Skip to content

GitLab

peregrine
peregrine
Commit a628d37c authored by Philipp Meyer's avatar Philipp Meyer
Browse files
Options

Add more protocols for benchmarking

parent 251d80c3
Hide whitespace changes
Inline Side-by-side
Showing with 599 additions and 6 deletions
benchmarks/generate_protocols.sh
View file @ a628d37c
......@@ -11,6 +11,11 @@ echo "generating majority protocol"
mkdir -p $benchmark_dir/$protocols_dir/majority
$executable majorityPP >$benchmark_dir/$protocols_dir/majority/majority_m1_.pp
# fast majority
echo "generating fast majority protocol"
mkdir -p $benchmark_dir/$protocols_dir/fast_majority
cp ../examples/fast_majority.pp $benchmark_dir/$protocols_dir/fast_majority/fast_majority_m1_.pp
# broadcast
echo "generating broadcast protocol"
mkdir -p $benchmark_dir/$protocols_dir/broadcast
......@@ -25,25 +30,37 @@ for l in 2 3 4 5 6 7 8 9 10 11 12 13 14 15; do
done
done
# modulo
echo "generating modulo protocols"
mkdir -p $benchmark_dir/$protocols_dir/modulo
# Remainder
echo "generating remainder protocols"
mkdir -p $benchmark_dir/$protocols_dir/remainder
for m in 2 3 4 5 6 7 8 9 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150; do
for c in 1; do
$executable moduloPP $m $c >$benchmark_dir/$protocols_dir/modulo/modulo_m${m}_c${c}_.pp
$executable moduloPP $m $c >$benchmark_dir/$protocols_dir/remainder/remainder_m${m}_c${c}_.pp
done
done
# flock of birds
# Flock of Birds from Chatzigianniks, Michail, Spirakis: Algorithmic verification of population protocols
echo "generating flockofbirds protocols"
mkdir -p $benchmark_dir/$protocols_dir/flockofbirds
for c in 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100; do
$executable flockOfBirdsPP $c >$benchmark_dir/$protocols_dir/flockofbirds/flockofbirds_c${c}_.pp
done
# new birds
# new birds from Clément, Delporte-Gallet, Fauconnier, Sighireanu: Guidelines for the verification of population protocols
echo "generating newbirds protocols"
mkdir -p $benchmark_dir/$protocols_dir/newbirds
for c in 10 20 25 30 40 50 75 100 125 150 175 200 225 250 275 300 325 350; do
$executable newBirdsPP $c >$benchmark_dir/$protocols_dir/newbirds/newbirds_c${c}_.pp
done
# Logarithmic Flock of Birds
mkdir -p $benchmark_dir/$protocols_dir/logflockofbirds
for c in 10 100 1000 10000 100000 1000000 10000000 100000000 1000000000; do
python python/execprotocol.py python/flock_log.py "{\"scheme\": { \"threshold\": { \"value\": ${c} } } }" >$benchmark_dir/$protocols_dir/logflockofbirds/logflockofbirds_c${c}_.pp
done
# Tower Flock of Birds
mkdir -p $benchmark_dir/$protocols_dir/towerflockofbirds
for c in 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100; do
python python/execprotocol.py python/flock_of_birds_-_tower.py "{\"scheme\": { \"c\": { \"value\": ${c} } } }" >$benchmark_dir/$protocols_dir/towerflockofbirds/towerflockofbirds_c${c}_.pp
done
benchmarks/python/avc.py 0 → 100644
View file @ a628d37c
# -*- coding: utf-8 -*-
params = {
'scheme': {
'm': {
'descr': "m",
'values': range(3, 22, 2),
'value': 3
},
'd': {
'descr': "d",
'values': range(1, 16),
'value': 1
}
}
}
def generateProtocol(params):
m = params["scheme"]["m"]["value"]
d = params["scheme"]["d"]["value"]
# Generate states
states = [("weak", ("+", 0)), ("weak", ("-", 0))]
for n in range(3, m + 1, 2):
states.append(("strong", n))
states.append(("strong", -n))
for n in range(1, d + 1):
states.append(("intermediate", ("+", n)))
states.append(("intermediate", ("-", n)))
# Helper functions
def strong(x):
return x[0] == "strong"
def intermediate(x):
return x[0] == "intermediate"
