Intel® Math Kernel Library (Intel® MKL) 2018 Bug Fixes list

By Gennady Fedorov, Published: 08/25/2017, Last Updated: 05/29/2018

Intel(R) MKL 2018 Update 4 ( September 2018)

MKLD-3796 Fixed an issue with the ?GEEV routine producing segmentation fault when user allocates required workspace. 
MKLD-3794 Renamed macro name DEPRECATED to MKL_DEPRECATED.
MKLD-3793 Fixed an issue with the ?GEEV routine producing an error related to incorrect value of Parameter 2 for ZGEHD2.
MKLD-3786 Fixed an issue with the P[S,D,C,Z]GEMM routine returning incorrect result under specific conditions.
MKLD-3736 Fixed an issue with the P[S,D,C,Z]GESVD routine producing an error related to illegal value for Parameter 19. 
MKLD-3725 Fixed an issue with the [S,D,C,Z]]SYEVR routine producing an error related to incorrect value of Parameter 6 for DSTEIN. 
MKLD-3719 Improved the performance of Parallel Direct Sparse Solver for Clusters in the backsubstitution phase for multiple right-hand sides.
MKLD-3690 Fixed an issue with  2D FFT routine producing incorrect results when MKL_DYNAMIC is set to ‘false’.
MKLD-3686 Fixed an issue with ScaLAPACK SVD single precision (PSGESVD) routine returning incorrect size of the work array. 
 MKLD-3684 Fixed an issue with build operation of custom DLL for cluster components requiring modifications to makefile depending on OS.
MKLD-3681 Fixed an issue with [S,D,C,Z]GESDD routines not  converging for specific input matrices.
MKLD-3613 Fixed an issue with building FFTW3 interfaces when makefile does not allow multiple jobs using the -j X command line argument.

Intel(R) MKL 2018 Update 3 ( May 2018)

MKLD-3534 Added descriptions of MKL_LAPACK_[s,d,c,z]_TREVC3 routines in the MKL Developer Reference manual.
MKLD-3528 Fixed an issue with the DSYEVR routine producing "Intel MKL ERROR: Parameter 12 was incorrect on entry to DSBRDB" message when problem size >= 6144
MKLD-3459 Introduced support of cross index-layout combinations for BSR matrix format of MV, ADD, SPMM, SP2M, SYRK, and SYPR Sparse BLAS Routines.
MKLD-3570 Fixed the performance regression of DGEMV function for specific problem sizes.
MKLD-3645 Fixed a performance problem of MKL_SPARSE_[S,D,C,Z]_ADD routine.
MKLD-3378 Fixed an issue with Intel PARDISO returning the wrong number line of the first pivot elements when number of threads > 1.
MKLD-2729 Introduced Schur complement matrix in sparse representation format by Intel PARDISO.

Intel(R) MKL 2018 Update 2 ( March 2018)

MKLD-3390 Fixed inconsistent behavior in MKL_LINK_TOOL routine.
MKLD-3532 Fixed failure in CLUSTER_SPARSE_SOLVER routine with IPARM(2) = 10
MKLD-3531 Introduced triple product functionality A*B*AT with sparse/dense matrix B.  
MKLD-3458 Fixed crash on exit after calling Intel® MKL Vector Mathematics. Functions.
MKLD-3497 Fixed issue of P?HEEV routine producing incorrect size of LRWORK. 
MKLD-3446 Improved performance of MKL FFT routines in the case of few threads. 
MKLD-3445 Fixed run-time failure of P?GEMM routine for specific problem sizes. 
MKLD-3424 Improved performance of MKL LAPACK_CPOSV_WORK routine for Intel® Xeon® processor supporting Intel® Advanced Vector Extensions 512 (Intel® AVX-512)  instruction set.
MKLD-3414 Fixed issue with MKL BLACS_EXIT() routine causing unstable behavior. 
MKLD-3364 Fixed runtime problem failure of vslSaveStreamM and vslLoadStreamM with MKL Niederreiter BRNG routine.
MKLD-3362 Improved performance of MKL DGBTRS routine for multiple right hand sides (RHS).
MKLD-3229 Fixed issue with Parallel Direct Sparse Solver for Clusters routine failures on the release all internal solver memory stage. 
MKLD-3217 Fixed hanging problem with MKL PDTRAN routine when number of MPI processes exceeds 10000.
MKLD-2846 Introduced support for Schur complement (dense), partial solving, and customer reordering in Parallel Direct Sparse Solver for Clusters.
MKLD-2776 Improved the performance of Intel MKL Pardiso routine in the case of multiple RHS and linking with Intel® Threaded Building Blocks (Intel® TBB) threading layer.
MKLD-2750 Improved performance of MKL FFT routines for specific problem sizes on Intel® Xeon Phi™ processors based on Intel® Advanced Vector Extensions 512 (Intel® AVX-512).
MKLD-2997 Improved performance of MKL ?SYTRI routine.
MKLD-240 Introduced support of OOC mode for Parallel Direct Sparse Solver for Clusters routine. 
MKLD-209 Improved performance of MKL ?SYGVX routine.
MKLD-60  Improved performance of MKL P?SYEVR routine.
  Fixed issue when double precision complex SVD workspace query returns too small workspace size

