pingcap/tidb

design vectorized aggregation functions

Open

#14,217 建立於 2019年12月25日

在 GitHub 查看
 (0 留言) (0 反應) (0 負責人)Go (40,090 star) (6,186 fork)batch import
help wantedtype/enhancement

描述

Description

All our aggregation functions (in executor/aggfuncs/aggfuncs.go) used by the execution engine are row-based. We can vectorize them to improve their performance like what we do for all scalar functions.

Difficulty

  • Medium

Score

  • 3000

Mentor(s)

  • @SunRunAway
  • @XuHuaiyu
  • @qw4990

Recommended Skills

  • Go programming

貢獻者指南

技術棧
go
領域
database
議題類型
performance
難度面向新貢獻者的預計實作難度，1 表示很小改動，5 表示專家級工作。
3
預計時間有經驗貢獻者完成調查、實作、測試並準備 pull request 的粗略時間範圍。
3-5 days
活動狀態議題目前的可參與程度：新鮮、活躍、陳舊、阻塞或等待維護者輸入。
stale
清晰度議題是否清楚說明預期改動、驗收標準和下一步。
clear
前置要求
Go programmingfamiliarity with TiDB executor
新手友善度1-100 的估計分數，表示該議題對首次貢獻者的友善程度。
25
研究方向
Begin by studying the existing row based aggregation functions in executor/aggfuncs/aggfuncs.go to understand the current interface and logic. Look at already vectorized scalar functions (e.g., in expressions/scalar function) for patterns. Then design a vectorized version, possibly starting with a single aggregation function like sum or count. Consider performance implications and write benchmarks. No existing PRs or comments on this issue, so the design is open.
design vectorized aggregation functions · pingcap/tidb#14217 | Good First Issue