x and y position calculation in kernel

.clang-format

@@ -1,4 +1,4 @@
----
+macs -nw 2	---
 Language:        Cpp
 BasedOnStyle:  Google
 ColumnLimit:     120

CUDADataFormats/HGCal/BuildFile.xml

@@ -0,0 +1,8 @@
+<use name="FWCore/Utilities"/>
+<use name="Geometry/HGCalGeometry"/>
+<use name="HeterogeneousCore/CUDACore"/>
+<use name="cuda"/>
+
+<export>
+    <lib name="1"/>
+</export>

CUDADataFormats/HGCal/interface/HGCConditions.h

@@ -4,20 +4,21 @@
 class HeterogeneousHGCSiliconDetId {
  public:
   constexpr HeterogeneousHGCSiliconDetId(uint32_t id): id_(id) {}
-  constexpr uint32_t type() { return (id_ >> kHGCalTypeOffset) & kHGCalTypeMask; }
-  constexpr int32_t zside() { return (((id_ >> kHGCalZsideOffset) & kHGCalZsideMask) ? -1 : 1); }
-  constexpr uint32_t layer() { return (id_ >> kHGCalLayerOffset) & kHGCalLayerMask; }
+  constexpr uint32_t type()     { return (id_ >> kHGCalTypeOffset) & kHGCalTypeMask; }
+  constexpr int32_t zside()     { return (((id_ >> kHGCalZsideOffset) & kHGCalZsideMask) ? -1 : 1); }
+  constexpr uint32_t layer()    { return (id_ >> kHGCalLayerOffset) & kHGCalLayerMask; }
   constexpr int32_t waferUAbs() { return (id_ >> kHGCalWaferUOffset) & kHGCalWaferUMask; }
   constexpr int32_t waferVAbs() { return (id_ >> kHGCalWaferVOffset) & kHGCalWaferVMask; }
-  constexpr int32_t waferU() { return (((id_ >> kHGCalWaferUSignOffset) & kHGCalWaferUSignMask) ? -waferUAbs() : waferUAbs()); }
-  constexpr int32_t waferV() { return (((id_ >> kHGCalWaferVSignOffset) & kHGCalWaferVSignMask) ? -waferVAbs() : waferVAbs()); }
-  constexpr int32_t waferX() { return (-2 * waferU() + waferV()); }
-  constexpr int32_t waferY() { return (2 * waferV()); }
-  constexpr int32_t cellU() { return (id_ >> kHGCalCellUOffset) & kHGCalCellUMask; }
-  constexpr int32_t cellV() { return (id_ >> kHGCalCellVOffset) & kHGCalCellVMask; }
-  constexpr int32_t cellX() { int N = (type() == HGCalFine) ? HGCalFineN : HGCalCoarseN;  return (3 * (cellV() - N) + 2); }
-  constexpr int32_t cellY() { int N = (type() == HGCalFine) ? HGCalFineN : HGCalCoarseN;  return (2 * cellU() - (N + cellV())); }
-
+  constexpr int32_t waferU()    { return (((id_ >> kHGCalWaferUSignOffset) & kHGCalWaferUSignMask) ? -waferUAbs() : waferUAbs()); }
+  constexpr int32_t waferV()    { return (((id_ >> kHGCalWaferVSignOffset) & kHGCalWaferVSignMask) ? -waferVAbs() : waferVAbs()); }
+  constexpr int32_t waferX()    { return (-2 * waferU() + waferV()); }
+  constexpr int32_t waferY()    { return (2 * waferV()); }
+  constexpr uint32_t cellU()    { return (id_ >> kHGCalCellUOffset) & kHGCalCellUMask; }
+  constexpr uint32_t cellV()    { return (id_ >> kHGCalCellVOffset) & kHGCalCellVMask; }
+  constexpr uint32_t nCells()   { return (type() == HGCalFine) ? HGCalFineN : HGCalCoarseN; }
+  constexpr int32_t cellX()     { const uint32_t N = nCells(); return (3 * (cellV() - N) + 2); }
+  constexpr int32_t cellY()     { const uint32_t N = nCells(); return (2 * cellU() - (N + cellV())); }
+
  private:
   uint32_t id_;
   enum waferType { HGCalFine = 0, HGCalCoarseThin = 1, HGCalCoarseThick = 2 };
@@ -116,20 +117,6 @@ namespace hgcal_conditions {
   } //namespace parameters
 
   namespace positions {
-    //stores the positions taken from the detid's in the CPU
-    struct HGCalPositions {
-      std::vector<float> x;
-      std::vector<float> y;
-      std::vector<float> z;
-    };
-
-    //stores the positions taken from the detid's in the GPU
-    //it is the same for all three subdetectors
-    struct HeterogeneousHGCalPositions {
-      float* x;
-      float* y;
-      float* z;
-    };
 
     enum class HeterogeneousHGCalPositionsType {Float, Int32_t, Uint32_t};
 
