On Wed, Nov 28, 2018 at 10:02:22AM +0200, Jani Nikula wrote: > On Tue, 06 Nov 2018, Lucas De Marchi <lucas.demarchi@xxxxxxxxx> wrote: > > From: Rodrigo Vivi <rodrigo.vivi@xxxxxxxxx> > > > > RANGE makes it longer, but clear. We are also going to add a check for > > the display part, so make rename to GT. > > I also still have my doubts about this patch I'm afraid. I've expressed > the concern before, but here goes again. > > How much is the split of gen to GT gen and display gen going to help us > in the long run? The only current platform that would benefit from this > is GLK. However, not all IS_GEMINILAKE() can be replaced with > IS_DISPLAY_GEN() >= 10 or similar. We also have VLV/CHV display that is > better represented by HAS_GMCH_DISPLAY() and AFAICT can't usefully be > replaced by a display gen check. > > My main fear is that the split adds a lot of confusion. (Where should I > use GT gen, where should I use display gen, patches to change between > one and the other. It's not 100% clear cut.) > > Here too I wonder if we're better off adding more has_feature flags that > describe what gt or display features a platform has. It will never be a clear cut. Adding more flags make perfect sense when there is such a feature to check for. Sometimes registers just change location based on what is the display gen. Having the display gen serves to group small differences together into a "single flag" if you will. And today what we do is to actually check for gen rather than display gen. Or a mix of platform names. Lucas De Marchi _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
