change node index from int16_t to int64_t (#94)

include/pando-lib-galois/containers/dist_array.hpp

@@ -153,7 +153,7 @@ class DistArray {
     pando::Status err = hostsPlaces.initialize(pando::getPlaceDims().node.id);
     PANDO_CHECK_RETURN(err);
 
-    for (std::int16_t i = 0; i < pando::getPlaceDims().node.id; i++) {
+    for (std::int64_t i = 0; i < pando::getPlaceDims().node.id; i++) {
       hostsPlaces[i] = PlaceType{pando::Place{pando::NodeIndex{i}, pando::anyPod, pando::anyCore},
                                  pando::MemoryType::Main};
     }
@@ -321,7 +321,7 @@ class DistArray {
    * @brief A beginning local iterator for a specified node `n` that
    * points to the first local item of this distributed array.
    */
-  pando::GlobalPtr<T> localBegin(std::int16_t n) noexcept {
+  pando::GlobalPtr<T> localBegin(std::int64_t n) noexcept {
     pando::Array<T> arr = m_data[n];
     return pando::GlobalPtr<T>(arr.begin());
   }
@@ -330,7 +330,7 @@ class DistArray {
    * @brief A ending local iterator for a specified node `n` that
    * points to (the last local item of this distributed array + 1).
    */
-  pando::GlobalPtr<T> localEnd(std::int16_t n) noexcept {
+  pando::GlobalPtr<T> localEnd(std::int64_t n) noexcept {
     pando::Array<T> arr = m_data[n];
     return pando::GlobalPtr<T>(arr.end());
   }

include/pando-lib-galois/containers/host_indexed_map.hpp

@@ -176,16 +176,16 @@ class HostIndexedMap {
 template <typename T>
 class HostIndexedMapIt {
   pando::GlobalPtr<T> m_curr;
-  std::int16_t m_loc;
+  std::int64_t m_loc;
 
 public:
   using iterator_category = std::random_access_iterator_tag;
-  using difference_type = std::int16_t;
+  using difference_type = std::int64_t;
   using value_type = T;
   using pointer = pando::GlobalPtr<T>;
   using reference = pando::GlobalRef<T>;
 
-  HostIndexedMapIt(pando::GlobalPtr<T> curr, std::int16_t loc) : m_curr(curr), m_loc(loc) {}
+  HostIndexedMapIt(pando::GlobalPtr<T> curr, std::int64_t loc) : m_curr(curr), m_loc(loc) {}
 
   constexpr HostIndexedMapIt() noexcept = default;
   constexpr HostIndexedMapIt(HostIndexedMapIt&&) noexcept = default;

include/pando-lib-galois/containers/host_local_storage.hpp

@@ -107,11 +107,11 @@ class HostLocalStorage {
   }
 
   pando::GlobalPtr<T> get(std::uint64_t i) noexcept {
-    return m_items.getPointerAt(pando::NodeIndex(static_cast<std::int16_t>(i)));
+    return m_items.getPointerAt(pando::NodeIndex(static_cast<std::int64_t>(i)));
   }
 
   pando::GlobalPtr<const T> get(std::uint64_t i) const noexcept {
-    return m_items.getPointerAt(pando::NodeIndex(static_cast<std::int16_t>(i)));
+    return m_items.getPointerAt(pando::NodeIndex(static_cast<std::int64_t>(i)));
   }
 
   pando::GlobalRef<T> operator[](std::uint64_t i) noexcept {
@@ -184,16 +184,16 @@ class HostLocalStorage {
 template <typename T>
 class HostLocalStorageIt {
   HostLocalStorage<T> m_curr;
-  std::int16_t m_loc;
+  std::int64_t m_loc;
 
 public:
   using iterator_category = std::random_access_iterator_tag;
-  using difference_type = std::int16_t;
+  using difference_type = std::int64_t;
   using value_type = T;
   using pointer = pando::GlobalPtr<T>;
   using reference = pando::GlobalRef<T>;
 