@@ -140,27 +127,34 @@ namespace hgcal_conditions {
 								 HeterogeneousHGCalPositionsType::Uint32_t };
 
     struct HGCalPositionsMapping {
+      std::vector<float> z_per_layer;
       std::vector<int32_t> numberCellsHexagon;
       std::vector<uint32_t> detid;
+      //variables required for calculating the positions (x,y) from the detid in the GPU
+      float waferSize;
+      float sensorSeparation;
       //variables required for the mapping of detid -> cell in the geometry
-      int firstLayer;
-      int lastLayer;
-      int waferMax;
-      int waferMin;
+      int32_t firstLayer;
+      int32_t lastLayer;
+      int32_t waferMax;
+      int32_t waferMin;
     };
 
     struct HeterogeneousHGCalPositionsMapping {
       //the x, y and z positions will not be filled in the CPU
       float* x;
       float* y;
-      float* z;
+      float* z_per_layer;
       int32_t *numberCellsHexagon;
       uint32_t *detid;
+      //variables required for calculating the positions (x,y) from the detid in the GPU
+      float waferSize;
+      float sensorSeparation;
       //variables required for the mapping of detid -> cell in the geometry
-      int firstLayer;
-      int lastLayer;
-      int waferMax;
-      int waferMin;
+      int32_t firstLayer;
+      int32_t lastLayer;
+      int32_t waferMax;
+      int32_t waferMin;
     };
 
   } //namespace positions
@@ -170,13 +164,18 @@ namespace hgcal_conditions {
   };
   struct HeterogeneousHEFConditionsESProduct {
     parameters::HeterogeneousHGCalHEFParameters params;
-    positions::HeterogeneousHGCalPositionsMapping posmap;
-    size_t nelems_posmap;
+    //positions::HeterogeneousHGCalPositionsMapping posmap;
+    //size_t nelems_posmap;
   };
   struct HeterogeneousHEBConditionsESProduct {
     parameters::HeterogeneousHGCalHEBParameters params;
   };
 
+  struct HeterogeneousHEFCellPositionsConditionsESProduct {
+    positions::HeterogeneousHGCalPositionsMapping posmap;
+    size_t nelems_posmap;
+  };
+
 } //namespace conditions
 
 #endif //CUDADataFormats_HGCal_HGCConditions_h

CUDADataFormats/HGCal/interface/HGCUncalibratedRecHitsToRecHitsConstants.h

@@ -6,19 +6,19 @@
 //maximum sizes for SoA's arrays holding configuration data ("constants")
 namespace maxsizes_constants {
   //EE
-  constexpr size_t ee_fCPerMIP = 6; //number of elements pointed by hgcEE_fCPerMIP_
-  constexpr size_t ee_cce = 6; //number of elements posize_ted by hgcEE_cce_
-  constexpr size_t ee_noise_fC = 6; //number of elements posize_ted by hgcEE_noise_fC_
-  constexpr size_t ee_rcorr = 6; //number of elements posize_ted by rcorr_
-  constexpr size_t ee_weights = 53; //number of elements posize_ted by weights_
+  constexpr size_t ee_fCPerMIP = 3; //number of elements pointed by hgcEE_fCPerMIP_
+  constexpr size_t ee_cce = 3; //number of elements pointed by hgcEE_cce_
+  constexpr size_t ee_noise_fC = 3; //number of elements pointed by hgcEE_noise_fC_
+  constexpr size_t ee_rcorr = 3; //number of elements pointed by rcorr_
+  constexpr size_t ee_weights = 51; //number of elements posize_ted by weights_
   //HEF
-  constexpr size_t hef_fCPerMIP = 6; //number of elements pointed by hgcEE_fCPerMIP_
-  constexpr size_t hef_cce = 6; //number of elements posize_ted by hgcEE_cce_
-  constexpr size_t hef_noise_fC = 6; //number of elements posize_ted by hgcEE_noise_fC_
-  constexpr size_t hef_rcorr = 6; //number of elements posize_ted by rcorr_
-  constexpr size_t hef_weights = 53; //number of elements posize_ted by weights_
+  constexpr size_t hef_fCPerMIP = 3; //number of elements pointed by hgcEE_fCPerMIP_
+  constexpr size_t hef_cce = 3; //number of elements pointed by hgcEE_cce_
+  constexpr size_t hef_noise_fC = 3; //number of elements pointed by hgcEE_noise_fC_
+  constexpr size_t hef_rcorr = 3; //number of elements pointed by rcorr_
+  constexpr size_t hef_weights = 51; //number of elements pointed by weights_
   //HEB
-  constexpr size_t heb_weights = 53; //number of elements posize_ted by weights_
+  constexpr size_t heb_weights = 51; //number of elements pointed by weights_
 }
 
