This thesis is first and foremost a story about compilers. It is interesting to note that whereas the term quantum compilation has been in use for the longest part of the existence of quantum computing as a field, it is only recently that our community has started to adopt tools, ideas and results from our classical counterparts {% footnote(id=“test”) %} We use the word classical as a derogatory term to refer to any form of computing that is not advanced enough to be quantum. {% end %}.

Strengthening the bridge between classical and quantum compilation research is one of the main motivations for this thesis – and arguably its most ambitious goal.

Other similar text #

This chapter is the story of how the expression and transformation of quantum computations is changing. As is often the case, this journey started with physicists in Oxford capturing thought experiments in quantum information theory as diagrams (Deutsch 1989, Quantum computational networks). The idea caught on, and soon software tools were created to facilitate building such diagrams (and printing them in papers) {% footnote(id=“qcl”) %} The Quantum Computation Language (QCL, Omer, 1998) was in many ways ahead of its time, offering an advanced programming interface including loops and control flow, similar to some of the languages developed today. It presumably failed to gain traction as it did not support printing diagrams for papers :D {% end %}.

As the possibility of actually performing these experiments on quantum hardware became more tangible, the need to automatically transform and optimize these diagrams became apparent. The result were software packages for quantum computing that for the first time aimed to create programs to be executed on real quantum hardware (projectq, qiskit, cirq, TKET). We called them quantum compilers.

The analogy between traditional and quantum compilers holds up to the extent that both are concerned with abstracting away some of the low-level assembly-like details of the underlying hardware {% footnote(id=“test3”) %} I am tempted to say these quantum compilers were more akin to assemblers of the 1960s (IBM FAP 1960), but that might be pushing it too far. {% end %} – more on this below. However, there are also important differences; the majority of the quantum software stacks were developped within the Python ecosystem, with limited concerns for performance on large-scale quantum programs (Ittah, 2021) and interoperability with other toolchains, and in particular with the mature classical compiler ecosystem (LLVM, MLIR).

A new generation of quantum compilers is starting to address these concerns (QCOR, QIRO, Catalyst, CUDAQ, Guppy, HUGR). We will see that this is opening up new ways to program quantum computers, that enable quantum error correction as well as seamless integration with heterogenous classical computing (CPUs, GPUs, etc) used in high-performance computing (HPC).

At the beginning was the Quantum Circuit #

s

Qiskit & Co: Static compilation #

Dynamic compilation #

This is a test of the footnote shortcode. {% footnote(id=“test2”) %} This is a test of the footnote shortcode. {% end %}