Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add runtime allocation of PF rechit fraction SoA #46135

Merged
merged 3 commits into from
Oct 3, 2024
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 12 additions & 0 deletions HLTrigger/Configuration/python/customizeHLTforCMSSW.py
Original file line number Diff line number Diff line change
Expand Up @@ -48,6 +48,16 @@ def customiseForOffline(process):

return process

def customizeHLTfor46135(process):
"""Remove pfRecHitFractionAllocation from PFClusterSoAProducer config"""
for producer in producers_by_type(process, "PFClusterSoAProducer@alpaka"):
if hasattr(producer, 'pfRecHitFractionAllocation'):
delattr(producer, 'pfRecHitFractionAllocation')
for producer in producers_by_type(process, "alpaka_serial_sync::PFClusterSoAProducer"):
if hasattr(producer, 'pfRecHitFractionAllocation'):
delattr(producer, 'pfRecHitFractionAllocation')
return process

# CMSSW version specific customizations
def customizeHLTforCMSSW(process, menuType="GRun"):

Expand All @@ -56,4 +66,6 @@ def customizeHLTforCMSSW(process, menuType="GRun"):
# add call to action function in proper order: newest last!
# process = customiseFor12718(process)

process = customizeHLTfor46135(process)

return process
Original file line number Diff line number Diff line change
Expand Up @@ -8,55 +8,83 @@
#include "FWCore/Utilities/interface/EDGetToken.h"
#include "FWCore/Utilities/interface/InputTag.h"
#include "FWCore/Utilities/interface/StreamID.h"
#include "HeterogeneousCore/AlpakaCore/interface/alpaka/stream/EDProducer.h"
#include "HeterogeneousCore/AlpakaCore/interface/alpaka/stream/SynchronizingEDProducer.h"
#include "HeterogeneousCore/CUDACore/interface/JobConfigurationGPURecord.h"
#include "RecoParticleFlow/PFClusterProducer/interface/PFCPositionCalculatorBase.h"
#include "RecoParticleFlow/PFClusterProducer/interface/alpaka/PFClusterParamsDeviceCollection.h"
#include "RecoParticleFlow/PFClusterProducer/plugins/alpaka/PFClusterSoAProducerKernel.h"
#include "RecoParticleFlow/PFRecHitProducer/interface/PFRecHitTopologyRecord.h"

namespace ALPAKA_ACCELERATOR_NAMESPACE {
class PFClusterSoAProducer : public stream::EDProducer<> {
class PFClusterSoAProducer : public stream::SynchronizingEDProducer<> {
public:
PFClusterSoAProducer(edm::ParameterSet const& config)
: pfClusParamsToken(esConsumes(config.getParameter<edm::ESInputTag>("pfClusterParams"))),
topologyToken_(esConsumes(config.getParameter<edm::ESInputTag>("topology"))),
inputPFRecHitSoA_Token_{consumes(config.getParameter<edm::InputTag>("pfRecHits"))},
outputPFClusterSoA_Token_{produces()},
outputPFRHFractionSoA_Token_{produces()},
synchronise_(config.getParameter<bool>("synchronise")),
pfRecHitFractionAllocation_(config.getParameter<int>("pfRecHitFractionAllocation")) {}
numRHF_{cms::alpakatools::make_host_buffer<uint32_t, Platform>()},
synchronise_(config.getParameter<bool>("synchronise")) {}

void produce(device::Event& event, device::EventSetup const& setup) override {
void acquire(device::Event const& event, device::EventSetup const& setup) override {
const reco::PFClusterParamsDeviceCollection& params = setup.getData(pfClusParamsToken);
const reco::PFRecHitHCALTopologyDeviceCollection& topology = setup.getData(topologyToken_);
const reco::PFRecHitHostCollection& pfRecHits = event.get(inputPFRecHitSoA_Token_);
int nRH = 0;
if (pfRecHits->metadata().size() != 0)
nRH = pfRecHits->size();

reco::PFClusteringVarsDeviceCollection pfClusteringVars{nRH, event.queue()};
reco::PFClusteringEdgeVarsDeviceCollection pfClusteringEdgeVars{(nRH * 8), event.queue()};
reco::PFClusterDeviceCollection pfClusters{nRH, event.queue()};
reco::PFRecHitFractionDeviceCollection pfrhFractions{nRH * pfRecHitFractionAllocation_, event.queue()};
pfClusteringVars_.emplace(nRH, event.queue());
pfClusteringEdgeVars_.emplace(nRH * 8, event.queue());
pfClusters_.emplace(nRH, event.queue());

