Indigo 1.1.14

[OnoSendai]

The big change here is the introduction of the BVH (Bounding volume hierarchy) to replace the BIH. The BVH is generally a lot faster in terms of rendering speed than the BIH (> 80% the speed of the kd-tree), while preserving the same low memory usage and build times of the BIH.
The BVH still builds low quality trees in some circumstances, resulting in low rendering performance , such as the Cornell box scene, but generally it's pretty fast. This will hopefully be fixed in future.

Let me know any performance results, especially comparing the BVH to kd-tree on different scenes.

Switching between the BVH and kd-tree will be done with the bih_tri_threshold setting. (Which is now a misnomer)


32 bit:
http://www.indigorenderer.com/dist/indigo_v1.1.14.zip

64 bit:
http://www.indigorenderer.com/dist/indi ... 1.1.14.zip

changelog:
1.1.14
* fixed fmod linking in ISL
* added OpenMP Dll + manifest to distributions
* replaced BIH with BVH

[Whaat]

Thanks ono!

[Marcofly]

thank you Ono! rendertime speed improvement sounds very interesting!!

[Zom-B]

Thanks Ono.... speedups are always very welcome

***EDIT***
OK, now I get it... so Heavy Poly (displacement+subdivision) will benefit from this...
Any numbers how much BVH is faster than BIH for rendering?!

[PureSpider]

<@PureSpider> will only the build speed be faster or render speed in general?
<+CTZn> both if I understood ok
<@OnoSendai> of what vs what?
<@PureSpider> BVH vs "normal" kd
<@OnoSendai> build is faster, render is slower
<@PureSpider> eeeeeh :@
<+CTZn> ah, necessarily slower ?
<@OnoSendai> yes
<+CTZn> ok, good to know
<@OnoSendai> ok, in some cases it may be faster
...
<@OnoSendai> render speed: kdtree > BVH > BIH
<@OnoSendai> mem usage: kdtree > BVH = BIH

Hope that got things clear

[Borgleader]

Should the exporters for 1.1.13 work with this build?

[Camox]

Thanks Ono !

[CTZn]

Expectedly 100% borgleader

[Pinko5]

wowwww tnx Master!!!
Luca.

[Zom-B]

Any numbers how much BVH is faster than BIH for rendering?!

My small 5Min Benchmark with some fancy displacement & subdivision was 2.4% faster for BVH than BIH... but BVH consumed about 7% more RAM (1.384GB vs 1.290 GB)

... so far my number crunching test.

[Big Fan]

what happened to the clay material experiment?
postponed or scrapped?

[vmuriel]

oooooooooh, again no linux release.

[fused]

@ono: couldnt find out what the uv set name was, so i rendered it with SSS. im not even sure if there is any...

doubt this is of interest for you, but i find 60s for 28million poly very exciting

Building Mesh 'lucy.ply'...
14027870 vertices (321.072 MB)
28055737 triangles (749.169 MB)
BVH::build()
Calcing root AABB.
Done.
Root AABB min: (225.841080,-388.417145,-605.892822)
Root AABB max: (1155.670166,145.354340,991.145996)
Sorting...
Done (8.44906 s).
Allocating intersect triangles...
Done. Intersect_tris mem usage: 963.217 MB
Build Stats:
Total nodes used: 9408392 (574.243 MB)
Total tri indices used: 37624372 (143.526 MB)
Num sub threshold leaves: 9408272
Num max depth leaves: 120
Num cheaper no-split leaves: 0
Mean tris per leaf: 2.981991
Max tris per leaf: 5
Mean leaf depth: 23.645914
Max leaf depth: 63
Finished building tree.
Done Building Mesh 'lucy.ply'. (Time taken: 60.20800 s)

[delle]

Thank you Ono.

I wait for a new Indigo release as a child waits for Christmas...

Delle

[filippo]

thaks ono!!!
filippo