Worked on the ROCm/rocm-systems repository to enhance the reliability and clarity of HIP performance testing. Migrated existing HIP performance tests for Compute and Stream modules to the Catch2 framework using C++ and HIP, unifying test reporting and improving maintainability. Developed new kernel latency and dispatch speed tests with advanced timing modes and iterative measurements, providing more accurate and actionable performance metrics for GPU computing. Improved test diagnostics by introducing unified logging macros and conditional debug output, streamlining debugging and performance analysis. These efforts standardized test reporting and accelerated optimization cycles, supporting more efficient benchmarking and kernel development across the HIP stack.