我正在使用MPI在C中编写并行图像处理算法。
将图像解析为2d数组，根据处理器进行切片并正确发送到每个进程！
奇怪的问题是，当我打印出数组时，我将最后8个数组元素作为0。
我尝试添加sliceSize * width + 8， 工作了！
但是当我尝试使用不同（更大）的尺寸时，0还有更多。
int main(int argc, char *argv[]) {
TGA_Header header;
char output_filename[100];
if (argc != 2) {
printf("\nUsage:\n");
printf("\t./tga <filename>.tga\n\n");
printf(
"The input file should be a greyscale image with an 8-bit depth.\n\n");
exit(0);
}
TGA *image = load_tga(argv[1], &header);
/*
* Image processing starts here
*/
int my_rank = 0, size = 0;
MPI_Init(&argc, &argv);
MPI_Comm_rank(MPI_COMM_WORLD, &my_rank);
MPI_Comm_size(MPI_COMM_WORLD, &size);
// Variable Initialization
int i = 0, k = 0, l = 0, j = 0, c = 0, d = 0, avg = 0, count = 0, sum = 0,
sliceSizeExt = 0, upper = 0, start = 0, end = 0, extrapixels = 0,
sliceSize = 0, height = 0, width = 0, **subArray, **subResult,
**subArrayFext, **subResultFext, **array, **result, **s, **r,
SLICEWIDTH = 0;
MPI_Status status;
MPI_Request request;
// Set Height and width!
width = image->w;
height = image->h;
// Dynamic allocation for both arrays
sliceSize = height / size;
SLICEWIDTH = sliceSize * width;
subArray = (int **)malloc(sizeof(int *) * sliceSize);
for (int i = 0; i < sliceSize; i++) {
subArray[i] = (int *)malloc(sizeof(int) * width);
}
r = s = malloc(sizeof(int *) * sliceSize);
for (i = 0; i < sliceSize; i++) {
s[i] = malloc(sizeof(int) * width);
}
/*
First and last arrays should be allocated with +1 line for the respected
ghost borders
Arrays for processes with rank 0 > my_rank > size should keep the same
allocation
Exception:
If the height/size modulo is (>) Greater than (0) Zeroprintf("\n Computing
bounds for process ")
}
Then it should be processed by the rank 0
Just to keep the communication dynamic.
*/
extrapixels = height % size;
sliceSizeExt = sliceSize + extrapixels;
if (my_rank == 0) {
array = malloc(sizeof(int *) * height);
for (i = 0; i < height; i++) {
array[i] = malloc(sizeof(int) * width);
}
printf(
"\nThese are the extra pixels! : %d\n This is the Ext dimension:%d\n",
extrapixels, sliceSizeExt);
subArrayFext = malloc(sizeof(int *) * sliceSizeExt);
for (i = 0; i < sliceSizeExt; i++) {
subArrayFext[i] = malloc(sizeof(int) * width);
}
printf("IM RANK %d , Array init!\n", my_rank);
printf("\n***%d : Image dimensions %d %d\n", my_rank, height, width);
printf("\n***********\n%d Some statistics: \nSlave Processes:%d\nArray "
"Slice Size(height): %d\nMoving on to Compute slice "
"bounds!\n************\n",
my_rank, size - 1, sliceSize);
printf("\n***%d Parsing pixel Array to 2d\n", my_rank);
// Parse image pixel array to 2d array
for (i = 0; i < height; i++) {
for (j = 0; j < width; j++) {
