Optimized Parallel Tiled Approach to perform 2D Convolution by taking advantage of the lower latency, higher bandwidth shared memory as well as global constant memory cached aggresively within GPU thread blocks.
Matrix convolution is primarily used in image processing for tasks such as image enhancing, blurring, etc. A standard image convolution formula for a 5x5 convolution filter M with an Image N is:
P (i, j) = sum over a=0->4(sum over b=0->4(M [a][b] ∗ N [i + a − 2][j + b − 2])), where 0 ≤ i < N.height and 0 ≤ j < N.width
In this code, elements that are “outside” Matrix N are treated as if they had the value zero.
• No arguments: The application will create a randomized Filter M and Image N. A CPU implementation of the convolution algorithm will be used to generate a correct solution which will be compared with your program’s output. If it matches (within a certain tolerance), it will print out “Test PASSED” to the screen before exiting.
• One argument: The application will create a randomized Filter M and Image N, and write the device-computed output to the file specified by the argument.
• Three arguments: The application will read the filter and image from provided files. The first argument should be a file containing two integers representing the image height and width respectively. The second and third function arguments should be files which have exactly enough entries to fill the Filter M and Image N respectively. No output is written to file.
• Four arguments: The application will read its inputs using the files pro- vided by the first three arguments, and write its output to the file provided in the fourth.
The (optional) input files should have a single line containing whitespace- separated floating point numbers representing the matrix data. There should be m · n numbers on this line for a m × n matrix, where the first n numbers are the first row, the second n numbers are the second row, etc. For example, to represent the following matrix:
[ 1 2 3 ]
[ 4 5 6 ]
[ 7 8 9 ]
the corresonding input file should contain the following line (without quotes):
“1 2 3 4 5 6 7 8 9”
Note that if you wish to use the output file from one run of the application as an input in a later run, you must delete the first line in the output file, which displays the accuracy of the values within the file.