ESPResSo
Extensible Simulation Package for Research on Soft Matter Systems
Loading...
Searching...
No Matches
init.hpp File Reference
#include "config/config.hpp"
#include "utils.hpp"
#include <boost/serialization/access.hpp>
#include <cstddef>
#include <string>
#include <utility>
#include <vector>
+ Include dependency graph for cuda/init.hpp:
+ This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Classes

struct  EspressoGpuDevice
 Struct to hold information relevant to ESPResSo about GPUs. More...
 

Functions

void cuda_init ()
 Initializes the CUDA stream.
 
int cuda_get_n_gpus ()
 Get the number of CUDA devices on the local host.
 
bool cuda_check_gpu_compute_capability (int dev)
 Check that a given GPU has compute capability.
 
std::string cuda_get_gpu_name (int dev)
 Get the name of a CUDA device.
 
void cuda_set_device (int dev)
 Choose a device for future CUDA computations.
 
int cuda_get_device ()
 Get the current CUDA device.
 
bool cuda_test_device_access ()
 Test if communication to the CUDA device works.
 
void cuda_check_device ()
 Check that a device is available, that its compute capability is sufficient for ESPResSo, and that data can be written to and read from it.
 
std::vector< EspressoGpuDevicecuda_gather_gpus ()
 Gather unique list of CUDA devices on all nodes.
 
EspressoGpuDevice cuda_get_device_props (int dev)
 Get properties of a CUDA device.
 
void cuda_on_program_start ()
 Called on program start.
 

Function Documentation

◆ cuda_check_device()

void cuda_check_device ( )

Check that a device is available, that its compute capability is sufficient for ESPResSo, and that data can be written to and read from it.

Otherwise, throw an exception.

Definition at line 102 of file init_cuda.cu.

References cuda_check_gpu_compute_capability(), cuda_get_device(), cuda_get_n_gpus(), cuda_test_device_access(), and stream.

◆ cuda_check_gpu_compute_capability()

bool cuda_check_gpu_compute_capability ( int  dev)

Check that a given GPU has compute capability.

The minimal compute capability required by ESPResSo is computeCapabilityMinMajor . computeCapabilityMinMinor .

Parameters
devCUDA device number
Returns
false if the GPU meets the requirements, else true.

Definition at line 50 of file init_cuda.cu.

References computeCapabilityMinMajor, computeCapabilityMinMinor, CUDA_CHECK, and stream.

Referenced by cuda_check_device().

◆ cuda_gather_gpus()

std::vector< EspressoGpuDevice > cuda_gather_gpus ( )

Gather unique list of CUDA devices on all nodes.

Returns
vector of device properties.

Gather unique list of CUDA devices on all nodes.

It relies on MPI_Get_processor_name() to get a unique identifier of the physical node, as opposed to the logical rank of which there can be more than one per node.

Definition at line 56 of file cuda/init.cpp.

References comm_cart, cuda_get_device_props(), cuda_get_n_gpus(), Utils::Mpi::gather_buffer(), invoke_skip_cuda_exceptions(), stream, and this_node.

Referenced by ScriptInterface::System::CudaInitHandle::do_call_method().

◆ cuda_get_device()

int cuda_get_device ( )

Get the current CUDA device.

Returns
the current device's number.

Definition at line 84 of file init_cuda.cu.

References CUDA_CHECK, and stream.

Referenced by cuda_check_device(), and ScriptInterface::System::CudaInitHandle::CudaInitHandle().

◆ cuda_get_device_props()

EspressoGpuDevice cuda_get_device_props ( int  dev)

Get properties of a CUDA device.

Parameters
devCUDA device number

Definition at line 64 of file init_cuda.cu.

References CUDA_CHECK, and stream.

Referenced by cuda_gather_gpus().

◆ cuda_get_gpu_name()

std::string cuda_get_gpu_name ( int  dev)

Get the name of a CUDA device.

Parameters
[in]devthe CUDA device number to ask the name for

Definition at line 58 of file init_cuda.cu.

References CUDA_CHECK, and stream.

Referenced by ScriptInterface::System::CudaInitHandle::do_call_method().

◆ cuda_get_n_gpus()

int cuda_get_n_gpus ( )

Get the number of CUDA devices on the local host.

Returns
the number of GPUs.

Definition at line 44 of file init_cuda.cu.

References CUDA_CHECK, and stream.

Referenced by cuda_check_device(), cuda_gather_gpus(), and ScriptInterface::System::CudaInitHandle::do_call_method().

◆ cuda_init()

void cuda_init ( )

Initializes the CUDA stream.

Definition at line 42 of file init_cuda.cu.

References CUDA_CHECK, and stream.

Referenced by cuda_on_program_start().

◆ cuda_on_program_start()

void cuda_on_program_start ( )

Called on program start.

Definition at line 91 of file cuda/init.cpp.

References communicator, cuda_init(), invoke_skip_cuda_exceptions(), and this_node.

Referenced by CommunicationEnvironment::CommunicationEnvironment().

◆ cuda_set_device()

void cuda_set_device ( int  dev)

Choose a device for future CUDA computations.

Parameters
devthe device to use

Definition at line 78 of file init_cuda.cu.

References CUDA_CHECK, and stream.

Referenced by ScriptInterface::System::CudaInitHandle::CudaInitHandle().

◆ cuda_test_device_access()

bool cuda_test_device_access ( )

Test if communication to the CUDA device works.

Returns
false on success, else true.

Definition at line 90 of file init_cuda.cu.

References CUDA_CHECK, and stream.

Referenced by cuda_check_device().