Eric Holk
This paper is about a DSL for ray tracing on GPUs: http://graphics.cs.williams.edu/papers/OptiXSIGGRAPH10/Parker10OptiX.pdf They use a trampolining transformation to allow recursion, which we'll probably be using before long. We should see...
I'd propose something like `range-kernel`, which might look like this: ``` (range-kernel (100 100) (+ 1 2)) ``` This would be a 2D kernel, with 100 threads in each direction,...
This was basically to let us inline `iota`, which we managed to do already. I'm closing the issue for now, and we can reopen it if it seems like we...
It's also possible to implement ropes as an ADT and some associated libraries, perhaps with a new kernel macro to enable easy kerneling over ropes. To make this really useful,...
We have a test case with some example syntax here: https://github.com/eholk/harlan/blob/master/test/async-kernel.kfc
This would be easy enough to add as a macro now. We may not get all the optimizations, but we're trending more towards an "optimize from first principles" approach, which...
mandelbrot.kfc is one of them.
I think the main thing that defeats optimization is having a let. We can inline obvious things like `(kernel ((i (kernel ...)) ...)`, but if it's `(let ((x (kernel ...)))...
I think that's the basic idea for what we wanted to do. There's just the question of whether we always want to say inline if a variable is bound to...