Skip to content

Commit

Permalink
Addressed review comments
Browse files Browse the repository at this point in the history
  • Loading branch information
@ariostas
ariostas committed Aug 30, 2024
1 parent 0340221 commit 3eb7a56
Showing 1 changed file with 65 additions and 129 deletions.
194 changes: 65 additions & 129 deletions RecoTracker/LSTCore/src/alpaka/Event.dev.cc
View file
Original file line number Diff line number Diff line change
Expand Up @@ -12,26 +12,6 @@ using namespace ALPAKA_ACCELERATOR_NAMESPACE::lst;
void Event::initSync(bool verbose) {
alpaka::wait(queue); // other calls can be asynchronous
addObjects = verbose;
hitsInGPU = std::nullopt;
mdsInGPU = std::nullopt;
segmentsInGPU = std::nullopt;
tripletsInGPU = std::nullopt;
quintupletsInGPU = std::nullopt;
trackCandidatesInGPU = std::nullopt;
pixelTripletsInGPU = std::nullopt;
pixelQuintupletsInGPU = std::nullopt;
rangesInGPU = std::nullopt;

hitsInCPU = std::nullopt;
rangesInCPU = std::nullopt;
mdsInCPU = std::nullopt;
segmentsInCPU = std::nullopt;
tripletsInCPU = std::nullopt;
trackCandidatesInCPU = std::nullopt;
modulesInCPU = std::nullopt;
quintupletsInCPU = std::nullopt;
pixelTripletsInCPU = std::nullopt;
pixelQuintupletsInCPU = std::nullopt;

//reset the arrays
for (int i = 0; i < 6; i++) {
Expand Down Expand Up @@ -71,73 +51,35 @@ void Event::resetEventSync() {
n_quintuplets_by_layer_endcap_[i] = 0;
}
}
if (hitsInGPU) {
hitsInGPU = std::nullopt;
hitsBuffers = std::nullopt;
}
if (mdsInGPU) {
mdsInGPU = std::nullopt;
miniDoubletsBuffers = std::nullopt;
}
if (rangesInGPU) {
rangesInGPU = std::nullopt;
rangesBuffers = std::nullopt;
}
if (segmentsInGPU) {
segmentsInGPU = std::nullopt;
segmentsBuffers = std::nullopt;
}
if (tripletsInGPU) {
tripletsInGPU = std::nullopt;
tripletsBuffers = std::nullopt;
}
if (quintupletsInGPU) {
quintupletsInGPU = std::nullopt;
quintupletsBuffers = std::nullopt;
}
if (trackCandidatesInGPU) {
trackCandidatesInGPU = std::nullopt;
trackCandidatesBuffers = std::nullopt;
}
if (pixelTripletsInGPU) {
pixelTripletsInGPU = std::nullopt;
pixelTripletsBuffers = std::nullopt;
}
if (pixelQuintupletsInGPU) {
pixelQuintupletsInGPU = std::nullopt;
pixelQuintupletsBuffers = std::nullopt;
}

if (hitsInCPU) {
hitsInCPU = std::nullopt;
}
if (rangesInCPU) {
rangesInCPU = std::nullopt;
}
if (mdsInCPU) {
mdsInCPU = std::nullopt;
}
if (segmentsInCPU) {
segmentsInCPU = std::nullopt;
}
if (tripletsInCPU) {
tripletsInCPU = std::nullopt;
}
if (quintupletsInCPU) {
quintupletsInCPU = std::nullopt;
}
if (pixelTripletsInCPU) {
pixelTripletsInCPU = std::nullopt;
}
if (pixelQuintupletsInCPU) {
pixelQuintupletsInCPU = std::nullopt;
}
if (trackCandidatesInCPU) {
trackCandidatesInCPU = std::nullopt;
}
if (modulesInCPU) {
modulesInCPU = std::nullopt;
}
hitsInGPU.reset();
hitsBuffers.reset();
mdsInGPU.reset();
miniDoubletsBuffers.reset();
rangesInGPU.reset();
rangesBuffers.reset();
segmentsInGPU.reset();
segmentsBuffers.reset();
tripletsInGPU.reset();
tripletsBuffers.reset();
quintupletsInGPU.reset();
quintupletsBuffers.reset();
trackCandidatesInGPU.reset();
trackCandidatesBuffers.reset();
pixelTripletsInGPU.reset();
pixelTripletsBuffers.reset();
pixelQuintupletsInGPU.reset();
pixelQuintupletsBuffers.reset();

