Interface to accelerators API. More...

Classes
struct	sirius::GPU_Complex< double >

struct	sirius::GPU_Complex< float >

class	sirius::acc::stream_id
	Helper class to wrap stream id (integer number). More...

Namespaces
module	sirius
	Namespace of the SIRIUS library.

namespace	sirius::acc
	Namespace for accelerator-related functions.

Macros
#define	GPU_PREFIX(x) cuda##x

#define	make_accDoubleComplex make_cuDoubleComplex

#define	make_accFloatComplex make_cuFloatComplex

#define	accCadd cuCadd

#define	accCsub cuCsub

#define	accCmul cuCmul

#define	accCdiv cuCdiv

#define	accConj cuConj

#define	accCaddf cuCaddf

#define	accCsubf cuCsubf

#define	accCmulf cuCmulf

#define	accCdivf cuCdivf

#define	accConjf cuConjf

#define	ACC_DYNAMIC_SHARED(type, var) extern __shared__ type var[];

#define	CALL_DEVICE_API(func__, args__)

Typedefs
using	sirius::acc_stream_t = cudaStream_t

using	sirius::acc_error_t = cudaError_t

using	sirius::acc_complex_float_t = cuFloatComplex

using	sirius::acc_complex_double_t = cuDoubleComplex

template<typename T >
using	sirius::gpu_complex_type = typename GPU_Complex< T >::type

Functions
void	sirius::acc::stack_backtrace ()

int	sirius::acc::num_devices ()
	Get the number of devices. More...

void	sirius::acc::set_device_id (int id__)
	Set the GPU id. More...

int	sirius::acc::get_device_id ()
	Get current device ID. More...

std::vector< acc_stream_t > &	sirius::acc::streams ()
	Vector of device streams. More...

acc_stream_t	sirius::acc::stream (stream_id sid__)
	Return a single device stream. More...

int	sirius::acc::num_streams ()
	Get number of streams. More...

void	sirius::acc::create_streams (int num_streams__)
	Create CUDA streams. More...

void	sirius::acc::destroy_streams ()
	Destroy CUDA streams. More...

void	sirius::acc::sync_stream (stream_id sid__)
	Synchronize a single stream. More...

void	sirius::acc::reset ()
	Reset device. More...

void	sirius::acc::sync ()
	Synchronize device. More...

size_t	sirius::acc::get_free_mem ()

void	sirius::acc::print_device_info (int device_id__, std::ostream &out__)

template<typename T >
void	sirius::acc::copy (T target__, T const source__, size_t n__)
	Copy memory inside a device. More...

template<typename T >
void	sirius::acc::copy (T target__, int ld1__, T const source__, int ld2__, int nrow__, int ncol__)
	2D copy inside a device. More...

template<typename T >
void	sirius::acc::copyin (T target__, T const source__, size_t n__)
	Copy memory from host to device. More...

template<typename T >
void	sirius::acc::copyin (T target__, T const source__, size_t n__, stream_id sid__)
	Asynchronous copy from host to device. More...

template<typename T >
void	sirius::acc::copyin (T target__, int ld1__, T const source__, int ld2__, int nrow__, int ncol__)
	2D copy to the device. More...

template<typename T >
void	sirius::acc::copyin (T target__, int ld1__, T const source__, int ld2__, int nrow__, int ncol__, stream_id sid__)
	Asynchronous 2D copy to the device. More...

template<typename T >
void	sirius::acc::copyout (T target__, T const source__, size_t n__)
	Copy memory from device to host. More...

template<typename T >
void	sirius::acc::copyout (T target__, T const source__, size_t n__, stream_id sid__)
	Asynchronous copy from device to host. More...

template<typename T >
void	sirius::acc::copyout (T target__, int ld1__, T const source__, int ld2__, int nrow__, int ncol__)
	2D copy from device to host. More...

template<typename T >
void	sirius::acc::copyout (T target__, int ld1__, T const source__, int ld2__, int nrow__, int ncol__, stream_id sid__)
	Asynchronous 2D copy from device to host. More...

