1
0
mirror of https://github.com/mariadb-corporation/mariadb-columnstore-engine.git synced 2025-07-30 19:23:07 +03:00
Files
mariadb-columnstore-engine/dbcon/execplan/arithmeticoperator.h
Gagan Goel 1fc399451a MCOL-4957 Fix performance slowdown for processing TIMESTAMP columns.
Part 1:
 As part of MCOL-3776 to address synchronization issue while accessing
 the fTimeZone member of the Func class, mutex locks were added to the
 accessor and mutator methods. However, this slows down processing
 of TIMESTAMP columns in PrimProc significantly as all threads across
 all concurrently running queries would serialize on the mutex. This
 is because PrimProc only has a single global object for the functor
 class (class derived from Func in utils/funcexp/functor.h) for a given
 function name. To fix this problem:

   (1) We remove the fTimeZone as a member of the Func derived classes
   (hence removing the mutexes) and instead use the fOperationType
   member of the FunctionColumn class to propagate the timezone values
   down to the individual functor processing functions such as
   FunctionColumn::getStrVal(), FunctionColumn::getIntVal(), etc.

   (2) To achieve (1), a timezone member is added to the
   execplan::CalpontSystemCatalog::ColType class.

Part 2:
 Several functors in the Funcexp code call dataconvert::gmtSecToMySQLTime()
 and dataconvert::mySQLTimeToGmtSec() functions for conversion between seconds
 since unix epoch and broken-down representation. These functions in turn call
 the C library function localtime_r() which currently has a known bug of holding
 a global lock via a call to __tz_convert. This significantly reduces performance
 in multi-threaded applications where multiple threads concurrently call
 localtime_r(). More details on the bug:
   https://sourceware.org/bugzilla/show_bug.cgi?id=16145

 This bug in localtime_r() caused processing of the Functors in PrimProc to
 slowdown significantly since a query execution causes Functors code to be
 processed in a multi-threaded manner.

 As a fix, we remove the calls to localtime_r() from gmtSecToMySQLTime()
 and mySQLTimeToGmtSec() by performing the timezone-to-offset conversion
 (done in dataconvert::timeZoneToOffset()) during the execution plan
 creation in the plugin. Note that localtime_r() is only called when the
 time_zone system variable is set to "SYSTEM".

 This fix also required changing the timezone type from a std::string to
 a long across the system.
2022-02-11 19:03:32 -05:00

