Re: [PATCH] block: blk-merge: fast-clone bio when splitting rw bios

[Jens Axboe <axboe@kernel.dk>]

On 09/17/2015 09:50 AM, Ming Lei wrote:
> On Thu, Sep 17, 2015 at 11:19 PM, Jens Axboe <axboe@kernel.dk> wrote:
>> On 09/17/2015 09:13 AM, Ming Lei wrote:
>>>
>>> biovecs has become immutable since v3.13, so it isn't necessary
>>> to allocate biovecs for the new cloned bios, then we can save
>>> one extra biovecs allocation/copy, and the allocation is often
>>> not fixed-length and a bit more expensive.
>>>
>>> For example, if the 'max_sectors_kb' of null blk's queue is set
>>> as 16(32 sectors) via sysfs just for making more splits, this patch
>>> can increase throught about ~70% in the sequential read test over
>>> null_blk(direct io, bs: 1M).
>>
>>
>> I'd be curious how this compares to before we did the splitting, not
>> exceeding the limits through bio_add_page() instead?
>
> Let me show these test results:
>
> ----------------------------------------------------------------------------------
> kernel                                                    | throught
> ----------------------------------------------------------------------------------
> 4.3.0-rc1-next-20150916                   | bw=12227MB/s, iops=12227
> ----------------------------------------------------------------------------------
> 4.3.0-rc1-next-20150916 with patch | bw=21011MB/s, iops=21011
> ----------------------------------------------------------------------------------
> v4.2                                                       |
> bw=18959MB/s, iops=18958
> ----------------------------------------------------------------------------------
>
> So from the above, looks this patch is kind of fix for performance regression
> introduced by 54efd50bfd(block: make generic_make_request handle
> arbitrarily sized bios), :-)

So that's 1MB user IO, and 16KB device limit, correct? If that is the 
case, then the results make sense. And looks like we're still ahead of 
the older bio_add_page() approach, which is what I mostly cared about. 
Thanks! I'll apply this for -rc2.

-- 
Jens Axboe