Enabling Accuracy-Aware Quantum Compilers using Symbolic Resource Estimation
summary
Claims an automatic accuracy manager that extracts closed-form cost and error formulas from source code and uses them to tune approximation parameters to minimize T-count subject to a target error. Implements a LLVM prototype with a ClangTool, AST-to-C++ extraction, LLVM loop passes and symbolic simplifications to produce symbolic resource estimation for T-count and total epsilon. Evaluation uses target error 5e-3 and shows constant evaluation time for the symbolic method versus growing runtimes (linear for QPE, cubic for Shor), so the optimizer runs hundreds of times feasibly.
AI-written summary · about
chapters (13)
- 0:00Intro
- 0:15Quantum computers
- 1:04Quantum programming languages
- 3:28Quantum circuits
- 4:47Errors in quantum compilation
- 5:14Composing approximation errors
- 6:08Approximation-aware programming
- 8:57Language support With larguage support
- 9:25Automatic accuracy management
- 11:56Symbolic cost/constraint functions
- 13:16LLVM prototype
- 14:29LLVM implementation
- 15:53Conclusions
related talks
19mTraq: Estimating the Quantum Cost of Classical Programs
19mThe T-Complexity Costs of Error Correction for Control Flow in Quantum Computation
26mUREQA: Leveraging Operation-Aware Error Rates for Effective Quantum Circuit Mapping
19mFlexible Type-Based Resource Estimation in Quantum Circuit Description Languages
46mUnlocking quantum potential with software and compilers
19mCausalTuner: Feature-Aware Causal Guidance for Compiler Auto-tuning
53mLocknote: How Badly Do We Want Correct Compilers?
15mAccelerQ: Accelerating Quantum Eigensolvers With Machine Learning on Quantum Simulators