Malloc mode

GPU memory allocation

Description

App Configuration Pane: GPU Code

Configuration Objects: coder.gpuConfig

The Malloc Mode parameter specifies the memory allocation mode to use in the generated CUDA^® code. For more information, see Discrete and Managed Modes.

Note

Setting the Malloc Mode parameter to unified is not recommended when targeting NVIDIA^® devices on the host development computer. See unified memory allocation mode on host being removed.

Settings

discrete

This value is the default value.

Uses the cudaMalloc API to transfer data between the CPU and the GPU. This mode has a traditional memory architecture with separate CPU and GPU global memory address space.

unified

Uses the cudaMallocManaged API, which uses a unified CPU and GPU global memory address space.

Programmatic Use

Property: MallocMode

Values: "discrete" | "unified"

Default: "discrete"

Version History

Introduced in R2018a