/* Optimize by combining instructions for GNU compiler.
- Copyright (C) 1987-2017 Free Software Foundation, Inc.
+ Copyright (C) 1987-2019 Free Software Foundation, Inc.
This file is part of GCC.
insn as having a logical link to the preceding insn. The same is true
for an insn explicitly using CC0.
- We check (with use_crosses_set_p) to avoid combining in such a way
+ We check (with modified_between_p) to avoid combining in such a way
as to move a computation to a place where its value would be different.
Combination is done by mathematically substituting the previous
#include "explow.h"
#include "insn-attr.h"
#include "rtlhooks-def.h"
-#include "params.h"
+#include "expr.h"
#include "tree-pass.h"
#include "valtrack.h"
#include "rtl-iter.h"
#include "print-rtl.h"
+#include "function-abi.h"
/* Number of attempts to combine instructions in this function. */
static rtx_insn *added_links_insn;
+/* And similarly, for notes. */
+
+static rtx_insn *added_notes_insn;
+
/* Basic block in which we are performing combines. */
static basic_block this_basic_block;
static bool optimize_this_for_speed_p;
static void record_dead_and_set_regs (rtx_insn *);
static int get_last_value_validate (rtx *, rtx_insn *, int, int);
static rtx get_last_value (const_rtx);
-static int use_crosses_set_p (const_rtx, int);
static void reg_dead_at_p_1 (rtx, const_rtx, void *);
static int reg_dead_at_p (rtx, rtx_insn *);
static void move_deaths (rtx, rtx, int, rtx_insn *, rtx *);
}
/* Try to split PATTERN found in INSN. This returns NULL_RTX if
- PATTERN can not be split. Otherwise, it returns an insn sequence.
+ PATTERN cannot be split. Otherwise, it returns an insn sequence.
This is a wrapper around split_insns which ensures that the
reg_stat vector is made larger if the splitter creates a new
register. */
}
-/* Delete any insns that copy a register to itself. */
+/* Delete any insns that copy a register to itself.
+ Return true if the CFG was changed. */
-static void
+static bool
delete_noop_moves (void)
{
rtx_insn *insn, *next;
basic_block bb;
+ bool edges_deleted = false;
+
FOR_EACH_BB_FN (bb, cfun)
{
for (insn = BB_HEAD (bb); insn != NEXT_INSN (BB_END (bb)); insn = next)
if (dump_file)
fprintf (dump_file, "deleting noop move %d\n", INSN_UID (insn));
- delete_insn_and_edges (insn);
+ edges_deleted |= delete_insn_and_edges (insn);
}
}
}
+
+ return edges_deleted;
}
\f
/* Main entry point for combiner. F is the first insn of the function.
NREGS is the first unused pseudo-reg number.
- Return nonzero if the combiner has turned an indirect jump
- instruction into a direct jump. */
+ Return nonzero if the CFG was changed (e.g. if the combiner has
+ turned an indirect jump instruction into a direct jump). */
static int
combine_instructions (rtx_insn *f, unsigned int nregs)
{
subst_low_luid = DF_INSN_LUID (insn);
subst_insn = insn;
- note_stores (PATTERN (insn), set_nonzero_bits_and_sign_copies,
- insn);
+ note_stores (insn, set_nonzero_bits_and_sign_copies, insn);
record_dead_and_set_regs (insn);
if (AUTO_INC_DEC)
insn);
/* Record the current insn_cost of this instruction. */
- if (NONJUMP_INSN_P (insn))
- INSN_COST (insn) = insn_cost (insn, optimize_this_for_speed_p);
+ INSN_COST (insn) = insn_cost (insn, optimize_this_for_speed_p);
if (dump_file)
{
fprintf (dump_file, "insn_cost %d for ", INSN_COST (insn));
init_reg_last ();
setup_incoming_promotions (first);
last_bb = ENTRY_BLOCK_PTR_FOR_FN (cfun);
- int max_combine = PARAM_VALUE (PARAM_MAX_COMBINE_INSNS);
+ int max_combine = param_max_combine_insns;
FOR_EACH_BB_FN (this_basic_block, cfun)
{
default_rtl_profile ();
clear_bb_flags ();
new_direct_jump_p |= purge_all_dead_edges ();
- delete_noop_moves ();
+ new_direct_jump_p |= delete_noop_moves ();
/* Clean up. */
obstack_free (&insn_link_obstack, NULL);
function lie within the current compilation unit. (This does
take into account the exporting of a function via taking its
address, and so forth.) */
- strictly_local = cgraph_node::local_info (current_function_decl)->local;
+ strictly_local
+ = cgraph_node::local_info_node (current_function_decl)->local;
/* The mode and signedness of the argument before any promotions happen
(equal to the mode of the pseudo holding it at that stage). */
/* Don't call nonzero_bits if it cannot change anything. */
if (rsp->nonzero_bits != HOST_WIDE_INT_M1U)
{
- bits = nonzero_bits (src, nonzero_bits_mode);
+ machine_mode mode = GET_MODE (x);
+ if (GET_MODE_CLASS (mode) == MODE_INT
+ && HWI_COMPUTABLE_MODE_P (mode))
+ mode = nonzero_bits_mode;
+ bits = nonzero_bits (src, mode);
if (reg_equal && bits)
- bits &= nonzero_bits (reg_equal, nonzero_bits_mode);
+ bits &= nonzero_bits (reg_equal, mode);
rsp->nonzero_bits |= bits;
}
if (rsp->sign_bit_copies != 1)
{
num = num_sign_bit_copies (SET_SRC (set), GET_MODE (x));
- if (reg_equal && num != GET_MODE_PRECISION (GET_MODE (x)))
+ if (reg_equal && maybe_ne (num, GET_MODE_PRECISION (GET_MODE (x))))
{
unsigned int numeq = num_sign_bit_copies (reg_equal, GET_MODE (x));
if (num == 0 || numeq > num)
&& (reg_used_between_p (dest, succ2, i3)
|| reg_used_between_p (dest, succ, succ2)))
|| (!succ2 && succ && reg_used_between_p (dest, succ, i3))
+ || (!succ2 && !succ && reg_used_between_p (dest, insn, i3))
|| (succ
/* SUCC and SUCC2 can be split halves from a PARALLEL; in
that case SUCC is not in the insn stream, so use SUCC2
|| (! all_adjacent
&& (((!MEM_P (src)
|| ! find_reg_note (insn, REG_EQUIV, src))
- && use_crosses_set_p (src, DF_INSN_LUID (insn)))
+ && modified_between_p (src, insn, i3))
|| (GET_CODE (src) == ASM_OPERANDS && MEM_VOLATILE_P (src))
|| GET_CODE (src) == UNSPEC_VOLATILE))
/* Don't combine across a CALL_INSN, because that would possibly
is_volatile_p = volatile_refs_p (PATTERN (insn))
? volatile_refs_p
: volatile_insn_p;
-
+
for (p = NEXT_INSN (insn); p != i3; p = NEXT_INSN (p))
if (INSN_P (p) && p != succ && p != succ2 && is_volatile_p (PATTERN (p)))
return 0;
/* If INSN contains an autoincrement or autodecrement, make sure that
register is not used between there and I3, and not already used in
I3 either. Neither must it be used in PRED or SUCC, if they exist.
- Also insist that I3 not be a jump; if it were one
- and the incremented register were spilled, we would lose. */
+ Also insist that I3 not be a jump if using LRA; if it were one
+ and the incremented register were spilled, we would lose.
+ Reload handles this correctly. */
if (AUTO_INC_DEC)
for (link = REG_NOTES (insn); link; link = XEXP (link, 1))
if (REG_NOTE_KIND (link) == REG_INC
- && (JUMP_P (i3)
+ && ((JUMP_P (i3) && targetm.lra_p ())
|| reg_used_between_p (XEXP (link, 0), insn, i3)
|| (pred != NULL_RTX
&& reg_overlap_mentioned_p (XEXP (link, 0), PATTERN (pred)))
if (REG_P (src) && REG_P (dest)
&& ((HARD_REGISTER_P (src)
&& ! TEST_HARD_REG_BIT (fixed_reg_set, REGNO (src))
- && targetm.class_likely_spilled_p (REGNO_REG_CLASS (REGNO (src))))
+#ifdef LEAF_REGISTERS
+ && ! LEAF_REGISTERS [REGNO (src)])
+#else
+ )
+#endif
|| (HARD_REGISTER_P (dest)
&& ! TEST_HARD_REG_BIT (fixed_reg_set, REGNO (dest))
&& targetm.class_likely_spilled_p (REGNO_REG_CLASS (REGNO (dest))))))
info.mask = mask;
for (p = PREV_INSN (use); info.mask && p != insn; p = PREV_INSN (p))
if (INSN_P (p))
- note_stores (PATTERN (p), likely_spilled_retval_1, &info);
+ note_stores (p, likely_spilled_retval_1, &info);
mask = info.mask;
/* Check if any of the (probably) live return value registers is
/* Don't change between modes with different underlying register sizes,
since this could lead to invalid subregs. */
- if (REGMODE_NATURAL_SIZE (mode)
- != REGMODE_NATURAL_SIZE (GET_MODE (x)))
+ if (maybe_ne (REGMODE_NATURAL_SIZE (mode),
+ REGMODE_NATURAL_SIZE (GET_MODE (x))))
return false;
regno = REGNO (x);
|| !REG_P (SET_DEST (XVECEXP (pat, 0, i))))
return false;
for ( ; i < len; i++)
- if (GET_CODE (XVECEXP (pat, 0, i)) != CLOBBER
- || XEXP (XVECEXP (pat, 0, i), 0) == const0_rtx)
- return false;
-
+ switch (GET_CODE (XVECEXP (pat, 0, i)))
+ {
+ case CLOBBER:
+ if (XEXP (XVECEXP (pat, 0, i), 0) == const0_rtx)
+ return false;
+ break;
+ default:
+ return false;
+ }
return true;
}
return true;
}
+/* Return whether X is just a single set, with the source
+ a general_operand. */
+static bool
+is_just_move (rtx x)
+{
+ if (INSN_P (x))
+ x = PATTERN (x);
+
+ return (GET_CODE (x) == SET && general_operand (SET_SRC (x), VOIDmode));
+}
+
+/* Callback function to count autoincs. */
+
+static int
+count_auto_inc (rtx, rtx, rtx, rtx, rtx, void *arg)
+{
+ (*((int *) arg))++;
+
+ return 0;
+}
+
/* Try to combine the insns I0, I1 and I2 into I3.
Here I0, I1 and I2 appear earlier than I3.