 class HGCConstantVectorData {
@@ -60,6 +60,7 @@ class HGChefUncalibratedRecHitConstantData {
   float xmax_;         //used for computing the time resolution error
   float aterm_;        //used for computing the time resolution error
   float cterm_;        //used for computing the time resolution error
+  uint32_t layerOffset_;  //layer offset relative to layer#1 of the EE subsetector
 };
 
 class HGChebUncalibratedRecHitConstantData {
@@ -69,6 +70,7 @@ class HGChebUncalibratedRecHitConstantData {
   double keV2DIGI_;   //energy to femto coloumb conversion: 1000 eV/3.62 (eV per e) / 6.24150934e3 (e per fC)
   double uncalib2GeV_; //sets the ADC; obtained by dividing 1e-6 by hgcHEB_keV2DIGI_
   double noise_MIP_;  //noise
+  uint32_t layerOffset_;  //layer offset relative to layer#1 of the EE subsetector
 };
 
 #endif

CondFormats/DataRecord/BuildFile.xml

@@ -1,4 +1,5 @@
 <use name="FWCore/Framework"/>
+<use name="FWCore/Utilities"/>
 <use name="Geometry/Records"/>
 <export>
   <lib name="1"/>

CondFormats/DataRecord/interface/HeterogeneousHGCalHEFCellPositionsConditionsRecord.h

@@ -0,0 +1,12 @@
+#ifndef CondFormats_DataRecord_HeterogeneousHGCalHEFCellPositionsConditionsRecord_h
+#define CondFormats_DataRecord_HeterogeneousHGCalHEFCellPositionsConditionsRecord_h
+
+#include "FWCore/Framework/interface/DependentRecordImplementation.h"
+//#include "Geometry/HGCalGeometry/interface/HGCalGeometry.h"
+#include "Geometry/Records/interface/IdealGeometryRecord.h"
+#include "FWCore/Utilities/interface/mplVector.h"
+
+class HeterogeneousHGCalHEFCellPositionsConditionsRecord
+    : public edm::eventsetup::DependentRecordImplementation<HeterogeneousHGCalHEFCellPositionsConditionsRecord, edm::mpl::Vector<IdealGeometryRecord>> {};
+
+#endif //CondFormats_DataRecord_HeterogeneousHGCalHEFCellPositionsConditionsRecord_h

CondFormats/DataRecord/src/HeterogeneousHGCalHEFCellPositionsConditionsRecord.cc

@@ -0,0 +1,4 @@
+#include "CondFormats/DataRecord/interface/HeterogeneousHGCalHEFCellPositionsConditionsRecord.h"
+#include "FWCore/Framework/interface/eventsetuprecord_registration_macro.h"
+
+EVENTSETUP_RECORD_REG(HeterogeneousHGCalHEFCellPositionsConditionsRecord);

CondFormats/HGCalObjects/BuildFile.xml

@@ -0,0 +1,8 @@
+<use   name="FWCore/Utilities"/>
+<use   name="FWCore/MessageLogger"/>
+<use   name="HeterogeneousCore/CUDACore"/>
+<use   name="CUDADataFormats/HGCal"/>
+<use   name="Geometry/HGCalCommonData"/>
+<export>
+  <lib   name="1"/>
+</export>

CondFormats/HGCalObjects/interface/HeterogeneousHGCalHEFCellPositionsConditions.h

