您好我在创建它们后释放了一些变量似乎存在问题。
我的计划背景很少。我的程序正在读取MatrixMarket文件并根据读入的信息构建矩阵结构。然后我使用矩阵进行标量和向量乘法。
这是我的矩阵结构。
typedef struct {
double *values; // Value of each non-zero element.
unsigned int *col_indices; // Corresponding column coordinate.
unsigned int *row_ptr; // Initial position of each row.
unsigned int nz; // Total non-zero elements.
unsigned int row_count; // Total row_count of matrix.
unsigned int cols_count; // Total columns of matrix.
} CSRMatrix;
我的目标是能够操纵我的矩阵并能够将矢量乘以CSR矩阵。 （CSR是压缩稀疏行矩阵）。
以下是我初始化内存分配的方法。
/**
* Set the correct values to the struct and create the memory allocation.
*/
csrMatrix.row_count = numberOfRows;
csrMatrix.cols_count = numberOfColumn;
csrMatrix.nz = numberOfNonZeros;
csrMatrix.row_ptr = calloc(numberOfRows, 1);
csrMatrix.values = (double *)malloc(numberOfRows * numberOfRows * 1);
csrMatrix.col_indices = calloc(numberOfRows * numberOfRows, 1);
这是我认为我的问题所在，我可能错了。我在排除内存问题时非常糟糕。
在我进一步深入讨论矩阵的主要乘法和迭代之前：
// Initialize vectors for multiplication
int x_size = numberOfRows * numberOfColumn;
int csr_x[x_size];
memset(csr_x, 0, sizeof(csr_x));
for(int h = 0; h < x_size; h++){
csr_x[h] = 1;
}
double* csr_y;
printf("\nCalculating matrix-vector product of the CSR for %u iterations...\n", iterations);
clock_t begin = clock();
for(i=0; i < iterations; i++){
csr_y = (double *)malloc(csrMatrix.row_count * csrMatrix.cols_count * 1);
csrMVP(&csrMatrix, csr_x,csr_y);
free(csr_y);
}
clock_t end = clock();
double time_spent = (double)(end - begin) / CLOCKS_PER_SEC;
printf("for loop used %f seconds to complete the execution\n", time_spent);
fclose(file);
free(csr_x);
free(csr_y);
exit(EXIT_SUCCESS);
当我运行我的程序时，奇怪的事情正在发生。程序将运行几个循环（总循环为1000次迭代），然后使用Process finished with exit code 0或exit code 11退出不完整。这是我的故障排除开始的地方。
由于intellij没有抛出任何内存问题或错误，我必须更深入。
我决定问我的一位朋友谁曾使用Valgrind对我的节目进行记忆。这就是结果：
==5607== Memcheck, a memory error detector
==5607== Copyright (C) 2002-2017, and GNU GPL'd, by Julian Seward et al.
==5607== Using Valgrind-3.13.0 and LibVEX; rerun with -h for copyright info
==5607== Command: ./code
/home/tron/src/workspace/classes/cda5110/matrix/mat/ibm32/ibm32.mtx
==5607==
==5607== Invalid write of size 4
==5607== at 0x109537: loadCSRMatrix (mainCSR.c:173)
==5607== by 0x108F3A: main (mainCSR.c:72)
==5607== Address 0x51f23b0 is 0 bytes after a block of size 128 alloc'd
==5607== at 0x4C2EEF5: calloc (vg_replace_malloc.c:711)
==5607== by 0x108EC5: main (mainCSR.c:64)
==5607==
==5607== Invalid read of size 4
==5607== at 0x109300: csrMVP (mainCSR.c:140)
==5607== by 0x109085: main (mainCSR.c:92)
==5607== Address 0x51f23b0 is 0 bytes after a block of size 128 alloc'd
==5607== at 0x4C2EEF5: calloc (vg_replace_malloc.c:711)
==5607== by 0x108EC5: main (mainCSR.c:64)
==5607==
==5607== Invalid free() / delete / delete[] / realloc()
==5607== at 0x4C2E10B: free (vg_replace_malloc.c:530)
==5607== by 0x10910C: main (mainCSR.c:99)
==5607== Address 0x1ffefff5a0 is on thread 1's stack
