naive tbb

fix win build
calcite enable
512 kb is optimal

Author: Devjiu

omniscidb/QueryEngine/ColumnFetcher.cpp

@@ -147,8 +147,8 @@ JoinColumn ColumnFetcher::makeJoinColumn(
         data_provider,
         column_cache);
     if (col_buff != nullptr) {
+      join_chunk_array[num_chunks] = JoinChunk{col_buff, elem_count, num_elems};
       num_elems += elem_count;
-      join_chunk_array[num_chunks] = JoinChunk{col_buff, elem_count};
     } else {
       continue;
     }
@@ -359,12 +359,12 @@ const int8_t* ColumnFetcher::getAllTableColumnFragments(
 
     std::vector<int8_t*> raw_write_ptrs;
     raw_write_ptrs.reserve(frag_count);
-    for (uint i = 0; i < frag_count; i++) {
+    for (size_t i = 0; i < frag_count; i++) {
       raw_write_ptrs.emplace_back(write_ptrs[i].first);
     }
 
-    std::unique_ptr<ColumnarResults> merged_results(
-        new ColumnarResults(raw_write_ptrs, total_row_count, col_info->type, thread_idx));
+    std::unique_ptr<ColumnarResults> merged_results(new ColumnarResults(
+        std::move(raw_write_ptrs), total_row_count, col_info->type, thread_idx));
 
     table_column = merged_results.get();
     columnarized_scan_table_cache_.emplace(std::make_pair(table_id, col_id),

omniscidb/QueryEngine/JoinHashTable/Builders/PerfectHashTableBuilder.h

@@ -166,8 +166,6 @@ class PerfectJoinHashTableBuilder {
                                            0);
 
     auto cpu_hash_table_buff = reinterpret_cast<int32_t*>(hash_table_->getCpuBuffer());
-    // const int thread_count = cpu_threads();
-    // std::vector<std::thread> init_cpu_buff_threads;
 
     {
       auto timer_init = DEBUG_TIMER("CPU One-To-One Perfect-Hash: init_hash_join_buff");
@@ -176,13 +174,11 @@ class PerfectJoinHashTableBuilder {
                           hash_join_invalid_val);
     }
     const bool for_semi_join = for_semi_anti_join(join_type);
-    // std::atomic<int> err{0};
     {
       auto timer_fill =
-          DEBUG_TIMER("CPU One-To-One Perfect-Hash: fill_hash_join_buff_bucketized");
+          DEBUG_TIMER("CPU One-To-One Perfect-Hash: fill_hash_join_buff_bucketized_cpu");
 
       {
-        // int partial_err = 0;
         JoinColumnTypeInfo type_info{static_cast<size_t>(type->size()),
                                      col_range.getIntMin(),
                                      col_range.getIntMax(),
@@ -204,7 +200,6 @@ class PerfectJoinHashTableBuilder {
         if (error) {
           // Too many hash entries, need to retry with a 1:many table
           hash_table_ = nullptr;  // clear the hash table buffer
-          // LOG(ERROR) << "cpu hash_tabel error: " << error;
           throw NeedsOneToManyHash();
         }
       }