Re: [PATCH 2/2] block/mq-deadline: fallback to per-cpu insertion buckets under contention

[Jens Axboe <axboe@kernel.dk>]

On 1/18/24 1:46 PM, Bart Van Assche wrote:
> On 1/18/24 10:56, Jens Axboe wrote:
>> Do you need me to link the cover letter that you were CC'ed on?
> 
> There is no reason to use such an aggressive tone in your emails.

I'm getting frustrated with you because I need to say the same things
multiple times, and it doesn't seem like it's getting received at the
other end. And you had clearly seen the results, which means that rather
than the passive aggressive question, you could have said

"It'd probably be useful to include some performance numbers in
 the commmit messages themselves as well."

which would be received way differently than asking a question that you
already have the answer to.

> In the cover letter I see performance numbers for the patch series in
> its entirety but not for the individual patches. I'd like to know by
> how much this patch by itself improves performance because whether or
> not I will review this patch will depend on that data.

You can't really split them up, as you need both to see the real
benefit. The only reason they are split is because it makes sense to do
so in terms of the code, as they are two different paths and points of
contention. This obviously makes patch 2 look better than it is, and
that would be the case even if I swapped the order of them as well. As
mentioned in a previous reply, I'll be editing the commit messages and
probably just include the various performance results in patch 2 and
reference patch 1 from that too.

-- 
Jens Axboe