Figures out starting and ending positions for each CUDA thread in each block for use by a parallel sequential searching algorithm.

calculate_bounds.cpp

#include <string>
#include <string.h>

int **starts, **ends;
const int NUM_BLOCKS = 9, THREADS_PER_BLOCK = 10;
const std::string input = "assign4input1";

// figures out the start and end position for every thread and puts them in 2D arrays.
void calculateStarts(int textLength, int patternLength)
{
	int majorChunkSize = textLength / NUM_BLOCKS;
	int minorChunkSize = majorChunkSize / THREADS_PER_BLOCK;

	int chunkStart = 0;
	starts = new int*[NUM_BLOCKS];
	ends = new int*[NUM_BLOCKS];

	for (int i = 0; i<NUM_BLOCKS; i++)
	{
		// todo: this will be created on the heap... deallocate it later!
		starts[i] = new int[THREADS_PER_BLOCK];
		ends[i] = new int[THREADS_PER_BLOCK];

		chunkStart = i*majorChunkSize;

		for(int j=0; j<THREADS_PER_BLOCK; j++)
		{
			starts[i][j] = chunkStart + j*minorChunkSize;

			// sometimes ends will be > textLength... the matching algorithm will only check to textLength-1!
			ends[i][j] = starts[i][j] + minorChunkSize + patternLength - 1;
		}

		// remainders are added to the end of the last chunk
		int remainingForChunk = majorChunkSize % THREADS_PER_BLOCK;
		ends[i][THREADS_PER_BLOCK-1] += remainingForChunk;
	}

	int remainingForWhole = textLength%NUM_BLOCKS;
	ends[NUM_BLOCKS - 1][THREADS_PER_BLOCK - 1] += remainingForWhole;
}


int main()
{
	calculateStarts(1000, 20);

	return 0;
}