optimization of while(!(PINA & 0x40));

Log in or register to post comments
Go To Last Post
5 posts / 0 new
Author
Message
shadowbane
shadowbane's picture
Level: New Member
Joined: Thu. Aug 26, 2004
Posts: 11 View posts
#1
Posted by shadowbane: Sat. Sep 10, 2005 - 07:57 AM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hello,

Perhaps I have an option wrong. Its been quite a while since I've done any AVR stuff, so its entirely possible... but...

while(!(PINA & 0x80));

produces:

sbis	0x19, 7
rjmp	.-4

as it should, but

while(!(PINA & 0x40));

produces:

ldi	r18, 0x01
ldi	r19, 0x00
in	r24, 0x19
eor	r25, r25
ldi	r20, 0x06
lsr	r25
ror	r24
dec	r20
brne	.-8 
eor	r24, r18
eor	r25, r19
sbrc	r24, 0
rjmp	.-22

Some kind of promotion problem?, or perhaps the code just didnt get optimized?

All bits except 7 cause the problem. I have tried several different registers. Im using a clean install of WinAVR 20050214 -Os. For now Im just ignoring the issue, and if necessary I can switch to assembly for the loop.

Anyone know what Im doing wrong?

Tags:

Tools, Compilers and General Programming
Top
abcminiuser
abcminiuser's picture
Level: Raving Lunatic
Joined: Fri. Jan 23, 2004
Posts: 9845 View posts
Location: Melbourne, Australia
Posted by abcminiuser: Sat. Sep 10, 2005 - 08:32 AM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

C promotes all constants to an int, or somthing like that. Try typcasting:

while(!(PINA & (unsigned char)0x80));

- Dean :twisted:

Make Atmel Studio better with my free extensions. Open source and feedback welcome!

Top
shadowbane
shadowbane's picture
Level: New Member
Joined: Thu. Aug 26, 2004
Posts: 11 View posts
Posted by shadowbane: Sat. Sep 10, 2005 - 08:39 AM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Thanks for the response..

note that I said that the value 0x80 optimizes correctly, and the value 0x40 does not.

But just to be on the safe side, I tried your suggestion, and recieved the same results: 0x80 = ok, 0x40 (and any other bit) = poor code...

Top
alandras
alandras's picture
Level: New Member
Joined: Tue. May 24, 2005
Posts: 17 View posts
Location: Transylvania
Posted by alandras: Sat. Sep 10, 2005 - 10:26 AM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I can compile corectly! What is your context?
Andras

int main( void )
	{
		while(!(PINB & 0x80));
		while(!(PINB & 0x40));
		while(!(PINB & 0x20));
		while(!(PINB & 0x02));
		while(!(PINB & 0x01));


000003d8 
:
 3d8:	cf ef       	ldi	r28, 0xFF	; 255
 3da:	d2 e0       	ldi	r29, 0x02	; 2
 3dc:	de bf       	out	0x3e, r29	; 62
 3de:	cd bf       	out	0x3d, r28	; 61
 3e0:	1f 9b       	sbis	0x03, 7	; 3
 3e2:	fe cf       	rjmp	.-4      	; 0x3e0
 3e4:	1e 9b       	sbis	0x03, 6	; 3
 3e6:	fe cf       	rjmp	.-4      	; 0x3e4
 3e8:	1d 9b       	sbis	0x03, 5	; 3
 3ea:	fe cf       	rjmp	.-4      	; 0x3e8
 3ec:	19 9b       	sbis	0x03, 1	; 3
 3ee:	fe cf       	rjmp	.-4      	; 0x3ec
 3f0:	18 9b       	sbis	0x03, 0	; 3
 3f2:	fe cf       	rjmp	.-4      	; 0x3f0
Top
shadowbane
shadowbane's picture
Level: New Member
Joined: Thu. Aug 26, 2004
Posts: 11 View posts
Posted by shadowbane: Thu. Sep 15, 2005 - 12:11 AM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Sorry for the slow response. Just found a key to the puzzle a few seconds ago. It appears to be the c++ compiler only.

Here's how to reproduce it. First make tst.c:

#include 
void tst()
{
  while(!(PINB & 0x80));
  while(!(PINB & 0x40));
}

not much to it.

This line produces the problem:
avr-g++ -c -Os -mmcu=atmega128 -o tst.o tst.c
This line works correctly:
avr-gcc -c -Os -mmcu=atmega128 -o tst.o tst.c

I use this to check if the assembly:
avr-objdump --disassemble tst.o

Its strange that it works correctly for 0x80, but incorrectly otherwise...

Top