=back
+=head3 Changes since isl-0.09
+
+=over
+
+=item * The C<schedule_split_parallel> option has been replaced
+by the C<schedule_split_scaled> option.
+
+=item * The first argument of C<isl_pw_aff_cond> is now
+an C<isl_pw_aff> instead of an C<isl_set>.
+A call C<isl_pw_aff_cond(a, b, c)> can be replaced by
+
+ isl_pw_aff_cond(isl_set_indicator_function(a), b, c)
+
+=back
+
=head1 Installation
The source of C<isl> can be obtained either as a tarball
=back
+=head2 Error Handling
+
+C<isl> supports different ways to react in case a runtime error is triggered.
+Runtime errors arise, e.g., if a function such as C<isl_map_intersect> is called
+with two maps that have incompatible spaces. There are three possible ways
+to react on error: to warn, to continue or to abort.
+
+The default behavior is to warn. In this mode, C<isl> prints a warning, stores
+the last error in the corresponding C<isl_ctx> and the function in which the
+error was triggered returns C<NULL>. An error does not corrupt internal state,
+such that isl can continue to be used. C<isl> also provides functions to
+read the last error and to reset the memory that stores the last error. The
+last error is only stored for information purposes. Its presence does not
+change the behavior of C<isl>. Hence, resetting an error is not required to
+continue to use isl, but only to observe new errors.
+
+ #include <isl/ctx.h>
+ enum isl_error isl_ctx_last_error(isl_ctx *ctx);
+ void isl_ctx_reset_error(isl_ctx *ctx);
+
+Another option is to continue on error. This is similar to warn on error mode,
+except that C<isl> does not print any warning. This allows a program to
+implement its own error reporting.
+
+The last option is to directly abort the execution of the program from within
+the isl library. This makes it obviously impossible to recover from an error,
+but it allows to directly spot the error location. By aborting on error,
+debuggers break at the location the error occurred and can provide a stack
+trace. Other tools that automatically provide stack traces on abort or that do
+not want to continue execution after an error was triggered may also prefer to
+abort on error.
+
+The on error behavior of isl can be specified by calling
+C<isl_options_set_on_error> or by setting the command line option
+C<--isl-on-error>. Valid arguments for the function call are
+C<ISL_ON_ERROR_WARN>, C<ISL_ON_ERROR_CONTINUE> and C<ISL_ON_ERROR_ABORT>. The
+choices for the command line option are C<warn>, C<continue> and C<abort>.
+It is also possible to query the current error mode.
+
+ #include <isl/options.h>
+ int isl_options_set_on_error(isl_ctx *ctx, int val);
+ int isl_options_get_on_error(isl_ctx *ctx);
+
=head2 Identifiers
Identifiers are used to identify both individual dimensions
__isl_keep isl_pw_multi_aff *pma);
__isl_give isl_space *isl_pw_multi_aff_get_space(
__isl_keep isl_pw_multi_aff *pma);
+ __isl_give isl_space *isl_union_pw_multi_aff_get_space(
+ __isl_keep isl_union_pw_multi_aff *upma);
#include <isl/point.h>
__isl_give isl_space *isl_point_get_space(
__isl_give isl_id *isl_space_get_dim_id(
__isl_keep isl_space *space,
enum isl_dim_type type, unsigned pos);
- __isl_give isl_space *isl_space_set_dim_name(__isl_take isl_space *space,
- enum isl_dim_type type, unsigned pos,
- __isl_keep const char *name);
- __isl_keep const char *isl_space_get_dim_name(__isl_keep isl_space *space,
- enum isl_dim_type type, unsigned pos);
+ __isl_give isl_space *isl_space_set_dim_name(
+ __isl_take isl_space *space,
+ enum isl_dim_type type, unsigned pos,
+ __isl_keep const char *name);
+ int isl_space_has_dim_name(__isl_keep isl_space *space,
+ enum isl_dim_type type, unsigned pos);
+ __isl_keep const char *isl_space_get_dim_name(
+ __isl_keep isl_space *space,
+ enum isl_dim_type type, unsigned pos);
Note that C<isl_space_get_name> returns a pointer to some internal
data structure, so the result can only be used while the
__isl_give isl_space *isl_space_set_tuple_name(
__isl_take isl_space *space,
enum isl_dim_type type, const char *s);
+ int isl_space_has_tuple_name(__isl_keep isl_space *space,
+ enum isl_dim_type type);
const char *isl_space_get_tuple_name(__isl_keep isl_space *space,
enum isl_dim_type type);
unsigned n);
__isl_give isl_space *isl_space_map_from_set(
__isl_take isl_space *space);
+ __isl_give isl_space *isl_space_map_from_domain_and_range(
+ __isl_take isl_space *domain,
+ __isl_take isl_space *range);
__isl_give isl_space *isl_space_zip(__isl_take isl_space *space);
+ __isl_give isl_space *isl_space_curry(
+ __isl_take isl_space *space);
Note that if dimensions are added or removed from a space, then
the name and the internal structure are lost.
