Skip to content

[PWGCF] Speed up triplet mixing in femto framework #16965

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.

Sign up for GitHub

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

Merged
ariedel-cern merged 2 commits into from
Jul 8, 2026
+45 −18
Merged

[PWGCF] Speed up triplet mixing in femto framework #16965

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
f80000d
Feat: speed up event mixing in triplet tasks
ariedel-cern Jul 8, 2026
8a144cd
Fix: add missing include
ariedel-cern Jul 8, 2026
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
63 changes: 45 additions & 18 deletions PWGCF/Femto/Core/tripletProcessHelpers.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,6 +23,7 @@
#include <Framework/Logger.h>

#include <cstdint>
#include <optional>

namespace o2::analysis::femto::tripletprocesshelpers
{
Expand Down Expand Up @@ -459,20 +460,34 @@ void processMixedEvent(T1 const& Collisions,
T10& CtrManager,
T11& TcManager)
{
int64_t lastCollisionIndex = -1;
int64_t lastCollisionIndex1 = -1;
int64_t lastCollisionIndex2 = -1;
int windowSizeRaw = 0;
int windowSizeEffective = 0;

// collision1 stays fixed across the outer mixing window, and collision2 stays fixed across each inner sub-window (it only advances once collision3 wraps
// back to its start Both slices are therefore materialized once per window/sub-window and reused, instead of being re-sliced, i.e. a fresh arrow Slice + selection copy
std::optional<decltype(Partition1->sliceByCached(o2::aod::femtobase::stored::fColId, 0, cache))> sliceParticle1;
std::optional<decltype(Partition2->sliceByCached(o2::aod::femtobase::stored::fColId, 0, cache))> sliceParticle2;

for (auto const& [collision1, collision2, collision3] : o2::soa::selfCombinations(policy, depth, -1, Collisions, Collisions, Collisions)) {

// --- new window ---
if (collision1.globalIndex() != lastCollisionIndex) {
if (lastCollisionIndex != -1) {
// outer window
if (collision1.globalIndex() != lastCollisionIndex1) {
if (lastCollisionIndex1 != -1) {
TripletHistManager.fillMixingQaMePerMixingBin(windowSizeRaw, windowSizeEffective);
}
windowSizeRaw = 0;
windowSizeEffective = 0;
lastCollisionIndex = collision1.globalIndex();
lastCollisionIndex1 = collision1.globalIndex();
lastCollisionIndex2 = -1; // force sliceParticle2 to refresh below
sliceParticle1.emplace(Partition1->sliceByCached(o2::aod::femtobase::stored::fColId, collision1.globalIndex(), cache));
}

// inner sub-window
if (collision2.globalIndex() != lastCollisionIndex2) {
lastCollisionIndex2 = collision2.globalIndex();
sliceParticle2.emplace(Partition2->sliceByCached(o2::aod::femtobase::stored::fColId, collision2.globalIndex(), cache));
}

++windowSizeRaw;
Expand All @@ -486,21 +501,19 @@ void processMixedEvent(T1 const& Collisions,

CtrManager.setMagField(collision1.magField());

auto sliceParticle1 = Partition1->sliceByCached(o2::aod::femtobase::stored::fColId, collision1.globalIndex(), cache);
auto sliceParticle2 = Partition2->sliceByCached(o2::aod::femtobase::stored::fColId, collision2.globalIndex(), cache);
auto sliceParticle3 = Partition3->sliceByCached(o2::aod::femtobase::stored::fColId, collision3.globalIndex(), cache);

TripletHistManager.resetTrackedParticlesPerEvent();

if (sliceParticle1.size() == 0 || sliceParticle2.size() == 0 || sliceParticle3.size() == 0) {
if (sliceParticle1->size() == 0 || sliceParticle2->size() == 0 || sliceParticle3.size() == 0) {
TripletHistManager.fillMixingQaMePerEvent();
continue;
}

bool hasValidTriplet = false;
TripletHistManager.fillMixingQaMe(collision1, collision2, collision3);

for (auto const& [p1, p2, p3] : o2::soa::combinations(o2::soa::CombinationsFullIndexPolicy(sliceParticle1, sliceParticle2, sliceParticle3))) {
for (auto const& [p1, p2, p3] : o2::soa::combinations(o2::soa::CombinationsFullIndexPolicy(*sliceParticle1, *sliceParticle2, sliceParticle3))) {
// pair cleaning
if (!TcManager.isCleanTriplet(p1, p2, p3, TrackTable)) {
continue;
Expand Down Expand Up @@ -528,7 +541,7 @@ void processMixedEvent(T1 const& Collisions,
TripletHistManager.fillMixingQaMePerEvent();
}

