<div dir="ltr"><div class="gmail_quote"><div dir="ltr">On Tue, Nov 27, 2018 at 9:17 PM Ilia Mirkin <<a href="mailto:imirkin@alum.mit.edu">imirkin@alum.mit.edu</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">On Tue, Nov 27, 2018 at 9:00 PM Marek Olšák <<a href="mailto:maraeo@gmail.com" target="_blank">maraeo@gmail.com</a>> wrote:<br>
><br>
> From: Marek Olšák <<a href="mailto:marek.olsak@amd.com" target="_blank">marek.olsak@amd.com</a>><br>
><br>
> For format fallbacks like ETC and ASTC, switching between sRGB and linear<br>
> decoding is undefined, or at least is not bit-exact. Same as<br>
> EXT_texture_sRGB_decode on GLES.<br>
><br>
> There are no piglit or dEQP regresssions.<br>
<br>
Just checking -- did you run dEQP on hardware that uses the ASTC<br>
fallbacks? If not, could you?<br></blockquote><div><br></div><div>radeonsi doesn't support ASTC, so I always use the fallback. Like I said, no regressions in newly enabled tests.<br></div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
Also, what are the conditions for enabling ASTC fallbacks? Basically<br>
the issue is that OES_texture_view requires ASTC, so I want to make<br>
sure it'll be enabled for any drivers that<br>
PIPE_CAP_SAMPLER_VIEW_TARGET is enabled for.<br></blockquote><div><br></div><div>ASTC is always enabled.</div><div><br></div><div>The existing behavior being talked about is already exposed by EXT_texture_sRGB_decode on GLES. The extension only adds another way to access that behavior. I'm open to discussing how to make EXT_texture_sRGB_decode bit-exact for compressed format fallbacks. I may also let it slide if test suites don't care.<br></div><div><br></div><div>Marek<br></div></div></div>