topic Hi Ying, in Intel® oneAPI Math Kernel Library

Example does not run

Georgios_S_ — Mon, 15 Jun 2015 14:41:32 GMT

I am trying to run ScaLAPACK, through MKL. Since I have the student version, I had to install MPICH2, since what I want to do is to apply an inverse of a matrix, that can be done with Cholesky in a distributed environment.

I got this ex1.c from Intell 11. It compliles fine, but it fails to execute:

gsamaras@pythagoras:~/konstantis$ ../mpich-install/bin/mpicc -o test ex1.c -I../intel/mkl/include  ../intel/mkl/lib/intel64/libmkl_scalapack_ilp64.a -Wl,--start-group ../intel/mkl/lib/intel64/libmkl_intel_ilp64.a ../intel/mkl/lib/intel64/libmkl_core.a ../intel/mkl/lib/intel64/libmkl_sequential.a -Wl,--end-group ../intel/mkl/lib/intel64/libmkl_blacs_intelmpi_ilp64.a -lpthread -lm -ldl
gsamaras@pythagoras:~/konstantis$ mpiexec -n 4 ./test
{    0,    0}:  On entry to 
DESCI{    1,    0}:  On entry to NIT parameter number    9 had an illegal value 
{    0,    0}:  On entry to 
DESCINIT parameter number    4 had an illegal value 
  0  0 vals   1.00000   0.20100   0.50100   0.60100   0.10200   2.00000   0.50200   0.60200   0.10500   0.20500   5.00000   0.60500   0.10600   0.20600   0.50600   6.00000

DESCINIT parameter number    4 had an illegal value 
  0  0 vals   1.00000   0.20100   0.50100   0.60100   0.10200   2.00000   0.50200   0.60200   0.10500   0.20500   5.00000   0.60500   0.10600   0.20600   0.50600   6.00000
{    0,    1}:  On entry to 
DESCINIT parameter number    6 had an illegal value 
{    0,    1}:  On entry to 
DESCINIT parameter number    6 had an illegal value 
  0  1 vals   0.10300   0.20300   0.50300   0.60300   0.10400   0.20400   0.50400   0.60400   0.10700   0.20700   0.50700   0.60700   0.10800   0.20800   0.50800   0.60800
{    1,    1}:  On entry to {   -1,   -1}:  On entry to 
PSGESV
 parameter number  602 had an illegal value 
*** Error in `./test': free(): invalid next size (fast): 0x00000000021a6ae0 ***
DESCINIT parameter number    6 had an illegal value 
{    1,    1}:  On entry to 
DESCINIT parameter number    6 had an illegal value 
  0  1 vals   0.10300   0.20300   0.50300   0.60300   0.10400   0.20400   0.50400   0.60400   0.10700   0.20700   0.50700   0.60700   0.10800   0.20800   0.50800   0.60800
{   -1,   -1}:  On entry to 
PSGESV parameter number  602 had an illegal value 
*** Error in `./test': free(): invalid next size (fast): 0x000000000195dae0 ***

===================================================================================
=   BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
=   PID 13140 RUNNING AT pythagoras
=   EXIT CODE: 6
=   CLEANING UP REMAINING PROCESSES
=   YOU CAN IGNORE THE BELOW CLEANUP MESSAGES
===================================================================================
YOUR APPLICATION TERMINATED WITH THE EXIT STRING: Aborted (signal 6)

Moreover, the same happens with Ying's example from here, which terrifies me:

gsamaras@pythagoras:~/konstantis$ ../mpich-install/bin/mpic++ -o test n.cpp -I../intel/mkl/include  ../intel/mkl/lib/intel64/libmkl_scalapack_ilp64.a -Wl,--start-group ../intel/mkl/lib/intel64/libmkl_intel_ilp64.a ../intel/mkl/lib/intel64/libmkl_core.a ../intel/mkl/lib/intel64/libmkl_sequential.a -Wl,--end-group ../intel/mkl/lib/intel64/libmkl_blacs_intelmpi_ilp64.a -lpthread -lm -ldl
n.cpp: In function ‘int main(int, char**)’:
n.cpp:96:48: warning: deprecated conversion from string constant to ‘char*’ [-Wwrite-strings]
     Cblacs_gridinit(&ictxt, "Row", nprow, npcol);
                                                ^
gsamaras@pythagoras:~/konstantis$ mpiexec -n 1 ./test
Returned:  Hello World! I am 0 of 1
here

===================================================================================
=   BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
=   PID 13605 RUNNING AT pythagoras
=   EXIT CODE: 8
=   CLEANING UP REMAINING PROCESSES
=   YOU CAN IGNORE THE BELOW CLEANUP MESSAGES
===================================================================================
YOUR APPLICATION TERMINATED WITH THE EXIT STRING: Floating point exception (signal 8)

and this happens at this line of the code:

printf("here\n");
    locR = numroc_(&matrix_size, &block, &myrow, &izero, &nprow);
printf("here1\n");

I wanted to execute these examples, so that I know that I am good to go for developing my code. Is there a problem with the examples (which I do not really need to run), or is there a problem with MKL's ScaLAPACK?

