Framegraph Pipeline Hardening

Hardening objectives

This pass focused on making render pass dependencies explicit and easier to reason about under rapid feature iteration. Framegraph structure now better reflects real resource lifetimes.

The aim is fewer hidden coupling issues and safer pass-level optimization.

• Pass ordering was reviewed for dependency clarity.
• Resource transitions were tightened around actual usage.
• Toggle paths were aligned with graph execution boundaries.

Diagnostics and control improvements

Runtime toggles now provide more predictable A/B behavior for heavy features, enabling cleaner profiling sessions. This helps isolate expensive stages without invasive code edits.

The framegraph pass also improved observability for future regression checks.

• Feature toggles map directly to graph pass activation.
• Profiling runs can disable targeted features safely.
• Reduced ambiguity in pass-level cost attribution.

Near-term follow-up

The next iteration is pass fusion and bandwidth reduction where compatible stages can be merged. This should reduce synchronization and improve frame consistency.

Any fusion work will remain behind parity checks to avoid visual regressions.

• Identify merge candidates among post and lighting stages.
• Track per-pass timing before and after fusion.
• Gate changes with scripted visual comparisons.

Key metrics snapshot

Framegraph hardening is evaluated through pass-level profiling and resource-lifetime analysis. The goal is fewer expensive passes with predictable behavior.

Metrics are captured under consistent camera paths to keep comparisons stable.

• Per-pass CPU and GPU timing contribution.
• Active pass count and merge-opportunity inventory.
• Transient attachment lifetime overlap and bandwidth pressure.
• Toggle-path consistency across repeated profiling runs.

Next Steps