Re: [PATCH v5] Subject: io_uring: releasing CPU resources when polling

From: Jens Axboe
Date: Fri Jun 28 2024 - 17:14:13 EST


On 6/25/24 2:39 AM, hexue wrote:
> On 6/19/24 15:51, Jens Axboe wrote:
>> On 6/19/24 08:18, hexue wrote:
>>
>> While I do suspect there are cases where hybrid polling will be more
>> efficient, not sure there are many of them. And you're most likely
>> better off just doing IRQ driven IO at that point? Particularly with the
>> fairly substantial overhead of maintaining the data you need, and time
>> querying.
>
> I rebuilt the test cases based on your information, that is, each drive has
> only one thread with high pressure, there is a significant performance loss.
> I previously provided test data for a single drive with multi-threaded, which
> can achieve performance stable and CPU savings.
>
> Thanks for your data, I will reduce the cost of each IO and submit the next
> version.

Sounds good, thanks. I don't doubt it's useful for some cases as-is, but
I think there's the potential for it to be more useful.

--
Jens Axboe