// use incremental counter for 1d array(: image->pixels[width*height])
// to 2d transformation
array[i][j] = image->pixels[upper];
upper++;
}
}
for (int i = 0; i < height; i++) {
for (int j = 0; j < width; j++) {
printf("%d\t", array[i][j]);
}
printf("\n");
}
// Compute the bounds for processes 1-2-3
for (int target = 0; target < size; target++) {
printf("\n\nBounds for PROCESS:%d\n", target);
if (target == 0) {
start = target * sliceSize;
end = (start + sliceSize - 1) + extrapixels;
} else {
start = (target * sliceSize) + extrapixels;
end = start + sliceSize - 1;
}
printf("\n\n START : %d \t END : %d\n", start, end);
printf("\n\n*****SLICE FOR PROCESS %d******\n\n", target);
for (i = start; i <= end; i++) {
for (j = 0; j < width; j++) {
if (target == 0) {
subArrayFext[i - start][j] = array[i][j];
} else {
subArray[i - start][j] = array[i][j];
}
}
}
printf("\nSlice Done!\n");
if (target > 0) {
for (int k = 0; k < sliceSize; k++) {
for (int l = 0; l < width; l++) {
printf("%d\t", subArray[k][l]);
}
printf("\n");
}
printf("\n Sending array to process %d\n", target);
MPI_Send(*subArray, sliceSize * width, MPI_INT, target, 10,
MPI_COMM_WORLD);
}
}
} else {
printf("\n-----------------\nrank %d\nmy size:%d\n------------------\n",
my_rank, sliceSize * width);
MPI_Recv(*subArray, sliceSize * width, MPI_INT, 0, 10, MPI_COMM_WORLD,
&status);
for (k = 0; k < sliceSize; k++) {
for (l = 0; l < width; l++) {
printf("R:%d-%d\t", my_rank, subArray[k][l]);
}
printf("\n");
}
}
free(image);
MPI_Finalize();
// return (0);
}
这就是打印输出的样子，它与发送的数组不同。
R:1-64 R:1-65 R:1-66 R:1-67 R:1-69 R:1-71 R:1-68 R:1-94 R:1-133 R:1-151 R:1-130 R:1-80 R:1-58 R:1-65 R:1-71 R:1-72 R:1-73 R:1-76 R:1-76 R:1-74 R:1-73 R:1-196 R:1-221 R:1-218 R:1-231 R:1-216 R:1-84 R:1-70 R:1-73 R:1-72
R:1-62 R:1-64 R:1-66 R:1-68 R:1-71 R:1-66 R:1-88 R:1-170 R:1-181 R:1-180 R:1-185 R:1-171 R:1-90 R:1-67 R:1-73 R:1-69 R:1-64 R:1-65 R:1-66 R:1-73 R:1-66 R:1-123 R:1-214 R:1-198 R:1-208 R:1-143 R:1-67 R:1-74 R:1-73 R:1-73
R:1-62 R:1-62 R:1-58 R:1-57 R:1-63 R:1-61 R:1-109 R:1-178 R:1-179 R:1-181 R:1-176 R:1-188 R:1-131 R:1-61 R:1-69 R:1-78 R:1-113 R:1-123 R:1-106 R:1-71 R:1-69 R:1-66 R:1-89 R:1-107 R:1-90 R:1-69 R:1-75 R:1-74 R:1-72 R:1-71
R:1-59 R:1-64 R:1-99 R:1-119 R:1-98 R:1-61 R:1-82 R:1-169 R:1-183 R:1-185 R:1-179 R:1-187 R:1-108 R:1-59 R:1-86 R:1-164 R:1-172 R:1-179 R:1-182 R:1-149 R:1-71 R:1-71 R:1-68 R:1-66 R:1-66 R:1-64 R:1-62 R:1-60 R:1-68 R:1-70
R:1-67 R:1-145 R:1-201 R:1-204 R:1-198 R:1-159 R:1-68 R:1-94 R:1-153 R:1-158 R:1-152 R:1-115 R:1-66 R:1-63 R:1-119 R:1-185 R:1-174 R:1-182 R:1-178 R:1-187 R:1-105 R:1-64 R:1-0 R:1-0 R:1-0 R:1-0 R:1-0 R:1-0 R:1-0 R:1-0
R：1代表Rank：1