A local space is essentially a space with
zero or more existentially quantified variables.
-The local space of a basic set or relation can be obtained
+The local space of a (constraint of a) basic set or relation can be obtained
using the following functions.
+ #include <isl/constraint.h>
+ __isl_give isl_local_space *isl_constraint_get_local_space(
+ __isl_keep isl_constraint *constraint);
+
#include <isl/set.h>
__isl_give isl_local_space *isl_basic_set_get_local_space(
__isl_keep isl_basic_set *bset);
int isl_local_space_is_set(__isl_keep isl_local_space *ls);
int isl_local_space_dim(__isl_keep isl_local_space *ls,
enum isl_dim_type type);
+ int isl_local_space_has_dim_name(
+ __isl_keep isl_local_space *ls,
+ enum isl_dim_type type, unsigned pos)
const char *isl_local_space_get_dim_name(
__isl_keep isl_local_space *ls,
enum isl_dim_type type, unsigned pos);
__isl_give isl_local_space *isl_local_space_domain(
__isl_take isl_local_space *ls);
+ __isl_give isl_local_space *isl_local_space_range(
+ __isl_take isl_local_space *ls);
__isl_give isl_local_space *isl_local_space_from_domain(
__isl_take isl_local_space *ls);
__isl_give isl_local_space *isl_local_space_intersect(
__isl_give char *isl_printer_get_str(
__isl_keep isl_printer *printer);
+The printer can be inspected using the following functions.
+
+ FILE *isl_printer_get_file(
+ __isl_keep isl_printer *printer);
+ int isl_printer_get_output_format(
+ __isl_keep isl_printer *p);
+
The behavior of the printer can be modified in various ways
__isl_give isl_printer *isl_printer_set_output_format(
of C<isl_dim_cst>, C<isl_dim_param>,
C<isl_dim_in>, C<isl_dim_out> and C<isl_dim_div> for relations.
-A (basic) set or relation can also be constructed from a (piecewise)
-(multiple) affine expression
+A (basic or union) set or relation can also be constructed from a
+(union) (piecewise) (multiple) affine expression
or a list of affine expressions
(See L<"Piecewise Quasi Affine Expressions"> and
L<"Piecewise Multiple Quasi Affine Expressions">).
__isl_give isl_basic_map *isl_basic_map_from_aff(
__isl_take isl_aff *aff);
+ __isl_give isl_map *isl_map_from_aff(
+ __isl_take isl_aff *aff);
__isl_give isl_set *isl_set_from_pw_aff(
__isl_take isl_pw_aff *pwaff);
__isl_give isl_map *isl_map_from_pw_aff(
__isl_take isl_aff_list *list);
__isl_give isl_basic_map *isl_basic_map_from_multi_aff(
__isl_take isl_multi_aff *maff)
+ __isl_give isl_map *isl_map_from_multi_aff(
+ __isl_take isl_multi_aff *maff)
__isl_give isl_set *isl_set_from_pw_multi_aff(
__isl_take isl_pw_multi_aff *pma);
__isl_give isl_map *isl_map_from_pw_multi_aff(
__isl_take isl_pw_multi_aff *pma);
+ __isl_give isl_union_map *
+ isl_union_map_from_union_pw_multi_aff(
+ __isl_take isl_union_pw_multi_aff *upma);
The C<domain_dim> argument describes the domain of the resulting
basic relation. It is required because the C<list> may consist
#include <isl/constraint.h>
+ int isl_basic_set_n_constraint(
+ __isl_keep isl_basic_set *bset);
+ int isl_basic_set_foreach_constraint(
+ __isl_keep isl_basic_set *bset,
+ int (*fn)(__isl_take isl_constraint *c, void *user),
+ void *user);
int isl_basic_map_foreach_constraint(
__isl_keep isl_basic_map *bmap,
int (*fn)(__isl_take isl_constraint *c, void *user),
The coefficients of the constraints can be inspected using
the following functions.
+ int isl_constraint_is_lower_bound(
+ __isl_keep isl_constraint *constraint,
+ enum isl_dim_type type, unsigned pos);
+ int isl_constraint_is_upper_bound(
+ __isl_keep isl_constraint *constraint,
+ enum isl_dim_type type, unsigned pos);
void isl_constraint_get_constant(
__isl_keep isl_constraint *constraint, isl_int *v);
void isl_constraint_get_coefficient(
__isl_keep isl_basic_set *bset);
__isl_give isl_basic_set *isl_basic_set_set_tuple_name(
__isl_take isl_basic_set *set, const char *s);
+ int isl_set_has_tuple_name(__isl_keep isl_set *set);
const char *isl_set_get_tuple_name(
__isl_keep isl_set *set);
const char *isl_basic_map_get_tuple_name(
__isl_keep isl_basic_map *bmap,
enum isl_dim_type type);
+ __isl_give isl_basic_map *isl_basic_map_set_tuple_name(
+ __isl_take isl_basic_map *bmap,
+ enum isl_dim_type type, const char *s);
const char *isl_map_get_tuple_name(
__isl_keep isl_map *map,
enum isl_dim_type type);
The identifiers, positions or names of individual dimensions can be
read off using the following functions.
