add isl_qpolynomial_fold_scale_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_qpolynomial_scale_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_qpolynomial_val_on_domain

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_qpolynomial_get_constant_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_term_get_coefficient_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_aff_mod_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_aff_scale_down_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_aff_scale_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_aff_get_denominator_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_aff_get_coefficient_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_aff_get_constant_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_aff_add_coefficient_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_aff_add_constant_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_aff_set_coefficient_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_aff_set_constant_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_constraint_get_coefficient_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_constraint_get_constant_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_constraint_set_coefficient_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_constraint_set_constant_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_point_get_coordinate_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_point_set_coordinate_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_mat_get_element_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_mat_set_element_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_vec_set_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_vec_get_element_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_vec_set_element_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add GMP specific isl_val functions

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_multi_val_mod_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_multi_val_add_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_multi_val

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_multi_templ.c: isl_multi_*_set_*: extract out check for matching space

We will be adding isl_multi_val next, where such a check does not make sense.

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_val_list

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_val abstraction

An isl_val can represent integer and rational values as well as some
special values.  isl_vals are meant to replace isl_ints in the external
interface as well as some abuses of isl_qpolynomials.

The advantage of isl_vals is that they behave like other isl objects and
that they do not expose how integer values are represented internally.
This means that the user will not have to take into account that GMP
is being used and that GMP can be replaced by another library without
affecting the user.

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_int_submul_ui

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_int_addmul_ui

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_int_mul_si

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_int_fits_ulong

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_int_fits_slong

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

remove isl_int_gcdext from public isl/int.h

isl_int_gcdext was introduced after the latest release, specifically
in e191579 (add isl_int_gcdext, Wed Apr 17 17:51:28 2013 +0200), so
we do not need to worry about backward compatibility.

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

create private copy of isl/int.h

We will shortly be introducing a new isl_val abstraction
that is meant to replace isl_int as well as some abuses of isl_qpolynomial.
To implement this new abstraction, we will need some additional
isl_int_* functions.  However, we do not want to publicly expose
these isl_int_* functions since we do not want to encourage their
use by external users.
We therefore create a private copy of isl/int.h where we can
add the additional functions.  Once the transition is complete,
the public version can be removed.

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

Merge branch 'maint'

isl_union_set_compute_schedule: fix check for progress in feautrier scheduler

In carry_dependences we check that we have been able to carry at least
(part of) one dependence relation, by checking that the number of
edges that are not carried is smaller than the total number of edges.
However, we were only looking at the numerator.  If the optimal LP
solution is integral then this happens to work out, but otherwise
we could overestimate the number of non-carried edges, possibly
leading to the wrong conclusion.

Reported-by: Tomofumi Yuki <tomofumi.yuki@gmail.com>
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_vertex_get_domain: return parameter domain

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_ast_expr_from_constraint: split off constant term if it avoids operation

If one hand of the comparison operation that is constructed would be
exactly zero, then we put the constant term on that side.
This avoids the operation that adds the constant term to the other side,
resulting in slightly smaller expressions that are easier to read and
that may also reduce the risk of overflows.

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

Merge branch 'maint'

isl_basic_map_foreach_lexopt: properly merge partial empty solutions

b9f95db (isl_basic_map_foreach_lexopt: properly merge partial solutions,
Wed Mar 20 09:55:50 2013 +0100) fixed the merging of partial non-empty
solutions, but did not take into account that partial solutions may
be empty.  If the partial solutions are empty, then M is NULL and we
cannot dereference it.

Note that the new test case only triggers on top of d7e5adb
(isl_tab_basic_map_partial_lexopt: better exploit partial solution cache,
Tue Mar 12 17:25:04 2013 +0100) in the master branch.

Reported-by: Tomofumi Yuki <tomofumi.yuki@gmail.com>
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

merge shared isl_multi_* declarations

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

declare isl_multi_* together with isl_*

This will make it easier to add new isl_multi_* types.

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

make isl_multi_* files more self-contained

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_space_private.h: protect against multiple inclusions

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

detect missing gmp.h or missing/old gmp library during configure

Until now a missing gmp.h was only detected during compilation
and a missing/old library only during linking.

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_ast_codegen.c: add_nodes: use more transparent way to break up ordering SCC

The original method for breaking up an ordering SCC depended on
isl_set_make_disjoint making the basic sets rationally disjoint.
We prefer not to depend on core isl internals in the AST generator.