def weak(x):
return x[0] == "weak"
def value(x):
if strong(x):
return x[1]
elif intermediate(x):
return 1 if x[1][0] == "+" else -1
elif weak(x):
return 0
def weight(x):
return abs(value(x))
def sgn(x):
if strong(x) or intermediate(x):
return 1 if value(x) > 0 else -1
else:
return 1 if x[1][0] == "+" else -1
def phi(r):
if abs(r) > 1:
return ("strong", r)
else:
return ("intermediate", ("+" if r > 0 else "-", 1))
def r_down(k):
r = k if abs(k) % 2 == 1 else k - 1
return phi(r);
def r_up(k):
r = k if abs(k) % 2 == 1 else k + 1
return phi(r)
def shift_to_zero(x):
if intermediate(x):
j = x[1][1]
if j < d:
sign = "+" if value(x) == 1 else "-"
return ("intermediate", (sign, j + 1))
return x
def sign_to_zero(x):
sign = "+" if sgn(x) > 0 else "-"
return ("weak", (sign, 0))
def update(x, y):
if ((weight(x) > 0 and weight(y) > 1) or (weight(y) > 0 and weight(x) > 1)):
x_ = r_down((value(x) + value(y)) // 2)
y_ = r_up((value(x) + value(y)) // 2)
elif (weight(x) * weight(y) == 0 and value(x) + value(y) != 0):
if weight(x) != 0:
x_ = shift_to_zero(x)
y_ = sign_to_zero(x)
else:
y_ = shift_to_zero(y)
x_ = sign_to_zero(y)
elif ((intermediate(x) and x[1][1] == d and weight(y) and sgn(x) != sgn(y)) or
(intermediate(y) and y[1][1] == d and weight(x) and sgn(y) != sgn(x))):
x_ = ("weak", ("-", 0))
y_ = ("weak", ("+", 0))
else:
x_ = shift_to_zero(x)
y_ = shift_to_zero(y)
return (x_, y_)
def state_repr(x):
sign = "G" if sgn(x) > 0 else "B"
if strong(x):
return "{}{}".format(sign, weight(x))
elif intermediate(x):
return "{}1_{}".format(sign, x[1][1])
else:
return "{}0".format(sign)
def pair_repr(pair):
return list(map(state_repr, pair))
def generate_states():
return list(map(state_repr, states))
def generate_transitions():
transitions = []
for x in range(0, len(states)):
for y in range(x, len(states)):
p = states[x]
q = states[y]
pre = pair_repr((p, q))
post = pair_repr(update(p, q))
if not Utils.silent(pre, post):
transitions.append(Utils.transition(pre, post))
return transitions
def generate_initial_states():
return ["G{}".format(m), "B{}".format(m)]
def generate_true_states():
return [state_repr(q) for q in states if sgn(q) == 1]
def generate_predicate():
return "C[G{0}] >= C[B{0}]".format(m)
def generate_precondition():
return "C[G{0}] != C[B{0}]".format(m)
def generate_pescription():
return """This protocol is a fast and exact majority protocol. It computes
whether there are initially more agents voting yes (represented by
state Gm) than agents voting no (represented by state Bm). The protocol
does not handle ties. Parameters m and d control the running time
of the protocol. Described in Dan Alistarh, Rati Gelashvili,
Milan Vojnović. Fast and Exact Majority in Population Protocols.
PODC 2015."""
return {
"title": "Average and Conquer protocol",
"states": generate_states(),
"transitions": generate_transitions(),
"initialStates": generate_initial_states(),
"trueStates": generate_true_states(),
"predicate": generate_predicate(),
"precondition": generate_precondition(),
"description": generate_pescription()
}
benchmarks/python/broadcast.py 0 → 100644
View file @ a628d37c
# EDIT_WYSIWYG_ONLY
# -*- coding: utf-8 -*-
# This file has been created by Peregrine.
# Do not edit manually!
def generateProtocol(params):
return {
"title": "Broadcast protocol",
"states": ["yes", "no"],
"transitions": [Utils.transition(["yes", "no"], ["yes", "yes"])],
"initialStates": ["yes", "no"],
"trueStates": ["yes"],
"predicate": "C[yes] >= 1",
"description": """This protocol computes whether at least one agent is
initially in state yes."""
