/* Interchange heuristics and transform for loop interchange on polyhedral representation. Copyright (C) 2009 Free Software Foundation, Inc. Contributed by Sebastian Pop and Harsha Jagasia . This file is part of GCC. GCC is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 3, or (at your option) any later version. GCC is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with GCC; see the file COPYING3. If not see . */ #include "config.h" #include "system.h" #include "coretypes.h" #include "tm.h" #include "ggc.h" #include "tree.h" #include "rtl.h" #include "output.h" #include "basic-block.h" #include "diagnostic.h" #include "tree-flow.h" #include "toplev.h" #include "tree-dump.h" #include "timevar.h" #include "cfgloop.h" #include "tree-chrec.h" #include "tree-data-ref.h" #include "tree-scalar-evolution.h" #include "tree-pass.h" #include "domwalk.h" #include "value-prof.h" #include "pointer-set.h" #include "gimple.h" #include "params.h" #ifdef HAVE_cloog #include "cloog/cloog.h" #include "ppl_c.h" #include "sese.h" #include "graphite-ppl.h" #include "graphite.h" #include "graphite-poly.h" /* Builds a linear expression, of dimension DIM, representing PDR's memory access: L = r_{n}*r_{n-1}*...*r_{1}*s_{0} + ... + r_{n}*s_{n-1} + s_{n}. For an array A[10][20] with two subscript locations s0 and s1, the linear memory access is 20 * s0 + s1: a stride of 1 in subscript s0 corresponds to a memory stride of 20. */ static ppl_Linear_Expression_t build_linearized_memory_access (poly_dr_p pdr) { ppl_Linear_Expression_t res; ppl_Linear_Expression_t le; ppl_dimension_type i; ppl_dimension_type first = pdr_subscript_dim (pdr, 0); ppl_dimension_type last = pdr_subscript_dim (pdr, PDR_NB_SUBSCRIPTS (pdr)); Value size, sub_size; graphite_dim_t dim = pdr_dim (pdr); ppl_new_Linear_Expression_with_dimension (&res, dim); value_init (size); value_set_si (size, 1); value_init (sub_size); value_set_si (sub_size, 1); for (i = last - 1; i >= first; i--) { ppl_set_coef_gmp (res, i, size); ppl_new_Linear_Expression_with_dimension (&le, dim); ppl_set_coef (le, i, 1); ppl_max_for_le (PDR_ACCESSES (pdr), le, sub_size); value_multiply (size, size, sub_size); ppl_delete_Linear_Expression (le); } value_clear (sub_size); value_clear (size); return res; } /* Set STRIDE to the stride of PDR in memory by advancing by one in loop DEPTH. */ static void memory_stride_in_loop (Value stride, graphite_dim_t depth, poly_dr_p pdr) { ppl_Linear_Expression_t le, lma; ppl_Constraint_t new_cstr; ppl_Pointset_Powerset_C_Polyhedron_t p1, p2; graphite_dim_t nb_subscripts = PDR_NB_SUBSCRIPTS (pdr); ppl_dimension_type i, *map; ppl_dimension_type dim = pdr_dim (pdr); ppl_dimension_type dim_i = pdr_iterator_dim (pdr, depth); ppl_dimension_type dim_k = dim; ppl_dimension_type dim_L1 = dim + nb_subscripts + 1; ppl_dimension_type dim_L2 = dim + nb_subscripts + 2; ppl_dimension_type new_dim = dim + nb_subscripts + 3; /* Add new dimensions to the polyhedron corresponding to k, s0', s1',..., L1, and L2. These new variables are at dimensions dim, dim + 1,... of the polyhedron P1 