+ __isl_give isl_id *isl_basic_set_get_dim_id(
+ __isl_keep isl_basic_set *bset,
+ enum isl_dim_type type, unsigned pos);
__isl_give isl_set *isl_set_set_dim_id(
__isl_take isl_set *set, enum isl_dim_type type,
unsigned pos, __isl_take isl_id *id);
enum isl_dim_type type, __isl_keep isl_id *id);
int isl_map_find_dim_by_id(__isl_keep isl_map *map,
enum isl_dim_type type, __isl_keep isl_id *id);
+ int isl_set_find_dim_by_name(__isl_keep isl_set *set,
+ enum isl_dim_type type, const char *name);
int isl_map_find_dim_by_name(__isl_keep isl_map *map,
enum isl_dim_type type, const char *name);
const char *isl_basic_set_get_dim_name(
__isl_keep isl_basic_set *bset,
enum isl_dim_type type, unsigned pos);
+ int isl_set_has_dim_name(__isl_keep isl_set *set,
+ enum isl_dim_type type, unsigned pos);
const char *isl_set_get_dim_name(
__isl_keep isl_set *set,
enum isl_dim_type type, unsigned pos);
=item * Single-valuedness
+ int isl_basic_map_is_single_valued(
+ __isl_keep isl_basic_map *bmap);
+ int isl_map_plain_is_single_valued(
+ __isl_keep isl_map *map);
int isl_map_is_single_valued(__isl_keep isl_map *map);
int isl_union_map_is_single_valued(__isl_keep isl_union_map *umap);
can be computed,
i.e., whether both domain and range are nested relations.
+=item * Currying
+
+ int isl_basic_map_can_curry(
+ __isl_keep isl_basic_map *bmap);
+ int isl_map_can_curry(__isl_keep isl_map *map);
+
+Check whether the domain of the (basic) relation is a wrapped relation.
+
=back
=head3 Binary Properties
=item * Subset
+ int isl_basic_set_is_subset(
+ __isl_keep isl_basic_set *bset1,
+ __isl_keep isl_basic_set *bset2);
int isl_set_is_subset(__isl_keep isl_set *set1,
__isl_keep isl_set *set2);
int isl_set_is_strict_subset(
__isl_give isl_set *isl_set_complement(
__isl_take isl_set *set);
+ __isl_give isl_map *isl_map_complement(
+ __isl_take isl_map *map);
=item * Inverse map
__isl_take isl_basic_map *bmap,
enum isl_dim_type type,
unsigned first, unsigned n);
+ __isl_give isl_map *isl_map_eliminate(
+ __isl_take isl_map *map, enum isl_dim_type type,
+ unsigned first, unsigned n);
Eliminate the coefficients for the given dimensions from the constraints,
without removing the dimensions.
__isl_give isl_basic_map *isl_basic_map_lower_bound_si(
__isl_take isl_basic_map *bmap,
enum isl_dim_type type, unsigned pos, int value);
+ __isl_give isl_set *isl_set_lower_bound(
+ __isl_take isl_set *set,
+ enum isl_dim_type type, unsigned pos,
+ isl_int value);
__isl_give isl_set *isl_set_lower_bound_si(
__isl_take isl_set *set,
enum isl_dim_type type, unsigned pos, int value);
__isl_give isl_map *isl_map_lower_bound_si(
__isl_take isl_map *map,
enum isl_dim_type type, unsigned pos, int value);
+ __isl_give isl_set *isl_set_upper_bound(
+ __isl_take isl_set *set,
+ enum isl_dim_type type, unsigned pos,
+ isl_int value);
__isl_give isl_set *isl_set_upper_bound_si(
__isl_take isl_set *set,
enum isl_dim_type type, unsigned pos, int value);
enum isl_dim_type type, unsigned pos, int value);
Intersect the set or relation with the half-space where the given
-dimension has a value bounded the fixed given value.
+dimension has a value bounded by the fixed given value.
__isl_give isl_set *isl_set_equate(__isl_take isl_set *set,
enum isl_dim_type type1, int pos1,
enum isl_dim_type type2, int pos2);
+ __isl_give isl_basic_map *isl_basic_map_equate(
+ __isl_take isl_basic_map *bmap,
+ enum isl_dim_type type1, int pos1,
+ enum isl_dim_type type2, int pos2);
__isl_give isl_map *isl_map_equate(__isl_take isl_map *map,
enum isl_dim_type type1, int pos1,
enum isl_dim_type type2, int pos2);
Intersect the relation with the hyperplane where the given
dimensions have opposite values.
+ __isl_give isl_map *isl_map_order_gt(__isl_take isl_map *map,
+ enum isl_dim_type type1, int pos1,
+ enum isl_dim_type type2, int pos2);
+
+Intersect the relation with the half-space where the given
+dimensions satisfy the given ordering.
