for a memory access in the given MODE. */
static bool
-should_replace_address (rtx old, rtx new, enum machine_mode mode)
+should_replace_address (rtx old_rtx, rtx new_rtx, enum machine_mode mode,
+ bool speed)
{
int gain;
- if (rtx_equal_p (old, new) || !memory_address_p (mode, new))
+ if (rtx_equal_p (old_rtx, new_rtx) || !memory_address_p (mode, new_rtx))
return false;
/* Copy propagation is always ok. */
- if (REG_P (old) && REG_P (new))
+ if (REG_P (old_rtx) && REG_P (new_rtx))
return true;
/* Prefer the new address if it is less expensive. */
- gain = address_cost (old, mode) - address_cost (new, mode);
+ gain = address_cost (old_rtx, mode, speed) - address_cost (new_rtx, mode, speed);
/* If the addresses have equivalent cost, prefer the new address
if it has the highest `rtx_cost'. That has the potential of
eliminating the most insns without additional costs, and it
is the same that cse.c used to do. */
if (gain == 0)
- gain = rtx_cost (new, SET) - rtx_cost (old, SET);
+ gain = rtx_cost (new_rtx, SET, speed) - rtx_cost (old_rtx, SET, speed);
return (gain > 0);
}
+
+/* Flags for the last parameter of propagate_rtx_1. */
+
+enum {
+ /* If PR_CAN_APPEAR is true, propagate_rtx_1 always returns true;
+ if it is false, propagate_rtx_1 returns false if, for at least
+ one occurrence OLD, it failed to collapse the result to a constant.
+ For example, (mult:M (reg:M A) (minus:M (reg:M B) (reg:M A))) may
+ collapse to zero if replacing (reg:M B) with (reg:M A).
+
+ PR_CAN_APPEAR is disregarded inside MEMs: in that case,
+ propagate_rtx_1 just tries to make cheaper and valid memory
+ addresses. */
+ PR_CAN_APPEAR = 1,
+
+ /* If PR_HANDLE_MEM is not set, propagate_rtx_1 won't attempt any replacement
+ outside memory addresses. This is needed because propagate_rtx_1 does
+ not do any analysis on memory; thus it is very conservative and in general
+ it will fail if non-read-only MEMs are found in the source expression.
+
+ PR_HANDLE_MEM is set when the source of the propagation was not
+ another MEM. Then, it is safe not to treat non-read-only MEMs as
+ ``opaque'' objects. */
+ PR_HANDLE_MEM = 2,
+
+ /* Set when costs should be optimized for speed. */
+ PR_OPTIMIZE_FOR_SPEED = 4
+};
+
+
/* Replace all occurrences of OLD in *PX with NEW and try to simplify the
resulting expression. Replace *PX with a new RTL expression if an
occurrence of OLD was found.
- If CAN_APPEAR is true, we always return true; if it is false, we
- can return false if, for at least one occurrence OLD, we failed to
- collapse the result to a constant. For example, (mult:M (reg:M A)
- (minus:M (reg:M B) (reg:M A))) may collapse to zero if replacing
- (reg:M B) with (reg:M A).
-
- CAN_APPEAR is disregarded inside MEMs: in that case, we always return
- true if the simplification is a cheaper and valid memory address.
