/* Statement Analysis and Transformation for Vectorization
- Copyright (C) 2003-2015 Free Software Foundation, Inc.
+ Copyright (C) 2003-2016 Free Software Foundation, Inc.
Contributed by Dorit Naishlos <dorit@il.ibm.com>
and Ira Rosen <irar@il.ibm.com>
#include "tree-scalar-evolution.h"
#include "tree-vectorizer.h"
#include "builtins.h"
+#include "internal-fn.h"
/* For lang_hooks.types.type_for_mode. */
#include "langhooks.h"
{
if (!types_compatible_p (TREE_TYPE (type), TREE_TYPE (val)))
{
- if (CONSTANT_CLASS_P (val))
+ /* Scalar boolean value should be transformed into
+ all zeros or all ones value before building a vector. */
+ if (VECTOR_BOOLEAN_TYPE_P (type))
+ {
+ tree true_val = build_all_ones_cst (TREE_TYPE (type));
+ tree false_val = build_zero_cst (TREE_TYPE (type));
+
+ if (CONSTANT_CLASS_P (val))
+ val = integer_zerop (val) ? false_val : true_val;
+ else
+ {
+ new_temp = make_ssa_name (TREE_TYPE (type));
+ init_stmt = gimple_build_assign (new_temp, COND_EXPR,
+ val, true_val, false_val);
+ vect_init_vector_1 (stmt, init_stmt, gsi);
+ val = new_temp;
+ }
+ }
+ else if (CONSTANT_CLASS_P (val))
val = fold_convert (TREE_TYPE (type), val);
else
{
add_stmt_to_eh_lp (vec_stmt, lp_nr);
}
-/* Checks if CALL can be vectorized in type VECTYPE. Returns
- a function declaration if the target has a vectorized version
- of the function, or NULL_TREE if the function cannot be vectorized. */
+/* We want to vectorize a call to combined function CFN with function
+ decl FNDECL, using VECTYPE_OUT as the type of the output and VECTYPE_IN
+ as the types of all inputs. Check whether this is possible using
+ an internal function, returning its code if so or IFN_LAST if not. */
-tree
-vectorizable_function (gcall *call, tree vectype_out, tree vectype_in)
+static internal_fn
+vectorizable_internal_function (combined_fn cfn, tree fndecl,
+ tree vectype_out, tree vectype_in)
{
- tree fndecl = gimple_call_fndecl (call);
-
- /* We only handle functions that do not read or clobber memory -- i.e.
- const or novops ones. */
- if (!(gimple_call_flags (call) & (ECF_CONST | ECF_NOVOPS)))
- return NULL_TREE;
-
- if (!fndecl
- || TREE_CODE (fndecl) != FUNCTION_DECL
- || !DECL_BUILT_IN (fndecl))
- return NULL_TREE;
-
- return targetm.vectorize.builtin_vectorized_function (fndecl, vectype_out,
- vectype_in);
+ internal_fn ifn;
+ if (internal_fn_p (cfn))
+ ifn = as_internal_fn (cfn);
+ else
+ ifn = associated_internal_fn (fndecl);
+ if (ifn != IFN_LAST && direct_internal_fn_p (ifn))
+ {
+ const direct_internal_fn_info &info = direct_internal_fn (ifn);
+ if (info.vectorizable)
+ {
+ tree type0 = (info.type0 < 0 ? vectype_out : vectype_in);
+ tree type1 = (info.type1 < 0 ? vectype_out : vectype_in);
+ if (direct_internal_fn_supported_p (ifn, tree_pair (type0, type1),
+ OPTIMIZE_FOR_SPEED))
+ return ifn;
+ }
+ }
+ return IFN_LAST;
}
bool nested_in_vect_loop = nested_in_vect_loop_p (loop, stmt);
