HW2: Add the rest of the report (w/o the prof)

homework_2/include/config.h

@@ -12,6 +12,17 @@
 
 #include <cstdint>
 
+/*
+ * Versioning:
+ * - RC1: Model version
+ * - RC2: Parallel full sort
+ * - RC3: Vanila
+ * - RC3a: Exchange optimization
+ * - RC3b: MPI Pipeline
+ * - RC4:  Include all version
+ */
+static constexpr char version[] = "0.4";
+
 /*
  * Defines for different version of the exercise
  */

homework_2/include/distsort.hpp

@@ -25,11 +25,11 @@
 /*
  * Exported timers
  */
-extern Timing Ttotal;
-extern Timing TfullSort;
-extern Timing Texchange;
-extern Timing Tminmax;
-extern Timing TelbowSort;
+extern Timing Timer_total;
+extern Timing Timer_fullSort;
+extern Timing Timer_exchange;
+extern Timing Timer_minmax;
+extern Timing Timer_elbowSort;
 
 
 /*!
@@ -257,7 +257,7 @@ void elbowSort(ShadowedDataT& data, bool ascending) noexcept {
  */
 template<typename StatT>
 bool needsExchange(const StatT& lstat, StatT& rstat, mpi_id_t part, int tag, bool keepSmall) {
-    timeCall(Texchange, mpi.exchange_it, lstat, rstat, part, tag);
+    timeCall(Timer_exchange, mpi.exchange_it, lstat, rstat, part, tag);
     return (keepSmall) ?
            rstat.min < lstat.max   // Lmin: rstat.min - Smax: lstat.max
          : lstat.min < rstat.max;  // Lmin: lstat.min - Smax: rstat.max
@@ -351,25 +351,25 @@ void exchange(ShadowedDataT& data, mpi_id_t partner, bool keepSmall, int tag) {
 
     if (config.pipeline > 1) {
         // Pipeline case - use async MPI
-        Texchange.start();
+        Timer_exchange.start();
         mpi.exchange_start(active, shadow, count, partner, tag);
         for (size_t stage = 0; stage < config.pipeline; active += count, shadow += count) {
             // Wait previous chunk
             mpi.exchange_wait();
-            Texchange.stop();
+            Timer_exchange.stop();
             if (++stage < config.pipeline) {
                 // Start next chunk if there is a next one
-                Texchange.start();
+                Timer_exchange.start();
                 mpi.exchange_start(active + count, shadow + count, count, partner, ++tag);
             }
             // process the arrived data
-            timeCall(Tminmax, keepMinOrMax, active, shadow, count, keepSmall);
+            timeCall(Timer_minmax, keepMinOrMax, active, shadow, count, keepSmall);
         }
     }
     else {
         // No pipeline - use blocking MPI
-        timeCall(Texchange, mpi.exchange, active, shadow, count, partner, tag);
-        timeCall(Tminmax, keepMinOrMax, active, shadow, count, keepSmall);
+        timeCall(Timer_exchange, mpi.exchange, active, shadow, count, partner, tag);
+        timeCall(Timer_minmax, keepMinOrMax, active, shadow, count, keepSmall);
     }
     if (config.exchangeOpt)
         updateMinMax(localStat, data);
@@ -390,7 +390,7 @@ void exchange(ShadowedDataT& data, mpi_id_t partner, bool keepSmall, int tag) {
 template<typename ShadowedDataT>
 void distBubbletonic(ShadowedDataT& data, mpi_id_t Processes, mpi_id_t rank) {
     // Initially sort to create a half part of a bitonic sequence
-    timeCall(TfullSort,  fullSort, data, ascending<SortMode::Bubbletonic>(rank, 0));
+    timeCall(Timer_fullSort, fullSort, data, ascending<SortMode::Bubbletonic>(rank, 0));
 
     // Sort network (O(N) iterations)
     for (size_t step = 0; step < static_cast<size_t>(Processes); ++step) {
@@ -403,7 +403,7 @@ void distBubbletonic(ShadowedDataT& data, mpi_id_t Processes, mpi_id_t rank) {
             int tag = static_cast<int>(tagGenerator(0, step));
             if (!config.exchangeOpt || needsExchange(localStat, remoteStat, part, tag++, ks)) {
                 exchange(data, part, ks, tag);
-                timeCall(TelbowSort, elbowSort, data, ascending<SortMode::Bubbletonic>(rank, Processes));
+                timeCall(Timer_elbowSort, elbowSort, data, ascending<SortMode::Bubbletonic>(rank, Processes));
             }
         }
     }
@@ -430,7 +430,7 @@ void distBubbletonic(ShadowedDataT& data, mpi_id_t Processes, mpi_id_t rank) {
 template<typename ShadowedDataT>
 void distBitonic(ShadowedDataT& data, mpi_id_t Processes, mpi_id_t rank) {
     // Initially sort to create a half part of a bitonic sequence
-    timeCall(TfullSort, fullSort, data, ascending<SortMode::Bitonic>(rank, 0));
+    timeCall(Timer_fullSort, fullSort, data, ascending<SortMode::Bitonic>(rank, 0));
 
     // Run through sort network using elbow-sort ( O(LogN * LogN) iterations )
     auto p = static_cast<uint32_t>(std::log2(Processes));
@@ -447,7 +447,7 @@ void distBitonic(ShadowedDataT& data, mpi_id_t Processes, mpi_id_t rank) {
             }
         }
         // sort - O(N)
-        timeCall(TelbowSort, elbowSort, data, ascending<SortMode::Bitonic>(rank, depth));
+        timeCall(Timer_elbowSort, elbowSort, data, ascending<SortMode::Bitonic>(rank, depth));
     }
 }