Porting HPC Applications to AMD Instinct MI300A Using Unified Memory and OpenMP

I was talking with a friend in HPC lately who said that AMD is actually quite competitive in the HPC space these days. For example, Frontier (<a href="https:&#x2F;&#x2F;docs.olcf.ornl.gov&#x2F;systems&#x2F;frontier_user_guide.html" rel="nofollow">https:&#x2F;&#x2F;docs.olcf.ornl.gov&#x2F;systems&#x2F;frontier_user_guide.html</a>) is an all-AMD installation. Do scientists actually use ROCm in their code or does AMD have another programming framework for their Instinct chips?

APU’s for HPC are going to be a wild ride. Accelerated computing in shared memory. Get CPU-focused folks will actually get access to some high throughput compute accessible on the sort of timescales that we can actually reason about (the GPU is so far away).

I’ve been waiting for something like that in the HPC space for years - that’s what I wanted when HSA first came out.<p><a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Heterogeneous_System_Architecture" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Heterogeneous_System_Architect...</a>

Having looked briefly at the code I still think C++17 parallel algorithms are more ergonomic compared to OpenMP: <a href="https:&#x2F;&#x2F;rocm.blogs.amd.com&#x2F;software-tools-optimization&#x2F;hipstdpar&#x2F;README.html" rel="nofollow">https:&#x2F;&#x2F;rocm.blogs.amd.com&#x2F;software-tools-optimization&#x2F;hipst...</a>