+
=item * Identity
__isl_give isl_map *isl_set_identity(
__isl_give isl_union_map *isl_union_map_coalesce(
__isl_take isl_union_map *umap);
+One of the methods for combining pairs of basic sets or relations
+can result in coefficients that are much larger than those that appear
+in the constraints of the input. By default, the coefficients are
+not allowed to grow larger, but this can be changed by unsetting
+the following option.
+
+ int isl_options_set_coalesce_bounded_wrapping(
+ isl_ctx *ctx, int val);
+ int isl_options_get_coalesce_bounded_wrapping(
+ isl_ctx *ctx);
+
=item * Detecting equalities
__isl_give isl_basic_set *isl_basic_set_detect_equalities(
In case of union sets and relations, the polyhedral hull is computed
per space.
+=item * Feasibility
+
+ __isl_give isl_basic_set *isl_basic_set_sample(
+ __isl_take isl_basic_set *bset);
+ __isl_give isl_basic_set *isl_set_sample(
+ __isl_take isl_set *set);
+ __isl_give isl_basic_map *isl_basic_map_sample(
+ __isl_take isl_basic_map *bmap);
+ __isl_give isl_basic_map *isl_map_sample(
+ __isl_take isl_map *map);
+
+If the input (basic) set or relation is non-empty, then return
+a singleton subset of the input. Otherwise, return an empty set.
+
=item * Optimization
#include <isl/ilp.h>
=item * Power
+ __isl_give isl_map *isl_map_fixed_power(
+ __isl_take isl_map *map, isl_int exp);
+ __isl_give isl_union_map *isl_union_map_fixed_power(
+ __isl_take isl_union_map *umap, isl_int exp);
+
+Compute the given power of C<map>, where C<exp> is assumed to be non-zero.
+If the exponent C<exp> is negative, then the -C<exp> th power of the inverse
+of C<map> is computed.
+
__isl_give isl_map *isl_map_power(__isl_take isl_map *map,
int *exact);
__isl_give isl_union_map *isl_union_map_power(
Given a relation with nested relations for domain and range,
interchange the range of the domain with the domain of the range.
+=item * Currying
+
+ __isl_give isl_basic_map *isl_basic_map_curry(
+ __isl_take isl_basic_map *bmap);
+ __isl_give isl_map *isl_map_curry(
+ __isl_take isl_map *map);
+ __isl_give isl_union_map *isl_union_map_curry(
+ __isl_take isl_union_map *umap);
+
+Given a relation with a nested relation for domain,
+move the range of the nested relation out of the domain
+and use it as the domain of a nested relation in the range,
+with the original range as range of this nested relation.
+
=item * Aligning parameters
__isl_give isl_set *isl_set_align_params(
__isl_give isl_union_map *isl_union_map_gist_domain(
__isl_take isl_union_map *umap,
__isl_take isl_union_set *uset);
+ __isl_give isl_union_map *isl_union_map_gist_range(
+ __isl_take isl_union_map *umap,
+ __isl_take isl_union_set *uset);
The gist operation returns a set or relation that has the
same intersection with the context as the input set or relation.
__isl_take isl_printer *p,
__isl_keep isl_set_list *list);
+=head2 Vectors
+
+Vectors can be created, copied and freed using the following functions.
+
+ #include <isl/vec.h>
+ __isl_give isl_vec *isl_vec_alloc(isl_ctx *ctx,
+ unsigned size);
+ __isl_give isl_vec *isl_vec_copy(__isl_keep isl_vec *vec);
+ void isl_vec_free(__isl_take isl_vec *vec);
+
+Note that the elements of a newly created vector may have arbitrary values.
+The elements can be changed and inspected using the following functions.
+
+ isl_ctx *isl_vec_get_ctx(__isl_keep isl_vec *vec);
+ int isl_vec_size(__isl_keep isl_vec *vec);
+ int isl_vec_get_element(__isl_keep isl_vec *vec,
+ int pos, isl_int *v);
+ __isl_give isl_vec *isl_vec_set_element(
+ __isl_take isl_vec *vec, int pos, isl_int v);
+ __isl_give isl_vec *isl_vec_set_element_si(
+ __isl_take isl_vec *vec, int pos, int v);
+ __isl_give isl_vec *isl_vec_set(__isl_take isl_vec *vec,
+ isl_int v);
+ __isl_give isl_vec *isl_vec_set_si(__isl_take isl_vec *vec,
+ int v);
+
+C<isl_vec_get_element> will return a negative value if anything went wrong.
+In that case, the value of C<*v> is undefined.
+
=head2 Matrices
Matrices can be created, copied and freed using the following functions.
__isl_give isl_pw_aff *isl_pw_aff_from_aff(
__isl_take isl_aff *aff);
+A piecewise quasi affine expression that is equal to 1 on a set
+and 0 outside the set can be created using the following function.
