Description

Why Performance Analysis?

Without the right data, you’re guessing about how to improve software performance and are unlikely to make the most effective improvements. Intel® VTune™ Profiler collects key profiling data and presents it with a powerful interface that simplifies its analysis and interpretation. Optimize software for:

• High-performance computing (HPC) in weather forecasting, finite element analysis, and bioinformatics
• Embedded applications for IoT, transportation, and manufacturing
• Media software for video transcoding and image processing
• Cloud applications or Java* services in containers
• Device drivers
• Game engines
• Storage that includes Storage Performance Development Kit (SPDK) and Data Plane Development Kit (DPDK) polled software

Additional Capabilities
	Optimize single-threaded performance.
	Multithreaded - Effectively use all available cores.
	System - See a system-level view of application performance.
	Media & OpenCL™ Applications - Deliver high-performance image and video processing pipelines.
	HPC & Cloud - Access specialized, in-depth analyses for HPC and cloud computing.
	Memory & Storage Management - Diagnose memory, storage, and data plane bottlenecks
	Analyze & Filter Data - Mine data for answers.
	Environment - Fits your environment and workflow.

The Intel® VTune™ Amplifier is available as part of the Intel® oneAPI Toolkits.

---

Using OpenMP & Intel® VTune™ Profiler for Faster Code and Improved ROI

Our

training video

shows the performance effect adding a single OpenMP statement to some Fortran code which is then compiled with Intel® Fortran.

Since the example being used contains a load balance problem, Intel® VTune™ Profiler is used to analyse the performance of this demonstration program.

VTune pin-points the hot-spot that causes the imbalance.

Then, after explaining the imbalance situation, it is shown how to change the OpenMP statement, so that the run-time performance of the example program is further increased.

Steps taken to improve run-time performance

Finally it is shown that the performance could be maximised by using the optimization possibilities of Intel Fortran.

The maximum speed-up on an Intel Core i7 (4 cores): 3.6 faster than the regular serialised running code.