/* Generate code for transformation 1 (with MODE and OPERATION, operands OP1
and OP2, whose value is expected to be VALUE, result TARGET and
probability of taking the optimal path PROB). */
static rtx
gen_divmod_fixed_value (enum machine_mode mode, enum rtx_code operation,
rtx target, rtx op1, rtx op2, gcov_type value,
int prob)
{
rtx tmp, tmp1, jump;
rtx neq_label = gen_label_rtx ();
rtx end_label = gen_label_rtx ();
rtx sequence;
start_sequence ();
if (!REG_P (op2))
{
tmp = gen_reg_rtx (mode);
emit_move_insn (tmp, copy_rtx (op2));
}
else
tmp = op2;
do_compare_rtx_and_jump (tmp, GEN_INT (value), NE, 0, mode, NULL_RTX,
NULL_RTX, neq_label);
/* Add branch probability to jump we just created. */
jump = get_last_insn ();
REG_NOTES (jump) = gen_rtx_EXPR_LIST (REG_BR_PROB,
GEN_INT (REG_BR_PROB_BASE - prob),
REG_NOTES (jump));
tmp1 = simplify_gen_binary (operation, mode,
copy_rtx (op1), GEN_INT (value));
tmp1 = force_operand (tmp1, target);
if (tmp1 != target)
emit_move_insn (copy_rtx (target), copy_rtx (tmp1));
emit_jump_insn (gen_jump (end_label));
emit_barrier ();
emit_label (neq_label);
tmp1 = simplify_gen_binary (operation, mode,
copy_rtx (op1), copy_rtx (tmp));
tmp1 = force_operand (tmp1, target);
if (tmp1 != target)
emit_move_insn (copy_rtx (target), copy_rtx (tmp1));
emit_label (end_label);
sequence = get_insns ();
end_sequence ();
rebuild_jump_labels (sequence);
return sequence;
}
static rtx
gen_speculative_prefetch (rtx address, gcov_type delta, int write)
{
rtx tmp;
rtx sequence;
/* TODO: we do the prefetching for just one iteration ahead, which
often is not enough. */
start_sequence ();
if (offsettable_address_p (0, VOIDmode, address))
tmp = plus_constant (copy_rtx (address), delta);
else
{
tmp = simplify_gen_binary (PLUS, Pmode,
copy_rtx (address), GEN_INT (delta));
tmp = force_operand (tmp, NULL);
}
if (! (*insn_data[(int)CODE_FOR_prefetch].operand[0].predicate)
(tmp, insn_data[(int)CODE_FOR_prefetch].operand[0].mode))
tmp = force_reg (Pmode, tmp);
emit_insn (gen_prefetch (tmp, GEN_INT (write), GEN_INT (3)));
sequence = get_insns ();
end_sequence ();
return sequence;
}
/* Generate code for transformations 3 and 4 (with MODE and OPERATION,
operands OP1 and OP2, result TARGET, at most SUB subtractions, and
probability of taking the optimal path(s) PROB1 and PROB2). */
static rtx
gen_mod_subtract (enum machine_mode mode, enum rtx_code operation,
rtx target, rtx op1, rtx op2, int sub, int prob1, int prob2)
{
rtx tmp, tmp1, jump;
rtx end_label = gen_label_rtx ();
rtx sequence;
int i;
start_sequence ();
if (!REG_P (op2))
{
tmp = gen_reg_rtx (mode);
emit_move_insn (tmp, copy_rtx (op2));
}
else
tmp = op2;
emit_move_insn (target, copy_rtx (op1));
do_compare_rtx_and_jump (target, tmp, LTU, 0, mode, NULL_RTX,
NULL_RTX, end_label);
/* Add branch probability to jump we just created. */
jump = get_last_insn ();
REG_NOTES (jump) = gen_rtx_EXPR_LIST (REG_BR_PROB,
GEN_INT (prob1), REG_NOTES (jump));
for (i = 0; i < sub; i++)
{
tmp1 = expand_simple_binop (mode, MINUS, target, tmp, target,
0, OPTAB_WIDEN);
if (tmp1 != target)
emit_move_insn (target, tmp1);
do_compare_rtx_and_jump (target, tmp, LTU, 0, mode, NULL_RTX,