I0 and I1 can be zero; then we combine just I2 into I3, or I1 and I2 into
/* Notes that I1, I2 or I3 is a MULT operation. */
int have_mult = 0;
int swap_i2i3 = 0;
+ int split_i2i3 = 0;
int changed_i3_dest = 0;
+ bool i2_was_move = false, i3_was_move = false;
+ int n_auto_inc = 0;
int maxreg;
rtx_insn *temp_insn;
else
fprintf (dump_file, "\nTrying %d -> %d:\n",
INSN_UID (i2), INSN_UID (i3));
+
+ if (i0)
+ dump_insn_slim (dump_file, i0);
+ if (i1)
+ dump_insn_slim (dump_file, i1);
+ dump_insn_slim (dump_file, i2);
+ dump_insn_slim (dump_file, i3);
}
/* If multiple insns feed into one of I2 or I3, they can be in any
std::swap (i1, i2);
added_links_insn = 0;
+ added_notes_insn = 0;
/* First check for one important special case that the code below will
not handle. Namely, the case where I1 is zero, I2 is a PARALLEL
&& is_parallel_of_n_reg_sets (PATTERN (i2), 2)
&& can_split_parallel_of_n_reg_sets (i2, 2)
&& !reg_used_between_p (SET_DEST (XVECEXP (PATTERN (i2), 0, 0)), i2, i3)
- && !reg_used_between_p (SET_DEST (XVECEXP (PATTERN (i2), 0, 1)), i2, i3))
+ && !reg_used_between_p (SET_DEST (XVECEXP (PATTERN (i2), 0, 1)), i2, i3)
+ && !reg_set_between_p (SET_DEST (XVECEXP (PATTERN (i2), 0, 0)), i2, i3)
+ && !reg_set_between_p (SET_DEST (XVECEXP (PATTERN (i2), 0, 1)), i2, i3))
{
/* If there is no I1, there is no I0 either. */
i0 = i1;
SUBST (PATTERN (i2), XVECEXP (PATTERN (i2), 0, 1));
}
- /* Verify that I2 and I1 are valid for combining. */
- if (! can_combine_p (i2, i3, i0, i1, NULL, NULL, &i2dest, &i2src)
- || (i1 && ! can_combine_p (i1, i3, i0, NULL, i2, NULL,
- &i1dest, &i1src))
- || (i0 && ! can_combine_p (i0, i3, NULL, NULL, i1, i2,
- &i0dest, &i0src)))
+ /* Verify that I2 and maybe I1 and I0 can be combined into I3. */
+ if (!can_combine_p (i2, i3, i0, i1, NULL, NULL, &i2dest, &i2src))
+ {
+ if (dump_file && (dump_flags & TDF_DETAILS))
+ fprintf (dump_file, "Can't combine i2 into i3\n");
+ undo_all ();
+ return 0;
+ }
+ if (i1 && !can_combine_p (i1, i3, i0, NULL, i2, NULL, &i1dest, &i1src))
{
+ if (dump_file && (dump_flags & TDF_DETAILS))
+ fprintf (dump_file, "Can't combine i1 into i3\n");
undo_all ();
return 0;
}
+ if (i0 && !can_combine_p (i0, i3, NULL, NULL, i1, i2, &i0dest, &i0src))
+ {
+ if (dump_file && (dump_flags & TDF_DETAILS))
+ fprintf (dump_file, "Can't combine i0 into i3\n");
+ undo_all ();
+ return 0;
+ }
+
+ /* Record whether i2 and i3 are trivial moves. */
+ i2_was_move = is_just_move (i2);
+ i3_was_move = is_just_move (i3);
/* Record whether I2DEST is used in I2SRC and similarly for the other
cases. Knowing this will help in register status updating below. */
return 0;
}
+ /* Count how many auto_inc expressions there were in the original insns;
+ we need to have the same number in the resulting patterns. */
+
+ if (i0)
+ for_each_inc_dec (PATTERN (i0), count_auto_inc, &n_auto_inc);
+ if (i1)
+ for_each_inc_dec (PATTERN (i1), count_auto_inc, &n_auto_inc);
+ for_each_inc_dec (PATTERN (i2), count_auto_inc, &n_auto_inc);
+ for_each_inc_dec (PATTERN (i3), count_auto_inc, &n_auto_inc);
+
/* If the set in I2 needs to be kept around, we must make a copy of
PATTERN (I2), so that when we substitute I1SRC for I1DEST in
PATTERN (I2), we are only substituting for the original I1DEST, not into
{
/* Replace cc_use_loc with entire new RTX. */
SUBST (*cc_use_loc,
- gen_rtx_fmt_ee (compare_code, compare_mode,
+ gen_rtx_fmt_ee (compare_code, GET_MODE (*cc_use_loc),
newpat_dest, const0_rtx));
undobuf.other_insn = cc_use_insn;
}
/* Just replace the CC reg with a new mode. */
SUBST (XEXP (*cc_use_loc, 0), newpat_dest);
undobuf.other_insn = cc_use_insn;
- }
+ }
}
/* Now we modify the current newpat:
if (i1 && GET_CODE (newpat) != CLOBBER)
{
- /* Check that an autoincrement side-effect on I1 has not been lost.
- This happens if I1DEST is mentioned in I2 and dies there, and
- has disappeared from the new pattern. */
- if ((FIND_REG_INC_NOTE (i1, NULL_RTX) != 0
- && i1_feeds_i2_n
- && dead_or_set_p (i2, i1dest)
- && !reg_overlap_mentioned_p (i1dest, newpat))
- /* Before we can do this substitution, we must redo the test done
- above (see detailed comments there) that ensures I1DEST isn't
- mentioned in any SETs in NEWPAT that are field assignments. */
- || !combinable_i3pat (NULL, &newpat, i1dest, NULL_RTX, NULL_RTX,
- 0, 0, 0))
+ /* Before we can do this substitution, we must redo the test done
+ above (see detailed comments there) that ensures I1DEST isn't
+ mentioned in any SETs in NEWPAT that are field assignments. */
+ if (!combinable_i3pat (NULL, &newpat, i1dest, NULL_RTX, NULL_RTX,
+ 0, 0, 0))
{
undo_all ();
return 0;
if (i0 && GET_CODE (newpat) != CLOBBER)
{
- if ((FIND_REG_INC_NOTE (i0, NULL_RTX) != 0
- && ((i0_feeds_i2_n && dead_or_set_p (i2, i0dest))
- || (i0_feeds_i1_n && dead_or_set_p (i1, i0dest)))
- && !reg_overlap_mentioned_p (i0dest, newpat))
- || !combinable_i3pat (NULL, &newpat, i0dest, NULL_RTX, NULL_RTX,
- 0, 0, 0))
+ if (!combinable_i3pat (NULL, &newpat, i0dest, NULL_RTX, NULL_RTX,
+ 0, 0, 0))
{
undo_all ();
return 0;
substed_i0 = 1;
}
+ if (n_auto_inc)
+ {
+ int new_n_auto_inc = 0;
+ for_each_inc_dec (newpat, count_auto_inc, &new_n_auto_inc);
+
+ if (n_auto_inc != new_n_auto_inc)
+ {
+ if (dump_file && (dump_flags & TDF_DETAILS))
+ fprintf (dump_file, "Number of auto_inc expressions changed\n");
+ undo_all ();
+ return 0;
+ }
+ }
+
/* Fail if an autoincrement side-effect has been duplicated. Be careful
to count all the ways that I2SRC and I1SRC can be used. */
if ((FIND_REG_INC_NOTE (i2, NULL_RTX) != 0
}
else if (m_split_insn && NEXT_INSN (NEXT_INSN (m_split_insn)) == NULL_RTX
&& (next_nonnote_nondebug_insn (i2) == i3
- || ! use_crosses_set_p (PATTERN (m_split_insn), DF_INSN_LUID (i2))))
+ || !modified_between_p (PATTERN (m_split_insn), i2, i3)))
{
rtx i2set, i3set;
rtx newi3pat = PATTERN (NEXT_INSN (m_split_insn));
|| can_change_dest_mode (i2dest, added_sets_2,
GET_MODE (*split)))
&& (next_nonnote_nondebug_insn (i2) == i3
- || ! use_crosses_set_p (*split, DF_INSN_LUID (i2)))
+ || !modified_between_p (*split, i2, i3))
/* We can't overwrite I2DEST if its value is still used by
NEWPAT. */
&& ! reg_referenced_p (i2dest, newpat))
/* *SPLIT may be part of I2SRC, so make sure we have the
original expression around for later debug processing.
We should not need I2SRC any more in other cases. */
- if (MAY_HAVE_DEBUG_INSNS)
+ if (MAY_HAVE_DEBUG_BIND_INSNS)
i2src = copy_rtx (i2src);
else
i2src = NULL;
&& INTVAL (XEXP (*split, 1)) > 0
&& (i = exact_log2 (UINTVAL (XEXP (*split, 1)))) >= 0)
{
+ rtx i_rtx = gen_int_shift_amount (split_mode, i);
SUBST (*split, gen_rtx_ASHIFT (split_mode,
- XEXP (*split, 0), GEN_INT (i)));
+ XEXP (*split, 0), i_rtx));
/* Update split_code because we may not have a multiply
anymore. */
split_code = GET_CODE (*split);
&& (i = exact_log2 (UINTVAL (XEXP (XEXP (*split, 0), 1)))) >= 0)
{
rtx nsplit = XEXP (*split, 0);
+ rtx i_rtx = gen_int_shift_amount (GET_MODE (nsplit), i);
SUBST (XEXP (*split, 0), gen_rtx_ASHIFT (GET_MODE (nsplit),
- XEXP (nsplit, 0), GEN_INT (i)));
+ XEXP (nsplit, 0),
+ i_rtx));
/* Update split_code because we may not have a multiply
anymore. */
split_code = GET_CODE (*split);
&& GET_CODE (XVECEXP (newpat, 0, 1)) == SET
&& rtx_equal_p (SET_SRC (XVECEXP (newpat, 0, 1)),
XEXP (SET_SRC (XVECEXP (newpat, 0, 0)), 0))
- && ! use_crosses_set_p (SET_SRC (XVECEXP (newpat, 0, 1)),
- DF_INSN_LUID (i2))
+ && !modified_between_p (SET_SRC (XVECEXP (newpat, 0, 1)), i2, i3)
&& GET_CODE (SET_DEST (XVECEXP (newpat, 0, 1))) != ZERO_EXTRACT
&& GET_CODE (SET_DEST (XVECEXP (newpat, 0, 1))) != STRICT_LOW_PART
&& ! (temp_expr = SET_DEST (XVECEXP (newpat, 0, 1)),
(REG_P (temp_expr)
&& reg_stat[REGNO (temp_expr)].nonzero_bits != 0
- && GET_MODE_PRECISION (GET_MODE (temp_expr)) < BITS_PER_WORD
- && GET_MODE_PRECISION (GET_MODE (temp_expr)) < HOST_BITS_PER_INT
+ && known_lt (GET_MODE_PRECISION (GET_MODE (temp_expr)),
+ BITS_PER_WORD)
+ && known_lt (GET_MODE_PRECISION (GET_MODE (temp_expr)),
+ HOST_BITS_PER_INT)
&& (reg_stat[REGNO (temp_expr)].nonzero_bits
!= GET_MODE_MASK (word_mode))))
&& ! (GET_CODE (SET_DEST (XVECEXP (newpat, 0, 1))) == SUBREG
&& (temp_expr = SUBREG_REG (SET_DEST (XVECEXP (newpat, 0, 1))),
(REG_P (temp_expr)
&& reg_stat[REGNO (temp_expr)].nonzero_bits != 0
- && GET_MODE_PRECISION (GET_MODE (temp_expr)) < BITS_PER_WORD
- && GET_MODE_PRECISION (GET_MODE (temp_expr)) < HOST_BITS_PER_INT
+ && known_lt (GET_MODE_PRECISION (GET_MODE (temp_expr)),
+ BITS_PER_WORD)
+ && known_lt (GET_MODE_PRECISION (GET_MODE (temp_expr)),
+ HOST_BITS_PER_INT)
&& (reg_stat[REGNO (temp_expr)].nonzero_bits
!= GET_MODE_MASK (word_mode)))))
&& ! reg_overlap_mentioned_p (SET_DEST (XVECEXP (newpat, 0, 1)),
other insns to combine, but the destination of that SET is still live.