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_basic_map_order_gt

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

bound.c: verify_point: use isl_printer

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_input.c: read_factor: declare variables at start of code block

Reported-by: William Bader <william.bader@gmail.com>
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

Merge branch 'maint'

isl_*alloc*: return NULL if ctx argument is NULL

Whenever an isl function needs to perform an allocation, it calls
one of the isl_*alloc* functions on an isl_ctx obtained either from
the user or from a call to one of the *_get_ctx functions.
If this isl_ctx pointer is NULL, this means that either
the original isl_ctx_alloc failed or that the pointer on which
the *_get_ctx function was called was NULL.
In both cases, we are in an error state and we should not
allocate any more memory, but instead return as soon as possible.
We therefore return NULL from the isl_*alloc* functions,
an error condition which should already be handled by the caller.

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_space.c: add missing include

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

declare isl_tab_get_ctx in isl_tab.h

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

declare isl_*_list together with isl_*

Some of the isl_*_list types were declared together in isl/list.h.
It seems cleaner to declare them together with the corresponding isl_* instead.
Similarly, these types were defined together in isl_list_private.h and
are now defined together with the corresponding isl_*.

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_ast_graft.c: remove spurious include

The include has been redundant ever since it was introduced in
5888ac1 (add support for generating ASTs from schedule relations,
Mon Sep 17 22:11:06 2012 +0200).

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_ast_codegen.c: remove dependence on internals of isl_constraint_list

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

extract generic isl_*_list_foreach_scc from isl_ast_codegen.c

isl_ast_codegen.c contained two implementations of this function
that have now been replaced by two calls to the extracted functions.
This makes isl_ast_codegen.c less dependent on the internals of
isl_basic_set_list.  We should also be able to later reuse the
extracted function in other contexts.  Once all users of isl_tarjan
have been converted to this interface, we probably want to remove
the old interface and perform the calls to callback as soon as
each SCC is found rather than waiting until all of them have been found.

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_map_affine_hull: make stride information explicit before dropping divs

In a2cbe44 (isl_map_affine_hull: avoid computing explicit representations
for divs, Sun Jan 15 16:12:59 2012 +0100), isl_map_affine_hull was changed
to drop unknown divs rather than computing an explicit representation
for them.  This caused some stride information that involves multiple
unknown divs to get dropped.
Try and recover this lost information by making stride information
explicitly available before dropping the unknown divs.

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_morph.c: extract out isl_mat_parameter_compression_ext

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_ast_build_get_stride_constraint: combine multiple strides

If we find multiple constraints from which can derive a stride,
then we would only keep the stride information from the last
of these constraints.  We now combine multiple strides into
a single stride.  Note that this may not have any effect in practice
because multiple strides are usually already combined into one
at the level of the constraints of a basic set.

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_ast_build.c: use isl_int_gcdext instead of open-coded version

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_int_gcdext

We will use it in the next commit.

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_ast_build_ast_from_schedule: also add stride guard in generic case

Any guard implied by the stride constraint in the AST build domain
needs to be enforced by the generated AST and may not necessarily
be implied by the strided loop itself.
We were already adding this guard to the graft in the degenerate case,
but not in the generic case.

The implied guard is usually already available in loop bounds constraints,
so it is not clear if this fixes any bug that can appear in practice,
However, this may just be a consequence of the internal representation of sets
and we should not depend too much on this internal representation.

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_ast_build_set_loop_bounds: compute divs after eliminating strided dimension

Although it is unlikely that the elimination would result in any unknown divs,
we had better make sure to preserve the invariant that build->domain does not
have any unknown divs.

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_map_simplify.c: normalize_divs_in_context: improve error handling

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_ast_build_expr.c: remove left-over from refactoring

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_ast_codegen.c: create_node_scaled: reflect current domain slice in schedule

If we detect that a given schedule dimension has a fixed (possibly
parametric) value, then this value is recorded in the build and
the schedule dimension is eliminated.  This elimination process may
drop constraints on the part of the domain that we are currently considering,
which may cause all sorts of problems.  In particular, we may fail
to discover at the end of the AST generation that the original schedule
was single valued and insert spurious loops.

Plug in the value we found instead of simply eliminating the schedule
dimension.

Reported-by: Tobias Grosser <tobias@grosser.es>
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_union_map_preimage_domain_multi_aff

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add isl_map_preimage_domain_multi_aff

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_basic_map_preimage_multi_aff: micro-refactoring

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_map_simplify.c: div_is_redundant: ignore coefficients of unknown divs

If a div is unknown, then its coefficients are undefined so we should
not take them into account while checking if some other div is redundant.
In rare cases, looking at those undefined coefficients could result
in system dependent behavior.

