Differences

This shows you the differences between two versions of the page.

--- build:linear_algebra [2017/04/11 08:45] – Yann Pouillon
+++ build:linear_algebra [2019/08/26 12:29] (current) – Jean-Michel Beuken
@@ Line 1: / Line 1: @@
-<WRAP important>**IMPORTANT** \\ The information provided in this page is valid for Abinit 8.1 and future versions only.</WRAP>
+<WRAP important>**IMPORTANT** \\ The information provided in this page is valid for Abinit 8.5 and future versions only.</WRAP>
 ====== Configuring linear algebra ======
@@ Line 18: / Line 18: @@
 ===== Linear algebra flavors =====
-For sake of clarity, we provide here 3 different perspectives on the linear algebra flavors supported by Abinit and their relationships with the other concepts. The build system lets you configure linear algebra from each of these 3 perspectives, mainly by setting its //with_linalg_flavor// option in different ways (explained in detail below).
+For sake of clarity, we provide here 3 different perspectives on the linear algebra flavors supported by Abinit and their relationships with the other concepts. The build system lets you configure linear algebra from each of these 3 perspectives, mainly by setting its //with_linalg_flavor// option in different ways (explained in detail below). You can choose between:
+  * a vendor-first approach, where the criterion is to use as many components as possible from the same vendor;
+  * an architecture-first approach, where the components are selected depending on which elements of the computer architecture are enabled (e.g. OpenMP, MPI, GPU, ...);
+  * a component-first approach, where you specify explicitly each linear algebra component (e.g. BLAS, LAPACK, ScaLAPACK, ...) you want to use.
 The following table lists the available flavors supported by Abinit and their relationships with the other relevant concepts:
@@ Line 27: / Line 30: @@
 | elpa2014 | [[http://www.mpcdf.mpg.de/|MPCDF]] | [[http://elpa.mpcdf.mpg.de/|ELPA 2014]] | mpi | elpa |
 | elpa2015 | [[http://www.mpcdf.mpg.de/|MPCDF]] | [[http://elpa.mpcdf.mpg.de/|ELPA 2015]] | mpi | elpa |
+| elpa2016 | [[http://www.mpcdf.mpg.de/|MPCDF]] | [[http://elpa.mpcdf.mpg.de/|ELPA 2016]] | mpi | elpa |
+| elpa2017 | [[http://www.mpcdf.mpg.de/|MPCDF]] | [[http://elpa.mpcdf.mpg.de/|ELPA 2017]] | mpi | elpa |
 | essl     | [[http://www.ibm.com/|IBM]] | [[https://www-03.ibm.com/systems/power/software/essl/|ESSL]] | serial, openmp, mpi | blas, lapack, scalapack |
 | magma10  | [[http://www.icl.cs.utk.edu/|UTK]] | [[http://icl.cs.utk.edu/magma/|MAGMA 1.0]] | gpu | magma |
 | magma15  | [[http://www.icl.cs.utk.edu/|UTK]] | [[http://icl.cs.utk.edu/magma/|MAGMA 1.5]] | gpu | magma |
-| mkl      | [[http://developer.intel.com/|Intel]] | [[https://software.intel.com/en-us/intel-mkl|MKL]] | serial, mpi | blas, lapack, scalapack |
+| magma25  | [[http://www.icl.cs.utk.edu/|UTK]] | [[http://icl.cs.utk.edu/magma/|MAGMA 2.5]] | gpu | magma |
+| mkl      | [[http://developer.intel.com/|Intel]] | [[https://software.intel.com/en-us/intel-mkl|MKL]] | serial, openmp, mpi | blas, lapack, scalapack |
 | netlib   | [[http://netlib.org/|Netlib]] | [[http://netlib.org/liblist.html|Netlib Repository]] | serial, mpi | blas, lapack, lapacke, scalapack |
 | openblas | [[https://github.com/xianyi|Zhang Xianyi]] | [[http://www.openblas.net/|OpenBLAS]] | serial | blas |
 | plasma1  | [[http://www.icl.cs.utk.edu/|UTK]] | [[http://icl.cs.utk.edu/plasma/|PLASMA]] | openmp | plasma |
+Please refer to the instructions of each vendor listed above by clicking on the corresponding link to obtain information on how to install their linear algebra implementation.
 ===== Automatic detection =====
@@ Line 54: / Line 62: @@
 You can tune the detection mechanism through the //with_linalg_flavor// option in 3 ways:
-  * by just specifying a plus-separated list of flavors, in which case the build system will figure out how to assign which flavor to which component;
+  * by just specifying a plus-separated list of flavors, in which case the build system will figure out how to assign which flavor to which component and complain about inconsistencies;
   * architecture-wise, by specifying a comma-separated list of //arch:value// statements, where //arch// can be //serial//, //openmp//, //mpi//, or //gpu//, and //value// is a plus-separated list of flavors;
   * component-wise, by specifying a comma-separated list of //component:value// statements, where //component// is one of the keywords from the first column of the above table, and //value// is a corresponding flavor -- only one is permitted in this case.