What's new in Engineering and Scientific Subroutine Library (ESSL) Version 6.2

March 2020

• ESSL 6.2.1.1 supports IBM® POWER8® servers with Red Hat Enterprise Linux® 7.7 (Little Endian (LE) mode)
• CUDA Toolkit 10.2 (POWER8 P100 only)

November 2019

• CUDA Toolkit 10.1 Update 2
• Performance improvements for some single precision Fourier transform lengths when using Power9 CPUs.
• BLAS and CBLAS subroutines
  • DGEMMT (Combined Matrix Multiplication and Addition for General Matrices, Their Transposes, or Conjugate Transposes but update only the upper or lower triangular portion of the Array C)
• LAPACK and LAPACKE subroutines
  • SGEQP3, DGEQP3, CGEQP3, and ZGEQP3 (General Matrix QR Factorization)
• Added information about installing ESSL Version 6.2.1 in the following topics:
  • Install all ESSL 6.2.1 packages when any previous version of ESSL is installed, and upgrade all previous versions of ESSL to ESSL 6.2.1
  • Install all ESSL 6.2.1 packages when any previous version of ESSL is installed, and retain multiple releases or program temporary fixes of ESSL
  • Install all ESSL 6.2.1 packages except the license package to a different directory
• Updated installation information in the following topics:
  • Software requirements
  • Disk space requirement
  • Installing ESSL for Linux on POWER® xCAT kits
  • List of product files

June 2019

A new product, the IBM Engineering and Scientific Subroutine Library on OpenPOWER, V6.2 (product number: 5765-R13). The IBM Engineering and Scientific Subroutine Library on OpenPOWER,V6.2 provides support for OpenPower servers. For more information on supported hardware products, see Hardware Products Supported by ESSL.

May 2019

April 2019

August 2018

The following information is a summary of the updates that are made to the ESSL Version 6.1 documentation:

• ESSL Version 6.1 supports the following IBM Power Systems:
  • IBM POWER9 servers
    Notes:

    • The ESSL SMP CUDA Library is supported only on the IBM Power Systems AC922 (8335-GTC and 8335-GTW) servers that include Tesla V100 with NVLink GPUs.
    • Red Hat Enterprise Linux 7.5 for Power Little Endian (LE) (POWER9)
  • IBM POWER8™ servers
    Notes:

    • The ESSL SMP CUDA Library is supported only on IBM Power System S822LC (8335-GTB) servers with NVIDIA P100 GPUs.
    • Red Hat Enterprise Linux 7.5 (Little Endian (LE) mode)
• ESSL Version 6.1 supports clusters of IBM Power System AC922 (8335-GTC and 8335-GTW) servers that are running Red Hat Enterprise Linux 7.5 for Power Little Endian (Power9) with or without Tesla V100 with NVLink GPUs and dual port InfiniBand EDR (ConnectX-5, PCIe Gen4) adapters that are interconnected with EDR InfiniBand switches.
• Added information about upgrading in the Upgrading from ESSL 5.5 topic.
• Added information about installing ESSL Version 6.1 in the following topics:
  • Install all ESSL 6.1 packages when any previous version of ESSL is installed, and upgrade all previous versions of ESSL to ESSL 6.1
  • Install all ESSL 6.1 packages when any previous version of ESSL is installed, and retain multiple releases (or program temporary fixes (PTFs)) of ESSL
  • Installing ESSL xCAT kits
• Updated various information in the following topics:
  • Software requirements
  • Disk space requirement
  • Uninstalling ESSL
  • List of product files
• BLAS and CBLAS subroutines:
  • SAXPBY and DAXPBY (Scales Vector X and Vector Y by two Scalars, Adds them to one another, and Stores in the Vector Y)
• Added information about the following error messages:
  • 2538-2401 and 2538-2402 error messages in the Resource error messages topic.
  • 2538-2701 error message in the Miscellaneous error messages topic.
• LAPACK and LAPACKE subroutines:
  • CLANSY and ZLANSY (Complex Symmetric Matrix Norm)
  • SSYEVR, DSYEVR, CHEEVR, and ZHEEVR (Eigenvalues, and Optionally, the Eigenvectors of a Real Symmetric or Complex Hermitian Matrix)
  • SSYGVD, DSYGVD, CHEGVD, and ZHEGVD (Eigenvalues, and Optionally, the Eigenvectors of a Positive Definite Real Symmetric or Complex Hermitian Generalized Eigenproblem Using a Divide-and-Conquer Algorithm)
  • SSYGV, DSYGV, CHEGV, and ZHEGV (Eigenvalues, and Optionally, the Eigenvectors of a Positive Definite Real Symmetric or Complex Hermitian Generalized Eigenproblem)
• LAPACK GPU support:
  • SGESV, DGESV, CGESV, and ZGESV (General Matrix Factorization and Multiple Right-Hand Side Solve)
  • SGETRF, DGETRF, CGETRF, and ZGETRF (General Matrix Factorization)
  • SGETRS, DGETRS, CGETRS, and ZGETRS (General Matrix Multiple Right-Hand Side Solve)
  • SPOSV, DPOSV, CPOSV, and ZPOSV (Positive Definite Real Symmetric or Complex Hermitian Matrix Factorization and Multiple Right-Hand Side Solve)
  • SPOTRF, DPOTRF, CPOTRF, and ZPOTRF (Positive Definite Real Symmetric or Complex Hermitian Matrix Factorization)
  • SPOTRS, DPOTRS, CPOTRS, and ZPOTRS (Positive Definite Real Symmetric or Complex Hermitian Matrix Multiple Right-Hand Side Solve)
  • SPPSV, DPPSV, CPPSV, and ZPPSV (Positive Definite Real Symmetric or Complex Hermitian Matrix Factorization and Multiple Right-Hand Side Solve)
  • SPPTRF, DPPTRF, CPPTRF, and ZPPTRF (Positive Definite Real Symmetric or Complex Hermitian Matrix Factorization)
  • SPPTRS, DPPTRS, CPPTRS, and ZPPTRS (Positive Definite Real Symmetric or Complex Hermitian Matrix Multiple Right-Hand Side Solve)
  • SGEQRF, DGEQRF, CGEQRF, and ZGEQRF (General Matrix QR Factorization)
  • SGELS, DGELS, CGELS, and ZGELS (Linear Least Squares Solution for a General Matrix)
• Fourier Transforms GPU support:
  • SCFTD and DCFTD (Multidimensional Complex Fourier Transform)
  • SRCFTD and DRCFTD (Multidimensional Real-to-Complex Fourier Transform)
  • SCRFTD and DCRFTD (Multidimensional Complex-to-Real Fourier Transform)