Statements (23)
| Predicate | Object |
|---|---|
| gptkbp:instanceOf |
GPU programming feature
|
| gptkbp:API |
gptkb:CUDA_Graph_API
|
| gptkbp:benefit |
improved performance
better scheduling of GPU tasks lower CPU utilization |
| gptkbp:category |
parallel computing
|
| gptkbp:component |
gptkb:CUDA_Toolkit
|
| gptkbp:documentation |
https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#cuda-graphs
|
| gptkbp:enables |
recording and replaying GPU work
|
| https://www.w3.org/2000/01/rdf-schema#label |
CUDA Graphs
|
| gptkbp:introduced |
gptkb:NVIDIA
|
| gptkbp:introducedIn |
gptkb:CUDA_10.0
|
| gptkbp:purpose |
reduce CPU overhead in GPU workloads
|
| gptkbp:relatedTo |
gptkb:CUDA_Events
gptkb:CUDA_Kernels gptkb:CUDA_Streams |
| gptkbp:supports |
explicit graph construction
stream capture |
| gptkbp:usedIn |
deep learning
high-performance computing scientific computing |
| gptkbp:bfsParent |
gptkb:CUDA
|
| gptkbp:bfsLayer |
6
|