Sven Verdoolaege [Fri, 20 May 2011 10:02:03 +0000 (12:02 +0200)]
isl_qpolynomial_from_aff: create private copy of divs
Otherwise, reduce_divs could end up modifying the divs of the input isl_aff.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Fri, 20 May 2011 09:50:27 +0000 (11:50 +0200)]
isl_aff_expand_divs: fix typo
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Fri, 20 May 2011 09:23:49 +0000 (11:23 +0200)]
isl_aff_alloc: check that all divs are known
Sven Verdoolaege [Fri, 20 May 2011 09:23:31 +0000 (11:23 +0200)]
add isl_local_space_divs_known
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Fri, 20 May 2011 13:07:16 +0000 (15:07 +0200)]
Merge branch 'maint'
Sven Verdoolaege [Fri, 20 May 2011 09:24:20 +0000 (11:24 +0200)]
isl_qpolynomial_* constructors: check dim before accessing fields
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 18 May 2011 13:03:11 +0000 (15:03 +0200)]
add isl_qpolynomial_from_aff
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 18 May 2011 12:46:16 +0000 (14:46 +0200)]
isl_qpolynomial_involves_dims: fix typo
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 18 May 2011 11:25:04 +0000 (13:25 +0200)]
isl_polynomial.c: fix typo in comment
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 18 May 2011 11:07:52 +0000 (13:07 +0200)]
add isl_aff_get_dim
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Tue, 17 May 2011 13:29:16 +0000 (15:29 +0200)]
isl_basic_map_from_local_space: don't finalize result
This function is only called internally during the construction
of isl_constraints and isl_divs and we don't want to remove redundant
existentially quantified variables in these cases.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 12 May 2011 13:59:09 +0000 (15:59 +0200)]
add isl_set_add_constraint
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 12 May 2011 13:55:15 +0000 (15:55 +0200)]
add isl_equality_from_aff
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 12 May 2011 11:32:38 +0000 (13:32 +0200)]
add isl_set_max
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 11 May 2011 16:26:05 +0000 (18:26 +0200)]
add isl_constraint_get_bound
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 11 May 2011 16:01:59 +0000 (18:01 +0200)]
add isl_{basic_,}map_may_be_set
Sets and maps share the same internal representation, but only
some of those internal representations can be sets.
Add specific functions to test this condition, replacing some
open coded variants.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sat, 16 Apr 2011 14:23:57 +0000 (16:23 +0200)]
add isl_aff
Affine expressions will be used in a subsequent commit to specify
the objective function for an ILP problem, but they should hopefully
be more generally useful.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 12 May 2011 11:03:10 +0000 (13:03 +0200)]
add isl_basic_set_expand_divs
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 12 May 2011 10:24:16 +0000 (12:24 +0200)]
isl_polynomial.c: move merge_divs to isl_local_space.c
In time, we will want to define isl_qpolynomials over isl_local_spaces.
We start by moving some of the handling of divs to isl_local_space.c
as it also useful for local spaces in general.
The merging is adapted to handle possibly unknown divs, which should
never be merged.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sat, 16 Apr 2011 17:47:13 +0000 (19:47 +0200)]
add isl_local_space
Local spaces should hopefully replaced the duplicate functionality
in basic sets and quasi-polynomials.
In a later commit, we will start by using it in affine expressions.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sat, 16 Apr 2011 17:41:16 +0000 (19:41 +0200)]
isl_dim_size: handle isl_dim_all
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 11 May 2011 13:32:34 +0000 (15:32 +0200)]
add isl_constraint_involves_dims
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Fri, 6 May 2011 13:33:38 +0000 (15:33 +0200)]
isl_printer_print_qpolynomial: support proper isl output
That is, print description of the space associated to the isl_qpolynomial
in the isl format.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Fri, 6 May 2011 13:41:31 +0000 (15:41 +0200)]
add isl_qpolynomial_align_params
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Fri, 6 May 2011 13:15:00 +0000 (15:15 +0200)]
add isl_map_align_params
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 11 May 2011 14:34:43 +0000 (16:34 +0200)]
export isl_basic_set_remove_redundancies
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 9 May 2011 14:37:56 +0000 (16:37 +0200)]
export isl_qpolynomial_gist
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sun, 8 May 2011 13:37:43 +0000 (15:37 +0200)]
export isl_set_eliminate
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 16 May 2011 14:07:11 +0000 (16:07 +0200)]
isl/div.h: drop dependence on isl/set.h
Conceptually, a set involves divs and not the other way around.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 16 May 2011 13:55:37 +0000 (15:55 +0200)]
hide isl_div internals
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 16 May 2011 13:58:09 +0000 (15:58 +0200)]
add isl_div_get_ctx
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 12 May 2011 10:52:57 +0000 (12:52 +0200)]
add isl_mat_get_ctx
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 16 May 2011 10:26:50 +0000 (12:26 +0200)]
doc: drop documentation for isl-polylib
The documentation for isl-polylib doesn't belong here and
we don't want to encourage the use of isl-polylib.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sun, 15 May 2011 18:22:43 +0000 (20:22 +0200)]
add gmp flags to pkg-config file
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sun, 15 May 2011 18:12:35 +0000 (20:12 +0200)]
explicitly link in gmp in applications that reference gmp symbols
They do this through the use of isl_int_* macros.
