Block-triangular LU-factorization was implemented to be used
on computing an initial factorization of the basis matrix.
A new version of the Schur-complement-based factorization
module was included in the package. Now it can be used along
with plain as well as with block-triangular LU-factorization.
Currently the following flags can be used to specify the type
of the basis matrix factorization (glp_bfcp.type):
GLP_BF_LUF + GLP_BF_FT LUF, Forrest-Tomlin update (default)
GLP_BF_LUF + GLP_BF_BG LUF, Schur complement, Bartels-Golub
update
GLP_BF_LUF + GLP_BF_GR LUF, Schur complement, Givens rotation
update
GLP_BF_BTF + GLP_BF_BG BTF, Schur complement, Bartels-Golub
update
GLP_BF_BTF + GLP_BF_GR BTF, Schur complement, Givens rotation
update
In case of GLP_BF_FT the update is applied to matrix U, while
in cases of GLP_BF_BG and GLP_BF_GR the update is applied to
the Schur complement.
Corresponding new options --luf and --btf were added to glpsol.
For more details please see a new edition of the GLPK reference
manual included in the distribution.
A minor bug (in reporting the mip solution status) was fixed.
A call to "iodbc-config --cflags" was added in configure.ac
to correctly detect iodbc flags.
This changes the buildlink3.mk files to use an include guard for the
recursive include. The use of BUILDLINK_DEPTH, BUILDLINK_DEPENDS,
BUILDLINK_PACKAGES and BUILDLINK_ORDER is handled by a single new
variable BUILDLINK_TREE. Each buildlink3.mk file adds a pair of
enter/exit marker, which can be used to reconstruct the tree and
to determine first level includes. Avoiding := for large variables
(BUILDLINK_ORDER) speeds up parse time as += has linear complexity.
The include guard reduces system time by avoiding reading files over and
over again. For complex packages this reduces both %user and %sys time to
half of the former time.
A tentative implementation of the "exact" simplex method based
on bignum (rational) arithmetic was included in the package.
On API level this new feature is available through the routine
lpx_exact, which is similar to the routine lpx_simplex.
In the solver glpsol this feature is available through two new
command-line options: --exact and --xcheck. If the '--exact'
option is specified, glpsol solves LP instance using the exact
simplex method; in case of MIP it is used to obtain optimal
solution of LP relaxation. If the --xcheck option is specified,
LP instance (or LP relaxation) is solved using the standard
(floating-point) simplex method, however, then glpsol calls the
exact simplex routine to make sure that the final LP basis is
exactly optimal, and if it is not, to perform some additional
simplex iterations in exact arithmetic.
Changes 4.12:
A tentative implementation of some simplex method routines
based on exact (bignum) arithmetic was included in the package.
Currently these routines provide computing LU-factorization of
the basis matrix and computing components of basic solution.
These routines were used to implement a routine, which checks
primal and dual feasibility of basic solution exactly, i.e. in
rational numbers, without round-off errors. In glpsol this
feature is available through the command-line option --xcheck.
GLPK has its own low-level routines implementing operations on
integer and rational numbers that makes it independent on other
software packages. However, to attain a much better performance
it is highly recommended to install (before configuring GLPK)
the GNU Multiple Precision Arithmetic Library (GMP). Using GMP
makes computations 100-200 times faster.
and add a new helper target and script, "show-buildlink3", that outputs
a listing of the buildlink3.mk files included as well as the depth at
which they are included.
For example, "make show-buildlink3" in fonts/Xft2 displays:
zlib
fontconfig
iconv
zlib
freetype2
expat
freetype2
Xrender
renderproto
Cutting planes of two new classes were implemented: mixed cover
cuts and clique cuts. On API level this feature can be enabled
by setting control parameter LPX_K_USECUTS passed to the routine
lpx_intopt. In glpsol this feature is available through the
command-line options --cover and --clique. For more details see
the reference manual.
Now the routines lpx_read_mps and lpx_read_freemps support LI
bound type. It is similar to LO, however, indicates the column
as of integer kind.
RECOMMENDED is removed. It becomes ABI_DEPENDS.
BUILDLINK_RECOMMENDED.foo becomes BUILDLINK_ABI_DEPENDS.foo.
BUILDLINK_DEPENDS.foo becomes BUILDLINK_API_DEPENDS.foo.
BUILDLINK_DEPENDS does not change.
IGNORE_RECOMMENDED (which defaulted to "no") becomes USE_ABI_DEPENDS
which defaults to "yes".
Added to obsolete.mk checking for IGNORE_RECOMMENDED.
I did not manually go through and fix any aesthetic tab/spacing issues.
I have tested the above patch on DragonFly building and packaging
subversion and pkglint and their many dependencies.
I have also tested USE_ABI_DEPENDS=no on my NetBSD workstation (where I
have used IGNORE_RECOMMENDED for a long time). I have been an active user
of IGNORE_RECOMMENDED since it was available.
As suggested, I removed the documentation sentences suggesting bumping for
"security" issues.
As discussed on tech-pkg.
I will commit to revbump, pkglint, pkg_install, createbuildlink separately.
Note that if you use wip, it will fail! I will commit to pkgsrc-wip
later (within day).
