[Bf-committers] BIH progress update

[Yves Poissant]

From: "Joe Eagar"

> I thought Altivec and SSE both used the same memory layout structure 
> (though Altivec does have more SIMD registers then SSE)?  If I remember 
> right, they both require the same 16-byte alignment in memory as well, 
> though I could be wrong.

I'm not an altivec specialist but as far as I know, they do use the same 16 
byte alignment. You are right. But in order to be non architecture specific, 
the abstraction should not assume that every SIMD use the same memory layout 
structure right?

>> Look at post #4 for the c source file. It is cool indeed. I may try it 
>> just for my own education to see how well it can really accelerate the 
>> ray to scene intersection process.
>>
> I'd be interested to know how your tests work out.

I will post my results but I don't plan to try that until I have a nice 
really efficient BIH or BVH running.

>> If you are interested in SSE implementations, you should take a look at 
>> Arauna : http://igad.nhtv.nl/~bikker/ and look at his kdtree.c source 
>> code.
>>
>>
> I'll take a look, thanks.

His kdtree source code is a bit of a misnomer. It actually have code for 
traversing the kdtree or the BVH. One or the other are selectable. At one 
point, he had a BIH but he scrapped it and implemented a SAH BVH. 
Apparently, the BVH was 26% faster than the BIH. I still have the source 
with the BIH but as you will see, it is so heavily SSE that it is not 
usefull for my needs. It is still very interesting to look at though.

Yves