/* Statement Analysis and Transformation for Vectorization
- Copyright (C) 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010
+ Copyright (C) 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011
Free Software Foundation, Inc.
Contributed by Dorit Naishlos <dorit@il.ibm.com>
and Ira Rosen <irar@il.ibm.com>
#include "recog.h"
#include "optabs.h"
#include "diagnostic-core.h"
-#include "toplev.h"
#include "tree-vectorizer.h"
#include "langhooks.h"
+/* Return a variable of type ELEM_TYPE[NELEMS]. */
+
+static tree
+create_vector_array (tree elem_type, unsigned HOST_WIDE_INT nelems)
+{
+ return create_tmp_var (build_array_type_nelts (elem_type, nelems),
+ "vect_array");
+}
+
+/* ARRAY is an array of vectors created by create_vector_array.
+ Return an SSA_NAME for the vector in index N. The reference
+ is part of the vectorization of STMT and the vector is associated
+ with scalar destination SCALAR_DEST. */
+
+static tree
+read_vector_array (gimple stmt, gimple_stmt_iterator *gsi, tree scalar_dest,
+ tree array, unsigned HOST_WIDE_INT n)
+{
+ tree vect_type, vect, vect_name, array_ref;
+ gimple new_stmt;
+
+ gcc_assert (TREE_CODE (TREE_TYPE (array)) == ARRAY_TYPE);
+ vect_type = TREE_TYPE (TREE_TYPE (array));
+ vect = vect_create_destination_var (scalar_dest, vect_type);
+ array_ref = build4 (ARRAY_REF, vect_type, array,
+ build_int_cst (size_type_node, n),
+ NULL_TREE, NULL_TREE);
+
+ new_stmt = gimple_build_assign (vect, array_ref);
+ vect_name = make_ssa_name (vect, new_stmt);
+ gimple_assign_set_lhs (new_stmt, vect_name);
+ vect_finish_stmt_generation (stmt, new_stmt, gsi);
+ mark_symbols_for_renaming (new_stmt);
+
+ return vect_name;
+}
+
+/* ARRAY is an array of vectors created by create_vector_array.
+ Emit code to store SSA_NAME VECT in index N of the array.
+ The store is part of the vectorization of STMT. */
+
+static void
+write_vector_array (gimple stmt, gimple_stmt_iterator *gsi, tree vect,
+ tree array, unsigned HOST_WIDE_INT n)
+{
+ tree array_ref;
+ gimple new_stmt;
+
+ array_ref = build4 (ARRAY_REF, TREE_TYPE (vect), array,
+ build_int_cst (size_type_node, n),
+ NULL_TREE, NULL_TREE);
+
+ new_stmt = gimple_build_assign (array_ref, vect);
+ vect_finish_stmt_generation (stmt, new_stmt, gsi);
+ mark_symbols_for_renaming (new_stmt);
+}
+
+/* PTR is a pointer to an array of type TYPE. Return a representation
+ of *PTR. The memory reference replaces those in FIRST_DR
+ (and its group). */
+
+static tree
+create_array_ref (tree type, tree ptr, struct data_reference *first_dr)
+{
+ struct ptr_info_def *pi;
+ tree mem_ref, alias_ptr_type;
+
+ alias_ptr_type = reference_alias_ptr_type (DR_REF (first_dr));
+ mem_ref = build2 (MEM_REF, type, ptr, build_int_cst (alias_ptr_type, 0));
+ /* Arrays have the same alignment as their type. */
+ pi = get_ptr_info (ptr);
+ pi->align = TYPE_ALIGN_UNIT (type);
+ pi->misalign = 0;
+ return mem_ref;
+}
+
/* Utility functions used by vect_mark_stmts_to_be_vectorized. */
/* Function vect_mark_relevant.
static void
vect_mark_relevant (VEC(gimple,heap) **worklist, gimple stmt,
- enum vect_relevant relevant, bool live_p)
+ enum vect_relevant relevant, bool live_p,
+ bool used_in_pattern)
{
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
enum vect_relevant save_relevant = STMT_VINFO_RELEVANT (stmt_info);
bool save_live_p = STMT_VINFO_LIVE_P (stmt_info);
+ gimple pattern_stmt;
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "mark relevant %d, live %d.", relevant, live_p);
+ /* If this stmt is an original stmt in a pattern, we might need to mark its
+ related pattern stmt instead of the original stmt. However, such stmts
+ may have their own uses that are not in any pattern, in such cases the
+ stmt itself should be marked. */
if (STMT_VINFO_IN_PATTERN_P (stmt_info))
{
- gimple pattern_stmt;
+ bool found = false;
+ if (!used_in_pattern)
+ {
+ imm_use_iterator imm_iter;
+ use_operand_p use_p;
+ gimple use_stmt;
+ tree lhs;
- /* This is the last stmt in a sequence that was detected as a
- pattern that can potentially be vectorized. Don't mark the stmt
- as relevant/live because it's not going to be vectorized.
- Instead mark the pattern-stmt that replaces it. */
+ if (is_gimple_assign (stmt))
+ lhs = gimple_assign_lhs (stmt);
+ else
+ lhs = gimple_call_lhs (stmt);
- pattern_stmt = STMT_VINFO_RELATED_STMT (stmt_info);
+ /* This use is out of pattern use, if LHS has other uses that are
+ pattern uses, we should mark the stmt itself, and not the pattern
+ stmt. */
+ if (TREE_CODE (lhs) == SSA_NAME)
+ FOR_EACH_IMM_USE_FAST (use_p, imm_iter, lhs)
+ {
+ if (is_gimple_debug (USE_STMT (use_p)))
+ continue;
+ use_stmt = USE_STMT (use_p);
- if (vect_print_dump_info (REPORT_DETAILS))
- fprintf (vect_dump, "last stmt in pattern. don't mark relevant/live.");
- stmt_info = vinfo_for_stmt (pattern_stmt);
- gcc_assert (STMT_VINFO_RELATED_STMT (stmt_info) == stmt);
- save_relevant = STMT_VINFO_RELEVANT (stmt_info);
- save_live_p = STMT_VINFO_LIVE_P (stmt_info);
- stmt = pattern_stmt;
+ if (vinfo_for_stmt (use_stmt)
+ && STMT_VINFO_IN_PATTERN_P (vinfo_for_stmt (use_stmt)))
+ {
+ found = true;
+ break;
+ }
+ }
+ }
+
+ if (!found)
+ {
+ /* This is the last stmt in a sequence that was detected as a
+ pattern that can potentially be vectorized. Don't mark the stmt
+ as relevant/live because it's not going to be vectorized.
+ Instead mark the pattern-stmt that replaces it. */
+
+ pattern_stmt = STMT_VINFO_RELATED_STMT (stmt_info);
+
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "last stmt in pattern. don't mark"
+ " relevant/live.");
+ stmt_info = vinfo_for_stmt (pattern_stmt);
+ gcc_assert (STMT_VINFO_RELATED_STMT (stmt_info) == stmt);
+ save_relevant = STMT_VINFO_RELEVANT (stmt_info);
+ save_live_p = STMT_VINFO_LIVE_P (stmt_info);
+ stmt = pattern_stmt;
+ }
}
STMT_VINFO_LIVE_P (stmt_info) |= live_p;
}
}
- vect_mark_relevant (worklist, def_stmt, relevant, live_p);
+ vect_mark_relevant (worklist, def_stmt, relevant, live_p,
+ is_pattern_stmt_p (stmt_vinfo));
return true;
}
}
if (vect_stmt_relevant_p (phi, loop_vinfo, &relevant, &live_p))
- vect_mark_relevant (&worklist, phi, relevant, live_p);
+ vect_mark_relevant (&worklist, phi, relevant, live_p, false);
}
for (si = gsi_start_bb (bb); !gsi_end_p (si); gsi_next (&si))
{
}
if (vect_stmt_relevant_p (stmt, loop_vinfo, &relevant, &live_p))
- vect_mark_relevant (&worklist, stmt, relevant, live_p);
+ vect_mark_relevant (&worklist, stmt, relevant, live_p, false);
}
}
break;
}
- FOR_EACH_PHI_OR_STMT_USE (use_p, stmt, iter, SSA_OP_USE)
- {
- tree op = USE_FROM_PTR (use_p);
- if (!process_use (stmt, op, loop_vinfo, live_p, relevant, &worklist))
- {
- VEC_free (gimple, heap, worklist);
- return false;
- }
- }
+ if (is_pattern_stmt_p (vinfo_for_stmt (stmt)))
+ {
+ /* Pattern statements are not inserted into the code, so
+ FOR_EACH_PHI_OR_STMT_USE optimizes their operands out, and we
+ have to scan the RHS or function arguments instead. */
+ if (is_gimple_assign (stmt))
+ {
+ enum tree_code rhs_code = gimple_assign_rhs_code (stmt);
+ tree op = gimple_assign_rhs1 (stmt);
+
+ i = 1;
+ if (rhs_code == COND_EXPR && COMPARISON_CLASS_P (op))
+ {
+ if (!process_use (stmt, TREE_OPERAND (op, 0), loop_vinfo,
+ live_p, relevant, &worklist)
+ || !process_use (stmt, TREE_OPERAND (op, 1), loop_vinfo,
+ live_p, relevant, &worklist))
+ {
+ VEC_free (gimple, heap, worklist);
+ return false;
+ }
+ i = 2;
+ }
+ for (; i < gimple_num_ops (stmt); i++)
+ {
+ op = gimple_op (stmt, i);
+ if (!process_use (stmt, op, loop_vinfo, live_p, relevant,
+ &worklist))
+ {
+ VEC_free (gimple, heap, worklist);
+ return false;
+ }
+ }
+ }
+ else if (is_gimple_call (stmt))
+ {
+ for (i = 0; i < gimple_call_num_args (stmt); i++)
+ {
+ tree arg = gimple_call_arg (stmt, i);
+ if (!process_use (stmt, arg, loop_vinfo, live_p, relevant,
+ &worklist))
+ {
+ VEC_free (gimple, heap, worklist);
+ return false;
+ }
+ }
+ }
+ }
+ else
+ FOR_EACH_PHI_OR_STMT_USE (use_p, stmt, iter, SSA_OP_USE)
+ {
+ tree op = USE_FROM_PTR (use_p);
+ if (!process_use (stmt, op, loop_vinfo, live_p, relevant,
+ &worklist))
+ {
+ VEC_free (gimple, heap, worklist);
+ return false;
+ }
+ }
} /* while worklist */
VEC_free (gimple, heap, worklist);
static int
vect_cost_strided_group_size (stmt_vec_info stmt_info)
{
- gimple first_stmt = DR_GROUP_FIRST_DR (stmt_info);
+ gimple first_stmt = GROUP_FIRST_ELEMENT (stmt_info);
if (first_stmt == STMT_VINFO_STMT (stmt_info))
- return DR_GROUP_SIZE (stmt_info);
+ return GROUP_SIZE (stmt_info);
return 1;
}
void
vect_model_store_cost (stmt_vec_info stmt_info, int ncopies,
- enum vect_def_type dt, slp_tree slp_node)
+ bool store_lanes_p, enum vect_def_type dt,
+ slp_tree slp_node)
{
int group_size;
unsigned int inside_cost = 0, outside_cost = 0;
outside_cost = vect_get_stmt_cost (scalar_to_vec);
/* Strided access? */
- if (DR_GROUP_FIRST_DR (stmt_info))
+ if (STMT_VINFO_STRIDED_ACCESS (stmt_info))
{
if (slp_node)
{
}
else
{
- first_stmt = DR_GROUP_FIRST_DR (stmt_info);
+ first_stmt = GROUP_FIRST_ELEMENT (stmt_info);
group_size = vect_cost_strided_group_size (stmt_info);
}
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. */
- if (group_size > 1)
+ /* We assume that the cost of a single store-lanes instruction is
+ equivalent to the cost of GROUP_SIZE separate stores. If a strided
+ access is instead being provided by a permute-and-store operation,
+ include the cost of the permutes. */
+ if (!store_lanes_p && group_size > 1)
{
/* Uses a high and low interleave operation for each needed permute. */
inside_cost = ncopies * exact_log2(group_size) * group_size
access scheme chosen. */
void
-vect_model_load_cost (stmt_vec_info stmt_info, int ncopies, slp_tree slp_node)
-
+vect_model_load_cost (stmt_vec_info stmt_info, int ncopies, bool load_lanes_p,
+ slp_tree slp_node)
{
int group_size;
gimple first_stmt;
return;
/* Strided accesses? */
- first_stmt = DR_GROUP_FIRST_DR (stmt_info);
- if (first_stmt && !slp_node)
+ first_stmt = GROUP_FIRST_ELEMENT (stmt_info);
+ if (STMT_VINFO_STRIDED_ACCESS (stmt_info) && first_stmt && !slp_node)
{
group_size = vect_cost_strided_group_size (stmt_info);
first_dr = STMT_VINFO_DATA_REF (vinfo_for_stmt (first_stmt));
first_dr = 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)
+ /* We assume that the cost of a single load-lanes instruction is
+ equivalent to the cost of GROUP_SIZE separate loads. If a strided
+ access is instead being provided by a load-and-permute operation,
+ include the cost of the permutes. */
+ if (!load_lanes_p && group_size > 1)
{
/* Uses an even and odd extract operations for each needed permute. */
inside_cost = ncopies * exact_log2(group_size) * group_size
/* The loads themselves. */
vect_get_load_cost (first_dr, ncopies,
- ((!DR_GROUP_FIRST_DR (stmt_info)) || group_size > 1 || slp_node),
+ ((!STMT_VINFO_STRIDED_ACCESS (stmt_info)) || group_size > 1
+ || slp_node),
&inside_cost, &outside_cost);
if (vect_print_dump_info (REPORT_COST))
gimple def_stmt;
stmt_vec_info def_stmt_info = NULL;
stmt_vec_info stmt_vinfo = vinfo_for_stmt (stmt);
- tree vectype = STMT_VINFO_VECTYPE (stmt_vinfo);
- unsigned int nunits = TYPE_VECTOR_SUBPARTS (vectype);
+ unsigned int nunits;
loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_vinfo);
tree vec_inv;
tree vec_cst;
{
vector_type = get_vectype_for_scalar_type (TREE_TYPE (op));
gcc_assert (vector_type);
+ nunits = TYPE_VECTOR_SUBPARTS (vector_type);
if (scalar_def)
*scalar_def = op;
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "Create vector_cst. nunits = %d", nunits);
- for (i = nunits - 1; i >= 0; --i)
- {
- t = tree_cons (NULL_TREE, op, t);
- }
- vec_cst = build_vector (vector_type, t);
+ vec_cst = build_vector_from_val (vector_type,
+ fold_convert (TREE_TYPE (vector_type),
+ op));
return vect_init_vector (stmt, vec_cst, vector_type, NULL);
}
/* Get the def from the vectorized stmt. */
def_stmt_info = vinfo_for_stmt (def_stmt);
+
vec_stmt = STMT_VINFO_VEC_STMT (def_stmt_info);
+ /* Get vectorized pattern statement. */
+ if (!vec_stmt
+ && STMT_VINFO_IN_PATTERN_P (def_stmt_info)
+ && !STMT_VINFO_RELEVANT (def_stmt_info))
+ vec_stmt = STMT_VINFO_VEC_STMT (vinfo_for_stmt (
+ STMT_VINFO_RELATED_STMT (def_stmt_info)));
gcc_assert (vec_stmt);
if (gimple_code (vec_stmt) == GIMPLE_PHI)
vec_oprnd = PHI_RESULT (vec_stmt);
/* Get the def from the vectorized stmt. */
def_stmt_info = vinfo_for_stmt (def_stmt);
vec_stmt = STMT_VINFO_VEC_STMT (def_stmt_info);
- gcc_assert (vec_stmt && gimple_code (vec_stmt) == GIMPLE_PHI);
- vec_oprnd = PHI_RESULT (vec_stmt);
+ if (gimple_code (vec_stmt) == GIMPLE_PHI)
+ vec_oprnd = PHI_RESULT (vec_stmt);
+ else
+ vec_oprnd = gimple_get_lhs (vec_stmt);
return vec_oprnd;
}
slp_tree slp_node)
{
if (slp_node)
- vect_get_slp_defs (slp_node, vec_oprnds0, vec_oprnds1, -1);
+ vect_get_slp_defs (op0, op1, slp_node, vec_oprnds0, vec_oprnds1, -1);
else
{
tree vec_oprnd;
print_gimple_stmt (vect_dump, vec_stmt, 0, TDF_SLIM);
}
- gimple_set_location (vec_stmt, gimple_location (gsi_stmt (*gsi)));
+ gimple_set_location (vec_stmt, gimple_location (stmt));
}
/* Checks if CALL can be vectorized in type VECTYPE. Returns
loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
tree fndecl, new_temp, def, rhs_type;
gimple def_stmt;
- enum vect_def_type dt[2] = {vect_unknown_def_type, vect_unknown_def_type};
+ enum vect_def_type dt[3]
+ = {vect_unknown_def_type, vect_unknown_def_type, vect_unknown_def_type};
gimple new_stmt = NULL;
int ncopies, j;
VEC(tree, heap) *vargs = NULL;
enum { NARROW, NONE, WIDEN } modifier;
size_t i, nargs;
+ tree lhs;
/* FORNOW: unsupported in basic block SLP. */
gcc_assert (loop_vinfo);
if (TREE_CODE (gimple_call_lhs (stmt)) != SSA_NAME)
return false;
- if (stmt_could_throw_p (stmt))
+ if (stmt_can_throw_internal (stmt))
return false;
vectype_out = STMT_VINFO_VECTYPE (stmt_info);
vectype_in = NULL_TREE;
nargs = gimple_call_num_args (stmt);
- /* Bail out if the function has more than two arguments, we
- do not have interesting builtin functions to vectorize with
- more than two arguments. No arguments is also not good. */
- if (nargs == 0 || nargs > 2)
+ /* Bail out if the function has more than three arguments, we do not have
+ interesting builtin functions to vectorize with more than two arguments
+ except for fma. No arguments is also not good. */
+ if (nargs == 0 || nargs > 3)
return false;
for (i = 0; i < nargs; i++)
/** Transform. **/
if (vect_print_dump_info (REPORT_DETAILS))
- fprintf (vect_dump, "transform operation.");
+ fprintf (vect_dump, "transform call.");
/* Handle def. */
scalar_dest = gimple_call_lhs (stmt);
}
else
{
- vec_oprnd1 = gimple_call_arg (new_stmt, 2*i);
+ vec_oprnd1 = gimple_call_arg (new_stmt, 2*i + 1);
vec_oprnd0
= vect_get_vec_def_for_stmt_copy (dt[i], vec_oprnd1);
vec_oprnd1
rhs of the statement with something harmless. */
type = TREE_TYPE (scalar_dest);
- new_stmt = gimple_build_assign (gimple_call_lhs (stmt),
- fold_convert (type, integer_zero_node));
+ if (is_pattern_stmt_p (stmt_info))
+ lhs = gimple_call_lhs (STMT_VINFO_RELATED_STMT (stmt_info));
+ else
+ lhs = gimple_call_lhs (stmt);
+ new_stmt = gimple_build_assign (lhs, build_zero_cst (type));
set_vinfo_for_stmt (new_stmt, stmt_info);
set_vinfo_for_stmt (stmt, NULL);
STMT_VINFO_STMT (stmt_info) = new_stmt;
return new_stmt;
}
-
/* Check if STMT performs a conversion operation, that can be vectorized.
