Revolutionizing Large-Scale Heterogeneous HPC Systems with AMD's ROCm Platform
SessionGPU's
Session ChairPhilip C. Roth
Presenter
Event Type
Exhibitor Forum
Accelerators
Architectures
TimeWednesday, November 16th3:30pm - 4pm
Location155-E
DescriptionDuring this discussion, AMD will talk about its revolutionary approach to heterogeneous compute for HPC with their Radeon Open Compute Platform (ROCm). AMD’s new approach to GPU computing is open source from the ground-up and optimized for node and rack scaling with multi-GPUs. ROCm provides users with a simplified path to GPU acceleration using main stream standard languages based on the rich single source ISO C++ 11/14 HCC compiler, which is optimized for scaling with accelerators. This talk will also cover how users can port codes developed under CUDA over to this new approach with AMD’s HIP tool, providing users with an open path to choice in hardware.
The stage is set with AMD’s ROCm.
The stage is set with AMD’s ROCm.
Presenter
