Embeded Tutorial?

CC wrote:

> msg wrote:
> 
>> Tim Wescott wrote:
>>
>> <snip>
>>
>>>
>>> My question is:  Does anyone know a good tutorial that covers all of 
>>> this, or at least techniques for implementing the task loop/state 
>>> machine method?
>>
>>
>> Believe it or not, LaBrosse's book covers each topic you cited with
>> example (pseudo) code and is quite tutorial in its approach in my
>> opinion.
> 
> 
> Considering that Jean LaBrosse appears to have 4 books in print:
> 
> http://www.amazon.com/s/ref=nb_ss_b/002-3183383-0799266?url=search-alias%3Dstripbooks&field-keywords=LaBrosse%2C+jean&x=0&y=0 
> 
> 
> Could you tell us to which one you are referring?
> 
> 
> Thanks.
> 

There is a section in the 2nd book (uC/OS-II) that is pretty good; I also
recommend the online tutorial at FreeRTOS.org.

Michael

Matthias Arndt schrieb:
> Lanarcam schrieb:
>> http://www.le.ac.uk/engineering/mjp9/pes1ohp_a4.pdf
> 
> That's very interesting! I'll keep this as a generic tutorial in case I
> meet someone who is interested in such a course. It supplemented my
> knowledge on Embedded stuffs as well :)

Hm... from page 29 of the pdf:

 > Problem
 >
 > How do you create a simple delay without
 > using any hardware (timer) resources?
 >
 > Solution:
 >
 > Loop_Delay()
 > {
 >   unsigned x,y;
 >   for (x=0; x<65536; x++)
 >   {
 >     y++;
 >   }
 > }


I don't have any experience with the 8051, but I would expect that any 
c-compiler that has been updated during the last 10 years would optimize 
this function away.

The only reliable way to create a delay using such a for-loop is to 
declare y as volatile.

Old and obsolete stuff is you ask me.. For beginners this may do more 
harm than good.

   Nils

Nils wrote:
> Matthias Arndt schrieb:
>> Lanarcam schrieb:
>>> http://www.le.ac.uk/engineering/mjp9/pes1ohp_a4.pdf
>>
>> That's very interesting! I'll keep this as a generic tutorial in case I
>> meet someone who is interested in such a course. It supplemented my
>> knowledge on Embedded stuffs as well :)
> 
> Hm... from page 29 of the pdf:
> 
>  > Problem
>  >
>  > How do you create a simple delay without
>  > using any hardware (timer) resources?
>  >
>  > Solution:
>  >
>  > Loop_Delay()
>  > {
>  >   unsigned x,y;
>  >   for (x=0; x<65536; x++)
>  >   {
>  >     y++;
>  >   }
>  > }
> 
> 
> I don't have any experience with the 8051, but I would expect that any 
> c-compiler that has been updated during the last 10 years would optimize 
> this function away.
> 
> The only reliable way to create a delay using such a for-loop is to 
> declare y as volatile.
> 
> Old and obsolete stuff is you ask me.. For beginners this may do more 
> harm than good.
> 
>   Nils
> 
> 
> 
Why would this be specific to 8051 ??

Any C compiler would compile this code.

Old and obsolete, I take that personally !!! ;-)

donald

donald schrieb:

> 
> Any C compiler would compile this code.
> 

Sure it will compile, but why write such a null-code at all?

This is what GCC 4.2.2 makes out of the code. Compiled with -O1 for x86, 
but I'm sure it'll look similar for other architectures and compilers:

 > _delay  LABEL NEAR
 >         push    ebp              ; 0000 _ 55
 >         mov     ebp, esp         ; 0001 _ 89. E5
 >         pop     ebp              ; 0003 _ 5D
 >         ret                      ; 0004 _ C3


Not much delay there if you ask me..

> Old and obsolete, I take that personally !!! ;-)

:-)

   Nils

Nils wrote:
> donald schrieb:
> 
>>
>> Any C compiler would compile this code.
>>
> 
> Sure it will compile, but why write such a null-code at all?
> 
> This is what GCC 4.2.2 makes out of the code. Compiled with -O1 for x86, 
> but I'm sure it'll look similar for other architectures and compilers:
> 
>  > _delay  LABEL NEAR
>  >         push    ebp              ; 0000 _ 55
>  >         mov     ebp, esp         ; 0001 _ 89. E5
>  >         pop     ebp              ; 0003 _ 5D
>  >         ret                      ; 0004 _ C3
> 
> 
> Not much delay there if you ask me..
> 
>> Old and obsolete, I take that personally !!! ;-)
> 
> :-)
> 
>   Nils
Try making x and y static.

donald

donald schrieb:

> Try making x and y static.
> 
> donald

Doesn't makes it better :-)

This code:

 > void delay ()
 > {
 >   static unsigned x,y;
 >   for (x=0; x<65536; x++)
 >   {
 >     y++;
 >   }
 > }


Compiles into:

 > _delay  LABEL NEAR
 >         mov     eax, dword ptr [_y.1526]
 >         push    ebp
 >         mov     ebp, esp
 >         pop     ebp
 >         add     eax, 65536
 >         mov     dword ptr [_x.1525], 65536
 >         mov     dword ptr [_y.1526], eax
 >         ret

Still not much of a delay (and not much of clever code either...)

If you want the compiler to not optimize the loop away y must be 
declared as volatile. No other way around it.


I see such delay-loop b*shit way to often. Last time it was in a DSP 
example code for some init-code to wait some moment after powering on a 
piece of hardware. Good idea in principle.

I don't know what the engineer was smoking while he wrote that code 
though. The code even had comments and gave hints how many iterations 
would give what kinds of delays. How stupid is that? All compilers 
removes redundant loops unless you compile without optimizations or 
force them via volatiles.

Nils wrote:

<snip>
> I don't know what the engineer was smoking while he wrote that code 
> though. The code even had comments and gave hints how many iterations 
> would give what kinds of delays. How stupid is that? All compilers 
> removes redundant loops unless you compile without optimizations or 
> force them via volatiles.

Not true, for example certain versions of Hi-Tech C would compile
that code into a delay loop correctly without any special command line
switches. No need for 'volatile' declarations.

Michael

In message <13sp9ivef4avidf@corp.supernews.com>, msg 
<msg@_cybertheque.org_> writes
>Nils wrote:
>
><snip>
>> I don't know what the engineer was smoking while he wrote that code 
>>though. The code even had comments and gave hints how many iterations 
>>would give what kinds of delays. How stupid is that? All compilers 
>>removes redundant loops unless you compile without optimizations or 
>>force them via volatiles.
>
>Not true, for example certain versions of Hi-Tech C would compile
>that code into a delay loop correctly without any special command line
>switches. No need for 'volatile' declarations.
>
>Michael

As you say "certain versions" of some compilers will do it one way and 
lots will optimise it out.

So as a delay it is non portable and the delay is completely variable 
depending on the system it is running on.

In which case do a hardware delay using HW timers. It to will not be 
portable BUT it will give a precise and reliable delay.

-- 
\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\
\/\/\/\/\ Chris Hills  Staffs  England     /\/\/\/\/
/\/\/ chris@phaedsys.org      www.phaedsys.org \/\/\
\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/