438 lines
13 KiB
C++

/* Copyright (C) 2014 InfiniDB, Inc.
Copyright (C) 2019 MariaDB Corporation
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License
as published by the Free Software Foundation; version 2 of
the License.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston,
MA 02110-1301, USA. */
/***********************************************************************
* $Id$
*
*
***********************************************************************/
/** @file */
#ifndef ARITHMETICOPERATOR_H
#define ARITHMETICOPERATOR_H
#include <string>
#include <iosfwd>
#include <cmath>
#include <sstream>
#include "operator.h"
#include "parsetree.h"
namespace messageqcpp
{
class ByteStream;
}
namespace execplan
{
class ArithmeticOperator : public Operator
{
using cscType = execplan::CalpontSystemCatalog::ColType;
public:
ArithmeticOperator();
ArithmeticOperator(const std::string& operatorName);
ArithmeticOperator(const ArithmeticOperator& rhs);
virtual ~ArithmeticOperator();
/** return a copy of this pointer
*
* deep copy of this pointer and return the copy
*/
inline virtual ArithmeticOperator* clone() const
{
return new ArithmeticOperator(*this);
}
inline long timeZone() const
{
return fTimeZone;
}
inline void timeZone(const long timeZone)
{
fTimeZone = timeZone;
}
/**
* The serialization interface
*/
virtual void serialize(messageqcpp::ByteStream&) const;
virtual void unserialize(messageqcpp::ByteStream&);
/** @brief Do a deep, strict (as opposed to semantic) equivalence test
*
* Do a deep, strict (as opposed to semantic) equivalence test.
* @return true iff every member of t is a duplicate copy of every member of this; false otherwise
*/
virtual bool operator==(const TreeNode* t) const;
/** @brief Do a deep, strict (as opposed to semantic) equivalence test
*
* Do a deep, strict (as opposed to semantic) equivalence test.
* @return true iff every member of t is a duplicate copy of every member of this; false otherwise
*/
bool operator==(const ArithmeticOperator& t) const;
/** @brief Do a deep, strict (as opposed to semantic) equivalence test
*
* Do a deep, strict (as opposed to semantic) equivalence test.
* @return false iff every member of t is a duplicate copy of every member of this; true otherwise
*/
virtual bool operator!=(const TreeNode* t) const;
/** @brief Do a deep, strict (as opposed to semantic) equivalence test
*
* Do a deep, strict (as opposed to semantic) equivalence test.
* @return false iff every member of t is a duplicate copy of every member of this; true otherwise
*/
bool operator!=(const ArithmeticOperator& t) const;
/***********************************************************
* F&E framework *
***********************************************************/
using Operator::evaluate;
inline virtual void evaluate(rowgroup::Row& row, bool& isNull, ParseTree* lop, ParseTree* rop);
using Operator::getStrVal;
virtual const std::string& getStrVal(rowgroup::Row& row, bool& isNull, ParseTree* lop, ParseTree* rop)
{
evaluate(row, isNull, lop, rop);
return TreeNode::getStrVal(fTimeZone);
}
using Operator::getIntVal;
virtual int64_t getIntVal(rowgroup::Row& row, bool& isNull, ParseTree* lop, ParseTree* rop)
{
evaluate(row, isNull, lop, rop);
return TreeNode::getIntVal();
}
using Operator::getUintVal;
virtual uint64_t getUintVal(rowgroup::Row& row, bool& isNull, ParseTree* lop, ParseTree* rop)
{
evaluate(row, isNull, lop, rop);
return TreeNode::getUintVal();
}
using Operator::getFloatVal;
virtual float getFloatVal(rowgroup::Row& row, bool& isNull, ParseTree* lop, ParseTree* rop)
{
evaluate(row, isNull, lop, rop);
return TreeNode::getFloatVal();
}
using Operator::getDoubleVal;
virtual double getDoubleVal(rowgroup::Row& row, bool& isNull, ParseTree* lop, ParseTree* rop)
{
evaluate(row, isNull, lop, rop);
return TreeNode::getDoubleVal();
}
