Sponsored Post
Intel has long focused on supporting HPC software. But, as the years have gone by, much has changed — and the company’s offerings have grown and evolved.
A chapter from a recent edition of Parallel Universe Magazine, from this past July outlines this evolution and offers seven ways HPC software developers can benefit from Intel software investments.
According to the chapter in the latest Intel magazine, “HPC is at the dawn of a new golden age of hardware variety that recalls the early days of vector supercomputers, systolic arrays, and hypercubes.”
But the report says, for modern computing, there’s one big difference: the installed base of mission-critical scientific and engineering software.
In other words, today, success in HPC is defined by creating hardware that serves software needs.
In many ways, Intel and HPC have grown up together—and both have become very diverse and complex. Intel’s investments in software continue to expand the ways it helps HPC software developers. — Parallel Universe Magazine
Intel provides a list of seven ways Intel has contributed to software in ways that “truly matter” to software developers. This was outlined in further detail in the recent magazine issue that outlines evolution in Intel software:
- Community Support: Lifting Popular Community Codes to Higher Performance―and Letting Us Learn from It, Too: Intel established the Intel Parallel Computing Centers (IPCC) and the Intel Modern Code initiative to help HPC software harness new hardware.
- Deep Program Analysis Tools — Helping Experts Tune their Applications and Systems: The name VTune has become legendary among profiling tools. New innovations offer roofline analysis, application performance snapshots, storage performance snapshots, MPI communications analysis (Intel Trace Analyzer and Collector), and profiling of compiled code mixed with Python and Java.
- Highly-Tuned Libraries — Drop Them In, and Run Faster: Now, enterprises can use a vendor-optimized library to make a program run faster. Intel Math Kernel Library (Intel MKL) has done this for BLAS, FFTs, and solvers for years.
- Optimizing Compilers — Compile with Them and Run Faster: Augment development processes with compilers from Intel to create applications that can run faster and more efficiently.
- Software-Defined Visualization: Intel has invested heavily to support high-performance scientific visualization.
- High-Performance AI with Intel-Optimized Deep Learning Frameworks and Accelerated Python: Intel has invested in optimizing popular deep learning frameworks (e.g., TensorFlow and PyTorch) to do high-performance training and inference.
- Cross-Architecture Tools: Intel Distribution of OpenVINO Toolkit and a Vision to Help You Code Once and Run Faster Across a Variety of Hardware: The concept of code once and run everywhere isn’t new, but all the solutions seem to incur severe performance penalties on at least some platforms. Intel started an open-source project known as Intel Distribution of OpenVINO toolkit (which stands for open visual inferencing and neural network optimization). It spans CPUs, GPUs, FPGAs, and VPUs.
Intel has closely followed HPC as it’s grown, and both Intel’s offerings and the world of HPC has grown more complex over the years.
Download the new issue of Parallel Universe Magazine that outlines seven “concrete things to download, learn, and use” from Intel with the potential to benefit the HPC community.