Re: no-op delay loops

From: Rasmus Villemoes
Date: Mon Nov 30 2015 - 16:29:35 EST


On Fri, Nov 27 2015, Arnd Bergmann <arnd@xxxxxxxx> wrote:

> On Friday 27 November 2015 09:53:50 Rasmus Villemoes wrote:
>>
>> It seems that gcc happily compiles
>>
>> for (i = 0; i < 1000000000; ++i) ;
>>
>> into simply
>>
>> i = 1000000000;
>>
>> (which is then usually eliminated as a dead store). At least at -O2, and
>> when i is not declared volatile. So it would seem that the loops at
>>
>> arch/mips/pci/pci-rt2880.c:235
>> arch/mips/pmcs-msp71xx/msp_setup.c:80
>> arch/mips/sni/reset.c:35
>>
>> actually don't do anything. (In the middle one, i is 'register', but
>> that doesn't change anything.) Is mips compiled with some special flags
>> that would make gcc actually emit code for the above?
>>
>
> I remember that gcc used to not optimize code that looked like a
> delay loop such as the above, and my tests show that this was still
> the case in gcc-4.0.3, but starting with gcc-4.1 it opimtized away
> that loop.

OK, thanks. That's a very very long time ago.

FWIW, the remaining instances that my trivial coccinelle script found
are

./arch/alpha/boot/main.c:187:1-4: no-op delay loop
./arch/m68k/68000/m68VZ328.c:86:10-13: no-op delay loop
./arch/m68k/bvme6000/config.c:338:2-5: no-op delay loop
./arch/m68k/coldfire/m53xx.c:533:1-4: no-op delay loop
./drivers/cpufreq/cris-artpec3-cpufreq.c:85:3-6: no-op delay loop
./drivers/cpufreq/cris-etraxfs-cpufreq.c:85:3-6: no-op delay loop
./drivers/tty/hvc/hvc_opal.c:313:3-6: no-op delay loop
./drivers/tty/hvc/hvc_vio.c:289:3-6: no-op delay loop

(cc += a few people). The tty ones use volatile, so they probably work,
though one might still want to use the *delay API.

Rasmus
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/