// --- final window ---
// final window
if (windowSizeRaw > 0) {
TripletHistManager.fillMixingQaMePerMixingBin(windowSizeRaw, windowSizeEffective);
}
Expand Down Expand Up @@ -563,19 +576,34 @@ void processMixedEvent(T1 const& Collisions,
T12& CtrManager,
T13& TcManager)
{
int64_t lastCollisionIndex = -1;
int64_t lastCollisionIndex1 = -1;
int64_t lastCollisionIndex2 = -1;
int windowSizeRaw = 0;
int windowSizeEffective = 0;

// collision1 stays fixed across the outer mixing window, and collision2 stays fixed across each inner sub-window (it only advances once collision3 wraps
// back to its start Both slices are therefore materialized once per window/sub-window and reused, instead of being re-sliced, i.e. a fresh arrow Slice + selection copy
std::optional<decltype(Partition1->sliceByCached(o2::aod::femtobase::stored::fColId, 0, cache))> sliceParticle1;
std::optional<decltype(Partition2->sliceByCached(o2::aod::femtobase::stored::fColId, 0, cache))> sliceParticle2;

for (auto const& [collision1, collision2, collision3] : o2::soa::selfCombinations(policy, depth, -1, Collisions, Collisions, Collisions)) {

if (collision1.globalIndex() != lastCollisionIndex) {
if (lastCollisionIndex != -1) {
// outer window
if (collision1.globalIndex() != lastCollisionIndex1) {
if (lastCollisionIndex1 != -1) {
TripletHistManager.fillMixingQaMePerMixingBin(windowSizeRaw, windowSizeEffective);
}
windowSizeRaw = 0;
windowSizeEffective = 0;
lastCollisionIndex = collision1.globalIndex();
lastCollisionIndex1 = collision1.globalIndex();
lastCollisionIndex2 = -1; // force sliceParticle2 to refresh below
sliceParticle1.emplace(Partition1->sliceByCached(o2::aod::femtobase::stored::fColId, collision1.globalIndex(), cache));
}

// inner sub-window
if (collision2.globalIndex() != lastCollisionIndex2) {
lastCollisionIndex2 = collision2.globalIndex();
sliceParticle2.emplace(Partition2->sliceByCached(o2::aod::femtobase::stored::fColId, collision2.globalIndex(), cache));
}

++windowSizeRaw;
Expand All @@ -589,21 +617,19 @@ void processMixedEvent(T1 const& Collisions,

CtrManager.setMagField(collision1.magField());

auto sliceParticle1 = Partition1->sliceByCached(o2::aod::femtobase::stored::fColId, collision1.globalIndex(), cache);
auto sliceParticle2 = Partition2->sliceByCached(o2::aod::femtobase::stored::fColId, collision2.globalIndex(), cache);
auto sliceParticle3 = Partition3->sliceByCached(o2::aod::femtobase::stored::fColId, collision3.globalIndex(), cache);

TripletHistManager.resetTrackedParticlesPerEvent();

if (sliceParticle1.size() == 0 || sliceParticle2.size() == 0 || sliceParticle3.size() == 0) {
if (sliceParticle1->size() == 0 || sliceParticle2->size() == 0 || sliceParticle3.size() == 0) {
TripletHistManager.fillMixingQaMePerEvent();
continue;
}

bool hasValidTriplet = false;
TripletHistManager.fillMixingQaMe(collision1, collision2, collision3);

for (auto const& [p1, p2, p3] : o2::soa::combinations(o2::soa::CombinationsFullIndexPolicy(sliceParticle1, sliceParticle2, sliceParticle3))) {
for (auto const& [p1, p2, p3] : o2::soa::combinations(o2::soa::CombinationsFullIndexPolicy(*sliceParticle1, *sliceParticle2, sliceParticle3))) {
// pair cleaning
if (!TcManager.isCleanTriplet(p1, p2, p3, TrackTable)) {
continue;
Expand Down Expand Up @@ -631,6 +657,7 @@ void processMixedEvent(T1 const& Collisions,
TripletHistManager.fillMixingQaMePerEvent();
}

// --- final window ---
if (windowSizeRaw > 0) {
TripletHistManager.fillMixingQaMePerMixingBin(windowSizeRaw, windowSizeEffective);
}
Expand Down
Loading