hitsInCPU.reset();
rangesInCPU.reset();
mdsInCPU.reset();
segmentsInCPU.reset();
tripletsInCPU.reset();
quintupletsInCPU.reset();
pixelTripletsInCPU.reset();
pixelQuintupletsInCPU.reset();
trackCandidatesInCPU.reset();
modulesInCPU.reset();
}

void Event::addHitToEvent(std::vector<float> const& x,
Expand All @@ -150,14 +92,14 @@ void Event::addHitToEvent(std::vector<float> const& x,

// Initialize space on device/host for next event.
if (!hitsInGPU) {
hitsInGPU = Hits();
hitsBuffers = HitsBuffer<Device>(nModules_, nHits, devAcc, queue);
hitsInGPU.emplace();
hitsBuffers.emplace(nModules_, nHits, devAcc, queue);
hitsInGPU->setData(*hitsBuffers);
}

if (!rangesInGPU) {
rangesInGPU = ObjectRanges();
rangesBuffers = ObjectRangesBuffer<Device>(nModules_, nLowerModules_, devAcc, queue);
rangesInGPU.emplace();
rangesBuffers.emplace(nModules_, nLowerModules_, devAcc, queue);
rangesInGPU->setData(*rangesBuffers);
}

Expand Down Expand Up @@ -253,8 +195,8 @@ void Event::addPixelSegmentToEvent(std::vector<unsigned int> const& hitIndices0,
*nTotalMDs_buf_h.data() += n_max_pixel_md_per_modules;
unsigned int nTotalMDs = *nTotalMDs_buf_h.data();

mdsInGPU = MiniDoublets();
miniDoubletsBuffers = MiniDoubletsBuffer<Device>(nTotalMDs, nLowerModules_, devAcc, queue);
mdsInGPU.emplace();
miniDoubletsBuffers.emplace(nTotalMDs, nLowerModules_, devAcc, queue);
mdsInGPU->setData(*miniDoubletsBuffers);

alpaka::memcpy(queue, miniDoubletsBuffers->nMemoryLocations_buf, nTotalMDs_buf_h);
Expand All @@ -279,9 +221,8 @@ void Event::addPixelSegmentToEvent(std::vector<unsigned int> const& hitIndices0,

nTotalSegments_ += n_max_pixel_segments_per_module;

segmentsInGPU = Segments();
segmentsBuffers =
SegmentsBuffer<Device>(nTotalSegments_, nLowerModules_, n_max_pixel_segments_per_module, devAcc, queue);
segmentsInGPU.emplace();
segmentsBuffers.emplace(nTotalSegments_, nLowerModules_, n_max_pixel_segments_per_module, devAcc, queue);
segmentsInGPU->setData(*segmentsBuffers);

alpaka::memcpy(queue, segmentsBuffers->nMemoryLocations_buf, nTotalSegments_view);
Expand Down Expand Up @@ -378,8 +319,8 @@ void Event::createMiniDoublets() {
unsigned int nTotalMDs = *nTotalMDs_buf_h.data();

if (!mdsInGPU) {
mdsInGPU = MiniDoublets();
miniDoubletsBuffers = MiniDoubletsBuffer<Device>(nTotalMDs, nLowerModules_, devAcc, queue);
mdsInGPU.emplace();
miniDoubletsBuffers.emplace(nTotalMDs, nLowerModules_, devAcc, queue);
mdsInGPU->setData(*miniDoubletsBuffers);
}

Expand Down Expand Up @@ -413,9 +354,8 @@ void Event::createMiniDoublets() {

void Event::createSegmentsWithModuleMap() {
if (!segmentsInGPU) {
segmentsInGPU = Segments();
segmentsBuffers =
SegmentsBuffer<Device>(nTotalSegments_, nLowerModules_, n_max_pixel_segments_per_module, devAcc, queue);
segmentsInGPU.emplace();
segmentsBuffers.emplace(nTotalSegments_, nLowerModules_, n_max_pixel_segments_per_module, devAcc, queue);
segmentsInGPU->setData(*segmentsBuffers);
}

