Graphics Processing Units (GPUs):
Architecture and ProgrammingMohamed Zahran (aka Z)mzahran@cs.nyu.eduhttp://www.mzahran.comCSCI-GA.3033-004CUDA
Advanced Techniques 1Some Refreshing Exercises!• Suppose registers and shared memory
capacities were not an issue. When is it
still beneficial to put values fetched
from memory into the shared memory?Some Refreshing Exercises!• Assume a kernel is launched with 1000
blocks. Each block has 512 threads.
– If a variable is declared as local in thekern
