https://wiki.blender.jp/index.php?action=history&feed=atom&title=%E5%88%A9%E7%94%A8%E8%80%85%3ADingTo%2FGSoC_2014%2FWeekly_Reports%2FWeek5利用者:DingTo/GSoC 2014/Weekly Reports/Week5 - 版の履歴2026-07-31T02:10:15Zこのウィキのこのページに関する変更履歴MediaWiki 1.31.0https://wiki.blender.jp/index.php?title=%E5%88%A9%E7%94%A8%E8%80%85:DingTo/GSoC_2014/Weekly_Reports/Week5&diff=147361&oldid=prevYamyam: 1版 をインポートしました2018-06-28T21:08:49Z<p>1版 をインポートしました</p>
<table class="diff diff-contentalign-left" data-mw="interface">
<tr class="diff-title" lang="ja">
<td colspan="1" style="background-color: #fff; color: #222; text-align: center;">← 古い版</td>
<td colspan="1" style="background-color: #fff; color: #222; text-align: center;">2018年6月28日 (木) 21:08時点における版</td>
</tr><tr><td colspan="2" class="diff-notice" lang="ja"><div class="mw-diff-empty">(相違点なし)</div>
</td></tr></table>Yamyamhttps://wiki.blender.jp/index.php?title=%E5%88%A9%E7%94%A8%E8%80%85:DingTo/GSoC_2014/Weekly_Reports/Week5&diff=147360&oldid=prev2014年6月20日 (金) 14:40にwiki>DingToによる2014-06-20T14:40:52Z<p></p>
<p><b>新規ページ</b></p><div>= Week 5 =<br />
<br />
== What I did this week ==<br />
* Merged master into my branch (took some time, due to several conflicts) and updated QBVH Code in my Branch with the new SSE classes. ({{GitCommit|0e62c026ab3d}}, {{GitCommit|611d69adb98b}})<br />
<br />
* Did some research and experiments on improved QBVH / Triangle Intersection again. https://developer.blender.org/D599 I think the "Watertight Ray/Triangle Intersection" paper + the Embree sample implementation of that is a very nice candidate, but probably a bit over my head for now. It involves changes to BVH build, BVH traversal and Triangle Intersection. Nevertheless would it be a great addition.<br />
** Memory usage would be reduced (no more __tri_woop data, which would save us 4x float4 per triangle). <br />
** Improved performance due to Quad BVH + multiple triangle intersections with SIMD.<br />
** Also not sure how feasible that is for GPU. The algorithm itself does not use CPU only features as far as I can see that, but might need additional work. <br />
<br />
* Started to look into ways to avoid unneeded samples for black backgrounds. http://www.pasteall.org/52361/diff I quickly hacked it in for the case, where "Use Nodes" is still disabled for the world. In the final implementation, I need to check the world shader tree, and extract the Background node, to see if the color is black or the strength zero. I am not exactly sure yet, how to do that. Help here is welcome. <br />
<br />
* Today I spent time on the bugtracker.<br />
** Fixed: Issue with Checker Texture, GLSL error with Hair shader: ({{GitCommit|ead83a48f15c}}, {{GitCommit|a5a869c579de}})<br />
** Closed several other reports (Issues has been fixed, or not a bug). <br />
<br />
== Next week ==<br />
I will continue to work on the Background Sampling optimization or look into the Smooth Normals (as planned last week). <br />
<br />
University exams start in 2 weeks (7 July - 17 July), so I would like to focus on that.<br />
When I look at my proposal, half of the tasks are finished + WIP code for 2 more items is there. http://wiki.blender.org/index.php/User:DingTo/GSoC_2014/Proposal I also did additional work + fixes, that should be fine for Midterm I hope. :)<br />
I can work almost full time on Cycles from the 18th July to GSoC end (18th August), that should give me enough time to finish my Deliverables + some additional fun. But the upcoming 3-4 weeks will most likely be spend on University. Please tell me if that is ok. <br />
<br />
== Questions ==<br />
None.</div>wiki>DingTo