Bug in PPC inline assembly?

Bug in PPC inline assembly?

[Stefan]

I have some problems with using inline PowerPC assembly in GCC (4.0.1). 
Consider the following code:

    void save_fp_register(double* buffer)
    {
        asm("stfd F0,  0(%0)" : : "r" (buffer) );
    }

When compiled with -Os the generated code becomes

    save_fp_register:
        stfd F0,  0(3)
        blr

Which is OK; the input parameter comes in r3 which is used directly in 
the stfd operation. However, when compiled with -O0 (no optimization) 
the generated code instead becomes

    save_fp_register:
        stwu 1,-24(1)
        stw 31,20(1)
        mr 31,1
        stw 3,8(31)
        lwz 0,8(31)
        stfd F0,  0(0) <-- error
        lwz 11,0(1)
        lwz 31,-4(11)
        mr 1,11
        blr

The input parameter r3 is transferred to r0, which is used in the stfd 
operation. However, for this particular operation r0 means 0 -- not the 
value stored in r0 -- which means that the machine will actually try to 
store the value at address 0x0.

Is this a bug in GCC or is it my code that is messed up?

-- Stefan

Re: Bug in PPC inline assembly?

[David Edelsohn]

>>>>> Stefan  writes:

Stefan> I have some problems with using inline PowerPC assembly in GCC (4.0.1). 
Stefan> Consider the following code:

Stefan> void save_fp_register(double* buffer)
Stefan> {
Stefan> asm("stfd F0,  0(%0)" : : "r" (buffer) );
Stefan> }

	Use constraint "b".

David

Re: Bug in PPC inline assembly?

[Alan Lehotsky]

On Jul 17, 2005, at 19:15, Stefan wrote:

> I have some problems with using inline PowerPC assembly in GCC 
> (4.0.1). Consider the following code:
>
>    void save_fp_register(double* buffer)
>    {
>        asm("stfd F0,  0(%0)" : : "r" (buffer) );
>    }
>
Try using 'b' for the constraint - that selects for an "address base 
register", as opposed to 'r'
that is any of the general registers (including R0)

Re: Bug in PPC inline assembly?

[Stefan]

Thanks for your help: using constraint "b" instead of "r" solved my problem.

-- Stefan


 > Try using 'b' for the constraint - that selects for an "address base
 > register", as opposed to 'r' that is any of the general registers 
(including
 > R0)

 >> I have some problems with using inline PowerPC assembly in GCC (4.0.1).
 >> Consider the following code:
 >>
 >>    void save_fp_register(double* buffer)
 >>    {
 >>        asm("stfd F0,  0(%0)" : : "r" (buffer) );
 >>    }
 >>
 >> When compiled with -Os the generated code becomes
 >>
 >>    save_fp_register:
 >>        stfd F0,  0(3)
 >>        blr
 >>
 >> Which is OK; the input parameter comes in r3 which is used directly 
in the
 >> stfd operation. However, when compiled with -O0 (no optimization) the
 >> generated code instead becomes
 >>
 >>    save_fp_register:
 >>        stwu 1,-24(1)
 >>        stw 31,20(1)
 >>        mr 31,1
 >>        stw 3,8(31)
 >>        lwz 0,8(31)
 >>        stfd F0,  0(0) <-- error
 >>        lwz 11,0(1)
 >>        lwz 31,-4(11)
 >>        mr 1,11
 >>        blr
 >>
 >> The input parameter r3 is transferred to r0, which is used in the stfd
 >> operation. However, for this particular operation r0 means 0 -- not 
the value
 >> stored in r0 -- which means that the machine will actually try to 
store the
 >> value at address 0x0.
 >>
 >> Is this a bug in GCC or is it my code that is messed up?