#include <PFClusterSoAProducerKernel.h>

Public Member Functions
void	execute (Queue &queue, const reco::PFClusterParamsDeviceCollection &params, const reco::PFRecHitHCALTopologyDeviceCollection &topology, reco::PFClusteringVarsDeviceCollection &pfClusteringVars, reco::PFClusteringEdgeVarsDeviceCollection &pfClusteringEdgeVars, const reco::PFRecHitHostCollection &pfRecHits, reco::PFClusterDeviceCollection &pfClusters, reco::PFRecHitFractionDeviceCollection &pfrhFractions)

	PFClusterProducerKernel (Queue &queue, const reco::PFRecHitHostCollection &pfRecHits)

Private Attributes
cms::alpakatools::device_buffer< Device, float[]>	globalClusterEnergy

cms::alpakatools::device_buffer< Device, reco::pfClustering::Position4[]>	globalClusterPos

cms::alpakatools::device_buffer< Device, reco::pfClustering::Position4[]>	globalPrevClusterPos

cms::alpakatools::device_buffer< Device, int[]>	globalRechits

cms::alpakatools::device_buffer< Device, float[]>	globalRhFracSum

cms::alpakatools::device_buffer< Device, int[]>	globalSeeds

cms::alpakatools::device_buffer< Device, uint32_t >	nSeeds

Detailed Description

Definition at line 38 of file PFClusterSoAProducerKernel.h.

Constructor & Destructor Documentation

◆ PFClusterProducerKernel()

ALPAKA_ACCELERATOR_NAMESPACE::PFClusterProducerKernel::PFClusterProducerKernel	(	Queue &	queue,
		const reco::PFRecHitHostCollection &	pfRecHits
	)

Definition at line 1456 of file PFClusterSoAProducerKernel.dev.cc.

References nSeeds, and createBeamHaloJobs::queue.

       : nSeeds(cms::alpakatools::make_device_buffer<uint32_t>(queue)),
         globalClusterPos(
             cms::alpakatools::make_device_buffer<Position4[]>(queue, blocksForExoticClusters * maxTopoInput)),
         globalPrevClusterPos(
             cms::alpakatools::make_device_buffer<Position4[]>(queue, blocksForExoticClusters * maxTopoInput)),
         globalClusterEnergy(
             cms::alpakatools::make_device_buffer<float[]>(queue, blocksForExoticClusters * maxTopoInput)),
         globalRhFracSum(cms::alpakatools::make_device_buffer<float[]>(queue, blocksForExoticClusters * maxTopoInput)),
         globalSeeds(cms::alpakatools::make_device_buffer<int[]>(queue, blocksForExoticClusters * maxTopoInput)),
         globalRechits(cms::alpakatools::make_device_buffer<int[]>(queue, blocksForExoticClusters * maxTopoInput)) {
     alpaka::memset(queue, nSeeds, 0x00);  // Reset nSeeds
   }

Member Function Documentation

◆ execute()

void ALPAKA_ACCELERATOR_NAMESPACE::PFClusterProducerKernel::execute	(	Queue &	queue,
		const reco::PFClusterParamsDeviceCollection &	params,
		const reco::PFRecHitHCALTopologyDeviceCollection &	topology,
		reco::PFClusteringVarsDeviceCollection &	pfClusteringVars,
		reco::PFClusteringEdgeVarsDeviceCollection &	pfClusteringEdgeVars,
		const reco::PFRecHitHostCollection &	pfRecHits,
		reco::PFClusterDeviceCollection &	pfClusters,
		reco::PFRecHitFractionDeviceCollection &	pfrhFractions
	)

Definition at line 1470 of file PFClusterSoAProducerKernel.dev.cc.

