C＃GC.Collect（）和内存

Question

I am receiving a very large list as a method argument and would like to remove it from memory after it is used. Normally I would let the GC do its thing, but I need to be very careful of memory usage in this app.

Will this code accomplish my goal? I've read a lot of differing opinions and am confused.

public void Save(IList<Employee> employees)
    {
        // I've mapped the passed-in list
        var data = Mapper<Employee, EmployeeDTO>.MapList(employees);

        // ?????????????
        employees = null;
        GC.Collect();

        // Continues to process very long running methods....
        // I don't want this large list to stay in memory

   }

Maybe I should use another technique that I'm not aware of?

Answer 1

If the list is not used anymore the GC will automatically collect it when available memory is an issue .

However, if the caller uses the list after passing it to your function then the GC won't collect it even if you set it to null (all you have is a reference to the list - you can't do anything about other objects that hold references as wel).

Unless you have a measurable problem don't try and outsmart the GC.

Answer 2

This is not a direct answer to the question but some ideas how to handle the low-memory situations the poster (@Big Daddy) referred to.

If you're running into low memory situations on an x64 platform with 8 GB of memory, you should determine if your application is responsible for it. If it is, then run a memory profiler (CLR Profiler or something else or even get a full user dump and run WinDbg on it) to see what allocates the memory. It's possible that you have some objects that are not used anymore but are still referenced somewhere - this is not a true memory leak but it's memory that's not freed up in your application - most decent profilers will identify big objects (or objects with a lot of instances) along with their types.

I find it hard to believe that the list passed to this Save function would stress a server with 8 GB of memory but we don't know how much free memory is available to the process and what kind of a process it is (IIS, desktop, etc.)

If Save is called on several threads with huge inputs, it can potentially lead to memory stress but even then, it's not very likely and I'd check various counters and profile data to see when and why memory stress happens.