using Operator::getLongDoubleVal;
virtual long double getLongDoubleVal(rowgroup::Row& row, bool& isNull, ParseTree* lop, ParseTree* rop)
{
evaluate(row, isNull, lop, rop);
return TreeNode::getLongDoubleVal();
}
using Operator::getDecimalVal;
virtual IDB_Decimal getDecimalVal(rowgroup::Row& row, bool& isNull, ParseTree* lop, ParseTree* rop)
{
evaluate(row, isNull, lop, rop);
// @bug5736, double type with precision -1 indicates that this type is for decimal math,
// the original decimal scale is stored in scale field, which is no use for double.
if (fResultType.colDataType == CalpontSystemCatalog::DOUBLE && fResultType.precision == -1)
{
IDB_Decimal rv;
rv.scale = fResultType.scale;
rv.precision = 15;
rv.value = (int64_t)(TreeNode::getDoubleVal() * IDB_pow[rv.scale]);
return rv;
}
return TreeNode::getDecimalVal();
}
using Operator::getDateIntVal;
virtual int32_t getDateIntVal(rowgroup::Row& row, bool& isNull, ParseTree* lop, ParseTree* rop)
{
evaluate(row, isNull, lop, rop);
return TreeNode::getDateIntVal();
}
using Operator::getDatetimeIntVal;
virtual int64_t getDatetimeIntVal(rowgroup::Row& row, bool& isNull, ParseTree* lop, ParseTree* rop)
{
evaluate(row, isNull, lop, rop);
return TreeNode::getDatetimeIntVal();
}
using Operator::getTimestampIntVal;
virtual int64_t getTimestampIntVal(rowgroup::Row& row, bool& isNull, ParseTree* lop, ParseTree* rop)
{
evaluate(row, isNull, lop, rop);
return TreeNode::getTimestampIntVal();
}
using Operator::getTimeIntVal;
virtual int64_t getTimeIntVal(rowgroup::Row& row, bool& isNull, ParseTree* lop, ParseTree* rop)
{
evaluate(row, isNull, lop, rop);
return TreeNode::getTimeIntVal();
}
using Operator::getBoolVal;
virtual bool getBoolVal(rowgroup::Row& row, bool& isNull, ParseTree* lop, ParseTree* rop)
{
evaluate(row, isNull, lop, rop);
return TreeNode::getBoolVal();
}
void adjustResultType(const CalpontSystemCatalog::ColType& m);
inline bool getOverflowCheck() const
{
return fDecimalOverflowCheck;
}
inline void setOverflowCheck(bool check)
{
fDecimalOverflowCheck = check;
}
private:
template <typename result_t>
inline result_t execute(result_t op1, result_t op2, bool& isNull);
inline void execute(IDB_Decimal& result, IDB_Decimal op1, IDB_Decimal op2, bool& isNull);
long fTimeZone;
bool fDecimalOverflowCheck;
};
#include "parsetree.h"
inline void ArithmeticOperator::evaluate(rowgroup::Row& row, bool& isNull, ParseTree* lop, ParseTree* rop)
{
// fOpType should have already been set on the connector during parsing
switch (fOperationType.colDataType)
{
case execplan::CalpontSystemCatalog::BIGINT:
case execplan::CalpontSystemCatalog::INT:
case execplan::CalpontSystemCatalog::MEDINT:
case execplan::CalpontSystemCatalog::SMALLINT:
case execplan::CalpontSystemCatalog::TINYINT:
fResult.intVal = execute(lop->getIntVal(row, isNull), rop->getIntVal(row, isNull), isNull);
break;
case execplan::CalpontSystemCatalog::UBIGINT:
case execplan::CalpontSystemCatalog::UINT:
case execplan::CalpontSystemCatalog::UMEDINT:
case execplan::CalpontSystemCatalog::USMALLINT:
case execplan::CalpontSystemCatalog::UTINYINT:
fResult.uintVal = execute(lop->getUintVal(row, isNull), rop->getUintVal(row, isNull), isNull);
break;
case execplan::CalpontSystemCatalog::DOUBLE:
case execplan::CalpontSystemCatalog::FLOAT:
case execplan::CalpontSystemCatalog::UDOUBLE:
case execplan::CalpontSystemCatalog::UFLOAT:
fResult.doubleVal = execute(lop->getDoubleVal(row, isNull), rop->getDoubleVal(row, isNull), isNull);
break;
case execplan::CalpontSystemCatalog::LONGDOUBLE:
fResult.longDoubleVal =
execute(lop->getLongDoubleVal(row, isNull), rop->getLongDoubleVal(row, isNull), isNull);
break;
case execplan::CalpontSystemCatalog::DECIMAL:
case execplan::CalpontSystemCatalog::UDECIMAL:
execute(fResult.decimalVal, lop->getDecimalVal(row, isNull), rop->getDecimalVal(row, isNull), isNull);
