MSDN Magazine - October 2007 - (Page 27)

in msdnmag.py. However, in an expression such as“if x < 1:” you can generate a DynamicSite . The site’s Invoke method will be statically typed to take one argument as an integer and return a Boolean. The delegate inside the dynamic site is also statically typed in this way, which makes the resulting code faster. The delegate only has to test one argument in its caching, and as the stack unwinds, the JIT knows a Boolean value is ﬂowing to the right location. This eliminates any casts or checks surrounding the if test. Some dynamic languages have optional explicit typing. Even if they don’t,implementations can infer types from calls to .NET static methods,from literal constants,and so on.They can propagate this type information over lifetimes of variables,branches of conditionals, and the like.When languages have type information, they can generate DLRAST that in turn compiles to DynamicSite types with full type information.That can result in inline code generation for the Invoke method call and target code call in the cache. Today,the DLR generates generic DynamicSite types with six or fewer generic type parameters.The ﬁrst N-1 parameters represent the types of arguments, and the Nth represents the result type of the operation. After six types, you have to generate AST that just performs method calls to runtime helper functions. Those functions have to do the slow work to compute everything at run time. In the future, the DLR will allow you to use generic tuples to provide large numbers of statically typed arguments. We might also increase the number from six as we ﬁgure out the sweet spot for most programs. You should have a pretty good idea now how explore both language-speciﬁc ASTs and DLR ASTs.With some basic information on simple trees, you can examine the way certain languages have been built on the DLR. If you are trying to implement a language on the DLR yourself, you can use this technique. If you’re not, it might be fun to just look at what is happening under the covers. You also got to see some of the shared infrastructure the DLR offers to language implementers.There is a lot to leverage from the CLR and the DLR—garbage collection,JIT compilation,dynamic sites, common dynamic type model, and so forth. Implementing languages on the .NET runtime produces pretty good results now—and it will just keep getting better.Version 1 of the DLR will be great. Soon implementing dynamic languages on .NET will be amazingly easy and the results will be excellent. Send your questions and comments to clrinout@microsoft.com. Bill Chiles is a Senior Program Manager at Microsoft who has worked on dynamic language projects (CMU Common Lisp, Dylan) and tools for most of his career. Bill spent nine years on the Visual Studio Core Technologies team and now works on IronPython and the Dynamic Language Runtime. CLR Inside Out october2007 27 http://www.tedpattison.net http://www.tedpattison.net

Table of Contents Feed for the Digital Edition of MSDN Magazine - October 2007

Cover Contents Toolbox CLR Inside Out Basic Instincts Data Points Cutting Edge Pooled Threads WPF Threads Parallel Linq Parallel Performance Mobile Apps Test Run Foundations Windows with C++ Netting C++ .NET Matters { End Bracket } Net Nuptials

MSDN Magazine - October 2007

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.