GroupByAndAggregate.h File Reference

#include "BufferCompaction.h"
#include "ColumnarResults.h"
#include "CompilationOptions.h"
#include "GpuMemUtils.h"
#include "GpuSharedMemoryContext.h"
#include "InputMetadata.h"
#include "QueryExecutionContext.h"
#include "Rendering/RenderInfo.h"
#include "RuntimeFunctions.h"
#include "QueryEngine/Utils/DiamondCodegen.h"
#include "../Shared/sqltypes.h"
#include "Logger/Logger.h"
#include <llvm/IR/Function.h>
#include <llvm/IR/Instructions.h>
#include <llvm/IR/Value.h>
#include <boost/algorithm/string/join.hpp>
#include <boost/make_unique.hpp>
#include <stack>
#include <vector>

Include dependency graph for GroupByAndAggregate.h:

This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Classes
struct	ColRangeInfo
struct	KeylessInfo
class	GroupByAndAggregate

Functions
size_t	get_count_distinct_sub_bitmap_count (const size_t bitmap_sz_bits, const RelAlgExecutionUnit &ra_exe_unit, const ExecutorDeviceType device_type)

Variables
bool	g_enable_smem_group_by
bool	g_bigint_count

Function Documentation

size_t get_count_distinct_sub_bitmap_count ( const size_t  bitmap_sz_bits, const RelAlgExecutionUnit &  ra_exe_unit, const ExecutorDeviceType  device_type  )

inline

Definition at line 225 of file GroupByAndAggregate.h.

References g_cluster, GPU, and RelAlgExecutionUnit::groupby_exprs.

Referenced by anonymous_namespace{RelAlgExecutor.cpp}::decide_approx_count_distinct_implementation(), and anonymous_namespace{GroupByAndAggregate.cpp}::init_count_distinct_descriptors().

                                                                                        {
  // For count distinct on a column with a very small number of distinct values
  // contention can be very high, especially for non-grouped queries. We'll split
  // the bitmap into multiple sub-bitmaps which are unified to get the full result.
  // The threshold value for bitmap_sz_bits works well on Kepler.
  return bitmap_sz_bits < 50000 && ra_exe_unit.groupby_exprs.empty() &&
                 (device_type == ExecutorDeviceType::GPU || g_cluster)
             ? 64  // NB: must be a power of 2 to keep runtime offset computations cheap
             : 1;
}

Here is the caller graph for this function:

Variable Documentation

bool g_bigint_count

Definition at line 54 of file GroupByAndAggregate.cpp.

bool g_enable_smem_group_by