/* Swing Modulo Scheduling implementation.
- Copyright (C) 2004, 2005, 2006, 2007
+ Copyright (C) 2004, 2005, 2006, 2007, 2008
Free Software Foundation, Inc.
Contributed by Ayal Zaks and Mustafa Hagog <zaks,mustafa@il.ibm.com>
#include "timevar.h"
#include "tree-pass.h"
#include "dbgcnt.h"
+#include "df.h"
#ifdef INSN_SCHEDULING
Currently SMS relies on the do-loop pattern to recognize such loops,
where (1) the control part comprises of all insns defining and/or
using a certain 'count' register and (2) the loop count can be
- adjusted by modifying this register prior to the loop.
+ adjusted by modifying this register prior to the loop.
TODO: Rely on cfgloop analysis instead. */
\f
/* This page defines partial-schedule structures and functions for
};
-
+
static partial_schedule_ptr create_partial_schedule (int ii, ddg_ptr, int history);
static void free_partial_schedule (partial_schedule_ptr);
static void reset_partial_schedule (partial_schedule_ptr, int new_ii);
/* This page defines constants and structures for the modulo scheduling
driver. */
-/* As in haifa-sched.c: */
-/* issue_rate is the number of insns that can be scheduled in the same
- machine cycle. It can be defined in the config/mach/mach.h file,
- otherwise we set it to 1. */
-
-static int issue_rate;
-
-static int sms_order_nodes (ddg_ptr, int, int * result);
+static int sms_order_nodes (ddg_ptr, int, int *, int *);
static void set_node_sched_params (ddg_ptr);
static partial_schedule_ptr sms_schedule_by_order (ddg_ptr, int, int, int *);
-static void permute_partial_schedule (partial_schedule_ptr ps, rtx last);
-static void generate_prolog_epilog (partial_schedule_ptr, struct loop *loop,
+static void permute_partial_schedule (partial_schedule_ptr, rtx);
+static void generate_prolog_epilog (partial_schedule_ptr, struct loop *,
rtx, rtx);
-static void duplicate_insns_of_cycles (partial_schedule_ptr ps,
- int from_stage, int to_stage,
- int is_prolog, rtx count_reg);
+static void duplicate_insns_of_cycles (partial_schedule_ptr,
+ int, int, int, rtx);
#define SCHED_ASAP(x) (((node_sched_params_ptr)(x)->aux.info)->asap)
#define SCHED_TIME(x) (((node_sched_params_ptr)(x)->aux.info)->time)
code in order to use sched_analyze() for computing the dependencies.
They are used when initializing the sched_info structure. */
static const char *
-sms_print_insn (rtx insn, int aligned ATTRIBUTE_UNUSED)
+sms_print_insn (const_rtx insn, int aligned ATTRIBUTE_UNUSED)
{
static char tmp[80];
{
}
-static struct sched_info sms_sched_info =
+static struct common_sched_info_def sms_common_sched_info;
+
+static struct sched_deps_info_def sms_sched_deps_info =
+ {
+ compute_jump_reg_dependencies,
+ NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
+ NULL,
+ 0, 0, 0
+ };
+
+static struct haifa_sched_info sms_sched_info =
{
NULL,
NULL,
NULL,
sms_print_insn,
NULL,
- compute_jump_reg_dependencies,
+ NULL, /* insn_finishes_block_p */
NULL, NULL,
NULL, NULL,
- 0, 0, 0,
+ 0, 0,
- NULL, NULL, NULL, NULL, NULL,
+ NULL, NULL, NULL,
0
};
-
/* Given HEAD and TAIL which are the first and last insns in a loop;
return the register which controls the loop. Return zero if it has
more than one occurrence in the loop besides the control part or the
get_ebb_head_tail (pre_header, pre_header, &head, &tail);
for (insn = tail; insn != PREV_INSN (head); insn = PREV_INSN (insn))
