Does the .NET CLR JIT compile every method, every time?

I know that Java's HotSpot JIT will sometimes skip JIT compiling a method if it expects the overhead of compilation to be lower than the overhead of running the method in interpreted mode. Does the .NET CLR work based upon a similar heuristic?

Note: this answer is on a "per-run" context. The code is normally JITted each time you run the program. Using ngen or .NET Native changes that story, too...

Unlike HotSpot, the CLR JIT always compiles exactly once per run. It never interprets, and it never recompiles with heavier optimisation than before based on actual usage.

This may change, of course, but it's been that way since v1 and I don't expect it to change any time soon.

The advantage is that it makes the JIT a lot simpler - there's no need to consider "old" code which is already running, undo optimisations based on premises which are no longer valid etc.

One point in .NET's favour is that most CLR languages make methods non-virtual by default, which means a lot more inlining can be done. HotSpot can inline a method until it's first overridden at which point it undoes the optimisation (or does some clever stuff in some cases to conditionally still use the inlined code, based on actual type). With fewer virtual methods to worry about, .NET can largely ignore the pain of not being able to inline anything virtual.

EDIT: The above describes the desktop framework. The Compact Framework throws out native code when it wants to, JITting again as necessary. However, this still isn't like HotSpots adaptive optimisation.

The micro framework doesn't JIT at all apparently, interpreting the code instead. This makes sense for very constrained devices. (I can't say I know much about the micro framework.)

The .NET runtime always compiles code JIT before execution. So, it is never interpreted.

You can find some more interesting reading in CLR Design Choices with Anders Hejlsberg. Especially the part:

I read that Microsoft decided that IL will always be compiled, never interpreted. How does encoding type information in instructions help interpreters run more efficiently?

Anders Hejlsberg: If an interpreter can just blindly do what the instructions say without needing to track what's at the top of the stack, it can go faster. When it sees an iadd, for example, the interpreter doesn't first have to figure out which kind of add it is, it knows it's an integer add. Assuming someone has already verified that the stack looks correct, it's safe to cut some time there, and you care about that for an interpreter. In our case, though, we never intended to target an interpreted scenario with the CLR. We intended to always JIT [Just-in-time compile], and for the purposes of the JIT, we needed to track the type information anyway. Since we already have the type information, it doesn't actually buy us anything to put it in the instructions.

Bill Venners: Many modern JVMs [Java virtual machines] do adaptive optimization, where they start by interpreting bytecodes. They profile the app as it runs to find the 10% to 20% of the code that is executed 80% to 90% of the time, then they compile that to native. They don't necessarily just-in-time compile those bytecodes, though. A method's bytecodes can still be executed by the interpreter as they are being compiled to native and optimized in the background. When native code is ready, it can replace the bytecodes. By not targeting an interpreted scenario, have you completely ruled out that approach to execution in a CLR?

Anders Hejlsberg: No, we haven't completely ruled that out. We can still interpret. We're just not optimized for interpreting. We're not optimized for writing that highest performance interpreter that will only ever interpret. I don't think anyone does that any more. For a set top box 10 years ago, that might have been interesting. But it's no longer interesting. JIT technologies have gotten to the point where you can have multiple possible JIT strategies. You can even imagine using a fast JIT that just rips quickly, and then when we discover that we're executing a particular method all the time, using another JIT that spends a little more time and does a better job of optimizing. There's so much more you can do JIT-wise.

It will be nice to see some trace-based JITs in the future for devices with low memory. It would mainly interpret, find hot spots, and convert those into assembler and cache those. I think this is what Google does with their Android JIT and Microsoft Research has a research project ongoing for trace-based JIT.

I found an article, SPUR: A Trace-Based JIT Compiler for CIL.. Maybe some of this will make it into the CLR one day?