12 #ifndef PLSSVM_BACKENDS_CUDA_CSVM_HPP_
13 #define PLSSVM_BACKENDS_CUDA_CSVM_HPP_
22 #include <type_traits>
79 template <
typename... Args,
PLSSVM_REQUIRES(::plssvm::detail::has_only_parameter_named_args_v<Args...>)>
80 explicit csvm(Args &&...named_args) :
91 template <
typename... Args,
PLSSVM_REQUIRES(::plssvm::detail::has_only_parameter_named_args_v<Args...>)>
93 base_type{ std::forward<Args>(named_args)... } {
128 void run_q_kernel(std::
size_t device, const ::
plssvm::detail::execution_range &range, const ::
plssvm::detail::
parameter<
float> ¶ms,
device_ptr_type<
float> &q_d, const
device_ptr_type<
float> &data_d, const
device_ptr_type<
float> &data_last_d, std::
size_t num_data_points_padded, std::
size_t num_features) const final { this->
run_q_kernel_impl(device, range, params, q_d, data_d, data_last_d, num_data_points_padded, num_features); }
132 void run_q_kernel(std::size_t device, const ::plssvm::detail::execution_range &range, const ::plssvm::detail::parameter<double> ¶ms,
device_ptr_type<double> &q_d,
const device_ptr_type<double> &data_d,
const device_ptr_type<double> &data_last_d, std::size_t num_data_points_padded, std::size_t num_features)
const final { this->
run_q_kernel_impl(device, range, params, q_d, data_d, data_last_d, num_data_points_padded, num_features); }
136 template <
typename real_type>
141 void run_svm_kernel(std::size_t device, const ::plssvm::detail::execution_range &range, const ::plssvm::detail::parameter<float> ¶ms,
const device_ptr_type<float> &q_d,
device_ptr_type<float> &r_d,
const device_ptr_type<float> &x_d,
const device_ptr_type<float> &data_d,
float QA_cost,
float add, std::size_t num_data_points_padded, std::size_t num_features)
const final { this->
run_svm_kernel_impl(device, range, params, q_d, r_d, x_d, data_d, QA_cost, add, num_data_points_padded, num_features); }
145 void run_svm_kernel(std::size_t device, const ::plssvm::detail::execution_range &range, const ::plssvm::detail::parameter<double> ¶ms,
const device_ptr_type<double> &q_d,
device_ptr_type<double> &r_d,
const device_ptr_type<double> &x_d,
const device_ptr_type<double> &data_d,
double QA_cost,
double add, std::size_t num_data_points_padded, std::size_t num_features)
const final { this->
run_svm_kernel_impl(device, range, params, q_d, r_d, x_d, data_d, QA_cost, add, num_data_points_padded, num_features); }
149 template <
typename real_type>
150 void run_svm_kernel_impl(std::size_t device, const ::plssvm::detail::execution_range &range, const ::plssvm::detail::parameter<real_type> ¶ms,
const device_ptr_type<real_type> &q_d,
device_ptr_type<real_type> &r_d,
const device_ptr_type<real_type> &x_d,
const device_ptr_type<real_type> &data_d, real_type QA_cost, real_type add, std::size_t num_data_points_padded, std::size_t num_features)
const;
154 void run_w_kernel(std::size_t device, const ::plssvm::detail::execution_range &range,
device_ptr_type<float> &w_d,
const device_ptr_type<float> &alpha_d,
const device_ptr_type<float> &data_d,
const device_ptr_type<float> &data_last_d, std::size_t num_data_points, std::size_t num_features)
const final { this->
run_w_kernel_impl(device, range, w_d, alpha_d, data_d, data_last_d, num_data_points, num_features); }
158 void run_w_kernel(std::size_t device, const ::plssvm::detail::execution_range &range,
device_ptr_type<double> &w_d,
const device_ptr_type<double> &alpha_d,
const device_ptr_type<double> &data_d,
const device_ptr_type<double> &data_last_d, std::size_t num_data_points, std::size_t num_features)
const final { this->
run_w_kernel_impl(device, range, w_d, alpha_d, data_d, data_last_d, num_data_points, num_features); }
162 template <
typename real_type>
167 void run_predict_kernel(const ::plssvm::detail::execution_range &range, const ::plssvm::detail::parameter<float> ¶ms,
device_ptr_type<float> &out_d,
const device_ptr_type<float> &alpha_d,
const device_ptr_type<float> &point_d,
const device_ptr_type<float> &data_d,
const device_ptr_type<float> &data_last_d, std::size_t num_support_vectors, std::size_t num_predict_points, std::size_t num_features)
const final { this->
run_predict_kernel_impl(range, params, out_d, alpha_d, point_d, data_d, data_last_d, num_support_vectors, num_predict_points, num_features); }
171 void run_predict_kernel(const ::plssvm::detail::execution_range &range, const ::plssvm::detail::parameter<double> ¶ms,
device_ptr_type<double> &out_d,
const device_ptr_type<double> &alpha_d,
const device_ptr_type<double> &point_d,
const device_ptr_type<double> &data_d,
const device_ptr_type<double> &data_last_d, std::size_t num_support_vectors, std::size_t num_predict_points, std::size_t num_features)
const final { this->
run_predict_kernel_impl(range, params, out_d, alpha_d, point_d, data_d, data_last_d, num_support_vectors, num_predict_points, num_features); }
175 template <
typename real_type>
Base class for all C-SVM backends.
