If you're an HPC-focused developer, you likely program and use AI applications that run solely on high-end CPUs even if the compute power across multiple processing units might be a better performance fit.
So how can you simplify and expedite use of both a CPU and GPU on Xe Architecture?
One way is by using OpenMP* coupled with optimized Intel® C++ and Intel® Fortran Compilers.
Join senior principal engineer Xinmin Tian to unpack the possibilities of this CPU to GPU offloading model, including:
- A brief overview of the OpenMP 5.0 standard
- A closer look into LLVM compiler technology from Intel for OpenMP offloading support
- The performance impact on SPEC ACCEL* OpenMP workloads and kernels with different code-generation schemes
Get the Software
Download the Intel® oneAPI HPC Toolkit that contains six optimized tools and libraries for high-performance computing including Intel C++ and Intel Fortran compilers.
- Exponential Thinking: Xe Architecture
- The 7-Nanometer (NM) Fabrication Process for GPUs on Xe Architecture
Senior principal engineer and compiler architect, Intel Corporation
Xinmin promotes compiler OpenMP, offloading, vectorization, and parallelization technologies in current and future Intel® architecture. His current focus is on programming languages for oneAPI toolkits for CPUs and accelerators with Xe Architecture, compilers, and application performance tuning. Xinmin holds 27 US patents, has authored over 60 technical papers, and coauthored three books that span his expertise. Xinmin holds a PhD from the University of San Francisco.