ray-flag-force-opaque-with-anyhit
Status: stub. The full-length analysis is queued for a v1.0.x patch release per ADR 0018, section 5, criterion #6. The companion rule page at docs/rules/ray-flag-force-opaque-with-anyhit.md contains the canonical detection logic + GPU reasoning.
TL;DR
RAY_FLAG_FORCE_OPAQUE skips AnyHit invocation. Binding an AnyHit and then forcing opaque is dead code or a logic bug -- the AnyHit shader will never run. On RDNA 4, NVIDIA Ada/Blackwell and Intel Xe2, the AnyHit binding still costs scheduling table slots even when never invoked.
What the rule fires on
A TraceRay(...) call with RAY_FLAG_FORCE_OPAQUE set in a translation unit that also defines a [shader("anyhit")] entry point.
See the What it detects section of the rule page for the full pattern definition.
Why it matters
The full GPU-mechanism analysis lives in the Why it matters on a GPU section of the companion rule page.
Examples
The bad / good code snippets are kept canonical on the rule page; see ray-flag-force-opaque-with-anyhit.md -> Examples.
See also
- Rule page -- canonical reference + change log.
- dxr overview -- broader context.
- ADR 0018 -- v1.0 readiness plan.
This is a v1.0-ship stub. Full analysis pending; track issue link TBD.