struct data_reference *dr = STMT_VINFO_DATA_REF (stmt_info);
tree vectype = STMT_VINFO_VECTYPE (stmt_info);
+ tree rhs_vectype = NULL_TREE;
+ tree mask_vectype;
tree elem_type;
gimple *new_stmt;
tree dummy;
is_store = gimple_call_internal_fn (stmt) == IFN_MASK_STORE;
mask = gimple_call_arg (stmt, 2);
- if (TYPE_PRECISION (TREE_TYPE (mask))
- != GET_MODE_BITSIZE (TYPE_MODE (TREE_TYPE (vectype))))
+
+ if (TREE_CODE (TREE_TYPE (mask)) != BOOLEAN_TYPE)
return false;
/* FORNOW. This restriction should be relaxed. */
if (STMT_VINFO_STRIDED_P (stmt_info))
return false;
+ if (TREE_CODE (mask) != SSA_NAME)
+ return false;
+
+ if (!vect_is_simple_use (mask, loop_vinfo, &def_stmt, &dt, &mask_vectype))
+ return false;
+
+ if (!mask_vectype)
+ mask_vectype = get_mask_type_for_scalar_type (TREE_TYPE (vectype));
+
+ if (!mask_vectype || !VECTOR_BOOLEAN_TYPE_P (mask_vectype))
+ return false;
+
+ if (is_store)
+ {
+ tree rhs = gimple_call_arg (stmt, 3);
+ if (!vect_is_simple_use (rhs, loop_vinfo, &def_stmt, &dt, &rhs_vectype))
+ return false;
+ }
+
if (STMT_VINFO_GATHER_SCATTER_P (stmt_info))
{
gimple *def_stmt;
: DR_STEP (dr), size_zero_node) <= 0)
return false;
else if (!VECTOR_MODE_P (TYPE_MODE (vectype))
- || !can_vec_mask_load_store_p (TYPE_MODE (vectype), !is_store))
- return false;
-
- if (TREE_CODE (mask) != SSA_NAME)
- return false;
-
- if (!vect_is_simple_use (mask, loop_vinfo, &def_stmt, &dt))
+ || !can_vec_mask_load_store_p (TYPE_MODE (vectype),
+ TYPE_MODE (mask_vectype),
+ !is_store)
+ || (rhs_vectype
+ && !useless_type_conversion_p (vectype, rhs_vectype)))
return false;
- if (is_store)
- {
- tree rhs = gimple_call_arg (stmt, 3);
- if (!vect_is_simple_use (rhs, loop_vinfo, &def_stmt, &dt))
- return false;
- }
-
if (!vec_stmt) /* transformation not required. */
{
STMT_VINFO_TYPE (stmt_info) = call_vec_info_type;
/* Ensure that even with -fno-tree-dce the scalar MASK_LOAD is removed
from the IL. */
+ if (STMT_VINFO_RELATED_STMT (stmt_info))
+ {
+ stmt = STMT_VINFO_RELATED_STMT (stmt_info);
+ stmt_info = vinfo_for_stmt (stmt);
+ }
tree lhs = gimple_call_lhs (stmt);
new_stmt = gimple_build_assign (lhs, build_zero_cst (TREE_TYPE (lhs)));
set_vinfo_for_stmt (new_stmt, stmt_info);
misalign = DR_MISALIGNMENT (dr);
set_ptr_info_alignment (get_ptr_info (dataref_ptr), align,
misalign);
+ tree ptr = build_int_cst (TREE_TYPE (gimple_call_arg (stmt, 1)),
+ misalign ? misalign & -misalign : align);
new_stmt
= gimple_build_call_internal (IFN_MASK_STORE, 4, dataref_ptr,
- gimple_call_arg (stmt, 1),
- vec_mask, vec_rhs);
+ ptr, vec_mask, vec_rhs);
vect_finish_stmt_generation (stmt, new_stmt, gsi);
if (i == 0)
STMT_VINFO_VEC_STMT (stmt_info) = *vec_stmt = new_stmt;
misalign = DR_MISALIGNMENT (dr);
set_ptr_info_alignment (get_ptr_info (dataref_ptr), align,
misalign);
+ tree ptr = build_int_cst (TREE_TYPE (gimple_call_arg (stmt, 1)),