-  HostLocalStorageIt(HostLocalStorage<T> curr, std::int16_t loc) : m_curr(curr), m_loc(loc) {}
+  HostLocalStorageIt(HostLocalStorage<T> curr, std::int64_t loc) : m_curr(curr), m_loc(loc) {}
 
   constexpr HostLocalStorageIt() noexcept = default;
   constexpr HostLocalStorageIt(HostLocalStorageIt&&) noexcept = default;

include/pando-lib-galois/containers/per_thread.hpp

@@ -304,7 +304,7 @@ class PerThreadVector {
 
     // TODO(AdityaAtulTewari) Make this properly parallel.
     // Initialize the per host vectors
-    for (std::int16_t i = 0; i < static_cast<std::int16_t>(lift(flat, getNumHosts)); i++) {
+    for (std::int64_t i = 0; i < static_cast<std::int64_t>(lift(flat, getNumHosts)); i++) {
       auto place = pando::Place{pando::NodeIndex{i}, pando::anyPod, pando::anyCore};
       auto ref = fmap(flat, operator[], i);
       std::uint64_t start =
@@ -342,7 +342,7 @@ class PerThreadVector {
 
     // TODO(AdityaAtulTewari) Make this properly parallel.
     // Initialize the per host vectors
-    for (std::int16_t i = 0; i < static_cast<std::int16_t>(flat.getNumHosts()); i++) {
+    for (std::int64_t i = 0; i < static_cast<std::int64_t>(flat.getNumHosts()); i++) {
       auto ref = flat[i];
       std::uint64_t start =
           (i == 0) ? 0 : m_indices[static_cast<std::uint64_t>(i) * cores * threads - 1];

include/pando-lib-galois/graphs/dist_array_csr.hpp

@@ -169,7 +169,7 @@ class DistArrayCSR {
     err = vec.initialize(pando::getPlaceDims().node.id);
     PANDO_CHECK_RETURN(err);
 
-    for (std::int16_t i = 0; i < pando::getPlaceDims().node.id; i++) {
+    for (std::int64_t i = 0; i < pando::getPlaceDims().node.id; i++) {
       vec[i] = PlaceType{pando::Place{pando::NodeIndex{i}, pando::anyPod, pando::anyCore},
                          pando::MemoryType::Main};
     }
@@ -464,7 +464,7 @@ class DistArrayCSR {
     err = vec.initialize(pando::getPlaceDims().node.id);
     PANDO_CHECK_RETURN(err);
 
-    for (std::int16_t i = 0; i < pando::getPlaceDims().node.id; i++) {
+    for (std::int64_t i = 0; i < pando::getPlaceDims().node.id; i++) {
       vec[i] = PlaceType{pando::Place{pando::NodeIndex{i}, pando::anyPod, pando::anyCore},
                          pando::MemoryType::Main};
     }

include/pando-lib-galois/graphs/dist_local_csr.hpp

@@ -102,7 +102,7 @@ class DistLocalCSR {
       pointer ptr = m_pos + 1;
       CSR csrCurr = getCachedCSR(arrayOfCSRs, currNode);
       if (csrCurr.vertexEdgeOffsets.end() - 1 > ptr ||
-          (int16_t)currNode == pando::getPlaceDims().node.id - 1) {
+          (int64_t)currNode == pando::getPlaceDims().node.id - 1) {
         m_pos = ptr;
       } else {
         csrCurr = getCachedCSR(arrayOfCSRs, currNode + 1);
@@ -582,7 +582,7 @@ class DistLocalCSR {
     };
 
     for (std::uint64_t i = 0; i < numHosts; i++) {
-      pando::Place place = pando::Place{pando::NodeIndex{static_cast<std::int16_t>(i)},
+      pando::Place place = pando::Place{pando::NodeIndex{static_cast<std::int64_t>(i)},
                                         pando::anyPod, pando::anyCore};
       PANDO_CHECK(
           pando::executeOn(place, createCSRFuncs, this->arrayOfCSRs, vertexData, numEdges, i, wgh));
@@ -631,7 +631,7 @@ class DistLocalCSR {
           wgh.done();
         };
     for (std::uint64_t i = 0; i < numHosts; i++) {
-      pando::Place place = pando::Place{pando::NodeIndex{static_cast<std::int16_t>(i)},
+      pando::Place place = pando::Place{pando::NodeIndex{static_cast<std::int64_t>(i)},
                                         pando::anyPod, pando::anyCore};
       PANDO_CHECK_RETURN(
           pando::executeOn(place, fillCSRFuncs, *this, vertexData, edgeData, edgeMap, i, wgh));
@@ -791,7 +791,7 @@ class DistLocalCSR {
     PANDO_CHECK(wg.initialize(numHosts));
     auto wgh = wg.getHandle();
     for (std::uint64_t i = 0; i < numHosts; i++) {
-      pando::Place place = pando::Place{pando::NodeIndex{static_cast<std::int16_t>(i)},
+      pando::Place place = pando::Place{pando::NodeIndex{static_cast<std::int64_t>(i)},
                                         pando::anyPod, pando::anyCore};
       PANDO_CHECK(pando::executeOn(place, createMirrors, partEdges, mirrorList, V2PM, i, wgh));
     }
@@ -901,11 +901,11 @@ class DistLocalCSR {
       uint64_t numLocalEdges = edgesEnd - edgesStart;
       PANDO_CHECK(currentCSR.initializeTopologyMemory(
           vertex, numLocalEdges,
-          pando::Place{pando::NodeIndex{(int16_t)host}, pando::anyPod, pando::anyCore},
+          pando::Place{pando::NodeIndex{(int64_t)host}, pando::anyPod, pando::anyCore},
           pando::MemoryType::Main));
       PANDO_CHECK(currentCSR.initializeDataMemory(
           vertex, numLocalEdges,
-          pando::Place{pando::NodeIndex{(int16_t)host}, pando::anyPod, pando::anyCore},
+          pando::Place{pando::NodeIndex{(int64_t)host}, pando::anyPod, pando::anyCore},
           pando::MemoryType::Main));
 
