WET_Study/WETSort/WETSort.cpp

// WETSort.cpp : This file contains the 'main' function. Program execution begins and ends there.
//


#include "../arkiny/dclb/Sort.h"
#include "../ccm9026/JFSort/JFSort.h"
#include "../djarksnd/sort.h"
#include "../insun/Sort/Sort/SquickSort.h"
#define partition es_partition
#include "../prauser/ESSort.h"
#undef partition
#include "../witdrg/wc.h"

#include <random>
#include <chrono>
#include <ratio>
#include <algorithm>
#include <vector>
#include <functional>


std::string sha1(const void* p, size_t s);

const size_t testLength[] = {
#ifdef _DEBUG
	1ULL << 21,
	1ULL << 24,
	1ULL << 29,
#else
	1ULL << 31, // 2G (4G)
	1ULL << 32, // 4G (8G)
	1ULL << 33  // 8G (16G)
#endif
};

enum TestUnitId
{
	stl = -1,
	arkiny = 0,
	ccm9026,
	djarksnd,
	insun,
	prauser,
	witdrg,
	MAX_UID
};

const char* uidString[MAX_UID] = {
	"arkiny",
	"ccm9026",
	"djarksnd",
	"insun",
	"prauser",
	"witdrg"
};

struct TestScore
{
	bool error;
	double sum;
	int num;
};
TestScore testScore[MAX_UID] = { 0 };

struct TestResult
{
	double t;
	std::string hash;
};

enum GenerateDataTypes
{
	GenUniform32,
	GenUniform16,
	GenUniform16Xor,
	GenUniform8,
	GenUniform4,
	GenUniform1,
	GenUniform8Xor,
	GenNormal64,
	GenNormal32,
};

#pragma pack(push, 1)
template <size_t s>
struct Element
{
	int8_t buffer[s];
	bool operator < (const Element& other) const
	{
		for (size_t i = 0; i < s; ++i)
		{
			int n = int(buffer[i]) - int(other.buffer[i]);
			if (n)
				return n > 0;
		}
		return false;
	}
};
#pragma pack(pop)

static std::vector<uint8_t> data0, data1;

template <typename ElementType>
TestResult TestUnit(int index)
{
	std::function<void(ElementType*, size_t)> fn;
	switch (index)
	{
	case stl:
		fn = [](ElementType* items, size_t numItems)
		{
			std::sort(&items[0], &items[numItems], std::less<ElementType>());
		};
		break;
	case arkiny:
		fn = [](ElementType* items, size_t numItems)
		{
			dc::sort(&items[0], &items[numItems], std::less<ElementType>());
		};
		break;
	case ccm9026:
		fn = [](ElementType* items, size_t numItems)
		{
			JFFoundation::JFAlgorithm::_Sort(&items[0], &items[numItems], std::less<ElementType>());
		};
		break;
	case djarksnd:
		fn = [](ElementType* items, size_t numItems)
		{
			sort::sort(&items[0], &items[numItems], std::less<ElementType>());
		};
		break;
	case insun:
		fn = [](ElementType* items, size_t numItems)
		{
			SunSort::SSort(items, &items[0], &items[numItems - 1], std::less<ElementType>());
		};
		break;
	case prauser:
		fn = [](ElementType* items, size_t numItems)
		{
			// ERROR
			//std::less<ElementType> tmp;
			//ES::quickSort<ElementType*, std::less<ElementType>>(&items[0], &items[numItems], tmp);
		};
		break;
	case witdrg:
		fn = [](ElementType* items, size_t numItems)
		{
			// ERROR
			//wc::sort(&items[0], &items[numItems], std::less<ElementType>());
		};
		break;
	}

	if (fn)
	{
		std::chrono::high_resolution_clock clock;
		using sec = std::chrono::duration<double, std::ratio<1, 1>>;

		data1.resize(data0.size());
		memcpy(data1.data(), data0.data(), data0.size());

		size_t s = sizeof(ElementType);
		size_t numItems = data1.size() / s;

		auto start = clock.now();
		fn(reinterpret_cast<ElementType*>(data1.data()), numItems);
		auto end = clock.now();

