[Feat] Add fx.num_warp_threads() as constant accessor

[sjfeng1999]

Motivation

Technical Details

Test Plan

Test Result

Submission Checklist

• Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.

[Copilot]

Pull request overview

Adds a new constant accessor in the Python DSL GPU intrinsics layer so kernel authors can query the target warp (subgroup) lane count as a Python int.

Changes:

• Add fx.num_warp_threads() in flydsl.expr.gpu returning current_target().warp_size.
• Add current_target() convenience accessor to flydsl.compiler.backends and export it publicly.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
python/flydsl/expr/gpu.py	Exposes num_warp_threads() in the GPU intrinsics namespace (but currently introduces an import-time dependency that should be made lazy).
python/flydsl/compiler/backends/init.py	Adds current_target() helper returning the active backend’s GPUTarget.

Suppressed comments (1)

python/flydsl/expr/gpu.py:122

• num_warp_threads() relies on current_target, but current_target is imported at module scope. After removing the module-scope import, add a local import here (matching the existing pattern in expr/primitive.py) to avoid backend registry side effects during flydsl.expr import.

def num_warp_threads():
    """Lanes per warp on the target being compiled for, as a Python int."""
    return current_target().warp_size

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.