✳flâneur — a map of the web's best reading

NVIDIA/cuda-checkpoint: CUDA checkpoint and restore utility

github.com · 1,515 words · saved by 1 readers

Checkpoint and restore functionality for CUDA is exposed through a command-line utiity called cuda-checkpoint which is available in the bin directory of this repo. This utility can be used to transparently checkpoint and restore CUDA state within a running Linux process, and can be combined with CRIU (described below) to fully checkpoint CUDA applications. Display driver version 580 includes these features not present in 570: Demo programs show how to use GPU migration using the API and using cuda-checkpoint Display driver version 570 includes these features not present in 550: This demo program shows many of these features in action! Transparent, per-process checkpointing offers a middle ground between virtual machine checkpointing and application-driven checkpointing. Per-process checkpointing can be used in combination with containers to checkpoint the state of a complex application, facilitating uses-cases such as: The most popular utility for transparent per-process checkpointing

The cuda-checkpoint Utility Checkpoint and restore functionality for CUDA is exposed through a command-line utiity called cuda-checkpoint which is available in the bin directory of this repo. This utility can be used to transparently checkpoint and restore CUDA state within a running Linux process, and can be combined with CRIU (described below) to fully checkpoint CUDA applications. 610 Features Display driver version 610 includes this feature not present in 595: cuIpcGetMemHandle-based CUDA IPC support A demo program shows the two ways that this IPC support can be enabled: by launching the a

Explore this link on the map →

related reading