If VEC_STMT is also passed, vectorize the STMT: create a vectorized
stmt to replace it, put it in VEC_STMT, and insert it at BSI.
tree vectype_out, vectype_in;
int ncopies, j;
tree rhs_type;
- tree builtin_decl;
enum { NARROW, NONE, WIDEN } modifier;
int i;
VEC(tree,heap) *vec_oprnds0 = NULL;
/* Multiple types in SLP are handled by creating the appropriate number of
vectorized stmts for each SLP node. Hence, NCOPIES is always 1 in
case of SLP. */
- if (slp_node)
+ if (slp_node || PURE_SLP_STMT (stmt_info))
ncopies = 1;
/* Sanity check: make sure that at least one copy of the vectorized stmt
/* Supportable by target? */
if ((modifier == NONE
- && !targetm.vectorize.builtin_conversion (code, vectype_out, vectype_in))
+ && !supportable_convert_operation (code, vectype_out, vectype_in, &decl1, &code1))
|| (modifier == WIDEN
&& !supportable_widening_operation (code, stmt,
vectype_out, vectype_in,
else
vect_get_vec_defs_for_stmt_copy (dt, &vec_oprnds0, NULL);
- builtin_decl =
- targetm.vectorize.builtin_conversion (code,
- vectype_out, vectype_in);
FOR_EACH_VEC_ELT (tree, vec_oprnds0, i, vop0)
- {
- /* Arguments are ready. create the new vector stmt. */
- new_stmt = gimple_build_call (builtin_decl, 1, vop0);
- new_temp = make_ssa_name (vec_dest, new_stmt);
- gimple_call_set_lhs (new_stmt, new_temp);
- vect_finish_stmt_generation (stmt, new_stmt, gsi);
- if (slp_node)
- VEC_quick_push (gimple, SLP_TREE_VEC_STMTS (slp_node), new_stmt);
- }
+ {
+ /* Arguments are ready, create the new vector stmt. */
+ if (code1 == CALL_EXPR)
+ {
+ new_stmt = gimple_build_call (decl1, 1, vop0);
+ new_temp = make_ssa_name (vec_dest, new_stmt);
+ gimple_call_set_lhs (new_stmt, new_temp);
+ }
+ else
+ {
+ gcc_assert (TREE_CODE_LENGTH (code) == unary_op);
+ new_stmt = gimple_build_assign_with_ops (code, vec_dest, vop0,
+ NULL);
+ new_temp = make_ssa_name (vec_dest, new_stmt);
+ gimple_assign_set_lhs (new_stmt, new_temp);
+ }
+
+ vect_finish_stmt_generation (stmt, new_stmt, gsi);
+ if (slp_node)
+ VEC_quick_push (gimple, SLP_TREE_VEC_STMTS (slp_node), new_stmt);
+ }
if (j == 0)
STMT_VINFO_VEC_STMT (stmt_info) = *vec_stmt = new_stmt;
vec_oprnd1 = vect_get_vec_def_for_stmt_copy (dt[0], vec_oprnd0);
}
- /* Arguments are ready. Create the new vector stmt. */
+ /* Arguments are ready. Create the new vector stmt. */
new_stmt = gimple_build_assign_with_ops (code1, vec_dest, vec_oprnd0,
vec_oprnd1);
new_temp = make_ssa_name (vec_dest, new_stmt);
/* Multiple types in SLP are handled by creating the appropriate number of
vectorized stmts for each SLP node. Hence, NCOPIES is always 1 in
case of SLP. */
- if (slp_node)
+ if (slp_node || PURE_SLP_STMT (stmt_info))
ncopies = 1;
else
ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits;
else
return false;
+ if (code == VIEW_CONVERT_EXPR)
+ op = TREE_OPERAND (op, 0);
+
if (!vect_is_simple_use_1 (op, loop_vinfo, bb_vinfo,
&def_stmt, &def, &dt[0], &vectype_in))
{
/* We can handle NOP_EXPR conversions that do not change the number
of elements or the vector size. */
- if (CONVERT_EXPR_CODE_P (code)
+ if ((CONVERT_EXPR_CODE_P (code)
+ || code == VIEW_CONVERT_EXPR)
&& (!vectype_in
|| TYPE_VECTOR_SUBPARTS (vectype_in) != nunits
|| (GET_MODE_SIZE (TYPE_MODE (vectype))
!= GET_MODE_SIZE (TYPE_MODE (vectype_in)))))
return false;
+ /* We do not handle bit-precision changes. */
+ if ((CONVERT_EXPR_CODE_P (code)
+ || code == VIEW_CONVERT_EXPR)
+ && INTEGRAL_TYPE_P (TREE_TYPE (scalar_dest))
+ && ((TYPE_PRECISION (TREE_TYPE (scalar_dest))
+ != GET_MODE_PRECISION (TYPE_MODE (TREE_TYPE (scalar_dest))))
+ || ((TYPE_PRECISION (TREE_TYPE (op))
+ != GET_MODE_PRECISION (TYPE_MODE (TREE_TYPE (op))))))
+ /* But a conversion that does not change the bit-pattern is ok. */
+ && !((TYPE_PRECISION (TREE_TYPE (scalar_dest))
+ > TYPE_PRECISION (TREE_TYPE (op)))
+ && TYPE_UNSIGNED (TREE_TYPE (op))))
+ {
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "type conversion to/from bit-precision "
+ "unsupported.");
+ 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_EACH_VEC_ELT (tree, vec_oprnds, i, vop)
{
- if (CONVERT_EXPR_CODE_P (code))
+ if (CONVERT_EXPR_CODE_P (code)
+ || code == VIEW_CONVERT_EXPR)
vop = build1 (VIEW_CONVERT_EXPR, vectype, vop);
new_stmt = gimple_build_assign (vec_dest, vop);
new_temp = make_ssa_name (vec_dest, new_stmt);
return true;
}
-/* Function vectorizable_operation.
- Check if STMT performs a binary or unary operation that can be vectorized.
+/* Return TRUE if CODE (a shift operation) is supported for SCALAR_TYPE
+ either as shift by a scalar or by a vector. */
+
+bool
+vect_supportable_shift (enum tree_code code, tree scalar_type)
+{
+
+ enum machine_mode vec_mode;
+ optab optab;
+ int icode;
+ tree vectype;
+
+ vectype = get_vectype_for_scalar_type (scalar_type);
+ if (!vectype)
+ return false;
+
+ optab = optab_for_tree_code (code, vectype, optab_scalar);
+ if (!optab
+ || optab_handler (optab, TYPE_MODE (vectype)) == CODE_FOR_nothing)
+ {
+ optab = optab_for_tree_code (code, vectype, optab_vector);
+ if (!optab
+ || (optab_handler (optab, TYPE_MODE (vectype))
+ == CODE_FOR_nothing))
+ return false;
+ }
+
+ vec_mode = TYPE_MODE (vectype);
+ icode = (int) optab_handler (optab, vec_mode);
+ if (icode == CODE_FOR_nothing)
+ return false;
+
+ return true;
+}
+
+
+/* Function vectorizable_shift.
+
+ Check if STMT performs a shift operation that can be vectorized.
If VEC_STMT is also passed, vectorize the STMT: create a vectorized
stmt to replace it, put it in VEC_STMT, and insert it at BSI.
