Event Handler performance

Question

I have a performance problem. I create 100 new buttons and I want to assign an Click Event Handler. I execute this code for about 100 times:

Buttons[i].Button.Click += new System.EventHandler(Button_Click);

It takes about 2sec to complete. I have a lot of other event assignments in the same function, but they all take only some millisecond to execute. So I have transformed my code in

Buttons[i].Button.MouseUp += new System.Windows.Forms.MouseEventHandler(Button_Click);

Now the code is fast (some millisecond, like the others). Obviously I have modified the parameters of the function "Button_click" to fit the new event requirements, but no other changes were made.

I am wondering why this could happen. Is EventHandler that slow? Or am I doing something wrong? Or is there a best practice?

I am using VC2010 with C#, using .NET 4 in a Windows Form application.

EDIT:

Now I have "minified" my code and I put it there:

            Stopwatch stopWatch = new Stopwatch();
            stopWatch.Start();
            Button b;
            for(n=0;n<100;n++)
            {
                b = new Button();
                b.Location = new System.Drawing.Point(100, 0);
                b.Name = "btnGrid";
                b.Size = new System.Drawing.Size(50, 50);
                b.Text = b.Name;
                b.UseVisualStyleBackColor = true;
                b.Visible = false;
                b.Text = "..";
                b.Click += new EventHandler(this.Button_Click);
                //b.MouseUp += new System.Windows.Forms.MouseEventHandler(this.Button_ClickUP);
            }
            stopWatch.Stop();

            TimeSpan ts = stopWatch.Elapsed;
            string elapsedTime = String.Format("{0:00}:{1:00}:{2:00}.{3:00}", ts.Hours, ts.Minutes, ts.Seconds, ts.Milliseconds / 10);
            Log(elapsedTime, Color.Purple);

Button_Click and Button_Click are:

    private void Button_Click(object sender, EventArgs e)
    {            
    }

    private void Button_ClickUP(object sender, MouseEventArgs e)
    {
    }

I put this code in a button and the "Log" function display the result in a memo. When I enable "Click" the result is 01.05 sec, but when I enable "MouseUp" the result is 00.00.

Difference -> ONE SECOND!

why!?

== EDIT ==

I use .NET Framework 4. VS2010. Win XP. I found this: if I use .NET 3.5 or lower the speed changes: 0.5 sec. An Half. If I compile in debug or release mode it doesn't change.

If I use the executable without the debugger is blazing fast.

So I change my question: is .NET 4 slower then .NET 3? Why the Release mode works differently compared to the stand alone version?

Many thanks.

Answer 1

The code ".Click += ..." is transformed into ".add_Click( ... )". The "add_Click" method can have some logic checks.

You can little-bit speed up with no recreation of delegate:

EventHandler clickHandler = this.Button_Click;
foreach(Button btn in GetButtons()) {
   btn.Click += clicHandler;
}

EDIT:

Are you sure, the bottleneck is the attaching the handlers? I tried the for loop (100 loops) with attaching the eventhandler to Click event and I get this results:

/* only creation the button and attaching the handler */
button1_Click - A: 0 ms
button1_Click - B: 0 ms
button1_Click - A: 1 ms
button1_Click - B: 0 ms
button1_Click - A: 0 ms
button1_Click - B: 0 ms

/* creation the button, attaching the handler and add to the panel */
button2_Click - A: 223 ms
button2_Click - B: 202 ms
button2_Click - A: 208 ms
button2_Click - B: 201 ms
button2_Click - A: 204 ms
button2_Click - B: 230 ms

The source code:

    void button_Click(object sender, EventArgs e) {
        // do nothing
    }

    private void button1_Click(object sender, EventArgs e) {
        const int MAX_BUTTONS = 100;
        var stopWatch = new System.Diagnostics.Stopwatch();
        stopWatch.Start();
        for (int i = 0; i < MAX_BUTTONS; i++) {
            var button = new Button();
            button.Click += new EventHandler(button_Click);
        }
        stopWatch.Stop();
        System.Diagnostics.Debug.WriteLine(string.Format("button1_Click - A: {0} ms", stopWatch.ElapsedMilliseconds));

        stopWatch.Reset();
        stopWatch.Start();
        EventHandler clickHandler = this.button_Click;
        for (int i = 0; i < MAX_BUTTONS; i++) {
            var button = new Button();
            button.Click += clickHandler;
        }
        stopWatch.Stop();
        System.Diagnostics.Debug.WriteLine(string.Format("button1_Click - B: {0} ms", stopWatch.ElapsedMilliseconds));
    }

    private void button2_Click(object sender, EventArgs e) {
        const int MAX_BUTTONS = 100;

        var stopWatch = new System.Diagnostics.Stopwatch();

        this.panel1.Controls.Clear();
        stopWatch.Start();
        for (int i = 0; i < MAX_BUTTONS; i++) {
            var button = new Button();
            button.Click += new EventHandler(button_Click);
            this.panel1.Controls.Add(button);
        }
        stopWatch.Stop();
        System.Diagnostics.Debug.WriteLine(string.Format("button2_Click - A: {0} ms", stopWatch.ElapsedMilliseconds));

        stopWatch.Reset();

        this.panel1.Controls.Clear();
        stopWatch.Start();
        EventHandler clickHandler = this.button_Click;
        for (int i = 0; i < MAX_BUTTONS; i++) {
            var button = new Button();
            button.Click += clickHandler;
            this.panel1.Controls.Add(button);
        }
        stopWatch.Stop();
        System.Diagnostics.Debug.WriteLine(string.Format("button2_Click - B: {0} ms", stopWatch.ElapsedMilliseconds));
    }

EDIT 2: I tried compare time spent with attaching Click handler vs. attaching MouseUp handler. It does not seems, the attaching MouseUp event is faster than Click event.

