Further Reading: Framework Comparison and Interoperability

Tagged Tier 1 (confident it exists and recommended) and Tier 2 (real and worth seeking, but verify the current version or URL).

The primary sources

• The OpenQASM 3 specification. The interchange format this chapter rests on. Read specifically for what it does not specify — the state-vector ordering convention is not in there, and §18.4's central result follows from that omission. The classical-control features (which QASM 2 lacks) are the other section worth attention, since they determine whether Chapter 9's dynamic circuits survive a round trip. Tier 1.
• Cross, Javadi-Abhari, et al., "OpenQASM 3: A Broader and Deeper Quantum Assembly Language" (2022), ACM Transactions on Quantum Computing 3, 12. The design paper. Explains what QASM 3 adds over QASM 2 and why — timing, classical control, and extensibility. Tier 1.
• Each framework's own migration and interoperability documentation. Qiskit's qasm2/qasm3 modules, cirq.contrib.qasm_import, Braket's OpenQASM guide, qml.from_qasm, and qdk.openqasm. These change more often than the specification does, which is precisely why §18.3's table is dated and should be re-verified. Tier 1.
• The Qiskit QPY documentation. Qiskit's native serialization format, which does preserve global phase and parameter names — the properties §18.4 measured OpenQASM dropping. The right choice when you are round-tripping within Qiskit and the wrong one when you need to leave it. Exercise 18.25's subject. Tier 1.

On the conventions

• Any two quantum computing textbooks, compared on how they write a multi-qubit basis state. The endianness split is not a software quirk; it is a field-wide notational disagreement the frameworks inherited. Nielsen and Chuang, Kaye–Laflamme–Mosca, and Mermin do not all agree, and noticing that is more useful than memorizing any one of them. Tier 1.
• The classical literature on byte order. Cohen's "On Holy Wars and a Plea for Peace" (1981) is the canonical treatment, and the failure modes it describes — a convention mismatch at a boundary, invisible to symmetric test data, self-inverse so that double conversion hides it — are exactly Chapter 14's Case Study 1. Forty years older and no less relevant. Tier 2.

On interoperability and portability

• Any current work on quantum intermediate representations. QIR (the LLVM-based representation Q# compiles to), OpenQASM, and the various vendor IRs are competing answers to the same question. Reading two of them side by side clarifies what a "standard" is actually standardizing. Tier 2.
• The qiskit-braket-provider, pennylane-qiskit, and similar bridge packages. The practical alternative to a QASM pipeline: direct adapters. Worth reading the source of one to see how it handles the three losses §18.4 catalogues — most handle the endianness and quietly drop the rest. Tier 2.
• Literature on reproducibility in computational science. §18.4's losses are a reproducibility problem: two groups running "the same circuit" in different frameworks can get different answers for reasons neither reports. The general literature on computational reproducibility is more developed than anything quantum-specific and transfers directly. Tier 2.

On the frameworks themselves

The per-framework reading lives in the chapters:

• Chapter 14 — Cirq, and the Sycamore result it was built around.
• Chapter 15 — Q#, and Gidney and Ekerå on what fault tolerance costs.
• Chapter 16 — PennyLane, the parameter-shift rule, and the barren plateau literature.
• Chapter 17 — Braket, trapped ions, and neutral atoms.
• Chapters 7–13 — Qiskit, throughout.

Forward references

• Chapter 19 — Part IV opens, and the subject changes from how to run circuits to what to run.
• Chapter 27 — testing quantum programs, where this chapter's "a test that cannot fail is not a test" becomes a discipline rather than an anecdote.
• Appendix E — the gate-by-gate, call-by-call translation dictionary across all five frameworks.
• Appendix H — the wider ecosystem: plugins, bridges, and the tools that sit alongside these five.

Where to go next. If one thing: the OpenQASM 3 specification, read for its omissions. The state-vector convention is absent, and §18.4's entire result is a consequence of that single non-decision — which is a good lesson about reading standards for what they decline to standardize.

If two: add Cohen's byte-order paper, because the bug this Part spent two case studies on was thoroughly documented in 1981 and the field re-derived it from scratch.

Then Chapter 19 — Part IV, and the question this book has deferred for eighteen chapters: what do you actually run?