- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Dear all,
According to Intel there is a new support in MKL for largest eigenvalues.
I have upgraded to the latest Fortran compiler with MKL. I can compile the example folder for the largest eigenvalues but when I have tried to solve the following example I have got the following error.
Here is the minimal example:
PROGRAM sparse_exp USE MKL_SPBLAS USE MKL_SOLVERS_EE implicit none !!!!!!!!!!!!!!!!! Matrix declaration variables !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! integer n,i parameter (n=5) real val(25) integer rows(n+1), cols(25) ! Matrix descriptor TYPE(MATRIX_DESCR) descrA ! CSR matrix structure TYPE(SPARSE_MATRIX_T) csrA !!!!!!!!!!!!!!!!! Declaration of MKL_SPARSE_S_EV variables !!!!!!!!!!!!!!!!!!!!! !!!!!!!!!!!!!!!!! E - eigenvalues, X - eigenvectors, res - residual !!!!!!!!!!!! !!!!!!!!!!!!!!!!! XL - left singular vectors, XR - right singular vectors !!!!!! character*1 WHICH parameter (WHICH='L') !L - for largest eigenvalues to find integer pm(128) integer K0,K,info parameter (K0 = 1) !Required number of eigenvalues/singular values to find real E(n) real sigma(n) real X(n,n), XL(n,n), XR(n,n) real res(n) val(1) = 0.2124 val(2) = 0.1887 val(3) = 0.1838 val(4) = 0.2114 val(5) = 0.2016 val(6) = 0.1899 val(7) = 0.2614 val(8) = 0.2012 val(9) = 0.1883 val(10)= 0.1629 val(11)= 0.1845 val(12)= 0.2007 val(13)= 0.2373 val(14)= 0.1863 val(15)= 0.1926 val(16)= 0.2114 val(17)= 0.1871 val(18)= 0.1856 val(19)= 0.2106 val(20)= 0.2032 val(21)= 0.2016 val(22)= 0.1619 val(23)= 0.1919 val(24)= 0.2032 val(25)= 0.2395 rows(1)=1 rows(2)=6 rows(3)=11 rows(4)=16 rows(5)=21 rows(6)=26 cols(1) = 1 cols(2) = 2 cols(3) = 3 cols(4) = 4 cols(5) = 5 cols(6) = 1 cols(7) = 2 cols(8) = 3 cols(9) = 4 cols(10)= 5 cols(11)= 1 cols(12)= 2 cols(13)= 3 cols(14)= 4 cols(15)= 5 cols(16)= 1 cols(17)= 2 cols(18)= 3 cols(19)= 4 cols(20)= 5 cols(21)= 1 cols(22)= 2 cols(23)= 3 cols(24)= 4 cols(25)= 5 print *,'Sparse matrix size',n ! ! Task 1. Call MKL_SPARSE_C_CREATE_CSR to create matrix handle ! info = mkl_sparse_s_create_csr(csrA,SPARSE_INDEX_BASE_ZERO,N,N,rows,rows(2),cols,val) ! Create matrix descriptor descrA % TYPE = SPARSE_MATRIX_TYPE_GENERAL ! ! Task 2. Call MKL_SPARSE_EE_INIT to define the default values for the input ! parameters. ! info = mkl_sparse_ee_init(pm) pm(2) = 4 !Setting tolerance pm(8) = 1 ! Use absolute stopping criteria print *, ' Testing mkl_sparse_s_ev ' ! ! Task 3. Solve the standard eigenvalue problem Ax=ex. ! info = mkl_sparse_s_ev(WHICH,pm,csrA,descrA,k0,k,E,X,res) print *,' OUTPUT INFO ',info if(info.ne.0) stop 1 print *, 'Number of eigenvalues found ', k print *, ' Computed | Expected ' print *, ' Eigenvalues | Eigenvalues ' do i=1,K print *, E(i) enddo ! Release internal representation of CSR matrix info = MKL_SPARSE_DESTROY(csrA) end program
And here is the error:
Sparse matrix size 5 Testing mkl_sparse_s_ev forrtl: severe (174): SIGSEGV, segmentation fault occurred Image PC Routine Line Source eigen.exe 0000000000404B53 Unknown Unknown Unknown libpthread-2.28.s 00007FCE5012E3C0 Unknown Unknown Unknown libmkl_core.so 00007FCE508C4D0D mkl_serv_free Unknown Unknown libmkl_intel_thre 00007FCE561F80E1 mkl_sparse_s_kryl Unknown Unknown libmkl_intel_thre 00007FCE561F3FFA mkl_sparse_s_ev_i Unknown Unknown libmkl_intel_lp64 00007FCE573379D7 MKL_SPARSE_S_EV Unknown Unknown eigen.exe 0000000000403AFA Unknown Unknown Unknown eigen.exe 0000000000403862 Unknown Unknown Unknown libc-2.28.so 00007FCE4FF7C223 __libc_start_main Unknown Unknown eigen.exe 000000000040376E Unknown Unknown Unknown
Link Copied
0 Replies

Reply
Topic Options
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page