ROMCC problem

[YhLu]

Eric, 
the current romcc seems have problem to process "asm". Or bist cause the
problem?

Regards

YH

In Failover.E

normal_image:
        asm("jmp __normal_image"
                :
                : "a" (bist)
                :
                );
 cpu_reset:
        asm("jmp __cpu_reset"
                :
                : "a"(bist)
                :
                );
 fallback_image:

will produce 

Lfailover1063:
        /* main,failover.c:68.11 */
Lfailover1064:
        /* ,:0.0 */
 /* unknown %eax */
 /* unknown %ebx */
 /* unknown %ecx */
 /* unknown %edx */
#ASM

#NOT ASM

Old romcc result it right

Lfailover61:
        /* main,failover.c:21.28 */
        /* main,failover.c:67.18 */
        mov  $0 , %eax
#ASM
        jmp __normal_image
#NOT ASM
        /* main,failover.c:68.11 */
Lfailover62:
        /* main,failover.c:21.28 */
        /* main,failover.c:73.18 */
        mov  $0 , %eax
#ASM
        jmp __cpu_reset
#NOT ASM
        /* main,failover.c:74.16 */
Lfailover63:
        /* ,failover.c:82.1 */
Lfailover64:
/* max inline depth 6 */

YhLu <[e-mail]> writes:
> Eric, 
> the current romcc seems have problem to process "asm". Or bist cause the
> problem?

The optimizer took a look at the asm realized no one used
the output of it.  And optimized the instruction away.  
Correct but annoying.

Just mark the asm as volatile and the code will work fine.

Eric