Received-SPF: pass (google.com: domain of linux-kernel+bounces-130849-linux.lists.archive=gmail.com@vger.kernel.org designates 139.178.88.99 as permitted sender) client-ip=139.178.88.99;
Message-ID: <719745d9-cbb6-4691-a037-2ea7fff8494a@metaparadigm.com>
Date: Thu, 4 Apr 2024 15:59:36 +1300
Precedence: bulk
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: user-space concurrent pipe buffer scheduler interactions
To: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jens Axboe <axboe@kernel.dk>, Ingo Molnar <mingo@redhat.com>,
        Peter Zijlstra <peterz@infradead.org>, linux-kernel@vger.kernel.org
References: <969ccc0f-d909-4b45-908e-e98279777733@metaparadigm.com>
 <CAHk-=wj3khqjXnHmShPj+tZvC=SyUitTZ7BkDO-rqu_aCVPrXA@mail.gmail.com>
 <1a116a43-fd2e-4f03-8a17-75816fc62717@metaparadigm.com>
 <CAHk-=wiOqTAX7nwQqHJ2Ti3dbxM-HrUJq3FD_Gjua5qW=QAoTg@mail.gmail.com>
Content-Language: en-US
From: Michael Clark <michael@metaparadigm.com>
Organization: Metaparadigm
In-Reply-To: <CAHk-=wiOqTAX7nwQqHJ2Ti3dbxM-HrUJq3FD_Gjua5qW=QAoTg@mail.gmail.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit

On 4/4/24 09:57, Linus Torvalds wrote:
> On Wed, 3 Apr 2024 at 13:52, Michael Clark <michael@metaparadigm.com> wrote:
>>
>> On 4/4/24 05:56, Linus Torvalds wrote:
>>> On Tue, 2 Apr 2024 at 13:54, Michael Clark <michael@metaparadigm.com> wrote:
>>>>
>>>> I am working on a low latency cross-platform concurrent pipe buffer
>>>> using C11 threads and atomics.
>>>
>>> You will never get good performance doing spinlocks in user space
>>> unless you actually tell the scheduler about the spinlocks, and have
>>> some way to actually sleep on contention.
>>>
>>> Which I don't see you as having.
>>
>> We can work on this.
> 
> It's been tried.
> 
> Nobody ever found a use-case that is sufficiently convincing, but see
> the write-up at
> 
>     https://lwn.net/Articles/944895/

It's certainly interesting!

I wouldn't throw in the towel so soon until folks tried a bit harder. 
Even if we can't make it faster it would be neat to have an Austin Group 
meeting to give us user-space devels a waitlist_t so we don't hit that 
brain damaged deadlocky cond_wait foot-gun and realize its totally 
broken and that we must use cond_timedwait with some stupid delay until 
we can recheck the condition again in user-space in our "portable code".

Michael.