On Wed, 26 Oct 2011, Michel [ISO-8859-1] D?nzer wrote:
> On Die, 2011-10-25 at 22:40 -0400, Ilija Hadzic wrote: >> it's better that radeon_crtc_commit and radeon_crtc_prepare call >> crtc-specific dpms functions instead of hard-coding them to >> radeon_crtc_dpms. > > Is it really better? If it's always radeon_crtc_dpms anyway, this > obfuscates that fact (and introduces a guaranteed branch prediction > miss, but that probably doesn't matter here). > My reasoning for believing that it's better was that the functions are fairly generic: they loops through all CRTCs and disable/enable them all. Although at this time these functions are only used as helpers for legacy crtcs (so indeed the pointer always resolves to radeon_crtc_dpms), if they are ever called through some other path (in the context of atombios CRTC), it doesn't hurt to make them ready for that. Having said that, I don't really mind whatever the destiny of this patch ends up being. (and yes, branch prediction slot penalty really don't matter here; it's a slow path anyway) -- Ilija
