#include "tree.h"
#include "target.h"
#include "basic-block.h"
-#include "diagnostic.h"
#include "tree-pretty-print.h"
#include "gimple-pretty-print.h"
#include "tree-flow.h"
#include "expr.h"
#include "recog.h"
#include "optabs.h"
+#include "diagnostic-core.h"
#include "toplev.h"
#include "tree-vectorizer.h"
#include "langhooks.h"
}
+/* Get cost by calling cost target builtin. */
+
+static inline
+int vect_get_stmt_cost (enum vect_cost_for_stmt type_of_cost)
+{
+ tree dummy_type = NULL;
+ int dummy = 0;
+
+ return targetm.vectorize.builtin_vectorization_cost (type_of_cost,
+ dummy_type, dummy);
+}
+
int
cost_for_stmt (gimple stmt)
{
switch (STMT_VINFO_TYPE (stmt_info))
{
case load_vec_info_type:
- return TARG_SCALAR_LOAD_COST;
+ return vect_get_stmt_cost (scalar_load);
case store_vec_info_type:
- return TARG_SCALAR_STORE_COST;
+ return vect_get_stmt_cost (scalar_store);
case op_vec_info_type:
case condition_vec_info_type:
case assignment_vec_info_type:
case type_demotion_vec_info_type:
case type_conversion_vec_info_type:
case call_vec_info_type:
- return TARG_SCALAR_STMT_COST;
+ return vect_get_stmt_cost (scalar_stmt);
case undef_vec_info_type:
default:
gcc_unreachable ();
if (PURE_SLP_STMT (stmt_info))
return;
- inside_cost = ncopies * TARG_VEC_STMT_COST;
+ inside_cost = ncopies * vect_get_stmt_cost (vector_stmt);
/* FORNOW: Assuming maximum 2 args per stmts. */
for (i = 0; i < 2; i++)
{
if (dt[i] == vect_constant_def || dt[i] == vect_external_def)
- outside_cost += TARG_SCALAR_TO_VEC_COST;
+ outside_cost += vect_get_stmt_cost (vector_stmt);
}
if (vect_print_dump_info (REPORT_COST))
enum vect_def_type dt, slp_tree slp_node)
{
int group_size;
- int inside_cost = 0, outside_cost = 0;
+ unsigned int inside_cost = 0, outside_cost = 0;
+ struct data_reference *first_dr;
+ gimple first_stmt;
/* The SLP costs were already calculated during SLP tree build. */
if (PURE_SLP_STMT (stmt_info))
return;
if (dt == vect_constant_def || dt == vect_external_def)
- outside_cost = TARG_SCALAR_TO_VEC_COST;
+ outside_cost = vect_get_stmt_cost (scalar_to_vec);
/* Strided access? */
- if (DR_GROUP_FIRST_DR (stmt_info) && !slp_node)
- group_size = vect_cost_strided_group_size (stmt_info);
+ if (DR_GROUP_FIRST_DR (stmt_info))
+ {
+ if (slp_node)
+ {
+ first_stmt = VEC_index (gimple, SLP_TREE_SCALAR_STMTS (slp_node), 0);
+ group_size = 1;
+ }
+ else
+ {
+ first_stmt = DR_GROUP_FIRST_DR (stmt_info);
+ group_size = vect_cost_strided_group_size (stmt_info);
+ }
+
+ first_dr = STMT_VINFO_DATA_REF (vinfo_for_stmt (first_stmt));
+ }
/* Not a strided access. */
else
- group_size = 1;
+ {
+ group_size = 1;
+ first_dr = STMT_VINFO_DATA_REF (stmt_info);
+ }
/* Is this an access in a group of stores, which provide strided access?
