It is 2^32-1 right in the x-dimension? I am using 1D blocks and threads.
I have a problem with the kernel kicking me out with just a million of blocks…
It is weird and I have a GTX-Titan. The kernel is very light. I simply do not know…
Never mind I am moving to a two dimensional block…Even if I find a solution the application will not work on a Fermi GPU. So I am breaking things…
Its so interesting designing algorithms that push things to the limits…
There was no problem with the number of blocks…but with memory issues…
Limitations are there to make us think more efficient into doing things. I am concerned when some limitations are removed whether we will not think efficiently but lazily (lazily in a sense that we do not optimize)…I am very pleased that memory limitations prohibited me to do what I initially thought and made me think of an efficient plan, it greatly added to my experience.
This comes in accordance with a recent paper I made with my partners. If we had infinite computing power, bandwidth, memory we wouldn’t trouble. These limitations made us think of efficient ways to cope with a beautiful theory and optimized its computing, bandwidth, memory to the maximum extent we could.