- if (INSN_P (insn) && single_set (insn) &&
+ if (NONDEBUG_INSN_P (insn) && single_set (insn) &&
rtx_equal_p (count_reg, SET_DEST (single_set (insn))))
{
rtx pat = single_set (insn);
- if (GET_CODE (SET_SRC (pat)) == CONST_INT)
+ if (CONST_INT_P (SET_SRC (pat)))
{
*count = INTVAL (SET_SRC (pat));
return insn;
res_MII (ddg_ptr g)
{
if (targetm.sched.sms_res_mii)
- return targetm.sched.sms_res_mii (g);
-
- return (g->num_nodes / issue_rate);
+ return targetm.sched.sms_res_mii (g);
+
+ return ((g->num_nodes - g->num_debug) / issue_rate);
}
/* Now generate the reg_moves, attaching relevant uses to them. */
SCHED_NREG_MOVES (u) = nreg_moves;
old_reg = prev_reg = copy_rtx (SET_DEST (single_set (u->insn)));
- last_reg_move = u->insn;
+ /* Insert the reg-moves right before the notes which precede
+ the insn they relates to. */
+ last_reg_move = u->first_note;
for (i_reg_move = 0; i_reg_move < nreg_moves; i_reg_move++)
{
int i;
int last_stage = PS_STAGE_COUNT (ps) - 1;
edge e;
-
+
/* Generate the prolog, inserting its insns on the loop-entry edge. */
start_sequence ();
for (i = 0; i < last_stage; i++)
duplicate_insns_of_cycles (ps, 0, i, 1, count_reg);
-
+
/* Put the prolog on the entry edge. */
e = loop_preheader_edge (loop);
split_edge_and_insert (e, get_insns ());
for (i = 0; i < last_stage; i++)
duplicate_insns_of_cycles (ps, i + 1, last_stage, 0, count_reg);
-
+
/* Put the epilogue on the exit edge. */
gcc_assert (single_exit (loop));
e = single_exit (loop);
for (; head != NEXT_INSN (tail); head = NEXT_INSN (head))
{
if (NOTE_P (head) || LABEL_P (head)
- || (INSN_P (head) && JUMP_P (head)))
+ || (INSN_P (head) && (DEBUG_INSN_P (head) || JUMP_P (head))))
continue;
empty_bb = false;
break;
{
if (loop->inner || !loop_outer (loop))
+ {
+ if (dump_file)
+ fprintf (dump_file, "SMS loop inner or !loop_outer\n");
return false;
+ }
if (!single_exit (loop))
{
if (dump_file)
{
rtx insn = BB_END (loop->header);
-
+
fprintf (dump_file, "SMS loop many exits ");
fprintf (dump_file, " %s %d (file, line)\n",
insn_file (insn), insn_line (insn));
if (dump_file)
{
rtx insn = BB_END (loop->header);
-
+
fprintf (dump_file, "SMS loop many BBs. ");
fprintf (dump_file, " %s %d (file, line)\n",
insn_file (insn), insn_line (insn));
}
}
+/* Setup infos. */
+static void
+setup_sched_infos (void)
+{
+ memcpy (&sms_common_sched_info, &haifa_common_sched_info,
+ sizeof (sms_common_sched_info));
+ sms_common_sched_info.sched_pass_id = SCHED_SMS_PASS;
+ common_sched_info = &sms_common_sched_info;
+
+ sched_deps_info = &sms_sched_deps_info;
+ current_sched_info = &sms_sched_info;
+}
+
/* Probability in % that the sms-ed loop rolls enough so that optimized
version may be entered. Just a guess. */
#define PROB_SMS_ENOUGH_ITERATIONS 80
rtx insn;
ddg_ptr *g_arr, g;
int * node_order;
- int maxii;
+ int maxii, max_asap;
loop_iterator li;
partial_schedule_ptr ps;
basic_block bb = NULL;
issue_rate = 1;
/* Initialize the scheduler. */
- current_sched_info = &sms_sched_info;
-
- /* Init Data Flow analysis, to be used in interloop dep calculation. */
- df_set_flags (DF_LR_RUN_DCE);
- df_rd_add_problem ();
- df_note_add_problem ();
- df_chain_add_problem (DF_DU_CHAIN);
- df_analyze ();
- regstat_compute_calls_crossed ();
- sched_init ();
+ setup_sched_infos ();
+ haifa_sched_init ();
/* Allocate memory to hold the DDG array one entry for each loop.
