S3C2440裸机编程中汇编跳转到C函数失败

Question

//my_main.c
#define GPFCON (*(volatile unsigned long )0x56000050)
#define GPFDAT ((volatile unsigned long *)0x56000054)

#define GPF4_out (1<<(4*2))
#define GPF5_out (1<<(5*2))
#define GPF6_out (1<<(6*2))

void wait(volatile unsigned long dly)
{
for(; dly > 0; dly--);
}

int main(void)
{
unsigned long i = 0;

GPFCON = GPF4_out|GPF5_out|GPF6_out;        // 将LED1-3对应的GPF4/5/6三个引脚设为输出

while(1){
    wait(30000);
    GPFDAT = (~(i<<4));     // 根据i的值，点亮LED1-3

    if(++i == 8)
        i = 0;
}

return 0;

}

//start.s

.text
.global _start
_start:
ldr sp,=4096
bl disable_watchdog
bl set_sdram
bl init_nand
bl read_nand_to_sdram
ldr sp,=0x34000000
ldr lr,=loop
ldr pc,=main

loop:

b loop

//nand.lds

SECTIONS { 

firtst 0x00000000 : { head.o init.o nand.o}
second 0x30000000 : AT(4096) { my_main.o }
}

求教：为什么将wait函数改为内联函数，就能成功执行 ldr pc，=main
跳转到main中执行，
，如果不改，就跳转不到main中执行？