If so, add in the cost of the permutes. */
{
/* Uses a high and low interleave operation for each needed permute. */
inside_cost = ncopies * exact_log2(group_size) * group_size
- * TARG_VEC_STMT_COST;
+ * vect_get_stmt_cost (vector_stmt);
if (vect_print_dump_info (REPORT_COST))
fprintf (vect_dump, "vect_model_store_cost: strided group_size = %d .",
}
/* Costs of the stores. */
- inside_cost += ncopies * TARG_VEC_STORE_COST;
+ vect_get_store_cost (first_dr, ncopies, &inside_cost);
if (vect_print_dump_info (REPORT_COST))
fprintf (vect_dump, "vect_model_store_cost: inside_cost = %d, "
}
+/* Calculate cost of DR's memory access. */
+void
+vect_get_store_cost (struct data_reference *dr, int ncopies,
+ unsigned int *inside_cost)
+{
+ int alignment_support_scheme = vect_supportable_dr_alignment (dr, false);
+
+ switch (alignment_support_scheme)
+ {
+ case dr_aligned:
+ {
+ *inside_cost += ncopies * vect_get_stmt_cost (vector_store);
+
+ if (vect_print_dump_info (REPORT_COST))
+ fprintf (vect_dump, "vect_model_store_cost: aligned.");
+
+ break;
+ }
+
+ case dr_unaligned_supported:
+ {
+ gimple stmt = DR_STMT (dr);
+ stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
+ tree vectype = STMT_VINFO_VECTYPE (stmt_info);
+
+ /* Here, we assign an additional cost for the unaligned store. */
+ *inside_cost += ncopies
+ * targetm.vectorize.builtin_vectorization_cost (unaligned_store,
+ vectype, DR_MISALIGNMENT (dr));
+
+ if (vect_print_dump_info (REPORT_COST))
+ fprintf (vect_dump, "vect_model_store_cost: unaligned supported by "
+ "hardware.");
+
+ break;
+ }
+
+ default:
+ gcc_unreachable ();
+ }
+}
+
+
/* Function vect_model_load_cost
Models cost for loads. In the case of strided accesses, the last access
{
int group_size;
- int alignment_support_cheme;
gimple first_stmt;
struct data_reference *dr = STMT_VINFO_DATA_REF (stmt_info), *first_dr;
- int inside_cost = 0, outside_cost = 0;
+ unsigned int inside_cost = 0, outside_cost = 0;
/* The SLP costs were already calculated during SLP tree build. */
if (PURE_SLP_STMT (stmt_info))
first_dr = dr;
}
- alignment_support_cheme = vect_supportable_dr_alignment (first_dr);
-
/* Is this an access in a group of loads providing strided access?
If so, add in the cost of the permutes. */
if (group_size > 1)
{
/* Uses an even and odd extract operations for each needed permute. */
inside_cost = ncopies * exact_log2(group_size) * group_size
- * TARG_VEC_STMT_COST;
+ * vect_get_stmt_cost (vector_stmt);
if (vect_print_dump_info (REPORT_COST))
fprintf (vect_dump, "vect_model_load_cost: strided group_size = %d .",
group_size);
-
}
/* The loads themselves. */
- switch (alignment_support_cheme)
+ vect_get_load_cost (first_dr, ncopies,
+ ((!DR_GROUP_FIRST_DR (stmt_info)) || group_size > 1 || slp_node),
+ &inside_cost, &outside_cost);
+
+ if (vect_print_dump_info (REPORT_COST))
+ fprintf (vect_dump, "vect_model_load_cost: inside_cost = %d, "
+ "outside_cost = %d .", inside_cost, outside_cost);
+
+ /* Set the costs either in STMT_INFO or SLP_NODE (if exists). */
+ stmt_vinfo_set_inside_of_loop_cost (stmt_info, slp_node, inside_cost);
+ stmt_vinfo_set_outside_of_loop_cost (stmt_info, slp_node, outside_cost);
+}
+
+
+/* Calculate cost of DR's memory access. */
+void
