feat: Add float32 zero-copy angle encoding in qdp-core (single sample)

[viiccwen]

Related Issues

Closes #1248

Changes

• Bug fix
• New feature
• Refactoring
• Documentation
• Test
• CI/CD pipeline
• Other

Why

It adds single-sample float32 zero-copy angle encoding support in the Rust core and CUDA kernel layer, following the existing amplitude f32 rollout pattern.

How

• added a float32 CUDA kernel entrypoint for single-sample angle encoding
• added AngleEncoder::encode_from_gpu_ptr_f32_with_stream(...)
• added public QdpEngine APIs:
  • encode_angle_from_gpu_ptr_f32(...)
  • encode_angle_from_gpu_ptr_f32_with_stream(...)
• validated the new GPU-pointer path for:
  • empty input
  • null pointer
  • qubit/input-length mismatch
• added kernel-level numerical tests for the new float32 angle path
• kept engine precision conversion behavior consistent with the existing core API

Files Changed

• qdp/qdp-core/src/gpu/encodings/angle.rs
• qdp/qdp-core/src/lib.rs
• qdp/qdp-core/tests/gpu_ptr_encoding.rs
• qdp/qdp-kernels/src/angle.cu
• qdp/qdp-kernels/src/lib.rs
• qdp/qdp-kernels/tests/angle_encode.rs

Checklist

• Added or updated unit tests for all changes
• Added or updated documentation for all changes

[viiccwen]

cc @400Ping, @ryankert01

[ryankert01]

Overall lg, I think maybe can add a python test also?

[viiccwen]

Overall lg, I think maybe can add a python test also?

Currently, this PR doesn't include python binding, so it's still unsupported in python behavior.
I'll change python binding in the follow-up PR (#1250).

[guan404ming]

Looks great

[guan404ming]

I think this is missing validate_qubit_count(num_qubits)? call that all other encode paths use. Guards against overflow in 1 << num_qubits.