def convolved_table_list(self, tab1, tab2, tab3):
f_tab1a = bootstrap.ConvolvedBlockTable(tab1)
f_tab1s = bootstrap.ConvolvedBlockTable(tab1, symmetric=True)
f_tab2a = bootstrap.ConvolvedBlockTable(tab2)
f_tab2s = bootstrap.ConvolvedBlockTable(tab2, symmetric=True)
f_tab3 = bootstrap.ConvolvedBlockTable(tab3)
return [f_tab1a, f_tab1s, f_tab2a, f_tab2s, f_tab3]
def determine_grid(self):
#key = [self.inputs['dim'], self.inputs['kmax'], self.inputs['lmax'], self.inputs['mmax'], self.inputs['nmax']]
key = list(self.inputs.values())
tab1 = bootstrap.ConformalBlockTable(*key)
tab2 = bootstrap.ConvolvedBlockTable(tab1)
# Instantiate a Grid object with appropriate input values.
grid = Grid(*key, [], [])
for sig in self.sig_values:
for eps in self.eps_values:
sdp = bootstrap.SDP(sig, tab2)
sdp.set_bound(0, float(self.gap))
sdp.add_point(0, eps)
result = sdp.iterate()
if result:
grid.allowed_points.append((sig, eps))
else:
grid.disallowed_points.append((sig, eps))
# Now append this grid object to the IsingGap table.
# Note we will need to implement a look up table to retrieve desired data.
self.table.append(grid)
def iterate_k_max(k_range):
bootstrap.cutoff = 1e-10
dim = 3
l_max = 15
n_max = 4
m_max = 2
for k in k_range:
tab1 = bootstrap.ConformalBlockTable(dim, k_max, l_max, m_max, n_max)
tab2 = bootstrap.ConvolvedBlockTable(tab1)
plot_grid(dim, tab2, sig_set, eps_set)
def iterate_parameter(self, par, par_range):
if type(par_range)==int:
par_range=[par_range]
start_time=time.time()
start_cpu=time.clock()
for x in par_range:
self.inputs[par]=x
#Might make more sense to move tab1 and tab2 calculation to plot_grid()?
tab1=bootstrap.ConformalBlockTable(self.inputs['dim'],self.inputs['kmax'],self.inputs['lmax'],self.inputs['mmax'],self.inputs['nmax'])
tab2=bootstrap.ConvolvedBlockTable(tab1)
self.plot_grid(par,x,tab2)
end_time=time.time()
end_cpu=time.clock()
run_time=time.strftime("%H:%M:%S",time.gmtime(end_time-start_time))
cpu_time=time.strftime("%H:%M:%S",time.gmtime(end_cpu-start_cpu))
print("Run time "+run_time, "CPU time "+cpu_time)
def determine_grid(self, key):
#if self.get_grid_index(key) != -1:
start_time=time.time()
start_cpu=time.clock()
tab1 = bootstrap.ConformalBlockTable(self.dim, *key)
tab2 = bootstrap.ConvolvedBlockTable(tab1)
# Instantiate a Grid object with appropriate input values.
# grid=Grid(*key, [], [])
grid = Grid(*(key + [[], [], 0, 0]))
for sig in self.sig_values:
for eps in self.eps_values:
sdp = bootstrap.SDP(sig, tab2)
# SDPB will naturally try to parallelize across 4 cores / slots.
# To prevent this, we set its 'maxThreads' option to 1.
# See 'common.py' for the list of SDPB option strings, as well as their default values.
sdp.set_option("maxThreads", 1)
sdp.set_bound(0, float(self.gap))
sdp.add_point(0, eps)
result = sdp.iterate()
if result:
grid.allowed_points.append((sig, eps))
else:
grid.disallowed_points.append((sig, eps))
