RE: [avr-gcc-list] custom signal prologue

[Christian Troedhandl]

On Wed, 24 Nov 2004, Ben Mann wrote:

> I had attempted the same thing but also wasn't wild about
> reverse-engineering the register list after any changes.
>
> I quite like Christian's solution to this problem, which allows a much
> cleaner approach to managing the registers, albeit at the cost of an extra
> function call (ie call/ret overhead). For the time being I'm using this
> method.
>
> Now, if there were a way to make my function call inline, and thereby
> magically instantiate the prologue without the call/ret pair, ie like this:

Another idea...

If we exchange SIG_UART1_DATA() and MyInterruptHandler() and leave out the 
function call and the ret instruction, the processor will continue to 
MyInterruptHandler() after finishing SIG_UART1_DATA():

void SIG_UART1_DATA(void) __attribute__ ((naked));
void SIG_UART1_DATA(void) //SIGNAL(SIG_UART1_DATA)
{
   //Disable UDRIE
   BYTE UCSRBReg;
   asm volatile ("push %0\n\t" : "=r" (UCSRBReg) :);
   UCSRBReg= UCSR1B;
   UCSR1B = UCSRBReg& ~_BV(UDRIE);
   //Enable interrupts
   sei();
   asm volatile ("pop %0\n\t" : "=r" (UCSRBReg) :);
   // CPU will continue to MyInterruptHandler()
}

inline void MyInterruptHandler(void) __attribute__ ((signal,always_inline));
inline void MyInterruptHandler(void) {
   ...Do Something...
}


The only problem is, that I don't know if the compiler will always 
put the functions in the right order.

Christian

---------------------------------------------------------------------
Christian Troedhandl                   mailto:address@hidden
Embedded Computing Systems                  voice:+43 (1) 58801-18262
Vienna University of Technology
A-1040 Wien, Treitlstr. 3/182-2        http://www.vmars.tuwien.ac.at/