+ misalign ? misalign & -misalign : align);
new_stmt
= gimple_build_call_internal (IFN_MASK_LOAD, 3, dataref_ptr,
- gimple_call_arg (stmt, 1),
- vec_mask);
+ ptr, vec_mask);
gimple_call_set_lhs (new_stmt, make_ssa_name (vec_dest));
vect_finish_stmt_generation (stmt, new_stmt, gsi);
if (i == 0)
{
/* Ensure that even with -fno-tree-dce the scalar MASK_LOAD is removed
from the IL. */
+ if (STMT_VINFO_RELATED_STMT (stmt_info))
+ {
+ stmt = STMT_VINFO_RELATED_STMT (stmt_info);
+ stmt_info = vinfo_for_stmt (stmt);
+ }
tree lhs = gimple_call_lhs (stmt);
new_stmt = gimple_build_assign (lhs, build_zero_cst (TREE_TYPE (lhs)));
set_vinfo_for_stmt (new_stmt, stmt_info);
return true;
}
+/* Return true if vector types VECTYPE_IN and VECTYPE_OUT have
+ integer elements and if we can narrow VECTYPE_IN to VECTYPE_OUT
+ in a single step. On success, store the binary pack code in
+ *CONVERT_CODE. */
+
+static bool
+simple_integer_narrowing (tree vectype_out, tree vectype_in,
+ tree_code *convert_code)
+{
+ if (!INTEGRAL_TYPE_P (TREE_TYPE (vectype_out))
+ || !INTEGRAL_TYPE_P (TREE_TYPE (vectype_in)))
+ return false;
+
+ tree_code code;
+ int multi_step_cvt = 0;
+ auto_vec <tree, 8> interm_types;
+ if (!supportable_narrowing_operation (NOP_EXPR, vectype_out, vectype_in,
+ &code, &multi_step_cvt,
+ &interm_types)
+ || multi_step_cvt)
+ return false;
+
+ *convert_code = code;
+ return true;
+}
/* Function vectorizable_call.
else
return false;
+ /* We only handle functions that do not read or clobber memory. */
+ if (gimple_vuse (stmt))
+ {
+ if (dump_enabled_p ())
+ dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
+ "function reads from or writes to memory.\n");
+ return false;
+ }
+
/* For now, we only vectorize functions if a target specific builtin
is available. TODO -- in some cases, it might be profitable to
insert the calls for pieces of the vector, in order to be able
to vectorize other operations in the loop. */
- fndecl = vectorizable_function (stmt, vectype_out, vectype_in);
- if (fndecl == NULL_TREE)
+ fndecl = NULL_TREE;
+ internal_fn ifn = IFN_LAST;
+ combined_fn cfn = gimple_call_combined_fn (stmt);
+ tree callee = gimple_call_fndecl (stmt);
+
+ /* First try using an internal function. */
+ tree_code convert_code = ERROR_MARK;
+ if (cfn != CFN_LAST
+ && (modifier == NONE
+ || (modifier == NARROW
+ && simple_integer_narrowing (vectype_out, vectype_in,
+ &convert_code))))
+ ifn = vectorizable_internal_function (cfn, callee, vectype_out,
+ vectype_in);
+
+ /* If that fails, try asking for a target-specific built-in function. */
+ if (ifn == IFN_LAST)
+ {
+ if (cfn != CFN_LAST)
+ fndecl = targetm.vectorize.builtin_vectorized_function
+ (cfn, vectype_out, vectype_in);
+ else
+ fndecl = targetm.vectorize.builtin_md_vectorized_function
+ (callee, vectype_out, vectype_in);
+ }
+
+ if (ifn == IFN_LAST && !fndecl)
{
- if (gimple_call_internal_p (stmt)