NULL_RTX, end_label);
/* Add branch probability to jump we just created. */
jump = get_last_insn ();
REG_NOTES (jump) = gen_rtx_EXPR_LIST (REG_BR_PROB,
GEN_INT (prob2), REG_NOTES (jump));
}
tmp1 = simplify_gen_binary (operation, mode, copy_rtx (target), copy_rtx (tmp));
tmp1 = force_operand (tmp1, target);
if (tmp1 != target)
emit_move_insn (target, tmp1);
emit_label (end_label);
sequence = get_insns ();
end_sequence ();
rebuild_jump_labels (sequence);
return sequence;
}
/* Generate code for transformation 2 (with MODE and OPERATION, operands OP1
and OP2, result TARGET and probability of taking the optimal path PROB). */
static rtx
gen_mod_pow2 (enum machine_mode mode, enum rtx_code operation, rtx target,
rtx op1, rtx op2, int prob)
{
rtx tmp, tmp1, tmp2, tmp3, jump;
rtx neq_label = gen_label_rtx ();
rtx end_label = gen_label_rtx ();
rtx sequence;
start_sequence ();
if (!REG_P (op2))
{
tmp = gen_reg_rtx (mode);
emit_move_insn (tmp, copy_rtx (op2));
}
else
tmp = op2;
tmp1 = expand_simple_binop (mode, PLUS, tmp, constm1_rtx, NULL_RTX,
0, OPTAB_WIDEN);
tmp2 = expand_simple_binop (mode, AND, tmp, tmp1, NULL_RTX,
0, OPTAB_WIDEN);
do_compare_rtx_and_jump (tmp2, const0_rtx, NE, 0, mode, NULL_RTX,
NULL_RTX, neq_label);
/* Add branch probability to jump we just created. */
jump = get_last_insn ();
REG_NOTES (jump) = gen_rtx_EXPR_LIST (REG_BR_PROB,
GEN_INT (REG_BR_PROB_BASE - prob),
REG_NOTES (jump));
tmp3 = expand_simple_binop (mode, AND, op1, tmp1, target,
0, OPTAB_WIDEN);
if (tmp3 != target)
emit_move_insn (copy_rtx (target), tmp3);
emit_jump_insn (gen_jump (end_label));
emit_barrier ();
emit_label (neq_label);
tmp1 = simplify_gen_binary (operation, mode, copy_rtx (op1), copy_rtx (tmp));
tmp1 = force_operand (tmp1, target);
if (tmp1 != target)
emit_move_insn (target, tmp1);
emit_label (end_label);
sequence = get_insns ();
end_sequence ();
rebuild_jump_labels (sequence);
return sequence;
}
static void
gen_addr_rtx (enum machine_mode address_mode,
rtx symbol, rtx base, rtx index, rtx step, rtx offset,
rtx *addr, rtx **step_p, rtx **offset_p)
{
rtx act_elem;
*addr = NULL_RTX;
if (step_p)
*step_p = NULL;
if (offset_p)
*offset_p = NULL;
if (index)
{
act_elem = index;
if (step)
{
act_elem = gen_rtx_MULT (address_mode, act_elem, step);
if (step_p)
*step_p = &XEXP (act_elem, 1);
}
*addr = act_elem;
}
if (base && base != const0_rtx)
{
if (*addr)
*addr = simplify_gen_binary (PLUS, address_mode, base, *addr);
else
*addr = base;
}
if (symbol)
{
act_elem = symbol;
if (offset)
{
act_elem = gen_rtx_PLUS (address_mode, act_elem, offset);
if (offset_p)
*offset_p = &XEXP (act_elem, 1);
if (GET_CODE (symbol) == SYMBOL_REF
|| GET_CODE (symbol) == LABEL_REF
|| GET_CODE (symbol) == CONST)
act_elem = gen_rtx_CONST (address_mode, act_elem);
}
if (*addr)
*addr = gen_rtx_PLUS (address_mode, *addr, act_elem);
else
*addr = act_elem;
}
else if (offset)
{
if (*addr)
{
*addr = gen_rtx_PLUS (address_mode, *addr, offset);
if (offset_p)
*offset_p = &XEXP (*addr, 1);
}
else
{
*addr = offset;
if (offset_p)
*offset_p = addr;
}
}
if (!*addr)
*addr = const0_rtx;
}
static bool
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);
