Replace cub prefix scan with home-brewed one (#447)

Replace the use of the prefix scan from CUB with a home-brewed implementation, using dynamic instead of static shared memory. No changes to physics or timing performance.
cms-patatrack · Oct 20, 2020 · 8b365ad · 8b365ad
1 parent a848af7
commit 8b365ad
Show file tree

Hide file tree

Showing 2 changed files with 2 additions and 4 deletions.
diff --git a/RecoLocalTracker/SiPixelRecHits/plugins/PixelRecHits.cu b/RecoLocalTracker/SiPixelRecHits/plugins/PixelRecHits.cu
@@ -63,9 +63,7 @@ namespace pixelgpudetails {
     }
 
     if (nHits) {
-      auto hws = cms::cuda::make_device_unique<uint8_t[]>(TrackingRecHit2DSOAView::Hist::wsSize(), stream);
-      cms::cuda::fillManyFromVector(
-          hits_d.phiBinner(), hws.get(), 10, hits_d.iphi(), hits_d.hitsLayerStart(), nHits, 256, stream);
+      cms::cuda::fillManyFromVector(hits_d.phiBinner(), 10, hits_d.iphi(), hits_d.hitsLayerStart(), nHits, 256, stream);
       cudaCheck(cudaGetLastError());
     }
 

diff --git a/RecoLocalTracker/SiPixelRecHits/plugins/SiPixelRecHitSoAFromLegacy.cc b/RecoLocalTracker/SiPixelRecHits/plugins/SiPixelRecHitSoAFromLegacy.cc
@@ -251,7 +251,7 @@ void SiPixelRecHitSoAFromLegacy::produce(edm::StreamID streamID, edm::Event& iEv
     output->hitsLayerStart()[i] = hitsModuleStart[cpeView.layerGeometry().layerStart[i]];
   }
   cms::cuda::fillManyFromVector(
-      output->phiBinner(), nullptr, 10, output->iphi(), output->hitsLayerStart(), numberOfHits, 256, nullptr);
+      output->phiBinner(), 10, output->iphi(), output->hitsLayerStart(), numberOfHits, 256, nullptr);
 
   // std::cout << "created HitSoa for " <<  numberOfClusters << " clusters in " << numberOfDetUnits << " Dets" << std::endl;
   iEvent.put(std::move(output));