       uint64_t currLocalEdge = 0;
@@ -922,7 +922,7 @@ class DistLocalCSR {
       currentCSR.vertexEdgeOffsets[vertex] = Vertex{&currentCSR.edgeDestinations[currLocalEdge]};
 
       PANDO_CHECK(fmap(arrayOfCSRs[host], initialize,
-                       pando::Place{pando::NodeIndex{(int16_t)host}, pando::anyPod, pando::anyCore},
+                       pando::Place{pando::NodeIndex{(int64_t)host}, pando::anyPod, pando::anyCore},
                        pando::MemoryType::Main));
       fmap(arrayOfCSRs[host], operator[], host) = currentCSR;
       edgesStart = edgesEnd;
@@ -1087,7 +1087,7 @@ class DistLocalCSR {
 
       std::uint64_t hosts = static_cast<std::uint64_t>(pando::getPlaceDims().node.id);
       for (std::uint64_t i = 0; i < numThreads; i++) {
-        pando::Place place = pando::Place{pando::NodeIndex{static_cast<std::int16_t>(i % hosts)},
+        pando::Place place = pando::Place{pando::NodeIndex{static_cast<std::int64_t>(i % hosts)},
                                           pando::anyPod, pando::anyCore};
         PANDO_CHECK_RETURN(
             pando::executeOn(place, &galois::internal::loadVertexFilePerThread<VertexType>,
@@ -1101,7 +1101,7 @@ class DistLocalCSR {
 
       std::uint64_t hosts = static_cast<std::uint64_t>(pando::getPlaceDims().node.id);
       for (std::uint64_t i = 0; i < numThreads; i++) {
-        pando::Place place = pando::Place{pando::NodeIndex{static_cast<std::int16_t>(i % hosts)},
+        pando::Place place = pando::Place{pando::NodeIndex{static_cast<std::int64_t>(i % hosts)},
                                           pando::anyPod, pando::anyCore};
         PANDO_CHECK_RETURN(pando::executeOn(
             place, &galois::internal::loadEdgeFilePerThread<EdgeType>, dones.getHandle(i),

include/pando-lib-galois/graphs/mirror_dist_local_csr.hpp

@@ -728,7 +728,7 @@ class MirrorDistLocalCSR {
     };
     uint64_t local_mirror_size = 0;
     for (std::uint64_t i = 0; i < numHosts; i++) {
-      pando::Place place = pando::Place{pando::NodeIndex{static_cast<std::int16_t>(i)},
+      pando::Place place = pando::Place{pando::NodeIndex{static_cast<std::int64_t>(i)},
                                         pando::anyPod, pando::anyCore};
       PANDO_CHECK(pando::executeOn(place, mirrorAttach, vertexData, mirrorList, i, wgh));
       local_mirror_size = lift(mirrorList[i], size);
@@ -774,7 +774,7 @@ class MirrorDistLocalCSR {
     };
 
     for (std::uint64_t i = 0; i < numHosts; i++) {
-      pando::Place place = pando::Place{pando::NodeIndex{static_cast<std::int16_t>(i)},
+      pando::Place place = pando::Place{pando::NodeIndex{static_cast<std::int64_t>(i)},
                                         pando::anyPod, pando::anyCore};
       PANDO_CHECK(
           pando::executeOn(place, generateMetadata, *this, this->dlcsr, mirrorList, i, wgh));
@@ -812,11 +812,11 @@ class MirrorDistLocalCSR {
 