+
+ #include <isl/aff.h>
+ __isl_give isl_pw_aff *isl_set_indicator_function(
+ __isl_take isl_set *set);
+
Quasi affine expressions can be copied and freed using
#include <isl/aff.h>
const char *isl_pw_aff_get_dim_name(
__isl_keep isl_pw_aff *pa,
enum isl_dim_type type, unsigned pos);
+ int isl_pw_aff_has_dim_id(__isl_keep isl_pw_aff *pa,
+ enum isl_dim_type type, unsigned pos);
__isl_give isl_id *isl_pw_aff_get_dim_id(
__isl_keep isl_pw_aff *pa,
enum isl_dim_type type, unsigned pos);
__isl_give isl_aff *isl_aff_get_div(
__isl_keep isl_aff *aff, int pos);
+ int isl_pw_aff_n_piece(__isl_keep isl_pw_aff *pwaff);
int isl_pw_aff_foreach_piece(__isl_keep isl_pw_aff *pwaff,
int (*fn)(__isl_take isl_set *set,
__isl_take isl_aff *aff,
__isl_give isl_pw_aff *isl_pw_aff_coalesce(
__isl_take isl_pw_aff *pwqp);
+ __isl_give isl_aff *isl_aff_align_params(
+ __isl_take isl_aff *aff,
+ __isl_take isl_space *model);
__isl_give isl_pw_aff *isl_pw_aff_align_params(
__isl_take isl_pw_aff *pwaff,
__isl_take isl_space *model);
+ __isl_give isl_aff *isl_aff_project_domain_on_params(
+ __isl_take isl_aff *aff);
+
__isl_give isl_aff *isl_aff_gist_params(
__isl_take isl_aff *aff,
__isl_take isl_set *context);
#include <isl/aff.h>
__isl_give isl_pw_aff *isl_pw_aff_cond(
- __isl_take isl_set *cond,
+ __isl_take isl_pw_aff *cond,
__isl_take isl_pw_aff *pwaff_true,
__isl_take isl_pw_aff *pwaff_false);
The function C<isl_pw_aff_cond> performs a conditional operator
and returns an expression that is equal to C<pwaff_true>
-for elements
in C<cond> and equal to C<pwaff_false> for elements
-not in C<cond>.
+for elements
where C<cond> is non-zero and equal to C<pwaff_false> for elements
+where C<cond> is zero.
#include <isl/aff.h>
__isl_give isl_pw_aff *isl_pw_aff_union_min(
__isl_take isl_space *space,
__isl_take isl_aff_list *list);
-An empty piecewise multiple quasi affine expression (one with no cells) or
-a piecewise multiple quasi affine expression with a single cell can
-be created using the following functions.
+An empty piecewise multiple quasi affine expression (one with no cells),
+the zero piecewise multiple quasi affine expression (with value zero
+for each output dimension),
+a piecewise multiple quasi affine expression with a single cell (with
+either a universe or a specified domain) or
+a zero-dimensional piecewise multiple quasi affine expression
+on a given domain
+can be created using the following functions.
#include <isl/aff.h>
__isl_give isl_pw_multi_aff *isl_pw_multi_aff_empty(
__isl_take isl_space *space);
+ __isl_give isl_multi_aff *isl_multi_aff_zero(
+ __isl_take isl_space *space);
+ __isl_give isl_pw_multi_aff *
+ isl_pw_multi_aff_from_multi_aff(
+ __isl_take isl_multi_aff *ma);
__isl_give isl_pw_multi_aff *isl_pw_multi_aff_alloc(
__isl_take isl_set *set,
__isl_take isl_multi_aff *maff);
+ __isl_give isl_pw_multi_aff *isl_pw_multi_aff_from_domain(
+ __isl_take isl_set *set);
+
+ __isl_give isl_union_pw_multi_aff *
+ isl_union_pw_multi_aff_empty(
+ __isl_take isl_space *space);
+ __isl_give isl_union_pw_multi_aff *
+ isl_union_pw_multi_aff_add_pw_multi_aff(
+ __isl_take isl_union_pw_multi_aff *upma,
+ __isl_take isl_pw_multi_aff *pma);
+ __isl_give isl_union_pw_multi_aff *
+ isl_union_pw_multi_aff_from_domain(
+ __isl_take isl_union_set *uset);
A piecewise multiple quasi affine expression can also be initialized
from an C<isl_set> or C<isl_map>, provided the C<isl_set> is a singleton
void *isl_pw_multi_aff_free(
__isl_take isl_pw_multi_aff *pma);
+ __isl_give isl_union_pw_multi_aff *
+ isl_union_pw_multi_aff_copy(
+ __isl_keep isl_union_pw_multi_aff *upma);
+ void *isl_union_pw_multi_aff_free(
+ __isl_take isl_union_pw_multi_aff *upma);
+
The expression can be inspected using
#include <isl/aff.h>
__isl_keep isl_multi_aff *maff);
isl_ctx *isl_pw_multi_aff_get_ctx(
__isl_keep isl_pw_multi_aff *pma);
+ isl_ctx *isl_union_pw_multi_aff_get_ctx(
+ __isl_keep isl_union_pw_multi_aff *upma);
unsigned isl_multi_aff_dim(__isl_keep isl_multi_aff *maff,