machine_mode mode = GET_MODE (x);
machine_mode op_mode;
bool can_appear = (flags & PR_CAN_APPEAR) != 0;
bool valid_ops = true;
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 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_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_rtx, new_rtx, flags);
if (op0 == XEXP (x, 0))
return true;
tem = simplify_gen_unary (code, mode, op0, op_mode);
break;
case RTX_BIN_ARITH:
case RTX_COMM_ARITH:
op0 = XEXP (x, 0);
op1 = XEXP (x, 1);
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);
break;
case RTX_COMPARE:
case RTX_COMM_COMPARE:
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_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);
break;
case RTX_TERNARY:
case RTX_BITFIELD_OPS:
op0 = XEXP (x, 0);
op1 = XEXP (x, 1);
op2 = XEXP (x, 2);
op_mode = GET_MODE (op0);
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)
op_mode = GET_MODE (op0);
tem = simplify_gen_ternary (code, mode, op_mode, op0, op1, op2);
break;
case RTX_EXTRA:
/* The only case we try to handle is a SUBREG. */
if (code == SUBREG)
{
op0 = XEXP (x, 0);
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)),
SUBREG_BYTE (x));
}
break;
case RTX_OBJ:
if (code == MEM && x != new_rtx)
{
rtx new_op0;
op0 = XEXP (x, 0);
/* There are some addresses that we cannot work on. */
if (!can_simplify_addr (op0))
return true;
op0 = new_op0 = targetm.delegitimize_address (op0);
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
|| new_op0 == op0
|| !(GET_MODE (new_op0) == GET_MODE (op0)
|| GET_MODE (new_op0) == VOIDmode))
return true;
canonicalize_address (new_op0);
/* Copy propagations are always ok. Otherwise check the costs. */
if (!(REG_P (old_rtx) && REG_P (new_rtx))
&& !should_replace_address (op0, new_op0, GET_MODE (x),
MEM_ADDR_SPACE (x),
flags & PR_OPTIMIZE_FOR_SPEED))
return true;
tem = replace_equiv_address_nv (x, new_op0);
}
else if (code == LO_SUM)
{
op0 = XEXP (x, 0);
op1 = XEXP (x, 1);
/* 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_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;
/* (lo_sum (high x) x) -> x */
if (GET_CODE (op0) == HIGH && rtx_equal_p (XEXP (op0, 0), op1))
tem = op1;
else
tem = gen_rtx_LO_SUM (mode, op0, op1);
/* OP1 is likely not a legitimate address, otherwise there would have
been no LO_SUM. We want it to disappear if it is invalid, return
false in that case. */
return memory_address_p (mode, tem);
}
else if (code == REG)
{
if (rtx_equal_p (x, old_rtx))
{
*px = new_rtx;
return can_appear;
}
}
break;
default:
break;
}
/* No change, no trouble. */
if (tem == NULL_RTX)
return true;
*px = tem;
/* Allow replacements that simplify operations on a vector or complex
value to a component. The most prominent case is
(subreg ([vec_]concat ...)). */
if (REG_P (tem) && !HARD_REGISTER_P (tem)
&& (VECTOR_MODE_P (GET_MODE (new_rtx))
|| COMPLEX_MODE_P (GET_MODE (new_rtx)))
&& GET_MODE (tem) == GET_MODE_INNER (GET_MODE (new_rtx)))
return true;
/* The replacement we made so far is valid, if all of the recursive
replacements were valid, or we could simplify everything to
a constant. */
return valid_ops || can_appear || CONSTANT_P (tem);
}
static void
doloop_modify (struct loop *loop, struct niter_desc *desc,
