ompf2.com

The relative speed of floating point square root computations has improved a lot since FPU hardware was first integrated into processor cores. Nowadays, if you can tolerate the occasional imprecision in the least one or two mantissa bits, single precision floating point square roots are fully pipeli...

asyrov wrote:My next challenge is solid constructions.

My code above is not a good reference for constructive solid geometry. It does have a rather inefficient implementation; but that would only make sense in conjunction with spatial subdivision (where the voxels contained only pruned CSG subtrees).

The explanation behind general quadrics is not so simple and obvious. It does get quite tedious doing the math with 10 quadric coefficients and 12 more coefficients of a general affine transform. If you feel like digging through the source code of another toy raytracer, I can offer an older experime...

I gave golden ratio sampling along a Z-curve a try. It does cover the unit square nicely, but the sample spacing is less regular than true 2D low discrepancy sequences. I suspect that is due to the discontinuities of the Z-curve; the example images with the Hilbert curve looked a bit better in that ...

I cannot back this up with a citation or even only with plausible reasoning. But my first experiment would be bit interleaving, i.e. Z-order, also known as Morton code. In other words, instead of the Hilbert-Curve, I'd try a cheaper mapping from the unit interval to the unit square. I feel that the ...

With absolutely no pride whatsoever, I hereby announce the release of source code. There isn't much in it for you, dear readers, but if you want to see for yourself, surf to http://www.vectorizer.org/boar/index.html and scroll down a bit to the blog section. There you will find a tarball and a few s...

Pardon the late reply. Finally got around to doing an illustration of surface smoothness, showing the range from piecewise constant to twice continuously differentiable. Lighting and texturing also have strong effects on perceived smoothness that are being shown. If you are interested see: http://ve...

Both processor types will continue to converge somewhat. CPUs will continue to evolve more powerful SIMD ALUs (though not necessarily wider ones), and GPUs will further improve in terms of ALU utilization for a larger set of less regular algorithms. They will never ever meet in a single point, becau...

Latest details about Knights Landing as reported by c't magazin: The modified cores have two parallel SIMD FPUs; theoretical peak throughput is two 512 bit wide fused multiply adds per clock, that means 64 single precision FLOPs or 32 double precision FLOPs per clock cycle. The issue width is still ...

Some more information has cropped up: http://vr-zone.com/articles/xeon-phi-knights-series-continues-landing-2015/64112.html Apparently Knights Landing will use a derivative of the Silvermont (Atom) core. The above report speaks about "two 512b vector units per core", but it is still unclear if that ...