*numRHF_ = 0;

if (nRH != 0) {
PFClusterProducerKernel kernel(event.queue(), pfRecHits);
kernel.seedTopoAndContract(event.queue(),
params,
topology,
*pfClusteringVars_,
*pfClusteringEdgeVars_,
pfRecHits,
*pfClusters_,
numRHF_.data());
}
}

void produce(device::Event& event, device::EventSetup const& setup) override {
const reco::PFClusterParamsDeviceCollection& params = setup.getData(pfClusParamsToken);
const reco::PFRecHitHCALTopologyDeviceCollection& topology = setup.getData(topologyToken_);
const reco::PFRecHitHostCollection& pfRecHits = event.get(inputPFRecHitSoA_Token_);
int nRH = 0;

std::optional<reco::PFRecHitFractionDeviceCollection> pfrhFractions;

if (pfRecHits->metadata().size() != 0)
nRH = pfRecHits->size();

if (nRH != 0) {
pfrhFractions.emplace(*numRHF_.data(), event.queue());
PFClusterProducerKernel kernel(event.queue(), pfRecHits);
kernel.execute(event.queue(),
kernel.cluster(event.queue(),
params,
topology,
pfClusteringVars,
pfClusteringEdgeVars,
*pfClusteringVars_,
*pfClusteringEdgeVars_,
pfRecHits,
pfClusters,
pfrhFractions);
*pfClusters_,
*pfrhFractions);
} else {
pfrhFractions.emplace(0, event.queue());
}

if (synchronise_)
alpaka::wait(event.queue());

event.emplace(outputPFClusterSoA_Token_, std::move(pfClusters));
event.emplace(outputPFRHFractionSoA_Token_, std::move(pfrhFractions));
event.emplace(outputPFClusterSoA_Token_, std::move(*pfClusters_));
event.emplace(outputPFRHFractionSoA_Token_, std::move(*pfrhFractions));
}

static void fillDescriptions(edm::ConfigurationDescriptions& descriptions) {
Expand All @@ -65,7 +93,6 @@ namespace ALPAKA_ACCELERATOR_NAMESPACE {
desc.add<edm::ESInputTag>("pfClusterParams", edm::ESInputTag(""));
desc.add<edm::ESInputTag>("topology", edm::ESInputTag(""));
desc.add<bool>("synchronise", false);
desc.add<int>("pfRecHitFractionAllocation", 120);
descriptions.addWithDefaultLabel(desc);
}

Expand All @@ -75,8 +102,11 @@ namespace ALPAKA_ACCELERATOR_NAMESPACE {
const edm::EDGetTokenT<reco::PFRecHitHostCollection> inputPFRecHitSoA_Token_;
const device::EDPutToken<reco::PFClusterDeviceCollection> outputPFClusterSoA_Token_;
const device::EDPutToken<reco::PFRecHitFractionDeviceCollection> outputPFRHFractionSoA_Token_;
cms::alpakatools::host_buffer<uint32_t> numRHF_;
std::optional<reco::PFClusteringVarsDeviceCollection> pfClusteringVars_;
std::optional<reco::PFClusteringEdgeVarsDeviceCollection> pfClusteringEdgeVars_;
std::optional<reco::PFClusterDeviceCollection> pfClusters_;
const bool synchronise_;
const int pfRecHitFractionAllocation_;
};

} // namespace ALPAKA_ACCELERATOR_NAMESPACE
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -1090,7 +1090,6 @@ namespace ALPAKA_ACCELERATOR_NAMESPACE {
const reco::PFRecHitHCALTopologyDeviceCollection::ConstView topology,
const reco::PFRecHitHostCollection::ConstView pfRecHits,
reco::PFClusterDeviceCollection::View clusterView,
reco::PFRecHitFractionDeviceCollection::View fracView,
uint32_t* __restrict__ nSeeds) const {
const int nRH = pfRecHits.size();