}
\ No newline at end of file
benchmarks/python/execprotocol.py 0 → 100644
View file @ a628d37c
#!/usr/bin/env python
# -*- coding: utf-8 -*-
import sys
import json
import re
class Utils:
@staticmethod
def transition(pre, post):
return {
"name": "{}, {} -> {}, {}".format(pre[0], pre[1], post[0], post[1]),
"pre": [pre[0], pre[1]],
"post": [post[0], post[1]]
}
@staticmethod
def silent(pre, post):
return ((pre[0] == post[0] and pre[1] == post[1]) or
(pre[0] == post[1] and pre[1] == post[0]))
n = len(sys.argv)
def to_string(p):
return "s" + re.sub(r'[^a-zA-Z0-9_]', r'_', str(p))
if (n != 2 and n != 3):
print("Usage: program [file path] [param dict]")
else:
script = sys.argv[1]
exec(open(script).read())
if (n == 3):
params = json.loads(sys.argv[2])
protocol = generateProtocol(params)
# Convert states to strings
for p in ["states", "initialStates", "trueStates"]:
protocol[p] = list(map(to_string, protocol[p]))
for i in range(len(protocol["transitions"])):
for p in ["pre", "post"]:
protocol["transitions"][i][p] = list(map(to_string, protocol["transitions"][i][p]))
protocol["predicate"] = re.sub(r'C\[([^]]*)\]', lambda match: to_string(match.group(1)), protocol["predicate"])
if "statesStyle" in protocol:
style = protocol["statesStyle"]
protocol["statesStyle"] = {str(p): style[p] for p in style}
print(json.dumps(protocol))
else:
try:
params
except NameError:
params = {}
print(json.dumps(params))
benchmarks/python/flock_log.py 0 → 100644
View file @ a628d37c
# -*- coding: utf-8 -*-
import math
params = {
"scheme": {
"threshold": {
"descr": "Threshold c",
"values": range(1, 201),
"value": 10
}
}
}
def generateProtocol(params):
c = params["scheme"]["threshold"]["value"]
i = int(math.log(c, 2))
states = [0] + [2**j for j in range(0, i + 1)]
cur_state = 2**i
for j in range(i-1, 0, -1):
if c & 2**j:
cur_state += 2**j
states.append(cur_state)
if c not in states:
states.append(c)
transitions = []
for x in range(len(states)):
for y in range(x, len(states)):
q1 = states[x]
q2 = states[y]
sum = q1 + q2
pre = (q1, q2)
post = None
if sum < c and q1 != 0 and q2 != 0 and sum in states:
post = (sum, 0)
elif sum >= c:
post = (c, c)
if (post is not None) and (not Utils.silent(pre, post)):
transitions.append(Utils.transition(pre, post))
style = {q: {} for q in states}
for q in states:
if q < c:
style[q]["size"] = 0.5 + 0.5 * (q + 1) / (c + 1)
style[q]["color"] = "rgb({}, {}, {})".format(105 + 128 * (c - q) / c, 30, 99)
style[0]["color"] = "rgb(175, 175, 175)"
return {
"title": "Logarithmic flock-of-birds protocol",
"states": states,
"transitions": transitions,
"initialStates": [1],
"trueStates": [c],
"predicate": "C[1] >= {}".format(c),
"description": """This protocol computes whether the population size
is at least c. The protocol has O(log c) states.""",
"statesStyle": style
}
benchmarks/python/flock_of_birds_-_tower.py 0 → 100644
View file @ a628d37c
# -*- coding: utf-8 -*-
params = {
"scheme": {
"c": {
"descr": "Threshold c",
"values": range(1, 201),
"value": 3
}
}
}
def generateProtocol(params):
c = params["scheme"]["c"]["value"]
states = range(1, c + 1)
transitions = []
for i in states:
for j in states:
if i < c and i == j:
transitions.append(Utils.transition((i, j), (i, j+1)))
elif (i == c or j == c) and (i != c or j != c):
transitions.append(Utils.transition((i, j), (c, c)))
return {
"title": "Flock-of-birds protocol (Tower)",
"states": states,
"transitions": transitions,
"initialStates": [1],
"trueStates": [c],
"predicate": "C[1] >= {}".format(c),
"description": """This protocol computes whether a population of
birds contains at least c sick birds, i.e. whether
C[1] >= c. Described in Dana Angluin, James
Aspnes, David Eisenstat, Eric Ruppert.
The Computational Power of Population Protocols.
"""
