Re: [arm, patch] Remove redundant promotion of function arguments

[Mark Mitchell <mark at codesourcery dot com>]

Paul Brook wrote:

> Tested on sparc-sun-solaris2.8/-m64 and cross-compiler to arm-none-elf, and by 
> manually inspecting the generated assembly for the examples given previously 
> in this thread.
>
> Ok?
>  
This is OK for the mainline, modulo Eric's comments about ChangeLog typos.

This hunk:

> enum machine_mode
> promote_mode (tree type, enum machine_mode mode, int *punsignedp,
> 	      int for_call ATTRIBUTE_UNUSED)
> @@ -806,17 +810,28 @@ promote_mode (tree type, enum machine_mo
>   enum tree_code code = TREE_CODE (type);
>   int unsignedp = *punsignedp;
>
> -#ifdef PROMOTE_FOR_CALL_ONLY
> +#ifndef PROMOTE_MODE
>   if (! for_call)
>     return mode;
> #endif
>
>   switch (code)
>     {
> -#ifdef PROMOTE_MODE
> +#ifdef PROMOTE_FUNCTION_MODE
>     case INTEGER_TYPE:   case ENUMERAL_TYPE:   case BOOLEAN_TYPE:
>     case CHAR_TYPE:      case REAL_TYPE:       case OFFSET_TYPE:
> -      PROMOTE_MODE (mode, unsignedp, type);
> +#ifdef PROMOTE_MODE
> +      if (for_call)
> +	{
> +#endif
> +	  PROMOTE_FUNCTION_MODE (mode, unsignedp, type);
> +#ifdef PROMOTE_MODE
> +	}
> +      else
> +	{
> +	  PROMOTE_MODE (mode, unsignedp, type);
> +	}
> +#endif
>       break;
> #endif
>  
made my brain hurt a little.  I was wondering if the first "#ifndef 
PROMOTE_MODE" should be "if !defined(PROMOTE_MODE) && 
!defined(PROMOTE_FUNCTION_MODE)" -- but I guess it's correct the way it is.

Thanks,

--
Mark Mitchell
CodeSourcery, LLC
(916) 791-8304
mark@codesourcery.com