Definition: csvm.hpp:50
A C-SVM implementation using CUDA as backend.
Definition: csvm.hpp:39
void device_synchronize(const queue_type &queue) const final
Synchronize the device denoted by queue.
void run_predict_kernel(const ::plssvm::detail::execution_range &range, const ::plssvm::detail::parameter< double > ¶ms, device_ptr_type< double > &out_d, const device_ptr_type< double > &alpha_d, const device_ptr_type< double > &point_d, const device_ptr_type< double > &data_d, const device_ptr_type< double > &data_last_d, std::size_t num_support_vectors, std::size_t num_predict_points, std::size_t num_features) const final
Run the device kernel (only on the first device) to predict the new data points point_d.
Definition: csvm.hpp:171
void run_q_kernel(std::size_t device, const ::plssvm::detail::execution_range &range, const ::plssvm::detail::parameter< double > ¶ms, device_ptr_type< double > &q_d, const device_ptr_type< double > &data_d, const device_ptr_type< double > &data_last_d, std::size_t num_data_points_padded, std::size_t num_features) const final
Run the device kernel filling the q vector.
Definition: csvm.hpp:132
csvm(target_platform target, parameter params={})
Construct a new C-SVM using the CUDA backend on the target platform with the parameters given through...
csvm(csvm &&) noexcept=default
Default move-constructor since a virtual destructor has been declared. noexcept
void run_q_kernel_impl(std::size_t device, const ::plssvm::detail::execution_range &range, const ::plssvm::detail::parameter< real_type > ¶ms, device_ptr_type< real_type > &q_d, const device_ptr_type< real_type > &data_d, const device_ptr_type< real_type > &data_last_d, std::size_t num_data_points_padded, std::size_t num_features) const
Run the device kernel filling the q vector.
csvm(const csvm &)=delete
Delete copy-constructor since a CSVM is a move-only type.
void init(target_platform target)
Initialize all important states related to the CUDA backend.
void run_w_kernel_impl(std::size_t device, const ::plssvm::detail::execution_range &range, device_ptr_type< real_type > &w_d, const device_ptr_type< real_type > &alpha_d, const device_ptr_type< real_type > &data_d, const device_ptr_type< real_type > &data_last_d, std::size_t num_data_points, std::size_t num_features) const
Run the device kernel the calculate the w vector used to speed up the prediction when using the linea...
void run_svm_kernel_impl(std::size_t device, const ::plssvm::detail::execution_range &range, const ::plssvm::detail::parameter< real_type > ¶ms, const device_ptr_type< real_type > &q_d, device_ptr_type< real_type > &r_d, const device_ptr_type< real_type > &x_d, const device_ptr_type< real_type > &data_d, real_type QA_cost, real_type add, std::size_t num_data_points_padded, std::size_t num_features) const
Run the main device kernel used in the CG algorithm.
void run_predict_kernel_impl(const ::plssvm::detail::execution_range &range, const ::plssvm::detail::parameter< real_type > ¶ms, device_ptr_type< real_type > &out_d, const device_ptr_type< real_type > &alpha_d, const device_ptr_type< real_type > &point_d, const device_ptr_type< real_type > &data_d, const device_ptr_type< real_type > &data_last_d, std::size_t num_support_vectors, std::size_t num_predict_points, std::size_t num_features) const
Run the device kernel (only on the first device) to predict the new data points point_d.
csvm(Args &&...named_args)
Construct a new C-SVM using the CUDA backend and the optionally provided named_args.
