Subject: Re: [boost] Fw: [atomic] review results
From: Peter Dimov (pdimov_at_[hidden])
Date: 2011-11-08 18:14:19
Andrey Semashev wrote:
> If in your case modules are separately built and there is a possibility of
> code inconsistensies then surely inlining is a bad idea. But that doesn't
> that inlining is generally a bad thing when multiple modules are involved.
We're not discussing whether it's generally a bad thing. The specific use
case under discussion involves a function that does atomic operations.
Inlining is typically done for performance reasons, and the atomic
operations typically dominate, rendering the performance gain from the
inlining irrelevant. So you're only left with the drawbacks.
> Anyway, it's going slightly off-topic.