Sven Verdoolaege [Sat, 4 Dec 2010 13:47:36 +0000 (14:47 +0100)]
doc: emphasize the need for isl_int_get_gmp and isl_int_set_gmp
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 2 Dec 2010 14:38:01 +0000 (15:38 +0100)]
isl_qpolynomial_div: further normalize divs by reducing coefficients
This reduces the number of distinct divs that can appear inside
a quasi-polynomial description.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 2 Dec 2010 15:41:49 +0000 (16:41 +0100)]
isl_stream_read_pw_qpolynomial: accept powers on compound expressions
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 2 Dec 2010 15:40:13 +0000 (16:40 +0100)]
add isl_qpolynomial_pow
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 2 Dec 2010 16:56:09 +0000 (17:56 +0100)]
isl_qpolynomial: properly merge identical nested divs
When sort_divs encounters two identical that appear nested
inside some other div, it would simply drop one of the identical
divs, instead of summing the coefficients of both divs.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 2 Dec 2010 16:46:22 +0000 (17:46 +0100)]
add isl_mat_col_add
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Tue, 7 Dec 2010 09:34:58 +0000 (10:34 +0100)]
avoid using isl_basic_set and isl_basic_map internals in piplib dependent code
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 1 Dec 2010 15:12:19 +0000 (16:12 +0100)]
hide isl_map internals
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Tue, 30 Nov 2010 15:32:04 +0000 (16:32 +0100)]
hide isl_set internals
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Tue, 30 Nov 2010 14:30:54 +0000 (15:30 +0100)]
hide isl_basic_map internals
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Tue, 30 Nov 2010 13:53:13 +0000 (14:53 +0100)]
hide isl_basic_set internals
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sat, 27 Nov 2010 15:13:49 +0000 (16:13 +0100)]
add isl_basic_set_{in,}equalities_matrix
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sun, 21 Nov 2010 09:09:05 +0000 (10:09 +0100)]
isl_arg_parse: support footer
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 15 Nov 2010 22:28:07 +0000 (23:28 +0100)]
isl_arg_parse: avoid duplicate newline in help output
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 15 Nov 2010 22:25:03 +0000 (23:25 +0100)]
isl_arg_parse: support grouping of options within the same structure
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Fri, 12 Nov 2010 16:09:53 +0000 (17:09 +0100)]
isl_arg_parse: support int options
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sun, 21 Nov 2010 11:00:40 +0000 (12:00 +0100)]
isl_arg_parse: support ISL_ARG_HIDDEN flag
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sat, 13 Nov 2010 14:51:23 +0000 (15:51 +0100)]
isl_arg_parse: support ISL_ARG_BOOL_ARG flag
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sat, 13 Nov 2010 14:24:31 +0000 (15:24 +0100)]
isl_arg_parse: support ISL_ARG_SINGLE_DASH flag
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Fri, 12 Nov 2010 15:55:41 +0000 (16:55 +0100)]
isl_arg_parse: support aliases
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 15 Nov 2010 22:07:24 +0000 (23:07 +0100)]
isl_arg_parse: always print help option
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 15 Nov 2010 22:06:53 +0000 (23:06 +0100)]
isl_arg_parse: support phantom boolean options
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 22 Nov 2010 11:43:08 +0000 (12:43 +0100)]
isl_arg_parse: free string arguments before overwriting them
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sun, 21 Nov 2010 08:42:13 +0000 (09:42 +0100)]
isl_arg_parse: allow options after arguments
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 15 Nov 2010 22:10:43 +0000 (23:10 +0100)]
isl_arg_parse: only print --version option once
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Fri, 26 Nov 2010 14:22:43 +0000 (15:22 +0100)]
isl_arg_parse: print string option default
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 24 Nov 2010 19:36:30 +0000 (20:36 +0100)]
isl_arg_parse: delimit string option argument by angle brackets
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sat, 13 Nov 2010 14:20:26 +0000 (15:20 +0100)]
isl_arg_parse: allow options without long name
Sven Verdoolaege [Wed, 24 Nov 2010 11:01:51 +0000 (12:01 +0100)]
isl_vec_read_polylib: use isl_stream
The original implementation imposed an arbitrary limit on the line lengths,
without check.
