mysl 子查询带来的性能问题

一个带有多个子查询的sql，性能非常慢;通过减少子查询改写后性能大幅度提升。

• 原SQL&执行计划:

SELECT
  t1.version_no,
  t3.instance_id
FROM (SELECT
    v.version_no,
    v.version_id,
    v.message,
    v.state,
    v.start_time,
    v.end_time,
    t.task_id,
    t.data_freq_unit,
    t.team_id
  FROM tb_data_version v, tb_task t
  WHERE v.task_id = t.task_id AND v.task_id = 5757
) t1  LEFT JOIN (
    SELECT version_id, MAX(instance_id) AS mx
    FROM tb_instance
    WHERE task_id = 5757
    GROUP BY version_id
) t2  ON t2.version_id = t1.version_id
LEFT JOIN tb_instance t3 ON t3.instance_id = t2.mx

• 优化后的SQL&执行计划:

SELECT
	v.version_no,
  i.instance_id
FROM tb_data_version v
  JOIN tb_task t ON (v.task_id = t.task_id)
  LEFT JOIN tb_instance i ON (
  	v.version_id = i.version_id 
  	AND v.task_id = i.task_id 
  	and i.instance_id = (
  	    SELECT MAX(instance_id) FROM tb_instance WHERE version_id = v.version_id AND task_id = t.task_id
  	)
  )
WHERE v.task_id = 5757

• 疑惑：怀疑旧SQL中的第二个子查询，对于第一个子查询结果中的每行都会执行一次，但是还是不太确定，如果有类似经验的还请帮忙确认一下！