Not explicitly linking to gmp creates problems with some combinations
of libtool and ld, notably those that come with recent debians.
Reported-by: Albert Cohen <Albert.Cohen@inria.fr>
Tested-by: Albert Cohen <Albert.Cohen@inria.fr>
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sun, 15 May 2011 16:55:43 +0000 (18:55 +0200)]
configure.ac: don't depend on automake 1.11
AM_SILENT_RULES is not available in earlier versions of automake,
so test for its presence before calling it.
Reported-by: Louis-Noel Pouchet <pouchet@cse.ohio-state.edu>
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 9 May 2011 13:01:08 +0000 (15:01 +0200)]
isl_pw_qpolynomial_involves_dims: also check whether domains involve dims
Before, isl_pw_qpolynomial_involves_dims would only check whether the
polynomials associated to the different cells involved any of the indicated
dimensions, but didn't check if the decriptions of the cells involved
any of them.
If isl_pw_qpolynomial_involves_dims is used to check whether
isl_pw_qpolynomial_drop_dims can safely be applied, then clearly we need
to test the cell constraints too.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 4 May 2011 13:50:47 +0000 (15:50 +0200)]
add isl_union_map_plain_is_injective
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 4 May 2011 16:39:05 +0000 (18:39 +0200)]
isl_map_intersect_range: add special case for intersection with universe
If the user wants to intersect the range with the appropriate universe
set, we don't actually need to change anything.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 4 May 2011 13:46:49 +0000 (15:46 +0200)]
add isl_vec_sort
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 4 May 2011 10:59:23 +0000 (12:59 +0200)]
add isl_union_map_universe
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 4 May 2011 09:27:31 +0000 (11:27 +0200)]
rename *_fast_* functions to *_plain_*
Although the *_fast_* functions are certainly meant to be faster than
their unqualified alternatives, they are faster only because they are
not complete. The "plain" qualification is hopefully better at conveying
the idea that these functions only consider the obvious cases.
We keep some of the *_fast_* names for backward compatibility.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 5 May 2011 10:33:07 +0000 (12:33 +0200)]
Merge branch 'maint'
Sven Verdoolaege [Wed, 4 May 2011 13:03:27 +0000 (15:03 +0200)]
isl_map_range: correctly handle maps with 0D domains
Internally, sets are represented as maps with unnamed, unstructured
0D domains. isl_map_range takes advantage of this fact, but only
checked for 0D domains. This was correct when isl_map_range was added,
but it should have been updated when named and structured spaces were
introduced.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sat, 30 Apr 2011 14:37:10 +0000 (16:37 +0200)]
isl_list.h: allow inclusion from C++
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 27 Apr 2011 20:23:13 +0000 (22:23 +0200)]
isl_tab_pip: don't simplify domain in isl_for_add
As the comment above the function explains, the domain should not
be simplified. Unfortunately, an isl_basic_set_simplify snuck in
in 8b88ebf (isl_tab_pip.c: remove some code duplication between
isl_map_add and isl_for_add, Fri Oct 16 14:10:59 2009 +0200).
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Fri, 29 Apr 2011 09:52:39 +0000 (11:52 +0200)]
add isl_printer_print_constraint
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Fri, 29 Apr 2011 09:50:34 +0000 (11:50 +0200)]
add isl_constraint_get_ctx
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 27 Apr 2011 10:20:30 +0000 (12:20 +0200)]
add isl_mat_dump
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 27 Apr 2011 10:19:30 +0000 (12:19 +0200)]
rename isl_mat_dump to isl_mat_print_internal
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 25 Apr 2011 12:21:05 +0000 (14:21 +0200)]
add isl_constraint_get_dim
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 25 Apr 2011 15:15:48 +0000 (17:15 +0200)]
Merge branch 'maint'
Sven Verdoolaege [Mon, 25 Apr 2011 08:29:59 +0000 (10:29 +0200)]
isl_basic_map_affine_hull: finalize result
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Tobias Grosser [Mon, 25 Apr 2011 06:48:20 +0000 (08:48 +0200)]
Remove recently added unused variables
These variables produce warnings that may hide important warnings. Hence, remove
them.
