Hi,

Skimmed a bit on Phoenix - looks interesting; but what's POGO (Linux?)

Anyway I still don't think the JIT compiler would always optimize out the reference to the treeview node count - it would have to be *sure* that the treeview nodes were not changed anywhere within the body of the for loop. Seems to me I'm just forcing one obvious optimisation without ruling out others?

Regards,
Viv


>There are a few permutations of what will happen based on the optimizers analysis of your code ie. is the array on the stack etc. It could decide to unroll and inline the loop completely, or it could decide to re-order instructions for better cache coherence?
>
>The point I should have added to my original comment was that MS are putting a huge amount of time and effort into optimizing JITters (go look at Pheonix and POGO) to the point where they can actually generate code based on how the user "uses" the application!
>
>You should make sure that you optimize for algorithmic efficiency and not assume that the IL represents the actual native code being generated!
>
>Regards
>Neil