Differences between revisions 1 and 4 (spanning 3 versions)
Revision 1 as of 2007-11-12 15:01:25
Size: 1691
Comment: initial version
Revision 4 as of 2007-11-12 22:18:35
Size: 2243
Comment: mention netlib.org only BLAS & Lapack bug
Deletions are marked like this. Additions are marked like this.
Line 18: Line 18:
 * fix corner case when only netlib.org BLAS and Lapack are installed in default patch
Line 30: Line 31:
 * add and review Paul's fixes for a) string copy in case of exception b) missing inlines
Line 35: Line 37:
 * merge yet to be determined fixes/improvements by malb from Sage's givaro.spkg upstream
 * merge fixes/improvements by malb from Sage's givaro.spkg:
{{{
-- 2007-02-03 Martin Albrecht <malb@...>
  * new upstream release: 3.2.6
  * Changes to upstream (everything else below is irrelevant):
   ./src/library/poly1/givpoly1factor.h (2006-10-21 fix)
   ./src/kernel/zpz/givgfq.inl (2006-10-21 fix)
   ./src/kernel/zpz/givgfq.h (2006-10-21 fix)
   ./aclocal.m4 64-bit (2006-10-29 fix)
   ./src/library/poly1/givpoly1padic.h (2006-11-09 fix)
}}}

Working Group Members

Michael Abshoff, Clement Pernet

Goals

LinBox

At the Sage Days 6 coding spring:

  • fix bug fallout from the 1.1.4 release (mostly done)
  • fix Blas:MatrixHom memleak

  • determine which workarounds can be discarded from Sage's LinBox wrapper library

  • Sage class that wraps LinBox matrices directly - this should simplify automated testing of LinBox via Sage

  • wrap solve and other interesting BlackBoxes

  • sort out gmpxx.h issues, i.e. NTL version detect is broken in LinBox 1.1.4, but not in Sage

  • fix corner case when only netlib.org BLAS and Lapack are installed in default patch
  • fix test-zero-one segfault on Solaris
  • merge LinBox wrapper library into driver's directory in Linbox, making the build simpler

Short term, i.e. plans for a potential 1.1.5 LinBox release in the next 3-4 weeks:

  • support threaded GoToBLAS
  • support (threaded) SunPerfLib on Solaris

  • print configuration also to stdout so that it is logged in the Sage build log
  • investigate slow iterator test on OSX (10 or more times slow than under Linux on the same hardware)
  • proper lround fix for Solaris via ACTEST
  • set up build bots for Cygwin and Solaris, wider range of gcc releases (mabshoff)
  • add STLPort debug build target to general LinBox build

  • add and review Paul's fixes for a) string copy in case of exception b) missing inlines

Givaro

  • merge OSX 10.5 build fix by rpw
  • improved gcc 4.3 buidl fix: instead of string.h include cstring
  • merge fixes/improvements by malb from Sage's givaro.spkg:

-- 2007-02-03 Martin Albrecht <malb@...>
  * new upstream release: 3.2.6
  * Changes to upstream (everything else below is irrelevant): 
   ./src/library/poly1/givpoly1factor.h (2006-10-21 fix)
   ./src/kernel/zpz/givgfq.inl (2006-10-21 fix)
   ./src/kernel/zpz/givgfq.h  (2006-10-21 fix)
   ./aclocal.m4 64-bit (2006-10-29 fix)
   ./src/library/poly1/givpoly1padic.h (2006-11-09 fix)

ATLAS

  • build ATLAS with netlib's Lapack per default
  • use sage_fortran as fortran compiler
  • update to the 3.8.0 release, apply install-dynamic-libs patch
  • build multi-threaded version of ATLAS on demand, too

days6/sprint/packages (last edited 2008-11-14 13:42:00 by anonymous)