Intel® MKL 10.2 Fixes List

The tables below summarize specific customer issues or feature requests that have been addressed by the indicated product releases.  Some items span multiple architectures and/or operating systems and some relate only to a single architecture or operating system.

NOTE: The issues, defects, bug reports, and feature requests described below represent specific issues with specific test cases.  An item listed here does not imply that it necessarily applies to your application(s).  If your situation does not match the specific test case you may not have experienced the error or problem associated with that update.  It is not possible to describe the details of every issue and its specific test case in these tables.

Providing a complete description of each item in the list below is impractical. For that reason we ask that you post a message on the Intel MKL user forum, /en-us/forums/intel-math-kernel-library/, and reference this page and the "DPD" number associated with the item of interest, if you have questions regarding the information in these tables.

These tables are updated regularly to correct any errors and omissions.

Intel MKL 10.2 update 7 (14 Dec 2010)

DPD200150365 Change default threading for Poisson library to match Intel MKL default (threaded)
DPD200192258 MKL DSS API should support single precision inputs with F90 interface
DPD200193353 DSS out-of-core (OOC) fails for small matrices with message "MKL-DSS-DSS-Error, An unclassifiable error occurred, dss_factor_real failed, return code = -3".
DPD200193362 PARDISO OOC fails when solving large problems on the 64-bit Windows 7 operating system.


Intel MKL 10.2 update 6 (30 Aug 2010)

DPD200092519 mkl_dcscsv produces wrong results for implicit unit diagonal elements in a upper triangular matrix
DPD200186239 Problem with multi-threading in mkl_dcsrsymv()
DPD200094225 ScaLAPACK - undefined reference to `slasrt2'
DPD200092145 PARDISO - Metis doesn't support 64bit integer
DPD200095301 CGESDD returns non-unitary U matrix
DPD200096322 fftw2x_cdft wrapper problems for multidimensional transforms and use of fftwnd_f77_mpi_create_plan()
DPD200094153 Non-optimal descriptor commit performance FFTW2x wrapper use in loops
DPD200087228 PARDISO crashes while mtypes == 11 or -2 are solving when macfct > 1
DPD200156141 ZGELSD throws an unhandled exception for 100x100 matrix
DPD200157417 Intel MKL crash on 256 or greater core count systems for Windows and Linux


Intel MKL 10.2 update 5 (23 Apr 2010)

DPD200148543 Degradation in MKL/DGEMM performance on large-scale NUMA system based on Xeon/Nehalem
DPD200090099 Illegal instruction. 0x0000000000405406 in mkl_vml_core_e9_vml_dSub ()
DPD200085701 MKL leaks memory even when freeing buffers with mkl_free_buffers() after each MKL operation
DPD200141955 Multiple calls to CGGES return multiple results when multi-threading is enabled
DPD200087662 Incorrect prototype for dcgmrhs_init(..)
DPD200149262 OOC DSS crash in factorization phase
DPD200090882 Error code is misleading when running PARDISO out-of-core
DPD200090101 Illegal instruction Intel® Core™2 Duo CPU E8400 for VML function
DPD200091604 Threading functions crash on systems reporting greater than 32 cores
DPD200085780 Pardiso’s C prototype should be changed
DPD200083074 mpz_mul function produces wrong output for very large multiplicands
DPD200089153 PARDISO OOC solver doesn't return a correct error flag
DPD200150187 Error in algorithm for working with thread number in PARDISO
DPD200090490 PARDISO - incorrect behaviour in deciding whether to use in-core or OOC mode


Intel MKL 10.2 update 4 (19 Feb 2010)

DPD200087809 Memory leak in OOC PARDISO
DPD200087031 Memory leak in PDSYGVX
DPD200084671 zgeev code crash with MKL_DISABLE_FAST_MM on
DPD200087322 LAPACK dsygv fuction produces different results when run on one or two threads
DPD200130491 Compute convolution directly (VSL_CONV_MODE_DIRECT) - performance issue
DPD200087174 Duplicated symbols ( ippGenNumCoresOndie) caused the crash
DPD200141956 DGEQRF returns incorrect workspace size upon query
DPD200136191 Threaded BLAS library runs slower than sequential one
DPD200086305 mkl_dcsrmm produces different output when threading / sequential linking
DPD200134986 FFT 2D complex performance not what expected
DPD200149061 Wrong answers from FFTs for non-square matrices when using FFTW3 interface
DPD200134402 DGELSD never stops
DPD200087222 pardiso_unsym_complex_c.c throws an exception when iparm(4) is not a multiple of 10.
DPD200086951 Incorrect error code for VML function vdCosh


