google-cloud-cpp/plot2.py at data · dopiera/google-cloud-cpp · GitHub

History

125 lines (106 loc) · 3.73 KB

Raw

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

import math

import os

import os.path

import re

import subprocess

import sys

import tempfile

def usage(msg=None):

if msg is not None:

print msg

print "%s: directory filter_expr dimensions_to_plot" % sys.argv[0]

print "Fields available for filtering and plotting are: "

print ("op, file_sz, chunk_sz, buffer_sz, wall_time, cpu_use, bw, "

"cpu_per_byte, crc, md5, weight, progress")

sys.exit(1)

def read_file(f):

res = []

for line in open(f).readlines():

if not line.lstrip():

continue

if line.lstrip().startswith('#'):

continue

(op, file_size, chunk_size, buffer_size, crc, md5, wall_time, cpu_time,

prob, status, prog,

version) = line.rstrip('\n').split(',')

if status <> 'OK':

print "Warning: not OK status"

file_size = int(file_size)

chunk_size = int(chunk_size)

buffer_size = int(buffer_size)

wall_time = int(wall_time)

cpu_time = int(cpu_time)

crc = crc == 'true'

md5 = md5 == 'true'

weight = 1/float(prob)

prog = [[int(i) for i in time_point.split('|')] for time_point in

prog.split(';')]

sample = {}

sample["op"] = op

sample["file_sz"] = file_size / 1048576.

sample["chunk_sz"] = chunk_size / 1048576.

sample["buffer_sz"] = buffer_size / 1048576.

sample["wall_time"] = wall_time

sample["cpu_use"] = float(cpu_time) / wall_time

sample["bw"] = float(file_size) / int(wall_time)

sample["cpu_per_byte"] = float(cpu_time) / file_size

sample["crc"] = crc

sample["md5"] = md5

sample["weight"] = weight

sample["progress"] = prog

res.append(sample)

return res

def escape(s):

return s.replace('\\', '\\\\').replace('"', '\\"').replace('_', '\\_')

class Stats:

def __init__(self):

self.sum = 0.

self.num = 0

self.sum_square = 0.

self.weight_sum = 0

def add(self, x, weight):

self.weight_sum = self.weight_sum + weight

self.num = self.num + 1

self.sum = self.sum + x * weight

self.sum_square = self.sum_square + x*x * weight

def show(self):

ex = self.sum / self.weight_sum

print "Samples=%s EX=%.02f, SD=%.02f" % (self.num, ex,

math.sqrt(self.sum_square / self.weight_sum - ex * ex))

def run_gnuplot(res, fields, filt):

f = tempfile.NamedTemporaryFile()

print >>f, "$map1 << EOD"

stats = Stats()

for sample in res:

weight = sample[-1]

data = sample[:-1]

stats.add(data[-1], weight)

print >>f, ' '.join([str(s) for s in data])

print >>f, "EOD"

if len(fields) == 3:

print >>f, "set title \"(%s, %s) -> %s for %s\"" % (fields[0],

fields[1], fields[2], escape(filt))

print >>f, "splot $map1 with points pt 7 palette"

else:

print >>f, "set title \"%s -> %s for %s\"" % (fields[0],

fields[1], escape(filt))

print >>f, "set yrange[0:%s]" % (max([sample[1] for sample in res]) *

1.1)

print >>f, "plot $map1 with points pt 7 "

f.flush()

stats.show()

subprocess.Popen(['/usr/bin/gnuplot', f.name, '-']).wait()

def main():

if len(sys.argv) <> 4:

usage()

file = sys.argv[1]

filt = sys.argv[2]

fields = sys.argv[3].split(',')

if len(fields) <> 2 and len(fields) <> 3:

usage("expected 2 or 3 comma-separated fields, got %s", len(fields))

res = [[row[field] for field in fields] + [row["weight"]]

for row in read_file(file) if eval(filt, None, row)]

res.sort()

run_gnuplot(res, fields, filt)

if __name__== "__main__":

main()

google-cloud-cpp/plot2.py at data · dopiera/google-cloud-cpp · GitHub

Expand file tree

Breadcrumbs

plot2.py

Latest commit

History

Breadcrumbs

plot2.py

File metadata and controls