rtx doloop_seq, rtx condition, rtx count)
{
rtx counter_reg;
rtx tmp, noloop = NULL_RTX;
rtx sequence;
rtx jump_insn;
rtx jump_label;
int nonneg = 0, irr;
bool increment_count;
basic_block loop_end = desc->out_edge->src;
enum machine_mode mode;
jump_insn = BB_END (loop_end);
if (dump_file)
{
fprintf (dump_file, "Doloop: Inserting doloop pattern (");
if (desc->const_iter)
fprintf (dump_file, HOST_WIDEST_INT_PRINT_DEC, desc->niter);
else
fputs ("runtime", dump_file);
fputs (" iterations).\n", dump_file);
}
/* Discard original jump to continue loop. The original compare
result may still be live, so it cannot be discarded explicitly. */
delete_insn (jump_insn);
counter_reg = XEXP (condition, 0);
if (GET_CODE (counter_reg) == PLUS)
counter_reg = XEXP (counter_reg, 0);
mode = GET_MODE (counter_reg);
increment_count = false;
switch (GET_CODE (condition))
{
case NE:
/* Currently only NE tests against zero and one are supported. */
if (XEXP (condition, 1) == const1_rtx)
{
increment_count = true;
noloop = const1_rtx;
}
else if (XEXP (condition, 1) == const0_rtx)
noloop = const0_rtx;
else
abort ();
break;
case GE:
/* Currently only GE tests against zero are supported. */
if (XEXP (condition, 1) != const0_rtx)
abort ();
noloop = constm1_rtx;
/* The iteration count does not need incrementing for a GE test. */
increment_count = false;
/* Determine if the iteration counter will be non-negative.
Note that the maximum value loaded is iterations_max - 1. */
if (desc->niter_max
<= ((unsigned HOST_WIDEST_INT) 1
<< (GET_MODE_BITSIZE (mode) - 1)))
nonneg = 1;
break;
/* Abort if an invalid doloop pattern has been generated. */
default:
abort ();
}
if (increment_count)
count = simplify_gen_binary (PLUS, mode, count, const1_rtx);
/* Insert initialization of the count register into the loop header. */
start_sequence ();
tmp = force_operand (count, counter_reg);
convert_move (counter_reg, tmp, 1);
sequence = get_insns ();
end_sequence ();
emit_insn_after (sequence, BB_END (loop_preheader_edge (loop)->src));
if (desc->noloop_assumptions)
{
rtx ass = copy_rtx (desc->noloop_assumptions);
basic_block preheader = loop_preheader_edge (loop)->src;
basic_block set_zero
= loop_split_edge_with (loop_preheader_edge (loop), NULL_RTX);
basic_block new_preheader
= loop_split_edge_with (loop_preheader_edge (loop), NULL_RTX);
basic_block bb;
edge te;
gcov_type cnt;
/* Expand the condition testing the assumptions and if it does not pass,
reset the count register to 0. */
add_test (XEXP (ass, 0), preheader, set_zero);
EDGE_SUCC (preheader, 0)->flags &= ~EDGE_FALLTHRU;
cnt = EDGE_SUCC (preheader, 0)->count;
EDGE_SUCC (preheader, 0)->probability = 0;
EDGE_SUCC (preheader, 0)->count = 0;
irr = EDGE_SUCC (preheader, 0)->flags & EDGE_IRREDUCIBLE_LOOP;
te = make_edge (preheader, new_preheader, EDGE_FALLTHRU | irr);
te->probability = REG_BR_PROB_BASE;
te->count = cnt;
set_immediate_dominator (CDI_DOMINATORS, new_preheader, preheader);
set_zero->count = 0;
set_zero->frequency = 0;
for (ass = XEXP (ass, 1); ass; ass = XEXP (ass, 1))
{
bb = loop_split_edge_with (te, NULL_RTX);
te = EDGE_SUCC (bb, 0);
add_test (XEXP (ass, 0), bb, set_zero);
