solver_main.cu File Reference

the generic main file for all GPU solvers More...

#include <stdlib.h>
#include <stdio.h>
#include <math.h>
#include <string.h>
#include <stdbool.h>
#include <complex.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <cuda.h>
#include <cuda_runtime.h>
#include <helper_cuda.h>
#include <cuComplex.h>
#include "header.cuh"
#include "timer.h"
#include "solver.cuh"
#include "gpu_memory.cuh"
#include "read_initial_conditions.cuh"
#include "launch_bounds.cuh"
#include <assert.h>

Include dependency graph for solver_main.cu:

Go to the source code of this file.

Functions
void	write_log (int NUM, double t, const double *y_host, FILE *pFile)
	Writes state vectors to file. More...
void	memcpy2D_in (double *dst, const int pitch_dst, double const *src, const int pitch_src, const int offset, const size_t width, const int height)
	A convienience method to copy memory between host pointers of different pitches, widths and heights. Enables easier use of CUDA's cudaMemcpy2D functions. More...
void	memcpy2D_out (double *dst, const int pitch_dst, double const *src, const int pitch_src, const int offset, const size_t width, const int height)
	A convienience method to copy memory between host pointers of different pitches, widths and heights. Enables easier use of CUDA's cudaMemcpy2D functions. More...
int	main (int argc, char *argv[])

Variables
__device__ int	integrator_steps [DIVERGENCE_TEST] = {0}
	If DIVERGENCE_TEST is defined, this creates a device array for tracking. More...

Detailed Description

the generic main file for all GPU solvers

Author

Nicholas Curtis

Date

03/09/2015

Contains main function, setup, initialization, logging, timing and driver functions

Definition in file solver_main.cu.

Function Documentation

main()

int main	(	int	argc,
		char *	argv[]
	)

Main function

Parameters

[in]	argc	command line argument count
[in]	argv	command line argument vector

This allows running the integrators from the command line. The syntax is as follows:
./solver-name [num_threads] [num_IVPs]

• num_threads [Optional, Default:1]
  • The number OpenMP threads to utilize
  • The number of threads cannot be greater than recognized by OpenMP via omp_get_max_threads()
• num_IVPs [Optional, Default:1]
  • The number of initial value problems to solve.
  • This must be less than the number of conditions in the data file if SAME_IC is not defined.
  • If SAME_IC is defined, then the initial conditions in the mechanism files will be used.

Number of independent systems

Definition at line 150 of file solver_main.cu.

memcpy2D_in()

void memcpy2D_in ( double *  dst, const int  pitch_dst, double const *  src, const int  pitch_src, const int  offset, const size_t  width, const int  height  )

inline

A convienience method to copy memory between host pointers of different pitches, widths and heights. Enables easier use of CUDA's cudaMemcpy2D functions.

Parameters

[out]	dst	The destination array
[in]	pitch_dst	The width (in bytes) of the destination array. This corresponds to the padded number of IVPs to be solved.
[in]	src	The source pointer
[in]	pitch_src	The width (in bytes) of the source array. This corresponds to the (non-padded) number of IVPs read by read_initial_conditions
[in]	offset	The offset within the source array (IVP index) to copy from. This is useful in the case (for large models) where the solver and state vector memory will not fit in device memory and the integration must be split into multiple kernel calls.
[in]	width	The size (in bytes) of memory to copy for each entry in the state vector
[in]	height	The number of entries in the state vector

Definition at line 96 of file solver_main.cu.

memcpy2D_out()

void memcpy2D_out ( double *  dst, const int  pitch_dst, double const *  src, const int  pitch_src, const int  offset, const size_t  width, const int  height  )

inline

A convienience method to copy memory between host pointers of different pitches, widths and heights. Enables easier use of CUDA's cudaMemcpy2D functions.

Parameters

[out]	dst	The destination array
[in]	pitch_dst	The width (in bytes) of the source array. This corresponds to the (non-padded) number of IVPs read by read_initial_conditions
[in]	src	The source pointer
[in]	pitch_src	The width (in bytes) of the destination array. This corresponds to the padded number of IVPs to be solved.
[in]	offset	The offset within the destination array (IVP index) to copy to. This is useful in the case (for large models) where the solver and state vector memory will not fit in device memory and the integration must be split into multiple kernel calls.
[in]	width	The size (in bytes) of memory to copy for each entry in the state vector
[in]	height	The number of entries in the state vector

Definition at line 122 of file solver_main.cu.

write_log()

void write_log	(	int	NUM,
		double	t,
		const double *	y_host,
		FILE *	pFile
	)

Writes state vectors to file.

Parameters

[in]	NUM	the number of state vectors to write
[in]	t	the current system time
[in]	y_host	the current state vectors
[in]	pFile	the opened binary file object

The resulting file is updated as: system time
temperature, mass fractions (State #1)
temperature, mass fractions (State #2)...

Definition at line 59 of file solver_main.cu.

Variable Documentation

integrator_steps

__device__ int integrator_steps[DIVERGENCE_TEST] = {0}

If DIVERGENCE_TEST is defined, this creates a device array for tracking.

Definition at line 44 of file solver_main.cu.