Add NVTX annotations to cuda.compute user-facing APIs

[Copilot]

Description

closes #3395

Instruments all user-facing functions in cuda.compute with NVTX annotations for Nsight Systems profiling visibility.

Implementation

• Created _nvtx.py module with @annotate() decorator wrapping nvtx.annotate()
• Applied decorator to:
  • All algorithm functions: reduce_into, inclusive_scan, exclusive_scan, transforms, histogram, sorts, etc.
  • All make_<algorithm> factory functions
  • All __call__ methods of algorithm classes
  • All iterator factory functions: CountingIterator, ConstantIterator, TransformIterator, etc.
• Added nvtx dependency to pyproject.toml

Configuration

• Domain: "cuda.compute"
• Color: 0x76B900 (NVIDIA green)
• Message: Function name (automatic)

Example

import cuda.compute as compute

# Now visible in Nsight Systems under "cuda.compute" domain
compute.reduce_into(d_in, d_out, op, num_items, init, stream)

Checklist

• New or existing tests cover these changes.
• The documentation is up to date with these changes.

Original prompt

---

This section details on the original issue you should resolve

<issue_title>Add nvtx annotations around all functions in cuda.compute.</issue_title>
<issue_description>The user facing APIs in cuda.compute should be annotated using nvtx, so that they show up in Nsight Systems</issue_description>

<agent_instructions>Use https://nvtx.readthedocs.io/en/latest/index.html, introducing nvtx as a dependency in pyproject.toml. For most use cases, the high-level decorator @nvtx.annotate() should suffice. Use the domain name cuda.compute and use NVIDIA green (76B900) as the color. Be sure to also annotate the user-facing iterator creation functions. Be sure to also annotate the make_<algorithm> functions and their __call__ methods.</agent_instructions>

Comments on the Issue (you are @copilot in this section)

• Fixes Add nvtx annotations around all functions in cuda.compute. #6905

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.

[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.

[shwina]

/ok to test 4ad5fe0

[shwina]

/ok to test e3f280a

[github-actions]

🥳 CI Workflow Results

🟩 Finished in 1h 04m: Pass: 100%/48 | Total: 12h 16m | Max: 41m 52s

See results here.