+vect_get_load_cost (struct data_reference *dr, int ncopies,
+ bool add_realign_cost, unsigned int *inside_cost,
+ unsigned int *outside_cost)
+{
+ int alignment_support_scheme = vect_supportable_dr_alignment (dr, false);
+
+ switch (alignment_support_scheme)
{
case dr_aligned:
{
- inside_cost += ncopies * TARG_VEC_LOAD_COST;
+ *inside_cost += ncopies * vect_get_stmt_cost (vector_load);
if (vect_print_dump_info (REPORT_COST))
fprintf (vect_dump, "vect_model_load_cost: aligned.");
}
case dr_unaligned_supported:
{
- /* Here, we assign an additional cost for the unaligned load. */
- inside_cost += ncopies * TARG_VEC_UNALIGNED_LOAD_COST;
+ gimple stmt = DR_STMT (dr);
+ stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
+ tree vectype = STMT_VINFO_VECTYPE (stmt_info);
+ /* Here, we assign an additional cost for the unaligned load. */
+ *inside_cost += ncopies
+ * targetm.vectorize.builtin_vectorization_cost (unaligned_load,
+ vectype, DR_MISALIGNMENT (dr));
if (vect_print_dump_info (REPORT_COST))
fprintf (vect_dump, "vect_model_load_cost: unaligned supported by "
"hardware.");
}
case dr_explicit_realign:
{
- inside_cost += ncopies * (2*TARG_VEC_LOAD_COST + TARG_VEC_STMT_COST);
+ *inside_cost += ncopies * (2 * vect_get_stmt_cost (vector_load)
+ + vect_get_stmt_cost (vector_stmt));
/* FIXME: If the misalignment remains fixed across the iterations of
the containing loop, the following cost should be added to the
outside costs. */
if (targetm.vectorize.builtin_mask_for_load)
- inside_cost += TARG_VEC_STMT_COST;
+ *inside_cost += vect_get_stmt_cost (vector_stmt);
break;
}
access in the group. Inside the loop, there is a load op
and a realignment op. */
- if ((!DR_GROUP_FIRST_DR (stmt_info)) || group_size > 1 || slp_node)
+ if (add_realign_cost)
{
- outside_cost = 2*TARG_VEC_STMT_COST;
+ *outside_cost = 2 * vect_get_stmt_cost (vector_stmt);
if (targetm.vectorize.builtin_mask_for_load)
- outside_cost += TARG_VEC_STMT_COST;
+ *outside_cost += vect_get_stmt_cost (vector_stmt);
}
- inside_cost += ncopies * (TARG_VEC_LOAD_COST + TARG_VEC_STMT_COST);
-
+ *inside_cost += ncopies * (vect_get_stmt_cost (vector_load)
+ + vect_get_stmt_cost (vector_stmt));
break;
}
default:
gcc_unreachable ();
}
-
- if (vect_print_dump_info (REPORT_COST))
- fprintf (vect_dump, "vect_model_load_cost: inside_cost = %d, "
- "outside_cost = %d .", inside_cost, outside_cost);
-
- /* Set the costs either in STMT_INFO or SLP_NODE (if exists). */
- stmt_vinfo_set_inside_of_loop_cost (stmt_info, slp_node, inside_cost);
- stmt_vinfo_set_outside_of_loop_cost (stmt_info, slp_node, outside_cost);
}
tree def;
gimple def_stmt;
enum vect_def_type dt[2] = {vect_unknown_def_type, vect_unknown_def_type};
- int nunits = TYPE_VECTOR_SUBPARTS (vectype);
+ unsigned int nunits = TYPE_VECTOR_SUBPARTS (vectype);
int ncopies;
int i, j;
VEC(tree,heap) *vec_oprnds = NULL;
bb_vec_info bb_vinfo = STMT_VINFO_BB_VINFO (stmt_info);
gimple new_stmt = NULL;
stmt_vec_info prev_stmt_info = NULL;
+ enum tree_code code;
+ tree vectype_in;
/* Multiple types in SLP are handled by creating the appropriate number of
vectorized stmts for each SLP node. Hence, NCOPIES is always 1 in
if (TREE_CODE (scalar_dest) != SSA_NAME)
return false;
+ code = gimple_assign_rhs_code (stmt);
if (gimple_assign_single_p (stmt)
