Which piece of code is more performant?

Question

I have some code i'm revewing, which is used to convert some text into an MD5 Hash. Works great. It's used to create an MD5Hhash for a gravatar avatar. Here it is :-

static MD5CryptoServiceProvider md5CryptoServiceProvider = null;

public static string ToMD5Hash(this string value)
{
    //creating only when needed
    if (md5CryptoServiceProvider == null)
    {
        md5CryptoServiceProvider = new MD5CryptoServiceProvider();
    }

    byte[] newdata = Encoding.Default.GetBytes(value);
    byte[] encrypted = md5CryptoServiceProvider.ComputeHash(newdata);
    return BitConverter.ToString(encrypted).Replace("-", "").ToLower();
}

Notice how we create a MD5CryptoServiceProvider the first time this method is called? (lets not worry about race-conditions here, for simplicity).

I was wondering, is it more computationally expensive if i change the lines that are used to create the provider, to this...

using(var md5CryptoServiceProvider = new MD5CryptoServiceProvider())
{
    ... snip snip snip ....
}

Now, how is this method used/consumed? Well, imagine it's the home page of StackOverflow -> for each post, generate an md5 hash of the user, so we can generate their gravatar url. So the view could call this method a few dozen times.

Without trying to waste too much time stressing over premature optimzation, etc ... which would be better?

Answer 1

I'd be more interested in the thread-safefy... MSDN doesn't (unless I missed it) say that MD5CryptoServiceProvider is thread-safe, so IMO the best option is to have one per call...

It doesn't matter how quickly you can get the wrong answer ;-p

What you probably don't want to do (to fix the thread-safety issue) is have a static instance and lock around it... that'll serialize all your crypto code, when it could run in parallel on different requests.