}
\ No newline at end of file
benchmarks/python/majority_basic.py 0 → 100644
View file @ a628d37c
# EDIT_WYSIWYG_ONLY
#-*- coding: utf-8 -*-
# This file has been created by Peregrine.
# Do not edit manually!
def generateProtocol(params):
return {
"title": "Majority voting protocol",
"states": ["Y", "N", "y", "n"],
"transitions": [Utils.transition(("Y", "N"), ("y", "n")),
Utils.transition(("Y", "n"), ("Y", "y")),
Utils.transition(("N", "y"), ("N", "n")),
Utils.transition(("y", "n"), ("y", "y"))],
"initialStates": ["Y", "N"],
"trueStates": ["Y", "y"],
"predicate": "C[Y] >= C[N]",
"description": """This protocol takes a majority vote. More precisely,
it computes whether there are initially more agents
in state Y than N.""",
"statesStyle": {"Y": {"size": 1.0},
"N": {"size": 1.0},
"y": {"size": 0.5},
"n": {"size": 0.5}}
}
\ No newline at end of file
benchmarks/python/remainder.py 0 → 100644
View file @ a628d37c
# -*- coding: utf-8 -*-
params = {
"scheme": {
"remainder": {
"descr": "Remainder c",
"values": range(0, 101),
"value": 1
},
"modulo": {
"descr": "Modulo m",
"values": range(2, 101),
"value": 2
}
}
}
def generateProtocol(params):
c = params["scheme"]["remainder"]["value"]
m = params["scheme"]["modulo"]["value"]
# Helper functions
def label(q):
if q is True:
return "y";
elif q is False:
return "f";
else:
return q
# Generate states
numeric = range(m)
boolean = [False, True]
states = numeric + boolean
# Generate transitions
transitions = []
for i in range(len(numeric)):
p = numeric[i]
for j in range(i, len(numeric)):
q = numeric[j]
pre = map(label, (p, q))
post = map(label, ((p + q) % m, ((p + q) % m) == (c % m)))
if not Utils.silent(pre, post):
transitions.append(Utils.transition(pre, post))
for j in boolean:
q = boolean[j]
pre = map(label, (p, q))
post = map(label, (p, p == (c % m)))
if not Utils.silent(pre, post):
transitions.append(Utils.transition(pre, post))
# Generate predicate
expr = ["{0}*C[{0}]".format(q) for q in numeric]
predicate = "{} %=_{} {}".format(" + ".join(expr), m, c)
# Generate style
style = {label(q): {} for q in states}
for q in numeric:
style[label(q)]["size"] = 0.5 + 0.5 * q / (m - 1)
style[label(True)]["size"] = 0.5
style[label(False)]["size"] = 0.5
style[label(True)]["color"] = "blue"
style[label(False)]["color"] = "red"
return {
"title": "Remainder protocol",
"states": map(label, states),
"transitions": transitions,
"initialStates": map(label, numeric),
"trueStates": map(label, [c % m, True]),
"predicate": predicate,
"description": """This protocol tests whether
0·C[0] + 1·C[1] + ... + (m-1)·C[m-1] ≡ c (mod m).""",
"statesStyle": style
}
benchmarks/python/threshold.py 0 → 100644
View file @ a628d37c
# -*- coding: utf-8 -*-
params = {
"scheme": {
"minCoeff": {
"descr": "Min. coeff. a",
"values": range(-30, 31),
"value": -1
},
"maxCoeff": {
"descr": "Max. coeff. b",
"values": range(-30, 31),
"value": 1
},
"threshold": {
"descr": "Threshold c",
"values": range(1, 101),
"value": 2
}
}
}
def generateProtocol(params):
a = params["scheme"]["minCoeff"]["value"]
b = params["scheme"]["maxCoeff"]["value"]
c = params["scheme"]["threshold"]["value"]
s = max(abs(a), abs(b), abs(c) + 1)
# Helper functions
def is_leader(q):
return (q[0] == 1)
def output(q):
return (q[1] == 1)
def value(q):
return q[2]
def f(m, n):
return max(-s, min(s, m + n))
def g(m, n):
return m + n - f(m, n)
def o(m, n):
return 1 if f(m, n) < c else 0
# Generate states
states = [(x, y, u) for u in range(-s, s + 1)
for x in [0, 1] for y in [0, 1]]
# Generate transitions
transitions = []
for i in range(len(states)):
for j in range(i, len(states)):
p = states[i]
q = states[j]
if is_leader(p) or is_leader(q):
m = value(p)
n = value(q)
pre = (p, q)
post = ((1, o(m, n), f(m, n)), (0, o(m, n), g(m, n)))
t = Utils.transition(pre, post)
if (not Utils.silent(pre, post) and t not in transitions):
transitions.append(t)
# Generate initial states
initial_states = [q for q in states if is_leader(q) and
value(q) >= a and value(q) <= b and
output(q) == (value(q) < c)]
# Generate true states
true_states = [q for q in states if output(q)]
# Generate predicate
expr = ["{}*C[{}]".format(value(q), q) for q in initial_states]
predicate = "{} < {}".format(" + ".join(expr), c)
# Generate description
description = """This protocol evaluates the linear inequality
a·x_a + ... + b·x_b < c. Each state of the protocol
is of the form (l, o, v) where l is a bit indicating
whether the agent is active, o is the output of the
agent, and v is an integer indicating the value of
the agent. Described in Dana Angluin,
James Aspnes, Zoë Diamadi, Michael J. Fischer and
René Peralta. Computation in Networks of Passively
Mobile Finite-State Sensors. PODC 2004."""
# Generate style
style = {q: {} for q in states}
for q in states:
scale = (value(q) * 0.5 / s) + 0.5
style[q]["size"] = 1.0 if is_leader(q) else 0.5
if output(q):
style[q]["color"] = "rgb({}, {}, {})".format(63, int(81 + 200 * (1 - scale)),
int(100 + 81 * scale))
else:
style[q]["color"] = "rgb({}, {}, {})".format(int(105 + 128 * scale),
int(30 + 180 * (1 - scale)), 99)