Performance Pragma introduces a higher-level approach to optimization in Vitis HLS:
Define a system-wide throughput goal using
#pragma HLS performance target_ti.Use dataflow to enable parallel execution.
Let CSIM automatically determine loop trip counts.
Apply loop-level pragmas only for critical bottlenecks.
Validate results through synthesis reports and iterate as needed.
By following this methodology, developers can reduce manual effort, improve design maintainability, and achieve performance targets efficiently.
Copyright © 2020–2026 Advanced Micro Devices, Inc.