Definition: csvm.hpp:80
void run_w_kernel(std::size_t device, const ::plssvm::detail::execution_range &range, device_ptr_type< double > &w_d, const device_ptr_type< double > &alpha_d, const device_ptr_type< double > &data_d, const device_ptr_type< double > &data_last_d, std::size_t num_data_points, std::size_t num_features) const final
Run the device kernel the calculate the w vector used to speed up the prediction when using the linea...
Definition: csvm.hpp:158
void run_svm_kernel(std::size_t device, const ::plssvm::detail::execution_range &range, const ::plssvm::detail::parameter< float > ¶ms, const device_ptr_type< float > &q_d, device_ptr_type< float > &r_d, const device_ptr_type< float > &x_d, const device_ptr_type< float > &data_d, float QA_cost, float add, std::size_t num_data_points_padded, std::size_t num_features) const final
Run the main device kernel used in the CG algorithm.
Definition: csvm.hpp:141
void run_predict_kernel(const ::plssvm::detail::execution_range &range, const ::plssvm::detail::parameter< float > ¶ms, device_ptr_type< float > &out_d, const device_ptr_type< float > &alpha_d, const device_ptr_type< float > &point_d, const device_ptr_type< float > &data_d, const device_ptr_type< float > &data_last_d, std::size_t num_support_vectors, std::size_t num_predict_points, std::size_t num_features) const final
Run the device kernel (only on the first device) to predict the new data points point_d.
Definition: csvm.hpp:167
void run_w_kernel(std::size_t device, const ::plssvm::detail::execution_range &range, device_ptr_type< float > &w_d, const device_ptr_type< float > &alpha_d, const device_ptr_type< float > &data_d, const device_ptr_type< float > &data_last_d, std::size_t num_data_points, std::size_t num_features) const final
Run the device kernel the calculate the w vector used to speed up the prediction when using the linea...
Definition: csvm.hpp:154
csvm(parameter params={})
Construct a new C-SVM using the CUDA backend with the parameters given through params.
void run_svm_kernel(std::size_t device, const ::plssvm::detail::execution_range &range, const ::plssvm::detail::parameter< double > ¶ms, const device_ptr_type< double > &q_d, device_ptr_type< double > &r_d, const device_ptr_type< double > &x_d, const device_ptr_type< double > &data_d, double QA_cost, double add, std::size_t num_data_points_padded, std::size_t num_features) const final
Run the main device kernel used in the CG algorithm.
Definition: csvm.hpp:145
void run_q_kernel(std::size_t device, const ::plssvm::detail::execution_range &range, const ::plssvm::detail::parameter< float > ¶ms, device_ptr_type< float > &q_d, const device_ptr_type< float > &data_d, const device_ptr_type< float > &data_last_d, std::size_t num_data_points_padded, std::size_t num_features) const final
Run the device kernel filling the q vector.
Definition: csvm.hpp:128
csvm(const target_platform target, Args &&...named_args)
Construct a new C-SVM using the CUDA backend on the target platform and the optionally provided named...
Definition: csvm.hpp:92
Class specifying a backend independent execution range.
Definition: execution_range.hpp:31
A C-SVM implementation for all GPU backends to reduce code duplication.
Definition: gpu_csvm.hpp:46
std::vector< queue_type > devices_
The available/used backend devices.
Definition: gpu_csvm.hpp:280
int queue_type
The type of the device queue (dependent on the used backend).
Definition: gpu_csvm.hpp:52
detail::device_ptr< real_type > device_ptr_type
The type of the device pointer (dependent on the used backend).
Definition: gpu_csvm.hpp:50
Small wrapper around a CUDA device pointer.
Defines the base class for all C-SVM backends using a GPU. Used for code duplication reduction.
The main namespace containing all public API functions.
Definition: backend_types.hpp:24
target_platform
Enum class for all possible targets.
Definition: target_platforms.hpp:25
Implements the parameter class encapsulating all important C-SVM parameters.
Sets the value of the value member to true if T is a C-SVM using an available backend....
Definition: csvm.hpp:410
#define PLSSVM_REQUIRES(...)
A shorthand macro for the std::enable_if_t type trait.
Definition: type_traits.hpp:33