     // initialize localMirrorToRemoteMasterOrderedTable
     PANDO_CHECK_RETURN(localMasterToRemoteMirrorTable.initialize());
-    for (std::int16_t i = 0; i < dims.node.id; i++) {
+    for (std::int64_t i = 0; i < dims.node.id; i++) {
       pando::GlobalRef<pando::Vector<pando::Vector<MirrorToMasterMap>>>
           localMasterToRemoteMirrorMap = localMasterToRemoteMirrorTable[i];
       PANDO_CHECK_RETURN(fmap(localMasterToRemoteMirrorMap, initialize, dims.node.id));
-      for (std::int16_t i = 0; i < dims.node.id; i++) {
+      for (std::int64_t i = 0; i < dims.node.id; i++) {
         pando::GlobalRef<pando::Vector<MirrorToMasterMap>> mapVectorFromHost =
             fmap(localMasterToRemoteMirrorMap, operator[], i);
         PANDO_CHECK_RETURN(fmap(mapVectorFromHost, initialize, 0));
@@ -855,23 +855,23 @@ class MirrorDistLocalCSR {
 
   /** Testing Purpose **/
   pando::GlobalRef<pando::Array<MirrorToMasterMap>> getLocalMirrorToRemoteMasterOrderedMap(
-      int16_t hostId) {
+      int64_t hostId) {
     return localMirrorToRemoteMasterOrderedTable[hostId];
   }
   pando::GlobalRef<pando::Vector<pando::Vector<MirrorToMasterMap>>> getLocalMasterToRemoteMirrorMap(
       uint64_t hostId) {
     return localMasterToRemoteMirrorTable[hostId];
   }
-  pando::GlobalRef<pando::Array<bool>> getMasterBitSet(int16_t hostId) {
+  pando::GlobalRef<pando::Array<bool>> getMasterBitSet(int64_t hostId) {
     return masterBitSets[hostId];
   }
-  pando::GlobalRef<pando::Array<bool>> getMirrorBitSet(int16_t hostId) {
+  pando::GlobalRef<pando::Array<bool>> getMirrorBitSet(int64_t hostId) {
     return mirrorBitSets[hostId];
   }
-  pando::GlobalRef<LocalVertexRange> getMasterRange(int16_t hostId) {
+  pando::GlobalRef<LocalVertexRange> getMasterRange(int64_t hostId) {
     return masterRange[hostId];
   }
-  pando::GlobalRef<LocalVertexRange> getMirrorRange(int16_t hostId) {
+  pando::GlobalRef<LocalVertexRange> getMirrorRange(int64_t hostId) {
     return mirrorRange[hostId];
   }
 

include/pando-lib-galois/import/ingest_rmat_el.hpp

@@ -78,7 +78,7 @@ ReturnType initializeELDACSR(pando::Array<char> filename, std::uint64_t numVerti
 
   PANDO_MEM_STAT_NEW_KERNEL("loadELFilePerThread Start");
   for (std::uint64_t i = 0; i < numThreads; i++) {
-    pando::Place place = pando::Place{pando::NodeIndex{static_cast<std::int16_t>(i % hosts)},
+    pando::Place place = pando::Place{pando::NodeIndex{static_cast<std::int64_t>(i % hosts)},
                                       pando::anyPod, pando::anyCore};
     PANDO_CHECK(pando::executeOn(place, &galois::loadELFilePerThread, wgh, filename, 1, numThreads,
                                  i, localReadEdges, perThreadRename, numVertices));
@@ -145,7 +145,7 @@ ReturnType initializeELDLCSR(pando::Array<char> filename, std::uint64_t numVerti
 
   PANDO_MEM_STAT_NEW_KERNEL("loadELFilePerThread Start");
   for (std::uint64_t i = 0; i < numThreads; i++) {
-    pando::Place place = pando::Place{pando::NodeIndex{static_cast<std::int16_t>(i % hosts)},
+    pando::Place place = pando::Place{pando::NodeIndex{static_cast<std::int64_t>(i % hosts)},
                                       pando::anyPod, pando::anyCore};
     PANDO_CHECK(pando::executeOn(place, &galois::loadELFilePerThread, wgh, filename, 1, numThreads,
                                  i, localReadEdges, perThreadRename, numVertices));

include/pando-lib-galois/import/ingest_wmd_csv.hpp

@@ -91,7 +91,7 @@ galois::DistLocalCSR<VertexType, EdgeType> initializeWMDDLCSR(pando::Array<char>
 