Also do this if we started with two insns and (at least) one of the
- resulting sets is a noop; this noop will be deleted later. */
+ resulting sets is a noop; this noop will be deleted later.
+
+ Also do this if we started with two insns neither of which was a simple
+ move. */
else if (insn_code_number < 0 && asm_noperands (newpat) < 0
&& GET_CODE (newpat) == PARALLEL
&& XVECLEN (newpat, 0) == 2
&& GET_CODE (XVECEXP (newpat, 0, 0)) == SET
&& GET_CODE (XVECEXP (newpat, 0, 1)) == SET
- && (i1 || set_noop_p (XVECEXP (newpat, 0, 0))
- || set_noop_p (XVECEXP (newpat, 0, 1)))
+ && (i1
+ || set_noop_p (XVECEXP (newpat, 0, 0))
+ || set_noop_p (XVECEXP (newpat, 0, 1))
+ || (!i2_was_move && !i3_was_move))
&& GET_CODE (SET_DEST (XVECEXP (newpat, 0, 0))) != ZERO_EXTRACT
&& GET_CODE (SET_DEST (XVECEXP (newpat, 0, 0))) != STRICT_LOW_PART
&& GET_CODE (SET_DEST (XVECEXP (newpat, 0, 1))) != ZERO_EXTRACT
one which uses any regs/memory set in between i2 and i3 can't
be first. The PARALLEL might also have been pre-existing in i3,
so we need to make sure that we won't wrongly hoist a SET to i2
- that would conflict with a death note present in there. */
- if (!use_crosses_set_p (SET_SRC (set1), DF_INSN_LUID (i2))
+ that would conflict with a death note present in there, or would
+ have its dest modified between i2 and i3. */
+ if (!modified_between_p (SET_SRC (set1), i2, i3)
&& !(REG_P (SET_DEST (set1))
&& find_reg_note (i2, REG_DEAD, SET_DEST (set1)))
&& !(GET_CODE (SET_DEST (set1)) == SUBREG
&& find_reg_note (i2, REG_DEAD,
SUBREG_REG (SET_DEST (set1))))
+ && !modified_between_p (SET_DEST (set1), i2, i3)
&& (!HAVE_cc0 || !reg_referenced_p (cc0_rtx, set0))
/* If I3 is a jump, ensure that set0 is a jump so that
we do not create invalid RTL. */
newi2pat = set1;
newpat = set0;
}
- else if (!use_crosses_set_p (SET_SRC (set0), DF_INSN_LUID (i2))
+ else if (!modified_between_p (SET_SRC (set0), i2, i3)
&& !(REG_P (SET_DEST (set0))
&& find_reg_note (i2, REG_DEAD, SET_DEST (set0)))
&& !(GET_CODE (SET_DEST (set0)) == SUBREG
&& find_reg_note (i2, REG_DEAD,
SUBREG_REG (SET_DEST (set0))))
+ && !modified_between_p (SET_DEST (set0), i2, i3)
&& (!HAVE_cc0 || !reg_referenced_p (cc0_rtx, set1))
/* If I3 is a jump, ensure that set1 is a jump so that
we do not create invalid RTL. */
}
insn_code_number = recog_for_combine (&newpat, i3, &new_i3_notes);
+
+ if (insn_code_number >= 0)
+ split_i2i3 = 1;
}
}
return 0;
}
- if (MAY_HAVE_DEBUG_INSNS)
+ if (MAY_HAVE_DEBUG_BIND_INSNS)
{
struct undo *undo;
if (swap_i2i3)
{
- rtx_insn *insn;
- struct insn_link *link;
- rtx ni2dest;
-
/* I3 now uses what used to be its destination and which is now
I2's destination. This requires us to do a few adjustments. */
PATTERN (i3) = newpat;
adjust_for_new_dest (i3);
+ }
- /* We need a LOG_LINK from I3 to I2. But we used to have one,
- so we still will.
+ if (swap_i2i3 || split_i2i3)
+ {
+ /* We might need a LOG_LINK from I3 to I2. But then we used to
+ have one, so we still will.
However, some later insn might be using I2's dest and have
- a LOG_LINK pointing at I3. We must remove this link.
- The simplest way to remove the link is to point it at I1,
- which we know will be a NOTE. */
+ a LOG_LINK pointing at I3. We should change it to point at
+ I2 instead. */
/* newi2pat is usually a SET here; however, recog_for_combine might
have added some clobbers. */
- if (GET_CODE (newi2pat) == PARALLEL)
- ni2dest = SET_DEST (XVECEXP (newi2pat, 0, 0));
- else
- ni2dest = SET_DEST (newi2pat);
-
- for (insn = NEXT_INSN (i3);
- insn && (this_basic_block->next_bb == EXIT_BLOCK_PTR_FOR_FN (cfun)
- || insn != BB_HEAD (this_basic_block->next_bb));
+ rtx x = newi2pat;
+ if (GET_CODE (x) == PARALLEL)
+ x = XVECEXP (newi2pat, 0, 0);
+
+ /* It can only be a SET of a REG or of a SUBREG of a REG. */
+ unsigned int regno = reg_or_subregno (SET_DEST (x));
+
+ bool done = false;
+ for (rtx_insn *insn = NEXT_INSN (i3);
+ !done
+ && insn
+ && NONDEBUG_INSN_P (insn)
+ && BLOCK_FOR_INSN (insn) == this_basic_block;
insn = NEXT_INSN (insn))
{
- if (NONDEBUG_INSN_P (insn)
- && reg_referenced_p (ni2dest, PATTERN (insn)))
- {
- FOR_EACH_LOG_LINK (link, insn)
- if (link->insn == i3)
- link->insn = i1;
-
- break;
- }
+ struct insn_link *link;
+ FOR_EACH_LOG_LINK (link, insn)
+ if (link->insn == i3 && link->regno == regno)
+ {
+ link->insn = i2;
+ done = true;
+ break;
+ }
}
}
if (newi2pat)
{
- if (MAY_HAVE_DEBUG_INSNS && i2scratch)
+ if (MAY_HAVE_DEBUG_BIND_INSNS && i2scratch)
propagate_for_debug (i2, last_combined_insn, i2dest, i2src,
this_basic_block);
INSN_CODE (i2) = i2_code_number;
}
else
{
- if (MAY_HAVE_DEBUG_INSNS && i2src)
+ if (MAY_HAVE_DEBUG_BIND_INSNS && i2src)
propagate_for_debug (i2, last_combined_insn, i2dest, i2src,
this_basic_block);
SET_INSN_DELETED (i2);
{
LOG_LINKS (i1) = NULL;
REG_NOTES (i1) = 0;
- if (MAY_HAVE_DEBUG_INSNS)
+ if (MAY_HAVE_DEBUG_BIND_INSNS)
propagate_for_debug (i1, last_combined_insn, i1dest, i1src,
this_basic_block);
SET_INSN_DELETED (i1);
{
LOG_LINKS (i0) = NULL;
REG_NOTES (i0) = 0;
- if (MAY_HAVE_DEBUG_INSNS)
+ if (MAY_HAVE_DEBUG_BIND_INSNS)
propagate_for_debug (i0, last_combined_insn, i0dest, i0src,
this_basic_block);
SET_INSN_DELETED (i0);
been made to this insn. The order is important, because newi2pat
can affect nonzero_bits of newpat. */
if (newi2pat)
- note_stores (newi2pat, set_nonzero_bits_and_sign_copies, NULL);
- note_stores (newpat, set_nonzero_bits_and_sign_copies, NULL);
+ note_pattern_stores (newi2pat, set_nonzero_bits_and_sign_copies, NULL);
+ note_pattern_stores (newpat, set_nonzero_bits_and_sign_copies, NULL);
}
if (undobuf.other_insn != NULL_RTX)
combine_successes++;
undo_commit ();
- if (added_links_insn
- && (newi2pat == 0 || DF_INSN_LUID (added_links_insn) < DF_INSN_LUID (i2))
- && DF_INSN_LUID (added_links_insn) < DF_INSN_LUID (i3))
- return added_links_insn;
- else
- return newi2pat ? i2 : i3;
+ rtx_insn *ret = newi2pat ? i2 : i3;
+ if (added_links_insn && DF_INSN_LUID (added_links_insn) < DF_INSN_LUID (ret))
+ ret = added_links_insn;
+ if (added_notes_insn && DF_INSN_LUID (added_notes_insn) < DF_INSN_LUID (ret))
+ ret = added_notes_insn;
+
+ return ret;
}
\f
/* Get a marker for undoing to the current state. */
}
/* If we have a PLUS whose second operand is a constant and the
- address is not valid, perhaps will can split it up using
+ address is not valid, perhaps we can split it up using
the machine-specific way to split large constants. We use
the first pseudo-reg (one of the virtual regs) as a placeholder;
it will not remain in the result. */
/* This should have produced two insns, each of which sets our
placeholder. If the source of the second is a valid address,
- we can make put both sources together and make a split point
+ we can put both sources together and make a split point
in the middle. */
if (seq
}
}
+ /* If that didn't work and we have a nested plus, like:
+ ((REG1 * CONST1) + REG2) + CONST2 and (REG1 + REG2) + CONST2
+ is valid address, try to split (REG1 * CONST1). */
+ if (GET_CODE (XEXP (XEXP (x, 0), 0)) == PLUS
+ && !OBJECT_P (XEXP (XEXP (XEXP (x, 0), 0), 0))
+ && OBJECT_P (XEXP (XEXP (XEXP (x, 0), 0), 1))
+ && ! (GET_CODE (XEXP (XEXP (XEXP (x, 0), 0), 0)) == SUBREG
+ && OBJECT_P (SUBREG_REG (XEXP (XEXP (XEXP (x, 0),
+ 0), 0)))))
+ {
+ rtx tem = XEXP (XEXP (XEXP (x, 0), 0), 0);
+ XEXP (XEXP (XEXP (x, 0), 0), 0) = reg;
+ if (memory_address_addr_space_p (GET_MODE (x), XEXP (x, 0),
+ MEM_ADDR_SPACE (x)))
+ {
+ XEXP (XEXP (XEXP (x, 0), 0), 0) = tem;
+ return &XEXP (XEXP (XEXP (x, 0), 0), 0);
+ }
+ XEXP (XEXP (XEXP (x, 0), 0), 0) = tem;
+ }
+ else if (GET_CODE (XEXP (XEXP (x, 0), 0)) == PLUS
+ && OBJECT_P (XEXP (XEXP (XEXP (x, 0), 0), 0))
+ && !OBJECT_P (XEXP (XEXP (XEXP (x, 0), 0), 1))
+ && ! (GET_CODE (XEXP (XEXP (XEXP (x, 0), 0), 1)) == SUBREG
+ && OBJECT_P (SUBREG_REG (XEXP (XEXP (XEXP (x, 0),
+ 0), 1)))))
+ {
+ rtx tem = XEXP (XEXP (XEXP (x, 0), 0), 1);
+ XEXP (XEXP (XEXP (x, 0), 0), 1) = reg;
+ if (memory_address_addr_space_p (GET_MODE (x), XEXP (x, 0),
+ MEM_ADDR_SPACE (x)))
+ {
+ XEXP (XEXP (XEXP (x, 0), 0), 1) = tem;
+ return &XEXP (XEXP (XEXP (x, 0), 0), 1);
+ }
+ XEXP (XEXP (XEXP (x, 0), 0), 1) = tem;
+ }
+
/* If that didn't work, perhaps the first operand is complex and
needs to be computed separately, so make a split point there.