-
This is only a wrapper around simplify-rtx.c: do not add any pattern
matching code here. (The sole exception is the handling of LO_SUM, but
that is because there is no simplify_gen_* function for LO_SUM). */
static bool
-propagate_rtx_1 (rtx *px, rtx old, rtx new, bool can_appear)
+propagate_rtx_1 (rtx *px, rtx old_rtx, rtx new_rtx, int flags)
{
rtx x = *px, tem = NULL_RTX, op0, op1, op2;
enum rtx_code code = GET_CODE (x);
enum machine_mode mode = GET_MODE (x);
enum machine_mode op_mode;
+ bool can_appear = (flags & PR_CAN_APPEAR) != 0;
bool valid_ops = true;
- /* If X is OLD_RTX, return NEW_RTX. Otherwise, if this is an expression,
- try to build a new expression from recursive substitution. */
+ if (!(flags & PR_HANDLE_MEM) && MEM_P (x) && !MEM_READONLY_P (x))
+ {
+ /* If unsafe, change MEMs to CLOBBERs or SCRATCHes (to preserve whether
+ they have side effects or not). */
+ *px = (side_effects_p (x)
+ ? gen_rtx_CLOBBER (GET_MODE (x), const0_rtx)
+ : gen_rtx_SCRATCH (GET_MODE (x)));
+ return false;
+ }
- if (x == old)
+ /* If X is OLD_RTX, return NEW_RTX. But not if replacing only within an
+ address, and we are *not* inside one. */
+ if (x == old_rtx)
{
- *px = new;
+ *px = new_rtx;
return can_appear;
}
+ /* If this is an expression, try recursive substitution. */
switch (GET_RTX_CLASS (code))
{
case RTX_UNARY:
op0 = XEXP (x, 0);
op_mode = GET_MODE (op0);
- valid_ops &= propagate_rtx_1 (&op0, old, new, can_appear);
+ valid_ops &= propagate_rtx_1 (&op0, old_rtx, new_rtx, flags);
if (op0 == XEXP (x, 0))
return true;
tem = simplify_gen_unary (code, mode, op0, op_mode);
case RTX_COMM_ARITH:
op0 = XEXP (x, 0);
op1 = XEXP (x, 1);
- valid_ops &= propagate_rtx_1 (&op0, old, new, can_appear);
- valid_ops &= propagate_rtx_1 (&op1, old, new, can_appear);
+ valid_ops &= propagate_rtx_1 (&op0, old_rtx, new_rtx, flags);
+ valid_ops &= propagate_rtx_1 (&op1, old_rtx, new_rtx, flags);
if (op0 == XEXP (x, 0) && op1 == XEXP (x, 1))
return true;
tem = simplify_gen_binary (code, mode, op0, op1);
op0 = XEXP (x, 0);
op1 = XEXP (x, 1);
op_mode = GET_MODE (op0) != VOIDmode ? GET_MODE (op0) : GET_MODE (op1);
- valid_ops &= propagate_rtx_1 (&op0, old, new, can_appear);
- valid_ops &= propagate_rtx_1 (&op1, old, new, can_appear);
+ valid_ops &= propagate_rtx_1 (&op0, old_rtx, new_rtx, flags);
+ valid_ops &= propagate_rtx_1 (&op1, old_rtx, new_rtx, flags);
if (op0 == XEXP (x, 0) && op1 == XEXP (x, 1))
return true;
tem = simplify_gen_relational (code, mode, op_mode, op0, op1);
op1 = XEXP (x, 1);
op2 = XEXP (x, 2);
op_mode = GET_MODE (op0);
- valid_ops &= propagate_rtx_1 (&op0, old, new, can_appear);
- valid_ops &= propagate_rtx_1 (&op1, old, new, can_appear);
- valid_ops &= propagate_rtx_1 (&op2, old, new, can_appear);
+ valid_ops &= propagate_rtx_1 (&op0, old_rtx, new_rtx, flags);
+ valid_ops &= propagate_rtx_1 (&op1, old_rtx, new_rtx, flags);
+ valid_ops &= propagate_rtx_1 (&op2, old_rtx, new_rtx, flags);
if (op0 == XEXP (x, 0) && op1 == XEXP (x, 1) && op2 == XEXP (x, 2))
return true;
if (op_mode == VOIDmode)
if (code == SUBREG)
{
op0 = XEXP (x, 0);
- valid_ops &= propagate_rtx_1 (&op0, old, new, can_appear);
