Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Problem with MRRR and range=I #229

Open
ingutheil opened this issue Apr 11, 2017 · 0 comments
Open

Problem with MRRR and range=I #229

ingutheil opened this issue Apr 11, 2017 · 0 comments

Comments

@ingutheil
Copy link

I am testing the HermitianEig with range=I, D&C and QR work with the smallest 5%, the smallest 25%, and all eingenvalues and eigenvectors with range=I, il=0, iu=matrixheight-1,
MRRR only works with range=I, il=0, iu=matrixheight-1, with the other two options it gives

  • srun bench_hermitian_eig_openblas --gridHeigt 6 --range I --sort 2 --startheight 6000 --incr 4000 --steps 5 --nb 64 --nbLocal 16 --algInt 2 --calcMatrix true --pcev 5
    srun: error: jrc0093: tasks 0-2,4-8,11-23: Segmentation fault
    srun: error: jrc0093: tasks 3,9-10: Terminated
    srun: Force Terminated job step 3166195.0

  • JUBE_ERR_CODE=139

  • '[' 139 -ne 0 ']'

  • exit 139
    The only output is
    Optional arguments:
    --gridHeight [int,0,0,NOT found]
    height of process grid

    --colMajor [bool,1,1,NOT found]
    column-major ordering?

    --onlyEigvals [bool,0,0,NOT found]
    only compute eigenvalues?

    --range [char,A,I,found]
    range of eigenpairs: 'A' for all, 'I' for index range, 'V' for value range

    --il [int,0,0,NOT found]
    lower bound of index range

    --iu [int,99,99,NOT found]
    upper bound of index range

    --vl [double,0,0,NOT found]
    lower bound of value range

    --vu [double,100,100,NOT found]
    upper bound of value range

    --pcev [int,100,5,found]
    percenet of eigenspectrum wanted

    --sort [int,0,2,found]
    sort type 0: unsorted 1: descending 2: ascending

    --cluster [bool,0,0,NOT found]
    force clustered eigenvalues?

    --uplo [char,L,L,NOT found]
    upper or lower storage: L/U

    --startheight [int,100,6000,found]
    height of first matrix to test

    --incr [int,100,4000,found]
    increment

    --steps [int,1,5,found]
    steps

    --nb [int,96,64,found]
    algorithmic blocksize

    --nbLocal [int,32,16,found]
    local blocksize

    --avoidTrmv [bool,1,1,NOT found]
    avoid Trmv based Symv

    --useScaLAPACK [bool,0,0,NOT found]
    test ScaLAPACK?

    --algInt [int,1,2,found]
    0: QR, 1: D&C, 2: MRRR

    --sequential [bool,0,0,NOT found]
    test sequential?

    --distributed [bool,1,1,NOT found]
    test distributed?

    --correctness [bool,1,1,NOT found]
    test correctness?

    --progress [bool,0,0,NOT found]
    print progress?

    --print [bool,0,0,NOT found]
    print matrices?

    --timeStages [bool,1,1,NOT found]
    time stages?

    --readMatrix [bool,0,0,NOT found]
    read matrix?

    --calcMatrix [bool,1,true,found]
    read matrix?

Out of 0 required arguments, 0 were not specified.
Out of 28 optional arguments, 18 were not specified.

Grid height: 4 width: 6
mat_siz nro nco OMP blocksz nev time(solve) max deviation max devHemm max devHerk
Condense time: 5.78749 secs

bench_hermitian_eig_openblas
is a modification of the program HermitianEig in test which builds the matrix that I am using for the other tests and it calls HermitianEig steps times with matrix sizes startheigt and startheight+incr, startheight+2*incr...

(Just a question concerning the original HermitianEig program in test: The default for matrixheight is 100, if you choose range I the default for il is 0, but the default for iu is 100, shouldn't it be 99???)

Regards
Inge Gutheil

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant