Add two affine expressions that live in the same local space. */

Add constraints to graph->lp that bound the dependence distance for the given * dependence from node i to node j. * If s = 1, we add the constraint

Add constraints to graph->lp that bound the dependence distance for the given * dependence from a node i to itself. * If s = 1, we add the constrain

For each known div d = floor(f/m), add the constraints * * f - m d >= 0 * -(f-(m-1)) + m d >= 0 * * Do not finalize the result. */

Add encodings of the "n" loop AST generation options "type" to "options". * If "isolate" is set, then these options refer to the isolated part. * *

For a div d = floor(f/m), add the constraint * * -(f-(m-1)) + m d >= 0 */

Add a new node to the graph representing the given instance set. * "nvar" is the (possibly compressed) number of variables and * may be smaller than

Add a constraint to graph->lp that equates the value at position * "sum_pos" to the sum of the parameter coefficients of all nodes. */

Add "prefix" to the list of prefixes and return the updated * number of prefixes. */

Add terms to "expr" for each variable in "aff" with a coefficient * with sign equal to "sign". * The result is simplified in terms of data->build->d

For a div d = floor(f/m), add the constraint * * f - m d >= 0 */

Add the valid constraints of the basic map represented by "info" * to "bmap". "len" is the size of the constraints. * If only one of the pair of in

Add a constraint to graph->lp that equates the value at position * "sum_pos" to the sum of the variable coefficients of all nodes. */

Wrap "ineq" (or its opposite if "negate" is set) around "bound" * to include "set" and add the result in position "w" of "wraps". * "len" is the tot

Is "aff" a rational expression, i.e., does it have a denominator * different from one? */

Is any domain element of "umap" scheduled after any of * the corresponding image elements by the tree rooted at * the set node "node"? * * This is

For a given mapping between iterations of must source k and iterations * of the sink, compute all iterations of may source j preceding * the sink at

Does "test" hold for all pairs of basic maps in "map1" and "map2"? */

Compute all iterations of may source j that precedes the sink at the given * level for sink iterations in set_C. */

Allocate an array of groups of variables, one for each variable * in "context", initialized to zero. */

Ensure that "p" has a note identified by "id". * If there is no such note yet, then it is created by "note_create" and * scheduled do be freed by "n

Append an "extra" number of zeros to the range of "umap" and * return the result. */

Construct an isl_ast_expr from "list" within "build". * If "state" is isl_state_single, then "list" contains a single entry and * an isl_ast_expr is

Create an expression of type "type" with as arguments "arg0" followed * by "arguments". */

Combine the grafts in the list into a single graft. * * The guard is initialized to the shared guard of the list elements (if any), * provided it d

Insert an if node around "node" testing the condition encoded * in guard "guard". * * If the user does not want any disjunctions in the if conditio

If "node" contains an isl_ast_op_min, isl_ast_op_max or isl_ast_op_fdiv_q * then set the corresponding bit in "macros". */

Return the constraint * * i_"depth" = aff + offset */

Bound the given variable of "bmap" from below (or above is "upper" * is set) to "value". */

Given a basic map with one output dimension, compute the minimum or * maximum of that dimension as an isl_pw_aff. * * Compute the optimum as a lexi

Base case of isl_tab_basic_map_partial_lexopt, after removing * some obvious symmetries. * * We make sure the divs in the domain are properly order

Given two basic sets bset1 and bset2, compute the maximal difference * between the values of dimension pos in bset1 and those in bset2 * for any com

Return the optimal (rational) value of "obj" over "bset", assuming * that "obj" and "bset" have aligned parameters and divs. * If "max" is set, then

Base case of applying bernstein expansion. * * We compute the chamber decomposition of the parametric polytope "bset" * and then perform bernstein

Apply bernstein expansion recursively by working in on len[i] * set variables at a time, with i ranging from n_group - 1 to 0. */

Given a main tableau where more than one row requires a split, * determine and return the "best" row to split on. * * If any of the rows requiring

Call "fn" on each pair of isl_pw_multi_affs in "upma1" and "upma2". * The isl_pw_multi_aff from upma1 is stored in data->pma (where data is * passed

Given a map, compute the smallest superset of this map that is of the form * * { i -> j : L <= j - i <= U and exists a_p: j_p - i_p = M_p a_p } *

Given a map, compute the smallest superset of this map that is of the form * * { i -> j : L <= j - i <= U and exists a_p: j_p - i_p = M_p a_p } *

Generate an AST that visits the elements in the domain of "executed" * in the relative order specified by the node "node" and its descendants. * *

