validateGPU

Validate GPU devices

Since R2024b

Syntax

validateGPU

validateGPU("all")

validateGPU(ind)

Description

validateGPU validates the currently selected GPU device. If no GPU device is selected, then the function validates the default device.

validateGPU("all") validates all GPU devices detected in your system.

validateGPU(ind) validates the GPU devices specified by indices ind.

Tip

Validating your GPU device is not required. Use the validateGPU function to diagnose problems with your GPU setup.

Examples

Validate GPU Device

Validate the currently selected GPU device. If no GPU device is selected, the function validates the default device. The validateGPU function prints the validation steps performed and an indication of whether each step passed, failed, or was skipped. If a validation step fails, the function prints a diagnostic message.

validateGPU

# Beginning GPU validation
# Performing system validation
#    CUDA-supported platform .................................................PASSED
#    CUDA-enabled graphics driver exists .....................................PASSED
#        Version: 537.70
#    CUDA-enabled graphics driver load .......................................PASSED
#    CUDA environment variables ..............................................PASSED
#    CUDA device count .......................................................PASSED
#        Found 2 devices.
#    GPU libraries load ......................................................PASSED
# 
# Performing device validation for device index 1
#    Device exists ...........................................................PASSED
#        NVIDIA RTX A5000
#    Device supported ........................................................PASSED
#    Device available ........................................................PASSED
#        Device is in 'Default' compute mode.
#    Device selectable .......................................................PASSED
#    Device memory allocation ................................................PASSED
#    Device kernel launch ....................................................PASSED
# 
# Finished GPU validation with no failures.

Validate All Local GPU Devices

Validate all GPU devices present in your system. The validateGPU function performs a series of system validation checks once, and then a series of device validation steps once for each device detected.

validateGPU("all")

# Beginning GPU validation
# Performing system validation
#    CUDA-supported platform .................................................PASSED
#    CUDA-enabled graphics driver exists .....................................PASSED
#        Version: 570.133.07
#    CUDA-enabled graphics driver load .......................................PASSED
#    CUDA environment variables ..............................................PASSED
#        CUDA_VISIBLE_DEVICES: "0,1"
#    CUDA device count .......................................................PASSED
#        Found 2 devices.
#    GPU libraries load ......................................................PASSED
# 
# Performing device validation for device index 1
#    Device exists ...........................................................PASSED
#        NVIDIA RTX A5000
#    Device supported ........................................................PASSED
#    Device available ........................................................PASSED
#        Device is in 'Default' compute mode.
#    Device selectable .......................................................PASSED
#    Device memory allocation ................................................PASSED
#    Device kernel launch ....................................................PASSED
# 
# Performing device validation for device index 2
#    Device exists ...........................................................PASSED
#        NVIDIA RTX A5000
#    Device supported ........................................................PASSED
#    Device available ........................................................PASSED
#        Device is in 'Default' compute mode.
#    Device selectable .......................................................PASSED
#    Device memory allocation ................................................PASSED
#    Device kernel launch ....................................................PASSED
# 
# Finished GPU validation with no failures.

Choose GPU Device to Validate

Validate the GPU device with index 2.

validateGPU(2)

# Beginning GPU validation
# Performing system validation
#    CUDA-supported platform .................................................PASSED
#    CUDA-enabled graphics driver exists .....................................PASSED
#        Version: 570.133.07
#    CUDA-enabled graphics driver load .......................................PASSED
#    CUDA environment variables ..............................................PASSED
#        CUDA_VISIBLE_DEVICES: "0,1"
#    CUDA device count .......................................................PASSED
#        Found 2 devices.
#    GPU libraries load ......................................................PASSED
# 
# Performing device validation for device index 2
#    Device exists ...........................................................PASSED
#        NVIDIA RTX A5000
#    Device supported ........................................................PASSED
#    Device available ........................................................PASSED
#        Device is in 'Default' compute mode.
#    Device selectable .......................................................PASSED
#    Device memory allocation ................................................PASSED
#    Device kernel launch ....................................................PASSED
# 
# Finished GPU validation with no failures.

If you have multiple GPU devices, then you can validate several using a single call to validateGPU by passing a vector of device indices as input. For example, to validate GPU devices 1 and 2, call validateGPU([1 2]).

Input Arguments

ind — Indices of GPU devices

integer | vector of integers

Indices of GPU devices to validate, specified as an integer or vector of integers in the range 1 to gpuDeviceCount.

Example: 2

Example: [1 3]

Data Types: single | double | int8 | int16 | int32 | int64 | uint8 | uint16 | uint32 | uint64

Tips

• To validate remote GPUs, start a parallel pool using your cluster, then call validateGPU inside an spmd block. If the workers have more than one GPU, use the "all" option.

  cluster = parcluster("myRemoteGPUCluster");
  pool = parpool(cluster);
  
  spmd
    validateGPU("all")
  end

  For more information about using remote GPUs, see Work with Remote GPUs.

Algorithms

List of Validation Steps

The validateGPU performs these system-level validation steps.

Validation Step	Description
CUDA-supported platform	CUDA® is supported on the operating system.
CUDA-enabled graphics driver exists	A supported CUDA driver is found.
CUDA-enabled graphics driver load	MATLAB® can load the CUDA driver.
CUDA environment variables	No devices are hidden by environment variables such as CUDA_VISIBLE_DEVICES.
CUDA device count	There is at least one device.
GPU libraries load	MATLAB can load Parallel Computing Toolbox™ GPU libraries.

The validateGPU performs these device-level validation steps. If you validate more than one device, then these steps are repeated for each device.

Validation Step	Description
Device exists	MATLAB can detect the device.
Device supported	The device's compute capability is supported by MATLAB.
Device available	The device is in an available state.
Device selectable	MATLAB can select the device.
Device memory allocation	MATLAB can create a gpuArray.
Device kernel launch	MATLAB can launch a GPU kernel.

Version History

Introduced in R2024b