@@ -0,0 +1,73 @@
+#ifndef CondFormats_HGCalObjects_HeterogeneousHGCalHEFConditions_h
+#define CondFormats_HGCalObjects_HeterogeneousHGCalHEFConditions_h
+
+#include <numeric> //accumulate
+#include <typeinfo>
+
+#include "FWCore/MessageLogger/interface/MessageLogger.h"
+#include "HeterogeneousCore/CUDACore/interface/ESProduct.h"
+#include "HeterogeneousCore/CUDAUtilities/interface/cudaCheck.h"
+#include "CUDADataFormats/HGCal/interface/HGCConditions.h"
+#include "Geometry/HGCalCommonData/interface/HGCalDDDConstants.h"
+#include "Geometry/HGCalCommonData/interface/HGCalParameters.h"
+
+namespace cpar = hgcal_conditions::parameters;
+namespace cpos = hgcal_conditions::positions;  
+
+// Declare the wrapper ESProduct. The corresponding ESProducer should
+// produce objects of this type.
+class HeterogeneousHGCalHEFCellPositionsConditions {
+ public:
+  // Constructor takes the standard CPU ESProduct, and transforms the
+  // necessary data to array(s) in pinned host memory
+  HeterogeneousHGCalHEFCellPositionsConditions(cpos::HGCalPositionsMapping*);
+
+  // Deallocates all pinned host memory
+  ~HeterogeneousHGCalHEFCellPositionsConditions();
+
+  // Function to return the actual payload on the memory of the current device
+  hgcal_conditions::HeterogeneousHEFCellPositionsConditionsESProduct const *getHeterogeneousConditionsESProductAsync(cudaStream_t stream) const;
+
+ private:
+  // Holds the data in pinned CPU memory
+  // Contrary to its non-heterogeneous counterpart (constructor argument) it is *not* a pointer (so to avoid an extra allocation)
+  cpos::HeterogeneousHGCalPositionsMapping posmap_;
+  size_t nelems_posmap_;
+
+  std::vector<size_t> sizes_pos_;
+  size_t chunk_pos_;
+  const size_t number_position_arrays = 2; //x and y; required due to the assymetry between cpos::HeterogeneousHGCalPositionsMapping and cpos::HGCalPositionsMapping
+
+  std::vector<size_t> calculate_memory_bytes_pos_(cpos::HGCalPositionsMapping*);
+  size_t allocate_memory_pos_(const std::vector<size_t>&);
+  void transfer_data_to_heterogeneous_pointers_pos_(const std::vector<size_t>&, cpos::HGCalPositionsMapping*);
+  void transfer_data_to_heterogeneous_vars_pos_(const cpos::HGCalPositionsMapping*);
+
+  /*methods for managing SoA's pointers*/
+  //float
+  float*& select_pointer_f_(cpos::HeterogeneousHGCalPositionsMapping*, const unsigned int&) const;
+  std::vector<float>& select_pointer_f_(cpos::HGCalPositionsMapping*, const unsigned int&);
+  //int32_t
+  int32_t*& select_pointer_i_(cpos::HeterogeneousHGCalPositionsMapping*, const unsigned int&) const;
+  std::vector<int32_t>& select_pointer_i_(cpos::HGCalPositionsMapping*, const unsigned int&);
+  //uint32_t
+  uint32_t*& select_pointer_u_(cpos::HeterogeneousHGCalPositionsMapping*, const unsigned int&) const;
+  std::vector<uint32_t>& select_pointer_u_(cpos::HGCalPositionsMapping*, const unsigned int&);
+
+  // Helper struct to hold all information that has to be allocated and
+  // deallocated per device
+  struct GPUData {
+    // Destructor should free all member pointers
+    ~GPUData();
+    // internal pointers are on device, struct itself is on CPU
+    hgcal_conditions::HeterogeneousHEFCellPositionsConditionsESProduct *host = nullptr;
+    // internal pounters and struct are on device
+    hgcal_conditions::HeterogeneousHEFCellPositionsConditionsESProduct *device = nullptr;
+  };
+
+  // Helper that takes care of complexity of transferring the data to
+  // multiple devices
+  cms::cuda::ESProduct<GPUData> gpuData_;
+};
+
+#endif //CondFormats_HGCalObjects_HeterogeneousHGCalHEFConditions_h