Return FALSE if not a vectorizable STMT, TRUE otherwise. */
static bool
-vectorizable_operation (gimple stmt, gimple_stmt_iterator *gsi,
- gimple *vec_stmt, slp_tree slp_node)
+vectorizable_shift (gimple stmt, gimple_stmt_iterator *gsi,
+ gimple *vec_stmt, slp_tree slp_node)
{
tree vec_dest;
tree scalar_dest;
enum tree_code code;
enum machine_mode vec_mode;
tree new_temp;
- int op_type;
optab optab;
int icode;
enum machine_mode optab_op2_mode;
int nunits_in;
int nunits_out;
tree vectype_out;
+ tree op1_vectype;
int ncopies;
int j, i;
- VEC(tree,heap) *vec_oprnds0 = NULL, *vec_oprnds1 = NULL;
+ VEC (tree, heap) *vec_oprnds0 = NULL, *vec_oprnds1 = NULL;
tree vop0, vop1;
unsigned int k;
- bool scalar_shift_arg = false;
+ bool scalar_shift_arg = true;
bb_vec_info bb_vinfo = STMT_VINFO_BB_VINFO (stmt_info);
int vf;
code = gimple_assign_rhs_code (stmt);
- /* For pointer addition, we should use the normal plus for
- the vector addition. */
- if (code == POINTER_PLUS_EXPR)
- code = PLUS_EXPR;
+ if (!(code == LSHIFT_EXPR || code == RSHIFT_EXPR || code == LROTATE_EXPR
+ || code == RROTATE_EXPR))
+ return false;
- /* Support only unary or binary operations. */
- op_type = TREE_CODE_LENGTH (code);
- if (op_type != unary_op && op_type != binary_op)
+ scalar_dest = gimple_assign_lhs (stmt);
+ vectype_out = STMT_VINFO_VECTYPE (stmt_info);
+ if (TYPE_PRECISION (TREE_TYPE (scalar_dest))
+ != GET_MODE_PRECISION (TYPE_MODE (TREE_TYPE (scalar_dest))))
{
if (vect_print_dump_info (REPORT_DETAILS))
- fprintf (vect_dump, "num. args = %d (not unary/binary op).", op_type);
+ fprintf (vect_dump, "bit-precision shifts not supported.");
return false;
}
- scalar_dest = gimple_assign_lhs (stmt);
- vectype_out = STMT_VINFO_VECTYPE (stmt_info);
-
op0 = gimple_assign_rhs1 (stmt);
if (!vect_is_simple_use_1 (op0, loop_vinfo, bb_vinfo,
- &def_stmt, &def, &dt[0], &vectype))
+ &def_stmt, &def, &dt[0], &vectype))
{
if (vect_print_dump_info (REPORT_DETAILS))
fprintf (vect_dump, "use not simple.");
if (nunits_out != nunits_in)
return false;
- if (op_type == binary_op)
+ op1 = gimple_assign_rhs2 (stmt);
+ if (!vect_is_simple_use_1 (op1, loop_vinfo, bb_vinfo, &def_stmt, &def,
+ &dt[1], &op1_vectype))
{
- op1 = gimple_assign_rhs2 (stmt);
- if (!vect_is_simple_use (op1, loop_vinfo, bb_vinfo, &def_stmt, &def,
- &dt[1]))
- {
- if (vect_print_dump_info (REPORT_DETAILS))
- fprintf (vect_dump, "use not simple.");
- return false;
- }
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "use not simple.");
+ return false;
}
if (loop_vinfo)
/* Multiple types in SLP are handled by creating the appropriate number of
vectorized stmts for each SLP node. Hence, NCOPIES is always 1 in
case of SLP. */
- if (slp_node)
+ if (slp_node || PURE_SLP_STMT (stmt_info))
ncopies = 1;
else
ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits_in;
gcc_assert (ncopies >= 1);
- /* If this is a shift/rotate, determine whether the shift amount is a vector,
- or scalar. If the shift/rotate amount is a vector, use the vector/vector
- shift optabs. */
- if (code == LSHIFT_EXPR || code == RSHIFT_EXPR || code == LROTATE_EXPR
- || code == RROTATE_EXPR)
+ /* Determine whether the shift amount is a vector, or scalar. If the
+ shift/rotate amount is a vector, use the vector/vector shift optabs. */
+
+ if (dt[1] == vect_internal_def && !slp_node)
+ scalar_shift_arg = false;
+ else if (dt[1] == vect_constant_def
+ || dt[1] == vect_external_def
+ || dt[1] == vect_internal_def)
{
- /* vector shifted by vector */
- if (dt[1] == vect_internal_def)
+ /* In SLP, need to check whether the shift count is the same,
+ in loops if it is a constant or invariant, it is always
+ a scalar shift. */
+ if (slp_node)
{
- optab = optab_for_tree_code (code, vectype, optab_vector);
- if (vect_print_dump_info (REPORT_DETAILS))
- fprintf (vect_dump, "vector/vector shift/rotate found.");
- }
+ VEC (gimple, heap) *stmts = SLP_TREE_SCALAR_STMTS (slp_node);
+ gimple slpstmt;
- /* See if the machine has a vector shifted by scalar insn and if not
- then see if it has a vector shifted by vector insn */
- else if (dt[1] == vect_constant_def || dt[1] == vect_external_def)
- {
- optab = optab_for_tree_code (code, vectype, optab_scalar);
- if (optab
- && optab_handler (optab, TYPE_MODE (vectype)) != CODE_FOR_nothing)
- {
- scalar_shift_arg = true;
- if (vect_print_dump_info (REPORT_DETAILS))
- fprintf (vect_dump, "vector/scalar shift/rotate found.");
- }
- else
- {
- optab = optab_for_tree_code (code, vectype, optab_vector);
- if (optab
- && (optab_handler (optab, TYPE_MODE (vectype))
- != CODE_FOR_nothing))
- {
- if (vect_print_dump_info (REPORT_DETAILS))
- fprintf (vect_dump, "vector/vector shift/rotate found.");
-
- /* Unlike the other binary operators, shifts/rotates have
- the rhs being int, instead of the same type as the lhs,
- so make sure the scalar is the right type if we are
- dealing with vectors of short/char. */
- if (dt[1] == vect_constant_def)
- op1 = fold_convert (TREE_TYPE (vectype), op1);
- }
- }
+ FOR_EACH_VEC_ELT (gimple, stmts, k, slpstmt)
+ if (!operand_equal_p (gimple_assign_rhs2 (slpstmt), op1, 0))
+ scalar_shift_arg = false;
}
+ }
+ else
+ {
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "operand mode requires invariant argument.");
+ return false;
+ }
- else
+ /* Vector shifted by vector. */
+ if (!scalar_shift_arg)
+ {
+ optab = optab_for_tree_code (code, vectype, optab_vector);
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "vector/vector shift/rotate found.");
+ if (!op1_vectype)
+ op1_vectype = get_same_sized_vectype (TREE_TYPE (op1), vectype_out);
+ if (op1_vectype == NULL_TREE
+ || TYPE_MODE (op1_vectype) != TYPE_MODE (vectype))
{
if (vect_print_dump_info (REPORT_DETAILS))
- fprintf (vect_dump, "operand mode requires invariant argument.");
+ fprintf (vect_dump, "unusable type for last operand in"
+ " vector/vector shift/rotate.");
return false;
}
}
+ /* See if the machine has a vector shifted by scalar insn and if not
+ then see if it has a vector shifted by vector insn. */
else
- optab = optab_for_tree_code (code, vectype, optab_default);
+ {
+ optab = optab_for_tree_code (code, vectype, optab_scalar);
+ if (optab
+ && optab_handler (optab, TYPE_MODE (vectype)) != CODE_FOR_nothing)
+ {
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "vector/scalar shift/rotate found.");
+ }
+ else
+ {
+ optab = optab_for_tree_code (code, vectype, optab_vector);
+ if (optab
+ && (optab_handler (optab, TYPE_MODE (vectype))
+ != CODE_FOR_nothing))
+ {
+ scalar_shift_arg = false;
+
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "vector/vector shift/rotate found.");
+
+ /* Unlike the other binary operators, shifts/rotates have
+ the rhs being int, instead of the same type as the lhs,
+ so make sure the scalar is the right type if we are
+ dealing with vectors of long long/long/short/char. */
+ if (dt[1] == vect_constant_def)
+ op1 = fold_convert (TREE_TYPE (vectype), op1);
+ else if (!useless_type_conversion_p (TREE_TYPE (vectype),
+ TREE_TYPE (op1)))
+ {
+ if (slp_node
+ && TYPE_MODE (TREE_TYPE (vectype))
+ != TYPE_MODE (TREE_TYPE (op1)))
+ {
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "unusable type for last operand in"
+ " vector/vector shift/rotate.");
+ return false;
+ }
+ if (vec_stmt && !slp_node)
+ {
+ op1 = fold_convert (TREE_TYPE (vectype), op1);
+ op1 = vect_init_vector (stmt, op1,
+ TREE_TYPE (vectype), NULL);
+ }
+ }
+ }
+ }
+ }
/* Supportable by target? */
if (!optab)
{
if (vect_print_dump_info (REPORT_DETAILS))
- fprintf (vect_dump, "no optab.");
+ fprintf (vect_dump, "no optab.");
return false;
}
vec_mode = TYPE_MODE (vectype);
if (icode == CODE_FOR_nothing)
{
if (vect_print_dump_info (REPORT_DETAILS))
- fprintf (vect_dump, "op not supported by target.");
+ fprintf (vect_dump, "op not supported by target.");
/* Check only during analysis. */
if (GET_MODE_SIZE (vec_mode) != UNITS_PER_WORD
- || (vf < vect_min_worthwhile_factor (code)
+ || (vf < vect_min_worthwhile_factor (code)
&& !vec_stmt))
return false;
if (vect_print_dump_info (REPORT_DETAILS))
- fprintf (vect_dump, "proceeding using word mode.");
+ fprintf (vect_dump, "proceeding using word mode.");
}
/* Worthwhile without SIMD support? Check only during analysis. */
&& !vec_stmt)
{
if (vect_print_dump_info (REPORT_DETAILS))
- fprintf (vect_dump, "not worthwhile without SIMD support.");
+ fprintf (vect_dump, "not worthwhile without SIMD support.");
return false;
}
if (!vec_stmt) /* transformation not required. */
{
- STMT_VINFO_TYPE (stmt_info) = op_vec_info_type;
+ STMT_VINFO_TYPE (stmt_info) = shift_vec_info_type;
if (vect_print_dump_info (REPORT_DETAILS))
- fprintf (vect_dump, "=== vectorizable_operation ===");
+ fprintf (vect_dump, "=== vectorizable_shift ===");
vect_model_simple_cost (stmt_info, ncopies, dt, NULL);
return true;
}
if (!slp_node)
{
vec_oprnds0 = VEC_alloc (tree, heap, 1);
- if (op_type == binary_op)
- vec_oprnds1 = VEC_alloc (tree, heap, 1);
+ vec_oprnds1 = VEC_alloc (tree, heap, 1);
}
else if (scalar_shift_arg)
vec_oprnds1 = VEC_alloc (tree, heap, slp_node->vec_stmts_size);
- /* In case the vectorization factor (VF) is bigger than the number
- of elements that we can fit in a vectype (nunits), we have to generate
- more than one vector stmt - i.e - we need to "unroll" the
- vector stmt by a factor VF/nunits. In doing so, we record a pointer
- from one copy of the vector stmt to the next, in the field
- STMT_VINFO_RELATED_STMT. This is necessary in order to allow following
- stages to find the correct vector defs to be used when vectorizing
- stmts that use the defs of the current stmt. The example below
- illustrates the vectorization process when VF=16 and nunits=4 (i.e.,
- we need to create 4 vectorized stmts):
-
- before vectorization:
- RELATED_STMT VEC_STMT
- S1: x = memref - -
- S2: z = x + 1 - -
-
- step 1: vectorize stmt S1 (done in vectorizable_load. See more details
- there):
- RELATED_STMT VEC_STMT
- VS1_0: vx0 = memref0 VS1_1 -
- VS1_1: vx1 = memref1 VS1_2 -
- VS1_2: vx2 = memref2 VS1_3 -
- VS1_3: vx3 = memref3 - -
- S1: x = load - VS1_0
- S2: z = x + 1 - -
-
- step2: vectorize stmt S2 (done here):
- To vectorize stmt S2 we first need to find the relevant vector
- def for the first operand 'x'. This is, as usual, obtained from
- the vector stmt recorded in the STMT_VINFO_VEC_STMT of the stmt
- that defines 'x' (S1). This way we find the stmt VS1_0, and the
- relevant vector def 'vx0'. Having found 'vx0' we can generate
- the vector stmt VS2_0, and as usual, record it in the
- STMT_VINFO_VEC_STMT of stmt S2.
- When creating the second copy (VS2_1), we obtain the relevant vector
- def from the vector stmt recorded in the STMT_VINFO_RELATED_STMT of
- stmt VS1_0. This way we find the stmt VS1_1 and the relevant
- vector def 'vx1'. Using 'vx1' we create stmt VS2_1 and record a
- pointer to it in the STMT_VINFO_RELATED_STMT of the vector stmt VS2_0.
- Similarly when creating stmts VS2_2 and VS2_3. This is the resulting
- chain of stmts and pointers:
- RELATED_STMT VEC_STMT
- VS1_0: vx0 = memref0 VS1_1 -
- VS1_1: vx1 = memref1 VS1_2 -
- VS1_2: vx2 = memref2 VS1_3 -
- VS1_3: vx3 = memref3 - -
- S1: x = load - VS1_0
- VS2_0: vz0 = vx0 + v1 VS2_1 -
- VS2_1: vz1 = vx1 + v1 VS2_2 -
- VS2_2: vz2 = vx2 + v1 VS2_3 -
- VS2_3: vz3 = vx3 + v1 - -
- S2: z = x + 1 - VS2_0 */
-
prev_stmt_info = NULL;
for (j = 0; j < ncopies; j++)
{
/* Handle uses. */
if (j == 0)
- {
- if (op_type == binary_op && scalar_shift_arg)
- {
- /* Vector shl and shr insn patterns can be defined with scalar
- operand 2 (shift operand). In this case, use constant or loop
- invariant op1 directly, without extending it to vector mode
- first. */
- optab_op2_mode = insn_data[icode].operand[2].mode;
- if (!VECTOR_MODE_P (optab_op2_mode))
- {
- if (vect_print_dump_info (REPORT_DETAILS))
- fprintf (vect_dump, "operand 1 using scalar mode.");
- vec_oprnd1 = op1;
- VEC_quick_push (tree, vec_oprnds1, vec_oprnd1);
- if (slp_node)
- {
- /* Store vec_oprnd1 for every vector stmt to be created
- for SLP_NODE. We check during the analysis that all
- the shift arguments are the same.
- TODO: Allow different constants for different vector
- stmts generated for an SLP instance. */
- for (k = 0; k < slp_node->vec_stmts_size - 1; k++)
- VEC_quick_push (tree, vec_oprnds1, vec_oprnd1);
- }
- }
- }
+ {
+ if (scalar_shift_arg)
+ {
+ /* Vector shl and shr insn patterns can be defined with scalar
+ operand 2 (shift operand). In this case, use constant or loop
+ invariant op1 directly, without extending it to vector mode
+ first. */
+ optab_op2_mode = insn_data[icode].operand[2].mode;
+ if (!VECTOR_MODE_P (optab_op2_mode))
+ {
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "operand 1 using scalar mode.");
+ vec_oprnd1 = op1;
+ VEC_quick_push (tree, vec_oprnds1, vec_oprnd1);
+ if (slp_node)
+ {
+ /* Store vec_oprnd1 for every vector stmt to be created
+ for SLP_NODE. We check during the analysis that all
+ the shift arguments are the same.
+ TODO: Allow different constants for different vector
+ stmts generated for an SLP instance. */
+ for (k = 0; k < slp_node->vec_stmts_size - 1; k++)
+ VEC_quick_push (tree, vec_oprnds1, vec_oprnd1);
+ }
+ }
+ }
/* vec_oprnd1 is available if operand 1 should be of a scalar-type
(a special case for certain kind of vector shifts); otherwise,
operand 1 should be of a vector type (the usual case). */
- if (op_type == binary_op && !vec_oprnd1)
+ if (vec_oprnd1)
+ vect_get_vec_defs (op0, NULL_TREE, stmt, &vec_oprnds0, NULL,
+ slp_node);
+ else
+ vect_get_vec_defs (op0, op1, stmt, &vec_oprnds0, &vec_oprnds1,
+ slp_node);
+ }
+ else
+ vect_get_vec_defs_for_stmt_copy (dt, &vec_oprnds0, &vec_oprnds1);
+
+ /* Arguments are ready. Create the new vector stmt. */
+ FOR_EACH_VEC_ELT (tree, vec_oprnds0, i, vop0)
+ {
+ vop1 = VEC_index (tree, vec_oprnds1, i);
+ new_stmt = gimple_build_assign_with_ops (code, vec_dest, vop0, vop1);
+ new_temp = make_ssa_name (vec_dest, new_stmt);
+ gimple_assign_set_lhs (new_stmt, new_temp);
+ vect_finish_stmt_generation (stmt, new_stmt, gsi);
+ if (slp_node)
+ VEC_quick_push (gimple, SLP_TREE_VEC_STMTS (slp_node), new_stmt);
+ }
+
+ if (slp_node)
+ continue;
+
+ if (j == 0)
+ STMT_VINFO_VEC_STMT (stmt_info) = *vec_stmt = new_stmt;
+ else
+ STMT_VINFO_RELATED_STMT (prev_stmt_info) = new_stmt;
+ prev_stmt_info = vinfo_for_stmt (new_stmt);
+ }
+
+ VEC_free (tree, heap, vec_oprnds0);
+ VEC_free (tree, heap, vec_oprnds1);
+
+ return true;
+}
+
+
+/* Function vectorizable_operation.
+
+ Check if STMT performs a binary, unary or ternary operation that can
+ be vectorized.
+ If VEC_STMT is also passed, vectorize the STMT: create a vectorized
+ stmt to replace it, put it in VEC_STMT, and insert it at BSI.
