As excessive efficiency computing (HPC) continues to deal with probably the most intricate real-world issues in science and trade, AMD platforms are driving the highly effective compute, effectivity, and adaptableness necessities for giant scale HPC deployment.
ROCm™ is the AMD open supply software program stack designed for GPU computing that consists of a set of drivers, improvement instruments, and APIs that allow GPU programming from low-level kernel to end-user functions.
The ROCm platform is suitable with lots of the hottest HPC programming languages and frameworks, together with those who parallelize operations throughout a number of GPUs and servers, deal with reminiscence hierarchies, and remedy linear methods.
AMD invitations HPC fanatics to hitch a webinar devoted to the most recent model of this software program, ROCm 6.0, that may happen on Thursday, July 17 at 11:00 a.m. EDT. The most recent ROCm updates might be introduced by Joe Schoonover, Founder & Analysis Software program Engineer at Fluid Numerics.
Discover New ROCm Updates Supporting AMD Intuition™ MI300 GPUs
The 2-part webinar’s first section will delve into current updates launched in ROCm 6.0.0 and 6.0.2 associated to MI300 assist and adjustments to supported working methods and GPUs. The webinar may even analyze efficiency implications for these transitioning from ROCm 5.7 to six.0 with a particular give attention to the AMD CDNA™ 2 structure for real-world functions such because the Spectral Factor Library in Fortran, an API for fixing partial differential equations.
AMD says the ROCm 6.0.x enhancements are primarily geared toward bolstering assist for the corporate’s AMD CDNA 3 structure, the devoted compute structure underlying its AMD Intuition MI300 Collection accelerators. The webinar will examine CDNA 2 and CDNA 3 architectures, together with APU (MI300A) and discrete GPU (MI300X) architectures, with an emphasis on reminiscence areas and the potential nuances associated to reminiscence administration, significantly with MI300A.
Driving Portability with HIP
ROCm helps to deal with the widespread HPC constraints of vendor lock-in and ease of migration with its emphasis on portability. With ROCm, customers can port CUDA®-based code into the AMD native, open supply Heterogeneous Computing Interface for Portability (HIP), permitting builders to run their code in each NVIDIA and AMD environments with minimal adjustments.
The second a part of the webinar will take a deep dive into HIP and its significance throughout the ROCm ecosystem. The presentation will define the historic evolution of ROCm regarding HIP and the varied AMD compiler aliases which can information customers in facilitating portability of HPC functions throughout GPUs from numerous distributors.
Dr. Schoonover and different HPC specialists from AMD may even be accessible on the finish of the webinar for a stay Q&A with attendees.
This fascinating webinar provides a novel alternative to remain forward of the curve as AMD redefines the HPC panorama with its versatile method to versatile and environment friendly computing, which is paving the way in which for the groundbreaking analysis and discovery the HPC group is understood for. Join the July 17 webinar at present at this hyperlink.
Be taught with the AMD HPC Person Discussion board
This thrilling ROCm webinar is introduced in partnership with the AMD HPC Person Discussion board, a member-led group of shoppers and customers of AMD processor-powered HPC methods. Members can share technical content material, experiences, and finest practices on a discussion board that gives them the chance for direct reference to AMD management and to affect future AMD HPC merchandise and options. Be taught extra concerning the AMD HPC Person Discussion board right here.