		TestResult res;
		res.t = std::chrono::duration_cast<sec>(end - start).count();
		res.hash = sha1(data1.data(), data1.size()).c_str();
		return res;
	}
	throw std::runtime_error("unexpected error!");
}

template <size_t unitSize>
void DoSortTest(void)
{
	size_t numItems = data0.size() / unitSize;
	TestResult base = TestUnit<Element<unitSize>>(stl);
	printf("[%s] items: %zu / %zu, elapsed: %f, hash: %s\n", "STD", unitSize, numItems, base.t, base.hash.c_str());
	for (int i = 0; i < MAX_UID; ++i)
	{
		TestScore& score = testScore[i];
		if (!score.error)
		{
			//printf("[%s] sorting...\n", uidString[i]);
			TestResult r = TestUnit<Element<unitSize>>(i);
			if (base.hash == r.hash)
			{
				double s = base.t / r.t;
				score.sum += s;
				score.num++;
				printf("[%s] items: %zu, elapsed: %f, score:%d\n", uidString[i], numItems, r.t, int(s * 100));
			}
			else
			{
				printf("[%s] error!! [REJECTED]\n", uidString[i]);
				score.error = true;
			}
		}
	}
}

void RunTest(size_t size, GenerateDataTypes type)
{
	std::random_device rdev;
	std::default_random_engine re(rdev());

	if (1)
	{
		size_t s = 64;
		while (s < size)
			s <<= 1;
		size = s;
	}

	// generate random number.
	data0.clear();
	data0.resize(size);

	switch (type)
	{
	case GenUniform32:
		if (1)
		{
			printf("Generating random data... (%zu bytes, type: Uniform32)\n", size);
			std::uniform_int_distribution<uint32_t> dist(0, 0xffffffff);
			for (size_t i = 0, e = size / sizeof(uint32_t); i < e; ++i)
			{
				reinterpret_cast<uint32_t*>(data0.data())[i] = dist(re);
			}
		}
		break;
	case GenUniform16: 
		if (1)
		{
			printf("Generating random data... (%zu bytes, type: Uniform16)\n", size);
			std::uniform_int_distribution<uint16_t> dist(0, 0xffff);
			for (size_t i = 0, e = size / sizeof(uint32_t); i < e; ++i)
			{
				reinterpret_cast<uint32_t*>(data0.data())[i] = dist(re);
			}
		}
		break;
	case GenUniform16Xor: 
		if (1)
		{
			printf("Generating random data... (%zu bytes, type: Uniform16Xor)\n", size);
			std::uniform_int_distribution<uint16_t> dist(0, 0xffff);
			uint32_t tmp = dist(re);
			for (size_t i = 0, e = size / sizeof(uint32_t); i < e; ++i)
			{
				uint32_t x = dist(re);
				uint32_t y = (i ^ (tmp << 3)) & 0xff;
				tmp = y;
				reinterpret_cast<uint32_t*>(data0.data())[i] = (x << 16) | y;
			}
		}
		break;
	case GenUniform8:
		if (1)
		{
			printf("Generating random data... (%zu bytes, type: Uniform8)\n", size);
			std::uniform_int_distribution<uint16_t> dist(0, 0xff);
			for (size_t i = 0, e = size / sizeof(uint32_t); i < e; ++i)
			{
				reinterpret_cast<uint32_t*>(data0.data())[i] = dist(re);
			}
		}
		break;
	case GenUniform8Xor:
		if (1)
		{
			printf("Generating random data... (%zu bytes, type: Uniform8Xor)\n", size);
			std::uniform_int_distribution<uint16_t> dist(0, 0xff);
			uint32_t tmp = dist(re);
			for (size_t i = 0, e = size / sizeof(uint32_t); i < e; ++i)
			{
				uint32_t val = dist(re);
				val = val ^ (tmp << 1);
				tmp = val;
				reinterpret_cast<uint32_t*>(data0.data())[i] = val;
			}
		}
		break;
	case GenUniform4:
		if (1)
		{
			printf("Generating random data... (%zu bytes, type: Uniform8)\n", size);
			std::uniform_int_distribution<uint16_t> dist(0, 0xff);
			for (size_t i = 0, e = size / sizeof(uint32_t); i < e; ++i)
			{
				reinterpret_cast<uint32_t*>(data0.data())[i] = dist(re) & 0x7;
			}
		}
		break;
	case GenUniform1:
		if (1)
		{
			printf("Generating random data... (%zu bytes, type: Uniform8)\n", size);
			std::uniform_int_distribution<uint16_t> dist(0, 0xff);
			for (size_t i = 0, e = size / sizeof(uint32_t); i < e; ++i)
			{
				reinterpret_cast<uint32_t*>(data0.data())[i] = dist(re) & 0x1;
			}
		}
		break;
	case GenNormal64:
		if (1)
		{
			printf("Generating random data... (%zu bytes, type: Normal64)\n", size);
			std::normal_distribution<double> dist;
			for (size_t i = 0, e = size / sizeof(double); i < e; ++i)
			{
				reinterpret_cast<double*>(data0.data())[i] = dist(re);
			}
		}
		break;
	case GenNormal32:
		if (1)
		{
			printf("Generating random data... (%zu bytes, type: Normal32)\n", size);
			std::normal_distribution<float> dist;
			for (size_t i = 0, e = size / sizeof(float); i < e; ++i)
			{
				reinterpret_cast<float*>(data0.data())[i] = dist(re);
			}
		}
		break;
	}