+ Return FALSE if not a vectorizable STMT, TRUE otherwise. */
+
+static bool
+vectorizable_operation (gimple stmt, gimple_stmt_iterator *gsi,
+ gimple *vec_stmt, slp_tree slp_node)
+{
+ tree vec_dest;
+ tree scalar_dest;
+ tree op0, op1 = NULL_TREE, op2 = NULL_TREE;
+ stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
+ tree vectype;
+ loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
+ enum tree_code code;
+ enum machine_mode vec_mode;
+ tree new_temp;
+ int op_type;
+ optab optab;
+ int icode;
+ tree def;
+ gimple def_stmt;
+ enum vect_def_type dt[3]
+ = {vect_unknown_def_type, vect_unknown_def_type, vect_unknown_def_type};
+ gimple new_stmt = NULL;
+ stmt_vec_info prev_stmt_info;
+ int nunits_in;
+ int nunits_out;
+ tree vectype_out;
+ int ncopies;
+ int j, i;
+ VEC(tree,heap) *vec_oprnds0 = NULL, *vec_oprnds1 = NULL, *vec_oprnds2 = NULL;
+ tree vop0, vop1, vop2;
+ bb_vec_info bb_vinfo = STMT_VINFO_BB_VINFO (stmt_info);
+ int vf;
+
+ if (!STMT_VINFO_RELEVANT_P (stmt_info) && !bb_vinfo)
+ return false;
+
+ if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def)
+ return false;
+
+ /* Is STMT a vectorizable binary/unary operation? */
+ if (!is_gimple_assign (stmt))
+ return false;
+
+ if (TREE_CODE (gimple_assign_lhs (stmt)) != SSA_NAME)
+ return false;
+
+ code = gimple_assign_rhs_code (stmt);
+
+ /* For pointer addition, we should use the normal plus for
+ the vector addition. */
+ if (code == POINTER_PLUS_EXPR)
+ code = PLUS_EXPR;
+
+ /* Support only unary or binary operations. */
+ op_type = TREE_CODE_LENGTH (code);
+ if (op_type != unary_op && op_type != binary_op && op_type != ternary_op)
+ {
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "num. args = %d (not unary/binary/ternary op).",
+ op_type);
+ return false;
+ }
+
+ scalar_dest = gimple_assign_lhs (stmt);
+ vectype_out = STMT_VINFO_VECTYPE (stmt_info);
+
+ /* Most operations cannot handle bit-precision types without extra
+ truncations. */
+ if ((TYPE_PRECISION (TREE_TYPE (scalar_dest))
+ != GET_MODE_PRECISION (TYPE_MODE (TREE_TYPE (scalar_dest))))
+ /* Exception are bitwise binary operations. */
+ && code != BIT_IOR_EXPR
+ && code != BIT_XOR_EXPR
+ && code != BIT_AND_EXPR)
+ {
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "bit-precision arithmetic not supported.");
+ return false;
+ }
+
+ op0 = gimple_assign_rhs1 (stmt);
+ if (!vect_is_simple_use_1 (op0, loop_vinfo, bb_vinfo,
+ &def_stmt, &def, &dt[0], &vectype))
+ {
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "use not simple.");
+ return false;
+ }
+ /* If op0 is an external or constant def use a vector type with
+ the same size as the output vector type. */
+ if (!vectype)
+ vectype = get_same_sized_vectype (TREE_TYPE (op0), vectype_out);
+ if (vec_stmt)
+ gcc_assert (vectype);
+ if (!vectype)
+ {
+ if (vect_print_dump_info (REPORT_DETAILS))
+ {
+ fprintf (vect_dump, "no vectype for scalar type ");
+ print_generic_expr (vect_dump, TREE_TYPE (op0), TDF_SLIM);
+ }
+
+ return false;
+ }
+
+ nunits_out = TYPE_VECTOR_SUBPARTS (vectype_out);
+ nunits_in = TYPE_VECTOR_SUBPARTS (vectype);
+ if (nunits_out != nunits_in)
+ return false;
+
+ if (op_type == binary_op || op_type == ternary_op)
+ {
+ op1 = gimple_assign_rhs2 (stmt);
+ if (!vect_is_simple_use (op1, loop_vinfo, bb_vinfo, &def_stmt, &def,
+ &dt[1]))
+ {
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "use not simple.");
+ return false;
+ }
+ }
+ if (op_type == ternary_op)
+ {
+ op2 = gimple_assign_rhs3 (stmt);
+ if (!vect_is_simple_use (op2, loop_vinfo, bb_vinfo, &def_stmt, &def,
+ &dt[2]))
+ {
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "use not simple.");
+ return false;
+ }
+ }
+
+ if (loop_vinfo)
+ vf = LOOP_VINFO_VECT_FACTOR (loop_vinfo);
+ else
+ vf = 1;
+
+ /* Multiple types in SLP are handled by creating the appropriate number of
+ vectorized stmts for each SLP node. Hence, NCOPIES is always 1 in
+ case of SLP. */
+ if (slp_node || PURE_SLP_STMT (stmt_info))
+ ncopies = 1;
+ else
+ ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits_in;
+
+ gcc_assert (ncopies >= 1);
+
+ /* Shifts are handled in vectorizable_shift (). */
+ if (code == LSHIFT_EXPR || code == RSHIFT_EXPR || code == LROTATE_EXPR
+ || code == RROTATE_EXPR)
+ return false;
+
+ optab = optab_for_tree_code (code, vectype, optab_default);
+
+ /* Supportable by target? */
+ if (!optab)
+ {
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "no optab.");
+ return false;
+ }
+ vec_mode = TYPE_MODE (vectype);
+ icode = (int) optab_handler (optab, vec_mode);
+ if (icode == CODE_FOR_nothing)
+ {
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "op not supported by target.");
+ /* Check only during analysis. */
+ if (GET_MODE_SIZE (vec_mode) != UNITS_PER_WORD
+ || (vf < vect_min_worthwhile_factor (code)
+ && !vec_stmt))
+ return false;
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "proceeding using word mode.");
+ }
+
+ /* Worthwhile without SIMD support? Check only during analysis. */
+ if (!VECTOR_MODE_P (TYPE_MODE (vectype))
+ && vf < vect_min_worthwhile_factor (code)
+ && !vec_stmt)
+ {
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "not worthwhile without SIMD support.");
+ return false;
+ }
+
+ if (!vec_stmt) /* transformation not required. */
+ {
+ STMT_VINFO_TYPE (stmt_info) = op_vec_info_type;
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "=== vectorizable_operation ===");
+ vect_model_simple_cost (stmt_info, ncopies, dt, NULL);
+ return true;
+ }
+
+ /** Transform. **/
+
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "transform binary/unary operation.");
+
+ /* Handle def. */
+ vec_dest = vect_create_destination_var (scalar_dest, vectype);
+
+ /* Allocate VECs for vector operands. In case of SLP, vector operands are
+ created in the previous stages of the recursion, so no allocation is
+ needed, except for the case of shift with scalar shift argument. In that
+ case we store the scalar operand in VEC_OPRNDS1 for every vector stmt to
+ be created to vectorize the SLP group, i.e., SLP_NODE->VEC_STMTS_SIZE.
+ In case of loop-based vectorization we allocate VECs of size 1. We
+ allocate VEC_OPRNDS1 only in case of binary operation. */
+ if (!slp_node)
+ {
+ vec_oprnds0 = VEC_alloc (tree, heap, 1);
+ if (op_type == binary_op || op_type == ternary_op)
+ vec_oprnds1 = VEC_alloc (tree, heap, 1);
+ if (op_type == ternary_op)
+ vec_oprnds2 = VEC_alloc (tree, heap, 1);
+ }
+
+ /* In case the vectorization factor (VF) is bigger than the number
+ of elements that we can fit in a vectype (nunits), we have to generate
+ more than one vector stmt - i.e - we need to "unroll" the
+ vector stmt by a factor VF/nunits. In doing so, we record a pointer
+ from one copy of the vector stmt to the next, in the field
+ STMT_VINFO_RELATED_STMT. This is necessary in order to allow following
+ stages to find the correct vector defs to be used when vectorizing
+ stmts that use the defs of the current stmt. The example below
+ illustrates the vectorization process when VF=16 and nunits=4 (i.e.,
+ we need to create 4 vectorized stmts):
+
+ before vectorization:
+ RELATED_STMT VEC_STMT
+ S1: x = memref - -
+ S2: z = x + 1 - -
+
+ step 1: vectorize stmt S1 (done in vectorizable_load. See more details
+ there):
+ RELATED_STMT VEC_STMT
+ VS1_0: vx0 = memref0 VS1_1 -
+ VS1_1: vx1 = memref1 VS1_2 -
+ VS1_2: vx2 = memref2 VS1_3 -
+ VS1_3: vx3 = memref3 - -
+ S1: x = load - VS1_0
+ S2: z = x + 1 - -
+
+ step2: vectorize stmt S2 (done here):
+ To vectorize stmt S2 we first need to find the relevant vector
+ def for the first operand 'x'. This is, as usual, obtained from
+ the vector stmt recorded in the STMT_VINFO_VEC_STMT of the stmt
+ that defines 'x' (S1). This way we find the stmt VS1_0, and the
+ relevant vector def 'vx0'. Having found 'vx0' we can generate
+ the vector stmt VS2_0, and as usual, record it in the
+ STMT_VINFO_VEC_STMT of stmt S2.
+ When creating the second copy (VS2_1), we obtain the relevant vector
+ def from the vector stmt recorded in the STMT_VINFO_RELATED_STMT of
+ stmt VS1_0. This way we find the stmt VS1_1 and the relevant
+ vector def 'vx1'. Using 'vx1' we create stmt VS2_1 and record a
+ pointer to it in the STMT_VINFO_RELATED_STMT of the vector stmt VS2_0.
+ Similarly when creating stmts VS2_2 and VS2_3. This is the resulting
+ chain of stmts and pointers:
+ RELATED_STMT VEC_STMT
+ VS1_0: vx0 = memref0 VS1_1 -
+ VS1_1: vx1 = memref1 VS1_2 -
+ VS1_2: vx2 = memref2 VS1_3 -
+ VS1_3: vx3 = memref3 - -
+ S1: x = load - VS1_0
+ VS2_0: vz0 = vx0 + v1 VS2_1 -
+ VS2_1: vz1 = vx1 + v1 VS2_2 -
+ VS2_2: vz2 = vx2 + v1 VS2_3 -
+ VS2_3: vz3 = vx3 + v1 - -
+ S2: z = x + 1 - VS2_0 */
+
+ prev_stmt_info = NULL;
+ for (j = 0; j < ncopies; j++)
+ {
+ /* Handle uses. */
+ if (j == 0)
+ {
+ if (op_type == binary_op || op_type == ternary_op)
vect_get_vec_defs (op0, op1, stmt, &vec_oprnds0, &vec_oprnds1,
slp_node);
else
vect_get_vec_defs (op0, NULL_TREE, stmt, &vec_oprnds0, NULL,
slp_node);
+ if (op_type == ternary_op)
+ {
+ vec_oprnds2 = VEC_alloc (tree, heap, 1);
+ VEC_quick_push (tree, vec_oprnds2,
+ vect_get_vec_def_for_operand (op2, stmt, NULL));
+ }
}
else
- vect_get_vec_defs_for_stmt_copy (dt, &vec_oprnds0, &vec_oprnds1);
+ {
+ vect_get_vec_defs_for_stmt_copy (dt, &vec_oprnds0, &vec_oprnds1);
+ if (op_type == ternary_op)
+ {
+ tree vec_oprnd = VEC_pop (tree, vec_oprnds2);
+ VEC_quick_push (tree, vec_oprnds2,
+ vect_get_vec_def_for_stmt_copy (dt[2],
+ vec_oprnd));
+ }
+ }
- /* Arguments are ready. Create the new vector stmt. */
+ /* Arguments are ready. Create the new vector stmt. */
FOR_EACH_VEC_ELT (tree, vec_oprnds0, i, vop0)
{
- vop1 = ((op_type == binary_op)
- ? VEC_index (tree, vec_oprnds1, i) : NULL);
- new_stmt = gimple_build_assign_with_ops (code, vec_dest, vop0, vop1);
+ vop1 = ((op_type == binary_op || op_type == ternary_op)
+ ? VEC_index (tree, vec_oprnds1, i) : NULL_TREE);
+ vop2 = ((op_type == ternary_op)
+ ? VEC_index (tree, vec_oprnds2, i) : NULL_TREE);
+ new_stmt = gimple_build_assign_with_ops3 (code, vec_dest,
+ vop0, vop1, vop2);
new_temp = make_ssa_name (vec_dest, new_stmt);
gimple_assign_set_lhs (new_stmt, new_temp);
vect_finish_stmt_generation (stmt, new_stmt, gsi);
VEC_free (tree, heap, vec_oprnds0);
if (vec_oprnds1)
VEC_free (tree, heap, vec_oprnds1);
+ if (vec_oprnds2)
+ VEC_free (tree, heap, vec_oprnds2);
return true;
}
VEC (tree, heap) *vec_oprnds0 = NULL;
VEC (tree, heap) *vec_dsts = NULL, *interm_types = NULL, *tmp_vec_dsts = NULL;
tree last_oprnd, intermediate_type;
+ bb_vec_info bb_vinfo = STMT_VINFO_BB_VINFO (stmt_info);
- /* FORNOW: not supported by basic block SLP vectorization. */
- gcc_assert (loop_vinfo);
-
- if (!STMT_VINFO_RELEVANT_P (stmt_info))
+ if (!STMT_VINFO_RELEVANT_P (stmt_info) && !bb_vinfo)
return false;
if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def)
if (! ((INTEGRAL_TYPE_P (TREE_TYPE (scalar_dest))
&& INTEGRAL_TYPE_P (TREE_TYPE (op0)))
|| (SCALAR_FLOAT_TYPE_P (TREE_TYPE (scalar_dest))
- && SCALAR_FLOAT_TYPE_P (TREE_TYPE (op0))
- && CONVERT_EXPR_CODE_P (code))))
+ && SCALAR_FLOAT_TYPE_P (TREE_TYPE (op0)))))
return false;
- if (!vect_is_simple_use_1 (op0, loop_vinfo, NULL,
+
+ if (INTEGRAL_TYPE_P (TREE_TYPE (scalar_dest))
+ && ((TYPE_PRECISION (TREE_TYPE (scalar_dest))
+ != GET_MODE_PRECISION (TYPE_MODE (TREE_TYPE (scalar_dest))))
+ || ((TYPE_PRECISION (TREE_TYPE (op0))
+ != GET_MODE_PRECISION (TYPE_MODE (TREE_TYPE (op0)))))))
+ {
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "type demotion to/from bit-precision unsupported.");
+ return false;
+ }
+
+ if (!vect_is_simple_use_1 (op0, loop_vinfo, bb_vinfo,
&def_stmt, &def, &dt[0], &vectype_in))
{
if (vect_print_dump_info (REPORT_DETAILS))
/* Multiple types in SLP are handled by creating the appropriate number of
vectorized stmts for each SLP node. Hence, NCOPIES is always 1 in
case of SLP. */
- if (slp_node)
+ if (slp_node || PURE_SLP_STMT (stmt_info))
ncopies = 1;
else
ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits_out;
{
/* Handle uses. */
if (slp_node)
- vect_get_slp_defs (slp_node, &vec_oprnds0, NULL, -1);
+ vect_get_slp_defs (op0, NULL_TREE, slp_node, &vec_oprnds0, NULL, -1);
else
{
VEC_free (tree, heap, vec_oprnds0);
vect_pow2 (multi_step_cvt) - 1);
}
- /* Arguments are ready. Create the new vector stmts. */
+ /* Arguments are ready. Create the new vector stmts. */
tmp_vec_dsts = VEC_copy (tree, heap, vec_dsts);
vect_create_vectorized_demotion_stmts (&vec_oprnds0,
multi_step_cvt, stmt, tmp_vec_dsts,
int multi_step_cvt = 0;
VEC (tree, heap) *vec_oprnds0 = NULL, *vec_oprnds1 = NULL;
VEC (tree, heap) *vec_dsts = NULL, *interm_types = NULL, *tmp_vec_dsts = NULL;
+ bb_vec_info bb_vinfo = STMT_VINFO_BB_VINFO (stmt_info);
+ unsigned int k;
- /* FORNOW: not supported by basic block SLP vectorization. */
- gcc_assert (loop_vinfo);
-
- if (!STMT_VINFO_RELEVANT_P (stmt_info))
+ if (!STMT_VINFO_RELEVANT_P (stmt_info) && !bb_vinfo)
return false;
if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def)
code = gimple_assign_rhs_code (stmt);
if (!CONVERT_EXPR_CODE_P (code)
- && code != WIDEN_MULT_EXPR)
+ && code != WIDEN_MULT_EXPR
+ && code != WIDEN_LSHIFT_EXPR)
return false;
scalar_dest = gimple_assign_lhs (stmt);
&& SCALAR_FLOAT_TYPE_P (TREE_TYPE (op0))
&& CONVERT_EXPR_CODE_P (code))))
return false;
- if (!vect_is_simple_use_1 (op0, loop_vinfo, NULL,
+
+ if (INTEGRAL_TYPE_P (TREE_TYPE (scalar_dest))
+ && ((TYPE_PRECISION (TREE_TYPE (scalar_dest))
+ != GET_MODE_PRECISION (TYPE_MODE (TREE_TYPE (scalar_dest))))
+ || ((TYPE_PRECISION (TREE_TYPE (op0))
+ != GET_MODE_PRECISION (TYPE_MODE (TREE_TYPE (op0)))))))
+ {
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "type promotion to/from bit-precision "
+ "unsupported.");
+ return false;
+ }
+
+ if (!vect_is_simple_use_1 (op0, 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;
}
+
+ op_type = TREE_CODE_LENGTH (code);
+ if (op_type == binary_op)
+ {
+ bool ok;
+
+ op1 = gimple_assign_rhs2 (stmt);
+ if (code == WIDEN_MULT_EXPR || code == WIDEN_LSHIFT_EXPR)
+ {
+ /* For WIDEN_MULT_EXPR, if OP0 is a constant, use the type of
+ OP1. */
+ if (CONSTANT_CLASS_P (op0))
+ ok = vect_is_simple_use_1 (op1, loop_vinfo, NULL,
+ &def_stmt, &def, &dt[1], &vectype_in);
+ else
+ ok = vect_is_simple_use (op1, loop_vinfo, NULL, &def_stmt, &def,
+ &dt[1]);
+
+ if (!ok)
+ {
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "use not simple.");
+ return false;
+ }
+ }
+ }
+
/* If op0 is an external or constant def use a vector type with
the same size as the output vector type. */
if (!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
case of SLP. */
- if (slp_node)
+ if (slp_node || PURE_SLP_STMT (stmt_info))
ncopies = 1;
else
ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits_in;
gcc_assert (ncopies >= 1);
- op_type = TREE_CODE_LENGTH (code);
- if (op_type == binary_op)
- {
- op1 = gimple_assign_rhs2 (stmt);
- if (!vect_is_simple_use (op1, loop_vinfo, NULL, &def_stmt, &def, &dt[1]))
- {
- if (vect_print_dump_info (REPORT_DETAILS))
- fprintf (vect_dump, "use not simple.");
- return false;
- }
- }
-
/* Supportable by target? */
if (!supportable_widening_operation (code, stmt, vectype_out, vectype_in,
&decl1, &decl2, &code1, &code2,
fprintf (vect_dump, "transform type promotion operation. ncopies = %d.",
ncopies);
+ if (code == WIDEN_MULT_EXPR || code == WIDEN_LSHIFT_EXPR)
+ {
+ if (CONSTANT_CLASS_P (op0))
+ op0 = fold_convert (TREE_TYPE (op1), op0);
+ else if (CONSTANT_CLASS_P (op1))
+ op1 = fold_convert (TREE_TYPE (op0), op1);
+ }
+
/* Handle def. */
/* In case of multi-step promotion, we first generate promotion operations
to the intermediate types, and then from that types to the final one.