We use loop->num as index into this array. */
g_arr = XCNEWVEC (ddg_ptr, number_of_loops ());
+ if (dump_file)
+ {
+ fprintf (dump_file, "\n\nSMS analysis phase\n");
+ fprintf (dump_file, "===================\n\n");
+ }
+
/* Build DDGs for all the relevant loops and hold them in G_ARR
indexed by the loop index. */
FOR_EACH_LOOP (li, loop, 0)
break;
}
+ if (dump_file)
+ {
+ rtx insn = BB_END (loop->header);
+
+ fprintf (dump_file, "SMS loop num: %d, file: %s, line: %d\n",
+ loop->num, insn_file (insn), insn_line (insn));
+
+ }
+
if (! loop_canon_p (loop))
continue;
if (! loop_single_full_bb_p (loop))
+ {
+ if (dump_file)
+ fprintf (dump_file, "SMS not loop_single_full_bb_p\n");
continue;
+ }
bb = loop->header;
if (single_exit (loop)->count)
trip_count = latch_edge->count / single_exit (loop)->count;
- /* Perfrom SMS only on loops that their average count is above threshold. */
+ /* Perform SMS only on loops that their average count is above threshold. */
if ( latch_edge->count
&& (latch_edge->count < single_exit (loop)->count * SMS_LOOP_AVERAGE_COUNT_THRESHOLD))
/* Make sure this is a doloop. */
if ( !(count_reg = doloop_register_get (head, tail)))
+ {
+ if (dump_file)
+ fprintf (dump_file, "SMS doloop_register_get failed\n");
continue;
+ }
/* Don't handle BBs with calls or barriers, or !single_set insns,
or auto-increment insns (to avoid creating invalid reg-moves
- for the auto-increment insns).
+ for the auto-increment insns).
??? Should handle auto-increment insns.
??? Should handle insns defining subregs. */
for (insn = head; insn != NEXT_INSN (tail); insn = NEXT_INSN (insn))
if (CALL_P (insn)
|| BARRIER_P (insn)
- || (INSN_P (insn) && !JUMP_P (insn)
+ || (NONDEBUG_INSN_P (insn) && !JUMP_P (insn)
&& !single_set (insn) && GET_CODE (PATTERN (insn)) != USE)
|| (FIND_REG_INC_NOTE (insn, NULL_RTX) != 0)
|| (INSN_P (insn) && (set = single_set (insn))
fprintf (dump_file, "SMS loop-with-barrier\n");
else if (FIND_REG_INC_NOTE (insn, NULL_RTX) != 0)
fprintf (dump_file, "SMS reg inc\n");
- else if ((INSN_P (insn) && !JUMP_P (insn)
+ else if ((NONDEBUG_INSN_P (insn) && !JUMP_P (insn)
&& !single_set (insn) && GET_CODE (PATTERN (insn)) != USE))
fprintf (dump_file, "SMS loop-with-not-single-set\n");
else
if (! (g = create_ddg (bb, 0)))
{
if (dump_file)
- fprintf (dump_file, "SMS doloop\n");
+ fprintf (dump_file, "SMS create_ddg failed\n");
continue;
}
g_arr[loop->num] = g;
+ if (dump_file)
+ fprintf (dump_file, "...OK\n");
+
}
+ if (dump_file)
+ {
+ fprintf (dump_file, "\nSMS transformation phase\n");
+ fprintf (dump_file, "=========================\n\n");
+ }
/* We don't want to perform SMS on new loops - created by versioning. */
FOR_EACH_LOOP (li, loop, 0)
continue;
if (dump_file)
- print_ddg (dump_file, g);
+ {
+ rtx insn = BB_END (loop->header);
+
+ fprintf (dump_file, "SMS loop num: %d, file: %s, line: %d\n",
+ loop->num, insn_file (insn), insn_line (insn));
+
+ print_ddg (dump_file, g);
+ }
get_ebb_head_tail (loop->header, loop->header, &head, &tail);
node_order = XNEWVEC (int, g->num_nodes);
mii = 1; /* Need to pass some estimate of mii. */
- rec_mii = sms_order_nodes (g, mii, node_order);
+ rec_mii = sms_order_nodes (g, mii, node_order, &max_asap);
mii = MAX (res_MII (g), rec_mii);
- maxii = MAXII_FACTOR * mii;
+ maxii = MAX (max_asap, MAXII_FACTOR * mii);
if (dump_file)
fprintf (dump_file, "SMS iis %d %d %d (rec_mii, mii, maxii)\n",
ps = sms_schedule_by_order (g, mii, maxii, node_order);
- if (ps)
+ if (ps){
stage_count = PS_STAGE_COUNT (ps);
+ gcc_assert(stage_count >= 1);
+ }
/* Stage count of 1 means that there is no interleaving between
iterations, let the scheduling passes do the job. */
- if (stage_count < 1
+ if (stage_count <= 1
|| (count_init && (loop_count <= stage_count))
|| (flag_branch_probabilities && (trip_count <= stage_count)))
{
the closing_branch was scheduled and should appear in the last (ii-1)
row. Otherwise, we are free to schedule the branch, and we let nodes
that were scheduled at the first PS_MIN_CYCLE cycle appear in the first
- row; this should reduce stage_count to minimum.
