Re: [lwip-users] Choosing the right lwip API...

[DownyTif]

I do have some more questions on this topic... I decided to test the speed
improvements using the RAW API. This API allows me to use only one thread to
do whatever I want instead of two threads using the sequential API because
the "recv" is blocking and I need to send data in the idle time.

1- I just read from the Wiki that the RAW API can only be used by the main
thread?? In my project, I need the main thread to call
"vTaskStartScheduler()" to run FreeRTOS for multithreading, which never
returns. So, I can't use my main thread to manage the RAW API, but I can
call "tcp_new", "tcp_bind", "tcp_listen" and "tcp_accept" before. Is that
sufficient?

2- 
What do I need to call to use/init the RAW API? For the sequential, I was
using sys_init(), mem_init(), memp_init(), pbuf_init(), netif_init(),
tcpip_init(EthernetConfigInterface, NULL). I'm not sure if I still need
those... or if I'm missing something.

3-
I haven't found it, so I think it would be good to have a little resume of
what are the #defines needed by all the API. For example, if I want to use
the RAW API, I must absolutly define some values in the lwipopts.h file.
Personally, I'm not sure I have everything configured ok..

Thanks,
Downy





Jonathan Larmour wrote:
> 
> DownyTif wrote:
>> Hi all,
>> 
>> I was wondering what API I should use to do what I want to do. By the
>> way,
>> I'm new to the mail-list/forum and lwip world. I tried to find the
>> differences between the Raw, sockets or sequential (netconn) APIs, but I
>> haven't found anything useful.
> 
> http://lwip.wikia.com/wiki/LwIP_Wiki has a lot of stuff about the APIs. In 
> essence the raw API is event-driven, the sequential (netconn) API, is a 
> more traditional API with calls to e.g. netconn_write() which can block 
> until the operation completes, and the sockets API is the same sort of 
> thing as the sequential API, but using the well-known BSD sockets 
> functions (like socket(), bind(), listen() accept(), etc.).
> 
> Each API (raw->sequential->sockets) in turn has more overhead (both code 
> and data) and is slightly slower than the previous one, and for some 
> hardware, you can get much better performance with the raw or sequential 
> APIs than the sockets API in some situations, but that doesn't mean the 
> sequential or sockets APIs are slow. If performance is critical for you 
> (and your other mail implies this might be tree) it could be that the raw 
> API is a better choice. Although you have to do more things yourself with 
> the raw API.
> 
>> So, I was thinking that maybe the sequential API (netconn) was not the
>> way
>> to go.
> 
> It's unrelated to your other mail about large packet sizes. An 
> event-driven model such as the raw API affects your program design a lot. 
> Some people like the familiarity of the sockets API if they've used it 
> elsewhere, or if they are reusing existing code. It's a choice with 
> trade-offs between them, which is why lwIP has all three!
> 
> Jifl
> -- 
> eCosCentric Limited      http://www.eCosCentric.com/     The eCos experts
> Barnwell House, Barnwell Drive, Cambridge, UK.       Tel: +44 1223 245571
> Registered in England and Wales: Reg No 4422071.
> ------["The best things in life aren't things."]------      Opinions==mine
> 
> 
> _______________________________________________
> lwip-users mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/lwip-users
> 
> 

-- 
View this message in context: 
http://www.nabble.com/Choosing-the-right-lwip-API...-tp21575028p21713721.html
Sent from the lwip-users mailing list archive at Nabble.com.