- && gimple_call_internal_fn (stmt) == IFN_GOMP_SIMD_LANE
+ if (cfn == CFN_GOMP_SIMD_LANE
&& !slp_node
&& loop_vinfo
&& LOOP_VINFO_LOOP (loop_vinfo)->simduid
}
}
- gcc_assert (!gimple_vuse (stmt));
-
if (slp_node || PURE_SLP_STMT (stmt_info))
ncopies = 1;
- else if (modifier == NARROW)
+ else if (modifier == NARROW && ifn == IFN_LAST)
ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits_out;
else
ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits_in;
dump_printf_loc (MSG_NOTE, vect_location, "=== vectorizable_call ==="
"\n");
vect_model_simple_cost (stmt_info, ncopies, dt, NULL, NULL);
+ if (ifn != IFN_LAST && modifier == NARROW && !slp_node)
+ add_stmt_cost (stmt_info->vinfo->target_cost_data, ncopies / 2,
+ vec_promote_demote, stmt_info, 0, vect_body);
+
return true;
}
vec_dest = vect_create_destination_var (scalar_dest, vectype_out);
prev_stmt_info = NULL;
- switch (modifier)
+ if (modifier == NONE || ifn != IFN_LAST)
{
- case NONE:
+ tree prev_res = NULL_TREE;
for (j = 0; j < ncopies; ++j)
{
/* Build argument list for the vectorized call. */
vec<tree> vec_oprndsk = vec_defs[k];
vargs[k] = vec_oprndsk[i];
}
- new_stmt = gimple_build_call_vec (fndecl, vargs);
- new_temp = make_ssa_name (vec_dest, new_stmt);
- gimple_call_set_lhs (new_stmt, new_temp);
+ if (modifier == NARROW)
+ {
+ tree half_res = make_ssa_name (vectype_in);
+ new_stmt = gimple_build_call_internal_vec (ifn, vargs);
+ gimple_call_set_lhs (new_stmt, half_res);
+ vect_finish_stmt_generation (stmt, new_stmt, gsi);
+ if ((i & 1) == 0)
+ {
+ prev_res = half_res;
+ continue;
+ }
+ new_temp = make_ssa_name (vec_dest);
+ new_stmt = gimple_build_assign (new_temp, convert_code,
+ prev_res, half_res);
+ }
+ else
+ {
+ if (ifn != IFN_LAST)
+ new_stmt = gimple_build_call_internal_vec (ifn, vargs);
+ else
+ new_stmt = gimple_build_call_vec (fndecl, vargs);
+ 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);
SLP_TREE_VEC_STMTS (slp_node).quick_push (new_stmt);
}
new_temp = make_ssa_name (vec_dest);
new_stmt = gimple_build_assign (new_temp, new_var);
}
+ else if (modifier == NARROW)
+ {
+ tree half_res = make_ssa_name (vectype_in);
+ new_stmt = gimple_build_call_internal_vec (ifn, vargs);
+ gimple_call_set_lhs (new_stmt, half_res);
+ vect_finish_stmt_generation (stmt, new_stmt, gsi);
+ if ((j & 1) == 0)
+ {
+ prev_res = half_res;
+ continue;
+ }
+ new_temp = make_ssa_name (vec_dest);
+ new_stmt = gimple_build_assign (new_temp, convert_code,
+ prev_res, half_res);
+ }
else
{
- new_stmt = gimple_build_call_vec (fndecl, vargs);
+ if (ifn != IFN_LAST)
+ new_stmt = gimple_build_call_internal_vec (ifn, vargs);
+ else
+ new_stmt = gimple_build_call_vec (fndecl, vargs);
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 (j == 0)
+ if (j == (modifier == NARROW ? 1 : 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);
}
-
- break;
-
- case NARROW:
+ }
