scaling

#!/usr/bin/env python3

def cli():
    import os,sys,argparse
    cli = argparse.ArgumentParser(description='CLARA scaling test')
    cli.add_argument('-P',help='plot only',action='store_true')
    cli.add_argument('-y',help='YAML file',default=None)
    cli.add_argument('-c',help='CLARA_HOME path',default=os.getenv('CLARA_HOME',None))
    cli.add_argument('-t',help='threads (default=4,8)',default='4,8')
    cli.add_argument('-e',help='events per thread',default=100,type=int)
    cli.add_argument('-i',help='input data file',default=None)
    cfg = cli.parse_args()
    cfg.t = cfg.t.split(',')
    if cfg.P:
        cfg.i = 'scaling.txt'
    else:
        if cfg.y is None: sys.exit('-y YAML is required w/o -P')
        if cfg.c is None: sys.exit('-c or $CLARA_HOME is required w/o -P')
    if cfg.i is None: sys.exit('-i is required')
    return cfg

def run(cmd):
    import subprocess
    print('run >>> '+' '.join(cmd))
    p = subprocess.Popen(cmd,stdout=subprocess.PIPE,stderr=subprocess.STDOUT,universal_newlines=True,encoding='latin-1')
    for line in iter(p.stdout.readline, ''):
        line = line.strip()
        if len(line) > 0:
            yield line
    p.wait()
    if p.returncode != 0:
        pass

def benchmark(cfg, threads):
    import sys,shutil,collections
    if not shutil.which('run-clara'):
        sys.exit('run-clara is not in $PATH')
    exiting,benchmarks = False,collections.OrderedDict()
    cmd = ['run-clara',
            '-c',cfg.c,
            '-n',str(cfg.e*int(threads)),
            '-t',str(threads),
            '-y',cfg.y,
            '-o',f'tmp-scaling-{threads}',
            cfg.i]
    for line in run(cmd):
        cols = line.split()
        print(line)
        try:
            if line.find('Benchmark results:') >= 0:
                exiting = True
            elif line.find('Processing is complete') >= 0:
                exiting = False
            elif len(cols) > 20:
                if line.find('Processed') >= 0:
                    benchmarks['Event'] = float(cols[12])
            elif exiting:
                # catch-all for services:
                if len(cols) > 14:
                    if 'Services' not in benchmarks:
                        benchmarks['Services'] = collections.OrderedDict()
                    benchmarks['Services'][cols[2]] = float(cols[14])
                # FIXME:  what are these, why don't they add up?
                elif line.find('Average processing time') >= 0:
                    benchmarks['Avg'] = float(cols[6])
                elif line.find('Total processing time') >= 0:
                    benchmarks['Total'] = float(cols[6])
                elif line.find('Total orchestrator time') >= 0:
                    benchmarks['Orch'] = float(cols[6])
        except ValueError:
            pass
    return benchmarks

def table(benchmarks):
    table = []
    header = [ 'Threads' ]
    b = benchmarks[0][1]
    header.extend([x for x in b if x != 'Services'])
    if 'Services' in b:
        header.extend(b['Services'].keys())
    table.append(header)
    for b in benchmarks:
        threads,benchmark = b[0],b[1]
        row = [threads]
        for k in ['Avg','Total','Orch','Services']:
            if k in benchmark:
                if k == 'Services':
                    row.extend(benchmark[k].values())
                else:
                    row.append(benchmark[k])
        table.append(row)
    return table

def show(benchmarks):
    for row in table(benchmarks):
        print(' '.join([str(x) for x in row])) 

def save(benchmarks):
    with open('scaling.txt','w') as f:
        for row in table(benchmarks):
            f.write(' '.join([str(x) for x in row])+'\n') 

if __name__ == '__main__':
    cfg = cli()
    if not cfg.P:
        benchmarks = []
        for threads in cfg.t:
            benchmarks.append([threads, benchmark(cfg, threads)])
        show(benchmarks)
        save(benchmarks)