154 lines
5.6 KiB
Python
154 lines
5.6 KiB
Python
# vim: ft=python
|
|
|
|
from Compiler import types
|
|
|
|
from Compiler.util import *
|
|
|
|
from Compiler.oram import OptimalORAM
|
|
|
|
from Compiler.library import for_range, do_while, time, if_, print_ln, crash, print_str, break_loop
|
|
from Compiler.gs import OMatrix, OStack
|
|
|
|
|
|
class Matchmaker:
|
|
"""
|
|
Based on Matchmaker from Compiler/gs.py in MP-SPDZ, copyright (c) 2023,
|
|
Commonwealth Scientific and Industrial Research Organisation (CSIRO)
|
|
ABN 41 687 119 230, published under the BSD 3-Clause Licence
|
|
"""
|
|
|
|
def pair(self, patient, therapist, for_real):
|
|
self.paired_therapists.access(patient, therapist, for_real)
|
|
self.paired_patients.access(therapist, patient, for_real)
|
|
|
|
def unpair(self, patient, therapist, for_real):
|
|
self.paired_therapists.delete(patient, for_real)
|
|
self.paired_patients.delete(therapist, for_real)
|
|
self.unpaired.append(patient, for_real)
|
|
|
|
def request_therapist(self, patient, therapist, for_real):
|
|
experience = self.t_exps[therapist][0]
|
|
(old_patient,), free = self.paired_patients.read(therapist) # patient paired to therapist
|
|
paired = 1 - free
|
|
rank_patient = self.p_cases[patient][0]
|
|
rank_old_patient = self.p_cases[old_patient][0] * paired
|
|
matches_exp = self.int_type(rank_patient) == self.int_type(experience)
|
|
same_as_old = self.int_type(rank_patient) != self.int_type(rank_old_patient)
|
|
leaving = paired * matches_exp * same_as_old
|
|
|
|
print_str('therapist: %s, patient: %s, old patient: %s, ',
|
|
*(x.reveal() for x in (therapist, patient, old_patient)))
|
|
print_ln('rank patient: %s, rank old patient: %s, paired: %s, leaving: %s',
|
|
*(x.reveal() for x in
|
|
(rank_patient, rank_old_patient, paired, leaving)))
|
|
|
|
self.unpair(old_patient, therapist, paired * leaving * for_real)
|
|
self.pair(patient, therapist, (1 - (paired * (1 - leaving))) * for_real)
|
|
self.unpaired.append(patient, paired * (1 - leaving) * for_real)
|
|
|
|
def match(self, n_loops=None):
|
|
if n_loops is None or n_loops > self.N * self.M:
|
|
loop = do_while
|
|
init_rounds = self.N
|
|
else:
|
|
loop = for_range(n_loops)
|
|
init_rounds = n_loops / self.M
|
|
|
|
self.paired_therapists = \
|
|
self.oram_type(self.N, entry_size=log2(self.N),
|
|
init_rounds=0, value_type=self.basic_type)
|
|
self.paired_patients = \
|
|
self.oram_type(self.N, entry_size=log2(self.N),
|
|
init_rounds=0, value_type=self.basic_type)
|
|
self.unpaired = OStack(self.N, oram_type=self.oram_type,
|
|
int_type=self.int_type)
|
|
|
|
@for_range(init_rounds)
|
|
def _(i):
|
|
self.unpaired.append(i)
|
|
rounds = types.MemValue(types.regint(0))
|
|
|
|
@loop
|
|
def _(i=None):
|
|
rounds.iadd(1)
|
|
time()
|
|
patient = self.unpaired.pop()
|
|
pref = self.int_type(p_cases[patient][0]) # patient suffers from pref
|
|
therapist = types.MemValue(self.int_type(0))
|
|
|
|
# Get index of first free therapist who has experience with pref
|
|
@for_range(self.N)
|
|
def _(i):
|
|
(_,), free = self.paired_patients.read(i)
|
|
@if_(((pref == t_exps[i][0]) * free).reveal())
|
|
def _():
|
|
therapist.write(self.int_type(i))
|
|
break_loop()
|
|
|
|
@if_((i == self.N).reveal())
|
|
def _():
|
|
print_ln('run out of acceptable therapists')
|
|
crash()
|
|
|
|
self.request_therapist(patient, therapist.read(), True)
|
|
print_ln('patient: %s, pref: %s, left: %s',
|
|
*(x.reveal() for x in (patient, pref, self.unpaired.size)))
|
|
return types.regint((self.unpaired.size > 0).reveal())
|
|
|
|
print_ln('%s rounds', rounds)
|
|
print_ln('\nPRINTING PAIRS\n')
|
|
|
|
@for_range(init_rounds)
|
|
def _(i):
|
|
print_ln('patient %s : therapist %s', i, self.paired_therapists[i].reveal())
|
|
|
|
def __init__(self, p_cases, t_exps, N, M=None, reverse=False,
|
|
oram_type=OptimalORAM, int_type=types.sint):
|
|
self.N = N
|
|
self.M = N if M is None else M
|
|
self.p_cases = p_cases
|
|
self.t_exps = t_exps
|
|
self.reverse = reverse
|
|
self.oram_type = oram_type
|
|
self.int_type = int_type
|
|
self.basic_type = int_type.basic_type
|
|
|
|
|
|
# Constants
|
|
PLAYERS = 3
|
|
MATCHING_SIZE = 50
|
|
|
|
p_shares = Matrix(rows=PLAYERS, columns=MATCHING_SIZE, value_type=types.sint)
|
|
t_shares = Matrix(rows=PLAYERS, columns=MATCHING_SIZE, value_type=types.sint)
|
|
|
|
# Fill data from players into the matrices
|
|
@for_range(PLAYERS)
|
|
def _(i):
|
|
@for_range(2 * MATCHING_SIZE)
|
|
def _(j):
|
|
index = j % MATCHING_SIZE
|
|
typ = sint.get_input_from(i)
|
|
|
|
@if_((typ == -100).reveal())
|
|
def _():
|
|
p_shares[i][index] = sint.get_input_from(i)
|
|
|
|
@if_((typ == -200).reveal())
|
|
def _():
|
|
t_shares[i][index] = sint.get_input_from(i)
|
|
|
|
# Add entire columns together to recreate secret-shared input in ORAM
|
|
p_cases = OMatrix(N=MATCHING_SIZE, M=1, oram_type=OptimalORAM, int_type=types.sint)
|
|
t_exps = OMatrix(N=MATCHING_SIZE, M=1, oram_type=OptimalORAM, int_type=types.sint)
|
|
|
|
@for_range(MATCHING_SIZE)
|
|
def _(i):
|
|
p_cases[i][0] = sum(p_shares.get_column(i))
|
|
print_ln('patient %s suffers from case %s', i, p_cases[i][0].reveal())
|
|
t_exps[i][0] = sum(t_shares.get_column(i))
|
|
print_ln('therapist %s is experienced with case %s', i, t_exps[i][0].reveal())
|
|
|
|
# Run algorithm
|
|
mm = Matchmaker(p_cases, t_exps, N=MATCHING_SIZE, M=1)
|
|
mm.match()
|