Re: [lwip-users] alignment problems

[Bartlomiej Wicher]

Of course, you are right, I can address them indivudually:

        char foo[5] =   {0x12, 0x34, 0x56, 0x78, 0};
        x = foo[0]; //x = 0x12;
        x = foo[1]; //x = 0x34;
        x = *(foo); //x = 0x12
        x = *(foo+1); //x = 0x34;

I've just checked the way how compiler deals with that - it doubles the  
address of
an array and it's content:

        char foo[5] =   {0x12, 0x34, 0x56, 0x78, 0}; //address of foo is 0x2f91
        char *xP;
	xP = &foo[0]; //xP points to 0x5f22 (I see it in debugger) but it's value  
is 0x12
        xP = &foo[1]; //xP points to 0x5f23 but it's value is 0x34

I don't know how the compiler deals with overwriting the memory (maybe  
debugger shows some dummy values...).
The thing is that physically foo[0] and foo[1] have exactly the same  
address (0x2f91 in that case) and I'm not sure how it could influence  
lwIP. I've just checked how conversion from int* to char* works:

        int boo[5] =    {0x12, 0x34, 0x56, 0x78, 0}; //address 0x2f8c
        char *xP;
        xP = (char *)&boo[0]; //xP points to 0x5f18, value is 0x12
        xP = (char *)&boo[1]; //xP points to 0x5f1a, value is 0x34
        xP = (char *)&boo[2]; //xP points to 0x5f1c, value is 0x56
        xP = (char *)&boo[3]; //xP points to 0x5f1e, value is 0x78

Well, it seems to work fine, but I'm not absolutely sure if that behaviour  
is standard...Sorry for misunderstandings.
I'll set alignment to 2 and keep working with it. Right now I have some  
problems with memp_malloc - program crashes on that function...

Thank you for your response



> I've never seen a compiler that supported 8 bit chars that couldn't
> address each of them individually. The processor may not address them
> individually, but you should be able to access each char independently
> at the higher level. That's what higher level languages do - hide the
> details of the hardware.
>
> Access to individual char values on a word-based bus may result in the
> compiler generating a read+shift or read+mask, but you still can
> typically load/store even and odd addressed characters at the 'C' level.
>
>
> -----Original Message-----
> From: address@hidden
> [mailto:address@hidden On Behalf
> Of Kieran Mansley
> Sent: Thursday, November 08, 2007 6:27 AM
> To: Mailing list for lwIP users
> Subject: Re: [lwip-users] alignment problems
>
> On Thu, 2007-11-08 at 11:08 +0000, Jonathan Larmour wrote:
>
> > I'm concerned that you say that "two addresses" are only used with
> > long, and one for both char and int. Does this mean that for:
> > char foo[4];
> > that: &foo[0] == &foo[1] ?
> >
> > If that is the case, I think you may have an uphill struggle.
>
> Yeah, that confused me.  Sounds like it would be sensible to think of it
> as having 16-bit bytes rather than worrying about the word size.  If
> that's the case I think there may be problems for you in lwIP where
> we've confused bytes with octets (i.e. 8 bits).
>
> Kieran
>
>
>
> _______________________________________________
> lwip-users mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/lwip-users
>
>
> _______________________________________________
> lwip-users mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/lwip-users



--
Bartlomiej Wicher
University of Southern Denmark
Odense, Denmark