snganRT's channel
Loading...
5
subscribers
5,526
video views
-
snganRT
uploaded
4 months ago
super fast bvh builder on dp sandy bridge
here we have a partly sah bvh builder based on intels avx vector units
with a numa aware virtual memory sorting. it allows to build up bvh trees w...
About snganRT's channel
more
less
node parallel bvh builder
the boeing model built with 4 nodes (8 cpus) coupled thru gpi.
total build time: 2.94s ! the complete bvh node structure is available on each node ...
333 views
snganRT said:
@SuperGastrocnemius
The current status of our work is:
For small scenes like the fairy (<250k polygons, everything fits into L3) it is around 2x faster than hlbvh2.
For midrange scenes like the buddha/dragon (1m polygons) it is 5-15% behind. Large scenes like the boeing are a "no go" for gpu/gp...
pbvh.ogg
Streaming parallel BVH construction on a dual intel X5650 (2.66gHz). Top levels of the dragon dataset (871k triangles) are build by using a space f...
569 views
snganRT said:
@davemc0 ok, that is pretty fast. i am curios to know what the hlbvh2 really implements and for shure the results. btw have u seen my new node parallel builder constructing the boeing model ?
pbvh.ogg
Streaming parallel BVH construction on a dual intel X5650 (2.66gHz). Top levels of the dragon dataset (871k triangles) are build by using a space f...
569 views
snganRT said:
@davemc0 i am not allowed to publish results from a new dual socket sandy-bridge machine but believe me: 40-70% will not be enough. so i think that even this hlbvh2 thingy will be easily beaten by this algo.
pbvh.ogg
Streaming parallel BVH construction on a dual intel X5650 (2.66gHz). Top levels of the dragon dataset (871k triangles) are build by using a space f...
569 views
snganRT said:
oohhh youtube !!! sorry davemc0
>>Hey, this looks really good for a CPU. Nice job!
So if this is three times faster than Pantoleoni's HPG 2010 paper that would make it 40% to 70% slower than Garanzha's HPG 2011 HLBVH2 paper. And another 2X slower when comparing a single CPU to a single GPU.
So...
Realtime AO
AO pass using a new BVH multi frustum traversal
and ray pre-sorting. The algorithm is able to
produce HQ ambient occlusion effects in realtime
on m...
1,128 views
snganRT said:
@rouncer81 ao is special case of hdr lightning (f.e. a really cloudy day around noon => unique distribution
of rays over the hemisphere).by using real hdri the hemisphere gets splitted into a number of light-energy hotspots (e.g. 64-1024) => higher ray coherency for a shading point. so today it i...
+ 2 more videosshow less