GroupByAndAggregate.cpp File Reference

#include "GroupByAndAggregate.h"
#include "AggregateUtils.h"
#include "CardinalityEstimator.h"
#include "CodeGenerator.h"
#include "Descriptors/QueryMemoryDescriptor.h"
#include "ExpressionRange.h"
#include "ExpressionRewrite.h"
#include "GpuInitGroups.h"
#include "InPlaceSort.h"
#include "LLVMFunctionAttributesUtil.h"
#include "MaxwellCodegenPatch.h"
#include "OutputBufferInitialization.h"
#include "TargetExprBuilder.h"
#include "../CudaMgr/CudaMgr.h"
#include "../Shared/checked_alloc.h"
#include "../Shared/funcannotations.h"
#include "../Utils/ChunkIter.h"
#include "DataMgr/BufferMgr/BufferMgr.h"
#include "Execute.h"
#include "QueryTemplateGenerator.h"
#include "RuntimeFunctions.h"
#include "Shared/misc.h"
#include "StreamingTopN.h"
#include "TopKSort.h"
#include "WindowContext.h"
#include <llvm/Transforms/Utils/BasicBlockUtils.h>
#include <cstring>
#include <limits>
#include <numeric>
#include <string_view>
#include <thread>

Include dependency graph for GroupByAndAggregate.cpp:

Go to the source code of this file.

Namespaces
	anonymous_namespace{GroupByAndAggregate.cpp}

Macros
#define	LL_CONTEXT   executor_->cgen_state_->context_
#define	LL_BUILDER   executor_->cgen_state_->ir_builder_
#define	LL_BOOL(v)   executor_->cgen_state_->llBool(v)
#define	LL_INT(v)   executor_->cgen_state_->llInt(v)
#define	LL_FP(v)   executor_->cgen_state_->llFp(v)
#define	ROW_FUNC   executor_->cgen_state_->row_func_
#define	CUR_FUNC   executor_->cgen_state_->current_func_

Functions
std::ostream &	operator<< (std::ostream &out, const ColRangeInfo &info)
std::ostream &	operator<< (std::ostream &out, const CountDistinctImplType &type)
std::ostream &	operator<< (std::ostream &out, const CountDistinctDescriptor &desc)
int32_t	anonymous_namespace{GroupByAndAggregate.cpp}::get_agg_count (const std::vector< Analyzer::Expr * > &target_exprs)
bool	anonymous_namespace{GroupByAndAggregate.cpp}::expr_is_rowid (const Analyzer::Expr *expr)
bool	anonymous_namespace{GroupByAndAggregate.cpp}::has_count_distinct (const RelAlgExecutionUnit &ra_exe_unit)
bool	anonymous_namespace{GroupByAndAggregate.cpp}::is_column_range_too_big_for_perfect_hash (const ColRangeInfo &col_range_info, const int64_t max_entry_count)
bool	anonymous_namespace{GroupByAndAggregate.cpp}::cardinality_estimate_less_than_column_range (const int64_t cardinality_estimate, const ColRangeInfo &col_range_info)
ColRangeInfo	anonymous_namespace{GroupByAndAggregate.cpp}::get_expr_range_info (const RelAlgExecutionUnit &ra_exe_unit, const std::vector< InputTableInfo > &query_infos, const Analyzer::Expr *expr, Executor *executor)
int64_t	anonymous_namespace{GroupByAndAggregate.cpp}::get_bucketed_cardinality_without_nulls (const ColRangeInfo &col_range_info)
KeylessInfo	anonymous_namespace{GroupByAndAggregate.cpp}::get_keyless_info (const RelAlgExecutionUnit &ra_exe_unit, const std::vector< InputTableInfo > &query_infos, const bool is_group_by, Executor *executor)
CountDistinctDescriptors	anonymous_namespace{GroupByAndAggregate.cpp}::init_count_distinct_descriptors (const RelAlgExecutionUnit &ra_exe_unit, const std::vector< InputTableInfo > &query_infos, const ColRangeInfo &group_by_range_info, const ExecutorDeviceType device_type, Executor *executor)
RUNTIME_EXPORT void	agg_count_distinct (int64_t *agg, const int64_t val)
RUNTIME_EXPORT void	agg_count_distinct_skip_val (int64_t *agg, const int64_t val, const int64_t skip_val)
RUNTIME_EXPORT void	agg_approx_quantile (int64_t *agg, const double val)
RUNTIME_EXPORT void	agg_mode_func (int64_t *agg, const int64_t val)

Variables
bool g_cluster	false
bool	g_bigint_count {false}
int	g_hll_precision_bits {11}
size_t	g_watchdog_baseline_max_groups {120000000}
int64_t	g_bitmap_memory_limit
size_t	g_leaf_count

Macro Definition Documentation

#define CUR_FUNC   executor_->cgen_state_->current_func_

Definition at line 389 of file GroupByAndAggregate.cpp.

