Add SAR backprojection example using real data

[tbensonatl]

Add an example using the sar_bp and other MatX operators to reconstruct a real data set. Several commercial spaceborne SAR companies provide data sets with permissive licensing. This example includes supporting scripts and code to read and process one such data set. See the example README.md file for detailed instructions.

This example will download and install additional third-party open-source software projects and data sets. Review the license terms of these open-source projects and data sets before use.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[greptile-apps]

Greptile Summary

This PR adds a complete end-to-end SAR backprojection example using real spaceborne CPHD data: a Python preprocessing script (cphd_to_sarbp_input.py) that converts CPHD to a self-contained .sarbp binary, a CUDA driver (sarbp.cu) that performs GPU backprojection via the MatX sar_bp operator, and a visualisation helper (view_sarbp_image.py). All previously flagged P0/P1 issues (stack overflow on oversized pulse headers, bad fallback header size, byteswap correctness, TOA intersection edge case, vel_enu trimming) have been addressed in follow-up commits.

Confidence Score: 5/5

Safe to merge; all prior P0/P1 concerns are resolved and only minor P2 observations remain.

Every critical issue raised in previous review rounds (stack overflow, bad fallback size, byteswap correctness, TOA intersection, vel_enu post-trim mismatch) has been addressed. The two remaining comments are P2 quality-of-life items (silent write-error and an unvectorised Python loop) that do not affect correctness or security.

No files require special attention; sarbp.cu output-write error handling is a minor hardening opportunity.

Important Files Changed

Filename	Overview
examples/sarbp/sarbp.cu	Main CUDA backprojection driver; previous P0 (stack overflow, bad fallback size) and P1 (warmup semantics) issues resolved; output-write error check missing (P2).
examples/sarbp/cphd_to_sarbp_input.py	CPHD → .sarbp conversion script; previously flagged issues (byteswap, TOA intersection, vel_enu trimming) all resolved; Doppler mask construction is an unvectorised Python loop (P2).
examples/sarbp/view_sarbp_image.py	Visualisation helper; reads .sarbp header for dimensions, handles square-image fallback, and gracefully selects an interactive matplotlib backend. No issues found.
examples/CMakeLists.txt	Adds sarbp executable target and a post-build rule to copy the two Python helper scripts beside the binary. Straightforward and correct.
examples/sarbp/requirements.txt	Pinned dependency list for sarpy, matplotlib, numpy, scipy etc. Versions look reasonable; pinning aids reproducibility.
examples/sarbp/README.md	Clear step-by-step instructions covering data download, Python venv setup, conversion, build, GPU execution and visualisation. No issues.

Sequence Diagram

sequenceDiagram
    participant User
    participant Python as cphd_to_sarbp_input.py
    participant CPHD as CPHD File (sarpy)
    participant SARBP as .sarbp File
    participant CUDA as sarbp.cu (GPU)
    participant RAW as .raw Output

    User->>Python: run with .cphd path
    Python->>CPHD: open_phase_history() + read_pvp_array()
    CPHD-->>Python: metadata, PVP, signal data
    Python->>Python: aperture trim, ENU conversion, resolution auto-sizing
    Python->>SARBP: write header + per-pulse records (int16 or complex64)
    User->>CUDA: run sarbp executable
    CUDA->>SARBP: read 256-byte file header
    CUDA->>SARBP: stream per-pulse headers + samples
    CUDA->>CUDA: ifft-shift + zero-pad + IFFT/FFT (range compression)
    CUDA->>CUDA: sar_bp accumulate blocks → image tensor
    CUDA->>RAW: write complex float image
    User->>Python: view_sarbp_image.py output.raw

Loading

_{Reviews (6): Last reviewed commit: "Fix viewer orientation; update example S..." | Re-trigger Greptile}

[tbensonatl]

/build

[coveralls]

Coverage is 91.805% — add-sarbp-real-data-example into main. No base build found for main.

[tbensonatl]

/build

[greptile-apps]

Want your agent to iterate on Greptile's feedback? Try greploops.

[tbensonatl]

/build

[tbensonatl]

/build