• State
  • Locked Locked
  • Replies 3 replies
  • Subscribers 79 subscribers
  • Views 3541 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

Logic Shift Left MSP430F2274 and CCS v5

Pramod Aryal
Intellectual 265 points
Other Parts Discussed in Thread: MSP430F2274

I am trying to use logic shift left for MSP430F2274 using CCS v5 in following way. This crashes when it tries to do shift.  I suspect it is because of arithmetic overflow. If so how can I work around ? If not what is the issue? Logic shift right appears to work fine. Only left has issue

unsigned char shift_len;

unsigned long int outdata;

shift_len = 24;

outdata = outdata << shift_len;

thanks

  • 0
    TI__Guru 56650 points

    Hi,

    i don't think the shift left is the problem. I made a small test program using CCS, and i don't see the crash:

    have you turned off the watchdog? have you assigned the correct inverrupt vector if you are using any interrupt(s)?

  • 0 in reply to Leo Hendrawan
    Intellectual 265 points

    Thank you for the response. Watchdog is turn off and interrupts are not enabled. 

    After spending hours on it, I discovered that single stepping the code is the problem. If logic shift left is done in nested function calls (3 levels inside), program crash most of the time and it restarts if I step through the code. Note that I have turned off code optimisation too.

    If I use same bit of code in IAR, it steps through fine. It appears that single stepping in CCS  v5 (in my case using MSP-FET430UIF and MSP430F2274 hardware) can not be trusted.

  • 0 in reply to Pramod Aryal
    Guru 227245 points

    The MSP does not have a "<<n" instruction (well, the X2 core has one for n<=4).

    So the compiler generates a mixture of byte swap, byte move and shift operations. To not waste too much space on this, the MSPGCC compiler (where I once analyzed this) uses an instruction stack where it jumps into the stack depending on the number of shifts and at the end the function returns with the result of the shift. A clever but puzzling construct.
    What instruciton chain does your compiler generate?

    In case of a (fixed) <<24 on a long int, I would do a move from low byte to high word, then a byte swap and clear of the lower 16 bit.

    Assuming the value in R14 (bit 0-15) and R15 (bit 16-32), a <<24 would be
    MOV.B R14,R15
    SWAPB R15
    CLR R14

    Independent of the code, there may be some errata on your specific MSP related to signle stepping and CPU instructions. See teh EMM and CPU errata in the device errata sheet.

    Pramod Aryal said:
    It appears that single stepping in CCS  v5 can not be trusted.

    The debugger generally cannot be trusted at all. Usually, it works fine for abstract code. But as soon as 'real world effects' are part of the problem, the debugger may become useless or even turn into a bugger. unfortunately, real-world effects are what you usually have to deal with on a microcontroller.

    Personally, I never (!) used a debugger on an MSP. Not a single debug session ever in all these years. I use logic analyzers and port pins, LEDs and UART messages for debugging.

**Attention** This is a public forum