100 Days of CUDA development
Wrote CUDA code for 3 different operations; vector addition, matrix addition, matrix vector multiplication
I skimmed though the first 3 chapters(because I already have a little knowledge on CUDA), reading the basics of CUDA programming model, what threads, blocks and grids mean, how they are leveraged to create a workload inside a kernel.
I am unsure about the necessity about the limiting numbers for these aspects. For eg, 512 is alimit for the number fo threads in a block, WHY?
In Chapter 4, the author speaks about the dimensions of a kernel launch, denoting the number of blocks in a grid and the number of threads in a block