add scanb back to the barrier, now other threads will wait for finish of work submission

qdp_project/src/Benchmark.cpp

@@ -81,6 +81,9 @@ void scan_b(size_t gid, size_t tid) {
     }
 
     THREAD_TIMING_[SCANB_TIMING_INDEX][UniqueIndex(gid,tid)][0][TIME_STAMP_WAIT] = std::chrono::steady_clock::now();
+
+    BARRIERS_[gid]->arrive_and_drop();
+
     THREAD_TIMING_[SCANB_TIMING_INDEX][UniqueIndex(gid,tid)][0][TIME_STAMP_END] = std::chrono::steady_clock::now();
 }
 
@@ -155,7 +158,6 @@ void aggr_j(size_t gid, size_t tid) {
 
         uint64_t tmp = _mm512_reduce_add_epi64(aggregator);
         aggregator = aggregation::apply_masked(aggregator, data_ptr, mask_ptr_a, SUBCHUNK_SIZE_B);
-
     }
 
     THREAD_TIMING_[AGGRJ_TIMING_INDEX][UniqueIndex(gid,tid)][0][TIME_STAMP_END] = std::chrono::steady_clock::now();
@@ -209,7 +211,7 @@ int main() {
         std::vector<std::thread>  agg_pool;
 
         for(uint32_t gid = 0; gid < GROUP_COUNT; ++gid) {
-            BARRIERS_.emplace_back(new std::barrier<NopStruct>(TC_SCANA + TC_AGGRJ));
+            BARRIERS_.emplace_back(new std::barrier<NopStruct>(TC_SCANA + TC_AGGRJ + TC_SCANB));
 
             for(uint32_t tid = 0; tid < TC_SCANA; ++tid) {
                 filter_pool.emplace_back(scan_a, gid, tid);