hcq/oceanbase
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

[bugfix] get wrong db_time when subplan filter is used in the plan

Browse Source
This commit is contained in:
raywill
2023-01-10 08:38:04 +00:00
committed by ob-robot
parent b52f63ea88
commit 460fdf9fc4
3 changed files with 30 additions and 16 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
src/sql/engine/ob_operator.h
View File

@ -601,21 +601,28 @@ protected:
inline void begin_cpu_time_counting()
{
// begin with current operator
ObActiveSessionGuard::get_stat().plan_line_id_ = spec_.id_;
cpu_begin_time_ = rdtsc();
}
inline void end_cpu_time_counting()
{
total_time_ += (rdtsc() - cpu_begin_time_);
}
inline void begin_ash_line_id_reg()
{
// begin with current operator
ObActiveSessionGuard::get_stat().plan_line_id_ = spec_.id_;
}
inline void end_ash_line_id_reg()
{
// move back to parent operator
// known issue: when switch from batch to row in same op,
// we shift line id to parent op un-intently. but we tolerate this inaccuracy
if (OB_LIKELY(spec_.get_parent())) {
common::ObActiveSessionGuard::get_stat().plan_line_id_ = spec_.get_parent()->id_;
} else {
common::ObActiveSessionGuard::get_stat().plan_line_id_ = -1;
}
}
uint64_t cpu_begin_time_; // start of counting cpu time
uint64_t total_time_; // total time cost on this op, including io & cpu time
protected: