Quantum Development Kit (QDK) resources estimator
As the name implies, the ResourcesEstimator class estimates the resources required to run a given instance of a Q# operation on a quantum computer. It accomplishes this by running the quantum operation without actually simulating the state of a quantum computer; for this reason, it is able to estimate resources for Q# operations that use thousands of qubits, provided that the classical part of the code runs in a reasonable time.
The resources estimator is built on top of the Quantum trace simulator, which provides a richer set of metrics and tools to help debug Q# programs.
Invoking and running the resources estimator
The resources estimator can be thought of as another type of simulation target. As such, there exists a variety of methods with which to invoke this target, which are presented below. For additional details, you can also have a look at the Ways to run a Q# program.
Invoking the resources estimator from C#
As with other targets, you first create an instance of the ResourcesEstimator class and then pass it as the first parameter of an operation's Run method.
Note that, unlike the QuantumSimulator class, the ResourcesEstimator class does not implement the IDisposable interface, and thus you do not need to enclose it within a C# using statement.
using System;
using Microsoft.Quantum.Simulation.Simulators;
namespace Quantum.MyProgram
{
class Driver
{
static void Main(string[] args)
{
ResourcesEstimator estimator = new ResourcesEstimator();
MyOperation.Run(estimator).Wait();
Console.WriteLine(estimator.ToTSV());
}
}
}
As the example shows, ResourcesEstimator provides the ToTSV() method, which generates a table with tab-separated values (TSV). You can save the table to a file or display it to the console for analysis. The following is a sample output from the preceding program:
Metric Sum Max
CNOT 1000 1000
QubitClifford 1000 1000
R 0 0
Measure 4002 4002
T 0 0
Depth 0 0
Width 2 2
QubitCount 2 2
BorrowedWidth 0 0
For more information about the collected metrics, see the description of the Reported Metrics. The default configuration used for the ResourcesEstimator counts only T gates. Other configurations can be set by creating and customizing a QCTraceSimulator instance.
Note
A ResourcesEstimator instance does not reset its calculations on every run. If you use the same instance to run another operation, it aggregates the new results with the existing results in the Sum column. Additionally, the Max column will contain the highest value encountered so far for each metric. If you need to reset calculations between runs, create a new instance for every run.
Invoking the resources estimator from Python
Use the estimate_resources() method from the Python library with the imported Q# operation:
qubit_result = myOperation.estimate_resources()
Invoking the resources estimator from the command line
When running a Q# program from the command line, use the --simulator (or -s shortcut) parameter to specify the ResourcesEstimator target machine. The following command runs a program using the resources estimator:
dotnet run -s ResourcesEstimator
Invoking the resources estimator from Jupyter Notebooks
Use the IQ# magic command %estimate to run the Q# operation.
%estimate myOperation
Programmatically retrieving the estimated data
In addition to a TSV table, you can programmatically retrieve the resources estimated during the run via the Data property of the resources estimator. The Data property provides a System.DataTable instance with three columns: Metric, Sum, and Max, indexed by the metrics' names.
The following code shows how to retrieve and print the total number of QubitClifford, T and CNOT operations used by a Q# operation:
using System;
using Microsoft.Quantum.Simulation.Simulators;
namespace Quantum.MyProgram
{
class Driver
{
static void Main(string[] args)
{
ResourcesEstimator estimator = new ResourcesEstimator();
MyOperation.Run(estimator).Wait();
var data = estimator.Data;
Console.WriteLine($"QubitCliffords: {data.Rows.Find("QubitClifford")["Sum"]}");
Console.WriteLine($"Ts: {data.Rows.Find("T")["Sum"]}");
Console.WriteLine($"CNOTs: {data.Rows.Find("CNOT")["Sum"]}");
}
}
}
Metrics Reported
The resources estimator tracks the following metrics:
| Metric | Description |
|---|---|
| CNOT | The run count of CNOT operations (also known as Controlled Pauli X operations). |
| QubitClifford | The run count of any single qubit Clifford and Pauli operations. |
| Measure | The run count of any measurements. |
| R | The run count of any single-qubit rotations, excluding T, Clifford and Pauli operations. |
| T | The run count of T operations and their conjugates, including the T operations, T_x = H.T.H, and T_y = Hy.T.Hy. |
| Depth | Depth of the quantum circuit run by the Q# operation (see below). By default, the depth metric only counts T gates. For more details, see Depth Counter. |
| Width | Width of the quantum circuit run by the Q# operation (see below). By default, the depth metric only counts T gates. For more details, see Width Counter. |
| QubitCount | The lower bound for the maximum number of qubits allocated during the run of the Q# operation. This metric might not be compatible with Depth (see below). |
| BorrowedWidth | The maximum number of qubits borrowed inside the Q# operation. |
Depth, Width, and QubitCount
Reported Depth and Width estimates are compatible with each other. (Previously both numbers were achievable, but different circuits would be required for Depth and for Width.) Currently both metrics in this pair can be achieved by the same circuit at the same time.
The following metrics are reported:
Depth: For the root operation - time it takes to execute it assuming specific gate times. For operations called or subsequent operations - time difference between latest qubit availability time at the beginning and the end of the operation.
Width: For the root operation - number of qubits actually used to execute it (and operation it calls). For operations called or subsequent operations - how many more qubits were used in addition to the qubits already used at the beginning of the operation.
Please note, that reused qubits do not contribute to this number. For example, if a few qubits have been released before operation A starts, and all qubit demanded by this operation A (and operations called from A) were satisfied by reusing previously release qubits, the Width of operation A is reported as 0. Successfully borrowed qubits do not contribute to the Width either.
QubitCount: For the root operation - minimum number of qubits that would be sufficient to execute this operation (and operations called from it). For operations called or subsequent operations - minimum number of qubits that would be sufficient to execute this operation separately. This number doesn't include input qubits. It does include borrowed qubits.
Two modes of operation are supported. Mode is selected by setting QCTraceSimulatorConfiguration.OptimizeDepth.
OptimizeDepth=true: QubitManager is discouraged from qubit reuse and allocates new qubit every time it is asked for a qubit. For the root operation Depth becomes the minimum depth (lower bound). Compatible Width is reported for this depth (both can be achieved at the same time). Note that this width will likely be not optimal given this depth. QubitCount may be lower than Width for the root operation because it assumes reuse.
OptimizeDepth=false: QubitManager is encouraged to reuse qubits and will reuse released qubits before allocating new ones. For the root operation Width becomes the minimal width (lower bound). Compatible Depth is reported on which it can be achieved. QubitCount will be the same as Width for the root operation assuming no borrowing.
Providing the probability of measurement outcomes
You can use AssertMeasurementProbability operation from the Microsoft.Quantum.Diagnostics namespace namespace to provide information about the expected probability of a measurement operation. For more information, see Quantum Trace Simulator
See also
Povratne informacije
Pošalјite i prikažite povratne informacije za