Message du 23/03/11 17:47 De : "Oliver Kowalke" A : boost@lists.boost.org Copie à : Objet : Re: [boost] [context review] Review - alternative version available
Am 23.03.2011 17:35, schrieb Vicente BOTET:
Message du 23/03/11 14:51 De : "Oliver Kowalke" A : boost@lists.boost.org Copie à : Objet : Re: [boost] [context review] Review - alternative version available
Hello Vicente,
I would like to see a good rationale explaining the advantages of letting Boost.Context manage them. In any case the name asm_impl and native_impl are not adapted. I let you make some better suggestions than the current ones. What about dropping context< stack, impl> in favour of fast_context< stack> and native_context< stack> (nastive_context<> doesn't describe it well - do you have a better idea?). The user has the opertunity to choos which context he wants to use. Hi,
I'm still looking for the use case. What could make a user that has the possibility to get a fast context switch 13x faster to prefers to use the low context provided using ucontext?
You suggest that context<> should use fcontext if available and native OS facility otherwise?! If the other people agree I'll implement it. Maybe I should remove the stack template argument so that context only supports protected_stack (for security reasons -> overwriting memory of the app if stacksize is too small)?
I remember that protected_stack made use of mmap. I will prefer to been able to don't depend on mmap for this purpose and been able to use static allocation. I don't know if this is pre-optimization. Best, Vicente