Christian Gmeiner<p>Funny… I'm investigating some wide_line CTS failures, and for some unknown reason, the blob driver does a second draw with line width 1 after the normal wide line draw. etnaviv does not know about the existence/need of this special draw, which causes the CTS tests to fail.</p><p>Starting to question my life choices 🙃 </p><p><a href="https://fosstodon.org/tags/etnaviv" class="mention hashtag" rel="tag">#<span>etnaviv</span></a> <a href="https://fosstodon.org/tags/mesa3d" class="mention hashtag" rel="tag">#<span>mesa3d</span></a> <a href="https://fosstodon.org/tags/librem5" class="mention hashtag" rel="tag">#<span>librem5</span></a> <a href="https://fosstodon.org/tags/mntreform" class="mention hashtag" rel="tag">#<span>mntreform</span></a></p>
fosstodon.org is one of the many independent Mastodon servers you can use to participate in the fediverse.
Fosstodon is an invite only Mastodon instance that is open to those who are interested in technology; particularly free & open source software.
If you wish to join, contact us for an invite.
Administered by:
Server stats:
8.8Kactive users
fosstodon.org: About · Status · Profiles directory · Privacy policy
Mastodon: About · Get the app · Keyboard shortcuts · View source code · v4.4.2
#etnaviv
5 posts · 2 participants · 0 posts today
Christian Gmeiner<p>dEQP-GLES3.functional.multisample.*_alpha_to_coverage</p><p>Test run totals:<br /> Passed: 4/8 (50.0%)<br /> Failed: 0/8 (0.0%)<br /> Not supported: 4/8 (50.0%)<br /> Warnings: 0/8 (0.0%)<br /> Waived: 0/8 (0.0%)</p><p><a href="https://fosstodon.org/tags/etnaviv" class="mention hashtag" rel="tag">#<span>etnaviv</span></a> <a href="https://fosstodon.org/tags/mesa3d" class="mention hashtag" rel="tag">#<span>mesa3d</span></a> <a href="https://fosstodon.org/tags/librem5" class="mention hashtag" rel="tag">#<span>librem5</span></a> <a href="https://fosstodon.org/tags/mntreform" class="mention hashtag" rel="tag">#<span>mntreform</span></a></p>
Christian Gmeiner<p>Today's challenge: Fix failing dEQP-GLES3.functional.multisample.* tests.</p><p><a href="https://fosstodon.org/tags/etnaviv" class="mention hashtag" rel="tag">#<span>etnaviv</span></a> <a href="https://fosstodon.org/tags/mesa3d" class="mention hashtag" rel="tag">#<span>mesa3d</span></a> <a href="https://fosstodon.org/tags/librem5" class="mention hashtag" rel="tag">#<span>librem5</span></a> <a href="https://fosstodon.org/tags/mntreform" class="mention hashtag" rel="tag">#<span>mntreform</span></a></p>
Christian Gmeiner<p>If anyone’s interested: <a href="https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36303" target="_blank" rel="nofollow noopener" translate="no"><span class="invisible">https://</span><span class="ellipsis">gitlab.freedesktop.org/mesa/me</span><span class="invisible">sa/-/merge_requests/36303</span></a></p><p>Off to bed now.</p><p><a href="https://fosstodon.org/tags/etnaviv" class="mention hashtag" rel="tag">#<span>etnaviv</span></a> <a href="https://fosstodon.org/tags/mesa3d" class="mention hashtag" rel="tag">#<span>mesa3d</span></a> <a href="https://fosstodon.org/tags/librem5" class="mention hashtag" rel="tag">#<span>librem5</span></a> <a href="https://fosstodon.org/tags/mntreform" class="mention hashtag" rel="tag">#<span>mntreform</span></a></p>
Christian Gmeiner<p>Taking a break helped - it was a lot easier than I thought. 🙃</p><p><a href="https://fosstodon.org/tags/etnaviv" class="mention hashtag" rel="tag">#<span>etnaviv</span></a> <a href="https://fosstodon.org/tags/mesa3d" class="mention hashtag" rel="tag">#<span>mesa3d</span></a></p>
Christian Gmeiner<p>RE'ing fixed-point number representations with weird negative value encoding takes way more time than you'd expect.</p><p><a href="https://fosstodon.org/tags/etnaviv" class="mention hashtag" rel="tag">#<span>etnaviv</span></a> <a href="https://fosstodon.org/tags/mesa3d" class="mention hashtag" rel="tag">#<span>mesa3d</span></a></p>
