Re: [boost] [graph][heap][coroutine] interruptable dijkstra shortest path function

Subject: Re: [boost] [graph][heap][coroutine] interruptable dijkstra shortest path function
From: Oliver Kowalke (oliver.kowalke_at_[hidden])
Date: 2012-10-05 13:09:31

Am 05.10.2012 17:47, schrieb Giovanni Piero Deretta:
>
>
> But the biggest reason for the performance difference is that the stackless
> coroutine code can be completely inlined inside the caller as the compiler
> can "see thru" the yield and resume. That's the reason while, even without
> yielding, the stackless code is faster.  Having inlineable stack-full
> coroutines would require heroic compiler help.
>

heroic? :^)

I'm thinking about to provide inline asm in boost.context for compilers 
which support it - a t first it will be gcc.

Oliver