# Now append this grid object to the IsingGap table.
# Note we will need to implement a look up table to retrieve desired data.
end_time=time.time()
end_cpu=time.clock()
run_time=end_time-start_time
cpu_time=end_cpu-start_cpu
run_time = datetime.timedelta(seconds = int(end_time - start_time))
cpu_time = datetime.timedelta(seconds = int(end_cpu - start_cpu))
grid.run_time = run_time
grid.cpu_time = cpu_time
self.table.append(grid)
self.save_grid(grid, self.name)
bootstrap.cutoff = 0
reference_sdp = None
for i in range(len(row[0])):
sig = row[0][i]
eps = row[1][i]
global start_time
start_time = time.time()
global start_cpu
start_cpu = time.clock()
g_tab1 = bootstrap.ConformalBlockTable(3, *key)
g_tab2 = bootstrap.ConformalBlockTable(
3, *(key + [eps - sig, sig - eps, "odd_spins = True"]))
g_tab3 = bootstrap.ConformalBlockTable(
3, *(key + [sig - eps, sig - eps, "odd_spins = True"]))
f_tab1a = bootstrap.ConvolvedBlockTable(g_tab1)
f_tab1s = bootstrap.ConvolvedBlockTable(g_tab1, symmetric=True)
f_tab2a = bootstrap.ConvolvedBlockTable(g_tab2)
f_tab2s = bootstrap.ConvolvedBlockTable(g_tab2, symmetric=True)
f_tab3 = bootstrap.ConvolvedBlockTable(g_tab3)
tab_list = [f_tab1a, f_tab1s, f_tab2a, f_tab2s, f_tab3]
global now
global now_clock
global CB_time
global CB_cpu
now = time.time()
now_clock = time.clock()
CB_time = datetime.timedelta(seconds=int(now - start_time))
CB_cpu = datetime.timedelta(seconds=int(now_clock - start_cpu))
print(
"The calculation of the required conformal blocks has successfully completed."
l_max = 15
n_max = 4
m_max = 2
def find_bounds(table1, table2, lower, upper, tol, channel):
dim_phi = 0.5
x = []
y = []
while dim_phi < 0.6:
sdp = bootstrap.SDP(dim_phi, table2)
result = sdp.bisect(lower, upper, tol, channel)
x.append(dim_phi)
y.append(result)
dim_phi += 0.002
plt.plot(x, y)
tab1 = bootstrap.ConformalBlockTable(dim, k_max, l_max, m_max, n_max)
tab2 = bootstrap.ConvolvedBlockTable(tab1)
l = 0.9
u = 1.7
t = 0.01
c = 0
find_bounds(tab1, tab2, l, u, t, c)
plt.xlabel('$\Delta_{\sigma}$')
plt.ylabel('$\Delta_{\epsilon}$')
plt.show()
cprint("Finding basic bound at external dimension " + str(dim_phi) + "...")
# Spatial dimension.
dim = 3
# Dictates the number of poles to keep and therefore the accuracy of a conformal block.
k_max = 20
# Says that conformal blocks for spin-0 up to and including spin-14 should be computed.
l_max = 14
# Conformal blocks are functions of (a, b) and as many derivatives of each should be kept for strong bounds.
# This says to keep derivatives up to fourth order in b.
n_max = 4
# For a given n, this states how many a derivatives should be included beyond 2 * (n - n_max).
m_max = 2
# Generates the table.
table1 = bootstrap.ConformalBlockTable(dim, k_max, l_max, m_max, n_max)
# Computes the convolution.
table2 = bootstrap.ConvolvedBlockTable(table1)
# Sets up a semidefinite program that we can use to study this.
sdp = bootstrap.SDP(dim_phi, table2)
# We think it is perfectly find for all internal scalars coupling to our external one to have dimension above 0.7.
lower = 0.7
# Conversely, we think it is a problem for crossing symmetry if they all have dimension above 1.7.
upper = 1.7
# The boundary between these regions will be found within an error of 0.01.
tol = 0.01
# The 0.7 and 1.7 are our guesses for scalars, not some other type of operator.
channel = 0
# Calls SDPB to compute the bound.
result = sdp.bisect(lower, upper, tol, channel)
cprint(
"If crossing symmetry and unitarity hold, the maximum gap we can have for Z2-even scalars is: "
+ str(result))
def determine_row(self, key, row):