Christian Gmeiner<p>I'm currently in the process of cleaning up all the hacks I used to upstream this as quickly as possible. Along the way, I also fixed some issues with 3d, cube, and 2d_array support 😎</p><p>dEQP-GLES3.functional.texture.format.sized.*.rgba32*</p><p>Test run totals:<br /> Passed: 24/24 (100.0%)<br /> Failed: 0/24 (0.0%)<br /> Not supported: 0/24 (0.0%)<br /> Warnings: 0/24 (0.0%)<br /> Waived: 0/24 (0.0%)</p><p><a href="https://fosstodon.org/tags/etnaviv" class="mention hashtag" rel="tag">#<span>etnaviv</span></a> <a href="https://fosstodon.org/tags/mesa" class="mention hashtag" rel="tag">#<span>mesa</span></a> <a href="https://fosstodon.org/tags/Librem5" class="mention hashtag" rel="tag">#<span>Librem5</span></a></p>
Christian Gmeiner<p>With some more NIR shader magic and a bit of framebuffer state trickery, I can now render to 128-bit formats by using two render targets in parallel, each writing 64 bits. But that's enough hacking for today - time for bed.</p><p><a href="https://fosstodon.org/tags/etnaviv" class="mention hashtag" rel="tag">#<span>etnaviv</span></a> <a href="https://fosstodon.org/tags/mesa" class="mention hashtag" rel="tag">#<span>mesa</span></a> <a href="https://fosstodon.org/tags/Librem5" class="mention hashtag" rel="tag">#<span>Librem5</span></a></p>
Christian Gmeiner<p>Sweet! Using some NIR shader magic, I can also sample 128-bit formats. Under the hood, I'm sampling two separate 64-bit values and combining them within the shader.</p><p><a href="https://fosstodon.org/tags/etnaviv" class="mention hashtag" rel="tag">#<span>etnaviv</span></a> <a href="https://fosstodon.org/tags/mesa" class="mention hashtag" rel="tag">#<span>mesa</span></a> <a href="https://fosstodon.org/tags/Librem5" class="mention hashtag" rel="tag">#<span>Librem5</span></a></p>
Christian Gmeiner<p>My 128bit format emulation - needed for GLES3 - is starting to pass the first CTS.</p><p><a href="https://fosstodon.org/tags/etnaviv" class="mention hashtag" rel="tag">#<span>etnaviv</span></a> <a href="https://fosstodon.org/tags/mesa" class="mention hashtag" rel="tag">#<span>mesa</span></a> <a href="https://fosstodon.org/tags/Librem5" class="mention hashtag" rel="tag">#<span>Librem5</span></a></p>
Christian Gmeiner<p>I’m seeing every possible tiling pattern in my CTS result image 🙃 Time for a ☕ while I figure out what I’m doing wrong.</p><p><a href="https://fosstodon.org/tags/etnaviv" class="mention hashtag" rel="tag">#<span>etnaviv</span></a> <a href="https://fosstodon.org/tags/mesa" class="mention hashtag" rel="tag">#<span>mesa</span></a> <a href="https://fosstodon.org/tags/Librem5" class="mention hashtag" rel="tag">#<span>Librem5</span></a></p>
Christian Gmeiner<p>Rebuilding deqp to be faster finding the relevant sources of a failing test.</p><p><a href="https://github.com/KhronosGroup/VK-GL-CTS/commit/f805e24209ef7e25346f64aead0523b3027d9a14" target="_blank" rel="nofollow noopener" translate="no"><span class="invisible">https://</span><span class="ellipsis">github.com/KhronosGroup/VK-GL-</span><span class="invisible">CTS/commit/f805e24209ef7e25346f64aead0523b3027d9a14</span></a></p><p>Thanks <span class="h-card" translate="no"><a href="https://mastodon.online/@rg3" class="u-url mention">@<span>rg3</span></a></span></p><p><a href="https://fosstodon.org/tags/etnaviv" class="mention hashtag" rel="tag">#<span>etnaviv</span></a> <a href="https://fosstodon.org/tags/mesa" class="mention hashtag" rel="tag">#<span>mesa</span></a></p>
Christian Gmeiner<p>Found some unknown shader instructions and GPU states that need reversing... couldn't resist diving in 😅 </p><p>0x01105001, /* [01100] GS.UNK01100 := 0x1105001 */<br />0x00000005, /* [01104] GS.UNK01104 := 0x5 */<br />0x00010001, /* [01108] GS.UNK01108 := 0x10001 */<br />0x00000000, /* [01120] GS.UNK01120[0] := 0x0 */<br />0x00000000, /* [01124] GS.UNK01120[1] := 0x0 */<br />0x00000000, /* [01128] GS.UNK01120[2] := 0x0 */<br />...</p><p><a href="https://fosstodon.org/tags/etnaviv" class="mention hashtag" rel="tag">#<span>etnaviv</span></a> <a href="https://fosstodon.org/tags/mesa" class="mention hashtag" rel="tag">#<span>mesa</span></a></p>