Reported-by: "sinzerely.yorz" <slashdevslashzerr0@gmail.com>
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 21 Jan 2010 10:31:18 +0000 (11:31 +0100)]
Rename headers from isl_header.h to isl/header.h
We need to turn on the nostdinc option of automake as otherwise
it would include -I$(top_builddir)/include/isl in DEFAULT_INCLUDES
because of
AC_CONFIG_HEADERS(include/isl/config.h)
in configure.ac.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 18 Nov 2010 14:26:50 +0000 (15:26 +0100)]
isl_tab_basic_map_partial_lexopt: fix memory leak on error path
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 17 Nov 2010 15:28:35 +0000 (16:28 +0100)]
add isl_union_set_is_equal
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 17 Nov 2010 15:24:05 +0000 (16:24 +0100)]
add isl_union_map_read_from_str
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 17 Nov 2010 15:03:15 +0000 (16:03 +0100)]
isl_tab_basic_map_partial_lexopt: add all equalities before pivoting
During the construction of the initial tableau in tab_for_lexmin,
restore_lexmin would get called after the addition of each equality.
However, restore_lexmin performs some pivoting operations that could
invalidate the assumptions of last_var_col_or_int_par_col during
the addition of the next equality.
We therefore postpone the call to restore_lexmin until after
all equalities have been added.
Problem reported by Tobias Grosser <grosser@fim.uni-passau.de>
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 17 Nov 2010 15:00:19 +0000 (16:00 +0100)]
isl_pip: avoid empty domain during verification
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 17 Nov 2010 14:59:27 +0000 (15:59 +0100)]
add isl_basic_set_box_from_points
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 15 Nov 2010 10:39:19 +0000 (11:39 +0100)]
isl_basic_map_eliminate_vars: don't detect divs while removing duplicate constraints
We were careful to remove the definitions of divs that depend on the
eliminated variables at the start of isl_basic_map_eliminate_vars,
but remove_duplicate_constraints could rediscover some of them and
add them back in.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sun, 14 Nov 2010 13:18:55 +0000 (14:18 +0100)]
isl_dim.c: set_name: only print single error message on incorrect input
The global_pos function already prints an error message,
so there is no point in printing another error message inside
set_name when global_pos returns an error condition.
Signed-off-by: Sven Verdoolaege <sven@nestor.cs.kuleuven.be>
Sven Verdoolaege [Sun, 14 Nov 2010 13:14:55 +0000 (14:14 +0100)]
isl_pw_qpolynomial_fold_bound: fix handling or zero input with wrapped domain
That is, make sure the result has the right domain even if the input is zero.
Signed-off-by: Sven Verdoolaege <sven@nestor.cs.kuleuven.be>
Sven Verdoolaege [Sun, 14 Nov 2010 13:11:42 +0000 (14:11 +0100)]
isl_printer_print_pw_qpolynomial{,_fold}: always print special domains
If the domain of a piecewise quasipolynomial (fold) is the zero-dimensional
space, then we don't print the "[] -> " to avoid clutter.
However, if the domain has a name or if it has internal structure
then it should be printed explicitly as otherwise no distinction can
be made with the unnamed unstructured zero-dimensional space.
Signed-off-by: Sven Verdoolaege <sven@nestor.cs.kuleuven.be>
Sven Verdoolaege [Fri, 12 Nov 2010 13:25:00 +0000 (14:25 +0100)]
isl_qpolynomial_bound_on_domain_bernstein: optionally triangulate domains
Preliminary experiments show that triangulation may slow down
the computation but helps tremendously in keeping memory use down.
It is therefore enabled by default.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Fri, 12 Nov 2010 13:51:00 +0000 (14:51 +0100)]
isl_options_parse: print default on boolean options
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Fri, 12 Nov 2010 13:42:46 +0000 (14:42 +0100)]
isl_options_parse: accept prefix before or after "no" on boolean options
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 8 Nov 2010 17:47:17 +0000 (18:47 +0100)]
add isl_cell_foreach_simplex
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 8 Nov 2010 16:17:38 +0000 (17:17 +0100)]
isl_basic_set_compute_vertices: store input basic set in result
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 8 Nov 2010 15:34:07 +0000 (16:34 +0100)]
isl_vertices_foreach_cell: store individual vertex ids inside cell
Currently, each cell corresponds to a (possibly reduced) chamber
in the chamber decomposition. If we want to allow other kinds
of cells, e.g., triangulations of chambers, then we need to
store the vertex ids inside the individual cells.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 11 Nov 2010 12:05:06 +0000 (13:05 +0100)]
add isl_union_pw_qpolynomial_to_polynomial
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Fri, 12 Nov 2010 10:01:09 +0000 (11:01 +0100)]
doc: update isl_pw_qpolynomial_gist description
In particular, update the description to changes made in 661c374
(isl_pw_qpolynomial_gist: also try to eliminate integer divisions,
Fri Nov 5 16:44:54 2010 +0100).
