As high performance computing (HPC) moves far beyond its traditional users to the domains of AI, machine learning, enterprises and the cloud, and heterogeneous programming becomes the norm, Nico Mittenzwey, senior HPC engineer at Megware Computer, and Mike Lee, technical, enterprise and cloud segment marketing manager at Intel, discuss the considerations in optimizing HPC clusters and applications. Nico explains how the Intel MPI Library has helped Megware and its customers rise to these challenges, providing real-world examples of two of the company’s most recent projects. He also explores the challenges of fragmentation in heterogeneous computing environments, and how a single, common programming model can help address these challenges.
To learn more:
Data scientists spend 60% of their time cleaning and preprocessing data, transforming this dirty data into crystallized insights. Dataframes, such as Pandas, provide exceptional tooling to address data wrangling tasks, yet Pandas themselves increasingly lack ease and speed as they scale. Alex Baden, Technical Director at OmniSci, and Devin Petersohn, Machine Learning Engineer at Intel, dive into the challenges and considerations of dataframe scaling. They explore how the Intel Modin / OmniSci solution, part of the Intel AI Analytics Toolkit, offers an open road to quick, transparent scaling across heterogeneous architectures. They also explain how this solution’s integration with the rest of the Python ecosystem enables data scientists to focus on extracting value from data rather than provisioning and orchestrating resources.
To deliver increasingly captivating stories in the world of animated movies takes passion and determination. The passion to continually push the boundaries of what’s possible. The determination to continually seek ways to make technology easily approachable by artists, so that the artists can focus on what they’re best at—telling stories that capture our imaginations. Max Liani , Senior Lead Engineer of RenderMan at Pixar Animation Studios, and Alex Wells, Principal Engineer at Intel, talk about collaborating to push the boundaries of what’s possible at the intersection of technology and art, and the ramifications that pervade outside the realm of motion pictures.
DPC++ provide developers with the ability to write standard C++ code for heterogeneous systems, and accelerate execution using a range of different processors including CPUs, GPUs and FPGAs.
Software developers can use this implementation to target Nvidia GPUs using any SYCL code, without any porting or special tricks required. If you have existing SYCL code, or if you are writing new SYCL code, you can compile it and target Nvidia GPUs without modifications.
Doing a hands-on exercise will help you to learn how to invoke the compiler to compile your code for Nvidia hardware using DPC++.
In Part I of two episodes about GROMACS, one of the world’s most widely used open source molecular dynamics (MD) applications, Erik Lindahl, Biophysics Professor at Stockholm University & KTH Royal Institute of Technology, and Roland Schulz, Parallel Software Engineer at Intel, unpack the mysteries of MD. As Erik explains, MD helps us understand “everything from the proteins inside our bodies to the largest galaxies in the world.” Erik and Roland discuss MD simulation using GROMACS, the immense benefits and challenges it brings, its role in the fight against our global pandemic, and some of the most promising developments. They also explore the significance of the Folding@home project as it relates to the world’s supercomputers and exascale computing.
Krystian Ligenza, Autodesk Maya Software Architect, and Mike Voss, Intel Principal Engineer, talk about how the need for immersive experiences in VFX and 3D animated movies and games has driven the need for increasingly more capable (and complex) software and hardware. They discuss how Autodesk and Intel are addressing these challenges, and explain how the companies are collaborating to equip the most iconic animation studios and game developers with the tools needed to deliver ultra-realistic, award-winning entertainment experiences.
Is oneAPI living up to its promise? How do oneAPI and DPC++ relate to the Khronos Group SYCL and ISO C++ standards? Andrew Richards and Hal Finkel respond to these provocative questions and more from Sanjiv Shah, Intel GM and VP of Developer Software Engineering. They share their experiences with oneAPI and DPC++, and talk about oneAPI’s best aspects and what they would like to change about this cross-architecture programming model. Listen in.
Ronan Keryell and Jeff Hammond explain why open collaboration — modeled through open source and open standards — is key to solving some of today’s biggest challenges in research and industry, revealing some of the misconceptions, or least understood aspects, along the way. Then they explore the value of open languages and programming models, diving into ISO C++, Khronos Group SYCL, the amazing SYCL community, and what excites them most about the SYCL 2020 Provisional Specification.
Peter Wang explains the series of happy accidents that have led to organic adoption of Python as the number one language among developers in data science and machine learning in this episode of Code Together. He and David Liu talk about the latest advancements in hardware and software and how to make them more accessible to developers, as well as how to bridge the multiple layers of abstraction. They also explore collaboration on building a low-power, high-performance data science stack, democratization of data literacy, and the future of Python in this action-packed discussion between two data science experts.
Existing math kernel libraries have lacked portability across heterogeneous platforms—until now. A unifying programming model—and availability of standard library interfaces—enables development of performance-portable libraries among diverse hardware architectures. In this episode, we talk about a collaborative project to make this portability possible, and dive into Mehdi’s work to enable the first math library implementation for oneAPI on Nvidia GPUs.
Geoff and Andrew discuss the value of standards in enabling hardware and software developers to collaborate and innovate. From the early days of video games to modern AI, Andrew weaves an unexpected yet common thread between these, and explore how oneAPI builds on this foundation. Andrew talks about Codeplay’s contributions to oneAPI, focused on enabling development on Nvidia GPUs, while Geoff shares his vision of enabling broad data parallelism and portability across platforms, to provide a glimpse into what’s next for oneAPI.
In this episode of Code Together, we talk about how the industry is uniting to address the need for programming portability and performance across diverse architectures- particularly important with the rise of data-intensive workloads like AI and machine learning. We discuss the shift to modern C++ programming models and how the cross-industry oneAPI initiative and DPC++, bring much-needed portable performance to developers.