172 lines
5.7 KiB
Python
Raw Normal View History

2023-08-21 22:17:38 +05:30
import os
2023-08-22 20:12:00 +05:30
import sys
import math
2023-08-16 12:42:41 +05:30
import typer
2023-08-21 22:17:38 +05:30
import json
import pandas as pd
import numpy as np
2023-08-16 12:42:41 +05:30
import logging as log
from pathlib import Path
2023-08-21 22:17:38 +05:30
import matplotlib.pyplot as plt
2023-08-22 22:46:49 +05:30
from os import walk
2023-08-23 00:18:51 +05:30
from collections import defaultdict
2023-08-16 12:42:41 +05:30
2023-08-21 22:17:38 +05:30
def read_json(fname):
with open(fname) as f:
cdata = json.load(f)
return cdata
2023-08-16 12:42:41 +05:30
def write_json(dic, fname):
dic_str = {str(k): str(v) for k,v in dic.items()}
jdump = json.dumps(dic_str, indent=4, sort_keys=True)
with open(fname, 'w') as f:
f.write(jdump)
2023-08-16 12:42:41 +05:30
2023-08-24 18:46:18 +05:30
def read_dict(fname):
with open(fname, 'r') as f:
return eval(f.read())
def write_dict(dic, fname):
with open(fname, 'w') as f:
return f.write(str(dic))
2023-08-21 22:17:38 +05:30
def read_csv(fname):
df = pd.read_csv(fname, header=0, comment='#', skipinitialspace = True )
return df
2023-08-16 12:42:41 +05:30
2023-08-21 22:17:38 +05:30
def write_csv(df, fname):
df.to_csv(fname)
2023-08-16 12:42:41 +05:30
def compute_view_finalisation_times(df, conf, oprefix, tag="tag", plot=False):
2023-08-22 20:12:00 +05:30
num_nodes = conf["node_count"]
two3rd = math.floor(num_nodes * 2/3) + 1
views, view2fin_time = df.current_view.unique()[:-2], {}
log.debug(f'views: {views}')
for start_view in views:
end_view = start_view + 2
start_idx = df.index[(df['current_view'] == start_view)][0]
end_idx = df.index[(df['current_view'] == end_view)][two3rd-1]
start_step = df.iloc[start_idx].step_id
end_step = df.iloc[end_idx].step_id
view2fin_time[start_view] = end_step - start_step
log.debug(f'{start_view}({start_idx}), {end_view}({end_idx}) : {end_step} - {start_step} = {view2fin_time[start_view]}')
if not plot:
return view2fin_time
2023-08-22 20:12:00 +05:30
fig, axes = plt.subplots(1, 1, layout='constrained', sharey=False)
fig.set_figwidth(12)
fig.set_figheight(10)
fig.suptitle(f'View Finalisation Times - {tag}')
axes.set_ylabel("Number of Epochs to Finalise a View")
axes.set_xlabel("Views")
axes.set_xticks([x for x in view2fin_time.keys()])
axes.set_yticks([x for x in range(0, max(view2fin_time.values())+2)])
axes.plot(view2fin_time.keys(), view2fin_time.values(), linestyle='--', marker='o')
plt.show()
plt.savefig(f'{oprefix}-view-finalisation-times.pdf', format="pdf", bbox_inches="tight")
2023-08-16 12:42:41 +05:30
2023-08-24 18:46:18 +05:30
def compute_view_lengths(path, oprefix):
pathlen2vfins = {}
conf_fnames = next(walk(f'{path}/configs'), (None, None, []))[2]
for conf in conf_fnames:
tag = os.path.splitext(os.path.basename(conf))[0]
cfile, dfile = f'{path}/configs/{conf}', f'{path}/output/output/{tag}.csv'
conf, df = read_json(cfile), read_csv(dfile)
simtype = conf["stream_settings"]["path"].split("/")[1].split("_")[0]
view2fin = compute_view_finalisation_times(df, conf, oprefix, tag, plot=False)