Referenced by GroupByAndAggregate::codegenAggArg().

#define LL_BOOL

(

v

)

executor_->cgen_state_->llBool(v)

Definition at line 385 of file GroupByAndAggregate.cpp.

Referenced by GroupByAndAggregate::codegenOutputSlot().

#define LL_BUILDER   executor_->cgen_state_->ir_builder_

Definition at line 384 of file GroupByAndAggregate.cpp.

Referenced by GroupByAndAggregate::codegen(), GroupByAndAggregate::codegenAggArg(), GroupByAndAggregate::codegenAggCalls(), GroupByAndAggregate::codegenAggColumnPtr(), GroupByAndAggregate::codegenEstimator(), GroupByAndAggregate::codegenGroupBy(), GroupByAndAggregate::codegenMultiColumnBaselineHash(), GroupByAndAggregate::codegenMultiColumnPerfectHash(), GroupByAndAggregate::codegenOutputSlot(), GroupByAndAggregate::codegenWindowRowPointer(), GroupByAndAggregate::convertNullIfAny(), and GroupByAndAggregate::getAdditionalLiteral().

#define LL_CONTEXT   executor_->cgen_state_->context_

Definition at line 383 of file GroupByAndAggregate.cpp.

Referenced by GroupByAndAggregate::codegen(), GroupByAndAggregate::codegenAggArg(), GroupByAndAggregate::codegenAggCalls(), GroupByAndAggregate::codegenAggColumnPtr(), GroupByAndAggregate::codegenCountDistinct(), GroupByAndAggregate::codegenEstimator(), GroupByAndAggregate::codegenGroupBy(), GroupByAndAggregate::codegenMultiColumnBaselineHash(), GroupByAndAggregate::codegenMultiColumnPerfectHash(), GroupByAndAggregate::codegenPerfectHashFunction(), GroupByAndAggregate::codegenVarlenOutputBuffer(), GroupByAndAggregate::codegenWindowRowPointer(), and GroupByAndAggregate::getAdditionalLiteral().

#define LL_FP

(

v

)

executor_->cgen_state_->llFp(v)

Definition at line 387 of file GroupByAndAggregate.cpp.

Referenced by GroupByAndAggregate::codegenOutputSlot().

#define LL_INT

(

v

)

executor_->cgen_state_->llInt(v)

Definition at line 386 of file GroupByAndAggregate.cpp.

Referenced by GroupByAndAggregate::codegen(), GroupByAndAggregate::codegenAggColumnPtr(), GroupByAndAggregate::codegenCountDistinct(), GroupByAndAggregate::codegenEstimator(), GroupByAndAggregate::codegenGroupBy(), GroupByAndAggregate::codegenMultiColumnBaselineHash(), GroupByAndAggregate::codegenMultiColumnPerfectHash(), GroupByAndAggregate::codegenOutputSlot(), GroupByAndAggregate::codegenPerfectHashFunction(), GroupByAndAggregate::codegenSingleColumnPerfectHash(), GroupByAndAggregate::codegenWindowRowPointer(), and GroupByAndAggregate::getAdditionalLiteral().

#define ROW_FUNC   executor_->cgen_state_->row_func_

Definition at line 388 of file GroupByAndAggregate.cpp.

Referenced by GroupByAndAggregate::codegen(), GroupByAndAggregate::codegenEstimator(), GroupByAndAggregate::codegenGroupBy(), GroupByAndAggregate::codegenOutputSlot(), GroupByAndAggregate::codegenVarlenOutputBuffer(), GroupByAndAggregate::codegenWindowRowPointer(), and GroupByAndAggregate::getAdditionalLiteral().

Function Documentation

RUNTIME_EXPORT void agg_approx_quantile	(	int64_t *	agg,
		const double	val
	)

Definition at line 1845 of file GroupByAndAggregate.cpp.

References quantile::detail::TDigest< RealType, IndexType >::allocate().

                                                                                   {
  auto* t_digest = reinterpret_cast<quantile::TDigest*>(*agg);
  t_digest->allocate();
  t_digest->add(val);
}

Here is the call graph for this function:

RUNTIME_EXPORT void agg_count_distinct	(	int64_t *	agg,
		const int64_t	val
	)

Definition at line 1833 of file GroupByAndAggregate.cpp.

Referenced by agg_count_distinct_skip_val().

                                                                                   {
  reinterpret_cast<CountDistinctSet*>(*agg)->insert(val);
}

Here is the caller graph for this function:

RUNTIME_EXPORT void agg_count_distinct_skip_val	(	int64_t *	agg,
		const int64_t	val,
		const int64_t	skip_val
	)

Definition at line 1837 of file GroupByAndAggregate.cpp.

References agg_count_distinct().

                                                                                   {
  if (val != skip_val) {
    agg_count_distinct(agg, val);
  }
}

Here is the call graph for this function:

RUNTIME_EXPORT void agg_mode_func	(	int64_t *	agg,
		const int64_t	val
	)

Definition at line 1851 of file GroupByAndAggregate.cpp.