+ valid_ops &= propagate_rtx_1 (&op0, old_rtx, new_rtx, flags);
if (op0 == XEXP (x, 0))
return true;
tem = simplify_gen_subreg (mode, op0, GET_MODE (SUBREG_REG (x)),
break;
case RTX_OBJ:
- if (code == MEM && x != new)
+ if (code == MEM && x != new_rtx)
{
rtx new_op0;
op0 = XEXP (x, 0);
return true;
op0 = new_op0 = targetm.delegitimize_address (op0);
- valid_ops &= propagate_rtx_1 (&new_op0, old, new, true);
+ valid_ops &= propagate_rtx_1 (&new_op0, old_rtx, new_rtx,
+ flags | PR_CAN_APPEAR);
/* Dismiss transformation that we do not want to carry on. */
if (!valid_ops
canonicalize_address (new_op0);
/* Copy propagations are always ok. Otherwise check the costs. */
- if (!(REG_P (old) && REG_P (new))
- && !should_replace_address (op0, new_op0, GET_MODE (x)))
+ if (!(REG_P (old_rtx) && REG_P (new_rtx))
+ && !should_replace_address (op0, new_op0, GET_MODE (x),
+ flags & PR_OPTIMIZE_FOR_SPEED))
return true;
tem = replace_equiv_address_nv (x, new_op0);
/* The only simplification we do attempts to remove references to op0
or make it constant -- in both cases, op0's invalidity will not
make the result invalid. */
- propagate_rtx_1 (&op0, old, new, true);
- valid_ops &= propagate_rtx_1 (&op1, old, new, can_appear);
+ propagate_rtx_1 (&op0, old_rtx, new_rtx, flags | PR_CAN_APPEAR);
+ valid_ops &= propagate_rtx_1 (&op1, old_rtx, new_rtx, flags);
if (op0 == XEXP (x, 0) && op1 == XEXP (x, 1))
return true;
else if (code == REG)
{
- if (rtx_equal_p (x, old))
+ if (rtx_equal_p (x, old_rtx))
{
- *px = new;
+ *px = new_rtx;
return can_appear;
}
}
return valid_ops || can_appear || CONSTANT_P (tem);
}
+
+/* for_each_rtx traversal function that returns 1 if BODY points to
+ a non-constant mem. */
+
+static int
+varying_mem_p (rtx *body, void *data ATTRIBUTE_UNUSED)
+{
+ rtx x = *body;
+ return MEM_P (x) && !MEM_READONLY_P (x);
+}
+
+
/* Replace all occurrences of OLD in X with NEW and try to simplify the
resulting expression (in mode MODE). Return a new expression if it is
a constant, otherwise X.
Otherwise, we accept simplifications that have a lower or equal cost. */
static rtx
-propagate_rtx (rtx x, enum machine_mode mode, rtx old, rtx new)
+propagate_rtx (rtx x, enum machine_mode mode, rtx old_rtx, rtx new_rtx,
+ bool speed)
{
rtx tem;
bool collapsed;
+ int flags;
- if (REG_P (new) && REGNO (new) < FIRST_PSEUDO_REGISTER)
+ if (REG_P (new_rtx) && REGNO (new_rtx) < FIRST_PSEUDO_REGISTER)
return NULL_RTX;
- new = copy_rtx (new);
+ flags = 0;
+ if (REG_P (new_rtx) || CONSTANT_P (new_rtx))
+ flags |= PR_CAN_APPEAR;
+ if (!for_each_rtx (&new_rtx, varying_mem_p, NULL))
+ flags |= PR_HANDLE_MEM;
+
+ if (speed)
+ flags |= PR_OPTIMIZE_FOR_SPEED;
tem = x;
- collapsed = propagate_rtx_1 (&tem, old, new, REG_P (new) || CONSTANT_P (new));
+ collapsed = propagate_rtx_1 (&tem, old_rtx, copy_rtx (new_rtx), flags);
if (tem == x || !collapsed)
return NULL_RTX;
return true;
/* Check if the reg in USE has only one definition. We already
- know that this definition reaches use, or we wouldn't be here. */
+ know that this definition reaches use, or we wouldn't be here.
