RE: [avr-gcc-list] optimization and interrupt

[Ben Mann]

Hi,

The optimised code is correct.

It is actually doing both comparisons, because it:

subtracts 113 from diffcapt (values below 113 become greater than 0xFF8F)
compares diffcapt to 24 (137 - 113)
branches if same or higher.

It is faster and more efficient than two comparisons

Ben Mann


-----Original Message-----
From: address@hidden
[mailto:address@hidden On Behalf Of Jörg
Krein
Sent: Wednesday, 9 March 2005 11:44 PM
To: address@hidden
Subject: [avr-gcc-list] optimization and interrupt


Hi,
avr-gcc behaves strange when optimizing interrupt code.
In my interrupt routine I compare a variable against two constants. I added
the code at the end of this posting. After compiling I found only one
comparison (113) in the assembler code:

movw r24,r20          
sts (lastcapt)+1,r19  
sts lastcapt,r18      
subi r24,lo8(-(-113)) 
sbci r25,hi8(-(-113)) 
sbiw r24,24           
brsh .L1              

After changing optimization level to 0 i got:

std Y+3,r24
std Y+4,r25
ldd r24,Y+1
ldd r25,Y+2           
sts (lastcapt)+1,r25  
sts lastcapt,r24      
ldd r24,Y+3           
ldd r25,Y+4           
cpi r24,113           
cpc r25,__zero_reg__  
brlo .L1              
ldd r24,Y+3           
ldd r25,Y+4           
cpi r24,137           

So here a two comparisons (113, 137) as expected.
It helps to declare diffcapt to volatile but that doesn't
make sense to me as it is only used in the interrupt
I hope someone has a logical answer, let me believe
in gcc again.

Regards, Joerg


volatile uint16_t       lastcapt;

SIGNAL( SIG_INPUT_CAPTURE1 )
{
        uint16_t actcapt;
        uint16_t diffcapt;

        actcapt = ICR1;

        diffcapt = actcapt - lastcapt;
        lastcapt = actcapt;

        if( (diffcapt > 112)  &&  (diffcapt < 137) )
        {...}
}




_______________________________________________
AVR-GCC-list mailing list
address@hidden
http://lists.nongnu.org/mailman/listinfo/avr-gcc-list