Source code for qrisp.interface.batched_backend
"""
********************************************************************************
* Copyright (c) 2025 the Qrisp authors
*
* This program and the accompanying materials are made available under the
* terms of the Eclipse Public License 2.0 which is available at
* http://www.eclipse.org/legal/epl-2.0.
*
* This Source Code may also be made available under the following Secondary
* Licenses when the conditions for such availability set forth in the Eclipse
* Public License, v. 2.0 are satisfied: GNU General Public License, version 2
* with the GNU Classpath Exception which is
* available at https://www.gnu.org/software/classpath/license.html.
*
* SPDX-License-Identifier: EPL-2.0 OR GPL-2.0 WITH Classpath-exception-2.0
********************************************************************************
"""
import time
import threading
from qrisp.interface import VirtualBackend
[docs]
class BatchedBackend(VirtualBackend):
"""
This class tackles the problem that many physical backends have a high-overhead
regarding individual circuit execution. This overhead typically comes
from finite network latency, authentication procedures, compilation steps etc.
Typically this overhead is remedied through supporting the execution of
batches of circuits, which however doesn't really fit that well into the Qrisp
programming model, which shields the user from handling individual circuits
and automatically decodes the measurement results into human readable labels.
In order to bridge these worlds and still allow automatic decoding, the
``BatchedBackend`` allows Qrisp users to evaluate measurements from a
multi-threading perspective. The idea is here that the circuit
batch is collected through several threads, which each execute Qrisp code
until a individual backend call is required. This backend call is then saved
until the batch is complete. The batch can then be sent through the ``.dispatch``
method, which resumes each thread to execute the post-processing logic.
.. note::
Calling the ``.run`` method of a BatchedBackend from the
main thread will automatically dispatch all queries
(including the query set up by the main thread).
Parameters
----------
batch_run_func : function
A function that recieves a list of tuples in the form
list[tuple[QuantumCircuit, int]], which represents the quantum circuits
and the corresponding shots to execute on the backend. It should return
a list of dictionaries, where each dictionary corresponds to the measurement
results of the appropriate backend call.
Examples
--------
We set up a BatchedBackend, which sequentially executes the QuantumCircuits
on the Qrisp simulator.
::
from qrisp import *
from qrisp.interface import BatchedBackend
def run_func_batch(batch):
# Parameters
# ----------
# batch : list[tuple[QuantumCircuit, int]]
# The circuit and shot batch indicating the backend queries.
# Returns
# -------
# results : list[dict[string, int]]
# The list of results.
results = []
for i in range(len(batch)):
qc = batch[i][0]
shots = batch[i][1]
results.append(qc.run(shots = shots))
return results
# Set up batched backend
bb = BatchedBackend(run_func_batch)
Create some backend calls
::
a = QuantumFloat(4)
b = QuantumFloat(3)
a[:] = 1
b[:] = 2
c = a + b
d = QuantumFloat(4)
e = QuantumFloat(3)
d[:] = 2
e[:] = 3
f = d + e
Create threads
::
import threading
results = []
def eval_measurement(qv):
results.append(qv.get_measurement(backend = bb))
thread_0 = threading.Thread(target = eval_measurement, args = (c,))
thread_1 = threading.Thread(target = eval_measurement, args = (f,))
Start the threads and subsequently dispatch the batch.
::
# Start the threads
thread_0.start()
thread_1.start()
# Call the dispatch routine
# The min_calls keyword will make it wait
# until the batch has a size of 2
bb.dispatch(min_calls = 2)
# Wait for the threads to join
thread_0.join()
thread_1.join()
# Inspect the results
print(results)
This is automated by the :meth:`batched_measurement <qrisp.batched_measurement>`:
>>> batched_measurement([c,f], backend=bb)
[{3: 1.0}, {5: 1.0}]
"""
def __init__(self, batch_run_func):
self.batch_run_func = batch_run_func
self.batch = []
self.results_available = False
self.results = {}
def run(self, qc, shots):
self.batch.append((qc, shots))
if threading.current_thread() is threading.main_thread():
dispatching_thread = threading.Thread(target = self.dispatch)
dispatching_thread.start()
while not self.results_available:
time.sleep(0.01)
result = self.results[id(qc)]
del self.results[id(qc)]
if threading.current_thread() is threading.main_thread():
dispatching_thread.join()
return result
def dispatch(self, min_calls = 0):
"""
This method dispatches all collected queries and
subsequently resumes their threads.
Parameters
----------
min_calls : int, optional
If specified, the dispatch will be delayed until that
many queries have been collected. The default is 0.
"""
while len(self.batch) < min_calls:
time.sleep(0.01)
run_func_results = self.batch_run_func(self.batch)
self.results = {id(self.batch[i][0]) : run_func_results[i] for i in range(len(self.batch))}
self.batch = []
self.results_available = True
while len(self.results):
time.sleep(0.01)
self.results_available = False
