crashed in pardiso ( mkl 10.2)

[minifat]

Hello,

After update to mkl 10.2, the application crashed in pardiso, which was working well under 10.1.

When set MKL_NUM_THREADS=1 the application works also.

My system is Intel Core2 Quad CPU Q9400, CentOS release 5.3 (Final)

[power@localhost se2]$ gcc -v
Using built-in specs.
Target: x86_64-unknown-linux-gnu
Configured with: ../gcc-4.4.0/configure --prefix=/opt --enable-shared --enable-threads=posix --disable-checking --with-system-zlib --enable-__cxa_atexit --disable-libunwind-exceptions --enable-languages=c,c++ --with-mpfr=/usr/local
Thread model: posix
gcc version 4.4.0 (GCC)

The output information is as following

*** glibc detected *** ./app: free(): invalid next size (normal): 0x00002aaaac05bbd0 ***
======= Backtrace: =========
/lib64/libc.so.6[0x324da71ce2]
/lib64/libc.so.6(cfree+0x8c)[0x324da7590c]
./app(mkl_serv_mkl_free+0x1d)[0x528681]
./app[0x683dff]
./app(mkl_pds_lp64_sfinit_pardiso+0xf7)[0x550f73]
./app(mkl_pds_lp64_reorder1_pardiso+0xa42)[0x552816]
./app(mkl_pds_lp64_do_all_pardiso_fc+0x780e)[0x51eed2]
./app(mkl_pds_lp64_pardiso_c+0x305f)[0x4f5be7]
./app(mkl_pds_lp64_pardiso+0x44f)[0x4e2c8f]
./app(pardiso+0x2e)[0x4e283a]
/home/power/lib/libsparse_solve.so.1(_ZN13CSparseSolve213factorizationEPiS0_Pdi+0x1b3)[0x2b255b5626ef]
/home/power/lib/libsparse_solve.so.1(_ZN13CSparseSolve213factorizationER12CSparseStorei+0x40)[0x2b255b562534]
./app(_ZN15CSEFastDecouple5factQERSt3mapIiS0_IidSt4lessIiESaISt4pairIKidEEES2_SaIS3_IS4_S7_EEER12CSparseStore+0x1a1)[0x48d67f]
./app[0x49a6a5]
/opt/lib64/libgomp.so.1[0x2b255b9b0172]
/lib64/libpthread.so.0[0x324e606367]
/lib64/libc.so.6(clone+0x6d)[0x324dad2f7d]

[Gennady_F_Intel]

This is an unknown problem for us, how can we reproduce the problem?
What's your linking line? Can you get the test case? ( you can use the private thread for that).
I have to remind you that CentOS doesn't support officially by MKL but nevertheless it should works ..
--Gennady

[minifat]

Quoting - Gennady Fedorov (Intel)

Thank you for your information.
When I checked the linking line, I found a little diffrent with the example. After erase the -lgomp, It works::)

This is an unknown problem for us, how can we reproduce the problem?
What's your linking line? Can you get the test case? ( you can use the private thread for that).
I have to remind you that CentOS doesn't support officially by MKL but nevertheless it should works ..
--Gennady

[Gennady_F_Intel]

Hello minifat,
Can you please send us a reproducible test case?
--Gennady

[jbp1atdukeedu]

I came across the same problem when linking C code into the Fortran/BLAS interface (also using Centos, but with newer MKL code -- original post is now 1 year old). Gennady -- I can probably forward a reproducible test code if desired.

[Artem_V_Intel]

Hello,

Please sendyour test case, it will be very useful for reproducing this problem. Anyway, did you check your link line? This KB article may be helpful.

Best regards,
Artem