| | |
| | |
| |
|
| | #include <bitset> |
| | #include <ctime> |
| | #include <thread> |
| | #include <vector> |
| | #include "inc/Test.h" |
| | #include "inc/Core/Common/DistanceUtils.h" |
| |
|
| | template<typename T> |
| | static float ComputeCosineDistance(const T *pX, const T *pY, SPTAG::DimensionType length) { |
| | float diff = 0; |
| | const T* pEnd1 = pX + length; |
| | while (pX < pEnd1) diff += (*pX++) * (*pY++); |
| | return diff; |
| | } |
| |
|
| | template<typename T> |
| | static float ComputeL2Distance(const T *pX, const T *pY, SPTAG::DimensionType length) |
| | { |
| | float diff = 0; |
| | const T* pEnd1 = pX + length; |
| | while (pX < pEnd1) { |
| | float c1 = ((float)(*pX++) - (float)(*pY++)); diff += c1 * c1; |
| | } |
| | return diff; |
| | } |
| |
|
| | template<typename T> |
| | T random(int high = RAND_MAX, int low = 0) |
| | { |
| | return (T)(low + float(high - low)*(std::rand()/static_cast<float>(RAND_MAX + 1.0))); |
| | } |
| |
|
| | template<typename T> |
| | void test(int high) { |
| | SPTAG::DimensionType dimension = random<SPTAG::DimensionType>(256, 2); |
| | T *X = new T[dimension], *Y = new T[dimension]; |
| | BOOST_ASSERT(X != nullptr && Y != nullptr); |
| | for (SPTAG::DimensionType i = 0; i < dimension; i++) { |
| | X[i] = random<T>(high, -high); |
| | Y[i] = random<T>(high, -high); |
| | } |
| | BOOST_CHECK_CLOSE_FRACTION(ComputeL2Distance(X, Y, dimension), SPTAG::COMMON::DistanceUtils::ComputeDistance(X, Y, dimension, SPTAG::DistCalcMethod::L2), 1e-5); |
| | BOOST_CHECK_CLOSE_FRACTION(high * high - ComputeCosineDistance(X, Y, dimension), SPTAG::COMMON::DistanceUtils::ComputeDistance(X, Y, dimension, SPTAG::DistCalcMethod::Cosine), 1e-5); |
| |
|
| | delete[] X; |
| | delete[] Y; |
| | } |
| |
|
| | template <typename T> |
| | void test_dist_calc_performance( |
| | int high, |
| | SPTAG::DimensionType dimension = 256, |
| | SPTAG::SizeType size = 100, |
| | SPTAG::DistCalcMethod calc_method = SPTAG::DistCalcMethod::L2) |
| | { |
| | T **X = new T *[size]; |
| | T **Y = new T *[size]; |
| | for (SPTAG::SizeType i = 0; i < size; i++) |
| | { |
| | X[i] = new T[dimension]; |
| | Y[i] = new T[dimension]; |
| | for (SPTAG::DimensionType j = 0; j < dimension; j++) |
| | { |
| | X[i][j] = random<T>(high, -high); |
| | Y[i][j] = random<T>(high, -high); |
| | } |
| | } |
| |
|
| | double start, end; |
| | start = omp_get_wtime(); |
| | #pragma omp parallel for |
| | for (SPTAG::SizeType i = 0; i < size; i++) |
| | { |
| | SPTAG::COMMON::DistanceUtils::ComputeDistance(X[i], Y[i], dimension, calc_method); |
| | } |
| | end = omp_get_wtime(); |
| | std::cout << "Time to calculate distance (ms): " << (end - start) * 1000 << std::endl; |
| |
|
| | delete[] X; |
| | delete[] Y; |
| | } |
| |
|
| | BOOST_AUTO_TEST_SUITE(DistanceTest) |
| |
|
| | BOOST_AUTO_TEST_CASE(TestDistanceComputation) |
| | { |
| | test<float>(1); |
| | test<std::int8_t>(127); |
| | test<std::int16_t>(32767); |
| | } |
| |
|
| | BOOST_AUTO_TEST_CASE(TestDistanceComputationPerformance) |
| | { |
| | std::vector<SPTAG::DimensionType> dimensions{128, 256, 512, 1024}; |
| | std::vector<int> nums_threads{1, 16, 40}; |
| | std::vector<SPTAG::DistCalcMethod> calc_methods{SPTAG::DistCalcMethod::L2, SPTAG::DistCalcMethod::Cosine}; |
| | SPTAG::SizeType size = 100000; |
| | std::cout << "Testing DistanceComputationPerformance..." << std::endl; |
| | for (int num_threads : nums_threads) |
| | { |
| | std::cout << "num_thread: " << num_threads << std::endl; |
| | omp_set_num_threads(num_threads); |
| | for (SPTAG::DistCalcMethod calc_method : calc_methods) |
| | { |
| | std::cout << "calc_method: " << (calc_method == SPTAG::DistCalcMethod::L2 ? "L2" : "Cosine") << std::endl; |
| | for (auto dimension : dimensions) |
| | { |
| | std::cout << "type: int8, dimension: " << dimension << ", size: " << size << std::endl; |
| | test_dist_calc_performance<std::int8_t>(127, dimension, size, calc_method); |
| | } |
| | } |
| | } |
| | } |
| |
|
| | BOOST_AUTO_TEST_SUITE_END() |
| |
|