if (op_type == binary_op)
vec_oprnds1 = VEC_alloc (tree, heap, 1);
}
+ else if (code == WIDEN_LSHIFT_EXPR)
+ vec_oprnds1 = VEC_alloc (tree, heap, slp_node->vec_stmts_size);
/* In case the vectorization factor (VF) is bigger than the number
of elements that we can fit in a vectype (nunits), we have to generate
if (j == 0)
{
if (slp_node)
- vect_get_slp_defs (slp_node, &vec_oprnds0, &vec_oprnds1, -1);
- else
+ {
+ if (code == WIDEN_LSHIFT_EXPR)
+ {
+ vec_oprnd1 = op1;
+ /* Store vec_oprnd1 for every vector stmt to be created
+ for SLP_NODE. We check during the analysis that all
+ the shift arguments are the same. */
+ for (k = 0; k < slp_node->vec_stmts_size - 1; k++)
+ VEC_quick_push (tree, vec_oprnds1, vec_oprnd1);
+
+ vect_get_slp_defs (op0, NULL_TREE, slp_node, &vec_oprnds0, NULL,
+ -1);
+ }
+ else
+ vect_get_slp_defs (op0, op1, slp_node, &vec_oprnds0,
+ &vec_oprnds1, -1);
+ }
+ else
{
vec_oprnd0 = vect_get_vec_def_for_operand (op0, stmt, NULL);
VEC_quick_push (tree, vec_oprnds0, vec_oprnd0);
if (op_type == binary_op)
{
- vec_oprnd1 = vect_get_vec_def_for_operand (op1, stmt, NULL);
+ if (code == WIDEN_LSHIFT_EXPR)
+ vec_oprnd1 = op1;
+ else
+ vec_oprnd1 = vect_get_vec_def_for_operand (op1, stmt, NULL);
VEC_quick_push (tree, vec_oprnds1, vec_oprnd1);
}
}
VEC_replace (tree, vec_oprnds0, 0, vec_oprnd0);
if (op_type == binary_op)
{
- vec_oprnd1 = vect_get_vec_def_for_stmt_copy (dt[1], vec_oprnd1);
+ if (code == WIDEN_LSHIFT_EXPR)
+ vec_oprnd1 = op1;
+ else
+ vec_oprnd1 = vect_get_vec_def_for_stmt_copy (dt[1], vec_oprnd1);
VEC_replace (tree, vec_oprnds1, 0, vec_oprnd1);
}
}
- /* Arguments are ready. Create the new vector stmts. */
+ /* Arguments are ready. Create the new vector stmts. */
tmp_vec_dsts = VEC_copy (tree, heap, vec_dsts);
vect_create_vectorized_promotion_stmts (&vec_oprnds0, &vec_oprnds1,
multi_step_cvt, stmt,
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
struct data_reference *dr = STMT_VINFO_DATA_REF (stmt_info), *first_dr = NULL;
tree vectype = STMT_VINFO_VECTYPE (stmt_info);
+ tree elem_type;
loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
struct loop *loop = NULL;
enum machine_mode vec_mode;
int j;
gimple next_stmt, first_stmt = NULL;
bool strided_store = false;
+ bool store_lanes_p = false;
unsigned int group_size, i;
VEC(tree,heap) *dr_chain = NULL, *oprnds = NULL, *result_chain = NULL;
bool inv_p;
bool slp = (slp_node != NULL);
unsigned int vec_num;
bb_vec_info bb_vinfo = STMT_VINFO_BB_VINFO (stmt_info);
+ tree aggr_type;
if (loop_vinfo)
loop = LOOP_VINFO_LOOP (loop_vinfo);
/* Multiple types in SLP are handled by creating the appropriate number of
vectorized stmts for each SLP node. Hence, NCOPIES is always 1 in
case of SLP. */
- if (slp)
+ if (slp || PURE_SLP_STMT (stmt_info))
ncopies = 1;
else
ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits;
return false;
scalar_dest = gimple_assign_lhs (stmt);
+ if (TREE_CODE (scalar_dest) == VIEW_CONVERT_EXPR
+ && is_pattern_stmt_p (stmt_info))
+ scalar_dest = TREE_OPERAND (scalar_dest, 0);
if (TREE_CODE (scalar_dest) != ARRAY_REF
&& TREE_CODE (scalar_dest) != INDIRECT_REF
&& TREE_CODE (scalar_dest) != COMPONENT_REF
return false;
}
- /* The scalar rhs type needs to be trivially convertible to the vector
- component type. This should always be the case. */
- if (!useless_type_conversion_p (TREE_TYPE (vectype), TREE_TYPE (op)))
- {
- if (vect_print_dump_info (REPORT_DETAILS))
- fprintf (vect_dump, "??? operands of different types");
- return false;
- }
-
+ elem_type = TREE_TYPE (vectype);
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, vec_mode) == CODE_FOR_nothing)
if (!STMT_VINFO_DATA_REF (stmt_info))
return false;
+ if (tree_int_cst_compare (DR_STEP (dr), size_zero_node) < 0)
+ {
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "negative step for store.");
+ return false;
+ }
+
if (STMT_VINFO_STRIDED_ACCESS (stmt_info))
{
strided_store = true;
- first_stmt = DR_GROUP_FIRST_DR (stmt_info);
- if (!vect_strided_store_supported (vectype)
- && !PURE_SLP_STMT (stmt_info) && !slp)
- return false;
+ first_stmt = GROUP_FIRST_ELEMENT (stmt_info);
+ if (!slp && !PURE_SLP_STMT (stmt_info))
+ {
+ group_size = GROUP_SIZE (vinfo_for_stmt (first_stmt));
+ if (vect_store_lanes_supported (vectype, group_size))
+ store_lanes_p = true;
+ else if (!vect_strided_store_supported (vectype, group_size))
+ return false;
+ }
if (first_stmt == stmt)
{
/* STMT is the leader of the group. Check the operands of all the
stmts of the group. */
- next_stmt = DR_GROUP_NEXT_DR (stmt_info);
+ next_stmt = GROUP_NEXT_ELEMENT (stmt_info);
while (next_stmt)
{
gcc_assert (gimple_assign_single_p (next_stmt));
fprintf (vect_dump, "use not simple.");
return false;
}
- next_stmt = DR_GROUP_NEXT_DR (vinfo_for_stmt (next_stmt));
+ next_stmt = GROUP_NEXT_ELEMENT (vinfo_for_stmt (next_stmt));
}
}
}
if (!vec_stmt) /* transformation not required. */
{
STMT_VINFO_TYPE (stmt_info) = store_vec_info_type;
- vect_model_store_cost (stmt_info, ncopies, dt, NULL);
+ vect_model_store_cost (stmt_info, ncopies, store_lanes_p, dt, NULL);
return true;
}
if (strided_store)
{
first_dr = STMT_VINFO_DATA_REF (vinfo_for_stmt (first_stmt));
- group_size = DR_GROUP_SIZE (vinfo_for_stmt (first_stmt));
+ group_size = GROUP_SIZE (vinfo_for_stmt (first_stmt));
- DR_GROUP_STORE_COUNT (vinfo_for_stmt (first_stmt))++;
+ GROUP_STORE_COUNT (vinfo_for_stmt (first_stmt))++;
/* FORNOW */
gcc_assert (!loop || !nested_in_vect_loop_p (loop, stmt));
/* We vectorize all the stmts of the interleaving group when we
reach the last stmt in the group. */
- if (DR_GROUP_STORE_COUNT (vinfo_for_stmt (first_stmt))
- < DR_GROUP_SIZE (vinfo_for_stmt (first_stmt))
+ if (GROUP_STORE_COUNT (vinfo_for_stmt (first_stmt))
+ < GROUP_SIZE (vinfo_for_stmt (first_stmt))
&& !slp)
{
*vec_stmt = NULL;
alignment_support_scheme = vect_supportable_dr_alignment (first_dr, false);
gcc_assert (alignment_support_scheme);
+ /* Targets with store-lane instructions must not require explicit
+ realignment. */
+ gcc_assert (!store_lanes_p
+ || alignment_support_scheme == dr_aligned
+ || alignment_support_scheme == dr_unaligned_supported);
+
+ if (store_lanes_p)
+ aggr_type = build_array_type_nelts (elem_type, vec_num * nunits);
+ else
+ aggr_type = vectype;
/* In case the vectorization factor (VF) is bigger than the number
of elements that we can fit in a vectype (nunits), we have to generate
if (slp)
{
/* Get vectorized arguments for SLP_NODE. */
- vect_get_slp_defs (slp_node, &vec_oprnds, NULL, -1);
+ vect_get_slp_defs (NULL_TREE, NULL_TREE, slp_node, &vec_oprnds,
+ NULL, -1);
vec_oprnd = VEC_index (tree, vec_oprnds, 0);
}
NULL);
VEC_quick_push(tree, dr_chain, vec_oprnd);
VEC_quick_push(tree, oprnds, vec_oprnd);
- next_stmt = DR_GROUP_NEXT_DR (vinfo_for_stmt (next_stmt));
+ next_stmt = GROUP_NEXT_ELEMENT (vinfo_for_stmt (next_stmt));
}
}
/* We should have catched mismatched types earlier. */
gcc_assert (useless_type_conversion_p (vectype,
TREE_TYPE (vec_oprnd)));
- dataref_ptr = vect_create_data_ref_ptr (first_stmt, NULL, NULL_TREE,
- &dummy, &ptr_incr, false,
- &inv_p);
+ dataref_ptr = vect_create_data_ref_ptr (first_stmt, aggr_type, NULL,
+ NULL_TREE, &dummy, gsi,
+ &ptr_incr, false, &inv_p);
gcc_assert (bb_vinfo || !inv_p);
}
else
VEC_replace(tree, dr_chain, i, vec_oprnd);
VEC_replace(tree, oprnds, i, vec_oprnd);
}
- dataref_ptr =
- bump_vector_ptr (dataref_ptr, ptr_incr, gsi, stmt, NULL_TREE);
+ dataref_ptr = bump_vector_ptr (dataref_ptr, ptr_incr, gsi, stmt,
+ TYPE_SIZE_UNIT (aggr_type));
}
- if (strided_store)
+ if (store_lanes_p)
{
- result_chain = VEC_alloc (tree, heap, group_size);
- /* Permute. */
- if (!vect_permute_store_chain (dr_chain, group_size, stmt, gsi,
- &result_chain))
- return false;
- }
+ tree vec_array;
- next_stmt = first_stmt;
- for (i = 0; i < vec_num; i++)
- {
- struct ptr_info_def *pi;
-
- if (i > 0)
- /* Bump the vector pointer. */
- dataref_ptr = bump_vector_ptr (dataref_ptr, ptr_incr, gsi, stmt,
- NULL_TREE);
-
- if (slp)
- vec_oprnd = VEC_index (tree, vec_oprnds, i);
- else if (strided_store)
- /* For strided stores vectorized defs are interleaved in
- vect_permute_store_chain(). */
- vec_oprnd = VEC_index (tree, result_chain, i);
-
- data_ref = build2 (MEM_REF, TREE_TYPE (vec_oprnd), dataref_ptr,
- build_int_cst (reference_alias_ptr_type
- (DR_REF (first_dr)), 0));
- pi = get_ptr_info (dataref_ptr);
- pi->align = TYPE_ALIGN_UNIT (vectype);
- if (aligned_access_p (first_dr))
- pi->misalign = 0;
- else if (DR_MISALIGNMENT (first_dr) == -1)
- {
- TREE_TYPE (data_ref)
- = build_aligned_type (TREE_TYPE (data_ref),
- TYPE_ALIGN (TREE_TYPE (vectype)));
- pi->align = TYPE_ALIGN_UNIT (TREE_TYPE (vectype));
- pi->misalign = 0;
- }
- else
+ /* Combine all the vectors into an array. */
+ vec_array = create_vector_array (vectype, vec_num);
+ for (i = 0; i < vec_num; i++)
{
- TREE_TYPE (data_ref)
- = build_aligned_type (TREE_TYPE (data_ref),
- TYPE_ALIGN (TREE_TYPE (vectype)));
- pi->misalign = DR_MISALIGNMENT (first_dr);
+ vec_oprnd = VEC_index (tree, dr_chain, i);
+ write_vector_array (stmt, gsi, vec_oprnd, vec_array, i);
}
- /* Arguments are ready. Create the new vector stmt. */
- new_stmt = gimple_build_assign (data_ref, vec_oprnd);
+ /* Emit:
+ MEM_REF[...all elements...] = STORE_LANES (VEC_ARRAY). */
+ data_ref = create_array_ref (aggr_type, dataref_ptr, first_dr);
+ new_stmt = gimple_build_call_internal (IFN_STORE_LANES, 1, vec_array);
+ gimple_call_set_lhs (new_stmt, data_ref);
vect_finish_stmt_generation (stmt, new_stmt, gsi);
mark_symbols_for_renaming (new_stmt);
+ }
+ else
+ {
+ new_stmt = NULL;
+ if (strided_store)
+ {
+ result_chain = VEC_alloc (tree, heap, group_size);
+ /* Permute. */
+ vect_permute_store_chain (dr_chain, group_size, stmt, gsi,
+ &result_chain);
+ }
- if (slp)
- continue;
+ next_stmt = first_stmt;
+ for (i = 0; i < vec_num; i++)
+ {
+ struct ptr_info_def *pi;
+
+ if (i > 0)
+ /* Bump the vector pointer. */
+ dataref_ptr = bump_vector_ptr (dataref_ptr, ptr_incr, gsi,
+ stmt, NULL_TREE);
+
+ if (slp)
+ vec_oprnd = VEC_index (tree, vec_oprnds, i);
+ else if (strided_store)
+ /* For strided stores vectorized defs are interleaved in
+ vect_permute_store_chain(). */
+ vec_oprnd = VEC_index (tree, result_chain, i);
+
+ data_ref = build2 (MEM_REF, TREE_TYPE (vec_oprnd), dataref_ptr,
+ build_int_cst (reference_alias_ptr_type
+ (DR_REF (first_dr)), 0));
+ pi = get_ptr_info (dataref_ptr);
+ pi->align = TYPE_ALIGN_UNIT (vectype);
+ if (aligned_access_p (first_dr))
+ pi->misalign = 0;
+ else if (DR_MISALIGNMENT (first_dr) == -1)
+ {
+ TREE_TYPE (data_ref)
+ = build_aligned_type (TREE_TYPE (data_ref),
+ TYPE_ALIGN (elem_type));
+ pi->align = TYPE_ALIGN_UNIT (elem_type);
+ pi->misalign = 0;
+ }
+ else
+ {
+ TREE_TYPE (data_ref)
+ = build_aligned_type (TREE_TYPE (data_ref),
