我想在C中用LAPACK的dgels_来解决最小二乘问题| Ax-b | - > min,但是我得到了一个分段错误错误(我知道有{{ 3}},但代码完全不同,答案不适用于我的问题)。 我已经找到了问题,dgels_执行时是正确的。
代码:
#include <stdio.h>
#include <lapacke.h>
#define COL 3
#define ROW 4
int main()
{
char transa ='N';
lapack_int m, n, nrhs, lda, ldb, info;
m=ROW;
n=COL;
nrhs=1;
lda=ROW;
ldb=ROW;
double work[COL];
double A [COL*ROW] =
{ 1.1, 4.2, 1.7,
2.5, 2.1, 2.8,
3.4, 4.2, 8.5,
4.4, 5.2, 7.8 };
double b[ROW] =
{ 1.5,
2.1,
3.8,
3.4 };
printf("test 1 \n");
dgels_(&transa, &m, &n, &nrhs, A, &lda, b, &ldb, work, 0, &info);
printf("test 2 \n");
return 0;
}
第一个&#34;测试1&#34;打印然后有分段错误错误。
编辑: 我之前尝试使用值编译它,即使用dgels_(transa,m,n,nrhs,A,lda,b,ldb,work,0,info)。 但是,我收到了很多错误消息:
lapack_error.c: In function ‘main’:
lapack_error.c:32:13: warning: passing argument 1 of ‘dgels_’ makes pointer from integer without a cast [-Wint-conversion]
dgels_(transa, m, n, nrhs, A, lda, b, ldb, work, 0, info);
^~~~~~
In file included from /usr/include/lapacke.h:143:0,
from lapack_error.c:2:
/usr/include/lapacke.h:14793:6: note: expected ‘char *’ but argument is of type ‘char’
void LAPACK_dgels( char* trans, lapack_int* m, lapack_int* n, lapack_int* nrhs,
^
lapack_error.c:32:21: warning: passing argument 2 of ‘dgels_’ makes pointer from integer without a cast [-Wint-conversion]
dgels_(transa, m, n, nrhs, A, lda, b, ldb, work, 0, info);
^
In file included from /usr/include/lapacke.h:143:0,
from lapack_error.c:2:
/usr/include/lapacke.h:14793:6: note: expected ‘int *’ but argument is of type ‘int’
void LAPACK_dgels( char* trans, lapack_int* m, lapack_int* n, lapack_int* nrhs,
^
lapack_error.c:32:24: warning: passing argument 3 of ‘dgels_’ makes pointer from integer without a cast [-Wint-conversion]
dgels_(transa, m, n, nrhs, A, lda, b, ldb, work, 0, info);
^
In file included from /usr/include/lapacke.h:143:0,
from lapack_error.c:2:
/usr/include/lapacke.h:14793:6: note: expected ‘int *’ but argument is of type ‘int’
void LAPACK_dgels( char* trans, lapack_int* m, lapack_int* n, lapack_int* nrhs,
^
lapack_error.c:32:27: warning: passing argument 4 of ‘dgels_’ makes pointer from integer without a cast [-Wint-conversion]
dgels_(transa, m, n, nrhs, A, lda, b, ldb, work, 0, info);
^~~~
In file included from /usr/include/lapacke.h:143:0,
from lapack_error.c:2:
/usr/include/lapacke.h:14793:6: note: expected ‘int *’ but argument is of type ‘int’
void LAPACK_dgels( char* trans, lapack_int* m, lapack_int* n, lapack_int* nrhs,
^
lapack_error.c:32:36: warning: passing argument 6 of ‘dgels_’ makes pointer from integer without a cast [-Wint-conversion]
dgels_(transa, m, n, nrhs, A, lda, b, ldb, work, 0, info);
^~~
In file included from /usr/include/lapacke.h:143:0,
from lapack_error.c:2:
/usr/include/lapacke.h:14793:6: note: expected ‘int *’ but argument is of type ‘int’
void LAPACK_dgels( char* trans, lapack_int* m, lapack_int* n, lapack_int* nrhs,
^
lapack_error.c:32:44: warning: passing argument 8 of ‘dgels_’ makes pointer from integer without a cast [-Wint-conversion]
dgels_(transa, m, n, nrhs, A, lda, b, ldb, work, 0, info);
^~~
In file included from /usr/include/lapacke.h:143:0,
from lapack_error.c:2:
/usr/include/lapacke.h:14793:6: note: expected ‘int *’ but argument is of type ‘int’
void LAPACK_dgels( char* trans, lapack_int* m, lapack_int* n, lapack_int* nrhs,
^
lapack_error.c:32:58: warning: passing argument 11 of ‘dgels_’ makes pointer from integer without a cast [-Wint-conversion]
dgels_(transa, m, n, nrhs, A, lda, b, ldb, work, 0, info);
^~~~
In file included from /usr/include/lapacke.h:143:0,
from lapack_error.c:2:
/usr/include/lapacke.h:14793:6: note: expected ‘int *’ but argument is of type ‘int’
void LAPACK_dgels( char* trans, lapack_int* m, lapack_int* n, lapack_int* nrhs,
然后我尝试查找示例程序并找到a similar problem(使用sgesv,但我认为它可能与dgels类似)。并且在将dgels_(transa,m,n,nrhs,A,lda,b,ldb,work,0,info)重写为dgels _(&amp; transa,&amp; m,&amp; n,&amp; nrhs,A,&amp; lda,b,&amp; ldb,work,0,&amp; info)没有更多错误消息,所以我认为这是正确的方法。
答案 0 :(得分:1)
0
不是dgels的参数lwork
的合法值。实际上,它必须是指向c中整数的指针,因为所有参数都在fortran中通过引用调用,而所有参数都由c中的值调用。此外,lwork
的值指定数组work
的长度,must be larger than 1:LWORK >= max( 1, MN + max( MN, NRHS ) )
。唯一的例外是lwork=-1
:在这种特殊情况下,函数会返回work
中数组work[0]
的最佳大小。
例如,可以尝试以下几行:
lapack_int lwork=n;
if (m<n){lwork=m;}
lwork=-1;
double query;
dgels_(&transa, &m, &n, &nrhs, A, &lda, b, &ldb, &query, &lwork, &info);
lwork=(int)query;
printf("the optimal size is %d \n",lwork);
work=malloc(lwork*sizeof(double));
if(work==NULL){fprintf(stderr,"maloc failed\n");exit(1);}
dgels_(&transa, &m, &n, &nrhs, A, &lda, b, &ldb, work, &lwork, &info);
printf("test 2 \n");
printf("%g %g %g\n",b[0],b[1],b[2]);
dgels_
有两次调用:第一次返回work
的正确大小。然后,分配work
并再次调用dgels_
以执行最小二乘最小化。
结果可能与常规C开发人员所期望的结果不同:Fortran and C order of dimensions for multidimensional arrays are different .Lapacke包装器提供函数LAPACKE_dgels()
和LAPACKE_dgels_work()
参数LAPACK_COL_MAJOR
/ { {1}}和LAPACK_ROW_MAJOR
。始终确保通过执行测试为这些参数使用正确的值!如果失败,请尝试不同的值......
以下是基于您的示例代码,展示了通过Lapacke界面调用transa
的不同方法。它可以由dgels
编译。
gcc main.c -o main -llapacke -llapack -lblas -lm -Wall