我正在处理 block 对角矩阵(每个 block 具有相同的大小),当我使用 private
动态分配的二维数组时,我遇到了一个非法地址错误
。 .
// NB is the number of block
// N is the block size
// A is the main matrix (block diagonal)
double** B; // a block
B = new double*[N];
for (unsigned int i = 0; i < N; i++)
B[i] = new double[N];
#pragma acc parallel loop private(B[:N][:N]) copyin(A[:NB*N][:NB*N])
for (unsigned int b = 0; b < NB; b++) {
#pragma acc loop
for (unsigned int i = 0; i < N; i++) {
#pragma acc loop
for (unsigned int j = 0; j < N; j++) {
B[i][j] = A[b*N+i][b*N+j];
}
}
// process B
}
for (unsigned int i = 0; i < N; i++)
delete[] B[i];
delete[] B;
我得到的错误是:
call to cuStreamSynchronize returned error 700: Illegal address during kernel execution
如果我将数组展平为一维数组并使用字典索引或静态二维数组,效果很好,但我使用的函数需要 double**
作为参数,所以我宁愿坚持使用动态二维数组...
我已经阅读了规范中的 private
子句,但它并没有说不支持动态二维数组,所以我想我做错了什么......
最佳答案
抱歉,不支持在私有(private)子句中使用指针数组。问题在于编译器运行时必须为每个 gang、worker 或 vector(取决于带有 private 子句的循环)动态创建一个 private 并填充所有设备指针。这将带来极高的管理费用。
如果“B”是一个固定大小的静态数组,“double B[N][N]”,那么你可以在私有(private)子句中使用它。
否则,我会建议通过添加第三维来手动私有(private)化数组。
// NB is the number of block
// N is the block size
// A is the main matrix (block diagonal)
double*** B; // a block
B = new double**[NB];
for (unsigned int i = 0; i < NB; i++) {
B[i] = new double*[N];
for (unsigned int j = 0; j < N; j++) {
B[i][j] = new double[N];
}}
#pragma acc parallel loop create(B[:NB][:N][:N]) copyin(A[:NB*N][:NB*N])
for (unsigned int b = 0; b < NB; b++) {
#pragma acc loop
for (unsigned int i = 0; i < N; i++) {
#pragma acc loop
for (unsigned int j = 0; j < N; j++) {
B[b][i][j] = A[b*N+i][b*N+j];
}
}
// process B
}
for (unsigned int i = 0; i < NB; i++) {
for (unsigned int j = 0; j < N; j++) {
delete[] B[i][j];
}
delete[] B[i];
}
delete[] B;
关于c++ - OpenACC - 私有(private)二维数组,我们在Stack Overflow上找到一个类似的问题: https://stackoverflow.com/questions/44029157/