Reported-by: Tobias Grosser <tobias@grosser.es>
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

generalize isl_basic_set_preimage_multi_aff to isl_basic_map_preimage_multi_aff

This will allow us to implement isl_map_preimage_domain_multi_aff.

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_ast_build_ast_from_schedule: also add stride guard in generic case

Any guard implied by the stride constraint in the AST build domain
needs to be enforced by the generated AST and may not necessarily
be implied by the strided loop itself.
We were already adding this guard to the graft in the degenerate case,
but not in the generic case.

The implied guard is usually already available in loop bounds constraints,
so it is not clear if this fixes any bug that can appear in practice,
However, this may just be a consequence of the internal representation of sets
and we should not depend too much on this internal representation.

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_ast_build_set_loop_bounds: compute divs after eliminating strided dimension

Although it is unlikely that the elimination would result in any unknown divs,
we had better make sure to preserve the invariant that build->domain does not
have any unknown divs.

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_map_simplify.c: normalize_divs_in_context: improve error handling

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_map.c: fix typo in comment

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

extend isl_seq_preimage to allow taking preimage of a subset of dimensions

This will allow us to reuse isl_seq_preimage for computing the preimage
of the domain or range of a basic map.

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_map_simplify.c: div_is_redundant: ignore coefficients of unknown divs

If a div is unknown, then its coefficients are undefined so we should
not take them into account while checking if some other div is redundant.
In rare cases, looking at those undefined coefficients could result
in system dependent behavior.

Reported-by: Tobias Grosser <tobias@grosser.es>
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

Merge branch 'maint'

isl_ast_codegen.c: fix typo in comment

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_ast_build_ast_from_schedule: properly handle union atomic options

In particular, if the domain for which a given dimension should
be atomic is a union of basic sets, then the actual atomic domain
may be larger than this union and we need to make sure that this
larger set is removed from the remainder domain to preserve
the invariant that the elements of the list computed by compute_domains
are disjoint.

Reported-by: Tobias Grosser <tobias@grosser.es>
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_ast_codegen.c: update_unrolling_lower_bound: check for overflow

If the number of slices along a given lower bound exceeds the maximal
value of an integer, then it would mistakenly be replaced by its
least signficant bits, resulting in slices being dropped if the lower
bound ends up being selected as the best (based on the wrong number).
If the number of slices really exceeds the maximal value of an integer,
then we do not want to honor the request of the user anyway.
We should probably set an even smaller limit.

Reported-by: Tobias Grosser <tobias@grosser.es>
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_mat_free: return NULL

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_basic_map_insert_dims: plug memory leak when called on empty input

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

ChangeLog: drop some text that appears to have been copied by mistake

In particular, the text was copied by 9fe3e28 (isl 0.10,
Sun Jun 3 15:47:09 2012 +0200).

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl 0.11.2

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

add missing interface/all.h and interface/isl.py.top to distribution

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_schedule.c: sort_band_list: take into account that parent may be NULL

We missed this in 81bf4c2 (isl_schedule_get_band_forest: sort bands
in band list, Tue Feb 26 12:35:16 2013 +0100).

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_schedule_get_map: pad schedule map with zeros if reconstructed from bands

The schedule map extracted from the isl_schedule_nodes is padded with zeros
by construction.  However, if the schedule map is reconstructed from the
band forest, then the schedule map was not be padded with zeros.
Pad this schedule as well.

We may want to change isl_band_list_get_suffix_schedule to perform this
padded instead.

Reported-by: Ali El Moussawi <elmoussawi.ali@gmail.com>
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_ast_graft.c: store_guard: coalesce guard after taking gist

On rare occasions, coalescing may simplify the gisted guard.

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

Merge branch 'maint'

isl_ast_node_print: avoid duplicate declarations in printed code

If a loop has only one iteration but if the expression
for this single iteration is too complicated to be substituted into the AST,
then the expression is assigned to a variable corresponding to the loop
iterator.
If two such loops occur in the same block, then we would end up with
two declarations of the same variable.  Avoid this problem by introducing
a block around the declaration if it appears in a sequence of statements.

Reported-by: Tobias Grosser <tobias@grosser.es>
Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

isl_ast.c: extract out start_block and end_block from print_ast_node_c

We will be able to reuse them in the next commit.

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>

configure.ac: explicitly link in llvm bitreader component in extract_interface

Apparently, the bitreader component is required for ASTReader and
is no longer included implicitly.

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>