Except, it is legal to operate on non-type template parameters which, incidentally, may contain floating points (and that's exactly what metafloat does).
I'm curious, what are you doing that requires a floating-point compile-time constant?
To use them as non-type template parameters? Which is, of course, a wider problem.
Fear not, for there is *cough* http://ompf.org/stuff/metafloat/ *cough*
You're neglecting recent progress in the field; as stated in the FTA for traversal it's all about coherence. Note that coherent traversal (read SIMD friendly) has been transposed to other hierarchies, grid, bvh, bih... you name it.
Scan-line will remain mainstream as long as it will be the only method with a cheap specialised hardware implementation in town, even if objectively it doesn't make much sense.
Slashdot Mirror
That's quite sad indeed.
But i wonder, are you going to switch to the metric system this century or wait a few more?
Except, it is legal to operate on non-type template parameters which, incidentally, may contain floating points (and that's exactly what metafloat does).
I'm curious, what are you doing that requires a floating-point compile-time constant?
To use them as non-type template parameters? Which is, of course, a wider problem. Fear not, for there is *cough* http://ompf.org/stuff/metafloat/ *cough*You're neglecting recent progress in the field; as stated in the FTA for traversal it's all about coherence. Note that coherent traversal (read SIMD friendly) has been transposed to other hierarchies, grid, bvh, bih... you name it.
http://www.mpi-sb.mpg.de/~wald/PhD/wald_phd.pdfftp://download.intel.com/technology/computing/app
Etc...
Scan-line will remain mainstream as long as it will be the only method with a cheap specialised hardware implementation in town, even if objectively it doesn't make much sense.