this is for holding javascript data
James Shirley edited cuda_overview.tex
about 11 years ago
Commit id: 25730f4803265232022273012e1944c73e1d9960
deletions | additions
diff --git a/cuda_overview.tex b/cuda_overview.tex
index c4ead01..7f810a0 100644
--- a/cuda_overview.tex
+++ b/cuda_overview.tex
...
\subsection{CUDA Thread Model}
CUDA kernels are sets of instructions to be executed in parallel by a batch of threads. This batch of threads can exist as a one dimensional array, a two dimensional array or a three dimensional array. The threads can be further subdivided by splitting them into multiple
blocks thread-blocks however blocks can only exist as either part of a one or two dimensional array. The entirety of all threads doing work on a problem is called the grid, in other words, the grid is an (multi-dimensional) array of
blocks thread-blocks which are also each a (multi-dimensional) array of threads. Each of these threads traverse the exact same code however their behavior can be influenced using each threads identification data. This identification data consists of their thread index and their block index.
\subsection{CUDA Memory Model}