Expand Down Expand Up @@ -463,8 +403,8 @@ void Event::createTriplets() {
alpaka::memcpy(queue, maxTriplets_buf_h, rangesBuffers->device_nTotalTrips_buf);
alpaka::wait(queue); // wait to get the value before using it

tripletsInGPU = Triplets();
tripletsBuffers = TripletsBuffer<Device>(*maxTriplets_buf_h.data(), nLowerModules_, devAcc, queue);
tripletsInGPU.emplace();
tripletsBuffers.emplace(*maxTriplets_buf_h.data(), nLowerModules_, devAcc, queue);
tripletsInGPU->setData(*tripletsBuffers);

alpaka::memcpy(queue, tripletsBuffers->nMemoryLocations_buf, maxTriplets_buf_h);
Expand Down Expand Up @@ -537,9 +477,8 @@ void Event::createTriplets() {

void Event::createTrackCandidates(bool no_pls_dupclean, bool tc_pls_triplets) {
if (!trackCandidatesInGPU) {
trackCandidatesInGPU = TrackCandidates();
trackCandidatesBuffers =
TrackCandidatesBuffer<Device>(n_max_nonpixel_track_candidates + n_max_pixel_track_candidates, devAcc, queue);
trackCandidatesInGPU.emplace();
trackCandidatesBuffers.emplace(n_max_nonpixel_track_candidates + n_max_pixel_track_candidates, devAcc, queue);
trackCandidatesInGPU->setData(*trackCandidatesBuffers);
}

Expand Down Expand Up @@ -679,8 +618,8 @@ void Event::createTrackCandidates(bool no_pls_dupclean, bool tc_pls_triplets) {

void Event::createPixelTriplets() {
if (!pixelTripletsInGPU) {
pixelTripletsInGPU = PixelTriplets();
pixelTripletsBuffers = PixelTripletsBuffer<Device>(n_max_pixel_triplets, devAcc, queue);
pixelTripletsInGPU.emplace();
pixelTripletsBuffers.emplace(n_max_pixel_triplets, devAcc, queue);
pixelTripletsInGPU->setData(*pixelTripletsBuffers);
}

Expand Down Expand Up @@ -814,8 +753,8 @@ void Event::createQuintuplets() {
auto nTotalQuintuplets = *nTotalQuintuplets_buf.data();

if (!quintupletsInGPU) {
quintupletsInGPU = Quintuplets();
quintupletsBuffers = QuintupletsBuffer<Device>(nTotalQuintuplets, nLowerModules_, devAcc, queue);
quintupletsInGPU.emplace();
quintupletsBuffers.emplace(nTotalQuintuplets, nLowerModules_, devAcc, queue);
quintupletsInGPU->setData(*quintupletsBuffers);

alpaka::memcpy(queue, quintupletsBuffers->nMemoryLocations_buf, nTotalQuintuplets_buf);
Expand Down Expand Up @@ -876,14 +815,13 @@ void Event::pixelLineSegmentCleaning(bool no_pls_dupclean) {

void Event::createPixelQuintuplets() {
if (!pixelQuintupletsInGPU) {
pixelQuintupletsInGPU = PixelQuintuplets();
pixelQuintupletsBuffers = PixelQuintupletsBuffer<Device>(n_max_pixel_quintuplets, devAcc, queue);
pixelQuintupletsInGPU.emplace();
pixelQuintupletsBuffers.emplace(n_max_pixel_quintuplets, devAcc, queue);
pixelQuintupletsInGPU->setData(*pixelQuintupletsBuffers);
}
if (!trackCandidatesInGPU) {
trackCandidatesInGPU = TrackCandidates();
trackCandidatesBuffers =
TrackCandidatesBuffer<Device>(n_max_nonpixel_track_candidates + n_max_pixel_track_candidates, devAcc, queue);
trackCandidatesInGPU.emplace();
trackCandidatesBuffers.emplace(n_max_nonpixel_track_candidates + n_max_pixel_track_candidates, devAcc, queue);
trackCandidatesInGPU->setData(*trackCandidatesBuffers);
}