+ else if (modifier == NARROW)
+ {
for (j = 0; j < ncopies; ++j)
{
/* Build argument list for the vectorized call. */
vargs.quick_push (vec_oprndsk[i]);
vargs.quick_push (vec_oprndsk[i + 1]);
}
- new_stmt = gimple_build_call_vec (fndecl, vargs);
+ if (ifn != IFN_LAST)
+ new_stmt = gimple_build_call_internal_vec (ifn, vargs);
+ else
+ new_stmt = gimple_build_call_vec (fndecl, vargs);
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);
}
*vec_stmt = STMT_VINFO_VEC_STMT (stmt_info);
-
- break;
-
- case WIDEN:
- /* No current target implements this case. */
- return false;
}
+ else
+ /* No current target implements this case. */
+ return false;
vargs.release ();
&& SCALAR_FLOAT_TYPE_P (rhs_type))))
return false;
- if ((INTEGRAL_TYPE_P (lhs_type)
- && (TYPE_PRECISION (lhs_type)
- != GET_MODE_PRECISION (TYPE_MODE (lhs_type))))
- || (INTEGRAL_TYPE_P (rhs_type)
- && (TYPE_PRECISION (rhs_type)
- != GET_MODE_PRECISION (TYPE_MODE (rhs_type)))))
+ if (!VECTOR_BOOLEAN_TYPE_P (vectype_out)
+ && ((INTEGRAL_TYPE_P (lhs_type)
+ && (TYPE_PRECISION (lhs_type)
+ != GET_MODE_PRECISION (TYPE_MODE (lhs_type))))
+ || (INTEGRAL_TYPE_P (rhs_type)
+ && (TYPE_PRECISION (rhs_type)
+ != GET_MODE_PRECISION (TYPE_MODE (rhs_type))))))
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
return false;
}
+ if (VECTOR_BOOLEAN_TYPE_P (vectype_out)
+ && !VECTOR_BOOLEAN_TYPE_P (vectype_in))
+ {
+ if (dump_enabled_p ())
+ {
+ dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
+ "can't convert between boolean and non "
+ "boolean vectors");
+ dump_generic_expr (MSG_MISSED_OPTIMIZATION, TDF_SLIM, rhs_type);
+ dump_printf (MSG_MISSED_OPTIMIZATION, "\n");
+ }
+
+ return false;
+ }
+
nunits_in = TYPE_VECTOR_SUBPARTS (vectype_in);
nunits_out = TYPE_VECTOR_SUBPARTS (vectype_out);
if (nunits_in < nunits_out)
/* 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))))
+ && TYPE_UNSIGNED (TREE_TYPE (op)))
+ /* Conversion between boolean types of different sizes is
+ a simple assignment in case their vectypes are same
+ boolean vectors. */
+ && (!VECTOR_BOOLEAN_TYPE_P (vectype)
+ || !VECTOR_BOOLEAN_TYPE_P (vectype_in)))
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
/* 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))))
+ if (!VECTOR_BOOLEAN_TYPE_P (vectype_out)
+ && (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
group. */
vec_num = SLP_TREE_NUMBER_OF_VEC_STMTS (slp_node);
first_stmt = SLP_TREE_SCALAR_STMTS (slp_node)[0];
+ gcc_assert (GROUP_FIRST_ELEMENT (vinfo_for_stmt (first_stmt)) == first_stmt);
first_dr = STMT_VINFO_DATA_REF (vinfo_for_stmt (first_stmt));
op = gimple_assign_rhs1 (first_stmt);
}
bool grouped_load = false;
bool load_lanes_p = false;
gimple *first_stmt;
+ gimple *first_stmt_for_drptr = NULL;
bool inv_p;
bool negative = false;
bool compute_in_loop = false;
that leaves unused vector loads around punt - we at least create
very sub-optimal code in that case (and blow up memory,