template<typename T >
void	sirius::acc::zero (T *ptr__, size_t n__)
	Zero the device memory. More...

template<typename T >
void	sirius::acc::zero (T *ptr__, size_t n__, stream_id sid__)

template<typename T >
void	sirius::acc::zero (T *ptr__, int ld__, int nrow__, int ncol__)
	Zero the 2D block of device memory. More...

template<typename T >
T *	sirius::acc::allocate (size_t size__)
	Allocate memory on the GPU. More...

void	sirius::acc::deallocate (void *ptr__)
	Deallocate GPU memory. More...

template<typename T >
T *	sirius::acc::allocate_host (size_t size__)
	Allocate pinned memory on the host. More...

void	sirius::acc::deallocate_host (void *ptr__)
	Deallocate host memory. More...

void	sirius::acc::begin_range_marker (const char *label__)

void	sirius::acc::end_range_marker ()

template<typename T >
void	sirius::acc::register_host (T *ptr__, size_t size__)

void	sirius::acc::unregister_host (void *ptr)

bool	sirius::acc::check_last_error ()

bool	sirius::acc::check_device_ptr (void const *ptr__)

void	sirius::scale_matrix_rows_gpu (int nrow, int ncol, acc_complex_double_t mtrx, double const v)

void	sirius::scale_matrix_elements_gpu (acc_complex_double_t *ptr__, int ld__, int nrow__, int ncol__, double beta__)

Detailed Description

Interface to accelerators API.

Definition in file acc.hpp.

Macro Definition Documentation

◆ GPU_PREFIX

#define GPU_PREFIX ( x ) cuda##x

Definition at line 56 of file acc.hpp.

◆ make_accDoubleComplex

#define make_accDoubleComplex make_cuDoubleComplex

Definition at line 80 of file acc.hpp.

◆ make_accFloatComplex

#define make_accFloatComplex make_cuFloatComplex

Definition at line 81 of file acc.hpp.

◆ accCadd

#define accCadd cuCadd

Definition at line 82 of file acc.hpp.

◆ accCsub

#define accCsub cuCsub

Definition at line 83 of file acc.hpp.

◆ accCmul

#define accCmul cuCmul

Definition at line 84 of file acc.hpp.

◆ accCdiv

#define accCdiv cuCdiv

Definition at line 85 of file acc.hpp.

◆ accConj

#define accConj cuConj

Definition at line 86 of file acc.hpp.

◆ accCaddf

#define accCaddf cuCaddf

Definition at line 87 of file acc.hpp.

◆ accCsubf

#define accCsubf cuCsubf

Definition at line 88 of file acc.hpp.

◆ accCmulf

#define accCmulf cuCmulf

Definition at line 89 of file acc.hpp.

◆ accCdivf

#define accCdivf cuCdivf

Definition at line 90 of file acc.hpp.

◆ accConjf

#define accConjf cuConjf

Definition at line 91 of file acc.hpp.

◆ ACC_DYNAMIC_SHARED

#define ACC_DYNAMIC_SHARED	(	type,
		var
	)	extern __shared__ type var[];

Definition at line 92 of file acc.hpp.

◆ CALL_DEVICE_API

#define CALL_DEVICE_API	(	func__,
		args__
	)

Value:

{                                                                               \
    if (acc::num_devices()) {                                                   \
        acc_error_t error;                                                      \
        error = GPU_PREFIX(func__) args__;                                      \
        if (error != GPU_PREFIX(Success)) {                                     \
            char nm[1024];                                                      \
            gethostname(nm, 1024);                                              \
            std::printf("hostname: %s\n", nm);                                  \
            std::printf("Error in %s at line %i of file %s: %s\n",              \
              #func__, __LINE__, __FILE__, GPU_PREFIX(GetErrorString)(error));  \
            stack_backtrace();                                                  \
        }                                                                       \
    }                                                                           \
}

Definition at line 163 of file acc.hpp.

Classes

Namespaces

Macros

Typedefs

Functions