From Fedora Project Wiki
(→‎Current status: Retarget to F33.)
(→‎Dependencies: update packages status)
 
Line 67: Line 67:
  
 
== Dependencies ==
 
== Dependencies ==
atlas (35):
+
atlas (25/35):
* ceres-solver
+
* ceres-solver '''DONE'''
 
* CheMPS2 '''DONE'''
 
* CheMPS2 '''DONE'''
 
* cocoalib '''DONE'''
 
* cocoalib '''DONE'''
 
* cxsc '''DONE'''
 
* cxsc '''DONE'''
* DSDP
+
* DSDP '''DONE'''
 
* ergo ''test failures?''
 
* ergo ''test failures?''
 
* freefem++
 
* freefem++
Line 79: Line 79:
 
* iml '''DONE'''
 
* iml '''DONE'''
 
* jblas
 
* jblas
* libcint
+
* libcint '''DONE'''
 
* libghemical
 
* libghemical
 
* linbox '''DONE'''
 
* linbox '''DONE'''
Line 93: Line 93:
 
* python-cvxopt '''DONE'''
 
* python-cvxopt '''DONE'''
 
* python-nipy
 
* python-nipy
* python-scikit-learn
+
* python-scikit-learn '''''N/A'''''
 
* qcint
 
* qcint
 
* qm-vamp-plugins
 
* qm-vamp-plugins
 
* sagemath '''DONE'''
 
* sagemath '''DONE'''
* scamp
+
* scamp '''DONE'''
* sextractor
+
* sextractor '''DONE'''
 
* suitesparse '''DONE'''
 
* suitesparse '''DONE'''
 
* SuperLU '''DONE'''
 
* SuperLU '''DONE'''
 
* SuperLU43 '''''retired'''''
 
* SuperLU43 '''''retired'''''
* wannier90
+
* wannier90 '''''orphaned'''''
  
blas/lapack(1):
+
blas/lapack(1/1):
 
* shogun '''''retired'''''
 
* shogun '''''retired'''''
  

Latest revision as of 21:22, 5 April 2020

OpenBLAS as default BLAS implementation

Summary

Use OpenBLAS as the default BLAS library implementation.

Owner

Current status

Detailed Description

Most application use ATLAS because it's faster than reference BLAS. However, it doesn't support runtime CPU detection, so it must be built for a particular CPU. OpenBLAS is a competing BLAS implementation based on GotoBLAS2 that is and supports runtime CPU detection and all current Fedora primary arches.

Benefit to Fedora

Using a single default BLAS implementation will avoid bugs stemming from having two different BLAS libraries loaded at runtime that causes computation errors.

OpenBLAS detects CPU capabilities at runtime and selects the fastest computation kernel for the running hardware.

A lot of benchmarks around the net suggest openblas is faster than ATLAS in most workloads: https://github.com/tmolteno/necpp/issues/18 http://blog.nguyenvq.com/blog/2014/11/10/optimized-r-and-python-standard-blas-vs-atlas-vs-openblas-vs-mkl/ https://freddie.witherden.org/pages/blas-gemm-bench/ http://nghiaho.com/?p=1726

Scope

  • Proposal owners:

Modify the affected packages to BuildRequire: openblas-devel instead of atlas-devel or blas-devel and rebuild. Fix any failures.

  • Other developers:

N/A

  • Release engineering:

No help should be necessary. No mass rebuild necessary.

  • Policies and guidelines: N/A
  • Trademark approval: N/A

Upgrade/compatibility impact

How To Test

User Experience

Dependencies

atlas (25/35):

  • ceres-solver DONE
  • CheMPS2 DONE
  • cocoalib DONE
  • cxsc DONE
  • DSDP DONE
  • ergo test failures?
  • freefem++
  • ga DONE
  • gromacs DONE
  • iml DONE
  • jblas
  • libcint DONE
  • libghemical
  • linbox DONE
  • Macaulay2 DONE
  • mpqc
  • ncl DONE
  • numpy DONE
  • ocaml-gsl DONE
  • octave DONE
  • opengrm-ngram DONE
  • psfex
  • psi4 DONE
  • python-cvxopt DONE
  • python-nipy
  • python-scikit-learn N/A
  • qcint
  • qm-vamp-plugins
  • sagemath DONE
  • scamp DONE
  • sextractor DONE
  • suitesparse DONE
  • SuperLU DONE
  • SuperLU43 retired
  • wannier90 orphaned

blas/lapack(1/1):

  • shogun retired

Contingency Plan

  • Contingency mechanism: (What to do? Who will do it?) If switching to OpenBLAS breaks a particular package, changes to the package will be reverted and package rebuilt.
  • Contingency deadline: beta freeze
  • Blocks release? No
  • Blocks product? No

Documentation

Release Notes