Skip to content

v1.30.0

Latest
Latest

Choose a tag to compare

View all tags
@billti billti released this 09 Jul 20:57
v1.30.0
6f080b5
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Below are some of the highlights for the 1.30 release of the QDK.

Adaptive profile capabilities

QIR is the industry standard format that the QDK compiles programs into from various quantum languages (Q#, OpenQASM, Qiskit, etc), and is how programs are sent to quantum computers for execution, such as through Azure Quantum. It is also the format some of the QDK simulators use, such as the Stabilizer and GPU state vector simulators.

QIR specifies different profiles that dictate what instructions it may contain, and the profile may contain optional features. In this release, we have added a number of these "optional" features in the code generation for the "Adaptive" profile.

Several of these features don't directly affect what your code can express, but they can significantly impact performance. For example, by being able to directly express loops (rather than having to unroll them) and calls (rather than having to inline them) compilation time can be greatly reduced and compiled program size significantly decreased. In the most notable cases internally, we observed both improve by orders of magnitude.

One concrete example of a newly supported capability is unbounded loops, such as the "repeat-until-success" pattern. The below is a contrived but minimal example of a "repeat-until-success" loop that now compiles to QIR. (Previously, this would have given a "cannot have a loop with a dynamic condition" error).

@EntryPoint(Adaptive)
operation Main() : Int {
    mutable iterations = 0;
    use qubit = Qubit();

    // Loop until the measurement of the qubit in the Z basis returns One
    repeat {
        iterations += 1;
        Rx(0.1, qubit);
    } until MResetZ(qubit) == One;

    // Return the number of iterations it took to measure the desired state
    iterations
}

The Adaptive capabilities are a work in progress. Please check the wiki page at https://github.com/microsoft/qdk/wiki/QIR for the latest capabilities, limitations, and known issues. As always, please log an issue at https://github.com/microsoft/qdk/issues for any bugs, questions, or feature requests.

New quick-fixes

Several new Quick Fixes have been added this release. The first is to add missing import statements.

import-quickfix.mp4

Another common coding error is to pass a single qubit where a qubit array was expected, such as in Controlled functors.
For example, the code Controlled SX(qs[0], qs[1]); gives an error of "type error: expected Qubit[], found Qubit", and the "Convert to single element array" Quick Fix will change the code to be Controlled SX([qs[0]], qs[1]);, resolving the error.

Simulator loss policies

The noise model that can be applied to quantum simulations now support specifying a "loss policy", which describes the behavior of a two-qubit gate when one of the qubits is lost.

Previously the behavior was always to skip the two-qubit operation if one qubit is marked as "lost". This makes sense, for example, if mathematically you treat a lost qubit as being in the $\ket{0}$ state, then gates such as CX and CZ are effectively no-ops if one qubit is lost. On some quantum machines, the effect of a two-qubit gate may differ. The "loss policy" can now specify the desired behavior, for example:

from qdk.simulation import NoiseConfig, LossPolicy, run_qir
qir = ... # get the compiled program

noise = NoiseConfig()
noise.cz.on_loss   = LossPolicy.SKIP               # if one of the qubits is lost, skip the unitary
noise.cx.on_loss   = LossPolicy.PROPAGATE          # if one of the qubits is lost, lose the other one also
noise.rxx.on_loss  = LossPolicy.DEGRADE            # degrade to a single qubit gate, i.e. rx on the remaining qubit
noise.ryy.on_loss  = LossPolicy.RESIDUAL_S_DAGGER  # apply an S_DAG to the remaining qubits
noise.swap.on_loss = LossPolicy.APPLY_ANYWAY       # if swap is implemented as a relabel, then it still applies

# Works with all simulator types, in any profile.
run_qir(qir, shots=100, noise=noise, type="clifford")

Other notable changes

New Contributors

Full Changelog: v1.29.0...v1.30.0

Contributors

idavis, billti, and 13 other contributors
Assets 3
Loading