Signed-off-by: Tobias Grosser <tobias@grosser.es>
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Tobias Grosser [Mon, 25 Apr 2011 06:48:19 +0000 (08:48 +0200)]
Add extern C declaration to schedule.h
All headers except the recently added schedule.h header include an extern C
declaration. Add the same declaration to schedule.h to allow flawless inclusion
in C++ programs.
Signed-off-by: Tobias Grosser <tobias@grosser.es>
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Fri, 22 Apr 2011 16:12:17 +0000 (18:12 +0200)]
dataflow analysis: don't depend on word size during sorting of sources
The comparison routine would compute the difference between two
unsigned numbers, which would result in different signs depending
on the machine word size.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sat, 19 Mar 2011 13:47:47 +0000 (14:47 +0100)]
add isl_union_set_compute_schedule
This is still a very preliminary implementation.
The algorithm used is similar to that of Pluto, except that
it may compute parametric schedules and schedules with negative
coefficients. The current implementation probably tries to fuse too much.
This should be made tunable at some point.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Tue, 19 Apr 2011 13:45:05 +0000 (15:45 +0200)]
add isl_tab_basic_set_non_neg_lexmin
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Tue, 19 Apr 2011 12:51:43 +0000 (14:51 +0200)]
add isl_tab_basic_set_non_trivial_lexmin
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Tue, 19 Apr 2011 14:47:19 +0000 (16:47 +0200)]
isl_tab_pip.c: add_lexmin_eq: return int instead of isl_tab *
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Tue, 19 Apr 2011 14:41:58 +0000 (16:41 +0200)]
isl_tab_pip.c: restore_lexmin: return int instead of isl_tab *
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Fri, 15 Apr 2011 13:25:35 +0000 (15:25 +0200)]
isl_tab_pip.c: add some debugging code
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Fri, 15 Apr 2011 13:25:47 +0000 (15:25 +0200)]
isl_tab_pip.c: fix typo in comment
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 13 Apr 2011 14:21:00 +0000 (16:21 +0200)]
add isl_quicksort (copied from glibc)
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 13 Apr 2011 08:37:46 +0000 (10:37 +0200)]
isl_printer_print_basic_set: also print invalid basic sets
In particular, zero equality rows should not appear in a finalized
basic set, but they may appear during the construction of a basic set.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Tue, 12 Apr 2011 11:01:21 +0000 (13:01 +0200)]
add isl_union_map_is_injective
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Tue, 12 Apr 2011 10:05:46 +0000 (12:05 +0200)]
add isl_union_map_is_bijective
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Tue, 12 Apr 2011 09:57:58 +0000 (11:57 +0200)]
add isl_union_map_is_single_valued
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 11 Apr 2011 11:14:13 +0000 (13:14 +0200)]
add isl_mat_add_zero_rows
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sun, 10 Apr 2011 08:57:11 +0000 (10:57 +0200)]
isl_blk: don't reuse overly large blocks for small allocations
We usually allocate a large number of small objects, but occasionally
we may allocate one or more big objects. The original caching code
could reuse a cached huge object for a small allocation.
This is fine if the object would grow later, but if it remains
small, then we may end up consuming and wasting a lot of memory,
especially if these objects are long-lived.
Now we only a memory block if it is at most twice as big as the desired
size. This may result in some large objects sticking around in the cache,
so we evict objects from the cache after a while.
