Performance Pragma introduces a top-down optimization flow that starts with defining a global performance target rather than focusing on individual loops. The methodology includes:
flowchart TD
A["Define Throughput Goal (target_ti)"] --> B["Re-architect for Dataflow (Load → Compute → Store)"]
B --> C["Profile Loop Trip Counts via C-sim (optional manual refinement)"]
C --> D["Apply Performance Pragmas (Top-level & Loop-level)"]
D --> E["Iterate and Validate (Check synthesis report)"]
E -->|Refine pragmas| D
For more details on pragma syntax and methodology, see UG1399.