[burgholzer]

🤖 AI text below 🤖

Description

Build the clicker-operated MQSF keynote, System Software for Quantum Computing: From the Metal to the User. Seventeen slides connect the shared software stack to a hybrid QC-AFQMC application, device-aware compilation, structured programs, and adaptive execution. The static HTML runs offline with forward/backward controls.

Presentation preview · Presentation ZIP · Supporting evidence

• Build the architecture from users and classical compute through compiler, orchestrator, and QDMI. All frontends and their user connections appear before the program/runtime split. Shared geometry, vector logos, repository statistics, and community attribution carry through the conclusion.

• Run stretched H₂ QC-AFQMC through the Slurm playground: 4,096 programs with 1,024 shots each in one QDMI batch, followed by four-node walker propagation. Actual preprocessing, program completions, indexed retrievals, and walker observations drive synchronized replays and a continuous application clock.

• Compile the exact submitted application payload for IQM Emerald, IBM Nighthawk, and IonQ Forte-1. Recorded optimization, routing, native synthesis, metrics, and circuits accompany each target. The executable panel shows QIR for IQM, Qiskit's own SVG drawing plus qpy.dump for IBM, and native OpenQASM 3 for IonQ. IBM QPY is exported from the retained native IR and checked for circuit, phase, and layout preservation through a serialization round trip. QIR programs execute unchanged on DDSIM; measurement-instrument comparisons check the three targets.

• Show the complete syndrome-extraction QIR entry point with readable reset, measurement/feedback, and loop excerpts. Recorded Grover program sizes motivate preserving structure. These portable structured examples do not undergo hardware mapping or native synthesis.

• Execute iterative QPE for a four-qubit unitary and recycled control. Both the eight-round outer loop and runtime-dependent U-power loop remain in adaptive QIR. Four native workers execute 8,192 shots; indexed 64-shot completion observations synchronize their counters, the QDMI/JIT sequence, and the histogram. The complete circuit diagram comes from the submitted source because Qiskit export cannot represent this combination of dynamic bounds and classical indexing.

• End the feature overview with a Unitary Foundation QOSS survey invitation and vector QR. It fades in after the final feature card on the same click; the invitation includes the 20-minute estimate and 29 October deadline.

All four extracted compiler fixes are upstream: controlled-gate canonicalization #2752, GPI synthesis #2753, explicit CLI targets #2754, and scoped classical storage/source names #2755. Remaining native changes are optional mapping, synthesis, and DDSIM tracing with regression coverage. The compiler-change inventory records their scope, line counts, and limits. Capture paths and export tests explicitly request copies where upstream's consuming Qiskit export would otherwise invalidate a later use.

Read the Docs only packages static files. The browser loads a compact display projection. The 2.1 MB presentation ZIP contains four files: standalone HTML, slide-by-slide rehearsal notes, a quick-start guide, and combined notices. A separate evidence ZIP preserves original records, hashes, editable vectors, licenses, provenance, and technical documentation. Speaker notes cover all 17 slides with build cues and transitions; every slide also has concise on-screen cues. CI publishes the archives separately. Navigation includes slide jumps, overview, notes, reduced motion, and print output. Chart series have distinct line styles and markers, and active histogram bins have an outline so interpretation does not depend on color alone. Obsolete iteration plans have been removed.

Targets use SC QDMI models populated from retained provider snapshots. No hardware jobs are submitted. Four Slurm containers share one host. The bounded chemistry example and conditional walker-error bands do not establish convergence, cluster scaling, or quantum advantage. QIR timing observes completed batches and includes instrumentation overhead; this is not a throughput benchmark. Captures retain their original revisions rather than being relabeled after a rebase.

Validation (native results from the rebase to main ee77292f8; package and speaker-note checks repeated for this update):

• Optimized Clang/ThinLTO build and install; native CTest suite: 4,041 passed, one intentionally skipped.
• 535 Python tests and five subtests passed, including presentation capture and Qiskit translation; direct unitary-equivalence and syntax-export capture probes passed.
• Whole-file C++ lint passed for the rebase. The capture/output-format update passed all 83 presentation tests plus five subtests. This packaging cleanup passes all 41 build tests and repository lint; it changes no native implementation.
• Offline Chromium (extracted presentation ZIP) and Firefox (preceding output-format update): 17 slides, 55 builds, 24 finite replays, 709 circuit windows, 144 animation-layout samples, and 17 print pages. First paint, interrupted/restarted animation, result synchronization, navigation, and reduced motion passed. Sampled p95 frame intervals were 16.7 ms and 17.2 ms respectively.
• Chrome DevTools inspection covered all slides in the rebase and the revised output panels and survey reveal in this update, including console and network. The survey QR decodes at its displayed 176-pixel size. Both archives pass integrity checks. The presentation was extracted from its four-file ZIP for offline Chrome checks, including all 17 notes overlays; evidence and editable assets remain byte-identical to their source files.

Hosted CI for the updated head is pending. AI assistance covered implementation, independent review, validation, and this description within the user's authorization. Physical laptop/projector/clicker rehearsal remains.

Checklist

• The pull request only contains commits that are focused and relevant to this change.
• I have added appropriate tests that cover the new/changed functionality.
• I have updated the documentation to reflect these changes.
• The changes follow the project's style guidelines and introduce no new warnings.
• The changes are fully tested and pass the CI checks.
• I have reviewed my own code changes.

If PR contains AI-assisted content:

• Any agent that created, edited, or submitted GitHub content was explicitly authorized for that scope, as required by our AI Usage Guidelines.
• Every agent-authored or agent-edited public text body begins with the visible disclosure 🤖 *AI text below* 🤖 (titles are exempt).
• I have disclosed AI assistance in the PR description.
• I confirm that I have personally reviewed and understood all AI-generated content, and accept full responsibility for it.

[codecov]

Codecov Report

❌ Patch coverage is 97.92746% with 4 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
...QCO/Transforms/NativeSynthesis/TargetSynthesis.cpp	95.1%	3 Missing ⚠️
...lir/lib/Dialect/QCO/Transforms/Mapping/Mapping.cpp	99.2%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!

[burgholzer]

MQSF is over and was a success; archived this elsewhere. Closing here.