+ However, this is invalid for hard registers because if they are
+ live at the beginning of the function it does not mean that we
+ have an uninitialized access. */
regno = DF_REF_REGNO (use);
def = DF_REG_DEF_CHAIN (regno);
- if (def && (def->next_reg == NULL))
+ if (def
+ && def->next_reg == NULL
+ && regno >= FIRST_PSEUDO_REGISTER)
return false;
/* Check locally if we are in the same basic block. */
/* See if USE is killed between DEF_INSN and the last insn in the
basic block containing DEF_INSN. */
x = df_bb_regno_last_def_find (def_bb, regno);
- if (x && DF_INSN_LUID (x->insn) >= DF_INSN_LUID (def_insn))
+ if (x && DF_INSN_LUID (DF_REF_INSN (x)) >= DF_INSN_LUID (def_insn))
return true;
/* See if USE is killed between TARGET_INSN and the first insn in the
basic block containing TARGET_INSN. */
x = df_bb_regno_first_def_find (target_bb, regno);
- if (x && DF_INSN_LUID (x->insn) < DF_INSN_LUID (target_insn))
+ if (x && DF_INSN_LUID (DF_REF_INSN (x)) < DF_INSN_LUID (target_insn))
return true;
return false;
}
-/* for_each_rtx traversal function that returns 1 if BODY points to
- a non-constant mem. */
-
-static int
-varying_mem_p (rtx *body, void *data ATTRIBUTE_UNUSED)
-{
- rtx x = *body;
- return MEM_P (x) && !MEM_READONLY_P (x);
-}
-
/* Check if all uses in DEF_INSN can be used in TARGET_INSN. This
would require full computation of available expressions;
we check only restricted conditions, see use_killed_between. */
all_uses_available_at (rtx def_insn, rtx target_insn)
{
struct df_ref **use_rec;
+ struct df_insn_info *insn_info = DF_INSN_INFO_GET (def_insn);
rtx def_set = single_set (def_insn);
gcc_assert (def_set);
/* If the insn uses the reg that it defines, the substitution is
invalid. */
- for (use_rec = DF_INSN_USES (def_insn); *use_rec; use_rec++)
+ for (use_rec = DF_INSN_INFO_USES (insn_info); *use_rec; use_rec++)
{
struct df_ref *use = *use_rec;
if (rtx_equal_p (DF_REF_REG (use), def_reg))
return false;
}
- for (use_rec = DF_INSN_EQ_USES (def_insn); *use_rec; use_rec++)
+ for (use_rec = DF_INSN_INFO_EQ_USES (insn_info); *use_rec; use_rec++)
{
struct df_ref *use = *use_rec;
if (rtx_equal_p (use->reg, def_reg))
{
/* Look at all the uses of DEF_INSN, and see if they are not
killed between DEF_INSN and TARGET_INSN. */
- for (use_rec = DF_INSN_USES (def_insn); *use_rec; use_rec++)
+ for (use_rec = DF_INSN_INFO_USES (insn_info); *use_rec; use_rec++)
{
struct df_ref *use = *use_rec;
if (use_killed_between (use, def_insn, target_insn))
return false;
}
- for (use_rec = DF_INSN_EQ_USES (def_insn); *use_rec; use_rec++)
+ for (use_rec = DF_INSN_INFO_EQ_USES (insn_info); *use_rec; use_rec++)
{
struct df_ref *use = *use_rec;
if (use_killed_between (use, def_insn, target_insn))
}
}
- /* We don't do any analysis of memories or aliasing. Reject any
