[PATCH] um: clean up mm creation

[Anton Ivanov]

On 22/09/2023 21:12, Johannes Berg wrote:
> On Fri, 2023-09-22 at 21:08 +0100, Anton Ivanov wrote:
>> On 22/09/2023 20:41, Johannes Berg wrote:
>>> On Fri, 2023-09-22 at 14:41 +0100, Anton Ivanov wrote:
>>>> It is nearly twice slower than the current approach on a find /usr -type f -exec cat {} > /dev/null \;
>>>>
>>> Btw, I cannot reproduce that at all - seems about the same in my tests?
>>>
>>> Is there anything special in your setup?
>>>
>>> Or maybe it's because I'm using hostfs? But not sure why that would
>>> matter.
>> I am using a ubd with ext4 sitting on an nfs server. The host, however,
>> has more than enough memory to cache all of it, so it is pretty much
>> like reading off a ramdisk as it is fully cached.
>>
>> It is quite clear in that case.
>>
> Did you have your preempt patch applied also, btw? Because I was working
> off that now.

Yes.

>
> But even with PREEMPT turned off, I see basically no difference in such
> a benchmark. I was running only over /usr/bin/ because otherwise it's
> just too much time on my system overall, but there's basically no
> difference in the 4x4 matrix of
>
>   - preempt off / preempt voluntary
>   - with / without my patch

I can retest with hostfs and nfs over the weekend. For the /usr on the 
test system (Debian base + build deps for the kernel) it is at present:

9m without the patch, 15m with. On top of PREEMPT, PREEMPT is set to ON.

I have not tried without PREEMPT as that is somewhere around 30 mins :)

I can run smaller benches of course (f.e. /usr/bin).

Brgds,

>
> johannes
>

-- 
Anton R. Ivanov
Cambridgegreys Limited. Registered in England. Company Number 10273661
https://www.cambridgegreys.com/