- || gimple_assign_rhs_code (stmt) == PAREN_EXPR)
+ || code == PAREN_EXPR
+ || CONVERT_EXPR_CODE_P (code))
op = gimple_assign_rhs1 (stmt);
else
return false;
- if (!vect_is_simple_use (op, loop_vinfo, bb_vinfo, &def_stmt, &def, &dt[0]))
+ if (!vect_is_simple_use_1 (op, loop_vinfo, bb_vinfo,
+ &def_stmt, &def, &dt[0], &vectype_in))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "use not simple.");
return false;
}
+ /* We can handle NOP_EXPR conversions that do not change the number
+ of elements or the vector size. */
+ if (CONVERT_EXPR_CODE_P (code)
+ && (!vectype_in
+ || TYPE_VECTOR_SUBPARTS (vectype_in) != nunits
+ || (GET_MODE_SIZE (TYPE_MODE (vectype))
+ != GET_MODE_SIZE (TYPE_MODE (vectype_in)))))
+ return false;
+
if (!vec_stmt) /* transformation not required. */
{
STMT_VINFO_TYPE (stmt_info) = assignment_vec_info_type;
/* Arguments are ready. create the new vector stmt. */
for (i = 0; VEC_iterate (tree, vec_oprnds, i, vop); i++)
{
+ if (CONVERT_EXPR_CODE_P (code))
+ vop = build1 (VIEW_CONVERT_EXPR, vectype, vop);
new_stmt = gimple_build_assign (vec_dest, vop);
new_temp = make_ssa_name (vec_dest, new_stmt);
gimple_assign_set_lhs (new_stmt, new_temp);
{
optab = optab_for_tree_code (code, vectype, optab_scalar);
if (optab
- && (optab_handler (optab, TYPE_MODE (vectype))->insn_code
- != CODE_FOR_nothing))
+ && optab_handler (optab, TYPE_MODE (vectype)) != CODE_FOR_nothing)
{
scalar_shift_arg = true;
if (vect_print_dump_info (REPORT_DETAILS))
{
optab = optab_for_tree_code (code, vectype, optab_vector);
if (optab
- && (optab_handler (optab, TYPE_MODE (vectype))->insn_code
+ && (optab_handler (optab, TYPE_MODE (vectype))
!= CODE_FOR_nothing))
{
if (vect_print_dump_info (REPORT_DETAILS))
return false;
}
vec_mode = TYPE_MODE (vectype);
- icode = (int) optab_handler (optab, vec_mode)->insn_code;
+ icode = (int) optab_handler (optab, vec_mode);
if (icode == CODE_FOR_nothing)
{
if (vect_print_dump_info (REPORT_DETAILS))
&& TREE_CODE (scalar_dest) != INDIRECT_REF
&& TREE_CODE (scalar_dest) != COMPONENT_REF
&& TREE_CODE (scalar_dest) != IMAGPART_EXPR
- && TREE_CODE (scalar_dest) != REALPART_EXPR)
+ && TREE_CODE (scalar_dest) != REALPART_EXPR
+ && TREE_CODE (scalar_dest) != MEM_REF)
return false;
gcc_assert (gimple_assign_single_p (stmt));
vec_mode = TYPE_MODE (vectype);
/* FORNOW. In some cases can vectorize even if data-type not supported
(e.g. - array initialization with 0). */
- if (optab_handler (mov_optab, (int)vec_mode)->insn_code == CODE_FOR_nothing)
+ if (optab_handler (mov_optab, vec_mode) == CODE_FOR_nothing)
return false;
if (!STMT_VINFO_DATA_REF (stmt_info))
dr_chain = VEC_alloc (tree, heap, group_size);
oprnds = VEC_alloc (tree, heap, group_size);
- alignment_support_scheme = vect_supportable_dr_alignment (first_dr);
+ alignment_support_scheme = vect_supportable_dr_alignment (first_dr, false);
gcc_assert (alignment_support_scheme);
/* In case the vectorization factor (VF) is bigger than the number
vec_oprnd = VEC_index (tree, result_chain, i);
if (aligned_access_p (first_dr))
- data_ref = build_fold_indirect_ref (dataref_ptr);
+ data_ref
+ = build2 (MEM_REF, TREE_TYPE (vec_oprnd), dataref_ptr,
+ build_int_cst (reference_alias_ptr_type
+ (DR_REF (first_dr)), 0));
else
{
int mis = DR_MISALIGNMENT (first_dr);