- instruction that involves references to non-constant memory. */
- return !for_each_rtx (&SET_SRC (def_set), varying_mem_p, NULL);
+ return true;
}
\f
struct df_ref *orig_use = use, *new_use;
int width = -1;
int offset = -1;
+ enum machine_mode mode = 0;
rtx *new_loc = find_occurrence (loc, DF_REF_REG (orig_use));
use_rec++;
if (DF_REF_FLAGS_IS_SET (orig_use, DF_REF_SIGN_EXTRACT | DF_REF_ZERO_EXTRACT))
{
- width = DF_REF_WIDTH (orig_use);
- offset = DF_REF_OFFSET (orig_use);
+ width = DF_REF_EXTRACT_WIDTH (orig_use);
+ offset = DF_REF_EXTRACT_OFFSET (orig_use);
+ mode = DF_REF_EXTRACT_MODE (orig_use);
}
/* Add a new insn use. Use the original type, because it says if the
use was within a MEM. */
new_use = df_ref_create (DF_REF_REG (orig_use), new_loc,
insn, BLOCK_FOR_INSN (insn),
- type, DF_REF_FLAGS (orig_use) | new_flags, width, offset);
+ type, DF_REF_FLAGS (orig_use) | new_flags,
+ width, offset, mode);
/* Set up the use-def chain. */
df_chain_copy (new_use, DF_REF_CHAIN (orig_use));
performed. */
static bool
-try_fwprop_subst (struct df_ref *use, rtx *loc, rtx new, rtx def_insn, bool set_reg_equal)
+try_fwprop_subst (struct df_ref *use, rtx *loc, rtx new_rtx, rtx def_insn, bool set_reg_equal)
{
rtx insn = DF_REF_INSN (use);
enum df_ref_type type = DF_REF_TYPE (use);
int flags = DF_REF_FLAGS (use);
rtx set = single_set (insn);
- int old_cost = rtx_cost (SET_SRC (set), SET);
+ bool speed = optimize_bb_for_speed_p (BLOCK_FOR_INSN (insn));
+ int old_cost = rtx_cost (SET_SRC (set), SET, speed);
bool ok;
if (dump_file)
fprintf (dump_file, "\nIn insn %d, replacing\n ", INSN_UID (insn));
print_inline_rtx (dump_file, *loc, 2);
fprintf (dump_file, "\n with ");
- print_inline_rtx (dump_file, new, 2);
+ print_inline_rtx (dump_file, new_rtx, 2);
fprintf (dump_file, "\n");
}
- validate_unshare_change (insn, loc, new, true);
+ validate_unshare_change (insn, loc, new_rtx, true);
if (!verify_changes (0))
{
if (dump_file)
}
else if (DF_REF_TYPE (use) == DF_REF_REG_USE
- && rtx_cost (SET_SRC (set), SET) > old_cost)
+ && rtx_cost (SET_SRC (set), SET, speed) > old_cost)
{
if (dump_file)
fprintf (dump_file, "Changes to insn %d not profitable\n",
num_changes++;
df_ref_remove (use);
- if (!CONSTANT_P (new))
+ if (!CONSTANT_P (new_rtx))
{
- update_df (insn, loc, DF_INSN_USES (def_insn), type, flags);
- update_df (insn, loc, DF_INSN_EQ_USES (def_insn), type, flags);
+ struct df_insn_info *insn_info = DF_INSN_INFO_GET (def_insn);
+ update_df (insn, loc, DF_INSN_INFO_USES (insn_info), type, flags);
+ update_df (insn, loc, DF_INSN_INFO_EQ_USES (insn_info), type, flags);
}
}
else
cancel_changes (0);
/* Can also record a simplified value in a REG_EQUAL note,
- making a new one if one does not already exist.