break;
default:
{
std::ostringstream oss;
oss << "invalid arithmetic operand type: " << fOperationType.colDataType;
throw logging::InvalidArgumentExcept(oss.str());
}
}
}
template <typename result_t>
inline result_t ArithmeticOperator::execute(result_t op1, result_t op2, bool& isNull)
{
switch (fOp)
{
case OP_ADD: return op1 + op2;
case OP_SUB: return op1 - op2;
case OP_MUL: return op1 * op2;
case OP_DIV:
if (op2)
return op1 / op2;
else
isNull = true;
return 0;
default:
{
std::ostringstream oss;
oss << "invalid arithmetic operation: " << fOp;
throw logging::InvalidOperationExcept(oss.str());
}
}
}
inline void ArithmeticOperator::execute(IDB_Decimal& result, IDB_Decimal op1, IDB_Decimal op2, bool& isNull)
{
switch (fOp)
{
case OP_ADD:
if (fOperationType.colWidth == datatypes::MAXDECIMALWIDTH)
{
if (LIKELY(!fDecimalOverflowCheck))
{
datatypes::Decimal::addition<decltype(result.s128Value), false>(op1, op2, result);
}
else
{
datatypes::Decimal::addition<decltype(result.s128Value), true>(op1, op2, result);
}
}
else if (fOperationType.colWidth == utils::MAXLEGACYWIDTH)
{
if (LIKELY(!fDecimalOverflowCheck))
{
datatypes::Decimal::addition<decltype(result.value), false>(op1, op2, result);
}
else
{
datatypes::Decimal::addition<decltype(result.value), true>(op1, op2, result);
}
}
else
{
throw logging::InvalidArgumentExcept("Unexpected result width");
}
break;
case OP_SUB:
if (fOperationType.colWidth == datatypes::MAXDECIMALWIDTH)
{
if (LIKELY(!fDecimalOverflowCheck))
{
datatypes::Decimal::subtraction<decltype(result.s128Value), false>(op1, op2, result);
}
else
{
datatypes::Decimal::subtraction<decltype(result.s128Value), true>(op1, op2, result);
}
}
else if (fOperationType.colWidth == utils::MAXLEGACYWIDTH)
{
if (LIKELY(!fDecimalOverflowCheck))
{
datatypes::Decimal::subtraction<decltype(result.value), false>(op1, op2, result);
}
else
{
datatypes::Decimal::subtraction<decltype(result.value), true>(op1, op2, result);
}
}
else
{
throw logging::InvalidArgumentExcept("Unexpected result width");
}
break;
case OP_MUL:
if (fOperationType.colWidth == datatypes::MAXDECIMALWIDTH)
{
if (LIKELY(!fDecimalOverflowCheck))
{
datatypes::Decimal::multiplication<decltype(result.s128Value), false>(op1, op2, result);
}
else
{
datatypes::Decimal::multiplication<decltype(result.s128Value), true>(op1, op2, result);
}
}
else if (fOperationType.colWidth == utils::MAXLEGACYWIDTH)
{
if (LIKELY(!fDecimalOverflowCheck))
{
datatypes::Decimal::multiplication<decltype(result.value), false>(op1, op2, result);
}
else
{
datatypes::Decimal::multiplication<decltype(result.value), true>(op1, op2, result);
}
}
else
{
throw logging::InvalidArgumentExcept("Unexpected result width");
}
break;
case OP_DIV:
if (fOperationType.colWidth == datatypes::MAXDECIMALWIDTH)
{
if ((datatypes::Decimal::isWideDecimalTypeByPrecision(op2.precision) && op2.s128Value == 0) ||
(!datatypes::Decimal::isWideDecimalTypeByPrecision(op2.precision) && op2.value == 0))
{
isNull = true;
break;
}
if (LIKELY(!fDecimalOverflowCheck))
{
datatypes::Decimal::division<decltype(result.s128Value), false>(op1, op2, result);
}
else
{
datatypes::Decimal::division<decltype(result.s128Value), true>(op1, op2, result);
}
}
else if (fOperationType.colWidth == utils::MAXLEGACYWIDTH)
{
if (op2.value == 0)
{
isNull = true;
break;
}
if (LIKELY(!fDecimalOverflowCheck))
{
datatypes::Decimal::division<decltype(result.value), false>(op1, op2, result);
}
else
{
datatypes::Decimal::division<decltype(result.value), true>(op1, op2, result);
}
}
else
{
throw logging::InvalidArgumentExcept("Unexpected result width");
}
break;
default:
{
std::ostringstream oss;
oss << "invalid arithmetic operation: " << fOp;
throw logging::InvalidOperationExcept(oss.str());
}
}
}
std::ostream& operator<<(std::ostream& os, const ArithmeticOperator& rhs);
} // namespace execplan
#endif