This will occur on machines that just support REG + CONST
and have a constant moved through some previous computation. */
-
- else if (!OBJECT_P (XEXP (XEXP (x, 0), 0))
- && ! (GET_CODE (XEXP (XEXP (x, 0), 0)) == SUBREG
- && OBJECT_P (SUBREG_REG (XEXP (XEXP (x, 0), 0)))))
+ if (!OBJECT_P (XEXP (XEXP (x, 0), 0))
+ && ! (GET_CODE (XEXP (XEXP (x, 0), 0)) == SUBREG
+ && OBJECT_P (SUBREG_REG (XEXP (XEXP (x, 0), 0)))))
return &XEXP (XEXP (x, 0), 0);
}
is known to be on, this can be converted into a NEG of a shift. */
if (STORE_FLAG_VALUE == -1 && XEXP (SET_SRC (x), 1) == const0_rtx
&& GET_MODE (SET_SRC (x)) == GET_MODE (XEXP (SET_SRC (x), 0))
- && 1 <= (pos = exact_log2
- (nonzero_bits (XEXP (SET_SRC (x), 0),
- GET_MODE (XEXP (SET_SRC (x), 0))))))
+ && ((pos = exact_log2 (nonzero_bits (XEXP (SET_SRC (x), 0),
+ GET_MODE (XEXP (SET_SRC (x),
+ 0))))) >= 1))
{
machine_mode mode = GET_MODE (XEXP (SET_SRC (x), 0));
-
+ rtx pos_rtx = gen_int_shift_amount (mode, pos);
SUBST (SET_SRC (x),
gen_rtx_NEG (mode,
gen_rtx_LSHIFTRT (mode,
XEXP (SET_SRC (x), 0),
- GEN_INT (pos))));
+ pos_rtx)));
split = find_split_point (&SET_SRC (x), insn, true);
if (split && split != &SET_SRC (x))
break;
}
- if (len && pos >= 0
- && pos + len <= GET_MODE_PRECISION (GET_MODE (inner))
+ if (len
+ && known_subrange_p (pos, len,
+ 0, GET_MODE_PRECISION (GET_MODE (inner)))
&& is_a <scalar_int_mode> (GET_MODE (SET_SRC (x)), &mode))
{
/* For unsigned, we have a choice of a shift followed by an
{
unsigned HOST_WIDE_INT mask
= (HOST_WIDE_INT_1U << len) - 1;
+ rtx pos_rtx = gen_int_shift_amount (mode, pos);
SUBST (SET_SRC (x),
gen_rtx_AND (mode,
gen_rtx_LSHIFTRT
- (mode, gen_lowpart (mode, inner),
- GEN_INT (pos)),
+ (mode, gen_lowpart (mode, inner), pos_rtx),
gen_int_mode (mask, mode)));
split = find_split_point (&SET_SRC (x), insn, true);
}
else
{
+ int left_bits = GET_MODE_PRECISION (mode) - len - pos;
+ int right_bits = GET_MODE_PRECISION (mode) - len;
SUBST (SET_SRC (x),
gen_rtx_fmt_ee
(unsignedp ? LSHIFTRT : ASHIFTRT, mode,
gen_rtx_ASHIFT (mode,
gen_lowpart (mode, inner),
- GEN_INT (GET_MODE_PRECISION (mode)
- - len - pos)),
- GEN_INT (GET_MODE_PRECISION (mode) - len)));
+ gen_int_shift_amount (mode, left_bits)),
+ gen_int_shift_amount (mode, right_bits)));
split = find_split_point (&SET_SRC (x), insn, true);
if (split && split != &SET_SRC (x))
x = gen_rtx_CLOBBER (mode, const0_rtx);
}
else if (CONST_SCALAR_INT_P (new_rtx)
- && GET_CODE (x) == ZERO_EXTEND)
+ && (GET_CODE (x) == ZERO_EXTEND
+ || GET_CODE (x) == SIGN_EXTEND
+ || GET_CODE (x) == FLOAT
+ || GET_CODE (x) == UNSIGNED_FLOAT))
{
- x = simplify_unary_operation (ZERO_EXTEND, GET_MODE (x),
- new_rtx, GET_MODE (XEXP (x, 0)));
- gcc_assert (x);
+ x = simplify_unary_operation (GET_CODE (x), GET_MODE (x),
+ new_rtx,
+ GET_MODE (XEXP (x, 0)));
+ if (!x)
+ return gen_rtx_CLOBBER (VOIDmode, const0_rtx);
}
else
SUBST (XEXP (x, i), new_rtx);
/* If everything is a comparison, what we have is highly unlikely
to be simpler, so don't use it. */
&& ! (COMPARISON_P (x)
- && (COMPARISON_P (true_rtx) || COMPARISON_P (false_rtx))))
+ && (COMPARISON_P (true_rtx) || COMPARISON_P (false_rtx)))
+ /* Similarly, if we end up with one of the expressions the same
+ as the original, it is certainly not simpler. */
+ && ! rtx_equal_p (x, true_rtx)
+ && ! rtx_equal_p (x, false_rtx))
{
rtx cop1 = const0_rtx;
enum rtx_code cond_code = simplify_comparison (NE, &cond, &cop1);
mode, VOIDmode,
cond, cop1),
mode);
- else
- return gen_rtx_IF_THEN_ELSE (mode,
- simplify_gen_relational (cond_code,
- mode,
- VOIDmode,
- cond,
- cop1),
- true_rtx, false_rtx);
code = GET_CODE (x);
op0_mode = VOIDmode;
/* See if this can be moved to simplify_subreg. */
if (CONSTANT_P (SUBREG_REG (x))
- && subreg_lowpart_offset (mode, op0_mode) == SUBREG_BYTE (x)
+ && known_eq (subreg_lowpart_offset (mode, op0_mode), SUBREG_BYTE (x))
/* Don't call gen_lowpart if the inner mode
is VOIDmode and we cannot simplify it, as SUBREG without
inner mode is invalid. */
&& is_a <scalar_int_mode> (op0_mode, &int_op0_mode)
&& (GET_MODE_PRECISION (int_mode)
< GET_MODE_PRECISION (int_op0_mode))
- && (subreg_lowpart_offset (int_mode, int_op0_mode)
- == SUBREG_BYTE (x))
+ && known_eq (subreg_lowpart_offset (int_mode, int_op0_mode),
+ SUBREG_BYTE (x))
&& HWI_COMPUTABLE_MODE_P (int_op0_mode)
- && (nonzero_bits (SUBREG_REG (x), int_op0_mode)
- & GET_MODE_MASK (int_mode)) == 0)
+ && ((nonzero_bits (SUBREG_REG (x), int_op0_mode)
+ & GET_MODE_MASK (int_mode)) == 0)
+ && !side_effects_p (SUBREG_REG (x)))
return CONST0_RTX (int_mode);
}
GET_MODE_MASK (mode), 0));
/* We can truncate a constant value and return it. */
- if (CONST_INT_P (XEXP (x, 0)))
- return gen_int_mode (INTVAL (XEXP (x, 0)), mode);
+ {
+ poly_int64 c;
+ if (poly_int_rtx_p (XEXP (x, 0), &c))
+ return gen_int_mode (c, mode);
+ }
/* Similarly to what we do in simplify-rtx.c, a truncate of a register
whose value is a comparison can be replaced with a subreg if
&& (UINTVAL (XEXP (XEXP (XEXP (x, 0), 0), 1))
== (HOST_WIDE_INT_1U << (i + 1)) - 1))
|| (GET_CODE (XEXP (XEXP (x, 0), 0)) == ZERO_EXTEND
- && (GET_MODE_PRECISION (GET_MODE (XEXP (XEXP (XEXP (x, 0), 0), 0)))
- == (unsigned int) i + 1))))
+ && known_eq ((GET_MODE_PRECISION
+ (GET_MODE (XEXP (XEXP (XEXP (x, 0), 0), 0)))),
+ (unsigned int) i + 1))))
return simplify_shift_const
(NULL_RTX, ASHIFTRT, int_mode,
simplify_shift_const (NULL_RTX, ASHIFT, int_mode,
pc_rtx, pc_rtx, 0, 0, 0);
if (reg_mentioned_p (from, false_rtx))
false_rtx = subst (known_cond (copy_rtx (false_rtx), false_code,
- from, false_val),
+ from, false_val),
pc_rtx, pc_rtx, 0, 0, 0);
SUBST (XEXP (x, 1), swapped ? false_rtx : true_rtx);
|| reg_mentioned_p (true_rtx, false_rtx)
|| rtx_equal_p (false_rtx, XEXP (cond, 0))))
{
- true_code = reversed_comparison_code (cond, NULL);
SUBST (XEXP (x, 0), reversed_comparison (cond, GET_MODE (cond)));
SUBST (XEXP (x, 1), false_rtx);
SUBST (XEXP (x, 2), true_rtx);
if (z)
{
- temp = subst (simplify_gen_relational (true_code, m, VOIDmode,
+ machine_mode cm = m;
+ if ((op == ASHIFT || op == LSHIFTRT || op == ASHIFTRT)
+ && GET_MODE (c1) != VOIDmode)
+ cm = GET_MODE (c1);
+ temp = subst (simplify_gen_relational (true_code, cm, VOIDmode,
cond_op0, cond_op1),
pc_rtx, pc_rtx, 0, 0, 0);
- temp = simplify_gen_binary (MULT, m, temp,
- simplify_gen_binary (MULT, m, c1,
+ temp = simplify_gen_binary (MULT, cm, temp,
+ simplify_gen_binary (MULT, cm, c1,
const_true_rtx));
temp = subst (temp, pc_rtx, pc_rtx, 0, 0, 0);
temp = simplify_gen_binary (op, m, gen_lowpart (m, z), temp);
&& XEXP (cond, 1) == const0_rtx
&& false_rtx == const0_rtx
&& CONST_INT_P (true_rtx)
- && ((1 == nonzero_bits (XEXP (cond, 0), int_mode)
+ && ((nonzero_bits (XEXP (cond, 0), int_mode) == 1
&& (i = exact_log2 (UINTVAL (true_rtx))) >= 0)
|| ((num_sign_bit_copies (XEXP (cond, 0), int_mode)
== GET_MODE_PRECISION (int_mode))
if (GET_CODE (src) == SUBREG && subreg_lowpart_p (src)
&& !OBJECT_P (SUBREG_REG (src))
- && (((GET_MODE_SIZE (GET_MODE (src)) + (UNITS_PER_WORD - 1))
- / UNITS_PER_WORD)
- == ((GET_MODE_SIZE (GET_MODE (SUBREG_REG (src)))
- + (UNITS_PER_WORD - 1)) / UNITS_PER_WORD))
+ && (known_equal_after_align_up
+ (GET_MODE_SIZE (GET_MODE (src)),
+ GET_MODE_SIZE (GET_MODE (SUBREG_REG (src))),
+ UNITS_PER_WORD))
&& (WORD_REGISTER_OPERATIONS || !paradoxical_subreg_p (src))
&& ! (REG_P (dest) && REGNO (dest) < FIRST_PSEUDO_REGISTER
&& !REG_CAN_CHANGE_MODE_P (REGNO (dest),
/* If we have (set FOO (subreg:M (mem:N BAR) 0)) with M wider than N, this
would require a paradoxical subreg. Replace the subreg with a
- zero_extend to avoid the reload that would otherwise be required. */
+ zero_extend to avoid the reload that would otherwise be required.