respectively. */ ppl_new_Pointset_Powerset_C_Polyhedron_from_Pointset_Powerset_C_Polyhedron (&p1, PDR_ACCESSES (pdr)); ppl_Pointset_Powerset_C_Polyhedron_add_space_dimensions_and_embed (p1, nb_subscripts + 3); lma = build_linearized_memory_access (pdr); ppl_set_coef (lma, dim_L1, -1); ppl_new_Constraint (&new_cstr, lma, PPL_CONSTRAINT_TYPE_EQUAL); ppl_Pointset_Powerset_C_Polyhedron_add_constraint (p1, new_cstr); /* Build P2. */ ppl_new_Pointset_Powerset_C_Polyhedron_from_Pointset_Powerset_C_Polyhedron (&p2, p1); map = ppl_new_id_map (new_dim); ppl_interchange (map, dim_L1, dim_L2); ppl_interchange (map, dim_i, dim_k); for (i = 0; i < PDR_NB_SUBSCRIPTS (pdr); i++) ppl_interchange (map, pdr_subscript_dim (pdr, i), dim + i + 1); ppl_Pointset_Powerset_C_Polyhedron_map_space_dimensions (p2, map, new_dim); free (map); /* Add constraint k = i + 1. */ ppl_new_Linear_Expression_with_dimension (&le, new_dim); ppl_set_coef (le, dim_i, 1); ppl_set_coef (le, dim_k, -1); ppl_set_inhomogeneous (le, 1); ppl_new_Constraint (&new_cstr, le, PPL_CONSTRAINT_TYPE_EQUAL); ppl_Pointset_Powerset_C_Polyhedron_add_constraint (p2, new_cstr); ppl_delete_Linear_Expression (le); ppl_delete_Constraint (new_cstr); /* P1 = P1 inter P2. */ ppl_Pointset_Powerset_C_Polyhedron_intersection_assign (p1, p2); ppl_delete_Pointset_Powerset_C_Polyhedron (p2); /* Maximise the expression L2 - L1. */ ppl_new_Linear_Expression_with_dimension (&le, new_dim); ppl_set_coef (le, dim_L2, 1); ppl_set_coef (le, dim_L1, -1); ppl_max_for_le (p1, le, stride); ppl_delete_Linear_Expression (le); } /* Returns true when it is profitable to interchange loop at DEPTH1 and loop at DEPTH2 with DEPTH1 < DEPTH2 for PBB. Example: | int a[100][100]; | | int | foo (int N) | { | int j; | int i; | | for (i = 0; i < N; i++) | for (j = 0; j < N; j++) | a[j][2 * i] += 1; | | return a[N][12]; | } The data access A[j][i] is described like this: | i j N a s0 s1 1 | 0 0 0 1 0 0 -5 = 0 | 0 -1 0 0 1 0 0 = 0 |-2 0 0 0 0 1 0 = 0 | 0 0 0 0 1 0 0 >= 0 | 0 0 0 0 0 1 0 >= 0 | 0 0 0 0 -1 0 100 >= 0 | 0 0 0 0 0 -1 100 >= 0 The linearized memory access L to A[100][100] is: | i j N a s0 s1 1 | 0 0 0 0 100 1 0 Next, to measure the impact of iterating once in loop "i", we build a maximization problem: first, we add to DR accesses the dimensions k, s2, s3, L1 = 100 * s0 + s1, L2, and D1: polyhedron P1. | i j N a s0 s1 k s2 s3 L1 L2 D1 1 | 0 0 0 1 0 0 0 0 0 0 0 0 -5 = 0 alias = 5 | 0 -1 0 0 1 0 0 0 0 0 0 0 0 = 0 s0 = j |-2 0 0 0 0 1 0 0 0 0 0 0 0 = 0 s1 = 2 * i | 0 0 0 0 1 0 0 0 0 0 0 0 0 >= 0 | 0 0 0 0 0 1 0 0 0 0 0 0 0 >= 0 | 0 0 0 0 -1 0 0 0 0 0 0 0 100 >= 0 | 0 0 0 0 0 -1 0 0 0 0 0 0 100 >= 0 | 0 0 0 0 100 1 0 0 0 -1 0 0 0 = 0 L1 = 100 * s0 + s1 Then, we generate the polyhedron P2 by interchanging the dimensions (s0, s2), (s1, s3), (L1, L2), (i0, i) | i j N a s0 s1 k s2 s3 L1 L2 D1 1 | 0 0 0 1 0 0 0 0 0 0 0 0 -5 = 0 alias = 5 | 0 -1 0 0 0 0 0 1 0 0 0 0 0 = 0 s2 = j | 0 0 0 0 0 0 -2 0 1 0 0 0 0 = 0 s3 = 2 * k | 0 0 0 0 0 0 0 1 0 0 0 0 0 >= 0 | 0 0 0 0 0 0 0 0 1 0 0 0 0 >= 0 | 0 0 0 0 0 0 0 -1 0 0 0 0 100 >= 0 | 0 0 0 0 0 0 0 0 -1 0 0 0 100 >= 0 | 0 0 0 0 0 0 0 100 1 0 -1 0 0 = 0 L2 = 100 * s2 + s3 then we add to P2 the equality k = i + 1: |-1 0 0 0 0 0 1 0 0 0 0 0 -1 = 0 k = i + 1 and finally we maximize the expression "D1 = max (P1 inter P2, L2 - L1)". For determining the impact of one iteration on loop "j", we interchange (k, j), we add "k = j + 1", and we compute D2 the maximal value of the difference. Finally, the profitability test is D1 < D2: if in the outer loop the strides are smaller than in the inner loop, then it is profitable to interchange the loops at DEPTH1 and DEPTH2. */ static bool pbb_interchange_profitable_p (graphite_dim_t depth1, graphite_dim_t depth2, poly_bb_p pbb) { int i; poly_dr_p pdr; Value d1, d2, s; bool res; gcc_assert (depth1 < depth2); value_init (d1); value_set_si (d1, 0); value_init (d2); value_set_si (d2, 0); value_init (s); for (i = 0; VEC_iterate (poly_dr_p, PBB_DRS (pbb), i, pdr); i++) { memory_stride_in_loop (s, depth1, pdr); value_addto (d1, d1, s); memory_stride_in_loop (s, depth2, pdr); value_addto (d2, d2, s); } res = value_lt (d1, d2); value_clear (d1); value_clear (d2); value_clear (s); return res; } /* Interchanges the loops at DEPTH1 and DEPTH2 of the original scattering and assigns the resulting polyhedron to the transformed scattering. */ static void pbb_interchange_loop_depths (graphite_dim_t depth1, graphite_dim_t depth2, poly_bb_p pbb) { ppl_dimension_type i, dim; ppl_dimension_type *map; ppl_Polyhedron_t poly = PBB_TRANSFORMED_SCATTERING (pbb); ppl_dimension_type dim1 = psct_iterator_dim (pbb, depth1); ppl_dimension_type dim2 = psct_iterator_dim (pbb, depth2); ppl_Polyhedron_space_dimension (poly, &dim); map = (ppl_dimension_type *) XNEWVEC (ppl_dimension_type, dim); for (i = 0; i < dim; i++) map[i] = i; map[dim1] = dim2; map[dim2] = dim1; ppl_Polyhedron_map_space_dimensions (poly, map, dim); free (map); } /* Interchanges all the loop depths that are considered profitable for PBB. */ static bool pbb_do_interchange (poly_bb_p pbb, scop_p scop) { graphite_dim_t i, j; bool transform_done = false; for (i = 0; i < pbb_dim_iter_domain (pbb); i++) for (j = i + 1; j < pbb_dim_iter_domain (pbb); j++) if (pbb_interchange_profitable_p (i, j, pbb)) { pbb_interchange_loop_depths (i, j, pbb); if (graphite_legal_transform (scop)) { transform_done = true; if (dump_file && (dump_flags & TDF_DETAILS)) fprintf (dump_file, "PBB %d: loops at depths %d and %d will be interchanged.\n", GBB_BB (PBB_BLACK_BOX (pbb))->index, (int) i, (int) j); } else /* Undo the transform. */ pbb_interchange_loop_depths (j, i, pbb); } return transform_done; } /* Interchanges all the loop depths that are considered profitable for SCOP. */ bool scop_do_interchange (scop_p scop) { int i; poly_bb_p pbb; bool transform_done = false; store_scattering (scop); for (i = 0; VEC_iterate (poly_bb_p, SCOP_BBS (scop), i, pbb); i++) transform_done |= pbb_do_interchange (pbb, scop); if (!transform_done) return false; if (!graphite_legal_transform (scop)) { restore_scattering (scop); return false; } return transform_done; } #endif