Accelerator-Docker : provides common interface for automatic passthrough of heterogeneous hardware accelerators in docker
-
Updated
Oct 21, 2019 - C++
Accelerator-Docker : provides common interface for automatic passthrough of heterogeneous hardware accelerators in docker
A Metal-based CUDA compatibility framework for running Numba CUDA workloads on Apple Silicon.
Flux→PTX distributed GPU runtime. Evolving from Rust-to-PTX to agent-native Flux→MIR→PTX pipeline with cudaclaw, SmartCRDT, git-native constructs, fleet coordination. 6 new crates in the oxide stack.
CuTileForge v3.0.0: Native tile-fusion compiler & universal quantization engine on NVIDIA CuTile. Zero-ATen CUDA-graph decode at 172 tok/s, 2,377 tok/s batch serving, 13 model families, 70B hybrid on a single 32 GB GPU (sm_89 ↔ sm_120).
Modern C++20 host-side heterogeneous GPU runtime: coalescing aligned device memory pool, cache-line-isolated lock-free SPSC command queue, multi-stream DAG task scheduler, strided N-D tensor views with AVX2/scalar tiled SGEMM, and Chrome Trace profiler.
Transparent suspend/resume runtime enabling preemptible GPU workloads via memory snapshotting, UVM paging, and execution state orchestration.
To associate your repository with the gpu-runtime topic, visit your repo's landing page and select "manage topics."