   PANDO_MEM_STAT_NEW_KERNEL("loadWMDFilePerThread Start");
   for (std::uint64_t i = 0; i < numThreads; i++) {
-    pando::Place place = pando::Place{pando::NodeIndex{static_cast<std::int16_t>(i % hosts)},
+    pando::Place place = pando::Place{pando::NodeIndex{static_cast<std::int64_t>(i % hosts)},
                                       pando::anyPod, pando::anyCore};
     PANDO_CHECK(pando::executeOn(place, &galois::loadWMDFilePerThread, wgh, filename, 1, numThreads,
                                  i, localReadEdges, perThreadRename, localReadVertices, totVerts));

include/pando-lib-galois/loops/do_all.hpp

@@ -311,17 +311,17 @@ class DoAll {
     const auto end = range.end();
 
     for (auto curr = range.begin(); curr != end; curr++) {
-      int16_t assignedHost;
+      int64_t assignedHost;
       if (workPerHost > 0) {
-        assignedHost = (int16_t)(*curr / workPerHost);
+        assignedHost = (int64_t)(*curr / workPerHost);
       } else if (hosts % workItems == 0) {
         // Given that we have 8 pxns and we schedule 4 workItems, this will place the workers on the
         // even numbered pxns
-        assignedHost = (int16_t)(*curr * (hosts / workItems));
+        assignedHost = (int64_t)(*curr * (hosts / workItems));
       } else {
         // Given that workItems < hosts and workItems does not divide hosts, assign the work
         // sequentially
-        assignedHost = (int16_t)(*curr);
+        assignedHost = (int64_t)(*curr);
       }
       if (assignedHost >= pando::getPlaceDims().node.id) {
         assignedHost = pando::getPlaceDims().node.id - 1;

include/pando-lib-galois/utility/locality.hpp

@@ -64,7 +64,7 @@ namespace galois {
                                       static_cast<std::uint64_t>(placeDims.core.y);
   const std::uint64_t threadsPerHost = threadsPerPod * static_cast<std::uint64_t>(placeDims.pod.x) *
                                        static_cast<std::uint64_t>(placeDims.pod.y);
-  const pando::NodeIndex node(static_cast<int16_t>(idx / threadsPerHost));
+  const pando::NodeIndex node(static_cast<int64_t>(idx / threadsPerHost));
   const std::uint64_t threadPerHostIdx = idx % threadsPerHost;
   const std::uint64_t podPerHostIdx = threadPerHostIdx / threadsPerPod;
   const pando::PodIndex pod(podPerHostIdx / placeDims.pod.y, podPerHostIdx % placeDims.pod.y);

microbench/bfs/include/pando-bfs-galois/sssp.hpp

@@ -296,7 +296,7 @@ pando::Status SSSPMDLCSR(G& graph, std::uint64_t src, HostLocalStorage<MDWorkLis
         if (found) {
           graph.setDataOnly(srcID, 0);
 
-          std::int16_t srcHost = graph.getPhysicalHostID(src);
+          std::int64_t srcHost = graph.getPhysicalHostID(src);
           if (srcHost == pando::getCurrentPlace().node.id) {
             PANDO_CHECK(fmap(toRead[0], pushBack, srcID));
           }

pando-rt/examples/gups.cpp

@@ -48,7 +48,7 @@ int pandoMain(int argc, char** argv) {
     tablePtr =
         static_cast<pando::GlobalPtr<std::int64_t>>(memoryResource->allocate(tableByteCount));
 
-    for (std::int16_t nodeId = 0; nodeId < placeDims.node.id; ++nodeId) {
+    for (std::int64_t nodeId = 0; nodeId < placeDims.node.id; ++nodeId) {
       for (std::int8_t podX = 0; podX < placeDims.pod.x; ++podX) {
         for (std::int8_t podY = 0; podY < placeDims.pod.y; ++podY) {
           for (std::int8_t coreX = 0; coreX < placeDims.core.x; ++coreX) {

pando-rt/examples/helloworld.cpp

@@ -26,7 +26,7 @@ void nodeGreetings(int level) {
 
   if (level < 2) {
     pando::NodeIndex rightNode{
-        static_cast<std::int16_t>((thisPlace.node.id + 1) % placeDims.node.id)};
+        static_cast<std::int64_t>((thisPlace.node.id + 1) % placeDims.node.id)};
     PANDO_CHECK(
         pando::executeOn(pando::Place{rightNode, {}, pando::anyCore}, &nodeGreetings, level + 1));
   }