see PR65518). */
+ bool force_peeling = false;
if (first_stmt == stmt
- && !GROUP_NEXT_ELEMENT (stmt_info)
- && GROUP_SIZE (stmt_info) > TYPE_VECTOR_SUBPARTS (vectype))
+ && !GROUP_NEXT_ELEMENT (stmt_info))
+ {
+ if (GROUP_SIZE (stmt_info) > TYPE_VECTOR_SUBPARTS (vectype))
+ {
+ if (dump_enabled_p ())
+ dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
+ "single-element interleaving not supported "
+ "for not adjacent vector loads\n");
+ return false;
+ }
+
+ /* Single-element interleaving requires peeling for gaps. */
+ force_peeling = true;
+ }
+
+ /* If there is a gap in the end of the group or the group size cannot
+ be made a multiple of the vector element count then we access excess
+ elements in the last iteration and thus need to peel that off. */
+ if (loop_vinfo
+ && ! STMT_VINFO_STRIDED_P (stmt_info)
+ && (force_peeling
+ || GROUP_GAP (vinfo_for_stmt (first_stmt)) != 0
+ || (!slp && vf % GROUP_SIZE (vinfo_for_stmt (first_stmt)) != 0)))
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
- "single-element interleaving not supported "
- "for not adjacent vector loads\n");
- return false;
+ "Data access with gaps requires scalar "
+ "epilogue loop\n");
+ if (loop->inner)
+ {
+ if (dump_enabled_p ())
+ dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
+ "Peeling for outer loop is not supported\n");
+ return false;
+ }
+
+ LOOP_VINFO_PEELING_FOR_GAPS (loop_vinfo) = true;
}
if (slp && SLP_TREE_LOAD_PERMUTATION (slp_node).exists ())
if (grouped_load)
{
first_stmt = GROUP_FIRST_ELEMENT (stmt_info);
- if (slp
- && !SLP_TREE_LOAD_PERMUTATION (slp_node).exists ()
- && first_stmt != SLP_TREE_SCALAR_STMTS (slp_node)[0])
- first_stmt = SLP_TREE_SCALAR_STMTS (slp_node)[0];
+ /* For SLP vectorization we directly vectorize a subchain
+ without permutation. */
+ if (slp && ! SLP_TREE_LOAD_PERMUTATION (slp_node).exists ())
+ first_stmt = SLP_TREE_SCALAR_STMTS (slp_node)[0];
+ /* For BB vectorization always use the first stmt to base
+ the data ref pointer on. */
+ if (bb_vinfo)
+ first_stmt_for_drptr = 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))
(DR_REF (first_dr)), 0);
inv_p = false;
}
+ else if (first_stmt_for_drptr
+ && first_stmt != first_stmt_for_drptr)
+ {
+ dataref_ptr
+ = vect_create_data_ref_ptr (first_stmt_for_drptr, aggr_type,
+ at_loop, offset, &dummy, gsi,
+ &ptr_incr, simd_lane_access_p,
+ &inv_p, byte_offset);
+ /* Adjust the pointer by the difference to first_stmt. */
+ data_reference_p ptrdr
+ = STMT_VINFO_DATA_REF (vinfo_for_stmt (first_stmt_for_drptr));
+ tree diff = fold_convert (sizetype,
+ size_binop (MINUS_EXPR,
+ DR_INIT (first_dr),
+ DR_INIT (ptrdr)));
+ dataref_ptr = bump_vector_ptr (dataref_ptr, ptr_incr, gsi,
+ stmt, diff);
+ }
else
dataref_ptr
= vect_create_data_ref_ptr (first_stmt, aggr_type, at_loop,
unshare_expr
(gimple_assign_rhs1 (stmt))));
