topic meawiile , can quickly check in Intel® oneAPI Math Kernel Library

illegal value in DESCINIT and PSGESV

Arrigoni__Viviana — Mon, 01 Apr 2019 17:13:08 GMT

I am new to scaLapack and I am trying to execute the program ex1.c in this link:

http://geco.mines.edu/software/mkl/index.shtml

when I run it, these two messages are printed: "DESCINIT parameter number 9 had an illegal value", "and PSGESV parameter number 1 had an illegal value".
I don't see why those values are illegal, can someone help? Thanks.

Are you talking about psgesv

Gennady_F_Intel — Tue, 02 Apr 2019 02:28:07 GMT

Are you talking about psgesv.f90 example? How did you build this case and how did you run then? Do you still use MKL v.11?

No I am talking about ex1.c.

Arrigoni__Viviana — Tue, 02 Apr 2019 08:50:08 GMT

No I am talking about ex1.c. The program doesn't take any input, all variables are defined within the main. I simply compiled it using mpiicc as follows:

mpiicc -DMKL_ILP64 -I${MKLROOT}/include ex1_scalapack.c -o ex1_scalapack -L${MKLROOT}/lib/intel64 -lmkl_scalapack_ilp64 -lmkl_intel_ilp64 -lmkl_sequential -lmkl_core -lmkl_blacs_intelmpi_ilp64 -lpthread -lm -ldl

Since the code is designed to run using 4 processing cores, I ran it as:

mpirun -np ./ex1_scalapack

the program prints the two error messages that I said in the first post.
Thank you!

What version of mkl did you

Gennady_F_Intel — Tue, 02 Apr 2019 13:22:00 GMT

What version of mkl did you try?

It says MKL 2018, don't know

Arrigoni__Viviana — Tue, 02 Apr 2019 14:08:00 GMT

It ought to be MKL 2018.5.274, I am working on a cluster.

The latest available version

Gennady_F_Intel — Tue, 02 Apr 2019 15:06:46 GMT

The latest available version is 2019 u3. Usually we post the announcement of the latest release on the top of this forum:https://software.intel.com/en-us/forums/intel-math-kernel-library/topic/805788

Yes, we will check your case with this latest update.

meawiile , can quickly check

Gennady_F_Intel — Tue, 02 Apr 2019 15:16:34 GMT

meawiile , can quickly check if the problem still exists in the case if you will use LP64 API instead on ILP64?

-lmkl_scalapack_lp64 -lmkl_intel_lp64 -lmkl_sequential -lmkl_core -lmkl_blacs_intelmpi_lp64 -lpthread -lm -ldl

Yes! Thank you very much :)

Arrigoni__Viviana — Tue, 02 Apr 2019 15:53:46 GMT

Yes! Thank you very much :)