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 11 Nov 2010 12:04:22 +0000 (13:04 +0100)]
add isl_set_foreach_orthant
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 11 Nov 2010 12:03:51 +0000 (13:03 +0100)]
isl_range.c: privately export isl_qpolynomial_terms_of_sign
Outside users may call the function on quasi-polynomials, so allow
integer division, but assume the caller has made sure they are always
positive.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 11 Nov 2010 11:43:22 +0000 (12:43 +0100)]
isl_range.c: collect_fixed_sign_terms: drop dead code
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 10 Nov 2010 20:49:23 +0000 (21:49 +0100)]
isl_set_split_dims: extract out half-space creation
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Fri, 12 Nov 2010 09:15:32 +0000 (10:15 +0100)]
isl_printer_print_pw_qpolynomial_fold: use consistent names in C format
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 11 Nov 2010 13:31:47 +0000 (14:31 +0100)]
isl_dim_total: handle NULL input
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 11 Nov 2010 11:42:50 +0000 (12:42 +0100)]
add isl_vec_clr
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 10 Nov 2010 12:49:30 +0000 (13:49 +0100)]
isl_qpolynomial_div_pow: normalize given div
An isl_div obtained from an isl_set would normalize be expected
to have been normalized already, but the user may have constructed
the isl_div manually and then we want to make sure it gets normalized.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 8 Nov 2010 13:31:32 +0000 (14:31 +0100)]
isl_bound: accept folds as input
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 8 Nov 2010 12:44:02 +0000 (13:44 +0100)]
isl_bound: use isl_pw_qpolynomial_folds internally
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 8 Nov 2010 12:42:01 +0000 (13:42 +0100)]
export isl_pw_qpolynomial_fold_fix_dim
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sat, 6 Nov 2010 14:36:38 +0000 (15:36 +0100)]
add isl_union_set_polyhedral_hull
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Fri, 5 Nov 2010 15:44:54 +0000 (16:44 +0100)]
isl_pw_qpolynomial_gist: also try to eliminate integer divisions
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Fri, 5 Nov 2010 15:13:29 +0000 (16:13 +0100)]
hide some functions that were exposed by mistake
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 3 Nov 2010 20:15:25 +0000 (21:15 +0100)]
isl_qpolynomial_substitute_equalities: eliminate integer divisions by 1
In particular, replace them by the dividend.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 4 Nov 2010 16:40:42 +0000 (17:40 +0100)]
isl_stream_read_pw_qpolynomial: read nested divs
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 4 Nov 2010 16:39:11 +0000 (17:39 +0100)]
isl_qpolynomial_div_pow: allow nested divs
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 4 Nov 2010 16:56:02 +0000 (17:56 +0100)]
isl_printer_print_qpolynomial: properly print nested divs
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 3 Nov 2010 15:19:08 +0000 (16:19 +0100)]
isl_qpolynomial_substitute_equalities: ignore constant when removing factors
The constant term of the affine expression inside an integer division
should not be taking to account when trying to remove common factors
between numerator and denominator. For example, [(1 + 2 x)/4]
should be simplified to [x/2].