enum isl_dim_type type);
unsigned isl_pw_multi_aff_dim(
enum isl_dim_type type);
__isl_give isl_aff *isl_multi_aff_get_aff(
__isl_keep isl_multi_aff *multi, int pos);
+ __isl_give isl_pw_aff *isl_pw_multi_aff_get_pw_aff(
+ __isl_keep isl_pw_multi_aff *pma, int pos);
const char *isl_pw_multi_aff_get_dim_name(
__isl_keep isl_pw_multi_aff *pma,
enum isl_dim_type type, unsigned pos);
__isl_take isl_multi_aff *maff,
void *user), void *user);
+ int isl_union_pw_multi_aff_foreach_pw_multi_aff(
+ __isl_keep isl_union_pw_multi_aff *upma,
+ int (*fn)(__isl_take isl_pw_multi_aff *pma,
+ void *user), void *user);
+
It can be modified using
#include <isl/aff.h>
+ __isl_give isl_multi_aff *isl_multi_aff_set_aff(
+ __isl_take isl_multi_aff *multi, int pos,
+ __isl_take isl_aff *aff);
__isl_give isl_multi_aff *isl_multi_aff_set_dim_name(
__isl_take isl_multi_aff *maff,
enum isl_dim_type type, unsigned pos, const char *s);
+ __isl_give isl_multi_aff *isl_multi_aff_set_tuple_id(
+ __isl_take isl_multi_aff *maff,
+ enum isl_dim_type type, __isl_take isl_id *id);
__isl_give isl_pw_multi_aff *isl_pw_multi_aff_set_tuple_id(
__isl_take isl_pw_multi_aff *pma,
enum isl_dim_type type, __isl_take isl_id *id);
+ __isl_give isl_multi_aff *isl_multi_aff_drop_dims(
+ __isl_take isl_multi_aff *maff,
+ enum isl_dim_type type, unsigned first, unsigned n);
+
To check whether two multiple affine expressions are
obviously equal to each other, use
__isl_give isl_pw_multi_aff *isl_pw_multi_aff_add(
__isl_take isl_pw_multi_aff *pma1,
__isl_take isl_pw_multi_aff *pma2);
+ __isl_give isl_union_pw_multi_aff *isl_union_pw_multi_aff_add(
+ __isl_take isl_union_pw_multi_aff *upma1,
+ __isl_take isl_union_pw_multi_aff *upma2);
__isl_give isl_pw_multi_aff *isl_pw_multi_aff_union_add(
__isl_take isl_pw_multi_aff *pma1,
__isl_take isl_pw_multi_aff *pma2);
__isl_give isl_multi_aff *isl_multi_aff_lift(
__isl_take isl_multi_aff *maff,
__isl_give isl_local_space **ls);
+ __isl_give isl_pw_multi_aff *isl_pw_multi_aff_coalesce(
+ __isl_take isl_pw_multi_aff *pma);
__isl_give isl_multi_aff *isl_multi_aff_gist_params(
__isl_take isl_multi_aff *maff,
__isl_take isl_set *context);
__isl_give isl_pw_multi_aff *isl_pw_multi_aff_gist(
__isl_take isl_pw_multi_aff *pma,
__isl_take isl_set *set);
+ __isl_give isl_set *isl_pw_multi_aff_domain(
+ __isl_take isl_pw_multi_aff *pma);
+ __isl_give isl_union_set *isl_union_pw_multi_aff_domain(
+ __isl_take isl_union_pw_multi_aff *upma);
+ __isl_give isl_multi_aff *isl_multi_aff_flat_range_product(
+ __isl_take isl_multi_aff *ma1,
+ __isl_take isl_multi_aff *ma2);
+ __isl_give isl_pw_multi_aff *
+ isl_pw_multi_aff_flat_range_product(
+ __isl_take isl_pw_multi_aff *pma1,
+ __isl_take isl_pw_multi_aff *pma2);
+ __isl_give isl_union_pw_multi_aff *
+ isl_union_pw_multi_aff_flat_range_product(
+ __isl_take isl_union_pw_multi_aff *upma1,
+ __isl_take isl_union_pw_multi_aff *upma2);
If the C<ls> argument of C<isl_multi_aff_lift> is not C<NULL>,
then it is assigned the local space that lies at the basis of
__isl_give isl_printer *isl_printer_print_pw_multi_aff(
__isl_take isl_printer *p,
__isl_keep isl_pw_multi_aff *pma);
+ __isl_give isl_printer *isl_printer_print_union_pw_multi_aff(
+ __isl_take isl_printer *p,
+ __isl_keep isl_union_pw_multi_aff *upma);
=head2 Points
__isl_give isl_union_pw_qpolynomial *isl_union_pw_qpolynomial_copy(
__isl_keep isl_union_pw_qpolynomial *upwqp);
- void isl_union_pw_qpolynomial_free(
+ void *isl_union_pw_qpolynomial_free(
__isl_take isl_union_pw_qpolynomial *upwqp);
=head3 Inspecting (Piecewise) Quasipolynomials
__isl_take isl_qpolynomial_fold *fold);
void *isl_pw_qpolynomial_fold_free(
__isl_take isl_pw_qpolynomial_fold *pwf);
- void isl_union_pw_qpolynomial_fold_free(
+ void *isl_union_pw_qpolynomial_fold_free(
__isl_take isl_union_pw_qpolynomial_fold *upwf);
=head3 Printing Piecewise Quasipolynomial Reductions
or C<may_no_source> may be C<NULL>, but a C<NULL> value for
any of the other arguments is treated as an error.
