Optistruct MUMPs

Lamster
Lamster New Altair Community Member
edited May 2022 in Community Q&A

Hi all, 

I have a model which presents me with the following....

"MUMPS failed in factorization because the run-time memory
requirement was larger than the estimated amount. Solver will
retry with a larger memory relaxation coefficient"

 

Am trying to understand if this is an issue or not (is the model being slowed, but a simple command/parameter will improve matters). My run command is "-core IN -checkel NO -out -np 8 -ddm"

The PC is an I9-9980HK with 8 cores (hyperthreading off) + 128GB ram. 

Welcome!

It looks like you're new here. Sign in or register to get started.

Answers

  • Adriano_Koga
    Adriano_Koga
    Altair Employee
    edited January 2022

    i believe you shoud look at your memory estimatives in the .out file.

    Are they too large? Depending on the model, 128GB might not be enough.

  • Lamster
    Lamster New Altair Community Member
    edited January 2022

    Hi there, many thanks. This is from my .out file...

     

    MEMORY ESTIMATION INFORMATION :
    (FOR CURRENT MPI PROCESS)
    -------------------------------

    Solver Type is: Sparse-Matrix Solver
    Direct Method

    Memory (RAM) Allocated in Preprocessor : 6774 MB
    Estimated Memory (RAM) for Out of Core Solution : 8442 MB
    Estimated Memory (RAM) for In-Core Solution : 11171 MB
    Maximum Estimated Memory (RAM) for Any Single Host
    for Out of Core Solution : 65549 MB
    Maximum Estimated Memory (RAM) for Any Single Host
    for In-Core Solution : 84331 MB

    MUMPS Solver will Run with In-Core Solution.

     

     

    Do you think this would benefit from specifying available RAM (-fixlen)?

  • Casper Kruger
    Casper Kruger Altair Community Member
    edited May 2022

    Hi

     

    Did you resolve this issue in the end? 

  • rosh
    rosh Altair Community Member

    Any resolution for the problem?
    I am facing same issue with HM and OptiStruct 2023 version

Welcome!

It looks like you're new here. Sign in or register to get started.

Welcome!

It looks like you're new here. Sign in or register to get started.