new_temp = vect_init_vector (stmt, tem, vectype, NULL);
+ new_stmt = SSA_NAME_DEF_STMT (new_temp);
+ set_vinfo_for_stmt (new_stmt,
+ new_stmt_vec_info (new_stmt, vinfo));
}
else
{
gsi_next (&gsi2);
new_temp = vect_init_vector (stmt, scalar_dest,
vectype, &gsi2);
+ new_stmt = SSA_NAME_DEF_STMT (new_temp);
}
- new_stmt = SSA_NAME_DEF_STMT (new_temp);
- set_vinfo_for_stmt (new_stmt,
- new_stmt_vec_info (new_stmt, vinfo));
}
if (negative)
enum vect_def_type dt;
tree vectype1 = NULL_TREE, vectype2 = NULL_TREE;
+ /* Mask case. */
+ if (TREE_CODE (cond) == SSA_NAME
+ && TREE_CODE (TREE_TYPE (cond)) == BOOLEAN_TYPE)
+ {
+ gimple *lhs_def_stmt = SSA_NAME_DEF_STMT (cond);
+ if (!vect_is_simple_use (cond, vinfo, &lhs_def_stmt,
+ &dt, comp_vectype)
+ || !*comp_vectype
+ || !VECTOR_BOOLEAN_TYPE_P (*comp_vectype))
+ return false;
+ return true;
+ }
+
if (!COMPARISON_CLASS_P (cond))
return false;
vec<tree> vec_oprnds2 = vNULL;
vec<tree> vec_oprnds3 = vNULL;
tree vec_cmp_type;
+ bool masked = false;
if (reduc_index && STMT_SLP_TYPE (stmt_info))
return false;
if (!vect_is_simple_use (else_clause, stmt_info->vinfo, &def_stmt, &dt))
return false;
- vec_cmp_type = build_same_sized_truth_vector_type (comp_vectype);
+ if (VECTOR_BOOLEAN_TYPE_P (comp_vectype))
+ {
+ vec_cmp_type = comp_vectype;
+ masked = true;
+ }
+ else
+ vec_cmp_type = build_same_sized_truth_vector_type (comp_vectype);
if (vec_cmp_type == NULL_TREE)
return false;
auto_vec<tree, 4> ops;
auto_vec<vec<tree>, 4> vec_defs;
- ops.safe_push (TREE_OPERAND (cond_expr, 0));
- ops.safe_push (TREE_OPERAND (cond_expr, 1));
+ if (masked)
+ ops.safe_push (cond_expr);
+ else
+ {
+ ops.safe_push (TREE_OPERAND (cond_expr, 0));
+ ops.safe_push (TREE_OPERAND (cond_expr, 1));
+ }
ops.safe_push (then_clause);
ops.safe_push (else_clause);
vect_get_slp_defs (ops, slp_node, &vec_defs, -1);
vec_oprnds3 = vec_defs.pop ();
vec_oprnds2 = vec_defs.pop ();
- vec_oprnds1 = vec_defs.pop ();
+ if (!masked)
+ vec_oprnds1 = vec_defs.pop ();
vec_oprnds0 = vec_defs.pop ();
ops.release ();
else
{
gimple *gtemp;
- vec_cond_lhs =
- vect_get_vec_def_for_operand (TREE_OPERAND (cond_expr, 0),
- stmt, comp_vectype);
- vect_is_simple_use (TREE_OPERAND (cond_expr, 0),
- loop_vinfo, >emp, &dts[0]);
-
- vec_cond_rhs =
- vect_get_vec_def_for_operand (TREE_OPERAND (cond_expr, 1),
- stmt, comp_vectype);
- vect_is_simple_use (TREE_OPERAND (cond_expr, 1),
- loop_vinfo, >emp, &dts[1]);
+ if (masked)
+ {
+ vec_cond_lhs
+ = vect_get_vec_def_for_operand (cond_expr, stmt,
+ comp_vectype);
+ vect_is_simple_use (cond_expr, stmt_info->vinfo,
+ >emp, &dts[0]);
+ }
+ else
+ {
+ vec_cond_lhs =
+ vect_get_vec_def_for_operand (TREE_OPERAND (cond_expr, 0),
+ stmt, comp_vectype);
+ vect_is_simple_use (TREE_OPERAND (cond_expr, 0),
+ loop_vinfo, >emp, &dts[0]);
+
+ vec_cond_rhs =
+ vect_get_vec_def_for_operand (TREE_OPERAND (cond_expr, 1),
+ stmt, comp_vectype);