References AggMode::add().

                                                                              {
  auto* mode_map = reinterpret_cast<AggMode*>(*agg);
  mode_map->add(val);
}

Here is the call graph for this function:

std::ostream& operator<<	(	std::ostream &	out,
		const ColRangeInfo &	info
	)

Definition at line 64 of file GroupByAndAggregate.cpp.

References ColRangeInfo::bucket, ColRangeInfo::has_nulls, ColRangeInfo::hash_type_, ColRangeInfo::max, and ColRangeInfo::min.

                                                                  {
  out << "Hash Type = " << info.hash_type_ << " min = " << info.min
      << " max = " << info.max << " bucket = " << info.bucket
      << " has_nulls = " << info.has_nulls << "\n";
  return out;
}

std::ostream& operator<<	(	std::ostream &	out,
		const CountDistinctImplType &	type
	)

Definition at line 71 of file GroupByAndAggregate.cpp.

References Bitmap, Invalid, and UnorderedSet.

                                                                           {
  switch (type) {
    case CountDistinctImplType::Invalid:
      out << "Invalid";
      break;
    case CountDistinctImplType::Bitmap:
      out << "Bitmap";
      break;
    case CountDistinctImplType::UnorderedSet:
      out << "UnorderedSet";
      break;
    default:
      out << "<Unkown Type>";
      break;
  }
  return out;
}

std::ostream& operator<<	(	std::ostream &	out,
		const CountDistinctDescriptor &	desc
	)

Definition at line 89 of file GroupByAndAggregate.cpp.

References CountDistinctDescriptor::approximate, CountDistinctDescriptor::bitmap_sz_bits, CountDistinctDescriptor::device_type, CountDistinctDescriptor::impl_type_, CountDistinctDescriptor::min_val, and CountDistinctDescriptor::sub_bitmap_count.

                                                                             {
  out << "Type = " << desc.impl_type_ << " min val = " << desc.min_val
      << " bitmap_sz_bits = " << desc.bitmap_sz_bits
      << " bool approximate = " << desc.approximate
      << " device_type = " << desc.device_type
      << " sub_bitmap_count = " << desc.sub_bitmap_count;
  return out;
}

Variable Documentation

bool g_cluster false

Definition at line 53 of file GroupByAndAggregate.cpp.

bool g_bigint_count {false}

Definition at line 54 of file GroupByAndAggregate.cpp.

Referenced by QueryMemoryInitializer::allocateCountDistinctBuffers(), QueryMemoryInitializer::calculateCountDistinctBufferSize(), anonymous_namespace{QueryMemoryInitializer.cpp}::check_count_distinct_expr_metadata(), TargetExprCodegen::codegen(), GroupByAndAggregate::codegenApproxQuantile(), GroupByAndAggregate::codegenCountDistinct(), ColSlotContext::ColSlotContext(), anonymous_namespace{RelAlgExecutor.cpp}::decide_approx_count_distinct_implementation(), Executor::executePlanWithoutGroupBy(), anonymous_namespace{Execute.cpp}::fill_entries_for_empty_input(), CommandLineOptions::fillOptions(), get_agg_type(), anonymous_namespace{QueryMemoryDescriptor.cpp}::get_col_byte_widths(), get_heap_key_slot_index(), anonymous_namespace{GroupByAndAggregate.cpp}::get_keyless_info(), Executor::getDeviceTypeForTargets(), RelAlgExecutor::getFilteredCountAll(), RelAlgExecutor::getFilterSelectivity(), anonymous_namespace{GroupByAndAggregate.cpp}::has_count_distinct(), init_agg_val_vec(), anonymous_namespace{GroupByAndAggregate.cpp}::init_count_distinct_descriptors(), QueryMemoryInitializer::initColumnarGroups(), TargetExprCodegenBuilder::operator()(), anonymous_namespace{Execute.cpp}::GetTargetInfo::operator()(), QueryMemoryDescriptor::pick_target_compact_width(), target_exprs_to_infos(), and RelAlgTranslator::translateWindowFunction().

int64_t g_bitmap_memory_limit

Definition at line 28 of file QueryMemoryInitializer.cpp.

Referenced by anonymous_namespace{QueryMemoryInitializer.cpp}::check_total_bitmap_memory(), CommandLineOptions::fillDeveloperOptions(), anonymous_namespace{GroupByAndAggregate.cpp}::init_count_distinct_descriptors(), and InValuesBitmap::InValuesBitmap().

int g_hll_precision_bits {11}

Definition at line 55 of file GroupByAndAggregate.cpp.

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

size_t g_leaf_count

Definition at line 78 of file ParserNode.cpp.

size_t g_watchdog_baseline_max_groups {120000000}

Definition at line 56 of file GroupByAndAggregate.cpp.

Referenced by GroupByAndAggregate::initQueryMemoryDescriptorImpl().