Intel MKL 10.2 update 3 (14 Dec 2009)

DPD200082622 Segfault in vsRngUniform with sobol QRNG
DPD200083045 the file mkl_gmp.h -- While compiling C++ code with this header file included, compile time error occured
DPD200083471 1d FFT crashes with the error_message = MKL DFTI ERROR: Not enough memory to allocate
DPD200084053 MKL installation permission errors
DPD200084696 mkl_deallocate doesn't work if an double precision array of length more than 10,000,000 is allocated
DPD200084747 DFT not freeing internal memory in MKL 10.2.1
DPD200084777 mp_linpack seg fault with 16G memory per processors
DPD200084797 MKLROOT is an environment variable in the bourne shell configuration, but it is a shell variable in the csh configuration
DPD200085170 MKL java example for Intel 64 architecture is missing
DPD200085385 parallel double precision, complex, out-of-place FFT of size 1024 fails on 8-byte aligned data (Win/IA-32)
DPD200085935 vzArg returns wrong values
DPD200086205 segmentation fault when calling dsyevr
DPD200086305 mkl_dcsrmm produces different output when threading / sequential linking
DPD200086454 Unresolved reference in SGI MPT BLACS for Linux/Intel 64
DPD200086513 Solution of Nonlinear Least Squares Problem with boundary constraints goes out of bounds
DPD200086955 cblas_zhemv and cblas_zgerc error when x step is not 1
DPD200087348 segfaults with DSTEMR
DPD200134619 mkl_dcscmm computes all of C = alpha*A*B + beta* C even when beta=0
DPD200139153 makefile for mkl_lapack95.lib fails due to space in install_dir name
DPD200139584 Incorrect Poisson generator for 1D sequence
DPD200139924 LAPACK does not have any exceptional shift strategy for QZ
DPD200140261 MKL Dependency on Intel compiler library
DPD200141616 Eigenvector bug in multi-threaded code of ?STEQR
DPD200141956 DGEQRF returns incorrect workspace size upon query
DPD200084290 Low performance on 3D FFT with DFTI_NUMBER_OF_TRANSFORMS > 1
DPD200086472 Small-size DGEMM performance issue on for 8 threads
DPD200130881 Need optimzation for 32-bit version of LAPACK for medium size systems
DPD200138401 Distributed FFT doesn't pass verification on large vector sizes


Intel MKL 10.2 update 2 (16 Sep 2009)

DPD200082886 DFTI_ERROR_MESSAGE_EXTERNAL interface declaration issue
DPD200083603 Pardiso error performing several numerical factorizations in ooc mode
DPD200083759 Segmentation fault in dsyevr
DPD200083938 PARDISO: 32-bit in-core and out-of-core fail during numerical factorization when allocating > 2GB memory block
DPD200083947 DSPEV gives erroneous results
DPD200084033 function dgeqp3() will hang when the input matrix has NaN
DPD200121194 (z/c/s)gemm on IPF made incorrect stack modifications causing Global Register allocation errors
DPD200134123 Thread safety issue with DGELSD
DPD200134371 Fortran interface for mkl_?imatcopy does not work
DPD200134733 dss_reorder hangs program
DPD200135190 Cross process synchronisation causes slowdown in FFTs
DPD200135196 Zheevd produces different output 'info' on different systems.
DPD200135401 cgs iteration always crashes in pardiso
DPD200135409 MKL_DCSRCOO routine produces the wrong output
DPD200135919 DFTI returns incorrect results when descriptor is shared by more than one thread
DPD200136968 ZGEMM crash on Itanuim processors
DPD200140030 Intel MKL for Mac OS X is broken on latest OS tools
DPD200082802 Unexpected scalability results with mkl_csplblas_dcsrsymv
DPD200083870 CPU dispatch problem for P4/Xeon result in bad GFLOPS
DPD200132532 PARDISO OOC performance for very large problems
DPD200137664 bad performance on DPOTRF with input size 101



For more complete information about compiler optimizations, see our Optimization Notice.