Expand Down Expand Up @@ -1332,7 +1270,7 @@ HitsBuffer<alpaka_common::DevHost>& Event::getHits(bool sync) //std::shared_ptr
alpaka::wait(queue); // wait for the value before using

auto const nHits = *nHits_buf_h.data();
hitsInCPU = HitsBuffer<DevHost>(nModules_, nHits, devHost, queue);
hitsInCPU.emplace(nModules_, nHits, devHost, queue);
hitsInCPU->setData(*hitsInCPU);

alpaka::memcpy(queue, hitsInCPU->nHits_buf, hitsBuffers->nHits_buf);
Expand All @@ -1355,7 +1293,7 @@ HitsBuffer<alpaka_common::DevHost>& Event::getHitsInCMSSW(bool sync) {
alpaka::wait(queue); // wait for the value before using

auto const nHits = *nHits_buf_h.data();
hitsInCPU = HitsBuffer<DevHost>(nModules_, nHits, devHost, queue);
hitsInCPU.emplace(nModules_, nHits, devHost, queue);
hitsInCPU->setData(*hitsInCPU);

alpaka::memcpy(queue, hitsInCPU->nHits_buf, hitsBuffers->nHits_buf);
Expand All @@ -1368,7 +1306,7 @@ HitsBuffer<alpaka_common::DevHost>& Event::getHitsInCMSSW(bool sync) {

ObjectRangesBuffer<alpaka_common::DevHost>& Event::getRanges(bool sync) {
if (!rangesInCPU) {
rangesInCPU = ObjectRangesBuffer<DevHost>(nModules_, nLowerModules_, devHost, queue);
rangesInCPU.emplace(nModules_, nLowerModules_, devHost, queue);
rangesInCPU->setData(*rangesInCPU);

alpaka::memcpy(queue, rangesInCPU->hitRanges_buf, rangesBuffers->hitRanges_buf);
Expand All @@ -1390,7 +1328,7 @@ MiniDoubletsBuffer<alpaka_common::DevHost>& Event::getMiniDoublets(bool sync) {
alpaka::wait(queue); // wait for the value before using

auto const nMemHost = *nMemHost_buf_h.data();
mdsInCPU = MiniDoubletsBuffer<DevHost>(nMemHost, nLowerModules_, devHost, queue);
mdsInCPU.emplace(nMemHost, nLowerModules_, devHost, queue);
mdsInCPU->setData(*mdsInCPU);

alpaka::memcpy(queue, mdsInCPU->nMemoryLocations_buf, miniDoubletsBuffers->nMemoryLocations_buf);
Expand All @@ -1413,7 +1351,7 @@ SegmentsBuffer<alpaka_common::DevHost>& Event::getSegments(bool sync) {
alpaka::wait(queue); // wait for the value before using

auto const nMemHost = *nMemHost_buf_h.data();
segmentsInCPU = SegmentsBuffer<DevHost>(nMemHost, nLowerModules_, n_max_pixel_segments_per_module, devHost, queue);
segmentsInCPU.emplace(nMemHost, nLowerModules_, n_max_pixel_segments_per_module, devHost, queue);
segmentsInCPU->setData(*segmentsInCPU);

alpaka::memcpy(queue, segmentsInCPU->nMemoryLocations_buf, segmentsBuffers->nMemoryLocations_buf);
Expand Down Expand Up @@ -1449,7 +1387,7 @@ TripletsBuffer<alpaka_common::DevHost>& Event::getTriplets(bool sync) {
alpaka::wait(queue); // wait for the value before using

auto const nMemHost = *nMemHost_buf_h.data();
tripletsInCPU = TripletsBuffer<DevHost>(nMemHost, nLowerModules_, devHost, queue);
tripletsInCPU.emplace(nMemHost, nLowerModules_, devHost, queue);
tripletsInCPU->setData(*tripletsInCPU);

alpaka::memcpy(queue, tripletsInCPU->nMemoryLocations_buf, tripletsBuffers->nMemoryLocations_buf);
Expand Down Expand Up @@ -1486,7 +1424,7 @@ QuintupletsBuffer<alpaka_common::DevHost>& Event::getQuintuplets(bool sync) {
alpaka::wait(queue); // wait for the value before using

auto const nMemHost = *nMemHost_buf_h.data();
quintupletsInCPU = QuintupletsBuffer<DevHost>(nMemHost, nLowerModules_, devHost, queue);
quintupletsInCPU.emplace(nMemHost, nLowerModules_, devHost, queue);
quintupletsInCPU->setData(*quintupletsInCPU);