- Don't do this if the insn has a REG_RETVAL note, because the
- combined presence means that the REG_EQUAL note refers to the
- (full) contents of the libcall value. */
- if (set_reg_equal && !find_reg_note (insn, REG_RETVAL, NULL_RTX))
+ making a new one if one does not already exist. */
+ if (set_reg_equal)
{
if (dump_file)
fprintf (dump_file, " Setting REG_EQUAL note\n");
- set_unique_reg_note (insn, REG_EQUAL, copy_rtx (new));
+ set_unique_reg_note (insn, REG_EQUAL, copy_rtx (new_rtx));
/* ??? Is this still necessary if we add the note through
set_unique_reg_note? */
- if (!CONSTANT_P (new))
+ if (!CONSTANT_P (new_rtx))
{
- update_df (insn, loc, DF_INSN_USES (def_insn),
+ struct df_insn_info *insn_info = DF_INSN_INFO_GET (def_insn);
+ update_df (insn, loc, DF_INSN_INFO_USES (insn_info),
type, DF_REF_IN_NOTE);
- update_df (insn, loc, DF_INSN_EQ_USES (def_insn),
+ update_df (insn, loc, DF_INSN_INFO_EQ_USES (insn_info),
type, DF_REF_IN_NOTE);
}
}
{
rtx use_insn = DF_REF_INSN (use);
rtx use_set = single_set (use_insn);
- rtx src, reg, new, *loc;
+ rtx src, reg, new_rtx, *loc;
bool set_reg_equal;
enum machine_mode mode;
if (x != src)
{
rtx note = find_reg_note (use_insn, REG_EQUAL, NULL_RTX);
- rtx old = note ? XEXP (note, 0) : SET_SRC (use_set);
- rtx new = simplify_replace_rtx (old, src, x);
- if (old != new)
- set_unique_reg_note (use_insn, REG_EQUAL, copy_rtx (new));
+ rtx old_rtx = note ? XEXP (note, 0) : SET_SRC (use_set);
+ rtx new_rtx = simplify_replace_rtx (old_rtx, src, x);
+ if (old_rtx != new_rtx)
+ set_unique_reg_note (use_insn, REG_EQUAL, copy_rtx (new_rtx));
}
return false;
}
else
mode = GET_MODE (*loc);
- new = propagate_rtx (*loc, mode, reg, src);
+ new_rtx = propagate_rtx (*loc, mode, reg, src,
+ optimize_bb_for_speed_p (BLOCK_FOR_INSN (use_insn)));
- if (!new)
+ if (!new_rtx)
return false;
- return try_fwprop_subst (use, loc, new, def_insn, set_reg_equal);
+ return try_fwprop_subst (use, loc, new_rtx, def_insn, set_reg_equal);
}
else
parent = PATTERN (use_insn);
- if (!loc_mentioned_in_p (DF_REF_LOC (use), parent))
+ if (!reg_mentioned_p (DF_REF_REG (use), parent))
return;
def_insn = DF_REF_INSN (def);
struct df_ref *use = DF_USES_GET (i);
if (use)
if (DF_REF_TYPE (use) == DF_REF_REG_USE
- || DF_REF_BB (use)->loop_father == NULL)
+ || DF_REF_BB (use)->loop_father == NULL
+ /* The outer most loop is not really a loop. */
+ || loop_outer (DF_REF_BB (use)->loop_father) == NULL)
forward_propagate_into (use);
}
return 0;
}
-struct tree_opt_pass pass_rtl_fwprop =
+struct rtl_opt_pass pass_rtl_fwprop =
{
+ {
+ RTL_PASS,
"fwprop1", /* name */
gate_fwprop, /* gate */
fwprop, /* execute */
0, /* properties_destroyed */
0, /* todo_flags_start */
TODO_df_finish | TODO_verify_rtl_sharing |
- TODO_dump_func, /* todo_flags_finish */
- 0 /* letter */
+ TODO_dump_func /* todo_flags_finish */
+ }
};
static unsigned int
struct df_ref *use = DF_USES_GET (i);
if (use)
if (DF_REF_TYPE (use) != DF_REF_REG_USE
- && DF_REF_BB (use)->loop_father != NULL)
+ && DF_REF_BB (use)->loop_father != NULL
+ /* The outer most loop is not really a loop. */
+ && loop_outer (DF_REF_BB (use)->loop_father) != NULL)
forward_propagate_into (use);
}
return 0;
}
-struct tree_opt_pass pass_rtl_fwprop_addr =
+struct rtl_opt_pass pass_rtl_fwprop_addr =
{
+ {
+ RTL_PASS,
"fwprop2", /* name */
gate_fwprop, /* gate */
fwprop_addr, /* execute */
0, /* properties_destroyed */
0, /* todo_flags_start */
TODO_df_finish | TODO_verify_rtl_sharing |
- TODO_dump_func, /* todo_flags_finish */
- 0 /* letter */
+ TODO_dump_func /* todo_flags_finish */
+ }
};