Performance Quiz #4 -- Of Sets and Sundry -- Solution

re: Performance Quiz #4 -- Of Sets and Sundry -- Solution

Ryan Heath — Fri, 20 Aug 2004 07:02:00 GMT

Hi Rico,

Why the [System.Runtime.CompilerServices.MethodImplAttribute(System.Runtime.CompilerServices.MethodImplOptions.NoInlining)] before each test method?

// Ryan

re: Performance Quiz #4 -- Of Sets and Sundry -- Solution

Rico Mariani — Fri, 20 Aug 2004 18:27:00 GMT

Force of habit on the attribute. In other cases my test functions can be very small and they then get inlined directly into Main (i.e. the harness) which makes it harder to see clear results in the profile.

In this case there was basically no chance of inlining so it's moot. I thought of removing them but I thought it's good practice anyway in perf benchmarks so I left it.

re: Performance Quiz #4 -- Of Sets and Sundry -- Solution

Matthew W. Jackson — Fri, 20 Aug 2004 19:35:00 GMT

I'd leave them just in case future versions of the CLR decide to become more aggressive about inlining.

There's no clear spec in the CLR on which methods will be inlined (I suppose it's very machine- and implementation-dependant), although I think you posted some guidlines earlier regarding the MS implementation.

I'd still like a "MethodImplOptions.FavorInlining"
even if an implementation was free to ignore it completely. I wouldn't mind the MS implementation ignoring it, but I'd like to be able to at least think I have some control over the decision. <g>

re: Performance Quiz #4 -- Of Sets and Sundry -- Solution

Akshay Kumar — Wed, 01 Sep 2004 15:28:00 GMT

Do we get to download code for this solution somewhere ?
Regards,
Akshay

应用System.Runtime.CompilerServices.MethodImplOptions.NoInlining特性

Avatasia — Wed, 06 Aug 2008 08:40:34 GMT

在参考某些.Net代码的时候，我们有时会看到某些方法，甚至属性的get或set代码片断上会附加[System.Runtime.CompilerServices.MethodImpl(Syst...