data_ref = build2 (MISALIGNED_INDIRECT_REF, vectype, dataref_ptr, tmis);
}
- /* If accesses through a pointer to vectype do not alias the original
- memory reference we have a problem. This should never happen. */
- gcc_assert (alias_sets_conflict_p (get_alias_set (data_ref),
- get_alias_set (gimple_assign_lhs (stmt))));
-
/* Arguments are ready. Create the new vector stmt. */
new_stmt = gimple_build_assign (data_ref, vec_oprnd);
vect_finish_stmt_generation (stmt, new_stmt, gsi);
struct data_reference *dr = STMT_VINFO_DATA_REF (stmt_info), *first_dr;
tree vectype = STMT_VINFO_VECTYPE (stmt_info);
tree new_temp;
- int mode;
+ enum machine_mode mode;
gimple new_stmt = NULL;
tree dummy;
enum dr_alignment_support alignment_support_scheme;
&& code != INDIRECT_REF
&& code != COMPONENT_REF
&& code != IMAGPART_EXPR
- && code != REALPART_EXPR)
+ && code != REALPART_EXPR
+ && code != MEM_REF)
return false;
if (!STMT_VINFO_DATA_REF (stmt_info))
return false;
scalar_type = TREE_TYPE (DR_REF (dr));
- mode = (int) TYPE_MODE (vectype);
+ mode = TYPE_MODE (vectype);
/* FORNOW. In some cases can vectorize even if data-type not supported
(e.g. - data copies). */
- if (optab_handler (mov_optab, mode)->insn_code == CODE_FOR_nothing)
+ if (optab_handler (mov_optab, mode) == CODE_FOR_nothing)
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "Aligned load, but unsupported type.");
group_size = vec_num = 1;
}
- alignment_support_scheme = vect_supportable_dr_alignment (first_dr);
+ alignment_support_scheme = vect_supportable_dr_alignment (first_dr, false);
gcc_assert (alignment_support_scheme);
/* In case the vectorization factor (VF) is bigger than the number
{
case dr_aligned:
gcc_assert (aligned_access_p (first_dr));
- data_ref = build_fold_indirect_ref (dataref_ptr);
+ data_ref
+ = build2 (MEM_REF, vectype, dataref_ptr,
+ build_int_cst (reference_alias_ptr_type
+ (DR_REF (first_dr)), 0));
break;
case dr_unaligned_supported:
{
dr_explicit_realign,
dataref_ptr, NULL);
- data_ref = build1 (ALIGN_INDIRECT_REF, vectype, dataref_ptr);
+ new_stmt = gimple_build_assign_with_ops
+ (BIT_AND_EXPR, NULL_TREE, dataref_ptr,
+ build_int_cst
+ (TREE_TYPE (dataref_ptr),
+ -(HOST_WIDE_INT)TYPE_ALIGN_UNIT (vectype)));
+ ptr = make_ssa_name (SSA_NAME_VAR (dataref_ptr), new_stmt);
+ gimple_assign_set_lhs (new_stmt, ptr);
+ vect_finish_stmt_generation (stmt, new_stmt, gsi);
+ data_ref
+ = build2 (MEM_REF, vectype, ptr,
+ build_int_cst (reference_alias_ptr_type
+ (DR_REF (first_dr)), 0));
vec_dest = vect_create_destination_var (scalar_dest, vectype);
new_stmt = gimple_build_assign (vec_dest, data_ref);
new_temp = make_ssa_name (vec_dest, new_stmt);
bump = size_binop (MULT_EXPR, vs_minus_1,
TYPE_SIZE_UNIT (scalar_type));
ptr = bump_vector_ptr (dataref_ptr, NULL, gsi, stmt, bump);
- data_ref = build1 (ALIGN_INDIRECT_REF, vectype, ptr);
+ new_stmt = gimple_build_assign_with_ops
+ (BIT_AND_EXPR, NULL_TREE, ptr,
+ build_int_cst
+ (TREE_TYPE (ptr),
+ -(HOST_WIDE_INT)TYPE_ALIGN_UNIT (vectype)));
+ ptr = make_ssa_name (SSA_NAME_VAR (dataref_ptr), new_stmt);
+ gimple_assign_set_lhs (new_stmt, ptr);
+ vect_finish_stmt_generation (stmt, new_stmt, gsi);
+ data_ref
+ = build2 (MEM_REF, vectype, ptr,
+ build_int_cst (reference_alias_ptr_type
+ (DR_REF (first_dr)), 0));