Intel(R) MKL 2018 Update 1 ( November 2017) 

MKLD-3387 Fixed the problem with pdpotri routine which produces wrong results with some specific sizes and environment
MKLD-3350 Fixed the problem with LAPACKE_dgesvd routine with some specific input parameters (jobu = 'N' and jobvt = 'A') and when routine returned "Intel MKL ERROR: Parameter 11 was incorrect on entry to DGEMQR" in output
MKLD-3292 Fixed the overflow exception problem with dgemm routine when the product of doubles exceeds the maximum floating point value allowed
MKLD-3233 Fixed the MKL FATAL Error on loading function mkl_vml_serv_threader_c_1i_2o issue
MKLD-3182 Fixed the wrong result PZGETRF routine issue with some specific version of Intel MPI
MKLD_3169 Fixed the problem with Parallel Direct Sparse Solver for Clusters failure when using MPI Nested Dissection Algorithm ( error = -2) 
MKLD-3116 Introduced an option --quiet to mkl_link_tool routine
MKLD-3090 Introduced LAPACKE_set_nancheck routine for disabling/enabling nan checks in LAPACKE functions
MKLD-3072 Improved the MKL FFT performance with and without scaling factor across all domains
MKLD-2740 Improved the MKL FFT performance for batched real-to-complex 3D for  Intel® Xeon® Processor supporting Intel® Advanced Vector Extensions 512 (Intel® AVX-512) (codename Skylake Server) 

Intel(R) MKL 2018 ( August 25, 2017)

MKLD-3074 Added LAPACKE ?larcm function description to the Developer Reference for MKL 2018 - C
MKLD-3053 Fixed the wrong result issue of IGEMM_s16s16s32 routine when it linked with OpenMP library statically
MKLD-3054 Improved  the performance of cblas_gemm_s16s16s32 routine
MKLD-2999 Fixed a problem with LAPACKE_dstein routine when it uses nth element of the off-diagonal array (which is supposed to have n-1 elements)
MKLD-2769 Improved the performnce of SGEMM routine for some specific problem size on Intel® Xeon Phi™  processors based on Intel® Advanced Vector Extensions 512 (Intel® AVX-512)
MKLD-2767 Fixed the significant performance degradation of DGETRF routine for some specific problem sizes on Intel® AVX2 systems
MKLD-2753 Fixed a problem when MKL FFT is used via FFTW Interface
MKLD-2948 Fixed the wrong result issue when calling zgetrf by MKL_DIRECT_CALL
MKLD-2740 Improved a Intel MKL FFT 3D problem performance on Intel® Xeon Phi™ product family x200
MKLD-2733 Fixed the _MSC_VER mismatch again MKL installation problem
MKLD-2924 Fixed a problem when MKL Automatic Offload mode doesn’t work if LD_LIBRARY_PATH contains incompatibility paths
MKLD-271 Improved the performance of functions DORCSD2BY1 routine

 

Product and Performance Information

1

Intel's compilers may or may not optimize to the same degree for non-Intel microprocessors for optimizations that are not unique to Intel microprocessors. These optimizations include SSE2, SSE3, and SSSE3 instruction sets and other optimizations. Intel does not guarantee the availability, functionality, or effectiveness of any optimization on microprocessors not manufactured by Intel. Microprocessor-dependent optimizations in this product are intended for use with Intel microprocessors. Certain optimizations not specific to Intel microarchitecture are reserved for Intel microprocessors. Please refer to the applicable product User and Reference Guides for more information regarding the specific instruction sets covered by this notice.

Notice revision #20110804