Performance Tuning :: Compare Current Row Values With Previous One Until Mismatch

We have employee salary table which will have salary of an employee on daily basis (only working days). Below is the table structure:

emp_salary
----------
emp_id NUMBER(15) NOT NULL
effective_date DATE NOT NULL
salary NUMBER(15) NOT NULL
Primary key - emp_id, effective_date
This table is yearly partitioned

I have to find out how long the salary is not changed for an employee from given date, and last salary. I am using below query to do this:

WITH salary_tab AS
(SELECT effective_date, salary,
(CASE
WHEN (LAG (salary) OVER (PARTITION BY emp_id ORDER BY effective_date ASC) =
salary
)
THEN 0
ELSE 1
END
) changed_ind
FROM emp_salary
WHERE emp_id = 12345
[code]....

The cost of this query is 1677 and it is taking around 60 msec to complete. When I run this query for around 2000 employees in a loop it is taking around 3 minutes to complete.

The main bottleneck of this query is in the with clause where I am processing the entire history instead of stopping after first change.