https://lesitedefrancois.be/en/tags/rocm/Recent content in Rocm on Le Site de FrançoisHugo -- gohugo.ioen© 2026 Françoishttps://lesitedefrancois.be/en/ai/rocm/Mon, 01 Jan 0001 00:00:00 +0000https://lesitedefrancois.be/en/ai/rocm/<p><strong>ROCm (Radeon Open Compute)</strong> is AMD’s software stack for <strong>GPU compute</strong> on datacenter <strong>Instinct</strong> accelerators (and select consumer GPUs in community setups). Its objective mirrors <strong>CUDA</strong> for NVIDIA: provide kernel compilers (<strong>HIP</strong>), math libraries (rocBLAS, rocFFT), collective communication (<strong>RCCL</strong>, analogous to <strong>NCCL</strong>), and framework integrations so <strong>PyTorch</strong> and inference runtimes can execute training and <strong>inference</strong> on AMD hardware. ROCm is positioned as an open platform (Linux-first) for customers who want accelerator choice or who standardize on AMD in HPC and AI clusters.</p>