On Tue, Jan 15, 2013 at 3:49 PM, Ian Romanick <i...@freedesktop.org> wrote: > On 01/11/2013 08:54 PM, Eric Anholt wrote: >> >> The previous solution of transcoding to ARGB8888 gave exact results, but >> below the expected performance. Given that the user has gone out of >> their way to give us compressed data for performance reasons, try to >> respect their wishes by doing the best compression we can handle. We >> expect DXT1 compression to be better quality, but it may not always be >> available, in which case FXT1 seems like a reasonable fallback. >> >> This regresses OES_compressed_ETC1_RGB8_texture/miptree (which expects >> exact results, which we no longer provide), but it improves GLBenchmark >> 2.1 offscreen performance by 3.2% +/- 2.1% (n=11/10, throttling >> outliers removed). > > > There are ETC1 and ETC2 conformance tests. Do they pass?
In es3conform it regresses compressed_etc1_rgb8_texture.test and gl_compressed_rgb8_etc2. The first fails because of slight inaccuracies, eg (0.71, 0.698, 0.776) instead of (0.698, 0.698, 0.765). The second looks like it finds some potentially bad cases, eg (0.0 0.0 0.0 255.0) instead of (0.0 255.0 0.0 255.0). _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
