gpu

Documentation

Index

• func GetAbsEnergyFromGPU() []uint32
• func GetGpus() map[int]gpu_source.Device
• func GetMIGInstances() map[int]map[int]gpu_source.Device
• func GetProcessResourceUtilizationPerDevice(device gpu_source.Device, since time.Duration) (map[uint32]gpu_source.ProcessUtilizationSample, error)
• func GetSourceName() string
• func Init() error
• func IsGPUCollectionSupported() bool
• func SetGPUCollectionSupported(supported bool)
• func Shutdown() bool

Functions

func GetAbsEnergyFromGPU

func GetAbsEnergyFromGPU() []uint32

func GetGpus

func GetGpus() map[int]gpu_source.Device

func GetMIGInstances

func GetMIGInstances() map[int]map[int]gpu_source.Device

func GetProcessResourceUtilizationPerDevice

func GetProcessResourceUtilizationPerDevice(device gpu_source.Device, since time.Duration) (map[uint32]gpu_source.ProcessUtilizationSample, error)

GetProcessResourceUtilizationPerDevice tries to collect the GPU metrics. There is a known issue that some clusters the nvidia GPU can stop to respod and we need to start it again. See https://github.com/sustainable-computing-io/kepler/issues/610.

func GetSourceName

func GetSourceName() string

func Init

func Init() error

Init() only returns the erro regarding if the gpu collector was suceffully initialized or not The gpu.go file has an init function that starts and configures the gpu collector However this file is only included in the build if kepler is run with gpus support. This is necessary because nvidia libraries are not available on all systems

func IsGPUCollectionSupported

func IsGPUCollectionSupported() bool

func SetGPUCollectionSupported

func SetGPUCollectionSupported(supported bool)

func Shutdown

func Shutdown() bool