Skip to content

4.19 Extended GPU Memory ​

待翻译

本节已建立目录框架,正文尚未翻译。

4.19.1 Preliminaries ​

4.19.1.1 EGM Platforms: System topology ​

4.19.1.2 Socket Identifiers: What are they? How to access them? ​

4.19.1.3 Allocators and EGM support ​

4.19.1.4 Memory management extensions to current APIs ​

4.19.2 Using the EGM Interface ​

4.19.2.1 Single-Node, Single-GPU ​

4.19.2.2 Single-Node, Multi-GPU ​

4.19.2.2.1 Using VMM APIs ​

4.19.2.2.2 Using CUDA Memory Pool ​

4.19.2.3 Multi-Node, Multi-GPU ​