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 4 Nov 2010 21:16:20 +0000 (22:16 +0100)]
isl_qpolynomial_substitute_equalities: don't use equalities with existentials
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 1 Nov 2010 14:59:43 +0000 (15:59 +0100)]
isl_basic_map_underlying_set: remove any structure and names from the input
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 1 Nov 2010 10:01:20 +0000 (11:01 +0100)]
export isl_set_reset_dim
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sat, 30 Oct 2010 16:26:18 +0000 (18:26 +0200)]
isl_qpolynomial_morph: properly handle change in number of variables
The original implementation would only substitute the actual variables
and not the integer divisions, but if the number of variables changes,
then the positions of the integer divisions change too and so they have
to be adjusted as well.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 28 Oct 2010 18:49:06 +0000 (20:49 +0200)]
isl_stream_read_map: read_var_def: read var definition in "clean" basic map
The original code would mix up integer division already in bmap and newly
added integer divisions.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 27 Oct 2010 11:04:04 +0000 (13:04 +0200)]
add isl_set_remove_divs_involving_dims
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sat, 23 Oct 2010 12:33:19 +0000 (14:33 +0200)]
add isl_div_div
Sven Verdoolaege [Sat, 23 Oct 2010 12:41:59 +0000 (14:41 +0200)]
add isl_div_copy
Sven Verdoolaege [Tue, 19 Oct 2010 13:27:42 +0000 (15:27 +0200)]
export isl_pw_qpolynomial_fold_set_dim_name
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Tue, 26 Oct 2010 15:00:13 +0000 (17:00 +0200)]
doc: fix typo
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Tue, 19 Oct 2010 13:03:30 +0000 (15:03 +0200)]
export isl_pw_qpolynomial_fold_add
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Tue, 19 Oct 2010 12:50:40 +0000 (14:50 +0200)]
add isl_set_flatten_map
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 25 Oct 2010 11:49:22 +0000 (13:49 +0200)]
isl_pw_qpolynomial_gist: substitute equalities in polynomials
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 25 Oct 2010 11:37:41 +0000 (13:37 +0200)]
isl_pw_qpolynomial_gist: drop empty pieces in result
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Mon, 25 Oct 2010 11:37:04 +0000 (13:37 +0200)]
isl_map_gist_basic_map: drop empty pieces in result
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sat, 23 Oct 2010 11:42:21 +0000 (13:42 +0200)]
rename isl_map_remove to isl_map_remove_dims
The new name is more consistent with other functions and avoids confusion
as to what is being added.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Fri, 22 Oct 2010 19:42:50 +0000 (21:42 +0200)]
isl_pw_qpolynomial_print: properly print divs in ISL_FORMAT_C
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Tue, 19 Oct 2010 12:34:10 +0000 (14:34 +0200)]
isl_stream: allow tokens to start with "_"
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Tue, 19 Oct 2010 12:50:28 +0000 (14:50 +0200)]
isl_set_alloc_dim: fix memory leak on error path
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Tue, 26 Oct 2010 09:25:03 +0000 (11:25 +0200)]
fix isl_union_{set,map}_lex_g{e,t}_union_map
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Tue, 26 Oct 2010 09:21:26 +0000 (11:21 +0200)]
isl_stream_read_map: read_var_list: simplify and finalize result
read_var_list may add equalities, so at the very least we need
to gauss the constraints and finalize the result, but it doesn't
seem to hurt to also simplify the result.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 20 Oct 2010 20:16:15 +0000 (22:16 +0200)]
isl_dim_replace: fix handling of nested spaces
5ad911e (isl_dim_drop: also drop parameters from nested spaces,
Sat Oct 16 14:41:22 2010 +0200) changed isl_dim_drop to also
drop parameters from nested spaces, but it didn't change isl_dim_add
accordingly, breaking isl_dim_replace on isl_dims with nested spaces.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Fri, 15 Oct 2010 09:14:57 +0000 (11:14 +0200)]
support extended PolyLib output format
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sun, 17 Oct 2010 09:36:40 +0000 (11:36 +0200)]
doc: mention LaTeX output format
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 14 Oct 2010 13:49:52 +0000 (15:49 +0200)]
add isl_union_set_identity
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sat, 16 Oct 2010 13:05:24 +0000 (15:05 +0200)]
basic_map_read_polylib: clear newly created divs
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Thu, 14 Oct 2010 09:28:33 +0000 (11:28 +0200)]
compute explicit divs for context before calling isl_basic_map_gist
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Tue, 12 Oct 2010 09:18:27 +0000 (11:18 +0200)]
add isl_set_flatten
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sun, 17 Oct 2010 09:05:28 +0000 (11:05 +0200)]
isl_dim_flatten: remove space names when internal structure changes
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sat, 16 Oct 2010 12:50:59 +0000 (14:50 +0200)]
isl_stream_read_map: properly parse nested spaces
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sat, 16 Oct 2010 12:41:22 +0000 (14:41 +0200)]
isl_dim_drop: also drop parameters from nested spaces
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Sat, 16 Oct 2010 11:41:20 +0000 (13:41 +0200)]
privately export isl_basic_map_reset_dim
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
Sven Verdoolaege [Wed, 13 Oct 2010 20:02:25 +0000 (22:02 +0200)]
isl_map_coalesce: handle some cases of pairs of adjacent equalities
We don't want to handle all cases, because coalescing maps that
are too dissimilar may reduce the effectiveness of some decomposition
techniques during the computation of transitive closure.
Right now, we only perform such coalescing if there is exactly
one pair of adjacent equalities. It's not clear if this is the best
heuristic, but it makes all transitive closure test cases pass.
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
projects / platform / upstream / isl.git / log