46#ifndef MUELU_PERFMODELS_HPP
47#define MUELU_PERFMODELS_HPP
53#include <Teuchos_DefaultComm.hpp>
MueLu::DefaultLocalOrdinal LocalOrdinal
MueLu::DefaultScalar Scalar
MueLu::DefaultGlobalOrdinal GlobalOrdinal
void print_pingpong_table(std::ostream &out)
double launch_and_wait_latency_
void print_latency_corrected_stream_vector_table(std::ostream &out)
std::vector< int > pingpong_sizes_
void print_stream_vector_table(std::ostream &out)
void print_stream_vector_table_impl(std::ostream &out, bool use_latency_correction)
std::vector< int > stream_sizes_
double stream_vector_copy_lookup(int SIZE_IN_BYTES)
double launch_latency_lookup()
void stream_vector_make_table(int KERNEL_REPEATS, int LOG_MAX_SIZE=20)
double latency_corrected_stream_vector_lookup(int SIZE_IN_BYTES)
std::vector< double > latency_corrected_stream_copy_times_
void pingpong_make_table(int KERNEL_REPEATS, int LOG_MAX_SIZE, const RCP< const Teuchos::Comm< int > > &comm)
std::vector< double > pingpong_device_times_
double latency_corrected_stream_vector_copy_lookup(int SIZE_IN_BYTES)
double pingpong_device_lookup(int SIZE_IN_BYTES)
std::vector< double > latency_corrected_stream_add_times_
double pingpong_host_lookup(int SIZE_IN_BYTES)
double latency_corrected_stream_vector_add_lookup(int SIZE_IN_BYTES)
double stream_vector_add_lookup(int SIZE_IN_BYTES)
std::vector< double > stream_copy_times_
double stream_vector_lookup(int SIZE_IN_BYTES)
std::vector< double > stream_add_times_
std::vector< double > pingpong_host_times_
void launch_latency_make_table(int KERNEL_REPEATS)
void print_launch_latency_table(std::ostream &out)
Namespace for MueLu classes and methods.
KokkosClassic::DefaultNode::DefaultNodeType DefaultNode