Expand Down Expand Up @@ -1199,7 +1198,8 @@ namespace ALPAKA_ACCELERATOR_NAMESPACE {
const reco::PFRecHitHostCollection::ConstView pfRecHits,
reco::PFClusteringVarsDeviceCollection::View pfClusteringVars,
reco::PFClusterDeviceCollection::View clusterView,
uint32_t* __restrict__ nSeeds) const {
uint32_t* __restrict__ nSeeds,
uint32_t* __restrict__ nRHF) const {
const int nRH = pfRecHits.size();
int& totalSeedOffset = alpaka::declareSharedVar<int, __COUNTER__>(acc);
int& totalSeedFracOffset = alpaka::declareSharedVar<int, __COUNTER__>(acc);
Expand Down Expand Up @@ -1302,6 +1302,7 @@ namespace ALPAKA_ACCELERATOR_NAMESPACE {
pfClusteringVars.pcrhFracSize() = totalSeedFracOffset;
pfClusteringVars.nRHFracs() = totalSeedFracOffset;
clusterView.nRHFracs() = totalSeedFracOffset;
*nRHF = totalSeedFracOffset;
clusterView.nSeeds() = *nSeeds;
clusterView.nTopos() = pfClusteringVars.nTopos();

Expand Down Expand Up @@ -1467,14 +1468,14 @@ namespace ALPAKA_ACCELERATOR_NAMESPACE {
alpaka::memset(queue, nSeeds, 0x00); // Reset nSeeds
}

void 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) {
void PFClusterProducerKernel::seedTopoAndContract(Queue& queue,
const reco::PFClusterParamsDeviceCollection& params,
const reco::PFRecHitHCALTopologyDeviceCollection& topology,
reco::PFClusteringVarsDeviceCollection& pfClusteringVars,
reco::PFClusteringEdgeVarsDeviceCollection& pfClusteringEdgeVars,
const reco::PFRecHitHostCollection& pfRecHits,
reco::PFClusterDeviceCollection& pfClusters,
uint32_t* __restrict__ nRHF) {
const int nRH = pfRecHits->size();
const int threadsPerBlock = 256;
const int blocks = divide_up_by(nRH, threadsPerBlock);
Expand All @@ -1488,7 +1489,6 @@ namespace ALPAKA_ACCELERATOR_NAMESPACE {
topology.view(),
pfRecHits.view(),
pfClusters.view(),
pfrhFractions.view(),
nSeeds.data());
// prepareTopoInputs
alpaka::exec<Acc1D>(queue,
Expand Down Expand Up @@ -1524,7 +1524,19 @@ namespace ALPAKA_ACCELERATOR_NAMESPACE {
pfRecHits.view(),
pfClusteringVars.view(),
pfClusters.view(),
nSeeds.data());
nSeeds.data(),
nRHF);
}

void PFClusterProducerKernel::cluster(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) {
const int nRH = pfRecHits->size();

// fillRhfIndex
alpaka::exec<Acc2D>(queue,
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -39,7 +39,16 @@ namespace ALPAKA_ACCELERATOR_NAMESPACE {
public:
PFClusterProducerKernel(Queue& queue, const reco::PFRecHitHostCollection& pfRecHits);

void execute(Queue& queue,
void seedTopoAndContract(Queue& queue,
const reco::PFClusterParamsDeviceCollection& params,
const reco::PFRecHitHCALTopologyDeviceCollection& topology,
reco::PFClusteringVarsDeviceCollection& pfClusteringVars,
reco::PFClusteringEdgeVarsDeviceCollection& pfClusteringEdgeVars,
const reco::PFRecHitHostCollection& pfRecHits,
reco::PFClusterDeviceCollection& pfClusters,
uint32_t* __restrict__ nRHF);

void cluster(Queue& queue,
const reco::PFClusterParamsDeviceCollection& params,
const reco::PFRecHitHCALTopologyDeviceCollection& topology,
reco::PFClusteringVarsDeviceCollection& pfClusteringVars,
Expand Down