Differences

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

Link to this comparison view

doku:blas [2012/09/04 13:10] – external edit 127.0.0.1doku:blas [2014/07/25 10:04] (current) – sh
Line 2: Line 2:
  
 There are several implementations of the Basic Linear Algebra Subprograms (BLAS) libraries available. There are several implementations of the Basic Linear Algebra Subprograms (BLAS) libraries available.
-These provide highly optimized routines for matrix and vector operations and are a key to high performance applications.+These provide highly optimized routines for matrix and vector operations and are key to high performance applications.
  
 We recommend to use one of the fastest available libraries: We recommend to use one of the fastest available libraries:
-  * GOTO BLAS by Kazushige Goto (http://www.tacc.utexas.edu/tacc-projects/gotoblas2) available in the servers and compute nodes by <code>-L/opt/goto/ifort -lgoto2_barcelonap-r1.13</code> +<del>  * GOTO BLAS by Kazushige Goto (http://www.tacc.utexas.edu/tacc-projects/gotoblas2) available in the servers and compute nodes by <code>-L/opt/goto/ifort -lgoto2_barcelonap-r1.13</code> 
-    You may want to set ''THREADS=1''.+    You may want to set ''THREADS=1''.</del>
   * Intel Math Kernel Library (MKL): to use e.g. with <code>-L/opt/intel/composerxe/mkl/lib/intel64/ -lmkl_intel_lp64 -lmkl_sequential -lmkl_core</code> or -- with the Intel compiler suite -- by simply using <code>-mkl</code>   * Intel Math Kernel Library (MKL): to use e.g. with <code>-L/opt/intel/composerxe/mkl/lib/intel64/ -lmkl_intel_lp64 -lmkl_sequential -lmkl_core</code> or -- with the Intel compiler suite -- by simply using <code>-mkl</code>
   * AMD Core Math Library (ACML): to use e.g. with <code>-L/opt/acml5.1.0/ifort64/lib/ -lacml</code>   * AMD Core Math Library (ACML): to use e.g. with <code>-L/opt/acml5.1.0/ifort64/lib/ -lacml</code>
  
-The GOTO and MKL libraries exist in single and multi threaded versions. +MKL libraries exist in single and multi threaded versions.
-  * To use GOTO // multi threaded version// use ''-L/opt/goto/ifort -lgoto2_barcelonap-r1.13 -lpthread''+
   * To use the MKL //multi threaded version// use ''-lmkl_threads'' instead of ''-lmkl_sequential''   * To use the MKL //multi threaded version// use ''-lmkl_threads'' instead of ''-lmkl_sequential''
 +<del>  * To use GOTO // multi threaded version// use ''-L/opt/goto/ifort -lgoto2_barcelonap-r1.13 -lpthread''</del>
  
 The reference BLAS is installed on some nodes (''-lblas'') but significantly slower. We recommend not to use it, even if available. The reference BLAS is installed on some nodes (''-lblas'') but significantly slower. We recommend not to use it, even if available.
  
-If shared libraries are used, setting the variable ''LD_LIBRARY_PATH'' is required, e.g. <code>export LD_LIBRARY_PATH=/opt/goto/ifort:$LD_LIBRARY_PATH</code>+If shared libraries are used, resetting the variable ''LD_LIBRARY_PATH'' is required, e.g. <code>export LD_LIBRARY_PATH=/opt/somewhere/somelib/:$LD_LIBRARY_PATH</code>
  • doku/blas.1346764242.txt.gz
  • Last modified: 2014/07/25 10:04
  • (external edit)