我正在编写C代码并为PowerPC架构编译它。我想告诉编译器将所有类似返回的指令替换为函数代码末尾的分支。这是由于一些奇怪的要求，我不能在这段代码中使用任何返回来确保之后执行更多的程序集。
例如，我将C代码编译成以下汇编代码：
lis r9,4096
lis r8,255
lwz r10,0(r9)
ori r8,r8,65535
addi r9,r10,192
cmplw cr7,r9,r8
blelr- cr7 # Return if the cr7 register is "less than or equal"
lis r8,512
cmplw cr7,r9,r8
bgtlr- cr7 # Return if the cr7 register is "greater than or equal"
lwz r10,192(r10)
lis r8,303
ori r8,r8,65535
addi r9,r10,320
cmplw cr7,r9,r8
blelr- cr7 # Return if the cr7 register is "less than or equal"
lis r8,528
cmplw cr7,r9,r8
bgtlr- cr7 # Return if the cr7 register is "greater than or equal"
lwz r10,320(r10)
lis r8,287
ori r8,r8,65535
subi r9,r10,448
cmplw cr7,r9,r8
blelr- cr7
lis r8,544
cmplw cr7,r9,r8
bgtlr- cr7 # Return if the cr7 register is "greater than or equal"
lis r9,4919
ori r9,r9,4919
stw r9,-448(r10)
blr # Return
我想要的是所有类似返回的语句被一个always分支替换为函数代码的末尾，如下所示：
lis r9,4096
lis r8,255
lwz r10,0(r9)
ori r8,r8,65535
addi r9,r10,192
cmplw cr7,r9,r8
ble _END # Branch to the _END label if "less than"
lis r8,512
cmplw cr7,r9,r8
bgt _END # Branch to the _END label if "greater than"
lwz r10,192(r10)
lis r8,303
ori r8,r8,65535
addi r9,r10,320
cmplw cr7,r9,r8
ble cr7 # Branch to the _END label if "less than"
lis r8,528
cmplw cr7,r9,r8
bgt _END # Branch to the _END label if "greater than"
lwz r10,320(r10)
lis r8,287
ori r8,r8,65535
subi r9,r10,448
cmplw cr7,r9,r8
blelr- cr7
lis r8,544
cmplw cr7,r9,r8
bgt _END # Branch to the _END label if "greater than"
lis r9,4919
ori r9,r9,4919
stw r9,-448(r10)
_END:
blr # I guess it should stay otherwise the function call will continue executing "random" garbage such as the next function in the .text section. Via post-processing this would be easy to strip though!
是否可以自动化此编译首选项？
答案 0 :(得分：2)
来自@Jester和@fuz的评论提供了答案，我只是把它写下来。
你需要：
return替换为goto。goto end;和return;是一回事。内联汇编是一种告诉编译器的好方法，不要搞乱这个＆＃34;。使用一些宏：
#define RETURN goto _end
#define END _end: asm volatile("")
void foo(int x) {
if (x==1) RETURN;
if (x==2) RETURN;
printf("Hello, world\n");
END;
}
asm volatile("")指示编译器插入一些汇编代码。编译器不知道这个汇编代码的作用，因此它不能进行任何跳过它的优化（即使它是0的指令）。