break;
}
case dr_explicit_realign_optimized:
- data_ref = build1 (ALIGN_INDIRECT_REF, vectype, dataref_ptr);
+ new_stmt = gimple_build_assign_with_ops
+ (BIT_AND_EXPR, NULL_TREE, dataref_ptr,
+ build_int_cst
+ (TREE_TYPE (dataref_ptr),
+ -(HOST_WIDE_INT)TYPE_ALIGN_UNIT (vectype)));
+ new_temp = make_ssa_name (SSA_NAME_VAR (dataref_ptr), new_stmt);
+ gimple_assign_set_lhs (new_stmt, new_temp);
+ vect_finish_stmt_generation (stmt, new_stmt, gsi);
+ data_ref
+ = build2 (MEM_REF, vectype, new_temp,
+ build_int_cst (reference_alias_ptr_type
+ (DR_REF (first_dr)), 0));
break;
default:
gcc_unreachable ();
}
- /* If accesses through a pointer to vectype do not alias the original
- memory reference we have a problem. This should never happen. */
- gcc_assert (alias_sets_conflict_p (get_alias_set (data_ref),
- get_alias_set (gimple_assign_rhs1 (stmt))));
vec_dest = vect_create_destination_var (scalar_dest, vectype);
new_stmt = gimple_build_assign (vec_dest, data_ref);
new_temp = make_ssa_name (vec_dest, new_stmt);
return false;
vec_mode = TYPE_MODE (vectype);
- if ((icode1 = optab_handler (optab1, vec_mode)->insn_code) == CODE_FOR_nothing
- || (icode2 = optab_handler (optab2, vec_mode)->insn_code)
- == CODE_FOR_nothing)
+ if ((icode1 = optab_handler (optab1, vec_mode)) == CODE_FOR_nothing
+ || (icode2 = optab_handler (optab2, vec_mode)) == CODE_FOR_nothing)
return false;
/* Check if it's a multi-step conversion that can be done using intermediate
optab4 = optab_for_tree_code (c2, intermediate_type, optab_default);
if (!optab3 || !optab4
- || (icode1 = optab1->handlers[(int) prev_mode].insn_code)
- == CODE_FOR_nothing
+ || ((icode1 = optab_handler (optab1, prev_mode))
+ == CODE_FOR_nothing)
|| insn_data[icode1].operand[0].mode != intermediate_mode
- || (icode2 = optab2->handlers[(int) prev_mode].insn_code)
- == CODE_FOR_nothing
+ || ((icode2 = optab_handler (optab2, prev_mode))
+ == CODE_FOR_nothing)
|| insn_data[icode2].operand[0].mode != intermediate_mode
- || (icode1 = optab3->handlers[(int) intermediate_mode].insn_code)
- == CODE_FOR_nothing
- || (icode2 = optab4->handlers[(int) intermediate_mode].insn_code)
- == CODE_FOR_nothing)
+ || ((icode1 = optab_handler (optab3, intermediate_mode))
+ == CODE_FOR_nothing)
+ || ((icode2 = optab_handler (optab4, intermediate_mode))
+ == CODE_FOR_nothing))
return false;
VEC_quick_push (tree, *interm_types, intermediate_type);
return false;
vec_mode = TYPE_MODE (vectype);
- if ((icode1 = optab_handler (optab1, vec_mode)->insn_code)
- == CODE_FOR_nothing)
+ if ((icode1 = optab_handler (optab1, vec_mode)) == CODE_FOR_nothing)
return false;
/* Check if it's a multi-step conversion that can be done using intermediate
interm_optab = optab_for_tree_code (c1, intermediate_type,
optab_default);
if (!interm_optab
- || (icode1 = optab1->handlers[(int) prev_mode].insn_code)
- == CODE_FOR_nothing
+ || ((icode1 = optab_handler (optab1, prev_mode))
+ == CODE_FOR_nothing)
|| insn_data[icode1].operand[0].mode != intermediate_mode
- || (icode1
- = interm_optab->handlers[(int) intermediate_mode].insn_code)
- == CODE_FOR_nothing)
+ || ((icode1 = optab_handler (interm_optab, intermediate_mode))
+ == CODE_FOR_nothing))
return false;
VEC_quick_push (tree, *interm_types, intermediate_type);