+ TYPE_ALIGN (elem_type));
+ pi->misalign = DR_MISALIGNMENT (first_dr);
+ }
+
+ /* 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);
+ mark_symbols_for_renaming (new_stmt);
- if (j == 0)
- STMT_VINFO_VEC_STMT (stmt_info) = *vec_stmt = new_stmt;
+ if (slp)
+ continue;
+
+ next_stmt = GROUP_NEXT_ELEMENT (vinfo_for_stmt (next_stmt));
+ if (!next_stmt)
+ break;
+ }
+ }
+ if (!slp)
+ {
+ if (j == 0)
+ STMT_VINFO_VEC_STMT (stmt_info) = *vec_stmt = new_stmt;
else
STMT_VINFO_RELATED_STMT (prev_stmt_info) = new_stmt;
-
prev_stmt_info = vinfo_for_stmt (new_stmt);
- next_stmt = DR_GROUP_NEXT_DR (vinfo_for_stmt (next_stmt));
- if (!next_stmt)
- break;
}
}
return true;
}
+/* Given a vector type VECTYPE returns a builtin DECL to be used
+ for vector permutation and returns the mask that implements
+ reversal of the vector elements. If that is impossible to do,
+ returns NULL. */
+
+static tree
+perm_mask_for_reverse (tree vectype)
+{
+ tree mask_elt_type, mask_type, mask_vec;
+ int i, nunits;
+ unsigned char *sel;
+
+ nunits = TYPE_VECTOR_SUBPARTS (vectype);
+ sel = XALLOCAVEC (unsigned char, nunits);
+
+ for (i = 0; i < nunits; ++i)
+ sel[i] = nunits - 1 - i;
+
+ if (!can_vec_perm_p (TYPE_MODE (vectype), false, sel))
+ return NULL;
+
+ mask_elt_type
+ = lang_hooks.types.type_for_size
+ (TREE_INT_CST_LOW (TYPE_SIZE (TREE_TYPE (vectype))), 1);
+ mask_type = get_vectype_for_scalar_type (mask_elt_type);
+
+ mask_vec = NULL;
+ for (i = 0; i < nunits; i++)
+ mask_vec = tree_cons (NULL, build_int_cst (mask_elt_type, i), mask_vec);
+ mask_vec = build_vector (mask_type, mask_vec);
+
+ return mask_vec;
+}
+
+/* Given a vector variable X, that was generated for the scalar LHS of
+ STMT, generate instructions to reverse the vector elements of X,
+ insert them a *GSI and return the permuted vector variable. */
+
+static tree
+reverse_vec_elements (tree x, gimple stmt, gimple_stmt_iterator *gsi)
+{
+ tree vectype = TREE_TYPE (x);
+ tree mask_vec, perm_dest, data_ref;
+ gimple perm_stmt;
+
+ mask_vec = perm_mask_for_reverse (vectype);
+
+ perm_dest = vect_create_destination_var (gimple_assign_lhs (stmt), vectype);
+
+ /* Generate the permute statement. */
+ perm_stmt = gimple_build_assign_with_ops3 (VEC_PERM_EXPR, perm_dest,
+ x, x, mask_vec);
+ data_ref = make_ssa_name (perm_dest, perm_stmt);
+ gimple_set_lhs (perm_stmt, data_ref);
+ vect_finish_stmt_generation (stmt, perm_stmt, gsi);
+
+ return data_ref;
+}
+
/* vectorizable_load.
Check if STMT reads a non scalar data-ref (array/pointer/structure) that
bool nested_in_vect_loop = false;
struct data_reference *dr = STMT_VINFO_DATA_REF (stmt_info), *first_dr;
tree vectype = STMT_VINFO_VECTYPE (stmt_info);
+ tree elem_type;
tree new_temp;
enum machine_mode mode;
gimple new_stmt = NULL;
gimple phi = NULL;
VEC(tree,heap) *dr_chain = NULL;
bool strided_load = false;
+ bool load_lanes_p = false;
gimple first_stmt;
- tree scalar_type;
bool inv_p;
+ bool negative;
bool compute_in_loop = false;
struct loop *at_loop;
int vec_num;
enum tree_code code;
bb_vec_info bb_vinfo = STMT_VINFO_BB_VINFO (stmt_info);
int vf;
+ tree aggr_type;
if (loop_vinfo)
{
/* Multiple types in SLP are handled by creating the appropriate number of
vectorized stmts for each SLP node. Hence, NCOPIES is always 1 in
case of SLP. */
- if (slp)
+ if (slp || PURE_SLP_STMT (stmt_info))
ncopies = 1;
else
ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits;
&& code != COMPONENT_REF
&& code != IMAGPART_EXPR
&& code != REALPART_EXPR
- && code != MEM_REF)
+ && code != MEM_REF
+ && TREE_CODE_CLASS (code) != tcc_declaration)
return false;
if (!STMT_VINFO_DATA_REF (stmt_info))
return false;
- scalar_type = TREE_TYPE (DR_REF (dr));
+ negative = tree_int_cst_compare (DR_STEP (dr), size_zero_node) < 0;
+ if (negative && ncopies > 1)
+ {
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "multiple types with negative step.");
+ return false;
+ }
+
+ elem_type = TREE_TYPE (vectype);
mode = TYPE_MODE (vectype);
/* FORNOW. In some cases can vectorize even if data-type not supported
return false;
}
- /* The vector component type needs to be trivially convertible to the
- scalar lhs. This should always be the case. */
- if (!useless_type_conversion_p (TREE_TYPE (scalar_dest), TREE_TYPE (vectype)))
- {
- if (vect_print_dump_info (REPORT_DETAILS))
- fprintf (vect_dump, "??? operands of different types");
- return false;
- }
-
/* Check if the load is a part of an interleaving chain. */
if (STMT_VINFO_STRIDED_ACCESS (stmt_info))
{
/* FORNOW */
gcc_assert (! nested_in_vect_loop);
- /* Check if interleaving is supported. */
- if (!vect_strided_load_supported (vectype)
- && !PURE_SLP_STMT (stmt_info) && !slp)
- return false;
+ first_stmt = GROUP_FIRST_ELEMENT (stmt_info);
+ if (!slp && !PURE_SLP_STMT (stmt_info))
+ {
+ group_size = GROUP_SIZE (vinfo_for_stmt (first_stmt));
+ if (vect_load_lanes_supported (vectype, group_size))
+ load_lanes_p = true;
+ else if (!vect_strided_load_supported (vectype, group_size))
+ return false;
+ }
+ }
+
+ if (negative)
+ {
+ gcc_assert (!strided_load);
+ alignment_support_scheme = vect_supportable_dr_alignment (dr, false);
+ if (alignment_support_scheme != dr_aligned
+ && alignment_support_scheme != dr_unaligned_supported)
+ {
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "negative step but alignment required.");
+ return false;
+ }
+ if (!perm_mask_for_reverse (vectype))
+ {
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "negative step and reversing not supported.");
+ return false;
+ }
}
if (!vec_stmt) /* transformation not required. */
{
STMT_VINFO_TYPE (stmt_info) = load_vec_info_type;
- vect_model_load_cost (stmt_info, ncopies, NULL);
+ vect_model_load_cost (stmt_info, ncopies, load_lanes_p, NULL);
return true;
}
if (vect_print_dump_info (REPORT_DETAILS))
- fprintf (vect_dump, "transform load.");
+ fprintf (vect_dump, "transform load. ncopies = %d", ncopies);
/** Transform. **/
if (strided_load)
{
- first_stmt = DR_GROUP_FIRST_DR (stmt_info);
+ first_stmt = GROUP_FIRST_ELEMENT (stmt_info);
+ if (slp
+ && !SLP_INSTANCE_LOAD_PERMUTATION (slp_node_instance)
+ && first_stmt != VEC_index (gimple, SLP_TREE_SCALAR_STMTS (slp_node), 0))
+ first_stmt = VEC_index (gimple, SLP_TREE_SCALAR_STMTS (slp_node), 0);
+
/* Check if the chain of loads is already vectorized. */
if (STMT_VINFO_VEC_STMT (vinfo_for_stmt (first_stmt)))
{
return true;
}
first_dr = STMT_VINFO_DATA_REF (vinfo_for_stmt (first_stmt));
- group_size = DR_GROUP_SIZE (vinfo_for_stmt (first_stmt));
+ group_size = GROUP_SIZE (vinfo_for_stmt (first_stmt));
/* VEC_NUM is the number of vect stmts to be created for this group. */
if (slp)
}
else
vec_num = group_size;
-
- dr_chain = VEC_alloc (tree, heap, vec_num);
}
else
{
alignment_support_scheme = vect_supportable_dr_alignment (first_dr, false);
gcc_assert (alignment_support_scheme);
+ /* Targets with load-lane instructions must not require explicit
+ realignment. */
+ gcc_assert (!load_lanes_p
+ || alignment_support_scheme == dr_aligned
+ || alignment_support_scheme == dr_unaligned_supported);
/* In case the vectorization factor (VF) is bigger than the number
of elements that we can fit in a vectype (nunits), we have to generate
else
at_loop = loop;
+ if (negative)
+ offset = size_int (-TYPE_VECTOR_SUBPARTS (vectype) + 1);
+
+ if (load_lanes_p)
+ aggr_type = build_array_type_nelts (elem_type, vec_num * nunits);
+ else
+ aggr_type = vectype;
+
prev_stmt_info = NULL;
for (j = 0; j < ncopies; j++)
{
- /* 1. Create the vector pointer update chain. */
+ /* 1. Create the vector or array pointer update chain. */
if (j == 0)
- dataref_ptr = vect_create_data_ref_ptr (first_stmt,
- at_loop, offset,
- &dummy, &ptr_incr, false,
- &inv_p);
+ dataref_ptr = vect_create_data_ref_ptr (first_stmt, aggr_type, at_loop,
+ offset, &dummy, gsi,
+ &ptr_incr, false, &inv_p);
else
- dataref_ptr =
- bump_vector_ptr (dataref_ptr, ptr_incr, gsi, stmt, NULL_TREE);
+ dataref_ptr = bump_vector_ptr (dataref_ptr, ptr_incr, gsi, stmt,
+ TYPE_SIZE_UNIT (aggr_type));
- for (i = 0; i < vec_num; i++)
+ if (strided_load || slp_perm)
+ dr_chain = VEC_alloc (tree, heap, vec_num);
+
+ if (load_lanes_p)
{
- if (i > 0)
- dataref_ptr = bump_vector_ptr (dataref_ptr, ptr_incr, gsi, stmt,
- NULL_TREE);
+ tree vec_array;
- /* 2. Create the vector-load in the loop. */
- switch (alignment_support_scheme)
- {
- case dr_aligned:
- case dr_unaligned_supported:
- {
- struct ptr_info_def *pi;
- data_ref
- = build2 (MEM_REF, vectype, dataref_ptr,
- build_int_cst (reference_alias_ptr_type
- (DR_REF (first_dr)), 0));
- pi = get_ptr_info (dataref_ptr);
- pi->align = TYPE_ALIGN_UNIT (vectype);
- if (alignment_support_scheme == dr_aligned)
- {
- gcc_assert (aligned_access_p (first_dr));
- pi->misalign = 0;
- }
- else if (DR_MISALIGNMENT (first_dr) == -1)
- {
- TREE_TYPE (data_ref)
- = build_aligned_type (TREE_TYPE (data_ref),
- TYPE_ALIGN (TREE_TYPE (vectype)));
- pi->align = TYPE_ALIGN_UNIT (TREE_TYPE (vectype));
- pi->misalign = 0;
- }
- else
- {
- TREE_TYPE (data_ref)
- = build_aligned_type (TREE_TYPE (data_ref),
- TYPE_ALIGN (TREE_TYPE (vectype)));
- pi->misalign = DR_MISALIGNMENT (first_dr);
- }
- break;
- }
- case dr_explicit_realign:
- {
- tree ptr, bump;
- tree vs_minus_1 = size_int (TYPE_VECTOR_SUBPARTS (vectype) - 1);
-
- if (compute_in_loop)
- msq = vect_setup_realignment (first_stmt, gsi,
- &realignment_token,
- dr_explicit_realign,
- dataref_ptr, NULL);
-
- 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);
- gimple_assign_set_lhs (new_stmt, new_temp);
- gimple_set_vdef (new_stmt, gimple_vdef (stmt));
- gimple_set_vuse (new_stmt, gimple_vuse (stmt));
- vect_finish_stmt_generation (stmt, new_stmt, gsi);
- msq = new_temp;
-
- bump = size_binop (MULT_EXPR, vs_minus_1,
- TYPE_SIZE_UNIT (scalar_type));
- ptr = bump_vector_ptr (dataref_ptr, NULL, gsi, stmt, bump);
- 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:
- 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 ();
- }
- 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);
- gimple_assign_set_lhs (new_stmt, new_temp);
+ vec_array = create_vector_array (vectype, vec_num);
+
+ /* Emit:
+ VEC_ARRAY = LOAD_LANES (MEM_REF[...all elements...]). */
+ data_ref = create_array_ref (aggr_type, dataref_ptr, first_dr);
+ new_stmt = gimple_build_call_internal (IFN_LOAD_LANES, 1, data_ref);
+ gimple_call_set_lhs (new_stmt, vec_array);
vect_finish_stmt_generation (stmt, new_stmt, gsi);
mark_symbols_for_renaming (new_stmt);
- /* 3. Handle explicit realignment if necessary/supported. Create in
- loop: vec_dest = realign_load (msq, lsq, realignment_token) */
- if (alignment_support_scheme == dr_explicit_realign_optimized
- || alignment_support_scheme == dr_explicit_realign)
+ /* Extract each vector into an SSA_NAME. */