Finally, we don't reuse any cache elements for initially zero sized
allocations, but instead check the cache when the objects first growns
to a non-zero size.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sat, 9 Apr 2011 18:18:18 +0000 (20:18 +0200)]
add isl_hmap_map_basic_set
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Fri, 8 Apr 2011 07:47:04 +0000 (09:47 +0200)]
add isl_union_map_copy_map
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 7 Apr 2011 12:51:41 +0000 (14:51 +0200)]
add isl_union_map_contains
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 6 Apr 2011 13:19:10 +0000 (15:19 +0200)]
add isl_basic_set_transform_dims
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 6 Apr 2011 13:00:26 +0000 (15:00 +0200)]
add isl_basic_set_offset
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 6 Apr 2011 12:09:00 +0000 (14:09 +0200)]
change isl_mat_sub_alloc prototype
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Tue, 19 Apr 2011 13:18:10 +0000 (15:18 +0200)]
make isl_mat_sub_* functions private
They were never meant to be made public.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sat, 26 Mar 2011 14:58:08 +0000 (15:58 +0100)]
add isl_vec_get_ctx
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sat, 26 Mar 2011 14:18:17 +0000 (15:18 +0100)]
add isl_mat_initial_non_zero_cols
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sat, 26 Mar 2011 09:48:01 +0000 (10:48 +0100)]
add isl_mat_set_element_si
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 21 Mar 2011 12:01:18 +0000 (13:01 +0100)]
add isl_basic_set_add_constraints_dim_map
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 21 Mar 2011 11:59:18 +0000 (12:59 +0100)]
add isl_dim_map_range
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 21 Mar 2011 11:31:49 +0000 (12:31 +0100)]
isl_dim_map: allow signed mapping
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 21 Mar 2011 11:23:51 +0000 (12:23 +0100)]
isl_dim_map: extract from isl_map.c
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sun, 20 Mar 2011 15:46:51 +0000 (16:46 +0100)]
change prototype of isl_tab_dump
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sun, 20 Mar 2011 13:51:30 +0000 (14:51 +0100)]
add isl_map_remove_divs
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Tue, 19 Apr 2011 15:56:10 +0000 (17:56 +0200)]
doc: fix typo
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 23 Dec 2010 22:17:41 +0000 (23:17 +0100)]
doc: add a note on the accuracy of our approximation
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 22 Dec 2010 13:41:06 +0000 (14:41 +0100)]
add isl_union_set_coefficients and isl_union_set_solutions
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 22 Dec 2010 13:50:39 +0000 (14:50 +0100)]
add isl_union_set_lift
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Tue, 19 Apr 2011 12:58:39 +0000 (14:58 +0200)]
Merge branch 'maint'
Sven Verdoolaege [Sat, 26 Mar 2011 17:00:26 +0000 (18:00 +0100)]
isl_tab: alway free saved basis undo record
Before, it would only get freed if it was actually used.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Tue, 19 Apr 2011 12:32:21 +0000 (14:32 +0200)]
make sure destination for GDB module exists before installing
Usually, this directory already exists because the library has
been copied there, but during a parallel make install this may
not have happened yet.
Reported-by: dirtyepic <dirtyepic.sk@gmail.com>
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Fri, 15 Apr 2011 15:09:24 +0000 (17:09 +0200)]
isl_tab_pip.c: add_lexmin_eq: remove misguided optimizations
After adding a pair of inequalities corresponding to the equality,
the code tries to reduce the dimensionality of the problem by
removing a column. If one of the added constraints happens to have
been pivoted into a column, then it can easily be removed.
Otherwise, the code tries to pivot the row into a column and
then remove the column.
This pivot may disrupt the lexico-positiveness of the columns however.
At the first least, the code should check that the pivot entry
is negative, but it should also avoid pivoting out a column corresponding
to a parameter.
Since this optimization shouldn't have that much of an effect anyway,
we choose here to just remove it for now.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sun, 20 Mar 2011 15:40:41 +0000 (16:40 +0100)]
isl_basic_map_lexmin: turn off equality detection on non-parametric problems
Detecting equalities can be quite expensive and has mostly been shown
effective on parametric problems. For non-parametric problems, we
are effectively computing several solutions as a preprocessing step
to computing a single solution, albeit a special one.
Until we obtain more evidence that this is useful,
it seems prudent to turn off this equality detection on non-parametric
problems.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sat, 19 Mar 2011 16:25:29 +0000 (17:25 +0100)]
use silent make rules
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 11 Apr 2011 18:56:48 +0000 (20:56 +0200)]
isl_stream_read_map: fix handling of divs in tuples
Supoprt for them was broken in f1ea969 (isl_stream_read_map: treat divs
as output variables while adding constraints, Thu Mar 17 16:29:06 2011 +0100).
When treating divs as output variables, we need to make sure that
each of them is backed by a variable in the list of active variables.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sat, 26 Mar 2011 14:05:41 +0000 (15:05 +0100)]
isl_mat_left_hermite: plug memory leak on error path
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 7 Apr 2011 20:29:43 +0000 (22:29 +0200)]
plug memory leak in isl_mat_product
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Fri, 18 Mar 2011 15:02:33 +0000 (16:02 +0100)]
isl 0.06
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
projects / platform / upstream / isl.git / log