Uninitialized partial memory access with LAPACKE_cheev

[Chandrakant_19]

Hi Intel Support Team,

I am working on a project using mkl library. I am using the 'LAPACKE_cheev' mkl routine for eigenvalue(s) computation.

I have initilaized the input parameters before before calling the api.

The input matrix dimensions are 2x2 and Layout is LAPACK_ROW_MAJOR.

Code Snippet:
MKL_INT n = 2, lda = 2;
float w[2] = {};
MKL_Complex8 a[4] = {};

a is overwritten with actual data.

LAPACKE_cheev( LAPACK_ROW_MAJOR, 'N', 'U', n, a, lda, w )

Reference example code used: https://www.intel.com/content/www/us/en/docs/onemkl/code-samples-lapack/2024-1/lapacke-cheev-example-c-row.html

On valgrind-3.16.1 in Linux envirounment, we are observing following error

"Conditional jump or move depends on uninitialised value(s)" at mkl_lapack_ps_avx2_xclanhe.

On Intel Inspector in windows envirounment, we are observing following error

"Uninitialized partial memory access"

Thanking you.

[Gennady_F_Intel]

it might be the false positive message ...

What version of MKL do you use? is that 2024u1?

and let us know the command you used while calling Valgrind?

--Gennady

[Chandrakant_19]

Hi,

Please find below details about compilers and libraries.

IntelSWTools (Linux) - compilers_and_libraries_2016.2.181

IntelSWTools (Windows) - compilers_and_libraries_2017.1.143

Intel Inspector 2023.0

Command used while calling valgrind on linux,

valgrind --tool=memcheck --track-origins=yes --leak-check=full

Thanking You.

[Gennady_F_Intel]

I think this is kind of false positive issue. The version of MKL is no longer supported, Could You try to test the latest ( current ) one?

You could take this version from MKL Product Page as stand-alone package.

--Gennady

[Gennady_F_Intel · Accepted solution]

making the same experiment with the current version of oneMKL 2024.1

$ echo $MKLROOT

/opt/intel/oneapi/mkl/2024.1

and Valgrind version :

$ valgrind --version

valgrind-3.18.1

Compiling like as follow:

$ icx -qmkl lapacke_cheev_row.c

$ valgrind --tool=memcheck --track-origins=yes --leak-check=full ./a.out

==3398344== Memcheck, a memory error detector

==3398344== Copyright (C) 2002-2017, and GNU GPL'd, by Julian Seward et al.

==3398344== Using Valgrind-3.18.1 and LibVEX; rerun with -h for copyright i                                       nfo

==3398344== Command: ./a.out

==3398344==

LAPACKE_cheev (row-major, high-level) Example Program Results

 Eigenvalues

 -16.00 -6.76  6.67 25.51

 Eigenvectors (stored columnwise)

 ( 0.34, 0.00) ( -0.55, 0.00) ( 0.31, 0.00) ( -0.70, -0.00)

 ( 0.44, -0.54) ( 0.26, 0.18) ( 0.45, 0.29) ( 0.22, -0.28)

 ( -0.48, -0.37) ( -0.52, -0.02) ( -0.05, 0.57) ( 0.15, 0.08)

 ( 0.10, -0.12) ( -0.50, 0.28) ( -0.23, -0.48) ( 0.34, -0.49)

==3398344==

==3398344== HEAP SUMMARY:

==3398344==   in use at exit: 11,803 bytes in 31 blocks

==3398344==  total heap usage: 96 allocs, 65 frees, 175,735 bytes allocate                                       d

==3398344==

==3398344== LEAK SUMMARY:

==3398344==  definitely lost: 0 bytes in 0 blocks

==3398344==  indirectly lost: 0 bytes in 0 blocks

==3398344==   possibly lost: 0 bytes in 0 blocks

==3398344==  still reachable: 11,803 bytes in 31 blocks

==3398344==     suppressed: 0 bytes in 0 blocks

==3398344== Reachable blocks (those to which a pointer was found) are not s                                       hown.

==3398344== To see them, rerun with: --leak-check=full --show-leak-kinds=al                                       l

==3398344==

==3398344== For lists of detected and suppressed errors, rerun with: -s

==3398344== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 0 from 0)

[Gennady_F_Intel]

Compiler version used:

Intel(R) oneAPI DPC++/C++ Compiler 2024.1.0 (2024.1.0.20240308)

Target: x86_64-unknown-linux-gnu

Thread model: posix

InstalledDir: /opt/intel/oneapi/compiler/2024.1/bin/compiler

Configuration file: /opt/intel/oneapi/compiler/2024.1/bin/compiler/../icx.cfg

[Gennady_F_Intel]

the case is closing.