ASSIGNMENT 7 Histogram operation

CMPS 297S/396AA: GPU COMPUTING
ASSIGNMENT 7
In this assignment, you will implement a histogram operation using atomic operations, and optimize it
using privatization, shared memory, and thread coarsening.
Instructions
1. Place the files provided with this assignment in a single directory. The files are:
 main.cu: contains setup and sequential code
 kernel.cu: where you will implement your code (you should only modify this file)
 common.h: for shared declarations across main.cu and kernel.cu
 timer.h: to assist with timing
 Makefile: used for compilation
2. Edit kernel.cu where TODO is indicated as follows:
 Histogram with privatization and shared memory only:
o Implement the kernel (histogram_private_kernel):
 Declare a private copy of the histogram in shared memory and initialize
it to 0
 Have each thread load a single image pixel and atomically update the
corresponding histogram bin count in shared memory
 Commit the non-zero bin counts to the global copy of the histogram in
parallel
o Implement the host code (histogram_gpu_private):
 Launch the grid (Note: the image has already been copied to global
memory for you and the bins have already been initialized to 0)
 Histogram with privatization, shared memory, and thread coarsening:
o Implement the kernel (histogram_private_coarse_kernel):
 Similar to the previous implementation, but each thread loads multiple
image pixels based on a coarsening factor (make sure the loads are
coalesced)
o Implement the host code (histogram_gpu_private_coarse):
 Similar to the previous implementation, but remember to take the
coarsening factor into consideration when selecting the number of
thread blocks in the grid
3. Compile your code by running: make
4. Test your code by running: ./histogram
 If you are using the HPC cluster, do not forget to use the submission system. Do not run
on the head node!
 For testing on different input sizes, you can provide your own values for the input
dimensions as follows: ./histogram <height> <width>
Submission
Submit your modified kernel.cu file via Moodle by the due date. Do not submit any other files or
compressed folders.