Advisor is an advanced profiling tool which helps optimize code on Intel Xeon and Xe architectures by allowing developers to compare their code’s performance to the theoretical limits of the hardware and understand vectorization and offload efficiency. It helps ensure Fortran, C and C++ applications achive full performance potential on the Aurora system by:
- Offloading where it pays off the most - Use this feature to design code for efficient offloading to accelerators, even before you have hardware. Estimate code performance and compare it with data transfer costs. No recompilation required.
- Roofline analysis for CPU and GPU - Optimize CPU or GPU code for compute and memory. Locate bottlenecks and identify performance headroom for each loop or kernel to prioritize which optimizations will deliver the highest performance payoff.
- Design well-threaded & vectorized code - Develop performant software using vectorization optimization and thread prototyping capabilities.
- Modeling, tuning, and testing multiple threading designs.
- Creating and analyzing data flow and dependency-computation using heterogeneous algorithms.