Command line script

Partition

$ ./partition.py [-h] -f FILE [-p PARTITION_CLASS] [-d DATASET_SIZE] [-ng NGPU] [-nc NCPU] [-l] [-g] [-df]

                                            

short	sub-command	description
-h	--HELP	Show this help message and exit
-f	--FILE	Input the json file
-p	--PARTITION_CLASS	Inputs the partition ratio
-d	--DATASET_SIZE	Inputs the size of the dataset
-ng	--NGPU	Number of GPUs
-nc	--NCPU	Number of CPUs
-l	--LOG	Flag for turning on LOG
-g	--GRAPH	Flag for plotting GANTT chart for execution
-df	--DUMP_OUTPUT_FILE	Flag for dumping output file for a kernel

Scheduler

$ ./scheduler.py [-h] -f FILE [-s SELECT] [-ng NGPU] [-nc NCPU] [-l] [-g] [-df]
                        
                      

short	sub-command	description
-h	--HELP	Show this help message and exit
-f	--FILE	Input task file containing list of < json filename, partition class, dataset tuples >
-s	--SELECT	Scheduling heuristic (baseline, lookahead, adbias)
-ng	--NGPU	Number of GPUs
-nc	--NCPU	Number of CPUs
-l	--LOG	Flag for turning on LOG
-g	--GRAPH	Flag for plotting GANTT chart for execution
-df	--DUMP_OUTPUT_FILE	Flag for dumping output file for a kernel

Get Optimal Partition

$ ./get_optimal_partition.py [-h] -f FILE [-d DATASET_SIZE] [-nr RUNS]
                        
                      

short	sub-command	description
-h	--HELP	Show this help message and exit
-f	--FILE	Input the json file
-d	--DATASET_SIZE	Inputs the size of the dataset
-nr	--RUNS	Number of runs for executing each partitioned variant of the original program

Run Scheduler

$ ./run_scheduler.py [-h] [-f FILE] -t TASKS [-s SELECT] [-ng NGPU] [-nc NCPU] [-nr RUNS]

                        
                      

short	sub-command	description
-h	--HELP	Show this help message and exit
-f	--FILE	Input task file containing list of < json filename, partition class, dataset tuples >
-s	--SELECT	Scheduling heuristic (baseline, lookahead, adbias)
-ng	--NGPU	Number of GPUs
-nc	--NCPU	Number of CPUs
-nr	--RUNS	Number of runs for executing each scheduling algorithm

Log Parser

$ ./log_parser.py [-h] [-f FILE] [-d] [-p] [-c CALLBACK] [-he] [-ke] [-dvt DEVICE_TYPE] [-k KERNEL_NAME]

                                            

short	sub-command	description
-h	--HELP	Show this help message and exit
-d	--DISPATCH	Displays information regarding dispatch events for kernel
-p	--PARTITION_INFO	Displays information regarding partitioning for CPU/GPU/both
-c	--CALLBACK	Displays information regarding callback for PROBE, RESET, TRIGGERED depending on option
-he	--HOST_EVENT	Displays information regarding Host events for CPU/GPU/both
-dvt	--DEVICE_TYPE	Displays LOG information for specified device type - CPU/GPU/BOTH
-k	--KERNEL_NAME	Displays LOG information for the specified kernel

Parse Output Dump

$ ./parse_output_dump.py >filename.pickle<