class Interlocked
{
public:
    static inline void Lock()
    {
        NVIC_DisableIRQ(TIMER3_IRQn);
    }
    
    static inline void Unlock()
    {
        NVIC_EnableIRQ(TIMER3_IRQn);
    }
    
    static inline int Increment(int volatile *src)
    {
        Lock();
        int val = *++src;
        Unlock();
        return val;
    }
    
    static inline int Decrement(int volatile *src)
    {
        Lock();
        int val = *--src;
        Unlock();
        return val;
    }
    
    static inline int Exchange(int volatile *src, int val)
    {
        Lock();
        int ret = *src;
        *src = val;
        Unlock();
        return ret;
    }
    
    static inline int CompareExchange(int volatile *src, int Exchange, int Comparand)
    {
        Lock();
        int ret = *src;
        if (ret == Comparand) *src = Exchange;
        Unlock();
        return ret;
    }
    
};

Are there any portable ways to implmenent atomic operations on the mbed. I'd like to implmement the equivalent of the Windows API's Interlocked family of functions (InterlockedIncrement, InterlockedDecrement, InterlockedExchange, InterlockedCompareExchange, etc) I've come up with this (not yet debugged), as I'm using a Ticker to schedule callbacks: <<code>> class Interlocked { public: static inline void Lock() { NVIC_DisableIRQ(TIMER3_IRQn); } static inline void Unlock() { NVIC_EnableIRQ(TIMER3_IRQn); } static inline int Increment(int volatile *src) { Lock(); int val = *++src; Unlock(); return val; } static inline int Decrement(int volatile *src) { Lock(); int val = *--src; Unlock(); return val; } static inline int Exchange(int volatile *src, int val) { Lock(); int ret = *src; *src = val; Unlock(); return ret; } static inline int CompareExchange(int volatile *src, int Exchange, int Comparand) { Lock(); int ret = *src; if (ret == Comparand) *src = Exchange; Unlock(); return ret; } }; <</code>> But I'm not happy with the NVIC_DisableIRQ(TIMER3_IRQn)call, as it seems like overkill, when all I want is to atomically fetch or set the values. It looks like the {{{__ldrex and __strex}}} functions might do it, but is that portable? Anthony Wieser\\ Wieser Software Ltd

    volatile int t;
    __sync_add_and_fetch(&t, 1);       

" Undefined symbol __user_cmpxchg_4 (referred from arm_sync_4.o)." in file "/"

I came across this, "The ARM Compiler now supports the GNU Compiler atomic builtin functions in {{{--gnu}}} mode for atomic memory access." and includes {{{__synch_add_and_fetch}}} the following code compiles: <<code>> volatile int t; __sync_add_and_fetch(&t, 1); <</code>> but when linked for the LPC2368 I get this. {{{ " Undefined symbol __user_cmpxchg_4 (referred from arm_sync_4.o)." in file "/" }}} Do I need to implement this function, or am I missing a library?

Hi Anthony, You can certainly use ldrex/strex, but looking at your previous code example, perhaps the portable functions {{{__disable_irq()}}} and {{{__enable_irq()}}} are what you are after? Simon