References gather_cfg::blocks, ALPAKA_ACCELERATOR_NAMESPACE::blocksForExoticClusters, cms::alpakatools::divide_up_by(), globalClusterEnergy, globalClusterPos, globalPrevClusterPos, globalRechits, globalRhFracSum, globalSeeds, nSeeds, submitPVValidationJobs::params, HLT_2024v12_cff::pfRecHits, createBeamHaloJobs::queue, ALPAKA_ACCELERATOR_NAMESPACE::threadsPerBlockForClustering, and HLT_2024v12_cff::topology.

Referenced by ALPAKA_ACCELERATOR_NAMESPACE::PFClusterSoAProducer::produce().

                                                                                              {
     const int nRH = pfRecHits->size();
     const int threadsPerBlock = 256;
     const int blocks = divide_up_by(nRH, threadsPerBlock);
 
     // seedingTopoThresh
     alpaka::exec<Acc1D>(queue,
                         make_workdiv<Acc1D>(blocks, threadsPerBlock),
                         SeedingTopoThresh{},
                         pfClusteringVars.view(),
                         params.view(),
                         topology.view(),
                         pfRecHits.view(),
                         pfClusters.view(),
                         pfrhFractions.view(),
                         nSeeds.data());
     // prepareTopoInputs
     alpaka::exec<Acc1D>(queue,
                         make_workdiv<Acc1D>(blocks, threadsPerBlock),
                         PrepareTopoInputs{},
                         pfRecHits.view(),
                         pfClusteringVars.view(),
                         pfClusteringEdgeVars.view(),
                         nSeeds.data());
     // ECLCC
     alpaka::exec<Acc1D>(queue,
                         make_workdiv<Acc1D>(blocks, threadsPerBlock),
                         ECLCCInit{},
                         pfRecHits.view(),
                         pfClusteringVars.view(),
                         pfClusteringEdgeVars.view());
     alpaka::exec<Acc1D>(queue,
                         make_workdiv<Acc1D>(blocks, threadsPerBlock),
                         ECLCCCompute1{},
                         pfRecHits.view(),
                         pfClusteringVars.view(),
                         pfClusteringEdgeVars.view());
     alpaka::exec<Acc1D>(queue,
                         make_workdiv<Acc1D>(blocks, threadsPerBlock),
                         ECLCCFlatten{},
                         pfRecHits.view(),
                         pfClusteringVars.view(),
                         pfClusteringEdgeVars.view());
     // topoClusterContraction
     alpaka::exec<Acc1D>(queue,
                         make_workdiv<Acc1D>(1, threadsPerBlockForClustering),
                         TopoClusterContraction{},
                         pfRecHits.view(),
                         pfClusteringVars.view(),
                         pfClusters.view(),
                         nSeeds.data());
 
     // fillRhfIndex
     alpaka::exec<Acc2D>(queue,
                         make_workdiv<Acc2D>({divide_up_by(nRH, 32), divide_up_by(nRH, 32)}, {32, 32}),
                         FillRhfIndex{},
                         pfRecHits.view(),
                         pfClusteringVars.view(),
                         pfrhFractions.view());
 
     // Run fastCluster
     alpaka::exec<Acc1D>(queue,
                         make_workdiv<Acc1D>(nRH, threadsPerBlockForClustering),
                         FastCluster{},
                         pfRecHits.view(),
                         params.view(),
                         topology.view(),
                         pfClusteringVars.view(),
                         pfClusters.view(),
                         pfrhFractions.view());
     // exotic clustering kernel
     alpaka::exec<Acc1D>(queue,
                         make_workdiv<Acc1D>(blocksForExoticClusters,
                                             threadsPerBlockForClustering),  // uses 4 blocks to minimize memory usage
                         FastClusterExotic{},
                         pfRecHits.view(),
                         params.view(),
                         topology.view(),
                         pfClusteringVars.view(),
                         pfClusters.view(),
                         pfrhFractions.view(),
                         globalClusterPos.data(),
                         globalPrevClusterPos.data(),
                         globalClusterEnergy.data(),
                         globalRhFracSum.data(),
                         globalSeeds.data(),
                         globalRechits.data());
   }