Ben Sander, Senior AMD Fellow
In this video from SC16, Ben Sander from AMD presents: HIP and CAFFE Porting and Profiling with AMD’s ROCm.
“We are excited to present ROCm, the first open-source HPC/Hyperscale-class platform for GPU computing that’s also programming-language independent. We are bringing the UNIX philosophy of choice, minimalism and modular software development to GPU computing. The new ROCm foundation lets you choose or even develop tools and a language run time for your application. ROCm is built for scale; it supports multi-GPU computing in and out of server-node communication through RDMA. It also simplifies the stack when the driver directly incorporates RDMA peer-sync support. ROCm has a rich system run time with the critical features that large-scale application, compiler and language-run-time development requires.”
The ROCm Platform delivers new opportunities to GPU computing research. ROCm is comprised of three core foundation elements:
- New Linux Driver and Runtime Stack optimized for HPC & Ultra-scale class computing
- Heterogeneous C and C++ compilers which best address the whole system not just a single device
- HIP tool which allows for CUDA code conversion, acknowledging the need for platform choice when utilizing GPU computing API
Using our knowledge of the HSA standard and, more importantly, the HSA 1.0 Runtime, AMD has been able to successfully extend support to accelerators with critical features for NUMA class acceleration. As a result, the Radeon Open Compute Kernel (ROCK) driver is composed of several components based on our efforts to develop the Heterogeneous System Architecture for APUs, including the new AMD GPU driver, the Kernel Fusion Driver (KFD), the HSA+ Runtime and an LLVM-based compilation stack for the building of key language support. The ROCm platform now supports the FirePro S9300 x2, S9170, S9150 & W9100 graphics cards; as well as, the Radeon RX Series, R9 Nano and Pro Duo graphics cards.