Christian Gmeiner<p>Last night, I spent several hours debugging an issue with some CTS tests for a new feature. I couldn't pinpoint the problem and ended up staring at related GPU states until I eventually gave up for the night.</p><p>The next day, I took another look at my code - and within moments, I spotted the issue, fixed it, and the CTS passed without complaint.</p><p>dEQP-GLES31.functional.texture.gather.offset.*rgba*</p><p>Test run totals:<br /> Passed: 136/136 (100.0%)</p><p><a href="https://fosstodon.org/tags/etnaviv" class="mention hashtag" rel="tag">#<span>etnaviv</span></a> <a href="https://fosstodon.org/tags/mesa" class="mention hashtag" rel="tag">#<span>mesa</span></a> <a href="https://fosstodon.org/tags/Librem5" class="mention hashtag" rel="tag">#<span>Librem5</span></a></p>
Christian Gmeiner<p>I’ve made progress with hardware-based transform feedback. 😀 </p><p>dEQP-GLES3.functional.transform_feedback.*</p><p>Test run totals:<br /> Passed: 970/1320 (73.5%)<br /> Failed: 242/1320 (18.3%)<br /> Not supported: 108/1320 (8.2%)<br /> Warnings: 0/1320 (0.0%)<br /> Waived: 0/1320 (0.0%)</p><p><a href="https://fosstodon.org/tags/etnaviv" class="mention hashtag" rel="tag">#<span>etnaviv</span></a> <a href="https://fosstodon.org/tags/mesa" class="mention hashtag" rel="tag">#<span>mesa</span></a></p>
Christian Gmeiner<p>What can I say? All the texture offset lowerings are working — time to clean things up and open a merge request.</p><p>One step closer to GLES3 🥳 </p><p>dEQP-GLES3.functional.shaders.texture_functions.*</p><p>Test run totals:<br /> Passed: 609/609 (100.0%)<br /> Failed: 0/609 (0.0%)<br /> Not supported: 0/609 (0.0%)<br /> Warnings: 0/609 (0.0%)<br /> Waived: 0/609 (0.0%)</p><p><a href="https://fosstodon.org/tags/etnaviv" class="mention hashtag" rel="tag">#<span>etnaviv</span></a> <a href="https://fosstodon.org/tags/mesa" class="mention hashtag" rel="tag">#<span>mesa</span></a></p>
Christian Gmeiner<p>I took a short break from working on texture offset NIR lowerings to explore hardware-based transform feedback. The reverse engineering isn't fully complete yet, but there's already some progress.</p><p>dEQP-GLES3.functional.transform_feedback.position.points_*</p><p>Test run totals:<br /> Passed: 2/2 (100.0%)<br /> Failed: 0/2 (0.0%)<br /> Not supported: 0/2 (0.0%)<br /> Warnings: 0/2 (0.0%)<br /> Waived: 0/2 (0.0%)</p><p><a href="https://fosstodon.org/tags/etnaviv" class="mention hashtag" rel="tag">#<span>etnaviv</span></a> <a href="https://fosstodon.org/tags/mesa" class="mention hashtag" rel="tag">#<span>mesa</span></a></p>
Christian Gmeiner<p>Having fun writing texture offset lowerings in NIR - another step toward full GLES 3 support.</p><p><a href="https://fosstodon.org/tags/etnaviv" class="mention hashtag" rel="tag">#<span>etnaviv</span></a> <a href="https://fosstodon.org/tags/mesa" class="mention hashtag" rel="tag">#<span>mesa</span></a></p>
Christian Gmeiner<p>dEQP-GLES3.functional.shaders.texture_functions.texelfetch.*</p><p>Passed: 24/24 (100.0%)</p><p><a href="https://fosstodon.org/tags/etnaviv" class="mention hashtag" rel="tag">#<span>etnaviv</span></a> <a href="https://fosstodon.org/tags/mesa" class="mention hashtag" rel="tag">#<span>mesa</span></a></p>
Christian Gmeiner<p>🎉 After years of persistence, multi-planar YUV support has finally landed in etnaviv! Initially faced pushback, but now it's merged - bringing significant improvements to video playback performance. A big win for smoother multimedia experiences! 🚀</p><p>Merged MR: <a href="https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3418" target="_blank" rel="nofollow noopener" translate="no"><span class="invisible">https://</span><span class="ellipsis">gitlab.freedesktop.org/mesa/me</span><span class="invisible">sa/-/merge_requests/3418</span></a></p><p><a href="https://fosstodon.org/tags/etnaviv" class="mention hashtag" rel="tag">#<span>etnaviv</span></a> <a href="https://fosstodon.org/tags/mesa" class="mention hashtag" rel="tag">#<span>mesa</span></a></p>
TrendingLive feeds
Mastodon is the best way to keep up with what's happening.
Follow anyone across the fediverse and see it all in chronological order. No algorithms, ads, or clickbait in sight.
LoginDrag & drop to upload