+=head3 Interaction with Dependence Analysis
+
+During the dependence analysis, we frequently need to perform
+the following operation. Given a relation between sink iterations
+and potential soure iterations from a particular source domain,
+what is the last potential source iteration corresponding to each
+sink iteration. It can sometimes be convenient to adjust
+the set of potential source iterations before or after each such operation.
+The prototypical example is fuzzy array dataflow analysis,
+where we need to analyze if, based on data-dependent constraints,
+the sink iteration can ever be executed without one or more of
+the corresponding potential source iterations being executed.
+If so, we can introduce extra parameters and select an unknown
+but fixed source iteration from the potential source iterations.
+To be able to perform such manipulations, C<isl> provides the following
+function.
+
+ #include <isl/flow.h>
+
+ typedef __isl_give isl_restriction *(*isl_access_restrict)(
+ __isl_keep isl_map *source_map,
+ __isl_keep isl_set *sink, void *source_user,
+ void *user);
+ __isl_give isl_access_info *isl_access_info_set_restrict(
+ __isl_take isl_access_info *acc,
+ isl_access_restrict fn, void *user);
+
+The function C<isl_access_info_set_restrict> should be called
+before calling C<isl_access_info_compute_flow> and registers a callback function
+that will be called any time C<isl> is about to compute the last
+potential source. The first argument is the (reverse) proto-dependence,
+mapping sink iterations to potential source iterations.
+The second argument represents the sink iterations for which
+we want to compute the last source iteration.
+The third argument is the token corresponding to the source
+and the final argument is the token passed to C<isl_access_info_set_restrict>.
+The callback is expected to return a restriction on either the input or
+the output of the operation computing the last potential source.
+If the input needs to be restricted then restrictions are needed
+for both the source and the sink iterations. The sink iterations
+and the potential source iterations will be intersected with these sets.
+If the output needs to be restricted then only a restriction on the source
+iterations is required.
+If any error occurs, the callback should return C<NULL>.
+An C<isl_restriction> object can be created and freed using the following
+functions.
+
+ #include <isl/flow.h>
+
+ __isl_give isl_restriction *isl_restriction_input(
+ __isl_take isl_set *source_restr,
+ __isl_take isl_set *sink_restr);
+ __isl_give isl_restriction *isl_restriction_output(
+ __isl_take isl_set *source_restr);
+ __isl_give isl_restriction *isl_restriction_none(
+ __isl_keep isl_map *source_map);
+ __isl_give isl_restriction *isl_restriction_empty(
+ __isl_keep isl_map *source_map);
+ void *isl_restriction_free(
+ __isl_take isl_restriction *restr);
+
+C<isl_restriction_none> and C<isl_restriction_empty> are special
+cases of C<isl_restriction_input>. C<isl_restriction_none>
+is essentially equivalent to
+
+ isl_restriction_input(isl_set_universe(
+ isl_space_range(isl_map_get_space(source_map))),
+ isl_set_universe(
+ isl_space_domain(isl_map_get_space(source_map))));
+
+whereas C<isl_restriction_empty> is essentially equivalent to
+
+ isl_restriction_input(isl_set_empty(
+ isl_space_range(isl_map_get_space(source_map))),
+ isl_set_universe(
+ isl_space_domain(isl_map_get_space(source_map))));
+
=head2 Scheduling
B<The functionality described in this section is fairly new
and may be subject to change.>
The following function can be used to compute a schedule
-for a union of domains. The generated schedule respects
-all C<validity> dependences. That is, all dependence distances
-over these dependences in the scheduled space are lexicographically
-positive. The generated schedule schedule also tries to minimize
-the dependence distances over C<proximity> dependences.
+for a union of domains.
+By default, the algorithm used to construct the schedule is similar
+to that of C<Pluto>.
+Alternatively, Feautrier's multi-dimensional scheduling algorithm can
+be selected.
+The generated schedule respects all C<validity> dependences.
+That is, all dependence distances over these dependences in the
+scheduled space are lexicographically positive.
+The default algorithm tries to minimize the dependence distances over
+C<proximity> dependences.
Moreover, it tries to obtain sequences (bands) of schedule dimensions
for groups of domains where the dependence distances have only
non-negative values.
-The algorithm used to construct the schedule is similar to that
-of C<Pluto>.
+When using Feautrier's algorithm, the C<proximity> dependence
+distances are only minimized during the extension to a
+full-dimensional schedule.