alpaka::memcpy(queue, quintupletsInCPU->nMemoryLocations_buf, quintupletsBuffers->nMemoryLocations_buf);
Expand Down Expand Up @@ -1523,7 +1461,7 @@ PixelTripletsBuffer<alpaka_common::DevHost>& Event::getPixelTriplets(bool sync)
alpaka::wait(queue); // wait for the value before using

auto const nPixelTriplets = *nPixelTriplets_buf_h.data();
pixelTripletsInCPU = PixelTripletsBuffer<DevHost>(nPixelTriplets, devHost, queue);
pixelTripletsInCPU.emplace(nPixelTriplets, devHost, queue);
pixelTripletsInCPU->setData(*pixelTripletsInCPU);

alpaka::memcpy(queue, pixelTripletsInCPU->nPixelTriplets_buf, pixelTripletsBuffers->nPixelTriplets_buf);
Expand Down Expand Up @@ -1563,7 +1501,7 @@ PixelQuintupletsBuffer<alpaka_common::DevHost>& Event::getPixelQuintuplets(bool
alpaka::wait(queue); // wait for the value before using

auto const nPixelQuintuplets = *nPixelQuintuplets_buf_h.data();
pixelQuintupletsInCPU = PixelQuintupletsBuffer<DevHost>(nPixelQuintuplets, devHost, queue);
pixelQuintupletsInCPU.emplace(nPixelQuintuplets, devHost, queue);
pixelQuintupletsInCPU->setData(*pixelQuintupletsInCPU);

alpaka::memcpy(queue, pixelQuintupletsInCPU->nPixelQuintuplets_buf, pixelQuintupletsBuffers->nPixelQuintuplets_buf);
Expand Down Expand Up @@ -1597,8 +1535,7 @@ TrackCandidatesBuffer<alpaka_common::DevHost>& Event::getTrackCandidates(bool sy
// Get nTrackCanHost parameter to initialize host based trackCandidatesInCPU
auto nTrackCanHost_buf_h = cms::alpakatools::make_host_buffer<unsigned int[]>(queue, 1u);
alpaka::memcpy(queue, nTrackCanHost_buf_h, trackCandidatesBuffers->nTrackCandidates_buf);
trackCandidatesInCPU =
TrackCandidatesBuffer<DevHost>(n_max_nonpixel_track_candidates + n_max_pixel_track_candidates, devHost, queue);
trackCandidatesInCPU.emplace(n_max_nonpixel_track_candidates + n_max_pixel_track_candidates, devHost, queue);
trackCandidatesInCPU->setData(*trackCandidatesInCPU);
alpaka::wait(queue); // wait here before we get nTrackCanHost and trackCandidatesInCPU becomes usable

Expand Down Expand Up @@ -1636,8 +1573,7 @@ TrackCandidatesBuffer<alpaka_common::DevHost>& Event::getTrackCandidatesInCMSSW(
// Get nTrackCanHost parameter to initialize host based trackCandidatesInCPU
auto nTrackCanHost_buf_h = cms::alpakatools::make_host_buffer<unsigned int[]>(queue, 1u);
alpaka::memcpy(queue, nTrackCanHost_buf_h, trackCandidatesBuffers->nTrackCandidates_buf);
trackCandidatesInCPU =
TrackCandidatesBuffer<DevHost>(n_max_nonpixel_track_candidates + n_max_pixel_track_candidates, devHost, queue);
trackCandidatesInCPU.emplace(n_max_nonpixel_track_candidates + n_max_pixel_track_candidates, devHost, queue);
trackCandidatesInCPU->setData(*trackCandidatesInCPU);
alpaka::wait(queue); // wait for the value before using and trackCandidatesInCPU becomes usable

Expand All @@ -1663,7 +1599,7 @@ TrackCandidatesBuffer<alpaka_common::DevHost>& Event::getTrackCandidatesInCMSSW(
ModulesBuffer<alpaka_common::DevHost>& Event::getModules(bool isFull, bool sync) {
if (!modulesInCPU) {
// The last input here is just a small placeholder for the allocation.
modulesInCPU = ModulesBuffer<DevHost>(devHost, nModules_, nPixels_);
modulesInCPU.emplace(devHost, nModules_, nPixels_);

modulesInCPU->copyFromSrc(queue, modulesBuffers_, isFull);
if (sync)
Expand Down

0 comments on commit 3eb7a56

Please sign in to comment.