MSDN Magazine - April 2008 - (Page 19)

Figure 2 Measuring .NET Events, Delegates, Interfaces, and Virtual Methods Operation Measured MethodCalls: aClass.Interface() [count=1000 scale=10.0] MethodCalls: aClass.VirtualMethod() [count=1000 scale=10.0] Delegates: aInstanceDelegate() [count=1000 scale=10.0] Events: Fire Events [count=1000 scale=10.0] Median 1.651 1.228 1.151 1.228 Mean 1.660 1.175 1.159 1.195 StdDev 0.084 0.077 0.085 0.070 Min 1.579 1.083 1.075 1.088 Max 1.814 1.277 1.314 1.291 Samples 10 10 10 10 loop executed 1000 times). Each measurement is then performed 10 times and standard statistics (min, max, median, mean, standard deviation) are computed. To make the time measurements more meaningful, they are normalized so that the median time for calling (and returning) from an empty static function is one unit. It is not uncommon for benchmarks to have widely varying times, which is why all the statistical information is important. Note the large difference between the minimum (71.299) and maximum (953.864) times for the FinalizableClass benchmark. This variation needs to be explained before you can trust the data from that benchmark. In that particular case, it is the result of the runtime periodically executing slower code paths to allocate bookkeeping data structures in bulk. Already, having these statistics available is proving useful in validating the data. This table is a gold mine of useful performance data, detailing the costs of most of the primitive operations used by .NET- targeted code. I will go into detail in my next installment of this column, but here I want to explain an important feature of MeasureIt: it comes with its own source code. To unpack MeasureIt’s source code and launch Visual Studio® to browse it (if Visual Studio is available), type this: MeasureIt /edit Having the source means that you can quickly understand exactly what the benchmark is measuring. It also means that you can easily add a new benchmark to the suite. Again, my experience with the profile data processor is instructive. At one point on the design, I could do a certain common operation with either C# events, delegates, virtual methods, or interfaces. To make a decision, I needed to understand the performance tradeoff among these choices. Within minutes I had written the microbenchmark to measure the performance of each of the alternatives. Figure 2 displays the relevant rows and you can see that there is no substantial difference between the alternatives. This knowledge al- NCache TM Clustered Object Cache for .NET Boost your .NET application performance! If your .NET application runs in a server-farm and has large number of users, then use NCache to boost its performance and scalability. NCache is an extremely fast distributed object cache for .NET. Enterprise Distributed Object Cache Cache data, reduce database trips, and boost performance Scale up thru Replicated, Partitioned, and Client Cache topologies Distributed ASP.NET Session Cache Faster & more scalable session storage Integrate without any code change Distributed ASP.NET Output Cache Cache rendered pages and serve content without re-rendering Integrate without any code change Ph: +1-925-236-3830 Download FREE Trial CLR Inside Out april2008 19 www.alachisoft.com http://www.alachisoft.com http://www.alachisoft.com

Table of Contents Feed for the Digital Edition of MSDN Magazine - April 2008

MSDN Magazine - April 2008 Contents Toolbox CLR Inside Out Basic Instincts Cutting Edge Foundations Test Run Service Station Windows with C++ Going Places { End Bracket }

MSDN Magazine - April 2008

For optimal viewing of this digital publication, please enable JavaScript and then refresh the page. If you would like to try to load the digital publication without using Flash Player detection, please click here.