Re: [Mspgcc-users] ... iomacros.h, -fpack-struct

Stephen Douthit Sat, 25 Jul 2009 02:55:20 +0000

I think I've found a better solution than setting -fpack-struct, whichmight be undesired for other structures.

Packing only the __pin struct within ioregister_t union givessizeof(ioregister_t) as 1, which fixes all this incorrect portaddressing badness(at least for Rainer's example code).


Basing attribute usage on this link:
http://www.ohse.de/uwe/articles/gcc-attributes.html#type-packed

One liner patch attached.

Later,
Steve

Roberto Padovani wrote:

In packing the structure, just beware of ALWAYS aligning a word
variable to an even address, otherwise you'll end up in a mess.
As a matter of fact, if you call a word instruction (like mov.w) with
a word argument aligned at an odd address, the msp430 will zero out
the most significant byte of the word.

R#

2009/7/24 Frank Harvey <[email protected]>:

I added arg -fpack-struct to the msp430-gcc command and the addresses started 
incrementing by 1 as you would expect.

msp430-gcc -S -fpack-struct main.c

The problem smells like an old one I encountered on other machines where the structs are 
actually the size you'd think - but, when something goes to assign memory addresses to 
them, it pads the length to make it come out to the next word boundary. This is generally 
"a speed thing" but it plays havoc if you're trying to exactly map a set of 
structs over a just-arrived communications message.

frank


--- On Wed, 7/22/09, Stephen Douthit <[email protected]> wrote:

From: Stephen Douthit <[email protected]>
Subject: Re: [Mspgcc-users] problem with iomacros.h
To: "GCC for MSP430 - http://mspgcc.sf.net"; <[email protected]>
Date: Wednesday, July 22, 2009, 7:58 PM
On Wed, Jul 22, 2009 at 4:06 AM,
Rainer Haape <[email protected]>
wrote:

Hello,
I have a problem with iomacros.h under linux with the

latest Versions of

mspgcc.
All ioregisters seem to be 16 Bit instead of 8 Bit, so

all but one adresses

are wrong.
The following lines in c:

#include <msp430x16x.h>
#include <iomacros.h>

int main (void)
{
  char i;
  i = port1.in.pin1;

    // 0x0020

  port1.out.pin1 = 1;

   // 0x0021

  port1.dir.pin1 = 1;

   // 0x0022

  i = port1.ifg.pin1;

   // 0x0023

  port1.ie.pin1 = 1;

    // 0x0025

  port1.ies.pin1 = 1;

   // 0x0024

  port1.sel.pin1 = 1;

   // 0x0026

}

were compiled to:
--- snip ---

int main (void)
{
    4036:       31

40 00 39     mov
   #14592, r1      ;#0x3900

  char i;
  i = port1.in.pin1;
    403a:       e2

b3 20 00
   bit.b   #2,
   &0x0020 ;r3 As==10

    403e:       4f

43
   clr.b   r15

    4040:       4f

63
   adc.b   r15

  port1.out.pin1 = 1;
    4042:       e2

d3 22 00
   bis.b   #2,
   &0x0022 ;r3 As==10

  port1.dir.pin1 = 1;
    4046:       e2

d3 24 00
   bis.b   #2,
   &0x0024 ;r3 As==10

  i = port1.ifg.pin1;
    404a:       e2

b3 26 00
   bit.b   #2,
   &0x0026 ;r3 As==10

    404e:       4f

43
   clr.b   r15

    4050:       4f

63
   adc.b   r15

  port1.ie.pin1 = 1;
    4052:       e2

d3 2a 00
   bis.b   #2,
   &0x002a ;r3 As==10

  port1.ies.pin1 = 1;
    4056:       e2

d3 28 00
   bis.b   #2,
   &0x0028 ;r3 As==10

  port1.sel.pin1 = 1;
    405a:       e2

d3 2c 00
   bis.b   #2,
   &0x002c ;r3 As==10

}

--- snip ---

What is going wrong?

Thank you in advance,
Rainer Haape
[email protected]

------------------------------------------------------------------------------

_______________________________________________
Mspgcc-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/mspgcc-users

I built the mspgcc toolchain on a Debian(Lenny) setup a few
days ago, but
hadn't tried it out until I saw your message.

I got the same results you did, but after poking around the
code a little I
don't think there's anything wrong in iomacros.h.
Instead, it looks like
for some reason the ioregister_t union from
msp430/iostructures.h is getting
compled to twice the expected size.

#include <msp430/iostructures.h>

void main (void)
{
  char i;
  i=sizeof(char);
  i=sizeof(ioregister_t);
}

compiles to:

void main (void)
{
    fc36:       31 40 7e
02     mov
   #638,   r1
  ;#0x027e
    fc3a:       04
41           mov
   r1,     r4
  char i;
  i=sizeof(char);
    fc3c:       d4 43 00
00     mov.b   #1,
   0(r4)   ;r3 As==01,
0x0000(r4)
  i=sizeof(ioregister_t);
    fc40:       e4 43 00
00     mov.b   #2,
   0(r4)   ;r3 As==10,
0x0000(r4)
}

googling for mspgcc union errors gave me a bunch of reports
of misaligned
int reads from int/char unions, but nothing about the size
of the union
going haywire.

Has anyone else seen/fixed this?

Thanks,
Steve

-----Inline Attachment Follows-----

------------------------------------------------------------------------------

-----Inline Attachment Follows-----

_______________________________________________
Mspgcc-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/mspgcc-users

------------------------------------------------------------------------------
_______________________________________________
Mspgcc-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/mspgcc-users


------------------------------------------------------------------------------
_______________________________________________
Mspgcc-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/mspgcc-users

Re: [Mspgcc-users] ... iomacros.h, -fpack-struct

Reply via email to