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

[hexue]

On 6/19/24 16:22, Pavel Begunkov wrote:
>On 6/19/24 08:18, hexue wrote:
>
>> Test results:
>> set 8 poll queues, fio-3.35, Gen5 SSD, 8 CPU VM
>>
>> per CPU utilization:
>>      read(128k, QD64, 1Job)     53%   write(128k, QD64, 1Job)     45%
>>      randread(4k, QD64, 16Job)  70%   randwrite(4k, QD64, 16Job)  16%
>> performance reduction:
>>      read  0.92%     write  0.92%    randread  1.61%    randwrite  0%
>
>What are numbers for normal / non-IOPOLL runs?

for normal poll, per CPU utilization is 100%*8
performance like this:
				read			write			randread		randwrite
normal poll		BW=10.9GiB/s    BW=6202MiB/s 	IOPS=1682k		IOPS=255k
hybrid poll     BW=10.8GiB/s	BW=6145MiB/s	IOPS=1655k		IOPS=255k

--
hexue