// Copyright 2019 Google LLC // // This source code is licensed under the BSD-style license found in the // LICENSE file in the root directory of this source tree. #pragma once #include #include #include #include #include #include #include #include #include #include #include class VUnOpMicrokernelTester { public: enum class OpType { Abs, ELU, LeakyReLU, Negate, ReLU, RoundToNearestEven, RoundTowardsZero, RoundUp, RoundDown, Square, SquareRoot, Sigmoid, }; enum class Variant { Native, Scalar, }; inline VUnOpMicrokernelTester& batch_size(size_t batch_size) { assert(batch_size != 0); this->batch_size_ = batch_size; return *this; } inline size_t batch_size() const { return this->batch_size_; } inline VUnOpMicrokernelTester& inplace(bool inplace) { this->inplace_ = inplace; return *this; } inline bool inplace() const { return this->inplace_; } inline VUnOpMicrokernelTester& slope(float slope) { this->slope_ = slope; return *this; } inline float slope() const { return this->slope_; } inline VUnOpMicrokernelTester& prescale(float prescale) { this->prescale_ = prescale; return *this; } inline float prescale() const { return this->prescale_; } inline VUnOpMicrokernelTester& alpha(float alpha) { this->alpha_ = alpha; return *this; } inline float alpha() const { return this->alpha_; } inline VUnOpMicrokernelTester& beta(float beta) { this->beta_ = beta; return *this; } inline float beta() const { return this->beta_; } inline VUnOpMicrokernelTester& qmin(uint8_t qmin) { this->qmin_ = qmin; return *this; } inline uint8_t qmin() const { return this->qmin_; } inline VUnOpMicrokernelTester& qmax(uint8_t qmax) { this->qmax_ = qmax; return *this; } inline uint8_t qmax() const { return this->qmax_; } inline VUnOpMicrokernelTester& iterations(size_t iterations) { this->iterations_ = iterations; return *this; } inline size_t iterations() const { return this->iterations_; } void Test(xnn_f32_vunary_ukernel_function vunary, OpType op_type, Variant variant = Variant::Native) const { std::random_device random_device; auto rng = std::mt19937(random_device()); auto distribution = std::uniform_real_distribution(-125.0f, 125.0f); switch (op_type) { case OpType::ELU: distribution = std::uniform_real_distribution(-20.0f, 20.0f); break; case OpType::SquareRoot: distribution = std::uniform_real_distribution(0.0f, 10.0f); break; default: break; } auto f32rng = std::bind(distribution, std::ref(rng)); std::vector x(batch_size() + XNN_EXTRA_BYTES / sizeof(float)); std::vector y(batch_size() + (inplace() ? XNN_EXTRA_BYTES / sizeof(float) : 0)); std::vector y_ref(batch_size()); for (size_t iteration = 0; iteration < iterations(); iteration++) { if (inplace()) { std::generate(y.begin(), y.end(), std::ref(f32rng)); } else { std::generate(x.begin(), x.end(), std::ref(f32rng)); std::fill(y.begin(), y.end(), nanf("")); } const float* x_data = inplace() ? y.data() : x.data(); // Compute reference results. for (size_t i = 0; i < batch_size(); i++) { switch (op_type) { case OpType::Abs: y_ref[i] = std::abs(x_data[i]); break; case OpType::ELU: { y_ref[i] = std::signbit(x_data[i]) ? alpha() * std::expm1(double(x_data[i]) * prescale()) : double(x_data[i]) * beta(); break; } case OpType::LeakyReLU: y_ref[i] = std::signbit(x_data[i]) ? x_data[i] * slope() : x_data[i]; break; case OpType::Negate: y_ref[i] = -x_data[i]; break; case OpType::ReLU: y_ref[i] = std::max(x_data[i], 0.0f); break; case OpType::RoundToNearestEven: y_ref[i] = std::nearbyint(double(x_data[i])); break; case OpType::RoundTowardsZero: y_ref[i] = std::trunc(double(x_data[i])); break; case OpType::RoundUp: y_ref[i] = std::ceil(double(x_data[i])); break; case OpType::RoundDown: y_ref[i] = std::floor(double(x_data[i])); break; case OpType::Square: y_ref[i] = double(x_data[i]) * double(x_data[i]); break; case OpType::SquareRoot: y_ref[i] = std::sqrt(double(x_data[i])); break; case OpType::Sigmoid: { const double e = std::exp(double(x_data[i])); y_ref[i] = e / (1.0 + e); break; } } } // Prepare parameters. union { union xnn_f32_abs_params abs; union xnn_f32_elu_params elu; union xnn_f32_relu_params relu; union xnn_f32_lrelu_params lrelu; union xnn_f32_neg_params neg; union xnn_f32_rnd_params rnd; union xnn_f32_sqrt_params sqrt; } params; switch (op_type) { case OpType::Abs: switch (variant) { case Variant::Native: params.abs = xnn_init_f32_abs_params(); break; case Variant::Scalar: params.abs = xnn_init_scalar_f32_abs_params(); break; } break; case OpType::ELU: switch (variant) { case Variant::Native: params.elu = xnn_init_f32_elu_params(prescale(), alpha(), beta()); break; case Variant::Scalar: params.elu = xnn_init_scalar_f32_elu_params(prescale(), alpha(), beta()); break; } break; case OpType::LeakyReLU: switch (variant) { case Variant::Native: params.lrelu = xnn_init_f32_lrelu_params(slope()); break; case Variant::Scalar: params.lrelu = xnn_init_scalar_f32_lrelu_params(slope()); break; } break; case OpType::Negate: switch (variant) { case Variant::Native: params.neg = xnn_init_f32_neg_params(); break; case Variant::Scalar: params.neg = xnn_init_scalar_f32_neg_params(); break; } break; case OpType::RoundToNearestEven: case OpType::RoundTowardsZero: case OpType::RoundUp: case OpType::RoundDown: switch (variant) { case Variant::Native: params.rnd = xnn_init_f32_rnd_params(); break; case Variant::Scalar: params.rnd = xnn_init_scalar_f32_rnd_params(); break; } break; case OpType::ReLU: case OpType::Sigmoid: case OpType::Square: break; case OpType::SquareRoot: switch (variant) { case Variant::Native: params.sqrt = xnn_init_f32_sqrt_params(); break; case Variant::Scalar: params.sqrt = xnn_init_scalar_f32_sqrt_params(); break; } break; } // Call optimized micro-kernel. vunary(batch_size() * sizeof(float), x_data, y.data(), ¶ms); // Verify results. for (size_t i = 0; i < batch_size(); i++) { ASSERT_NEAR(y[i], y_ref[i], std::max(5.0e-6, std::abs(y_ref[i]) * 1.0e-5)) << "at " << i << " / " << batch_size() << ", x[" << i << "] = " << x[i]; } } } private: size_t batch_size_ = 1; bool inplace_ = false; float slope_ = 0.5f; float prescale_ = 1.0f; float alpha_ = 1.0f; float beta_ = 1.0f; uint8_t qmin_ = 0; uint8_t qmax_ = 255; size_t iterations_ = 15; };