Inheritance diagram for ProcessAcceleratorAlpaka.ProcessAcceleratorAlpaka:

Public Member Functions
def	__init__ (self)

def	apply (self, process, accelerators)

def	moduleTypeResolver (self, accelerators)

def	setBackend (self, backend)

Private Attributes
	_backend

Detailed Description

ProcessAcceleratorAlpaka itself does not define or inspect
availability of any accelerator devices. It merely sets up
necessary Alpaka infrastructure based on the availability of
accelerators that the concrete ProcessAccelerators (like
ProcessAcceleratorCUDA) define.

Definition at line 46 of file ProcessAcceleratorAlpaka.py.

Constructor & Destructor Documentation

◆ init()

def ProcessAcceleratorAlpaka.ProcessAcceleratorAlpaka.__init__ ( self )

Definition at line 53 of file ProcessAcceleratorAlpaka.py.

     def __init__(self):
         super(ProcessAcceleratorAlpaka, self).__init__()
         self._backend = None
 

Member Function Documentation

◆ apply()

def ProcessAcceleratorAlpaka.ProcessAcceleratorAlpaka.apply	(	self,
		process,
		accelerators
	)

Definition at line 65 of file ProcessAcceleratorAlpaka.py.

Referenced by heavyIonTools.ConfigureHeavyIons.__call__(), coreTools.RunOnData.__call__(), trackTools.MakeAODTrackCandidates.__call__(), runJetUncertainties.RunJetUncertainties.__call__(), metTools.AddMETCollection.__call__(), heavyIonTools.ProductionDefaults.__call__(), cmsswVersionTools.PickRelValInputFiles.__call__(), coreTools.RemoveMCMatching.__call__(), trackTools.MakePATTrackCandidates.__call__(), trigTools.SwitchOnTrigger.__call__(), runMETCorrectionsAndUncertainties.RunMETCorrectionsAndUncertainties.__call__(), heavyIonTools.SelectionDefaults.__call__(), heavyIonTools.DisbaleMonteCarloDeps.__call__(), trigTools.SwitchOnTriggerStandAlone.__call__(), tauTools.AddTauCollection.__call__(), trackTools.MakeTrackCandidates.__call__(), trigTools.SwitchOnTriggerMatching.__call__(), trigTools.SwitchOnTriggerMatchingStandAlone.__call__(), trigTools.SwitchOnTriggerMatchEmbedding.__call__(), jetTools.AddJetCollection.__call__(), jetTools.SwitchJetCollection.__call__(), jetTools.UpdateJetCollection.__call__(), jetTools.AddJetID.__call__(), and jetTools.SetTagInfos.__call__().

     def apply(self, process, accelerators):
         # Propagate the AlpakaService messages through the MessageLogger
         if not hasattr(process.MessageLogger, "AlpakaService"):
             process.MessageLogger.AlpakaService = cms.untracked.PSet()
 
         # Check if the CPU backend is available
         try:
             if not "cpu" in accelerators:
                 raise False
             from HeterogeneousCore.AlpakaServices.AlpakaServiceSerialSync_cfi import AlpakaServiceSerialSync
         except:
             # the CPU backend is not available, do not load the AlpakaServiceSerialSync
             if hasattr(process, "AlpakaServiceSerialSync"):
                 del process.AlpakaServiceSerialSync
         else:
             # the CPU backend is available, ensure the AlpakaServiceSerialSync is loaded
             if not hasattr(process, "AlpakaServiceSerialSync"):
                 process.add_(AlpakaServiceSerialSync)
 
         # Check if CUDA is available, and if the system has at least one usable NVIDIA GPU
         try:
             if not "gpu-nvidia" in accelerators:
                 raise False
             from HeterogeneousCore.AlpakaServices.AlpakaServiceCudaAsync_cfi import AlpakaServiceCudaAsync
         except:
             # CUDA is not available, do not load the AlpakaServiceCudaAsync
             if hasattr(process, "AlpakaServiceCudaAsync"):
                 del process.AlpakaServiceCudaAsync
         else:
             # CUDA is available, ensure the AlpakaServiceCudaAsync is loaded
             if not hasattr(process, "AlpakaServiceCudaAsync"):
                 process.add_(AlpakaServiceCudaAsync)
 
         # Check if ROCm is available, and if the system has at least one usable AMD GPU
         try:
             if not "gpu-amd" in accelerators:
                 raise False
             from HeterogeneousCore.AlpakaServices.AlpakaServiceROCmAsync_cfi import AlpakaServiceROCmAsync
         except:
             # ROCm is not available, do not load the AlpakaServiceROCmAsync
             if hasattr(process, "AlpakaServiceROCmAsync"):
                 del process.AlpakaServiceROCmAsync
         else:
             # ROCm is available, ensure the AlpakaServiceROCmAsync is loaded
             if not hasattr(process, "AlpakaServiceROCmAsync"):
                 process.add_(AlpakaServiceROCmAsync)
 
 
 # Ensure this module is kept in the configuration when dumping it
 cms.specialImportRegistry.registerSpecialImportForType(ProcessAcceleratorAlpaka, "from HeterogeneousCore.AlpakaCore.ProcessAcceleratorAlpaka import ProcessAcceleratorAlpaka")

◆ moduleTypeResolver()

def ProcessAcceleratorAlpaka.ProcessAcceleratorAlpaka.moduleTypeResolver	(	self,
		accelerators
	)

Definition at line 62 of file ProcessAcceleratorAlpaka.py.

References ProcessAcceleratorAlpaka.ProcessAcceleratorAlpaka._backend.

     def moduleTypeResolver(self, accelerators):
         return ModuleTypeResolverAlpaka(accelerators, self._backend)
 

◆ setBackend()

def ProcessAcceleratorAlpaka.ProcessAcceleratorAlpaka.setBackend	(	self,
		backend
	)

Definition at line 58 of file ProcessAcceleratorAlpaka.py.

References ProcessAcceleratorAlpaka.ProcessAcceleratorAlpaka._backend.

     def setBackend(self, backend):
         self._backend = backend
 

Member Data Documentation

◆ _backend

ProcessAcceleratorAlpaka.ProcessAcceleratorAlpaka._backend

private

Definition at line 55 of file ProcessAcceleratorAlpaka.py.

Referenced by ProcessAcceleratorAlpaka.ProcessAcceleratorAlpaka.moduleTypeResolver(), and ProcessAcceleratorAlpaka.ProcessAcceleratorAlpaka.setBackend().

Public Member Functions

Private Attributes