Re: Update doc/extend.texi regarding asm memory clobbers

[Andreas Jaeger <aj at suse dot de>]

Ok to commit for both mainline and 3.4?  Should this also go into 3.3?

Andreas

Andreas Jaeger <aj@suse.de> writes:

> Michael Matz <matz@suse.de> writes:
>
>> On Fri, 16 Jan 2004, Andreas Jaeger wrote:
>>
>>> +@example
>>> +@{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p @}) )@}.
>>                                                            ^^^
>> semicolon missing
>>
>>> +int foo (int *x)
>>> +@{
>>> +  int y = bar(x);
>>> +  int r;
>>> +  asm ("magic stuff accessing (%1)..."
>>> +        "=&d" (r) : "a" (y), "m" (*y));
>>> +  return r;
>>
>> 'y' must be a pointer to be dereferencable ;)
>
> Good idea ;-)
>
>> I suggest something like this:
>>
>> int foo ()
>> {
>>   int x = 42;
>>   int *y = &x;
>>   int result;
>>   asm ("stuff accessing an 'int' pointed to by '%1'"
>>        "=&r" (result) : "r" (y), "m" (*y));
>>   return result;
>> }
>>
>> And what could be optimized away is the store to x.
>
> Ok, adjusted.  What do you think of our ;-) patch now?
>
> Cheers,
> Andreas
>
> 2004-01-16  Andreas Jaeger  <aj@suse.de>
> 	    Michael Matz  <matz@suse.de>
>
> 	* doc/extend.texi (Extended Asm): Clarify memory clobber.
>
> ============================================================
> Index: gcc/doc/extend.texi
> --- gcc/doc/extend.texi	7 Jan 2004 22:24:39 -0000	1.177
> +++ gcc/doc/extend.texi	16 Jan 2004 17:04:23 -0000
> @@ -4006,13 +4006,35 @@ represents the condition codes as a spec
>  condition code is handled differently, and specifying @samp{cc} has no
>  effect.  But it is valid no matter what the machine.
>  
> -If your assembler instruction modifies memory in an unpredictable
> +If your assembler instructions access memory in an unpredictable
>  fashion, add @samp{memory} to the list of clobbered registers.  This
> -will cause GCC to not keep memory values cached in registers across
> -the assembler instruction.  You will also want to add the
> -@code{volatile} keyword if the memory affected is not listed in the
> -inputs or outputs of the @code{asm}, as the @samp{memory} clobber does
> -not count as a side-effect of the @code{asm}.
> +will cause GCC to not keep memory values cached in registers across the
> +assembler instruction and not optimize stores or loads to that memory.
> +You will also want to add the @code{volatile} keyword if the memory
> +affected is not listed in the inputs or outputs of the @code{asm}, as
> +the @samp{memory} clobber does not count as a side-effect of the
> +@code{asm}.  If you known how large the accessed memory is, you can add
> +it as input or output but if this is not known, you should add
> +@samp{memory}.  As an example, if you access ten bytes of a string, you
> +can use a memory input like:
> +
> +@example
> +@{"m"( (@{ struct @{ char x[10]; @} *p = (void *)ptr ; *p; @}) )@}.
> +@end example
> +
> +Note that in the following example the memory input is necessary,
> +otherwise GCC might optimize the store to @code{x} away:
> +@example
> +int foo ()
> +@{
> +  int x = 42;
> +  int *y = &x;
> +  int result;
> +  asm ("magic stuff accessing an 'int' pointed to by '%1'"
> +        "=&d" (r) : "a" (y), "m" (*y));
> +  return result;     
> +@}
> +@end example
>  
>  You can put multiple assembler instructions together in a single
>  @code{asm} template, separated by the characters normally used in assembly

Andreas
-- 
 Andreas Jaeger, aj@suse.de, http://www.suse.de/~aj
  SuSE Linux AG, Maxfeldstr. 5, 90409 Nürnberg, Germany
   GPG fingerprint = 93A3 365E CE47 B889 DF7F  FED1 389A 563C C272 A126

Attachment: pgp00000.pgp
Description: PGP signature