-
Notifications
You must be signed in to change notification settings - Fork 1k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Classical Control: Classical Data Registers #3231
Comments
Should this be added to the 2021 roadmap given that it's the precondition for the above 3 linked issues? |
After further consideration, I don't think these concerns are relevant to Cirq outside of the built-in simulators:
Cirq constructs only really matter during translation from and to the Cirq format; any handling of measurement results during circuit execution is the sole responsibility of the device / simulator. |
Adds a `ClassicalDataStore` class so we can keep track of which qubits are associated to which measurements. Closes #3232. Initially this was created as part 14 (of 14) of https://tinyurl.com/cirq-feedforward to enable qudits in classical conditions, by storing and using dimensions of the measured qubits when calculating the integer value of each measurement when resolving sympy expressions. However it may have broader applicability. This approach also sets us up to more easily add different types of measurements (#3233, #4274). It will also ease the path to #3002 and #4449., as we can eventually pass this into `Result` rather than the raw `log_of_measurement_results` dictionary. (The return type of `_run` will have to be changed to `Sequence[C;assicalDataStoreReader]`. Related: #887, #3231 (open question @95-martin-orion whether this closes those or not) This PR contains a `ClassicalDataStoreReader` and `ClassicalDataStoreBase` parent "interface" for the `ClassicalDataStore` class as well. This will allow us to swap in different representations that may have different performance characteristics. See #3808 for an example use case. This could be done by adding an optional `ClassicalDataStore` factory method argument to the `SimulatorBase` initializer, or separately to sampler classes. (Note this is an alternative to #4778 for supporting qudits in sympy classical control expressions, as discussed here: https://github.com/quantumlib/Cirq/pull/4778/files#r774816995. The other PR was simpler and less invasive, but a bit hacky. I felt even though bigger, this seemed like the better approach and especially fits better with our future direction, and closed the other one). **Breaking Changes**: 1. The abstract method `SimulatorBase._create_partial_act_on_args` argument `log_of_measurement_results: Dict` has been changed to `classical_data: ClassicalData`. Any third-party simulators that inherit `SimulatorBase` will need to update their implementation accordingly. 2. The abstract base class `ActOnArgs.__init__` argument `log_of_measurement_results: Dict` is now copied before use. For users that depend on the pass-by-reference semantics (this should be rare), they can use the new `classical_data: ClassicalData` argument instead, which is pass-by-reference.
@95-martin-orion is this before or after 1.0? |
This is functionally complete as of #4781. |
…mlib#4781) Adds a `ClassicalDataStore` class so we can keep track of which qubits are associated to which measurements. Closes quantumlib#3232. Initially this was created as part 14 (of 14) of https://tinyurl.com/cirq-feedforward to enable qudits in classical conditions, by storing and using dimensions of the measured qubits when calculating the integer value of each measurement when resolving sympy expressions. However it may have broader applicability. This approach also sets us up to more easily add different types of measurements (quantumlib#3233, quantumlib#4274). It will also ease the path to quantumlib#3002 and quantumlib#4449., as we can eventually pass this into `Result` rather than the raw `log_of_measurement_results` dictionary. (The return type of `_run` will have to be changed to `Sequence[C;assicalDataStoreReader]`. Related: quantumlib#887, quantumlib#3231 (open question @95-martin-orion whether this closes those or not) This PR contains a `ClassicalDataStoreReader` and `ClassicalDataStoreBase` parent "interface" for the `ClassicalDataStore` class as well. This will allow us to swap in different representations that may have different performance characteristics. See quantumlib#3808 for an example use case. This could be done by adding an optional `ClassicalDataStore` factory method argument to the `SimulatorBase` initializer, or separately to sampler classes. (Note this is an alternative to quantumlib#4778 for supporting qudits in sympy classical control expressions, as discussed here: https://github.com/quantumlib/Cirq/pull/4778/files#r774816995. The other PR was simpler and less invasive, but a bit hacky. I felt even though bigger, this seemed like the better approach and especially fits better with our future direction, and closed the other one). **Breaking Changes**: 1. The abstract method `SimulatorBase._create_partial_act_on_args` argument `log_of_measurement_results: Dict` has been changed to `classical_data: ClassicalData`. Any third-party simulators that inherit `SimulatorBase` will need to update their implementation accordingly. 2. The abstract base class `ActOnArgs.__init__` argument `log_of_measurement_results: Dict` is now copied before use. For users that depend on the pass-by-reference semantics (this should be rare), they can use the new `classical_data: ClassicalData` argument instead, which is pass-by-reference.
…mlib#4781) Adds a `ClassicalDataStore` class so we can keep track of which qubits are associated to which measurements. Closes quantumlib#3232. Initially this was created as part 14 (of 14) of https://tinyurl.com/cirq-feedforward to enable qudits in classical conditions, by storing and using dimensions of the measured qubits when calculating the integer value of each measurement when resolving sympy expressions. However it may have broader applicability. This approach also sets us up to more easily add different types of measurements (quantumlib#3233, quantumlib#4274). It will also ease the path to quantumlib#3002 and quantumlib#4449., as we can eventually pass this into `Result` rather than the raw `log_of_measurement_results` dictionary. (The return type of `_run` will have to be changed to `Sequence[C;assicalDataStoreReader]`. Related: quantumlib#887, quantumlib#3231 (open question @95-martin-orion whether this closes those or not) This PR contains a `ClassicalDataStoreReader` and `ClassicalDataStoreBase` parent "interface" for the `ClassicalDataStore` class as well. This will allow us to swap in different representations that may have different performance characteristics. See quantumlib#3808 for an example use case. This could be done by adding an optional `ClassicalDataStore` factory method argument to the `SimulatorBase` initializer, or separately to sampler classes. (Note this is an alternative to quantumlib#4778 for supporting qudits in sympy classical control expressions, as discussed here: https://github.com/quantumlib/Cirq/pull/4778/files#r774816995. The other PR was simpler and less invasive, but a bit hacky. I felt even though bigger, this seemed like the better approach and especially fits better with our future direction, and closed the other one). **Breaking Changes**: 1. The abstract method `SimulatorBase._create_partial_act_on_args` argument `log_of_measurement_results: Dict` has been changed to `classical_data: ClassicalData`. Any third-party simulators that inherit `SimulatorBase` will need to update their implementation accordingly. 2. The abstract base class `ActOnArgs.__init__` argument `log_of_measurement_results: Dict` is now copied before use. For users that depend on the pass-by-reference semantics (this should be rare), they can use the new `classical_data: ClassicalData` argument instead, which is pass-by-reference.
Problem:
When running on hardware Cirq supports two uses of classical data. The first is in the supplying of classical information to a circuit as a parameter. The second is as the measurement outcomes recorded in the measurement dictionary. This is not sufficient for future needs (see Feedforward #3232 and Flow Control #3234). Further the output of running a Cirq circuit, measurement dictionaries, is overly restrictive and more general measurement results (IQ points, for example) should be supported. Cirq also lacks the ability to specify classical data which can be loaded by a user, used during running of a circuit, and read out at the end of the circuit. This feature is to add the ability to support classical data.
Rough requirements:
The text was updated successfully, but these errors were encountered: