我想测试 Intel MKL
的矩阵乘法,所以我包含了 mkl.h
头文件,并使用了 cblas_dgemm
函数,但它总是报错。
undefined reference to `cblas_dgemm'
我也链接了-lmkl_core -lmkl_blas95_lp64 -lmkl_lapack95_lp64
,但是在$MKLROOT/lib/intel64/
目录下的库中测试了许多组合,错误仍然存在。请问有人可以给我一些建议吗?谢谢。
我想测试 Intel MKL
的矩阵乘法,所以我包含了 mkl.h
头文件,并使用了 cblas_dgemm
函数,但它总是报错。
undefined reference to `cblas_dgemm'
我也链接了-lmkl_core -lmkl_blas95_lp64 -lmkl_lapack95_lp64
,但是在$MKLROOT/lib/intel64/
目录下的库中测试了许多组合,错误仍然存在。请问有人可以给我一些建议吗?谢谢。
也许这是一个正确的答案,我们可以使用cblas_
:
在QT Creator的项目文件中:
unix {
INCLUDEPATH += /opt/intel/mkl/include
LIBS += -L/opt/intel/mkl/lib/intel64 \
-lmkl_intel_lp64 -lmkl_intel_thread -lmkl_core \
-L/opt/intel/lib/intel64 \
-liomp5 -lpthread -dl -lm
}
main.cpp
代码: #include <iostream>
using namespace std;
#include <cmath>
#include <cstdio>
#include <cstdlib>
#include <mkl.h>
#include <mkl_cblas.h>
#include <mkl_blas.h>
#include <mkl_lapack.h>
#include <mkl_lapacke.h>
template<typename T>
void printArray(T *data, char *name, int len){
cout << name << "\n";
for(int i=0;i<len;i++){
cout << data[i] << " ";
}
cout << "\n";
}
template<typename T>
void printMatrix(T *data, char *name, int m, int n){
cout << name << "\n";
for(int i=0;i<m;i++){
for(int j=0;j<n;j++){
cout << data[n*i+j] << " ";
}
cout << "\n";
}
cout << "\n";
}
int main()
{
int len=10;
double *x=new double[len];
double *y=new double[len];
for(int i=0;i<len;i++){
x[i]=(double)rand()/RAND_MAX;
y[i]=(double)rand()/RAND_MAX;
}
printArray<double>(x, "x", len);
printArray<double>(y, "y", len);
//sum(x)
double x_sum=cblas_dasum(len,x,1);
cout<< "sum(x): "<< x_sum <<"\n";
//y=a*x+y
double alpha=1;
cblas_daxpy(len,alpha,x,1,y,1);
printArray<double>(y,"y=a*x+y",len);
//y=x
cblas_dcopy(len,x,1,y,1);
printArray<double>(y,"y=x",len);
//x*y';
double xy_dot=cblas_ddot(len,x,1,y,1);
cout <<"x*y': "<<xy_dot<<"\n";
//extened x*y', not test
//cblas_sdsdot(),not test
//cbals_dsddot(),not test
//for complex vector
//cblas_dotc(),no this function
//cblas_dotu(),no this function
//norm(x), or ||x||_2
double x_norm=cblas_dnrm2(len,x,1);
cout<<"x_norm: "<<x_norm<<"\n";
//x(i)=c*x(i)+s*y(i);
//y(i)=c*y(i)-s*x(i);
//cblas_zrotg(); not test
//swap(x,y)
cblas_dswap(len,x,1,y,1);
printArray<double>(x,"x:",len);
printArray<double>(y,"y:",len);
//LEVEL 2 BLAS
//matrix and vector manipulation
int m=len;
int n=len;
double *A=new double[m*n];
for(int i=0;i<m*n;i++){
A[i]=(double)rand()/RAND_MAX;
}
printMatrix<double>(A,"A",m,n);
//matrix and vector multiplication
double alpha_dgemv=1.0;
double beta_dgemv=1.0;
//y=alpha*A*x+beta*y, if A is a mxn band matrix, then use cblas_dgbmv
cblas_dgemv(CblasRowMajor,CblasNoTrans,m,n,alpha_dgemv,A,m,x,1,beta_dgemv,y,1);
printArray<double>(x,"x:",len);
printArray<double>(y,"y=alpha*A*x+beta*y",len);
