topic Re: HC908GP __isflag_carry() in 8-bit Microcontrollers

HC908GP __isflag_carry()

CJH — Wed, 22 Nov 2006 06:26:57 GMT

I'm trying to check the carry bit using __isflag_carry() but keep getting the linker error

L1822: Symbol __isflag_carry in file ....c.o is undefined

According to the manual file,

"To avoid using HLI for this purpose, the Compiler offers a set of intrinsic functions. The code of these functions is inlined. The processor flags listed in Table 8.3 are read by the associated intrinsic function. "

Since it's supposed to be intrinsic, there does not appear to be a .h file to declare it.

What am I missing? (CW V5.7.0 Build 2015, HC908GP)

Thanks.

Re: HC908GP __isflag_carry()

CompilerGuru — Thu, 23 Nov 2006 00:58:56 GMT

The following sample builds/runs fine for me (default options)

Daniel

#include <hidef.h>
#include "derivative.h"

void Err() { for (;;) { /* wait here */ } }

char __isflag_carry(void);

char fun(unsigned char a) {
  __asm ADD #1;
  return __isflag_carry();
}
void main() {
  if (fun(1) != 0) {
    Err();
  }
  if (fun(255) != 1) {
    Err();
  }

  for(;;) {
    __RESET_WATCHDOG(); /* feeds the dog */
  }
}

    9:    __asm ADD #1;
  0000 ab01     [2]             ADD   #1
   10:    return __isflag_carry();
  0002 2502     [3]             BCS   L6 ;abs = 0006
  0004 4f       [1]             CLRA 
  0005 81       [4]             RTS  
  0006          [5]     L6:    
  0006 a601     [2]             LDA   #1
   11:  }
  0008 81       [4]             RTS  
   12:  void main() {

Re: HC908GP __isflag_carry()

Geezer — Thu, 23 Nov 2006 01:13:06 GMT

Check your compiler option "Disable optimizations".
Al

Re: HC908GP __isflag_carry()

CJH — Thu, 23 Nov 2006 02:07:20 GMT

Strange - I copied your code into CW and compiled and I get:

  12:     __asm ADD #1;
0000 ab01     [2]             ADD   #1
  13:     return __isflag_carry();
0002 cd0000   [5]             JSR   __isflag_carry
  14:     }
0005 81       [4]             RTS
  15:
  16: void main() {

Which of course results in the __isflag_carry undefined message from the linker...

What version of CW are you running?

Re: HC908GP __isflag_carry()

CJH — Thu, 23 Nov 2006 02:08:11 GMT

I do have "disable optimizations" checked in the compiler options tab.

Re: HC908GP __isflag_carry()

CompilerGuru — Thu, 23 Nov 2006 04:21:04 GMT

I see, it's a silly behavior of the -ont option.
If you disable the complete tree optimizer, then the intrinsics are not recognized :smileysad:
2 solutions:
- only disable the treeopt suboptimization you really do not want. So instead of -ont, use -ont=ab.......
- define the __isflag_carry function in another module.
As the function is only called with disabled optimizations, it wont affect the release build. And the debug build will just call it.

Daniel

Re: HC908GP __isflag_carry()

Geezer — Thu, 23 Nov 2006 04:39:11 GMT

CJH:

For younger generations checking something means clicking on a little square icon.

Sigh.Poor modern English on my part... when I said to "check" an item I meant it in the older sense of "go look at it".

Al

Re: HC908GP __isflag_carry()

CJH — Thu, 23 Nov 2006 05:07:57 GMT

Man, I would never have figured that out - disabling optimizations disables intrinsics too! Certainly not obvious to me, and I don't think a very desirable behavior...

Thank you!

Re: HC908GP __isflag_carry()

CJH — Thu, 23 Nov 2006 05:10:17 GMT

Yeah, I got the part about looking at it - I assumed that turning off optimizations was the desired setting, so I figured having it "checked" was right...

I'm not all _that_ young...

Thanks to all for the help.

Carlos

Re: HC908GP __isflag_carry()

bigmac — Thu, 23 Nov 2006 09:34:40 GMT

Hello,

As a matter of curiosity, do both the two different compiled versions produce the intended result? It appears to me that one version may make a macro substitution, and the other version makes a sub-routine call, with either version giving correct operation. Does it have something to do with the compiler optimising for speed, or code compactness?

Just a thought.

Regards,

Mac

Re: HC908GP __isflag_carry()

CJH — Mon, 27 Nov 2006 20:49:18 GMT

Mac,

The main problem I was having is that it's NOT defined anywhere else - there is no such subroutine, so the version compiled with a subroutine call produces an error in linking since the linker doesn't find that subroutine anywhere. So no, they don't produce equivalent results, at least not in my opinion...

Carlos