+ Don't do this unless we have a scalar integer mode, otherwise the
+ transformation is incorrect. */
enum rtx_code extend_op;
if (paradoxical_subreg_p (src)
&& MEM_P (SUBREG_REG (src))
+ && SCALAR_INT_MODE_P (GET_MODE (src))
&& (extend_op = load_extend_op (GET_MODE (SUBREG_REG (src)))) != UNKNOWN)
{
SUBST (SET_SRC (x),
{
rtx inner;
rtx pos; /* Always counts from low bit. */
- int len;
+ int len, inner_len;
rtx mask, cleared, masked;
scalar_int_mode compute_mode;
if (GET_CODE (SET_DEST (x)) == STRICT_LOW_PART
&& GET_CODE (XEXP (SET_DEST (x), 0)) == SUBREG)
{
+ rtx x0 = XEXP (SET_DEST (x), 0);
+ if (!GET_MODE_PRECISION (GET_MODE (x0)).is_constant (&len))
+ break;
inner = SUBREG_REG (XEXP (SET_DEST (x), 0));
- len = GET_MODE_PRECISION (GET_MODE (XEXP (SET_DEST (x), 0)));
- pos = GEN_INT (subreg_lsb (XEXP (SET_DEST (x), 0)));
+ pos = gen_int_mode (subreg_lsb (XEXP (SET_DEST (x), 0)),
+ MAX_MODE_INT);
}
else if (GET_CODE (SET_DEST (x)) == ZERO_EXTRACT
&& CONST_INT_P (XEXP (SET_DEST (x), 1)))
{
inner = XEXP (SET_DEST (x), 0);
+ if (!GET_MODE_PRECISION (GET_MODE (inner)).is_constant (&inner_len))
+ break;
+
len = INTVAL (XEXP (SET_DEST (x), 1));
pos = XEXP (SET_DEST (x), 2);
/* A constant position should stay within the width of INNER. */
- if (CONST_INT_P (pos)
- && INTVAL (pos) + len > GET_MODE_PRECISION (GET_MODE (inner)))
+ if (CONST_INT_P (pos) && INTVAL (pos) + len > inner_len)
break;
if (BITS_BIG_ENDIAN)
{
if (CONST_INT_P (pos))
- pos = GEN_INT (GET_MODE_PRECISION (GET_MODE (inner)) - len
- - INTVAL (pos));
+ pos = GEN_INT (inner_len - len - INTVAL (pos));
else if (GET_CODE (pos) == MINUS
&& CONST_INT_P (XEXP (pos, 1))
- && (INTVAL (XEXP (pos, 1))
- == GET_MODE_PRECISION (GET_MODE (inner)) - len))
+ && INTVAL (XEXP (pos, 1)) == inner_len - len)
/* If position is ADJUST - X, new position is X. */
pos = XEXP (pos, 0);
else
- {
- HOST_WIDE_INT prec = GET_MODE_PRECISION (GET_MODE (inner));
- pos = simplify_gen_binary (MINUS, GET_MODE (pos),
- gen_int_mode (prec - len,
- GET_MODE (pos)),
- pos);
- }
+ pos = simplify_gen_binary (MINUS, GET_MODE (pos),
+ gen_int_mode (inner_len - len,
+ GET_MODE (pos)),
+ pos);
}
}
bits outside of is_mode, don't look through
non-paradoxical SUBREGs. See PR82192. */
|| (pos_rtx == NULL_RTX
- && pos + len <= GET_MODE_PRECISION (is_mode))))
+ && known_le (pos + len, GET_MODE_PRECISION (is_mode)))))
{
/* If going from (subreg:SI (mem:QI ...)) to (mem:QI ...),
consider just the QI as the memory to extract from.
bits outside of is_mode, don't look through
TRUNCATE. See PR82192. */
&& pos_rtx == NULL_RTX
- && pos + len <= GET_MODE_PRECISION (is_mode))
+ && known_le (pos + len, GET_MODE_PRECISION (is_mode)))
inner = XEXP (inner, 0);
inner_mode = GET_MODE (inner);
/* We can't do this if we are widening INNER_MODE (it
may not be aligned, for one thing). */
&& !paradoxical_subreg_p (tmode, inner_mode)
+ && known_le (pos + len, GET_MODE_PRECISION (is_mode))
&& (inner_mode == tmode
|| (! mode_dependent_address_p (XEXP (inner, 0),
MEM_ADDR_SPACE (inner))
if (MEM_P (inner))
{
- HOST_WIDE_INT offset;
+ poly_int64 offset;
/* POS counts from lsb, but make OFFSET count in memory order. */
if (BYTES_BIG_ENDIAN)
- offset = (GET_MODE_PRECISION (is_mode) - len - pos) / BITS_PER_UNIT;
+ offset = bits_to_bytes_round_down (GET_MODE_PRECISION (is_mode)
+ - len - pos);
else
offset = pos / BITS_PER_UNIT;
return a new hard register. */
if (pos || in_dest)
{
- unsigned int offset
+ poly_uint64 offset
= subreg_offset_from_lsb (tmode, inner_mode, pos);
/* Avoid creating invalid subregs, for example when
other cases, we would only be going outside our object in cases when
an original shift would have been undefined. */
if (MEM_P (inner)
- && ((pos_rtx == 0 && pos + len > GET_MODE_PRECISION (is_mode))
+ && ((pos_rtx == 0 && maybe_gt (pos + len, GET_MODE_PRECISION (is_mode)))
|| (pos_rtx != 0 && len != 1)))
return 0;
For memory, assume that the desired extraction_mode and pos_mode
are the same as for a register operation, since at present we don't
have named patterns for aligned memory structures. */
- struct extraction_insn insn;
- if (get_best_reg_extraction_insn (&insn, pattern,
- GET_MODE_BITSIZE (inner_mode), mode))
+ class extraction_insn insn;
+ unsigned int inner_size;
+ if (GET_MODE_BITSIZE (inner_mode).is_constant (&inner_size)
+ && get_best_reg_extraction_insn (&insn, pattern, inner_size, mode))
{
wanted_inner_reg_mode = insn.struct_mode.require ();
pos_mode = insn.pos_mode;
&& partial_subreg_p (extraction_mode, mode))
extraction_mode = mode;
+ /* Punt if len is too large for extraction_mode. */
+ if (maybe_gt (len, GET_MODE_PRECISION (extraction_mode)))
+ return NULL_RTX;
+
if (!MEM_P (inner))
wanted_inner_mode = wanted_inner_reg_mode;
else
If it's a MEM we need to recompute POS relative to that.
However, if we're extracting from (or inserting into) a register,
we want to recompute POS relative to wanted_inner_mode. */
- int width = (MEM_P (inner)
- ? GET_MODE_BITSIZE (is_mode)
- : GET_MODE_BITSIZE (wanted_inner_mode));
+ int width;
+ if (!MEM_P (inner))
+ width = GET_MODE_BITSIZE (wanted_inner_mode);
+ else if (!GET_MODE_BITSIZE (is_mode).is_constant (&width))
+ return NULL_RTX;
if (pos_rtx == 0)
pos = width - len - pos;
&& ! mode_dependent_address_p (XEXP (inner, 0), MEM_ADDR_SPACE (inner))
&& ! MEM_VOLATILE_P (inner))
{
- int offset = 0;
+ poly_int64 offset = 0;
/* The computations below will be correct if the machine is big
endian in both bits and bytes or little endian in bits and bytes.
