This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] Fix small structure passing on x86-64


> Hi,
> 
> the structure
> 
> struct S { char c; char arr[4]; float f; };
> 
> is incorrect passed on x86-64/Linux with every C compiler I tried: only the 
> first 4 bytes and the float are passed (in registers), the 5th byte is lost.
> That's because the first word has partial integer class X86_64_INTEGERSI_CLASS 
> instead of full integer class X86_64_INTEGER_CLASS.
>  	    if (subclasses[0] == X86_64_SSESF_CLASS && bytes != 4)
>  	      subclasses[0] = X86_64_SSE_CLASS;
> -	    if (subclasses[0] == X86_64_INTEGERSI_CLASS && bytes != 4)
> +	    if (subclasses[0] == X86_64_INTEGERSI_CLASS
> +		&& !((bit_offset % 64) == 0 && bytes == 4))

the test here is still bit confused.  It should test if the whole array
fits in lower 4 bytes that would be somehting like
(bit_offset + 7) / 8 + bytes <= 4
There are very intersting problems related to this and reading past end
of the structure possibly causing segfault.  I am working on more
complette patch.

Honza
>  	      subclasses[0] = X86_64_INTEGER_CLASS;
>  
>  	    for (i = 0; i < words; i++)

> struct S { char c; char arr[4]; float f; };
> 
> char A[4] = { '1', '2', '3', '4' };
> 
> void foo (struct S s)
> {
>   if (__builtin_memcmp (s.arr, A, 4))
>     __builtin_abort ();
> }
> 
> int main (void)
> {
>   struct S s;
>   __builtin_memcpy (s.arr, A, 4);
>   foo (s);
>   return 0;
> }


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]