#include <isl/schedule.h>
__isl_give isl_schedule *isl_union_set_compute_schedule(
__isl_give isl_band_list *isl_schedule_get_band_forest(
__isl_keep isl_schedule *schedule);
+The individual bands can be visited in depth-first post-order
+using the following function.
+
+ #include <isl/schedule.h>
+ int isl_schedule_foreach_band(
+ __isl_keep isl_schedule *sched,
+ int (*fn)(__isl_keep isl_band *band, void *user),
+ void *user);
+
The list can be manipulated as explained in L<"Lists">.
The bands inside the list can be copied and freed using the following
functions.
int isl_band_member_is_zero_distance(
__isl_keep isl_band *band, int pos);
+ int isl_band_list_foreach_band(
+ __isl_keep isl_band_list *list,
+ int (*fn)(__isl_keep isl_band *band, void *user),
+ void *user);
+
Note that a scheduling dimension is considered to be ``zero
distance'' if it does not carry any proximity dependences
within its band.
That is, if the dependence distances of the proximity
dependences are all zero in that direction (for fixed
iterations of outer bands).
+Like C<isl_schedule_foreach_band>,
+the function C<isl_band_list_foreach_band> calls C<fn> on the bands
+in depth-first post-order.
+
+A band can be tiled using the following function.
+
+ #include <isl/band.h>
+ int isl_band_tile(__isl_keep isl_band *band,
+ __isl_take isl_vec *sizes);
+
+ int isl_options_set_tile_scale_tile_loops(isl_ctx *ctx,
+ int val);
+ int isl_options_get_tile_scale_tile_loops(isl_ctx *ctx);
+
+The C<isl_band_tile> function tiles the band using the given tile sizes
+inside its schedule.
+A new child band is created to represent the point loops and it is
+inserted between the modified band and its children.
+The C<tile_scale_tile_loops> option specifies whether the tile
+loops iterators should be scaled by the tile sizes.
A representation of the band can be printed using
=head3 Options
#include <isl/schedule.h>
+ int isl_options_set_schedule_max_coefficient(
+ isl_ctx *ctx, int val);
+ int isl_options_get_schedule_max_coefficient(
+ isl_ctx *ctx);
+ int isl_options_set_schedule_max_constant_term(
+ isl_ctx *ctx, int val);
+ int isl_options_get_schedule_max_constant_term(
+ isl_ctx *ctx);
+ int isl_options_set_schedule_maximize_band_depth(
+ isl_ctx *ctx, int val);
+ int isl_options_get_schedule_maximize_band_depth(
+ isl_ctx *ctx);
int isl_options_set_schedule_outer_zero_distance(
isl_ctx *ctx, int val);
int isl_options_get_schedule_outer_zero_distance(
isl_ctx *ctx);
+ int isl_options_set_schedule_split_scaled(
+ isl_ctx *ctx, int val);
+ int isl_options_get_schedule_split_scaled(
+ isl_ctx *ctx);
+ int isl_options_set_schedule_algorithm(
+ isl_ctx *ctx, int val);
+ int isl_options_get_schedule_algorithm(
+ isl_ctx *ctx);
+
=over
+=item * schedule_max_coefficient
+
+This option enforces that the coefficients for variable and parameter
+dimensions in the calculated schedule are not larger than the specified value.
+This option can significantly increase the speed of the scheduling calculation
+and may also prevent fusing of unrelated dimensions. A value of -1 means that
+this option does not introduce bounds on the variable or parameter
+coefficients.
+
+=item * schedule_max_constant_term
+
+This option enforces that the constant coefficients in the calculated schedule
+are not larger than the maximal constant term. This option can significantly
+increase the speed of the scheduling calculation and may also prevent fusing of
+unrelated dimensions. A value of -1 means that this option does not introduce
+bounds on the constant coefficients.
+
+=item * schedule_maximize_band_depth
+
+If this option is set, we do not split bands at the point
+where we detect splitting is necessary. Instead, we
+backtrack and split bands as early as possible. This
+reduces the number of splits and maximizes the width of
+the bands. Wider bands give more possibilities for tiling.
+
=item * schedule_outer_zero_distance
-It this option is set, then we try to construct schedules
+If this option is set, then we try to construct schedules
where the outermost scheduling dimension in each band
results in a zero dependence distance over the proximity
dependences.
+=item * schedule_split_scaled
+
+If this option is set, then we try to construct schedules in which the
+constant term is split off from the linear part if the linear parts of
+the scheduling rows for all nodes in the graphs have a common non-trivial
+divisor.
+The constant term is then placed in a separate band and the linear
+part is reduced.
+
+=item * schedule_algorithm
+
+Selects the scheduling algorithm to be used.
+Available scheduling algorithms are C<ISL_SCHEDULE_ALGORITHM_ISL>
+and C<ISL_SCHEDULE_ALGORITHM_FEAUTRIER>.
+
=back
=head2 Parametric Vertex Enumeration
projects / platform / upstream / isl.git / blobdiff