Re: [PATCH v2] mmc: fix async request mechanism for sequential read scenarios

[Konstantin Dorfman <kdorfman@codeaurora.org>]

On 11/05/2012 09:15 AM, Jaehoon Chung wrote:
Hello,
> Hi Konstantin,
> On 11/05/2012 03:20 PM, Per Förlin wrote:
>> Hi Konstantin,
>>
>> On 11/01/2012 03:40 PM, Konstantin Dorfman wrote:
>>> When current request is running on the bus and if next request fetched
>>> by mmcqd is NULL, mmc context (mmcqd thread) gets blocked until the
>>> current request completes. This means if new request comes in while
>>> the mmcqd thread is blocked, this new request can not be prepared in
>>> parallel to current ongoing request. This may result in latency to
>>> start new request.
>>>
>>> This change allows to wake up the MMC thread (which
>>> is waiting for the current running request to complete). Now once the
>>> MMC thread is woken up, new request can be fetched and prepared in
>>> parallel to current running request which means this new request can
>>> be started immediately after the current running request completes.
>>>
>>> With this change read throughput is improved by 16%.
>> What HW and what test cases have you been running?
> I also want to know which benchmark use?
> If you can share it, i will test with yours.
> 
> Best Regards,
> Jaehoon Chung

Our tests were done using lmdd and tiotest and iozone:
TIOTEST
-------
tiotest seq: ./data/tiotest -t 1 -d /data/mmc0 -f 800 -b $((512*1024))
-k 1 -k 3

tiotest rand: ./data/tiotest -t 4 -d /data/mmc0 -f 800 -b 4096 -k 2 -k 0
-p -s 2012 -r 12500 -n 20


LMDD
----
lmdd write: ./data/lmdd if=internal of=/data/mmc0/push1 bs=128k
count=2000 sync=1

lmdd read: ./data/lmdd of=internal if=/data/mmc0/push1 bs=128k count=2000



IOZONE
------
iozone (worst): ./data/iozone -i0 -i2 -r4k -s100m -O -o -I -f
/data/mmc0/file3

iozone (best): ./data/iozone -i0 -i2 -r4k -s100m -O -I -f /data/mmc0/file3

Thanks,
-- 
Konstantin Dorfman,
QUALCOMM ISRAEL, on behalf of Qualcomm Innovation Center,
Inc. is a member of Code Aurora Forum,
hosted by The Linux Foundation