4, 0x36379ae97a7e84fc, 0x68a4d0807b200c28, 0x1178895e2d328ef3};
for (size_t i = 0; i < sizeof(checked4) / sizeof(checked4[0]); ++i) {
CASE_EXPECT_EQ(gen1.random(), checked4[i]);
}
gen1.get_core().jump();
uint64_t checked5[] = {0x6892346243ec2224, 0x721f3bb7498cd45b, 0x4706ddfc3ac5a535, 0x1833b360cae1f78f, 0x49c783132a986f0f,
0x228581842db7c171, 0x575b41c3f6bafa53, 0xd54e667c5bdc8331, 0xd16f385b2194a5ee, 0xd35320c4a01add19};
for (size_t i = 0; i < sizeof(checked5) / sizeof(checked5[0]); ++i) {
CASE_EXPECT_EQ(gen1.random(), checked5[i]);
}
gen1.get_core().long_jump();
uint64_t checked6[] = {0xfc07bbe2a0ff49e3, 0x25b74d3c3e1395a4, 0x66c3b4e434a41253, 0xeef93c334db407df, 0xcbe33255433c267a,
0x1aeb5a580f8b97f7, 0xee0b16ebb05cc830, 0x1951fff956477d9e, 0xd586fc5de6068234, 0xb77c43707de92854};
for (size_t i = 0; i < sizeof(checked6) / sizeof(checked6[0]); ++i) {
CASE_EXPECT_EQ(gen1.random(), checked6[i]);
}
}
CASE_TEST(random_test, xoshiro256_plus) {
util::random::xoshiro256_plus gen1;
gen1.init_seed(0x12345678);
uint64_t checked1[] = {0xa6766515ded04cf7, 0x8b8ef459334146d2, 0xa5e132765ffaea1b, 0x42a32c7899960f6, 0xb55425c32b032ae,
0x4f5b510e66ed35c3, 0x86364d228b55294e, 0x15fee24568b9f2e4, 0x79172607cb7639c8, 0x380558a4c62a5df3};
for (size_t i = 0; i < sizeof(checked1) / sizeof(checked1[0]); ++i) {
CASE_EXPECT_EQ(gen1.random(), checked1[i]);
}
gen1.get_core().jump();
uint64_t checked2[] = {0x65e286c2682c7ede, 0x6fd589cda019d6da, 0x74ca58c461d9116, 0xaf78ff6a0b528fa, 0x532d4519bc355dff,
0x4fd9664f92848fd0, 0xe871531860e04f51, 0x9f66375cd3511195, 0xedb2c3bd1fad482d, 0xabd051255c4a95a4};
for (size_t i = 0; i < sizeof(checked2) / sizeof(checked2[0]); ++i) {
CASE_EXPECT_EQ(gen1.random(), checked2[i]);
}
gen1.get_core().long_jump();
uint64_t checked3[] = {0x823dc8ffe9ec3044, 0x2282021c3619f998, 0x58b7e7b6e4bf12c2, 0xba93e7ae8ffd60be, 0x7c5604182a5e5c,
0xfe2b1a4e6cc03509, 0x3cc13d59418b58b6, 0xee043f7276571ff3, 0xb0cefcacf43f423d, 0x2c06bd7d4b2a7b61};
for (size_t i = 0; i < sizeof(checked3) / sizeof(checked3[0]); ++i) {
CASE_EXPECT_EQ(gen1.random(), checked3[i]);
}
uint64_t arr_seeds[] = {0x01, 0x02, 0x03, 0x04};
uint64_t *ptr_seeds = &arr_seeds[0];
gen1.init_seed(ptr_seeds, ptr_seeds + 4);
uint64_t checked4[] = {0xf7d9f594a5851eca, 0x1554ccd2906e61d5, 0xf1a8da801e3b1052, 0xbf31cf72a94cc91b, 0x1a8b65873ac842c9,
0xd15befe036b53136, 0x8f8ff0898e5cca09, 0xf287e57ae52130ec, 0x4b93d1235a7a9687, 0x4a1e76c36a3f44f9};
for (size_t i = 0; i < sizeof(checked4) / sizeof(checked4[0]); ++i) {
CASE_EXPECT_EQ(gen1.random(), checked4[i]);
}
gen1.get_core().jump();
uint64_t checked5[] = {0x5ab99b574ab79179, 0x9dd72e9427eecf2d, 0x847348252d253c3a, 0x45e4172e49119e02, 0xfef4edb4819f165e,
0xecd2f9db3f6911a3, 0xaa21cb67c76ca6cf, 0x26a7e2345c58b741, 0xaf507d5180f1070c, 0x5c635df444a4f1};
for (size_t i = 0; i < sizeof(checked5) / sizeof(checked5[0]); ++i) {
CASE_EXPECT_EQ(gen1.random(), checked5[i]);
}
gen1.get_core().long_jump();
uint64_t checked6[] = {0x7f463ee9f76cdf70, 0xc0a587ad5ee6ec85, 0xb42e076175295e3a, 0x96d379025346fa5b, 0xa4de52d9b1d5e2e5,
0xa8790f06ce0d31de, 0x5d0eb9a768b4700f, 0xa4fbadd2392670bd, 0xc7b31a2f3b7819ee, 0xcca74cb594188bcd};
for (size_t i = 0; i < sizeof(checked6) / sizeof(checked6[0]); ++i) {
CASE_EXPECT_EQ(gen1.random(), checked6[i]);
}
}
int main() {
size_t success_count = 0;
size_t failed_count = 0;
for(std::list<test_case_base*>::iterator iter = g_all_tests.begin(); iter != g_all_tests.end(); ++ iter) {
g_current_test = *iter;
size_t old_fail = g_failed_count;
size_t old_succ = g_success_count;
std::cout<< "Start Test: "<< g_current_test->test_name_<< "."<< g_current_test->case_name_<< std::endl;
g_current_test->fn_();
if (old_fail < g_failed_count) {
++ failed_count;
std::cerr<< "Test Failed: "<< g_current_test->test_name_<< "."<< g_current_test->case_name_<< " with "<<
(g_failed_count - old_fail)<< " failed test(s) and "<<
(g_success_count - old_succ)<< " passed test(s)"<< std::endl;
} else {
++ success_count;
std::cout<< "Test Passed: "<< g_current_test->test_name_<< "."<< g_current_test->case_name_<< " with "<<
(g_success_count - old_succ)<< " passed test(s)"<< std::endl;
}
}
std::cout<< "All test case(s) finished, success: "<< success_count<< ", failed: "<< failed_count<< std::endl;
return 0;
}