# Pardiso problem with large equation systems

Hi guys,

I am facing a problem with Pardiso and I really appreciate it if someone could give me a hint on it.

The system I am trying to solve contains 1,016,451 number of equations with  569392326 nonzero elements. I am using my settings in pardiso which looks like the following

iparm[0] = 1; /* No solver default */
iparm[1] = 0;

# what's the real meaning of phase=331, 332, 333 in pardiso solver

In pardiso solver, the parameter phase means the solver execution steps.

If Ax=b, and A is a Real and symmetric indefinite matrix, then P*A*P'=LDL'.

For example, phase=332, only diagonal substitution is executed(Dx=b), and if b is an identity matrix, x will be inverse of D.

But, the result is not. What's the matter, and what's the real meaning of phase=331, 332, 333.

Thank you!

# Do I have Intel Compiler installed?

My company has a central license for this: w_mkl_11.0.0.089.exe

When I installed, it installed Intel Composer XE 2013 (according to the screen and the installation folder, which is: C:\Program Files (x86)\Intel\Composer XE 2013), to my surprise. However, I don't see any icl.exe in the bin folder.

In VS, I don't see the option to use Intel Compiler when right-clicking on the project, although I see options to use IPP/MKL/TBB in the project properties.

So I'm confused whether my installation package includes the compiler. I guess not.

# Sherman Morrisson algorithm : almost tridiagonal matrices

Hello all,

Let M be a real tridiagonal matrix of size n greater than or equal to 2. We perfectly know how to solve systems of the form MX = Y with dgttrf followed by dgttrs. Now imagine that I am not interested in solving MX = Y anymore, but rather NX = Y where N = M + u.Tv where u,v are columns vectors of size n and where Tv means "transposed of v". Solving NX = Y is equivalent to solve MX = Y, MX = u and MX = v, and is "almost" not more expensive than solving MX = Y when the rank of the matrix u.Tv is small wrt n. (In my case this rank is inferior or equal to 2.)

# Pardiso - phase 11 errors

Thank you for taking the time to look into this problem with me. I am trying to implement the Pardiso solver for a nonlinear finite element code, particularly fluid flow, and I seem to be getting stuck at the start.

When I compile and run I get three messages:

# install bombed immediately

unpacked the trial tar and tried to run install.sh - recieved the following error:

line 395: 19706 Segmentation fault      (core dumped) \$pset_engine_cli_binary --tmp_dir=\$user_tmp --TEMP_FOLDER=\$temp_folder --log_file=\$strings_log_file --string_ids="\$strings_list" --__get_string__=\$strings_file --LANG=\$user_lang --PSET_MODE=install
./install.sh: line 398: /tmp/intel.pset.strings.scmcduff.preston: No such file or directory

I do have write permissions to /tmp

Any help would be greatly appreciated - thanks

Sean

# Memory error in complex eigensolver (

SUMMARY:

===================

Deleting objects after successful calls to lapack_zhptrd, lapack_dstebz, lapack_zstein, and lapack_upmtr crashes with an error in free().

DESCRIPTION

============================

I'm writing an eigenvalue equation solver for complex hermitian matrices. Code is attached. It uses the sequence lapack_zhptrd, lapack_dstebz, lapack_zstein, and lapack_upmtr to

1) generate the symmetric matrix T = Q^H A Q, with a unitary matrix Q. (lapack_zhptrd)

2.) Get m smallest eigenvalues for T (lapack_dstebz)

# ddot function returning single precision result

Hi everyone,

I have the following function in a Fortran project (built in the Visual studio 2010 environment + Intel Compiler XE) that uses one of the MKL library functions (the ddot function that computes a vector-vector dot product):

*******************************************************************************************************************************************************