I think the problem will be somewhere else. Don't GC collect during your loop? Or don't you do something else there?

Results:

button1_Click - Click_A: 6 ms
button1_Click - Click_B: 6 ms
button1_Click - MouseUp_A: 15 ms
button1_Click - MousUp_B: 7 ms

button1_Click - Click_A: 16 ms
button1_Click - Click_B: 7 ms
button1_Click - MouseUp_A: 16 ms
button1_Click - MousUp_B: 10 ms

button1_Click - Click_A: 14 ms
button1_Click - Click_B: 19 ms
button1_Click - MouseUp_A: 27 ms
button1_Click - MousUp_B: 5 ms

button1_Click - Click_A: 17 ms
button1_Click - Click_B: 17 ms
button1_Click - MouseUp_A: 24 ms
button1_Click - MousUp_B: 8 ms

button1_Click - Click_A: 6 ms
button1_Click - Click_B: 5 ms
button1_Click - MouseUp_A: 14 ms
button1_Click - MousUp_B: 7 ms

button1_Click - Click_A: 14 ms
button1_Click - Click_B: 9 ms
button1_Click - MouseUp_A: 15 ms
button1_Click - MousUp_B: 7 ms

Code:

    private void button1_Click(object sender, EventArgs e) {
        const int MAX_BUTTONS = 1000;
        var stopWatch = new System.Diagnostics.Stopwatch();

        stopWatch.Start();
        for (int i = 0; i < MAX_BUTTONS; i++) {
            var button = new Button();
            button.Click += new EventHandler(button_Click);
        }
        stopWatch.Stop();
        System.Diagnostics.Debug.WriteLine(string.Format("button1_Click - Click_A: {0} ms", stopWatch.ElapsedMilliseconds));

        stopWatch.Reset();
        stopWatch.Start();
        EventHandler clickHandler = this.button_Click;
        for (int i = 0; i < MAX_BUTTONS; i++) {
            var button = new Button();
            button.Click += clickHandler;
        }
        stopWatch.Stop();
        System.Diagnostics.Debug.WriteLine(string.Format("button1_Click - Click_B: {0} ms", stopWatch.ElapsedMilliseconds));

        stopWatch.Start();
        for (int i = 0; i < MAX_BUTTONS; i++) {
            var button = new Button();
            button.MouseUp += new MouseEventHandler(button_MouseUp);
        }
        stopWatch.Stop();
        System.Diagnostics.Debug.WriteLine(string.Format("button1_Click - MouseUp_A: {0} ms", stopWatch.ElapsedMilliseconds));

        stopWatch.Reset();
        stopWatch.Start();
        MouseEventHandler mouseUpHandler = this.button_MouseUp;
        for (int i = 0; i < MAX_BUTTONS; i++) {
            var button = new Button();
            button.MouseUp += mouseUpHandler;
        }
        stopWatch.Stop();
        System.Diagnostics.Debug.WriteLine(string.Format("button1_Click - MousUp_B: {0} ms", stopWatch.ElapsedMilliseconds));
    }

EDIT : The body of add_Click method (= Click += ...) is rough:

public void add_Click(EventHandler value) {
   this.Events.AddHandler(ClickEventIdentifier, value);
}

The MouseUp events will looks similar. At least both events using Events property for holding lists of delegates for events.

But if I tried several things I can not get the problems with the events as you wrote :(. Can you reproduce same behaviour on another computers?

Answer 2

System.EventHandler is a delegate type and therefore doesn't do anything. This can't be the source of the difference in performance.
Adding a new Click handler and a new MouseUp event handler is the same internally. Both call Events.AddHandler.
In my eyes, the only difference can be, that either Click already has other event handlers attached and MouseUp hasn't or the other way around.
To check whether my assumption is correct, you could copy and paste the two code snippets and execute each twice and measure the duration for the first and second time.

If both runs for Click are slow and the second run for MouseUp is slow, the problem is, that there already are existing Click handlers and adding a handler when already one exists is slower than adding one when none exists.

If the first run for Click is slow and the second one is fast and both runs for MouseUp are fast, the problem is, that there are no existing Click handlers and adding a handler when already one exists is faster than adding one when none exists.

My answer assumes that the observations of the OP are side effect free. I didn't actually test whether his results are reproducible or plausible. My answer just wants to show that there really is nothing special to the Click event or System.EventHandler.