Anurag Dogra·Sep 4NCCL’s Tree Topology: How GPUs Reduce and Broadcast Data at ScaleIf you’ve trained a large model across dozens or hundreds of GPUs, you’ve relied on NCCL (NVIDIA Collective Communication Library). NCCL…
Anurag Dogra·Sep 2NCCL: The Math Behind Pointerless Binary Trees (GPU Topologies)Zero Traversal, Pure Bits: Decoding NVIDIA NCCL’s Logical Binary Tree Algorithm
Anurag Dogra·Mar 25, 2025LLAMA.CPP on NVIDIA Jetson Nano: A Complete GuideRunning LLAMA.cpp on Jetson Nano 4 GB with CUDA 10.2A response icon1A response icon1