+ row; this should reduce stage_count to minimum.
TODO: Revisit the issue of scheduling the insns of the
control part relative to the branch when the control part
has more than one insn. */
normalize_sched_times (ps);
rotate_partial_schedule (ps, PS_MIN_CYCLE (ps));
set_columns_for_ps (ps);
-
+
canon_loop (loop);
/* case the BCT count is not known , Do loop-versioning */
print_node_sched_params (dump_file, g->num_nodes, g);
/* Generate prolog and epilog. */
generate_prolog_epilog (ps, loop, count_reg, count_init);
-
+
free_undo_replace_buff (reg_move_replaces);
}
free_ddg (g);
}
- regstat_free_calls_crossed ();
free (g_arr);
/* Release scheduler data, needed until now because of DFA. */
- sched_finish ();
+ haifa_sched_finish ();
loop_optimizer_finalize ();
}
ddg_node_ptr v_node = e->src;
if (dump_file)
- {
+ {
fprintf (dump_file, "\nProcessing edge: ");
print_ddg_edge (dump_file, e);
fprintf (dump_file,
"\nScheduling %d (%d) in psp_not_empty,"
- " checking node %d (%d): ", u_node->cuid,
+ " checking p %d (%d): ", u_node->cuid,
INSN_UID (u_node->insn), v_node->cuid, INSN_UID
(v_node->insn));
}
if (TEST_BIT (sched_nodes, v_node->cuid))
{
- int node_st = SCHED_TIME (v_node)
- + e->latency - (e->distance * ii);
+ int p_st = SCHED_TIME (v_node);
- early_start = MAX (early_start, node_st);
+ early_start =
+ MAX (early_start, p_st + e->latency - (e->distance * ii));
+
+ if (dump_file)
+ fprintf (dump_file,
+ "pred st = %d; early_start = %d; latency: %d",
+ p_st, early_start, e->latency);
if (e->data_type == MEM_DEP)
end = MIN (end, SCHED_TIME (v_node) + ii - 1);
}
+ else if (dump_file)
+ fprintf (dump_file, "the node is not scheduled\n");
}
start = early_start;
end = MIN (end, early_start + ii);
+ /* Schedule the node close to it's predecessors. */
step = 1;
if (dump_file)
print_ddg_edge (dump_file, e);
fprintf (dump_file,
"\nScheduling %d (%d) in pss_not_empty,"
- " checking node %d (%d): ", u_node->cuid,
+ " checking s %d (%d): ", u_node->cuid,
INSN_UID (u_node->insn), v_node->cuid, INSN_UID
(v_node->insn));
}
if (TEST_BIT (sched_nodes, v_node->cuid))
{
- late_start = MIN (late_start,
- SCHED_TIME (v_node) - e->latency
- + (e->distance * ii));
- if (dump_file)
- fprintf (dump_file, "late_start = %d;", late_start);
+ int s_st = SCHED_TIME (v_node);
+
+ late_start = MIN (late_start,
+ s_st - e->latency + (e->distance * ii));
+
+ if (dump_file)
+ fprintf (dump_file,
+ "succ st = %d; late_start = %d; latency = %d",
+ s_st, late_start, e->latency);
if (e->data_type == MEM_DEP)
end = MAX (end, SCHED_TIME (v_node) - ii + 1);
}
start = late_start;
end = MAX (end, late_start - ii);
+ /* Schedule the node close to it's successors. */
step = -1;
if (dump_file)
{
int early_start = INT_MIN;
int late_start = INT_MAX;
+ int count_preds = 0;
+ int count_succs = 0;
start = INT_MIN;
