autocorrelation_cuda

Header file for CUDA accelerated algorithms

Currently, no algorithms are used in any other parts of the project, so if CUDA or CUDA-enabled devices are not available, this file can be skipped in compilation by commenting out the OBJECTSCUDA line in the makefile

Defines

THREADS_PER_BLOCK

Functions

void write_file_auto_corr_from_data_file_accel(std::string acfile, std::string chains_file, int dimension, int N_steps, int num_segments, double target_corr)

void write_file_auto_corr_from_data_accel(std::string acfile, double **output, int dimension, int N_steps, int num_segments, double target_corr)

void auto_corr_from_data_accel(double **output, int dimension, int N_steps, int num_segments, double target_corr, double **autocorr)

void launch_ac_gpu(int device, int element, double **data, int length, int dimension, double target_corr, int num_segments)

void ac_gpu_wrapper(int thread, int job_id)

void auto_correlation_spectral_accel(double *chains, int length, double *autocorr)

Variables

static double **chains_internal

static double target_corr_internal

static int dimension_internal

static int chain_length_internal

static int num_segments_internal

static double **autocorr_internal