Can "tab" be intersected with "bset" without resulting in * a lower-dimensional set. * "bset" itself is assumed to be full-dimensional. */

Given a basic set i with a constraint k that is adjacent to * basic set j, check if we can wrap * both the facet corresponding to k (if "wrap_facet"

Construct a schedule row for each node such that as many validity dependences * as possible are carried and then continue with the next band. * If "

Construct a schedule row for each node such that as many validity dependences * as possible are carried and then continue with the next band. * Do s

Construct a schedule row for each node such that as many validity dependences * as possible are carried and then continue with the next band. * Do s

Check if we can coalesce basic map "j" into basic map "i" after copying * those extra integer divisions in "i" that can be simplified away * using t

The two basic maps lie on adjacent hyperplanes. In particular, * basic map "i" has an equality that lies parallel to basic map "j". * Check if we c

Check whether the overapproximation of the power of "map" is exactly * the power of "map", possibly after projecting out the power (if "project" * i

Check whether app is the transitive closure of map. * In particular, check that app is acyclic and, if so, * check that * * app \subset (map \cup

Disjunct "j" lies on a hyperplane that is adjacent to disjunct "i". * In particular, disjunct "i" has an inequality constraint that is adjacent * to

Check that "pa" consists of a single expression. */

Return the filter of child "i" of "node". */

Clear all cached information in "map", either because it is about * to be modified or because it is being freed. * Always return the same pointer th

Free all the allocated memory in an array * of "n" isl_coalesce_info elements. */

Given a list of "n" if nodes, clear those starting at "first" * and return "first" (i.e., the updated size of the array). */

Free the memory associated to "data". */

Construct the identifier "cluster_i". */

Free all memory allocated for "c". */

Compare (known) divs. * Return non-zero if at least one of the two divs is unknown. * In particular, if both divs are unknown, we respect their * c

Check if the union of the given pair of basic maps * can be represented by a single basic map. * If so, replace the pair by the single basic map and

Pairwise coalesce the basic maps in the range [start1, end1[ of "info" * with those in the range [start2, end2[, skipping basic maps * that have bee

Check if basic map "j" is a subset of basic map "i" after * exploiting the extra equalities of "j" to simplify the divs of "i". * If so, remove basi

Construct a representation of the graph on the right of Figure 1 * in "Computing the Transitive Closure of a Union of * Affine Integer Tuple Relatio

Construct a representation of the power of the graph * on the right of Figure 1 in "Computing the Transitive Closure of * a Union of Affine Integer

For each (conditional) validity edge in "graph", * add the corresponding dependence relation using "add" * to a collection of dependence relations a

Return a map that is a union of the basic maps in "map", except i, * composed to left and right with qc based on the entries of "left" * and "right"

Compute a bound on the polynomial defined over the parametric polytope * using either range propagation or bernstein expansion and * store the resul

Factor bset, call fn on each of the factors and return the product. * * If no factors can be found, simply call fn on the input. * Otherwise, const

Construct an LP problem for finding schedule coefficients * such that the schedule carries as many of the "n_edge" groups of * dependences as possib

Compute the transitive closure of "map" incrementally by * computing * * map_i^+ \cup qc^+ * * or * * map_i^+ \cup ((id \cup map_i^) \circ qc^+

Compute the maximum of the current ("level") basis row over "tab" * and store the result in position "level" of "max". * * This function assumes th

Compute the minimum of the current ("level") basis row over "tab" * and store the result in position "level" of "min". * * This function assumes th

Split up the domain at the current depth into disjoint * basic sets for which code should be generated separately * for the given separation class d

Compute a schedule for a connected dependence graph and return * the updated schedule node. * * If Feautrier's algorithm is selected, we first recu

Compute a schedule for the given instance set, validity constraints, * proximity constraints and context and return a corresponding union map * repr

Is the inequality ineq (obviously) redundant with respect * to the constraints in "ci"? * * Look for an inequality in "ci" with the same coefficien

Return a string representation of the operator used when * printing a constraint where the LHS is greater than or equal to the LHS * (sign > 0) or s

Construct a constraint imposing that the value of the first dimension is * greater than or equal to that of the second. */

Check if the constraint "c" is a lower bound on dimension "pos", * an upper bound, or independent of dimension "pos". */

Construct an AST in case the schedule is specified by a union map. * * We read the context and the options from "s" and construct the AST. */

Make first row entries in column col of bset1 identical to * those of bset2, using the fact that entry bset1->eq[row][col]=a * is non-zero. Initial

Given a union of basic maps R = \cup_i R_i \subseteq D \times D * and a dimension specification (Z^{n+1} -> Z^{n+1}), * construct a map that is an o