==5607== in frame #1, created by main (mainCSR.c:24)
==5607==
==5607== Invalid free() / delete / delete[] / realloc()
==5607== at 0x4C2E10B: free (vg_replace_malloc.c:530)
==5607== by 0x10911B: main (mainCSR.c:100)
==5607== Address 0x59d3bc0 is 0 bytes inside a block of size 8,192 free'd
==5607== at 0x4C2E10B: free (vg_replace_malloc.c:530)
==5607== by 0x109094: main (mainCSR.c:93)
==5607== Block was alloc'd at
==5607== at 0x4C2CEDF: malloc (vg_replace_malloc.c:299)
==5607== by 0x109064: main (mainCSR.c:91)
==5607==
==5607==
==5607== HEAP SUMMARY:
==5607== in use at exit: 12,416 bytes in 3 blocks
==5607== total heap usage: 1,006 allocs, 1,005 frees, 8,213,160 bytes
allocated
==5607==
==5607== LEAK SUMMARY:
==5607== definitely lost: 0 bytes in 0 blocks
==5607== indirectly lost: 0 bytes in 0 blocks
==5607== possibly lost: 0 bytes in 0 blocks
==5607== still reachable: 12,416 bytes in 3 blocks
==5607== suppressed: 0 bytes in 0 blocks
==5607== Rerun with --leak-check=full to see details of leaked memory
==5607==
==5607== For counts of detected and suppressed errors, rerun with: -v
==5607== ERROR SUMMARY: 1003 errors from 4 contexts (suppressed: 0 from 0)
我突然发现了一些东西。有一些invaild写入和读取大小。这些是由错误调出的每个代码的以下行：
m->row_ptr[row_index + 1] = pairIndex; loadCSRMatrix(&csrMatrix, file); csrMatrix.row_ptr = calloc(numberOfRows, 1); y[i] += val[j] * x[col[j]]; free(csr_y); csrMatrix.row_ptr = calloc(numberOfRows, 1); free(csr_y); exit(EXIT_SUCCESS); } by 0x109064：main（mainCSR.c：91）：csrMVP(&csrMatrix, csr_x,csr_y);
从我提到的内容是我没有正确地使用malloc和calloc进行内存分配。
所以我在这里问我的问题，我不确定malloc或calloc是否正确使用。如果有人知道我可以在哪里开始寻找解决这个问题。
如果您想知道loadCSRMatrix和csrMVP功能是什么，我会在下面附上。
csrMVP
/**
* Stores on vector y the product of matrix M and vector x:
* y = Mx
*/
void csrMVP(CSRMatrix *m, int x[], double *y){
printf("\n\n\nValue of Y: \n ");
unsigned int i,j,end;
double *val = m->values;
unsigned int *col = m->col_indices;
unsigned int *ptr = m->row_ptr;
end = 0;
for(i = 0; i < m->row_count; i++){
y[i] = 0.0;
j = end;
end = ptr[i+1];
for( ; j < end; j++){
y[i] += val[j] * x[col[j]];
printf("%.f\t", y[i]);
}
}printf("\n");
}
loadCSRMatrix
/**
* Loads the matrix
*/
void loadCSRMatrix(CSRMatrix *m, FILE *f){
unsigned int pairIndex = 0;
for (int row_index = 0; row_index < m->row_count; row_index++) {
for (unsigned int column_index = 0; column_index < m->row_count; column_index++) {
double value;
if(fscanf(f, "%lg", &value) != EOF){
m->values[pairIndex] = value;
m->col_indices[pairIndex] = column_index+1;
//printf("Column: %d\t", matrixCSR.col_indices[pairIndex]);
printf("%.f\t", m->values[pairIndex]);
pairIndex++;
}
else {
m->values[pairIndex] = 0;
m->col_indices[pairIndex] = column_index+1;
//printf("Column: %d\t", matrixCSR.col_indices[pairIndex]);
printf("%.1f\t", m->values[pairIndex]);
pairIndex++;
}
}
if(pairIndex != 0) {
m->row_ptr[row_index + 1] = pairIndex;
}
//printf("%zu\n", matrixCSR.row_indices[row_index]);
printf("\n");
}
}