sub = XEXP (XEXP (x, 0), 0);
machine_mode sub_mode = GET_MODE (sub);
+ int sub_width;
if ((REG_P (sub) || MEM_P (sub))
- && GET_MODE_PRECISION (sub_mode) < mode_width)
+ && GET_MODE_PRECISION (sub_mode).is_constant (&sub_width)
+ && sub_width < mode_width)
{
unsigned HOST_WIDE_INT mode_mask = GET_MODE_MASK (sub_mode);
unsigned HOST_WIDE_INT mask;
if ((mask & mode_mask) == mode_mask)
{
new_rtx = make_compound_operation (sub, next_code);
- new_rtx = make_extraction (mode, new_rtx, 0, 0,
- GET_MODE_PRECISION (sub_mode),
+ new_rtx = make_extraction (mode, new_rtx, 0, 0, sub_width,
1, 0, in_code == COMPARE);
}
}
if (GET_CODE (x) == SUBREG
&& subreg_lowpart_p (x)
&& (partial_subreg_p (x)
- || (0 == (mask
- & GET_MODE_MASK (GET_MODE (x))
- & ~GET_MODE_MASK (GET_MODE (SUBREG_REG (x)))))))
+ || (mask
+ & GET_MODE_MASK (GET_MODE (x))
+ & ~GET_MODE_MASK (GET_MODE (SUBREG_REG (x)))) == 0))
return force_to_mode (SUBREG_REG (x), mode, mask, next_select);
scalar_int_mode int_mode, xmode;
int next_select = just_select || code == XOR || code == NOT || code == NEG;
unsigned HOST_WIDE_INT fuller_mask;
rtx op0, op1, temp;
+ poly_int64 const_op0;
/* When we have an arithmetic operation, or a shift whose count we
do not know, we need to assume that all bits up to the highest-order
case MINUS:
/* If X is (minus C Y) where C's least set bit is larger than any bit
in the mask, then we may replace with (neg Y). */
- if (CONST_INT_P (XEXP (x, 0))
- && least_bit_hwi (UINTVAL (XEXP (x, 0))) > mask)
+ if (poly_int_rtx_p (XEXP (x, 0), &const_op0)
+ && known_alignment (poly_uint64 (const_op0)) > mask)
{
x = simplify_gen_unary (NEG, xmode, XEXP (x, 1), xmode);
return force_to_mode (x, mode, mask, next_select);
mask = fuller_mask;
op0 = gen_lowpart_or_truncate (op_mode,
- force_to_mode (XEXP (x, 0), op_mode,
+ force_to_mode (XEXP (x, 0), mode,
mask, next_select));
if (op_mode != xmode || op0 != XEXP (x, 0))
/* Must be more sign bit copies than the mask needs. */
&& ((int) num_sign_bit_copies (XEXP (x, 0), GET_MODE (XEXP (x, 0)))
>= exact_log2 (mask + 1)))
- x = simplify_gen_binary (LSHIFTRT, xmode, XEXP (x, 0),
- GEN_INT (GET_MODE_PRECISION (xmode)
- - exact_log2 (mask + 1)));
-
+ {
+ int nbits = GET_MODE_PRECISION (xmode) - exact_log2 (mask + 1);
+ x = simplify_gen_binary (LSHIFTRT, xmode, XEXP (x, 0),
+ gen_int_shift_amount (xmode, nbits));
+ }
goto shiftrt;
case ASHIFTRT:
if (COMPARISON_P (cond0)
&& COMPARISON_P (cond1)
+ && SCALAR_INT_MODE_P (mode)
&& ((GET_CODE (cond0) == reversed_comparison_code (cond1, NULL)
&& rtx_equal_p (XEXP (cond0, 0), XEXP (cond1, 0))
&& rtx_equal_p (XEXP (cond0, 1), XEXP (cond1, 1)))
/* If X is a SUBREG, we can narrow both the true and false values
if the inner expression, if there is a condition. */
else if (code == SUBREG
- && 0 != (cond0 = if_then_else_cond (SUBREG_REG (x),
- &true0, &false0)))
+ && (cond0 = if_then_else_cond (SUBREG_REG (x), &true0,
+ &false0)) != 0)
{
true0 = simplify_gen_subreg (mode, true0,
GET_MODE (SUBREG_REG (x)), SUBREG_BYTE (x));
if (COMPARISON_P (x))
{
if (comparison_dominates_p (cond, code))
- return const_true_rtx;
+ return VECTOR_MODE_P (GET_MODE (x)) ? x : const_true_rtx;
code = reversed_comparison_code (x, NULL);
if (code != UNKNOWN
&& comparison_dominates_p (cond, code))
- return const0_rtx;
+ return CONST0_RTX (GET_MODE (x));
else
return x;
}
/* We must simplify subreg here, before we lose track of the
original inner_mode. */
new_rtx = simplify_subreg (GET_MODE (x), r,
- inner_mode, SUBREG_BYTE (x));
+ inner_mode, SUBREG_BYTE (x));
if (new_rtx)
return new_rtx;
else
/* We must simplify the zero_extend here, before we lose
track of the original inner_mode. */
new_rtx = simplify_unary_operation (ZERO_EXTEND, GET_MODE (x),
- r, inner_mode);
+ r, inner_mode);
if (new_rtx)
return new_rtx;
else
\f
/* Given a REG X of mode XMODE, compute which bits in X can be nonzero.
We don't care about bits outside of those defined in MODE.
+ We DO care about all the bits in MODE, even if XMODE is smaller than MODE.
For most X this is simply GET_MODE_MASK (GET_MODE (MODE)), but if X is
a shift, AND, or zero_extract, we can do better. */
rsp = ®_stat[REGNO (x)];
if (rsp->last_set_value != 0
&& (rsp->last_set_mode == mode
- || (GET_MODE_CLASS (rsp->last_set_mode) == MODE_INT
+ || (REGNO (x) >= FIRST_PSEUDO_REGISTER
+ && GET_MODE_CLASS (rsp->last_set_mode) == MODE_INT
&& GET_MODE_CLASS (mode) == MODE_INT))
&& ((rsp->last_set_label >= label_tick_ebb_start
&& rsp->last_set_label < label_tick)
{
enum rtx_code orig_code = code;
rtx orig_varop = varop;
- int count;
+ int count, log2;
machine_mode mode = result_mode;
machine_mode shift_mode;
scalar_int_mode tmode, inner_mode, int_mode, int_varop_mode, int_result_mode;
- unsigned int mode_words
- = (GET_MODE_SIZE (mode) + (UNITS_PER_WORD - 1)) / UNITS_PER_WORD;
/* We form (outer_op (code varop count) (outer_const)). */
enum rtx_code outer_op = UNKNOWN;
HOST_WIDE_INT outer_const = 0;
if (subreg_lowpart_p (varop)
&& is_int_mode (GET_MODE (SUBREG_REG (varop)), &inner_mode)
&& GET_MODE_SIZE (inner_mode) > GET_MODE_SIZE (int_varop_mode)
- && (unsigned int) ((GET_MODE_SIZE (inner_mode)
- + (UNITS_PER_WORD - 1)) / UNITS_PER_WORD)
- == mode_words
+ && (CEIL (GET_MODE_SIZE (inner_mode), UNITS_PER_WORD)
+ == CEIL (GET_MODE_SIZE (int_mode), UNITS_PER_WORD))
&& GET_MODE_CLASS (int_varop_mode) == MODE_INT)
{
varop = SUBREG_REG (varop);
is cheaper. But it is still better on those machines to
merge two shifts into one. */
if (CONST_INT_P (XEXP (varop, 1))
- && exact_log2 (UINTVAL (XEXP (varop, 1))) >= 0)
+ && (log2 = exact_log2 (UINTVAL (XEXP (varop, 1)))) >= 0)
{
- varop
- = simplify_gen_binary (ASHIFT, GET_MODE (varop),
- XEXP (varop, 0),
- GEN_INT (exact_log2 (
- UINTVAL (XEXP (varop, 1)))));
+ rtx log2_rtx = gen_int_shift_amount (GET_MODE (varop), log2);
+ varop = simplify_gen_binary (ASHIFT, GET_MODE (varop),
+ XEXP (varop, 0), log2_rtx);
continue;
}
break;
case UDIV:
/* Similar, for when divides are cheaper. */
if (CONST_INT_P (XEXP (varop, 1))
- && exact_log2 (UINTVAL (XEXP (varop, 1))) >= 0)
+ && (log2 = exact_log2 (UINTVAL (XEXP (varop, 1)))) >= 0)
{
- varop
- = simplify_gen_binary (LSHIFTRT, GET_MODE (varop),
- XEXP (varop, 0),
- GEN_INT (exact_log2 (
- UINTVAL (XEXP (varop, 1)))));
+ rtx log2_rtx = gen_int_shift_amount (GET_MODE (varop), log2);
+ varop = simplify_gen_binary (LSHIFTRT, GET_MODE (varop),
+ XEXP (varop, 0), log2_rtx);
continue;
}
break;
mask_rtx = gen_int_mode (nonzero_bits (varop, int_varop_mode),
int_result_mode);
-
+ rtx count_rtx = gen_int_shift_amount (int_result_mode, count);
mask_rtx
= simplify_const_binary_operation (code, int_result_mode,
- mask_rtx, GEN_INT (count));
+ mask_rtx, count_rtx);
/* Give up if we can't compute an outer operation to use. */
if (mask_rtx == 0
if (code == ASHIFTRT && int_mode != int_result_mode)
break;
+ rtx count_rtx = gen_int_shift_amount (int_result_mode, count);
rtx new_rtx = simplify_const_binary_operation (code, int_mode,
XEXP (varop, 0),
- GEN_INT (count));
+ count_rtx);
varop = gen_rtx_fmt_ee (code, int_mode, new_rtx, XEXP (varop, 1));
count = 0;
continue;
shift_unit_mode wider than result_mode. */
&& !(code == ASHIFTRT && GET_CODE (varop) == XOR
&& int_result_mode != shift_unit_mode
- && 0 > trunc_int_for_mode (INTVAL (XEXP (varop, 1)),
- shift_unit_mode))
+ && trunc_int_for_mode (INTVAL (XEXP (varop, 1)),
+ shift_unit_mode) < 0)
&& (new_rtx = simplify_const_binary_operation
(code, int_result_mode,
gen_int_mode (INTVAL (XEXP (varop, 1)), int_result_mode),
- GEN_INT (count))) != 0
+ gen_int_shift_amount (int_result_mode, count))) != 0
&& CONST_INT_P (new_rtx)
&& merge_outer_ops (&outer_op, &outer_const, GET_CODE (varop),
INTVAL (new_rtx), int_result_mode,
(ashiftrt (xor)) where we've widened the shift and the constant
changes the sign bit. */
if (CONST_INT_P (XEXP (varop, 1))
- && !(code == ASHIFTRT && GET_CODE (varop) == XOR
- && int_result_mode != shift_unit_mode
- && 0 > trunc_int_for_mode (INTVAL (XEXP (varop, 1)),
- shift_unit_mode)))
+ && !(code == ASHIFTRT && GET_CODE (varop) == XOR
+ && int_result_mode != shift_unit_mode
+ && trunc_int_for_mode (INTVAL (XEXP (varop, 1)),
+ shift_unit_mode) < 0))
{
rtx lhs = simplify_shift_const (NULL_RTX, code, shift_unit_mode,
XEXP (varop, 0), count);
else if ((code == ASHIFTRT || code == LSHIFTRT)
&& count < HOST_BITS_PER_WIDE_INT
&& HWI_COMPUTABLE_MODE_P (int_result_mode)
- && 0 == (nonzero_bits (XEXP (varop, 0), int_result_mode)
- >> count)
- && 0 == (nonzero_bits (XEXP (varop, 0), int_result_mode)
- & nonzero_bits (XEXP (varop, 1), int_result_mode)))
+ && (nonzero_bits (XEXP (varop, 0), int_result_mode)
