Path: csiph.com!fu-berlin.de!uni-berlin.de!individual.net!not-for-mail
From: Ian Collins <ian-news@hotmail.com>
Newsgroups: comp.lang.c
Subject: Re: Default values in #define
Date: Sun, 28 Aug 2016 08:41:46 +1200
Lines: 31
Message-ID: <e2ec8aFa03jU10@mid.individual.net>
References: <d6e97054-1a68-418a-a6ae-5807ffccabbe@googlegroups.com> <502669cd-f0fc-4b99-8c8a-0d806c8b5f20@googlegroups.com> <aa1be119-a537-4df5-97da-ade93afdba0c@googlegroups.com> <11e2bb75-ee3c-4904-b7c8-77023426fb45@googlegroups.com> <c61b98eb-0d30-40aa-986e-0914e2680d96@googlegroups.com> <ffdc9c87-29a2-4769-9399-667ee00491a1@googlegroups.com> <87vaynj0l9.fsf@bsb.me.uk> <4c14769d-f75d-4c8f-89f6-5290f6b97b32@googlegroups.com> <87poovitqy.fsf@bsb.me.uk> <62672946-dc1f-4d94-9c33-5434f51da85d@googlegroups.com> <eE2wz.36092$1b1.31741@fx43.iad> <71637eed-ce54-4950-9568-d053ff0abc84@googlegroups.com> <e2bttuFa03jU6@mid.individual.net> <e1cb23ea-8fe0-4f1d-9182-2661cd3dac1a@googlegroups.com> <e2c9okFa03jU7@mid.individual.net> <91a99d84-b481-4dae-949b-914cb5e8aab7@googlegroups.com> <npqvct$suc$1@jstuckle.eternal-september.org> <60c8cdba-0914-430a-999d-5543e627025e@googlegroups.com> <nps3ib$sbt$1@jstuckle.eternal-september.org> <2d96b3c5-8c6d-4aa5-a6f1-d6381a533e24@googlegroups.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: 7bit
X-Trace: individual.net ZHz++z+hiJRZX9odO0XfWAizIcUMhWgGnazqV5wm8lI3cceSKF
Cancel-Lock: sha1:n7G8L4D/YT1fWBZtmtX7equxPPY=
User-Agent: Mozilla/5.0 (X11; SunOS i86pc; rv:38.0) Gecko/20100101 Thunderbird/38.5.1
In-Reply-To: <2d96b3c5-8c6d-4aa5-a6f1-d6381a533e24@googlegroups.com>
Xref: csiph.com comp.lang.c:92820

On 08/28/16 08:10 AM, supercat@casperkitty.com wrote:
>
> If there are two threads with similar code, that design is clearly broken.
> Some similar designs, however, can be perfectly workable if a system has
> memory barriers around compiler accesses, or if everything that is used
> in the interrupt and mainline is "volatile" [note that memory barriers are
> often less costly than requiring that everything be volatile].
>
> For example:
>
>      volatile interrupt_item_count;
>
>      MAINLINE:
>      while(interrupt_item_count)
>        spin();
>      ... store data into an array
>      ... set pointer to start of array
>      interrupt_item_count = number_of_items;
>
>      INTERRUPT:
>      if (interrupt_item_count)
>      {
>        OUTPUT_DEVICE = *interrupt_data_ptr++;
>        interrupt_item_count--;
>      }

Why would you resort to such a scheme when there are standard (or 
obvious target specific) tools for the job?

-- 
Ian