+ vect_is_simple_use (TREE_OPERAND (cond_expr, 1),
+ loop_vinfo, >emp, &dts[1]);
+ }
if (reduc_index == 1)
vec_then_clause = reduc_def;
else
}
else
{
- vec_cond_lhs = vect_get_vec_def_for_stmt_copy (dts[0],
- vec_oprnds0.pop ());
- vec_cond_rhs = vect_get_vec_def_for_stmt_copy (dts[1],
- vec_oprnds1.pop ());
+ vec_cond_lhs
+ = vect_get_vec_def_for_stmt_copy (dts[0],
+ vec_oprnds0.pop ());
+ if (!masked)
+ vec_cond_rhs
+ = vect_get_vec_def_for_stmt_copy (dts[1],
+ vec_oprnds1.pop ());
+
vec_then_clause = vect_get_vec_def_for_stmt_copy (dts[2],
vec_oprnds2.pop ());
vec_else_clause = vect_get_vec_def_for_stmt_copy (dts[3],
if (!slp_node)
{
vec_oprnds0.quick_push (vec_cond_lhs);
- vec_oprnds1.quick_push (vec_cond_rhs);
+ if (!masked)
+ vec_oprnds1.quick_push (vec_cond_rhs);
vec_oprnds2.quick_push (vec_then_clause);
vec_oprnds3.quick_push (vec_else_clause);
}
/* Arguments are ready. Create the new vector stmt. */
FOR_EACH_VEC_ELT (vec_oprnds0, i, vec_cond_lhs)
{
- vec_cond_rhs = vec_oprnds1[i];
vec_then_clause = vec_oprnds2[i];
vec_else_clause = vec_oprnds3[i];
- vec_compare = build2 (TREE_CODE (cond_expr), vec_cmp_type,
- vec_cond_lhs, vec_cond_rhs);
+ if (masked)
+ vec_compare = vec_cond_lhs;
+ else
+ {
+ vec_cond_rhs = vec_oprnds1[i];
+ vec_compare = build2 (TREE_CODE (cond_expr), vec_cmp_type,
+ vec_cond_lhs, vec_cond_rhs);
+ }
vec_cond_expr = build3 (VEC_COND_EXPR, vectype,
vec_compare, vec_then_clause, vec_else_clause);
tree mask_type;
tree mask;
+ if (!STMT_VINFO_RELEVANT_P (stmt_info) && !bb_vinfo)
+ return false;
+
if (!VECTOR_BOOLEAN_TYPE_P (vectype))
return false;
ncopies = LOOP_VINFO_VECT_FACTOR (loop_vinfo) / nunits;
gcc_assert (ncopies >= 1);
- if (!STMT_VINFO_RELEVANT_P (stmt_info) && !bb_vinfo)
- return false;
-
if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def
&& !(STMT_VINFO_DEF_TYPE (stmt_info) == vect_nested_cycle
&& reduc_def))
}
else
{
- vec_rhs1 = vect_get_vec_def_for_operand (rhs1, stmt, NULL);
- vec_rhs2 = vect_get_vec_def_for_operand (rhs2, stmt, NULL);
+ vec_rhs1 = vect_get_vec_def_for_operand (rhs1, stmt, vectype);
+ vec_rhs2 = vect_get_vec_def_for_operand (rhs2, stmt, vectype);
}
}
else
gimple *patt_stmt = STMT_VINFO_STMT (patt_info);
gimple_set_bb (patt_stmt, NULL);
tree lhs = gimple_get_lhs (patt_stmt);
- if (TREE_CODE (lhs) == SSA_NAME)
+ if (lhs && TREE_CODE (lhs) == SSA_NAME)
release_ssa_name (lhs);
if (seq)
{
gimple *seq_stmt = gsi_stmt (si);
gimple_set_bb (seq_stmt, NULL);
lhs = gimple_get_lhs (seq_stmt);
- if (TREE_CODE (lhs) == SSA_NAME)
+ if (lhs && TREE_CODE (lhs) == SSA_NAME)
release_ssa_name (lhs);
free_stmt_vec_info (seq_stmt);
}
else
{
stmt_vec_info stmt_vinfo = vinfo_for_stmt (*def_stmt);
- if (is_a <bb_vec_info> (vinfo) && !STMT_VINFO_VECTORIZABLE (stmt_vinfo))
- *dt = vect_external_def;
- else
- *dt = STMT_VINFO_DEF_TYPE (stmt_vinfo);
+ *dt = STMT_VINFO_DEF_TYPE (stmt_vinfo);
}
if (dump_enabled_p ())