It's kind of weird to me that GPUs don't have planar textures now, I guess... maybe it shouldn't be weird to me, but it is?
@nothings Why is cache usage and bandwidth better if it is interleaved? You always have to fetch 16 texels at a time for the "warp"...
-
-
@cmuratori Take an unrealistic case to start with: those 16 texels are uncorrelated. Do you want to hit 16 cache lines or 64 cache lines? -
@cmuratori In practice they're highly correlated/overlapping, it's not 4x efficiency, it's 1.1x or something. - Show replies
New conversation -
-
-
@nothings I guess if you assume that you're always using all four most of the time, then yeah, you get less "overfill" in the cache. -
@cmuratori@nothings Some parts of texture units are built RGBA-wide, so using only a few channels wastes gates. - Show replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.