//y=alpha*A'*x+beta*y
cblas_dgemv(CblasRowMajor,CblasTrans,m,n,alpha_dgemv,A,m,x,1,beta_dgemv,y,1);
printArray<double>(x,"x:",len);
printArray<double>(y,"y=alpha*A'*x+beta*y",len);
//A=alpha*x*y'+A;
double alpha_dger=1.0;
cblas_dger(CblasRowMajor,m,n,alpha_dger,x,1,y,1,A,m);
printArray<double>(x,"x:",len);
printArray<double>(y,"y:",len);
printMatrix<double>(A,"A=alpha1*x*y'+A",m,n);
delete[] x;x=NULL;
delete[] y;y=NULL;
delete[] A;A=NULL;
//
m=10;
n=5;
int k=3;
double *Amxk=new double[m*k];
double *Bkxn=new double[k*n];
double *Cmxn=new double[m*n];
for(int i=0;i<m*k;i++){
Amxk[i]=(double)rand()/RAND_MAX;
}
for(int i=0;i<k*n;i++){
Bkxn[i]=(double)rand()/RAND_MAX;
}
for(int i=0;i<m*n;i++){
Cmxn[i]=0;
}
printMatrix<double>(Amxk,"Amxk",m,k);
printMatrix<double>(Bkxn,"Bkxn",k,n);
printMatrix<double>(Cmxn,"Cmxn",m,n);
double alpha_dgemm=1.0;
double beta_dgemm=1.0;
cblas_dgemm(CblasRowMajor,
CblasNoTrans,
CblasNoTrans,
m,
n,
k,
alpha_dgemm,
Amxk,
k,
Bkxn,
n,
beta_dgemm,
Cmxn,
n);
printMatrix<double>(Cmxn,"Cmxn",m,n);
delete[] Amxk;
delete[] Bkxn;
delete[] Cmxn;
//general symmetric matrix eigenvalue decomposition
/* Locals */
const MKL_INT N=5;
const MKL_INT LDA=5;
MKL_INT lda = LDA, info;
n=N;
/* Local arrays */
double w[N];
double a[LDA*N] = {
6.39, 0.13, -8.23, 5.71, -3.18,
0.00, 8.37, -4.46, -6.10, 7.21,
0.00, 0.00, -9.58, -9.25, -7.42,
0.00, 0.00, 0.00, 3.72, 8.54,
0.00, 0.00, 0.00, 0.00, 2.51
};
/* Executable statements */
printf( "LAPACKE_dsyevd (row-major, high-level) Example Program Results\n" );
/* Solve eigenproblem */
info = LAPACKE_dsyevd( LAPACK_ROW_MAJOR, 'V', 'U', n, a, lda, w );
/* Check for convergence */
if( info > 0 ) {
printf( "The algorithm failed to compute eigenvalues.\n" );
exit( 1 );
}
printArray<double>(w,"w:",N);
printMatrix<double>(a,"a:",N,N);
int N1=10;
int d=3;
double *A1=new double[N1*d];
for(int j=0;j<d;j++){
for(int i=0;i<N1;i++){
A1[j*N1+i]=(double)rand()/RAND_MAX;
}
}
printMatrix<double>(A1,"A:",d,N1);
double *A_mean=new double[1*d];
for(int i=0;i<d;i++){
A_mean[i]=cblas_dasum(N,A1+i*N1,1);
}
printArray<double>(A_mean,"A_mean",d);
delete[] A1;A1=NULL;
delete[] A_mean;A_mean=NULL;
return 0;
}
source /opt/intel/bin/compilervars.sh intel64
Intel® Math Kernel Library Link Line Advisor正是您所需要的。它可以为您生成正确的编译/链接选项。
http://software.intel.com/en-us/articles/intel-mkl-link-line-advisor
/home/ubuntu/Documents/3rdparty/include/opencv2
/home/ubuntu/Documents/3rdparty/include/opencv
/opt/intel/mkl/includeLIBS += -L/home/ubuntu/Documents/3rdparty/lib/
-L/opt/intel/mkl/lib/intel64
-lopencv_core -lopencv_imgproc -lopencv_highgui
-larmadillo -lblaze -lboost_system
-fopenmp -lpthread
-lmkl_core -lmkl_blas95_lp64 -lmkl_lapack95_lp64 -liomp5
-ldl -lm
} - mining