Hi Georgios,

Ying_H_Intel — Tue, 16 Jun 2015 08:50:30 GMT

Hi Georgios,

The int in C are supposed to be 32bit, you may try lp64 mode.

mpicc -o test_lp64 ex1.c -I/opt/intel/mkl/include /opt/intel/mkl/lib/intel64/libmkl_scalapack_lp64.a -L/opt/intel/mkl/lib/intel64 -Wl,--start-group /opt/intel/mkl/lib/intel64/libmkl_intel_lp64.a /opt/intel/mkl/lib/intel64/libmkl_core.a /opt/intel/mkl/lib/intel64/libmkl_sequential.a -Wl,--end-group /opt/intel/mkl/lib/intel64/libmkl_blacs_intelmpi_lp64.a -lpthread -lm -ldl

Best Regards,

Ying

[yhu5@prc-mic01 scalapack]$ mpirun -n 4 ./test_lp64 0 0 vals 1.00000 0.20100 0.50100 0.60100 0.10200 2.00000 0.50200 0.60200 0.10500 0.20500 5.00000 0.60500 0.10600 0.20600 0.50600 6.00000
1 0 vals 0.30100 0.40100 0.70100 0.80100 0.30200 0.40200 0.70200 0.80200 0.30500 0.40500 0.70500 0.80500 0.30600 0.40600 0.70600 0.80600
0 1 vals 0.10300 0.20300 0.50300 0.60300 0.10400 0.20400 0.50400 0.60400 0.10700 0.20700 0.50700 0.60700 0.10800 0.20800 0.50800 0.60800
1 1 vals 3.00000 0.40300 0.70300 0.80300 0.30400 4.00000 0.70400 0.80400 0.30700 0.40700 7.00000 0.80700 0.30800 0.40800 0.70800 8.00000
x=( 0 0) 0.9321 0.3773 0.0444 0.0074
x=( 1 0) 0.1924 0.0999 -0.0190 -0.0388

Hi Ying,

Georgios_S_ — Tue, 16 Jun 2015 10:40:34 GMT

Hi Ying,

thanks, that's solved the issue. Moreover, I was able to run your example, nice! Can you say me what we said to the compiler to do differently in your answer, so that I understand what I am doing.

Also I noticed that you had a part that I didn't have in the compilation command, i.e. -L/opt/intel/mkl/lib/intel64, which in my case had to be -L../intel/mkl/lib/intel64. I noticed that things were OK with and without it. Why should we use it?

Thanks,

George

PS - damn you are good, it would take me much time to find out! :)

If you have installed and set

TimP — Wed, 17 Jun 2015 00:58:07 GMT

If you have installed and set up the intel compiler paths, the L path is taken care of. Ilp64 libraries require long long int arguments while lp64 accepts 32 bit intel. N

Hi George,

Ying_H_Intel — Wed, 17 Jun 2015 01:12:35 GMT

Hi George,

You are welcome. You may find the expected link line by Link advisor: https://software.intel.com/en-us/articles/intel-mkl-link-line-advisor/

for example, it will give you

-m64 -I${MKLROOT}/include -Wl,--no-as-needed -L${MKLROOT}/lib/intel64 -lmkl_scalapack_lp64 -lmkl_intel_lp64 -lmkl_core -lmkl_sequential -lmkl_blacs_intelmpi_lp64 -lpthread -lm.

Here -L dir, means Add directory dir to the list of directories to be searched for -llibrary

And it is fine to use the explicit path of library as your command.

-I/opt/intel/mkl/include /opt/intel/mkl/lib/intel64/libmkl_scalapack_lp64.a -Wl,--start-group /opt/intel/mkl/lib/intel64/libmkl_intel_lp64.a /opt/intel/mkl/lib/intel64/libmkl_core.a /opt/intel/mkl/lib/intel64/libmkl_sequential.a -Wl,--end-group /opt/intel/mkl/lib/intel64/libmkl_blacs_intelmpi_lp64.a -lpthread -lm -ldl

So you are exactly right the "-L../intel/mkl/lib/intel6" is not needed with the explicit path, please ignore it.

Best Regards,

Ying