Hi all experts,
so I just made some tests on a ARM 400MHz with I2C hardware driver DS2482
getting temperature from a DS18B20. 
For me it's a surprise:
I've written a small C-program using the owcapi (see attached .c and
Makefile) http://old.nabble.com/file/p32244690/testowfs.c testowfs.c 
http://old.nabble.com/file/p32244690/Makefile Makefile 

I started the owcapi program without any owserver to read the uncached temp
10-times:
time ./testowfs uncached/28.018ACE010000.D8/temperature9 10
and counted average 670ms per 9-bit measurement, so far away from the 93ms
the conversion in the device costs.
A 12-bit measurement took 1180ms in average, so not much slower.

Later I started owserver and an owshell program 10 times to read the
uncached temp:
time for i in {1..10}; do /opt/owfs/bin/owread
/uncached/28.018ACE010000.D8/temperature9; done
and counted 1070ms per measurement, so you are right, this is a little bit
slower

An owfs call I can't measure at the moment, due to problems with the fuse
(other thread)
   Achim


Paul Alfille-2 wrote:
> 
> In all cases, actual 1-wire communication is the slowest part of the
> process. Each bit transmitted is 62usec
> 
> The owshell approach requires a new process for every 1-wire call. Linux
> processes are supposed to fairly light but it's probably the slowest
> approach. And getting to that data in a C program isn't convenient.
> 
> owfs should be pretty fast, the actual "filesystem" part of it does
> require
> a pass through the kernel for each transaction, but the fuse interface is
> pure subroutine calls. Getting to the data in C (or any language) is very
> convenient -- file access is natural.
> 
> The owserver approach requires network communication. Again pretty fast.
> 
> If you really want the lowest latency, use owcapi in your C program. There
> is no pass through the network or filesystem and the interface is simple.
> 
> If other processes need to access the 1-wire data at the same time, you
> need
> owserver or the filesystem. If they need to access remotely, you need
> owserver. All the methods, including owcapi, can access owserver.
> 
> All the methods are pretty fast and I'd choose the simplest approach for
> development.
> 
> Alias should be pretty fast, it's just a binary table lookup from memory.
> 
> Paul Alfille
> 
> On Wed, Aug 10, 2011 at 4:11 AM, ekgnkb3d <[email protected]> wrote:
> 
>>
>> Hello everybody,
>> I'm new to this forum and shortly want to introduce myself:
>> During my tests with 1-Wire for a hot-water controller, I used the Maxim
>> Public Domain Kit. But due to some limitations I try to switch to owfs
>> now.
>> My biggest fear at owfs is a low performance causing high load at my CPU
>> which is with 400MHz and ARM9 very limited... I'm using an hardware I2C
>> channel /dev/i2c-0 with DS2482-100. The aim is to have a closed loop
>> controlling rate of max 250ms for 4 DS18B20 temp sensing plus DS2408 dac.
>> So
>> therefore my (newbie) questions:
>>
>> The convenience of all the different owfs client technologies is great
>> (thanks to the developers for this amazing piece of code!!!), assuming
>> there
>> is not much difference in adapting my C-coded application. Which one of
>> the
>> following give the fastest access, creating the minimum CPU load:
>>
>> - owfs, accessing files in directories via fuse
>> - owshell, calling shell commands via local owserver
>> - owserver, adding code to my C-application acting as TCP client
>>
>> Or maybe it is not worth a discussion and all mentioned possibilities
>> have
>> almost the same speed/load?
>>
>> Another questions:
>> Does using alias slowing down the access??
>>
>> thanks a lot
>>  Achim
>> --
>> View this message in context:
>> http://old.nabble.com/different-access-performance-tp32230443p32230443.html
>> Sent from the OWFS - Dev mailing list archive at Nabble.com.
>>
>>
>>
>> ------------------------------------------------------------------------------
>> uberSVN's rich system and user administration capabilities and model
>> configuration take the hassle out of deploying and managing Subversion
>> and
>> the tools developers use with it. Learn more about uberSVN and get a free
>> download at:  http://p.sf.net/sfu/wandisco-dev2dev
>> _______________________________________________
>> Owfs-developers mailing list
>> [email protected]
>> https://lists.sourceforge.net/lists/listinfo/owfs-developers
>>
> 
> ------------------------------------------------------------------------------
> uberSVN's rich system and user administration capabilities and model 
> configuration take the hassle out of deploying and managing Subversion and 
> the tools developers use with it. Learn more about uberSVN and get a free 
> download at:  http://p.sf.net/sfu/wandisco-dev2dev
> 
> _______________________________________________
> Owfs-developers mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/owfs-developers
> 
> 



-- 
View this message in context: 
http://old.nabble.com/different-access-performance-tp32230443p32244690.html
Sent from the OWFS - Dev mailing list archive at Nabble.com.


------------------------------------------------------------------------------
Get a FREE DOWNLOAD! and learn more about uberSVN rich system, 
user administration capabilities and model configuration. Take 
the hassle out of deploying and managing Subversion and the 
tools developers use with it. 
http://p.sf.net/sfu/wandisco-dev2dev
_______________________________________________
Owfs-developers mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/owfs-developers

Reply via email to