+ >> count) == 0
+ && (nonzero_bits (XEXP (varop, 0), int_result_mode)
+ & nonzero_bits (XEXP (varop, 1), int_result_mode)) == 0)
{
varop = XEXP (varop, 1);
continue;
&& (new_rtx = simplify_const_binary_operation
(ASHIFT, int_result_mode,
gen_int_mode (INTVAL (XEXP (varop, 1)), int_result_mode),
- GEN_INT (count))) != 0
+ gen_int_shift_amount (int_result_mode, count))) != 0
&& CONST_INT_P (new_rtx)
&& merge_outer_ops (&outer_op, &outer_const, PLUS,
INTVAL (new_rtx), int_result_mode,
&& (new_rtx = simplify_const_binary_operation
(code, int_result_mode,
gen_int_mode (INTVAL (XEXP (varop, 1)), int_result_mode),
- GEN_INT (count))) != 0
+ gen_int_shift_amount (int_result_mode, count))) != 0
&& CONST_INT_P (new_rtx)
&& merge_outer_ops (&outer_op, &outer_const, XOR,
INTVAL (new_rtx), int_result_mode,
- GET_MODE_UNIT_PRECISION (GET_MODE (varop)))))
{
rtx varop_inner = XEXP (varop, 0);
-
- varop_inner
- = gen_rtx_LSHIFTRT (GET_MODE (varop_inner),
- XEXP (varop_inner, 0),
- GEN_INT
- (count + INTVAL (XEXP (varop_inner, 1))));
+ int new_count = count + INTVAL (XEXP (varop_inner, 1));
+ rtx new_count_rtx = gen_int_shift_amount (GET_MODE (varop_inner),
+ new_count);
+ varop_inner = gen_rtx_LSHIFTRT (GET_MODE (varop_inner),
+ XEXP (varop_inner, 0),
+ new_count_rtx);
varop = gen_rtx_TRUNCATE (GET_MODE (varop), varop_inner);
count = 0;
continue;
x = NULL_RTX;
if (x == NULL_RTX)
- x = simplify_gen_binary (code, shift_mode, varop, GEN_INT (count));
+ x = simplify_gen_binary (code, shift_mode, varop,
+ gen_int_shift_amount (shift_mode, count));
/* If we were doing an LSHIFTRT in a wider mode than it was originally,
turn off all the bits that the shift would have turned off. */
return tem;
if (!x)
- x = simplify_gen_binary (code, GET_MODE (varop), varop, GEN_INT (count));
+ x = simplify_gen_binary (code, GET_MODE (varop), varop,
+ gen_int_shift_amount (GET_MODE (varop), count));
if (GET_MODE (x) != result_mode)
x = gen_lowpart (result_mode, x);
return x;
if (BITS_BIG_ENDIAN)
start = GET_MODE_PRECISION (inner_mode) - size - start;
- if (start)
- x = gen_rtx_LSHIFTRT (inner_mode, XEXP (x, 0), GEN_INT (start));
+ if (start != 0)
+ x = gen_rtx_LSHIFTRT (inner_mode, XEXP (x, 0),
+ gen_int_shift_amount (inner_mode, start));
else
x = XEXP (x, 0);
+
if (mode != inner_mode)
- x = gen_lowpart_SUBREG (mode, x);
+ {
+ if (REG_P (x) && HARD_REGISTER_P (x)
+ && !can_change_dest_mode (x, 0, mode))
+ continue;
+
+ x = gen_lowpart_SUBREG (mode, x);
+ }
}
else if (GET_CODE (x) == ZERO_EXTEND
&& GET_CODE (XEXP (x, 0)) == SUBREG
size = GET_MODE_PRECISION (inner_mode);
x = SUBREG_REG (XEXP (x, 0));
if (GET_MODE (x) != mode)
- x = gen_lowpart_SUBREG (mode, x);
+ {
+ if (REG_P (x) && HARD_REGISTER_P (x)
+ && !can_change_dest_mode (x, 0, mode))
+ continue;
+
+ x = gen_lowpart_SUBREG (mode, x);
+ }
}
else if (GET_CODE (x) == ZERO_EXTEND
&& REG_P (XEXP (x, 0))
gen_lowpart_for_combine (machine_mode omode, rtx x)
{
machine_mode imode = GET_MODE (x);
- unsigned int osize = GET_MODE_SIZE (omode);
- unsigned int isize = GET_MODE_SIZE (imode);
rtx result;
if (omode == imode)
/* We can only support MODE being wider than a word if X is a
constant integer or has a mode the same size. */
- if (GET_MODE_SIZE (omode) > UNITS_PER_WORD
- && ! (CONST_SCALAR_INT_P (x) || isize == osize))
+ if (maybe_gt (GET_MODE_SIZE (omode), UNITS_PER_WORD)
+ && ! (CONST_SCALAR_INT_P (x)
+ || known_eq (GET_MODE_SIZE (imode), GET_MODE_SIZE (omode))))
goto fail;
/* X might be a paradoxical (subreg (mem)). In that case, gen_lowpart
if (imode == omode)
return x;
-
- isize = GET_MODE_SIZE (imode);
}
result = gen_lowpart_common (omode, x);
if (paradoxical_subreg_p (omode, imode))
return gen_rtx_SUBREG (omode, x, 0);
- HOST_WIDE_INT offset = byte_lowpart_offset (omode, imode);
+ poly_int64 offset = byte_lowpart_offset (omode, imode);
return adjust_address_nv (x, omode, offset);
}
/* If X is a comparison operator, rewrite it in a new mode. This
probably won't match, but may allow further simplifications. */
- else if (COMPARISON_P (x))
+ else if (COMPARISON_P (x)
+ && SCALAR_INT_MODE_P (imode)
+ && SCALAR_INT_MODE_P (omode))
return gen_rtx_fmt_ee (GET_CODE (x), omode, XEXP (x, 0), XEXP (x, 1));
/* If we couldn't simplify X any other way, just enclose it in a
&& HWI_COMPUTABLE_MODE_P (GET_MODE (SUBREG_REG (inner_op0)))
&& (GET_MODE (SUBREG_REG (inner_op0))
== GET_MODE (SUBREG_REG (inner_op1)))
- && (0 == ((~c0) & nonzero_bits (SUBREG_REG (inner_op0),
- GET_MODE (SUBREG_REG (inner_op0)))))
- && (0 == ((~c1) & nonzero_bits (SUBREG_REG (inner_op1),
- GET_MODE (SUBREG_REG (inner_op1))))))
+ && ((~c0) & nonzero_bits (SUBREG_REG (inner_op0),
+ GET_MODE (SUBREG_REG (inner_op0)))) == 0
+ && ((~c1) & nonzero_bits (SUBREG_REG (inner_op1),
+ GET_MODE (SUBREG_REG (inner_op1)))) == 0)
{
op0 = SUBREG_REG (inner_op0);
op1 = SUBREG_REG (inner_op1);
&& (equality_comparison_p || unsigned_comparison_p)
/* (A - C1) zero-extends if it is positive and sign-extends
if it is negative, C2 both zero- and sign-extends. */
- && ((0 == (nonzero_bits (a, inner_mode)
- & ~GET_MODE_MASK (mode))
+ && (((nonzero_bits (a, inner_mode)
+ & ~GET_MODE_MASK (mode)) == 0
&& const_op >= 0)
/* (A - C1) sign-extends if it is positive and 1-extends
if it is negative, C2 both sign- and 1-extends. */
this for equality comparisons due to pathological cases involving
overflows. */
if (equality_comparison_p
- && 0 != (tem = simplify_binary_operation (MINUS, mode,
- op1, XEXP (op0, 1))))
+ && (tem = simplify_binary_operation (MINUS, mode,
+ op1, XEXP (op0, 1))) != 0)
{
op0 = XEXP (op0, 0);
op1 = tem;
this for equality comparisons due to pathological cases involving
overflows. */
if (equality_comparison_p
- && 0 != (tem = simplify_binary_operation (PLUS, mode,
- XEXP (op0, 1), op1)))
+ && (tem = simplify_binary_operation (PLUS, mode,
+ XEXP (op0, 1), op1)) != 0)
{
op0 = XEXP (op0, 0);
op1 = tem;
}
if (equality_comparison_p
- && 0 != (tem = simplify_binary_operation (MINUS, mode,
- XEXP (op0, 0), op1)))
+ && (tem = simplify_binary_operation (MINUS, mode,
+ XEXP (op0, 0), op1)) != 0)
{
op0 = XEXP (op0, 1);
op1 = tem;
/* (eq (xor A B) C) -> (eq A (xor B C)). This is a simplification
if C is zero or B is a constant. */
if (equality_comparison_p
- && 0 != (tem = simplify_binary_operation (XOR, mode,
- XEXP (op0, 1), op1)))
+ && (tem = simplify_binary_operation (XOR, mode,
+ XEXP (op0, 1), op1)) != 0)
{
op0 = XEXP (op0, 0);
op1 = tem;
{
/* If there are two or more occurrences of REG in VALUE,
prevent the value from growing too much. */
- if (count_rtxs (tem) > MAX_LAST_VALUE_RTL)
+ if (count_rtxs (tem) > param_max_last_value_rtl)
tem = gen_rtx_CLOBBER (GET_MODE (tem), const0_rtx);
}
if (REG_P (dest))
{
/* If we are setting the whole register, we know its value. Otherwise
- show that we don't know the value. We can handle SUBREG in
- some cases. */
+ show that we don't know the value. We can handle a SUBREG if it's
+ the low part, but we must be careful with paradoxical SUBREGs on
+ RISC architectures because we cannot strip e.g. an extension around
+ a load and record the naked load since the RTL middle-end considers
+ that the upper bits are defined according to LOAD_EXTEND_OP. */
if (GET_CODE (setter) == SET && dest == SET_DEST (setter))
record_value_for_reg (dest, record_dead_insn, SET_SRC (setter));
else if (GET_CODE (setter) == SET
&& GET_CODE (SET_DEST (setter)) == SUBREG
&& SUBREG_REG (SET_DEST (setter)) == dest
- && GET_MODE_PRECISION (GET_MODE (dest)) <= BITS_PER_WORD
+ && known_le (GET_MODE_PRECISION (GET_MODE (dest)),
+ BITS_PER_WORD)
&& subreg_lowpart_p (SET_DEST (setter)))
record_value_for_reg (dest, record_dead_insn,
- gen_lowpart (GET_MODE (dest),
- SET_SRC (setter)));
+ WORD_REGISTER_OPERATIONS
+ && word_register_operation_p (SET_SRC (setter))
+ && paradoxical_subreg_p (SET_DEST (setter))
+ ? SET_SRC (setter)
+ : gen_lowpart (GET_MODE (dest),
+ SET_SRC (setter)));
else
record_value_for_reg (dest, record_dead_insn, NULL_RTX);
}
if (CALL_P (insn))
{
+ HARD_REG_SET callee_clobbers
+ = insn_callee_abi (insn).full_and_partial_reg_clobbers ();
hard_reg_set_iterator hrsi;
- EXECUTE_IF_SET_IN_HARD_REG_SET (regs_invalidated_by_call, 0, i, hrsi)
+ EXECUTE_IF_SET_IN_HARD_REG_SET (callee_clobbers, 0, i, hrsi)
{
reg_stat_type *rsp;
+ /* ??? We could try to preserve some information from the last
+ set of register I if the call doesn't actually clobber
+ (reg:last_set_mode I), which might be true for ABIs with
+ partial clobbers. However, it would be difficult to
+ update last_set_nonzero_bits and last_sign_bit_copies
+ to account for the part of I that actually was clobbered.