# Will be called with a given row_lists[i]
# Use generate_rows() method to build row_lists.
# row = row_lists[row_index]
reference_sdp = None
blocks_initiated = False
for i in range(len(row[0])):
phi = eval_mpfr(row[0][i], bootstrap.prec)
sing = eval_mpfr(row[1][i], bootstrap.prec)
# phi_sing = eval_mpfr(phi - sing, bootstrap.prec)
# sing_phi = eval_mpfr(sing - phi, bootstrap.prec)
start = time.time()
start_cpu = time.clock()
if blocks_initiated == False:
g_tab1 = bootstrap.ConformalBlockTable(
self.dim, *(key + [0, 0, "odd_spins = True"]))
g_tab2 = bootstrap.ConformalBlockTable(
self.dim,
*(key + [phi - sing, phi - sing, "odd_spins = True"]))
g_tab3 = bootstrap.ConformalBlockTable(
self.dim,
*(key + [sing - phi, phi - sing, "odd_spins = True"]))
f_tab1a = bootstrap.ConvolvedBlockTable(g_tab1)
f_tab1s = bootstrap.ConvolvedBlockTable(g_tab1, symmetric=True)
f_tab2a = bootstrap.ConvolvedBlockTable(g_tab2)
f_tab3a = bootstrap.ConvolvedBlockTable(g_tab3)
f_tab3s = bootstrap.ConvolvedBlockTable(g_tab3, symmetric=True)
tab_list = [f_tab1a, f_tab1s, f_tab2a, f_tab3a, f_tab3s]
for tab in [g_tab1, g_tab2, g_tab3]:
# tab.dump("tab_" + str(tab.delta_12) + "_" + str(tab.delta_34))
del tab
blocks_initiated = True
max_dimension = 0
for tab in tab_list:
max_dimension = max(max_dimension, len(tab.table[0].vector))
print("kmax should be around " + max_dimension.__str__() + ".")
dimension = (5 * len(f_tab1a.table[0].vector)) + (
2 * len(f_tab1s.table[0].vector))
bootstrap.cb_end = time.time()
bootstrap.cb_end_cpu = time.clock()
cb_time = datetime.timedelta(seconds=int(bootstrap.cb_end - start))
cb_cpu = datetime.timedelta(seconds=int(bootstrap.cb_end_cpu -
start_cpu))
print(
"The calculation of the required conformal blocks has successfully completed."
)
print("Time taken: " + str(cb_time))
print("CPU_time: " + str(cb_cpu))
if reference_sdp == None:
sdp = bootstrap.SDP([phi, sing],
tab_list,
vector_types=self.info)
reference_sdp = sdp
else:
sdp = bootstrap.SDP([phi, sing],
tab_list,
vector_types=self.info,
prototype=reference_sdp)
# We assume the continuum in both even vector and even singlet sectors begins at the dimension=3.
sdp.set_bound([0, 0], self.dim)
sdp.set_bound([0, 3], self.dim)
# Except for the two lowest dimension scalar operators in each sector.
sdp.add_point([0, 0], sing)
sdp.add_point([0, 3], phi)
sdp.set_option("maxThreads", 16)
sdp.set_option("dualErrorThreshold", 1e-15)
sdp.set_option("maxIterations", 1000)
# Run the SDP to determine if the current operator spectrum is permissable.
print("Testing point " + "(" + phi.__str__() + ", " +
sing.__str__() + ")" + " with " + dimension.__str__() +
" components.")
result = sdp.iterate()
end = time.time()
end_cpu = time.clock()
sdp_time = datetime.timedelta(seconds=int(end - bootstrap.xml_end))
sdp_cpu = datetime.timedelta(seconds=int(end_cpu -
bootstrap.xml_end_cpu))
run_time = datetime.timedelta(seconds=int(end - start))
cpu_time = datetime.timedelta(seconds=int(end_cpu - start_cpu))
print("The SDP has finished running.")
print("Time taken: " + str(sdp_time))
print("CPU_time: " + str(sdp_cpu))
print(
"See point file for more information. Check the times are consistent."
)
point = Point(*([phi, sing] + key + [
components, max_dimension, result, run_time, cpu_time, cb_time,
cb_cpu, bootstrap.xml_time, bootstrap.xml_cpu, sdp_time,
sdp_cpu
]))
self.point_table.append(point)
point.save(self.point_file)
