 |
 |
| View previous topic :: View next topic |
| Author |
Message |
grifcj
Joined: 12 Oct 2009
Posts: 4
|
| Fixed Point Integers with Nested Unions and Structures |
 Posted: Thu Nov 12, 2009 12:48 am |
 |
|
I'm trying to code 8.8 fixed point integers. Yes, I think I read somewhere that CCS has intrinsic support for this data type, but I'm doing this for a class, so I'd like to learn a method that will be portable.
My problem is that CCS doesn't seem to be handling the bit fields in my structure correctly. In fact, it looks like the generated assembly is faulty. My compiler version is 4.041.
Here's my nested union/struct definition. You may recognize it from an embedded.com article.
| Code: |
// Signed fixed point integer.
// Range -128 to (127 - 2^8)
// Granularity 1/2^8 = .0039
union SFIXED8_8tag {
int16_t full;
struct spart8_8tag {
int16_t decimal:8;
int16_t integer:8;
} part;
};
typedef union SFIXED8_8tag int8_8_t;
|
Here's the generated assembly listing for one of these unions in action. Variable nDegrees is of type int8_8_t (the above union) at addresses 0x18:0x19. wAccX is just some uin16_t variable located at 0x4B:0x4c.
For the first C line, it functions as expected, copying two bytes. For the second and third lines of C code, it seems to mess up. You'll notice from the definition that the bitfields are 8 bits long, so it shouldn't be moving 2 bytes for the assignment. Also, for the second C line, it grabs data from a completely different variable at address 0x1A.
| Code: |
.................... wAccX = nDegrees.full;
0162: MOVFF 19,4C
0166: MOVFF 18,4B
.................... wAccX = nDegrees.part.integer;
016A: MOVFF 1A,4C
016E: MOVFF 19,4B
.................... wAccX = nDegrees.part.decimal;
0172: MOVFF 19,4C
0176: MOVFF 18,4B
|
I suppose I might be able to come up with some workarounds, but I would like to know what's going here. |
|
 |
FvM
Joined: 27 Aug 2008
Posts: 2337
Location: Germany
|
|
| Posted: Thu Nov 12, 2009 3:57 am |
|
|
I agree, that the PCH handling of the int16 xx:8 bitfields isn't according to the C standard. You can even do worse by coding
| Code: | | nDegrees.part.integer = wAccX; |
It's writing out of bounds, which is a severe bug that must never happen to a C compiler, whatever possibly incompatible types have been assigned by the programmer. The bug is still present in V4.099, by the way.
Apart from this inacceptable bug, the below definition should work for your purpose
| Code: | struct {
int8 decimal;
int8 integer;
} part8; |
|
|
|
|
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
Powered by phpBB © 2001, 2005 phpBB Group
|
FAQ 
Search
Register
Profile 
Log in to check your private messages 
Log in