make_edge (bb, set_zero, irr);
}
start_sequence ();
convert_move (counter_reg, noloop, 0);
sequence = get_insns ();
end_sequence ();
emit_insn_after (sequence, BB_END (set_zero));
}
/* Some targets (eg, C4x) need to initialize special looping
registers. */
#ifdef HAVE_doloop_begin
{
rtx init;
unsigned level = get_loop_level (loop) + 1;
init = gen_doloop_begin (counter_reg,
desc->const_iter ? desc->niter_expr : const0_rtx,
desc->niter_max,
GEN_INT (level));
if (init)
{
start_sequence ();
emit_insn (init);
sequence = get_insns ();
end_sequence ();
emit_insn_after (sequence, BB_END (loop_preheader_edge (loop)->src));
}
}
#endif
/* Insert the new low-overhead looping insn. */
emit_jump_insn_after (doloop_seq, BB_END (loop_end));
jump_insn = BB_END (loop_end);
jump_label = block_label (desc->in_edge->dest);
JUMP_LABEL (jump_insn) = jump_label;
LABEL_NUSES (jump_label)++;
/* Ensure the right fallthru edge is marked, for case we have reversed
the condition. */
desc->in_edge->flags &= ~EDGE_FALLTHRU;
desc->out_edge->flags |= EDGE_FALLTHRU;
/* Add a REG_NONNEG note if the actual or estimated maximum number
of iterations is non-negative. */
if (nonneg)
{
REG_NOTES (jump_insn)
= gen_rtx_EXPR_LIST (REG_NONNEG, NULL_RTX, REG_NOTES (jump_insn));
}
}
/* Find values inside INSN for that we want to measure histograms for
division/modulo optimization and stores them to VALUES. */
static void
insn_divmod_values_to_profile (rtx insn, histogram_values *values)
{
rtx set, set_src, op1, op2;
enum machine_mode mode;
histogram_value hist;
if (!INSN_P (insn))
return;
set = single_set (insn);
if (!set)
return;
mode = GET_MODE (SET_DEST (set));
if (!INTEGRAL_MODE_P (mode))
return;
set_src = SET_SRC (set);
switch (GET_CODE (set_src))
{
case DIV:
case MOD:
case UDIV:
case UMOD:
op1 = XEXP (set_src, 0);
op2 = XEXP (set_src, 1);
if (side_effects_p (op2))
return;
/* Check for a special case where the divisor is power of 2. */
if ((GET_CODE (set_src) == UMOD) && !CONSTANT_P (op2))
{
hist = ggc_alloc (sizeof (*hist));
hist->value = op2;
hist->seq = NULL_RTX;
hist->mode = mode;
hist->insn = insn;
hist->type = HIST_TYPE_POW2;
hist->hdata.pow2.may_be_other = 1;
VEC_safe_push (histogram_value, *values, hist);
}
/* Check whether the divisor is not in fact a constant. */
if (!CONSTANT_P (op2))
{
hist = ggc_alloc (sizeof (*hist));
hist->value = op2;
hist->mode = mode;
hist->seq = NULL_RTX;
hist->insn = insn;
hist->type = HIST_TYPE_SINGLE_VALUE;
VEC_safe_push (histogram_value, *values, hist);
}
/* For mod, check whether it is not often a noop (or replaceable by
a few subtractions). */
if (GET_CODE (set_src) == UMOD && !side_effects_p (op1))
{
rtx tmp;
hist = ggc_alloc (sizeof (*hist));
start_sequence ();
tmp = simplify_gen_binary (DIV, mode, copy_rtx (op1), copy_rtx (op2));
hist->value = force_operand (tmp, NULL_RTX);
hist->seq = get_insns ();
end_sequence ();
hist->mode = mode;
hist->insn = insn;
hist->type = HIST_TYPE_INTERVAL;
hist->hdata.intvl.int_start = 0;
hist->hdata.intvl.steps = 2;
hist->hdata.intvl.may_be_less = 1;
hist->hdata.intvl.may_be_more = 1;
VEC_safe_push (histogram_value, *values, hist);
}
return;
default:
return;
}
}
