Device-specific Compilation

Once a target device has been selected, the quantum circuit — typically designed in a device-agnostic fashion that does not account for hardware limitations (such as a limited gate set or limited connectivity) — must be compiled so that it actually becomes executable on that device.

Compilation itself is a sequential process consisting of a series of compilation passes that step-by-step transform the original quantum circuit until it conforms to the limitations imposed by the target device. Since many of the underlying problems in compilation are computationally hard, an ever-growing variety of compilation passes is available across several quantum SDKs and software tools—each with their own advantages and disadvantages.

As a result of the sheer number of options, choosing the best sequence of compilation passes for a given application is nearly impossible. Consequently, most quantum SDKs (such as Qiskit and TKET) provide easy-to-use high-level function calls that encapsulate their sequence of compilation passes into a single compilation flow. While this allows convenient circuit compilation, it has several drawbacks:

  • It creates a kind of vendor lock that limits the available compilation passes to those provided by the SDK offering the compilation flow.

  • The respective compilation flows are designed to be broadly applicable and are neither device-specific nor circuit-specific.

  • No means are provided to optimize for a customizable figure of merit.

To overcome these limitations, the MQT Predictor framework provides device-specific quantum circuit compilers by combining compilation passes from various compiler tools and learning optimized sequences of those passes with respect to a customizable figure of merit. This mix-and-match approach eliminates vendor lock-in and enables optimized compilers that transcend the capabilities of individual tools.

The compilation process is modeled as a Markov Decision Process and takes three inputs to train a reinforcement learning (RL) model that acts as a compiler:

  1. Training circuits

  2. The targeted quantum device

  3. The figure of merit to optimize for

Illustration of the RL model

The trained model can then be used to compile any quantum circuit for the targeted device.