if not view2fin: # < 2 views
continue
num_nodes = conf["node_count"]
if simtype == "branch":
max_depth = conf["overlay_settings"]["branch_depth"]
else:
max_depth = math.log(num_nodes + 1, 2) - 1
if num_nodes in pathlen2vfins:
pathlen2vfins[num_nodes].append((simtype, max_depth, view2fin, tag))
else:
pathlen2vfins[num_nodes] = [(simtype, max_depth, view2fin, tag)]
return pathlen2vfins
2023-08-22 22:46:49 +05:30
2023-08-21 22:17:38 +05:30
app = typer.Typer()
2023-08-16 12:42:41 +05:30
2023-08-21 22:17:38 +05:30
@app.command()
2023-08-22 22:46:49 +05:30
def view(ctx: typer.Context,
2023-08-22 20:12:00 +05:30
data_file: Path = typer.Option("config.json",
exists=True, file_okay=True, readable=True,
help="Set the simulation config file"),
config_file: Path = typer.Option("simout.csv",
2023-08-16 12:42:41 +05:30
exists=True, file_okay=True, readable=True,
help="Set the simulation data file"),
oprefix: str = typer.Option("output",
help="Set the output prefix for the plots"),
):
2023-08-21 22:17:38 +05:30
log.basicConfig(level=log.INFO)
2023-08-22 20:12:00 +05:30
tag = os.path.splitext(os.path.basename(data_file))[0]
2023-08-22 22:46:49 +05:30
conf, df = read_json(config_file), read_csv(data_file)
compute_view_finalisation_times(df, conf, oprefix, tag, plot=True)
2023-08-21 22:17:38 +05:30
2023-08-22 22:46:49 +05:30
@app.command()
def views(ctx: typer.Context,
path: Path = typer.Option("../",
2023-08-22 20:12:00 +05:30
exists=True, dir_okay=True, readable=True,
help="Set the simulation config file"),
oprefix: str = typer.Option("output",
help="Set the output prefix for the plots")
):
2023-08-21 22:33:33 +05:30
log.basicConfig(level=log.INFO)
2023-08-24 18:46:18 +05:30
#pathlen2vfins = compute_view_lengths(path, oprefix)
#write_json(pathlen2vfins, f'{oprefix}-viewtimes.json')
#write_dict(pathlen2vfins, f'{oprefix}-viewtimes.dict')
pathlen2vfins = read_dict(f'{oprefix}-viewtimes.dict')
data, log4, log5 = [[], []], [], []
print("READ FROM FILE", pathlen2vfins)
for n in sorted(list(map(int, pathlen2vfins.keys()))):
vfin = pathlen2vfins[n]
for run in vfin:
if "default" in run[3] and "tree" in run[0]:
data[0].append(n)
data[1].append(int(run[2][0]))
log4.append(int(run[1])*1)
log5.append(int(run[1])*2)
fig, axes = plt.subplots(1, 1, layout='constrained', sharey=False)
fig.set_figwidth(12)
fig.set_figheight(10)
fig.suptitle(f'View Finalisation Times')
axes.set_ylabel("Number of Epochs")
axes.set_xlabel("Number of Nodes")
l1 = axes.plot(data[0], data[1], linestyle='-', marker='o', label='Carnot')
l2 = axes.plot(data[0], log4, linestyle='--', marker='x', label='1 * log(#nodes)')
l3 = axes.plot(data[0], log5, linestyle='--', marker='x', label='2 * log(#nodes)')
l = l1 + l2 + l3
labels = [curve.get_label() for curve in l]
axes.legend(l, labels, loc="lower right")
plt.show()
plt.savefig(f'{oprefix}-output.pdf', format="pdf", bbox_inches="tight")
2023-08-21 22:17:38 +05:30
@app.command()
def other_commands():
pass
2023-08-16 12:42:41 +05:30
if __name__ == "__main__":
2023-08-21 22:17:38 +05:30
app()