Main Content

Model Configuration Parameters: GPU Code

The Code Generation > GPU Code category includes parameters for configuring GPU-specific settings of the generated code.

These parameters require a GPU Coder™ license.

On the Configuration Parameters dialog box, the following configuration parameters are on the Code Generation > GPU Code pane.

ParameterDescription

GPU index

Specify the CUDA® device to target.

Compute capability

Specify the minimum compute capability for which CUDA code is generated.

Custom compute capability

Specify the name of the NVIDIA® virtual GPU architecture for code generation.

Memory mode

Specify the Memory allocation (malloc) mode to be used in the generated CUDA code.

Maximum blocks per kernel

Specify the maximum number of CUDA blocks created during a kernel launch.

Dynamic memory allocation threshold

Specify the size above which the private variables are allocated on the heap instead of the stack.

Memory Manager

Select the GPU memory manager for efficient memory allocation, management, and improving run-time performance.

Stack size per GPU thread

Specify the maximum stack limit per GPU thread.

Include error checks in generated code

Add run-time error-checking functionality to the generated CUDA code.

Kernel name prefix

Specify custom kernel name prefixes.

Additional compiler flags

Specify additional flags to the nvcc compiler.

cuBLAS

Replacement of math function calls with NVIDIA cuBLAS library calls.

cuSOLVER

Replacement of math function calls with NVIDIA cuSOLVER library calls.

cuFFT

Replacement of math function calls with NVIDIA cuFFT library calls.

Related Topics