Re: [RFC/patch] macro_removal_agp_mtrr.diff

[Arjan van de Ven <arjanv@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 1588 bytes --]

On Sat, 2004-09-04 at 12:02, Dave Airlie wrote:
> This patch is the first in the drmfntbl-0-0-2-branch of macro removals
> 
> This dumps __REALLY_HAVE_AGP and __REALLY_HAVE_MTRR and replaces them with
> a bitmask of driver features that is checked...
> 
> I've had to introduce __OS_HAS_AGP and __OS_HAS_MTRR, on Linux these map
> simply to (CONFIG_AGP || CONFIG_AGP_MODULE) and CONFIG_MTRR, however I
> can't remove them completely as we need to build on architectures that
> don't have the AGP or MTRR interfces... but they are much cleaner than the
> old code...


it looks you can nuke most of those.
> -#if __REALLY_HAVE_AGP
> +#if __OS_HAS_AGP
>  	drm_agp_head_t    *agp;	/**< AGP data */
>  #endif

for example does this extra pointer really hurt in the *really* unlikely
case you don't have AGP ?

> +static __inline__ int drm_core_check_feature(struct drm_device *dev, int feature)
> +{
> +	return ((dev->driver_features & feature) ? 1 : 0);
> +}

if you make a drm_core_has_AGP and drm_core_has_MTRR wrapper around
these then... 


> +#if __OS_HAS_MTRR
> +		if (drm_core_check_feature(dev, DRIVER_USE_MTRR)) {
> +			if ( map->type == _DRM_FRAME_BUFFER ||
> +			     (map->flags & _DRM_WRITE_COMBINING) ) {
> +				map->mtrr = mtrr_add( map->offset, map->size,
> +						      MTRR_TYPE_WRCOMB, 1 );
> +			}
>  		}

...this would NOT need an ifdef because drm_core_has_MTRR would be a
define to 0 for non-OS_HAS_MTRR machines at which point the compiler
removes the entire code. Eg no ifdef needed and the code gets more
readable.




[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]