[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: register declared variable for no optimization



If I remember correct the register keyword is an advise to the
compiler that the variable will be used heavy not that the compiler
always should use a register.

Anyway I think the GCC people can help you on this one.

Regards,

Søren Laursen

> Hi,
>
> I have a question on the GCC optimization. I am using
> GCC3.0.4 MIPS crosscompiler on Redhat 7.1. In a
> function I did the following:
>
> init()
> {
>   register unsigned int temp;
>       .
>       .
>   temp = devict.dev1.devtc
>       .
> }
>
> I compiled the code with optimization level 0 and
> found something weird about the register variable
> temp. This is shows in the disassembly file:
>
> for temp = devict.dev1.devtc
>     lui $3, 0xb801
>     lw  $3, 28($3)
>
> This is right since the CPU loads the
> devict.dev1.devtc value from memory to the register 3.
> However, after the above instructions, I found:
>
>      nop
>      sw $3, 0($30)
>      lw $2, 0($30)
>
> now the compiler asked the CPU to store the register
> in the stack. However, I declared the variable as a
> register, how come it still needs to go to the stack?
> I compiled the same code with gcc-2.8.1, optimization
> level 0, and did not find the same issue there.
>
> Why the gcc-3.0.4 did the weird stuff? Do I have to
> use at least level 1 to make the register declared
> work for it?
>
> Thanks a lot!
>
>
> Long
>
>
>
>
> __________________________________________________
> Do you Yahoo!?
> Yahoo! Mail Plus - Powerful. Affordable. Sign up now.
> http://mailplus.yahoo.com
>




Reply to: