fix prefetching subchunk indexing and adapt the weak access flag for join

qdp_project/src/Benchmark.cpp

@@ -41,6 +41,7 @@ uint64_t* DATA_DST_;
 void scan_b(size_t gid, size_t tid) {
     constexpr size_t SUBCHUNK_COUNT = TC_AGGRJ / (TC_SCANB == 0 ? 1 : TC_SCANB);
     constexpr size_t SUBCHUNK_SIZE_B = CHUNK_SIZE_B / SUBCHUNK_COUNT;
+    constexpr size_t SUBCHUNK_SIZE_ELEMENTS = CHUNK_SIZE_ELEMENTS / SUBCHUNK_COUNT;
 
     THREAD_TIMING_[SCANB_TIMING_INDEX][UniqueIndex(gid,tid)].clear();
     THREAD_TIMING_[SCANB_TIMING_INDEX][UniqueIndex(gid,tid)].resize(1);
@@ -55,7 +56,7 @@ void scan_b(size_t gid, size_t tid) {
             uint64_t* chunk_ptr = get_chunk<TC_SCANB>(DATA_B_, chunk_index, i);
 
             for (size_t j = 0; j < SUBCHUNK_COUNT; j++) {
-                uint64_t* sub_chunk_ptr = &chunk_ptr[j * CHUNK_SIZE_ELEMENTS];
+                uint64_t* sub_chunk_ptr = &chunk_ptr[j * SUBCHUNK_SIZE_ELEMENTS];
                 CACHE_.Access(reinterpret_cast<uint8_t*>(sub_chunk_ptr), SUBCHUNK_SIZE_B);
             }
         }
@@ -125,7 +126,7 @@ void aggr_j(size_t gid, size_t tid) {
         uint64_t* data_ptr;
 
         if constexpr (PERFORM_CACHING) {
-            data = CACHE_.Access(reinterpret_cast<uint8_t *>(chunk_ptr), CHUNK_SIZE_B / TC_AGGRJ);
+            data = CACHE_.Access(reinterpret_cast<uint8_t *>(chunk_ptr), CHUNK_SIZE_B / TC_AGGRJ, dsacache::FLAG_ACCESS_WEAK);
             data->WaitOnCompletion();
             data_ptr = reinterpret_cast<uint64_t*>(data->GetDataLocation());