Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

This was missing from barvinok-0.41.1-1-g094c67b706 (update isl
for move of interface/all.h, Tue Jul 31 07:54:05 2018 +0200).

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

This was missing from 7a84e1b5 (barvinok_{summate,bound}: take isl input,
Thu Nov 11 14:29:48 2010 +0100).

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

The isl.py target in itself is fairly harmless since isl.py
is not built by default and therefore only results in failures
when the bindings are being built explicitly (or on "make dist).
However, barvinok-0.40-18-gb2a61c464a (interface: make language bindings
depend on all relevant headers, Sun Feb 18 12:30:10 2018 +0100)
added a libdep.a target which does get built by default while
it can only be built with a bundled or build isl submodule.
Hide the entire isl.py machinery behind a conditional to avoid
failures when a system isl submodule is being used.

Reported-by: Marcin Copik <mcopik@gmail.com>
Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

Do so instead of relying on the headers getting included indirectly.

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

Do so instead of relying on the headers getting included indirectly.

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

Do so instead of relying on the headers getting included indirectly.

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

Reported-by: Uday Reddy B <udayreddy@gmail.com>
Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

This was missing from barvinok-0.40-13-g3f17e4d8 (AX_SUBMODULE: only check
submodule status in git repository, Fri Jun 2 11:32:40 2017 +0200).

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

This file is needed for "make distcheck".

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

This file is needed to recreate the documentation.

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

The typo was introduced in barvinok-0.40-13-g3f17e4d8db (AX_SUBMODULE:
only check submodule status in git repository, Fri Jun 2 11:32:40 2017 +0200).

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

The codegen operation prints an AST generated from a schedule,
given either as a schedule tree or a (union) map.
As a convenience, applying codegen to a set would result
in codegen being applied to the identity mapping on the set.
This seems to be causing confusion about the nature
of the codegen operation.  Drop the magic such that users
are aware that a schedule is needed to perform AST generation.
They can still explicitly call the "identity" operation on a set
if they really want the old behavior.

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

This was missing from f24cadb2 (introduce special version of isl.py.top
for use with pet, Tue Dec 8 15:58:18 2015 +0100).
Since a distribution always contains the pet submodule,
only packaging the pet version of interface/isl.py.top
should be fairly harmless, but it does result in an error message
about the missing file and it may break if the user decides
not to use the pet submodule.

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

Commit 5bd2d70c (AX_SUBMODULE: allow .git file in submodule,
Wed Dec 4 16:03:53 2013 +0100) changed the way an initialized
submodule is detected, but it did so in a way that can only
be applied when the toplevel is a git repository.
Applied it outside a git repository does not affect the outcome,
but it does result in the error message
"fatal: Not a git repository (or any of the parent directories): .git".
Only perform this check after checking that the toplevel is a git repository.

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

Without the cast, the pointer is treated as an integer and may get truncated.

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

Signed-off-by: Sven Verdoolaege <sven.verdoolaege@gmail.com>

The function isl_space_match is undocumented.  Use the recently
introduced isl_space_has_equal_params instead.

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

This function will be needed in the next commit.

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

The function verify_point_data_init turns parameters into
input dimensions in order to be able to count the number
of parameter values.  However, it does this without explicitly
introducing a set tuple, relying instead on isl_set_move_dims
introducing such a set tuple implicitly.
Introduce the set tuple explicitly.

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

The function basic_map_card turns input dimensions into parameters
in the input and then turns those parameters back into input dimensions
in the result.  The input tuple is completely removed in the input,
but it was not getting explicitly introduced again in the output.
Instead, it was relying on isl_pw_qpolynomial_move_dims implicitly
introducing the input tuple, but this is an undocumented feature
that is not guaranteed to persist.
Introduce the domain explicitly.

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

This function will be needed in the next commit.

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