+ for (i = 0; i < vec_num; i++)
+ {
+ new_temp = read_vector_array (stmt, gsi, scalar_dest,
+ vec_array, i);
+ VEC_quick_push (tree, dr_chain, new_temp);
+ }
+
+ /* Record the mapping between SSA_NAMEs and statements. */
+ vect_record_strided_load_vectors (stmt, dr_chain);
+ }
+ else
+ {
+ for (i = 0; i < vec_num; i++)
{
- tree tmp;
+ if (i > 0)
+ dataref_ptr = bump_vector_ptr (dataref_ptr, ptr_incr, gsi,
+ stmt, NULL_TREE);
- lsq = gimple_assign_lhs (new_stmt);
- if (!realignment_token)
- realignment_token = dataref_ptr;
+ /* 2. Create the vector-load in the loop. */
+ switch (alignment_support_scheme)
+ {
+ case dr_aligned:
+ case dr_unaligned_supported:
+ {
+ struct ptr_info_def *pi;
+ data_ref
+ = build2 (MEM_REF, vectype, dataref_ptr,
+ build_int_cst (reference_alias_ptr_type
+ (DR_REF (first_dr)), 0));
+ pi = get_ptr_info (dataref_ptr);
+ pi->align = TYPE_ALIGN_UNIT (vectype);
+ if (alignment_support_scheme == dr_aligned)
+ {
+ gcc_assert (aligned_access_p (first_dr));
+ pi->misalign = 0;
+ }
+ else if (DR_MISALIGNMENT (first_dr) == -1)
+ {
+ TREE_TYPE (data_ref)
+ = build_aligned_type (TREE_TYPE (data_ref),
+ TYPE_ALIGN (elem_type));
+ pi->align = TYPE_ALIGN_UNIT (elem_type);
+ pi->misalign = 0;
+ }
+ else
+ {
+ TREE_TYPE (data_ref)
+ = build_aligned_type (TREE_TYPE (data_ref),
+ TYPE_ALIGN (elem_type));
+ pi->misalign = DR_MISALIGNMENT (first_dr);
+ }
+ break;
+ }
+ case dr_explicit_realign:
+ {
+ tree ptr, bump;
+ tree vs_minus_1;
+
+ vs_minus_1 = size_int (TYPE_VECTOR_SUBPARTS (vectype) - 1);
+
+ if (compute_in_loop)
+ msq = vect_setup_realignment (first_stmt, gsi,
+ &realignment_token,
+ dr_explicit_realign,
+ dataref_ptr, NULL);
+
+ 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);
+ gimple_assign_set_lhs (new_stmt, new_temp);
+ gimple_set_vdef (new_stmt, gimple_vdef (stmt));
+ gimple_set_vuse (new_stmt, gimple_vuse (stmt));
+ vect_finish_stmt_generation (stmt, new_stmt, gsi);
+ msq = new_temp;
+
+ bump = size_binop (MULT_EXPR, vs_minus_1,
+ TYPE_SIZE_UNIT (elem_type));
+ ptr = bump_vector_ptr (dataref_ptr, NULL, gsi, stmt, bump);
+ 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:
+ 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 ();
+ }
vec_dest = vect_create_destination_var (scalar_dest, vectype);
- tmp = build3 (REALIGN_LOAD_EXPR, vectype, msq, lsq,
- realignment_token);
- new_stmt = gimple_build_assign (vec_dest, tmp);
+ new_stmt = gimple_build_assign (vec_dest, data_ref);
new_temp = make_ssa_name (vec_dest, new_stmt);
gimple_assign_set_lhs (new_stmt, new_temp);
vect_finish_stmt_generation (stmt, new_stmt, gsi);
+ mark_symbols_for_renaming (new_stmt);
- if (alignment_support_scheme == dr_explicit_realign_optimized)
+ /* 3. Handle explicit realignment if necessary/supported.
+ Create in loop:
+ vec_dest = realign_load (msq, lsq, realignment_token) */
+ if (alignment_support_scheme == dr_explicit_realign_optimized
+ || alignment_support_scheme == dr_explicit_realign)
{
- gcc_assert (phi);
- if (i == vec_num - 1 && j == ncopies - 1)
- add_phi_arg (phi, lsq, loop_latch_edge (containing_loop),
- UNKNOWN_LOCATION);
- msq = lsq;
+ lsq = gimple_assign_lhs (new_stmt);
+ if (!realignment_token)
+ realignment_token = dataref_ptr;
+ vec_dest = vect_create_destination_var (scalar_dest, vectype);
+ new_stmt
+ = gimple_build_assign_with_ops3 (REALIGN_LOAD_EXPR,
+ vec_dest, msq, lsq,
+ realignment_token);
+ new_temp = make_ssa_name (vec_dest, new_stmt);
+ gimple_assign_set_lhs (new_stmt, new_temp);
+ vect_finish_stmt_generation (stmt, new_stmt, gsi);
+
+ if (alignment_support_scheme == dr_explicit_realign_optimized)
+ {
+ gcc_assert (phi);
+ if (i == vec_num - 1 && j == ncopies - 1)
+ add_phi_arg (phi, lsq,
+ loop_latch_edge (containing_loop),
+ UNKNOWN_LOCATION);
+ msq = lsq;
+ }
}
- }
- /* 4. Handle invariant-load. */
- if (inv_p && !bb_vinfo)
- {
- gcc_assert (!strided_load);
- gcc_assert (nested_in_vect_loop_p (loop, stmt));
- if (j == 0)
+ /* 4. Handle invariant-load. */
+ if (inv_p && !bb_vinfo)
{
- int k;
- tree t = NULL_TREE;
- tree vec_inv, bitpos, bitsize = TYPE_SIZE (scalar_type);
-
- /* CHECKME: bitpos depends on endianess? */
- bitpos = bitsize_zero_node;
- vec_inv = build3 (BIT_FIELD_REF, scalar_type, new_temp,
- bitsize, bitpos);
- vec_dest =
- vect_create_destination_var (scalar_dest, NULL_TREE);
- new_stmt = gimple_build_assign (vec_dest, vec_inv);
- new_temp = make_ssa_name (vec_dest, new_stmt);
- gimple_assign_set_lhs (new_stmt, new_temp);
- vect_finish_stmt_generation (stmt, new_stmt, gsi);
+ tree tem, vec_inv;
+ gimple_stmt_iterator gsi2 = *gsi;
+ gcc_assert (!strided_load);
+ gsi_next (&gsi2);
+ tem = scalar_dest;
+ if (!useless_type_conversion_p (TREE_TYPE (vectype),
+ TREE_TYPE (tem)))
+ {
+ tem = fold_convert (TREE_TYPE (vectype), tem);
+ tem = force_gimple_operand_gsi (&gsi2, tem, true,
+ NULL_TREE, true,
+ GSI_SAME_STMT);
+ }
+ vec_inv = build_vector_from_val (vectype, tem);
+ new_temp = vect_init_vector (stmt, vec_inv,
+ vectype, &gsi2);
+ new_stmt = SSA_NAME_DEF_STMT (new_temp);
+ }
- for (k = nunits - 1; k >= 0; --k)
- t = tree_cons (NULL_TREE, new_temp, t);
- /* FIXME: use build_constructor directly. */
- vec_inv = build_constructor_from_list (vectype, t);
- new_temp = vect_init_vector (stmt, vec_inv, vectype, gsi);
+ if (negative)
+ {
+ new_temp = reverse_vec_elements (new_temp, stmt, gsi);
new_stmt = SSA_NAME_DEF_STMT (new_temp);
}
- else
- gcc_unreachable (); /* FORNOW. */
- }
- /* Collect vector loads and later create their permutation in
- vect_transform_strided_load (). */
- if (strided_load || slp_perm)
- VEC_quick_push (tree, dr_chain, new_temp);
+ /* Collect vector loads and later create their permutation in
+ vect_transform_strided_load (). */
+ if (strided_load || slp_perm)
+ VEC_quick_push (tree, dr_chain, new_temp);
- /* Store vector loads in the corresponding SLP_NODE. */
- if (slp && !slp_perm)
- VEC_quick_push (gimple, SLP_TREE_VEC_STMTS (slp_node), new_stmt);
+ /* Store vector loads in the corresponding SLP_NODE. */
+ if (slp && !slp_perm)
+ VEC_quick_push (gimple, SLP_TREE_VEC_STMTS (slp_node),
+ new_stmt);
+ }
}
if (slp && !slp_perm)
{
if (strided_load)
{
- if (!vect_transform_strided_load (stmt, dr_chain, group_size, gsi))
- return false;
-
+ if (!load_lanes_p)
+ vect_transform_strided_load (stmt, dr_chain, group_size, gsi);
*vec_stmt = STMT_VINFO_VEC_STMT (stmt_info);
- VEC_free (tree, heap, dr_chain);
- dr_chain = VEC_alloc (tree, heap, group_size);
}
else
{
prev_stmt_info = vinfo_for_stmt (new_stmt);
}
}
+ if (dr_chain)
+ VEC_free (tree, heap, dr_chain);
}
- if (dr_chain)
- VEC_free (tree, heap, dr_chain);
-
return true;
}
LOOP - the loop that is being vectorized.
COND - Condition that is checked for simple use.
+ Output:
+ *COMP_VECTYPE - the vector type for the comparison.
+
Returns whether a COND can be vectorized. Checks whether
condition operands are supportable using vec_is_simple_use. */
static bool
-vect_is_simple_cond (tree cond, loop_vec_info loop_vinfo)
+vect_is_simple_cond (tree cond, loop_vec_info loop_vinfo, tree *comp_vectype)
{
tree lhs, rhs;
tree def;
enum vect_def_type dt;
+ tree vectype1 = NULL_TREE, vectype2 = NULL_TREE;
if (!COMPARISON_CLASS_P (cond))
return false;
if (TREE_CODE (lhs) == SSA_NAME)
{
gimple lhs_def_stmt = SSA_NAME_DEF_STMT (lhs);
- if (!vect_is_simple_use (lhs, loop_vinfo, NULL, &lhs_def_stmt, &def,
- &dt))
+ if (!vect_is_simple_use_1 (lhs, loop_vinfo, NULL, &lhs_def_stmt, &def,
+ &dt, &vectype1))
return false;
}
else if (TREE_CODE (lhs) != INTEGER_CST && TREE_CODE (lhs) != REAL_CST
if (TREE_CODE (rhs) == SSA_NAME)
{
gimple rhs_def_stmt = SSA_NAME_DEF_STMT (rhs);
- if (!vect_is_simple_use (rhs, loop_vinfo, NULL, &rhs_def_stmt, &def,
- &dt))
+ if (!vect_is_simple_use_1 (rhs, loop_vinfo, NULL, &rhs_def_stmt, &def,
+ &dt, &vectype2))
return false;
}
else if (TREE_CODE (rhs) != INTEGER_CST && TREE_CODE (rhs) != REAL_CST
&& TREE_CODE (rhs) != FIXED_CST)
return false;
+ *comp_vectype = vectype1 ? vectype1 : vectype2;
return true;
}
{
tree scalar_dest = NULL_TREE;
tree vec_dest = NULL_TREE;
- tree op = NULL_TREE;
tree cond_expr, then_clause, else_clause;
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
tree vectype = STMT_VINFO_VECTYPE (stmt_info);
+ tree comp_vectype;
tree vec_cond_lhs = NULL_TREE, vec_cond_rhs = NULL_TREE;
tree vec_then_clause = NULL_TREE, vec_else_clause = NULL_TREE;
tree vec_compare, vec_cond_expr;
tree new_temp;
loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
- enum machine_mode vec_mode;
tree def;
enum vect_def_type dt, dts[4];
int nunits = TYPE_VECTOR_SUBPARTS (vectype);
/* FORNOW: unsupported in basic block SLP. */
gcc_assert (loop_vinfo);
+ /* FORNOW: SLP not supported. */
+ if (STMT_SLP_TYPE (stmt_info))
+ return false;
+
gcc_assert (ncopies >= 1);
if (reduc_index && ncopies > 1)
return false; /* FORNOW */
&& reduc_def))
return false;
- /* FORNOW: SLP not supported. */
- if (STMT_SLP_TYPE (stmt_info))
- return false;
-
/* FORNOW: not yet supported. */
if (STMT_VINFO_LIVE_P (stmt_info))
{
if (code != COND_EXPR)
return false;
- gcc_assert (gimple_assign_single_p (stmt));
- op = gimple_assign_rhs1 (stmt);
- cond_expr = TREE_OPERAND (op, 0);
- then_clause = TREE_OPERAND (op, 1);
- else_clause = TREE_OPERAND (op, 2);
-
- if (!vect_is_simple_cond (cond_expr, loop_vinfo))
- return false;
+ cond_expr = gimple_assign_rhs1 (stmt);
+ then_clause = gimple_assign_rhs2 (stmt);
+ else_clause = gimple_assign_rhs3 (stmt);
- /* We do not handle two different vector types for the condition
- and the values. */
- if (!types_compatible_p (TREE_TYPE (TREE_OPERAND (cond_expr, 0)),
- TREE_TYPE (vectype)))
+ if (!vect_is_simple_cond (cond_expr, loop_vinfo, &comp_vectype)
+ || !comp_vectype)
return false;
if (TREE_CODE (then_clause) == SSA_NAME)
&& TREE_CODE (else_clause) != FIXED_CST)
return false;
-
- vec_mode = TYPE_MODE (vectype);
-
if (!vec_stmt)
{
STMT_VINFO_TYPE (stmt_info) = condition_vec_info_type;
- return expand_vec_cond_expr_p (TREE_TYPE (op), vec_mode);
+ return expand_vec_cond_expr_p (vectype, comp_vectype);
}
/* Transform */
vec_else_clause);
}
- /* Arguments are ready. Create the new vector stmt. */
+ /* Arguments are ready. Create the new vector stmt. */
vec_compare = build2 (TREE_CODE (cond_expr), vectype,
vec_cond_lhs, vec_cond_rhs);
vec_cond_expr = build3 (VEC_COND_EXPR, vectype,
enum vect_relevant relevance = STMT_VINFO_RELEVANT (stmt_info);
bool ok;
tree scalar_type, vectype;
+ gimple pattern_stmt, pattern_def_stmt;
if (vect_print_dump_info (REPORT_DETAILS))
{
- any LABEL_EXPRs in the loop
- computations that are used only for array indexing or loop control.
In basic blocks we only analyze statements that are a part of some SLP
- instance, therefore, all the statements are relevant. */
+ instance, therefore, all the statements are relevant.
