Intel® MKL 11.1 Bug Fixes

Intel® MKL 11.1  Update 4 (21 Aug 2014)

DPD200523357pardiso_unsym_complex_c.c pardiso_unsym_csc_c.c are now changed according to CSR format
DPD200527600Description of the program mkl\examples\mic_offload\solverc\source\cg_jacobi_precon_c.c is now corrected
DPD200346477Added const modifier for ALL CBLAS level 2 routines
DPD200252589Improved MKL DGER Performance
DPD200532854DGEQRF now produces correct output when you set mkl_set_dynamic(false) similar to mkl_set_dynamic(true)
DPD200515841DCSRILU0 now gives correct results when -openmp and -check=stack flags are used
DPD200524415vRngBinomial now has ntrial > 0
DPD200524018Implemented 'rwork' parameter for C interfaces of ScaLapack routines PZGESVD/PCGESVD
DPD200399852Introduced threading version of daxpby
DPD200256290Fixed the problem of routine strnlspbc_solve giving divide by Zero error
DPD200521692Fixed the MKL Crash on machines with AMD Jaguar processor
DPD200530039Fixed the PARDISO solver hang
DPD200228633Fixed the SVD threading problem i.e, CPU utilization is low in non master threads
DPD200248270DSYGVD on Intel® Many Integrated Core Architecture (Intel® MIC Architecture) returns incorrect results when using selected numbers of threads

Intel® MKL 11.1  Update 3 (06 May 2014)

DPD200387939Extended Eigensolver routine message should be improved
DPD200387946Fixed dfeast_scsrgv returning error = -1
DPD200405570Fixed PARDISO crashing randomly with different messages
DPD200405571CDOTU produces segmentation fault when complied by gfortran linked with -lmkl_rt
DPD200518934Fixed the inconsistency of defining INTEGER DF_CHECK_FLAG in mkl_df_f90
DPD200347155Improved the performance of NaN Checker
DPD200253688Fixed the inconsistent behaviour of mkl_dspffrt2 and mkl_dspffrtx in handling zero pivot
DPD200401058Fixed the Error reporting in FFT OOC computation with pardiso
DPD200394086Added mkl_sdiamm example
DPD200522866Fixed build failure due to duplicate definitions of pstran and pdtran in both mkl_pblas.h and mkl_scalapack.h
DPD200407178Fixed - dfdConstruct1D returning error code _DF_NATURAL_CUBSPLINE_MIN_NX=-1004
DPD200401462Implemented Pivot control and diagonal extractions routines

Intel® MKL 11.1  Update 2 (02 Feb 2014)

DPD200398565Fixed Memory leak issue
DPD200390408ANSYS DMP support for multiple Intel® Many Integrated Core Architecture (Intel® MIC Architecture) cards
DPD200396128Performance - mkl_dcsrmm shows the scalablity degradation in 11.1 vs 11.0
DPD200248606FEAST random crashes on intel64
DPD200241919Provide optimized *GEMM (then other BLAS) for Intel® Atom™ processor
DPD200241920Provide optimized mkl_dsppfrt2/x (then other LAPACK) for Intel® Atom™ processor
DPD200247333Improve DGEMM native performance on KNC for non-square matrices
DPD200249529Improve small matrix GEMM performance for graphics and digital content creation customers
DPD200394473LAPACKE_dgesdd returns -11
DPD200249555MKL Link Line Advisor gives wrong linker flags for OS X
DPD200397430Pardiso OOC - fseek call during the backtransformation
DPD200401058FFT reports the Error in OOC computation with pardiso
DPD200349785Performance: spotrf produces the very different performance result with "U" and "L" cases
DPD200388403Replaced Matrix Order into other wording
DPD200247589Improve DGEMM performance for matrices with huge K size and tiny M and N size
DPD200388027there are no declarations of ?lagge, ?laghe and ?lagsy in any of header files
DPD200247749MKL_DSS_DIAGONAL_SOLVE opt_solve produces garbage results
DPD200395081OOC ( hybrid mode ) Pardiso should work with twol-level factorization algorithm in the case of small matrix
DPD200509794Fixed the problem of 1D DataFitting giving the wrong output
DPD200248198Fixed the Bug in MKL Examples
DPD200249444Improve GEMM performance for matrices with M=K=6 and large N
DPD200238399Unusual DGEMM performance
DPD200248270NAG Confirms DSYGVD on MIC returns incorrect results when using selected numbers of threads

Intel® MKL 11.1  Update 1 (24 Oct 2013)

DPD200385039DSS API - memory leakage on the factorization stage
DPD200247042DGEMM seg fault with PGI threading
DPD200344190pardiso problem with partial solution for right-hand
DPD200245542Need threaded DSDOT
DPD200351579Incorrect workspace size returned by query to DGESVD in MKL 11.0.5
DPD200243635Add service functions to query data fitting task structure
DPD200349748vsSub fails in compiler offload mode while trying to load CPU specific library (which MKL does not have for MIC native)
DPD200342324Statically linking to MKL creates executable stack security exposure
DPD200335069DF -- support of DF_APRIORI_MOST_LIKELY_CELL Computation Types
DPD200344380sgeqrf returning NaNs
DPD200243791out of place transpose slower than F intrinsic routine on tall skinny matrices
DPD200383963Defect: Sparse Matrix Structure leads Zero Pivot Error
DPD200346307DS: Heap Corruption happensin ?syrdb rouitine when the input size == 1
DPD200246843DSS solver gives incorrect results (each time different results)
DPD200333995optimal lwork too large in dormqr
DPD200317576pblas documentation needs to be amended to explain complex support
DPD200248173DFEAST_SCSREV crashes with OpenMP error

Intel® MKL 11.1  (3 Sept 2013)

DPD200342923Including MKL MIC library as the default installation in MKL 11.1
DPD200325646DS: zgelsd generates the error messages:"MKL INTERNAL ERROR: Condition 1 detected in function DLASD4
DPD200237716Performance regression on HPCC Star DGEMM and Star FFT on KNC
DPD200345891DS: dgeevx, dggev may return significantly different eigenvalues when eigenvectors are requested
DPD200349748vsSub fails in compiler offload mode while trying to load CPU specific library (which MKL does not have for MIC native)
DPD200245451MKL 11.0 Update 4 SIGSEGV regression vs update 2
DPD200242406unaligned daxpy is slow than plain C on an Intel® Advanced Vector Extensions 2 (Intel® AVX2) box
DPD200313133DS: performance of mkl's transposition should be improved set some un-existing paths
DPD200243422FFTW/FORTRAN: ability to set number of threads that execute a plan
DPD200344940ANSYS request for a query function
DPD200344945ANSYS Request for a "force this to offload" feature in MKL AO for MIC
DPD200245405Install Fortran 95 interfaces by default
DPD200338444Internal buffers grow unlimited while compiling 32 bit with sequential version
DPD200339085MKL memory leak of 120 bytes
For more complete information about compiler optimizations, see our Optimization Notice.