+ It wouldn't help much anyway, since we rarely see this
+ situation before RA. */
rsp = ®_stat[i];
rsp->last_set_invalid = 1;
rsp->last_set = insn;
the return value register is set at this LUID. We could
still replace a register with the return value from the
wrong subroutine call! */
- note_stores (PATTERN (insn), record_dead_and_set_regs_1, NULL_RTX);
+ note_stores (insn, record_dead_and_set_regs_1, NULL_RTX);
}
else
- note_stores (PATTERN (insn), record_dead_and_set_regs_1, insn);
+ note_stores (insn, record_dead_and_set_regs_1, insn);
}
/* If a SUBREG has the promoted bit set, it is in fact a property of the
/* If fewer bits were set than what we are asked for now, we cannot use
the value. */
- if (GET_MODE_PRECISION (rsp->last_set_mode)
- < GET_MODE_PRECISION (GET_MODE (x)))
+ if (maybe_lt (GET_MODE_PRECISION (rsp->last_set_mode),
+ GET_MODE_PRECISION (GET_MODE (x))))
return 0;
/* If the value has all its registers valid, return it. */
return 0;
}
\f
-/* Return nonzero if expression X refers to a REG or to memory
- that is set in an instruction more recent than FROM_LUID. */
-
-static int
-use_crosses_set_p (const_rtx x, int from_luid)
-{
- const char *fmt;
- int i;
- enum rtx_code code = GET_CODE (x);
-
- if (code == REG)
- {
- unsigned int regno = REGNO (x);
- unsigned endreg = END_REGNO (x);
-
-#ifdef PUSH_ROUNDING
- /* Don't allow uses of the stack pointer to be moved,
- because we don't know whether the move crosses a push insn. */
- if (regno == STACK_POINTER_REGNUM && PUSH_ARGS)
- return 1;
-#endif
- for (; regno < endreg; regno++)
- {
- reg_stat_type *rsp = ®_stat[regno];
- if (rsp->last_set
- && rsp->last_set_label == label_tick
- && DF_INSN_LUID (rsp->last_set) > from_luid)
- return 1;
- }
- return 0;
- }
-
- if (code == MEM && mem_last_set > from_luid)
- return 1;
-
- fmt = GET_RTX_FORMAT (code);
-
- for (i = GET_RTX_LENGTH (code) - 1; i >= 0; i--)
- {
- if (fmt[i] == 'E')
- {
- int j;
- for (j = XVECLEN (x, i) - 1; j >= 0; j--)
- if (use_crosses_set_p (XVECEXP (x, i, j), from_luid))
- return 1;
- }
- else if (fmt[i] == 'e'
- && use_crosses_set_p (XEXP (x, i), from_luid))
- return 1;
- }
- return 0;
-}
-\f
/* Define three variables used for communication between the following
routines. */
static unsigned int reg_dead_regno, reg_dead_endregno;
static int reg_dead_flag;
+rtx reg_dead_reg;
/* Function called via note_stores from reg_dead_at_p.
/* Set variables for reg_dead_at_p_1. */
reg_dead_regno = REGNO (reg);
reg_dead_endregno = END_REGNO (reg);
+ reg_dead_reg = reg;
reg_dead_flag = 0;
if (find_regno_note (insn, REG_UNUSED, reg_dead_regno))
return 1;
- note_stores (PATTERN (insn), reg_dead_at_p_1, NULL);
+ note_stores (insn, reg_dead_at_p_1, NULL);
if (reg_dead_flag)
return reg_dead_flag == 1 ? 1 : 0;
unsigned int regno = REGNO (x);
rtx_insn *where_dead = reg_stat[regno].last_death;
+ /* If we do not know where the register died, it may still die between
+ FROM_LUID and TO_INSN. If so, find it. This is PR83304. */
+ if (!where_dead || DF_INSN_LUID (where_dead) >= DF_INSN_LUID (to_insn))
+ {
+ rtx_insn *insn = prev_real_nondebug_insn (to_insn);
+ while (insn
+ && BLOCK_FOR_INSN (insn) == BLOCK_FOR_INSN (to_insn)
+ && DF_INSN_LUID (insn) >= from_luid)
+ {
+ if (dead_or_set_regno_p (insn, regno))
+ {
+ if (find_regno_note (insn, REG_DEAD, regno))
+ where_dead = insn;
+ break;
+ }
+
+ insn = prev_real_nondebug_insn (insn);
+ }
+ }
+
/* Don't move the register if it gets killed in between from and to. */
if (maybe_kill_insn && reg_set_p (x, maybe_kill_insn)
&& ! reg_referenced_p (x, maybe_kill_insn))
entire adjustment. Assert i3 contains at least some adjust. */
if (!noop_move_p (i3))
{
- int old_size, args_size = INTVAL (XEXP (note, 0));
+ poly_int64 old_size, args_size = get_args_size (note);
/* fixup_args_size_notes looks at REG_NORETURN note,
so ensure the note is placed there first. */
if (CALL_P (i3))
old_size = fixup_args_size_notes (PREV_INSN (i3), i3, args_size);
/* emit_call_1 adds for !ACCUMULATE_OUTGOING_ARGS
REG_ARGS_SIZE note to all noreturn calls, allow that here. */
- gcc_assert (old_size != args_size
+ gcc_assert (maybe_ne (old_size, args_size)
|| (CALL_P (i3)
&& !ACCUMULATE_OUTGOING_ARGS
&& find_reg_note (i3, REG_NORETURN, NULL_RTX)));
PUT_REG_NOTE_KIND (note, REG_DEAD);
place = i3;
}
+
+ /* A SET or CLOBBER of the REG_UNUSED reg has been removed,
+ but we can't tell which at this point. We must reset any
+ expectations we had about the value that was previously
+ stored in the reg. ??? Ideally, we'd adjust REG_N_SETS
+ and, if appropriate, restore its previous value, but we
+ don't have enough information for that at this point. */
+ else
+ {
+ record_value_for_reg (XEXP (note, 0), NULL, NULL_RTX);
+
+ /* Otherwise, if this register is now referenced in i2
+ then the register used to be modified in one of the
+ original insns. If it was i3 (say, in an unused
+ parallel), it's now completely gone, so the note can
+ be discarded. But if it was modified in i2, i1 or i0
+ and we still reference it in i2, then we're
+ referencing the previous value, and since the
+ register was modified and REG_UNUSED, we know that
+ the previous value is now dead. So, if we only
+ reference the register in i2, we change the note to
+ REG_DEAD, to reflect the previous value. However, if
+ we're also setting or clobbering the register as
+ scratch, we know (because the register was not
+ referenced in i3) that it's unused, just as it was
+ unused before, and we place the note in i2. */
+ if (from_insn != i3 && i2 && INSN_P (i2)
+ && reg_referenced_p (XEXP (note, 0), PATTERN (i2)))
+ {
+ if (!reg_set_p (XEXP (note, 0), PATTERN (i2)))
+ PUT_REG_NOTE_KIND (note, REG_DEAD);
+ if (! (REG_P (XEXP (note, 0))
+ ? find_regno_note (i2, REG_NOTE_KIND (note),
+ REGNO (XEXP (note, 0)))
+ : find_reg_note (i2, REG_NOTE_KIND (note),
+ XEXP (note, 0))))
+ place = i2;
+ }
+ }
+
break;
case REG_EQUAL:
&& CALL_P (from_insn)
&& find_reg_fusage (from_insn, USE, XEXP (note, 0)))
place = from_insn;
+ else if (i2 && reg_set_p (XEXP (note, 0), PATTERN (i2)))
+ {
+ /* If the new I2 sets the same register that is marked
+ dead in the note, we do not in general know where to
+ put the note. One important case we _can_ handle is
+ when the note comes from I3. */
+ if (from_insn == i3)
+ place = i3;
+ else
+ break;
+ }
else if (reg_referenced_p (XEXP (note, 0), PATTERN (i3)))
place = i3;
else if (i2 != 0 && next_nonnote_nondebug_insn (i2) == i3
|| rtx_equal_p (XEXP (note, 0), elim_i0))
break;
tem_insn = i3;
- /* If the new I2 sets the same register that is marked dead
- in the note, we do not know where to put the note.
- Give up. */
- if (i2 != 0 && reg_set_p (XEXP (note, 0), PATTERN (i2)))
- break;
}
if (place == 0)
{
XEXP (note, 1) = REG_NOTES (place);
REG_NOTES (place) = note;
+
+ /* Set added_notes_insn to the earliest insn we added a note to. */
+ if (added_notes_insn == 0
+ || DF_INSN_LUID (added_notes_insn) > DF_INSN_LUID (place))
+ added_notes_insn = place;
}
if (place2)
- add_shallow_copy_of_reg_note (place2, note);
+ {
+ add_shallow_copy_of_reg_note (place2, note);
+
+ /* Set added_notes_insn to the earliest insn we added a note to. */
+ if (added_notes_insn == 0
+ || DF_INSN_LUID (added_notes_insn) > DF_INSN_LUID (place2))
+ added_notes_insn = place2;
+ }
}
}
\f
|| GET_CODE (reg) == SUBREG)
reg = XEXP (reg, 0);
+ if (reg == pc_rtx)
+ continue;
+
/* A LOG_LINK is defined as being placed on the first insn that uses
a register and points to the insn that sets the register. Start
searching at the next insn after the target of the link and stop
total_attempts, total_merges, total_extras, total_successes);
}
\f
+/* Make pseudo-to-pseudo copies after every hard-reg-to-pseudo-copy, because
+ the reg-to-reg copy can usefully combine with later instructions, but we
+ do not want to combine the hard reg into later instructions, for that
+ restricts register allocation. */
+static void
+make_more_copies (void)
+{
+ basic_block bb;
+
+ FOR_EACH_BB_FN (bb, cfun)
+ {
+ rtx_insn *insn;
+
+ FOR_BB_INSNS (bb, insn)
+ {
+ if (!NONDEBUG_INSN_P (insn))
+ continue;
+
+ rtx set = single_set (insn);
+ if (!set)
+ continue;
+
+ rtx dest = SET_DEST (set);
+ if (!(REG_P (dest) && !HARD_REGISTER_P (dest)))
+ continue;
+
+ rtx src = SET_SRC (set);
+ if (!(REG_P (src) && HARD_REGISTER_P (src)))
+ continue;
+ if (TEST_HARD_REG_BIT (fixed_reg_set, REGNO (src)))
+ continue;
+
+ rtx new_reg = gen_reg_rtx (GET_MODE (dest));
+ rtx_insn *new_insn = gen_move_insn (new_reg, src);
+ SET_SRC (set) = new_reg;
+ emit_insn_before (new_insn, insn);
+ df_insn_rescan (insn);
+ }
+ }
+}
+
/* Try combining insns through substitution. */
static unsigned int
rest_of_handle_combine (void)
{
- int rebuild_jump_labels_after_combine;
+ make_more_copies ();
df_set_flags (DF_LR_RUN_DCE + DF_DEFER_INSN_RESCAN);
df_note_add_problem ();
regstat_init_n_sets_and_refs ();
reg_n_sets_max = max_reg_num ();
- rebuild_jump_labels_after_combine
+ int rebuild_jump_labels_after_combine
= combine_instructions (get_insns (), max_reg_num ());
/* Combining insns may have turned an indirect jump into a
projects / gcc.git / blobdiff