+
+ Pattern statement need to be analyzed instead of the original statement
+ if the original statement is not relevant. Otherwise, we analyze both
+ statements. */
+ pattern_stmt = STMT_VINFO_RELATED_STMT (stmt_info);
if (!STMT_VINFO_RELEVANT_P (stmt_info)
&& !STMT_VINFO_LIVE_P (stmt_info))
{
- if (vect_print_dump_info (REPORT_DETAILS))
- fprintf (vect_dump, "irrelevant.");
+ if (STMT_VINFO_IN_PATTERN_P (stmt_info)
+ && pattern_stmt
+ && (STMT_VINFO_RELEVANT_P (vinfo_for_stmt (pattern_stmt))
+ || STMT_VINFO_LIVE_P (vinfo_for_stmt (pattern_stmt))))
+ {
+ /* Analyze PATTERN_STMT instead of the original stmt. */
+ stmt = pattern_stmt;
+ stmt_info = vinfo_for_stmt (pattern_stmt);
+ if (vect_print_dump_info (REPORT_DETAILS))
+ {
+ fprintf (vect_dump, "==> examining pattern statement: ");
+ print_gimple_stmt (vect_dump, stmt, 0, TDF_SLIM);
+ }
+ }
+ else
+ {
+ if (vect_print_dump_info (REPORT_DETAILS))
+ fprintf (vect_dump, "irrelevant.");
- return true;
+ return true;
+ }
}
+ else if (STMT_VINFO_IN_PATTERN_P (stmt_info)
+ && pattern_stmt
+ && (STMT_VINFO_RELEVANT_P (vinfo_for_stmt (pattern_stmt))
+ || STMT_VINFO_LIVE_P (vinfo_for_stmt (pattern_stmt))))
+ {
+ /* Analyze PATTERN_STMT too. */
+ if (vect_print_dump_info (REPORT_DETAILS))
+ {
+ fprintf (vect_dump, "==> examining pattern statement: ");
+ print_gimple_stmt (vect_dump, stmt, 0, TDF_SLIM);
+ }
+
+ if (!vect_analyze_stmt (pattern_stmt, need_to_vectorize, node))
+ return false;
+ }
+
+ if (is_pattern_stmt_p (stmt_info)
+ && (pattern_def_stmt = STMT_VINFO_PATTERN_DEF_STMT (stmt_info))
+ && (STMT_VINFO_RELEVANT_P (vinfo_for_stmt (pattern_def_stmt))
+ || STMT_VINFO_LIVE_P (vinfo_for_stmt (pattern_def_stmt))))
+ {
+ /* Analyze def stmt of STMT if it's a pattern stmt. */
+ if (vect_print_dump_info (REPORT_DETAILS))
+ {
+ fprintf (vect_dump, "==> examining pattern def statement: ");
+ print_gimple_stmt (vect_dump, pattern_def_stmt, 0, TDF_SLIM);
+ }
+
+ if (!vect_analyze_stmt (pattern_def_stmt, need_to_vectorize, node))
+ return false;
+ }
+
switch (STMT_VINFO_DEF_TYPE (stmt_info))
{
ok = (vectorizable_type_promotion (stmt, NULL, NULL, NULL)
|| vectorizable_type_demotion (stmt, NULL, NULL, NULL)
|| vectorizable_conversion (stmt, NULL, NULL, NULL)
+ || vectorizable_shift (stmt, NULL, NULL, NULL)
|| vectorizable_operation (stmt, NULL, NULL, NULL)
|| vectorizable_assignment (stmt, NULL, NULL, NULL)
|| vectorizable_load (stmt, NULL, NULL, NULL, NULL)
else
{
if (bb_vinfo)
- ok = (vectorizable_operation (stmt, NULL, NULL, node)
+ ok = (vectorizable_type_promotion (stmt, NULL, NULL, node)
+ || vectorizable_type_demotion (stmt, NULL, NULL, node)
+ || vectorizable_shift (stmt, NULL, NULL, node)
+ || vectorizable_operation (stmt, NULL, NULL, node)
|| vectorizable_assignment (stmt, NULL, NULL, node)
|| vectorizable_load (stmt, NULL, NULL, node, NULL)
|| vectorizable_store (stmt, NULL, NULL, node));
return false;
}
- if (!PURE_SLP_STMT (stmt_info))
- {
- /* Groups of strided accesses whose size is not a power of 2 are not
- vectorizable yet using loop-vectorization. Therefore, if this stmt
- feeds non-SLP-able stmts (i.e., this stmt has to be both SLPed and
- loop-based vectorized), the loop cannot be vectorized. */
- if (STMT_VINFO_STRIDED_ACCESS (stmt_info)
- && exact_log2 (DR_GROUP_SIZE (vinfo_for_stmt (
- DR_GROUP_FIRST_DR (stmt_info)))) == -1)
- {
- if (vect_print_dump_info (REPORT_DETAILS))
- {
- fprintf (vect_dump, "not vectorized: the size of group "
- "of strided accesses is not a power of 2");
- print_gimple_stmt (vect_dump, stmt, 0, TDF_SLIM);
- }
-
- return false;
- }
- }
-
return true;
}
bool is_store = false;
gimple vec_stmt = NULL;
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
- gimple orig_stmt_in_pattern;
bool done;
switch (STMT_VINFO_TYPE (stmt_info))
gcc_assert (done);
break;
+ case shift_vec_info_type:
+ done = vectorizable_shift (stmt, gsi, &vec_stmt, slp_node);
+ gcc_assert (done);
+ break;
+
case op_vec_info_type:
done = vectorizable_operation (stmt, gsi, &vec_stmt, slp_node);
gcc_assert (done);
case call_vec_info_type:
gcc_assert (!slp_node);
done = vectorizable_call (stmt, gsi, &vec_stmt);
+ stmt = gsi_stmt (*gsi);
break;
case reduc_vec_info_type:
}
if (vec_stmt)
- {
- STMT_VINFO_VEC_STMT (stmt_info) = vec_stmt;
- orig_stmt_in_pattern = STMT_VINFO_RELATED_STMT (stmt_info);
- if (orig_stmt_in_pattern)
- {
- stmt_vec_info stmt_vinfo = vinfo_for_stmt (orig_stmt_in_pattern);
- /* STMT was inserted by the vectorizer to replace a computation idiom.
- ORIG_STMT_IN_PATTERN is a stmt in the original sequence that
- computed this idiom. We need to record a pointer to VEC_STMT in
- the stmt_info of ORIG_STMT_IN_PATTERN. See more details in the
- documentation of vect_pattern_recog. */
- if (STMT_VINFO_IN_PATTERN_P (stmt_vinfo))
- {
- gcc_assert (STMT_VINFO_RELATED_STMT (stmt_vinfo) == stmt);
- STMT_VINFO_VEC_STMT (stmt_vinfo) = vec_stmt;
- }
- }
- }
+ STMT_VINFO_VEC_STMT (stmt_info) = vec_stmt;
return is_store;
}
/* Free the attached stmt_vec_info and remove the stmt. */
next_si = gsi_for_stmt (next);
gsi_remove (&next_si, true);
- tmp = DR_GROUP_NEXT_DR (vinfo_for_stmt (next));
+ tmp = GROUP_NEXT_ELEMENT (vinfo_for_stmt (next));
free_stmt_vec_info (next);
next = tmp;
}
STMT_VINFO_VECTORIZABLE (res) = true;
STMT_VINFO_IN_PATTERN_P (res) = false;
STMT_VINFO_RELATED_STMT (res) = NULL;
+ STMT_VINFO_PATTERN_DEF_STMT (res) = NULL;
STMT_VINFO_DATA_REF (res) = NULL;
STMT_VINFO_DR_BASE_ADDRESS (res) = NULL;
STMT_VINFO_INSIDE_OF_LOOP_COST (res) = 0;
STMT_VINFO_OUTSIDE_OF_LOOP_COST (res) = 0;
STMT_SLP_TYPE (res) = loop_vect;
- DR_GROUP_FIRST_DR (res) = NULL;
- DR_GROUP_NEXT_DR (res) = NULL;
- DR_GROUP_SIZE (res) = 0;
- DR_GROUP_STORE_COUNT (res) = 0;
- DR_GROUP_GAP (res) = 0;
- DR_GROUP_SAME_DR_STMT (res) = NULL;
- DR_GROUP_READ_WRITE_DEPENDENCE (res) = false;
+ GROUP_FIRST_ELEMENT (res) = NULL;
+ GROUP_NEXT_ELEMENT (res) = NULL;
+ GROUP_SIZE (res) = 0;
+ GROUP_STORE_COUNT (res) = 0;
+ GROUP_GAP (res) = 0;
+ GROUP_SAME_DR_STMT (res) = NULL;
+ GROUP_READ_WRITE_DEPENDENCE (res) = false;
return res;
}
}
-/* Function get_vectype_for_scalar_type.
+/* Function get_vectype_for_scalar_type_and_size.
- Returns the vector type corresponding to SCALAR_TYPE as supported
+ Returns the vector type corresponding to SCALAR_TYPE and SIZE as supported
by the target. */
-tree
-get_vectype_for_scalar_type (tree scalar_type)
+static tree
+get_vectype_for_scalar_type_and_size (tree scalar_type, unsigned size)
{
enum machine_mode inner_mode = TYPE_MODE (scalar_type);
+ enum machine_mode simd_mode;
unsigned int nbytes = GET_MODE_SIZE (inner_mode);
int nunits;
tree vectype;
- if (nbytes == 0
- || (nbytes >= targetm.vectorize.units_per_simd_word (inner_mode)))
+ if (nbytes == 0)
return NULL_TREE;
/* We can't build a vector type of elements with alignment bigger than
if (nbytes < TYPE_ALIGN_UNIT (scalar_type))
return NULL_TREE;
- /* If we'd build a vector type of elements whose mode precision doesn't
- match their types precision we'll get mismatched types on vector
- extracts via BIT_FIELD_REFs. This effectively means we disable
- vectorization of bool and/or enum types in some languages. */
+ /* For vector types of elements whose mode precision doesn't
+ match their types precision we use a element type of mode
+ precision. The vectorization routines will have to make sure
+ they support the proper result truncation/extension. */
if (INTEGRAL_TYPE_P (scalar_type)
&& GET_MODE_BITSIZE (inner_mode) != TYPE_PRECISION (scalar_type))
+ scalar_type = build_nonstandard_integer_type (GET_MODE_BITSIZE (inner_mode),
+ TYPE_UNSIGNED (scalar_type));
+
+ if (GET_MODE_CLASS (inner_mode) != MODE_INT
+ && GET_MODE_CLASS (inner_mode) != MODE_FLOAT)
return NULL_TREE;
- /* FORNOW: Only a single vector size per mode
- (TARGET_VECTORIZE_UNITS_PER_SIMD_WORD) is expected. */
- nunits = targetm.vectorize.units_per_simd_word (inner_mode) / nbytes;
+ /* We shouldn't end up building VECTOR_TYPEs of non-scalar components.
+ When the component mode passes the above test simply use a type
+ corresponding to that mode. The theory is that any use that
+ would cause problems with this will disable vectorization anyway. */
+ if (!SCALAR_FLOAT_TYPE_P (scalar_type)
+ && !INTEGRAL_TYPE_P (scalar_type)
+ && !POINTER_TYPE_P (scalar_type))
+ scalar_type = lang_hooks.types.type_for_mode (inner_mode, 1);
+
+ /* If no size was supplied use the mode the target prefers. Otherwise
+ lookup a vector mode of the specified size. */
+ if (size == 0)
+ simd_mode = targetm.vectorize.preferred_simd_mode (inner_mode);
+ else
+ simd_mode = mode_for_vector (inner_mode, size / nbytes);
+ nunits = GET_MODE_SIZE (simd_mode) / nbytes;
+ if (nunits <= 1)
+ return NULL_TREE;
vectype = build_vector_type (scalar_type, nunits);
if (vect_print_dump_info (REPORT_DETAILS))
return vectype;
}
+unsigned int current_vector_size;
+
+/* Function get_vectype_for_scalar_type.
+
+ Returns the vector type corresponding to SCALAR_TYPE as supported
+ by the target. */
+
+tree
+get_vectype_for_scalar_type (tree scalar_type)
+{
+ tree vectype;
+ vectype = get_vectype_for_scalar_type_and_size (scalar_type,
+ current_vector_size);
+ if (vectype
+ && current_vector_size == 0)
+ current_vector_size = GET_MODE_SIZE (TYPE_MODE (vectype));
+ return vectype;
+}
+
/* Function get_same_sized_vectype
Returns a vector type corresponding to SCALAR_TYPE of size
VECTOR_TYPE if supported by the target. */
tree
-get_same_sized_vectype (tree scalar_type, tree vector_type ATTRIBUTE_UNUSED)
+get_same_sized_vectype (tree scalar_type, tree vector_type)
{
- return get_vectype_for_scalar_type (scalar_type);
+ return get_vectype_for_scalar_type_and_size
+ (scalar_type, GET_MODE_SIZE (TYPE_MODE (vector_type)));
}
/* Function vect_is_simple_use.
|| *dt == vect_nested_cycle)
{
stmt_vec_info stmt_info = vinfo_for_stmt (*def_stmt);
- if (STMT_VINFO_IN_PATTERN_P (stmt_info))
+
+ if (STMT_VINFO_IN_PATTERN_P (stmt_info)
+ && !STMT_VINFO_RELEVANT (stmt_info)
+ && !STMT_VINFO_LIVE_P (stmt_info))
stmt_info = vinfo_for_stmt (STMT_VINFO_RELATED_STMT (stmt_info));
+
*vectype = STMT_VINFO_VECTYPE (stmt_info);
gcc_assert (*vectype != NULL_TREE);
}
{
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
loop_vec_info loop_info = STMT_VINFO_LOOP_VINFO (stmt_info);
- struct loop *vect_loop = LOOP_VINFO_LOOP (loop_info);
+ struct loop *vect_loop = NULL;
bool ordered_p;
enum machine_mode vec_mode;
enum insn_code icode1, icode2;
tree wide_vectype = vectype_out;
enum tree_code c1, c2;
+ if (loop_info)
+ vect_loop = LOOP_VINFO_LOOP (loop_info);
+
/* The result of a vectorized widening operation usually requires two vectors
(because the widened results do not fit int one vector). The generated
vector results would normally be expected to be generated in the same
iterations in parallel). We therefore don't allow to change the order
of the computation in the inner-loop during outer-loop vectorization. */
- if (STMT_VINFO_RELEVANT (stmt_info) == vect_used_by_reduction
+ if (vect_loop
+ && STMT_VINFO_RELEVANT (stmt_info) == vect_used_by_reduction
&& !nested_in_vect_loop_p (vect_loop, stmt))
ordered_p = false;
else
}
break;
+ case WIDEN_LSHIFT_EXPR:
+ if (BYTES_BIG_ENDIAN)
+ {
+ c1 = VEC_WIDEN_LSHIFT_HI_EXPR;
+ c2 = VEC_WIDEN_LSHIFT_LO_EXPR;
+ }
+ else
+ {
+ c2 = VEC_WIDEN_LSHIFT_HI_EXPR;
+ c1 = VEC_WIDEN_LSHIFT_LO_EXPR;
+ }
+ break;
+
CASE_CONVERT:
if (BYTES_BIG_ENDIAN)
{
OSDN Git Service