	// 1 byte
	DoSortTest<1>();
	// 2 byte
	DoSortTest<2>();
	// 3 bytes
	DoSortTest<3>();
	//4 bytes
	DoSortTest<4>();
	// 5 bytes
	DoSortTest<5>();
	// 6
	DoSortTest<6>();
	// 7
	DoSortTest<7>();
	// 9 bytes
	DoSortTest<9>();
	// 10
	DoSortTest<10>();
	// 11
	DoSortTest<11>();
	// 12
	DoSortTest<12>();
	// 13
	DoSortTest<13>();
	// 15
	DoSortTest<15>();
	// 17 bytes
	DoSortTest<17>();
	//// 32 bytes
	//DoSortTest<32>();
	//// 33 bytes
	//DoSortTest<33>();
}

int main()
{
	// Set .error to '0' to run the test!
	testScore[arkiny].error   = 1;
	testScore[ccm9026].error  = 1;
	testScore[djarksnd].error = 1;
	testScore[insun].error    = 1;
	testScore[prauser].error  = 1;
	testScore[witdrg].error   = 1;

	auto dataTypes = {
				GenUniform32,
				GenUniform16,
				GenUniform16Xor,
				GenUniform8,
				GenUniform8Xor,
				GenUniform4,
				GenUniform1,
				GenNormal64,
				GenNormal32
	};

	auto PrintScore = [](int phase)
	{
		printf("------------ SCORE (%d/%d) --------------\n", phase, int(std::size(testLength)));
		for (int i = 0; i < MAX_UID; ++i)
		{
			const TestScore& score = testScore[i];
			if (score.error)
			{
				printf("UID(%s) rejected. (ERROR)\n", uidString[i]);
			}
			else
			{
				uint32_t s = (score.sum / double(score.num)) * 100.0;
				const char* t = "";
				if (s > 100)
					t = "(PERFECT)";
				else if (s > 90)
					t = "(GOOD)";
				else if (s < 60)
					t = "(POOR)";
				printf("UID(%s) score: %d. %s\n", uidString[i], s, t);
			}
		}
		printf("-----------------------------------------\n\n");
	};
	int index = 1;
	for (size_t n : testLength)
	{
		printf("========= Phase %d / %d (%zu) ==========\n", index, int(std::size(testLength)), n);
		for (auto t : dataTypes)
		{
			RunTest(n, t);
		}
		PrintScore(index);
		index++;
	}
}