CUDA-Q Logical by Example

The numbered Python programs and steane_bell_e2e_workflow.py under preview/logical/examples/ run in the test suite. The top-level sequence uses ordinary CUDA-Q kernels and target stacks; the standalone sequence exposes each CUDA-Q Logical compiler stage directly. The gadget sequence introduces verification and detector error models separately: start with one authored gadget, then build a physical surface-code memory experiment.

For one computation across both routes, follow the Steane Bell-pair walkthrough.

CUDA-Q kernel examples

• Logical Resource Estimate
• Clifford+T Synthesis
• CUDA-Q Logical Resource Estimate
• Fermi–Hubbard Estimation
• Carbon Code Target
• Gidney–Ekerå RSA

Standalone compiler examples

Examples 00–03 introduce authoring, placement, code and gadget definitions, and physical scheduling. Continue with encoded execution (06), or factory protocols and evaluation (07–08). Example 09 combines synthesis, factory provisioning, estimation, and scheduling in a lookup-addition capstone.

• Portable P0 Bell Program
• P1 Placement
• Code and Gadget
• Standalone Physical Schedule
• Steane-Encoded Bernstein–Vazirani on Stim
• T-State Distillation
• T-State Distillation with cudaq.run
• Gidney–Ekerå Lookup Addition

Gadget verification and error models

Start with Verify an authored gadget and Physical surface-code memory. The optional Stim-emission tutorial uses ideal MPP measurements, and the custom-noise tutorial derives models with an explicit fault factory. The reference comparison below adds individual model exports and two sampling paths to the physical-memory workflow.

• Verify an Authored Gadget
• Physical Surface-Code Memory
• Surface-Memory Stim Export
• Surface-Memory Error Model
• Surface-Memory Reference Comparison