Hi Vincent,
If I understand correctly, the word "volatile" itself is to tell the compiler not to ignore the statement. I am not sure if your statement is correct, but I have used
"asm volatile ("nop"::);"
and the compiler never removed the "nop"s from code.
Nayani
--- On Thu, 2/19/09, Vincent Trouilliez <address@hidden> wrote:
From: Vincent Trouilliez <address@hidden>
Subject: [avr-gcc-list] How to force GCC to not to remove "nop" statements ?
To: "address@hidden" <address@hidden>
Date: Thursday, February 19, 2009, 11:31 PM
Hi list,
Like many people I guess, I am using in-lined "nop" asm statements to
respect
the timing of external devices, when toggling their control
lines.
In the case at hand I am driving a text LCD module, and the
"Enable" line must be pulsed low for 250ns.
void lcd_send_nibble(uint8_t data)
{
...
//pulse Enable line
LCD_PORT |= LCD_E;
__asm__ volatile("nop");
__asm__ volatile("nop");
__asm__ volatile("nop");
__asm__ volatile("nop");
LCD_PORT &= ~LCD_E;
}
My AVR was running at 8MHz, so an instruction cycle of 125ns, so two nop
in-lined were spot on, just perfect, and it ran fine.
Today I pushed the AVR (ATmega32) up to its maximum of 16MHz.
So I now need 4 nops not two. So I did that.. and the LCD doesn't work
reliably. Suspecting the problem, I went straight to the generated
assembly, and horror, my 4 nop statements have disappeared !
I made a few trials. It appears that up to and including 3 nop
statements, GCC leaves them alone.
But if I put 4 of them, it removes
them all !
Any way to instruct GCC to leave my 4 nops ?
Am using avr-gcc 4.3.2
Regards,
--
Vince
_______________________________________________
AVR-GCC-list mailing list
address@hidden
http://lists.nongnu.org/mailman/listinfo/avr-gcc-list
|