[uurl]

Summary

• Map libm::j1 and libm::j1f to __nv_j1 and __nv_j1f.
• Normalize the result to +0 for either infinite input. CUDA returns -0 for j1(-inf), while libm 0.2.16 returns +0.
• Add generator validation, MIR lowering tests, and a GPU example covering both precisions, signed zero, finite inputs, infinities, NaN, and control calls.

The generated table now covers 120 of 124 f32/f64 libm exports. jn, jnf, yn, and ynf remain for a follow-up PR.

Validation

• cargo run -p cuda-intrinsics-gen --locked -- check
• cargo test -p cuda-intrinsics-gen -p dialect-mir -p mir-importer -p mir-lower --all-targets --locked
• cargo clippy -p cuda-intrinsics-gen -p dialect-mir -p mir-importer -p mir-lower --all-targets --locked -- -D warnings
• cargo oxide run math_bessel_j1: 84 checks passed across f32 and f64
• just check

Stacked on #1653. Refs #1224.

Testing

• just -f cuda-oxide/Justfile check passes (the local mirror of CI: fmt, clippy, tests, guards, docs)
• cargo oxide run <example> passes, or cuda-oxide/scripts/smoketest.sh -o '^<example>$'
• New example added (if applicable)

Checklist

• All commits signed off (git commit -s)
• SPDX headers on new source files

[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.