end = INT_MAX;
if (TEST_BIT (sched_nodes, v_node->cuid))
{
+ int p_st = SCHED_TIME (v_node);
+
early_start = MAX (early_start,
- SCHED_TIME (v_node) + e->latency
+ p_st + e->latency
- (e->distance * ii));
+
+ if (dump_file)
+ fprintf (dump_file,
+ "pred st = %d; early_start = %d; latency = %d",
+ p_st, early_start, e->latency);
+
+ if (e->type == TRUE_DEP && e->data_type == REG_DEP)
+ count_preds++;
+
if (e->data_type == MEM_DEP)
end = MIN (end, SCHED_TIME (v_node) + ii - 1);
}
+ else if (dump_file)
+ fprintf (dump_file, "the node is not scheduled\n");
+
}
for (e = u_node->out; e != 0; e = e->next_out)
{
if (TEST_BIT (sched_nodes, v_node->cuid))
{
+ int s_st = SCHED_TIME (v_node);
+
late_start = MIN (late_start,
- SCHED_TIME (v_node) - e->latency
+ s_st - e->latency
+ (e->distance * ii));
+
+ if (dump_file)
+ fprintf (dump_file,
+ "succ st = %d; late_start = %d; latency = %d",
+ s_st, late_start, e->latency);
+
+ if (e->type == TRUE_DEP && e->data_type == REG_DEP)
+ count_succs++;
+
if (e->data_type == MEM_DEP)
start = MAX (start, SCHED_TIME (v_node) - ii + 1);
}
+ else if (dump_file)
+ fprintf (dump_file, "the node is not scheduled\n");
+
}
start = MAX (start, early_start);
end = MIN (end, MIN (early_start + ii, late_start + 1));
step = 1;
+ /* If there are more successors than predecessors schedule the
+ node close to it's successors. */
+ if (count_succs >= count_preds)
+ {
+ int old_start = start;
+
+ start = end - 1;
+ end = old_start - 1;
+ step = -1;
+ }
}
else /* psp is empty && pss is empty. */
{
return 0;
}
+/* Calculate MUST_PRECEDE/MUST_FOLLOW bitmaps of U_NODE; which is the
+ node currently been scheduled. At the end of the calculation
+ MUST_PRECEDE/MUST_FOLLOW contains all predecessors/successors of
+ U_NODE which are (1) already scheduled in the first/last row of
+ U_NODE's scheduling window, (2) whose dependence inequality with U
+ becomes an equality when U is scheduled in this same row, and (3)
+ whose dependence latency is zero.
+
+ The first and last rows are calculated using the following parameters:
+ START/END rows - The cycles that begins/ends the traversal on the window;
+ searching for an empty cycle to schedule U_NODE.
+ STEP - The direction in which we traverse the window.
+ II - The initiation interval. */
+
+static void
+calculate_must_precede_follow (ddg_node_ptr u_node, int start, int end,
+ int step, int ii, sbitmap sched_nodes,
+ sbitmap must_precede, sbitmap must_follow)
+{
+ ddg_edge_ptr e;
+ int first_cycle_in_window, last_cycle_in_window;
+
+ gcc_assert (must_precede && must_follow);
+
+ /* Consider the following scheduling window:
+ {first_cycle_in_window, first_cycle_in_window+1, ...,
+ last_cycle_in_window}. If step is 1 then the following will be
+ the order we traverse the window: {start=first_cycle_in_window,
+ first_cycle_in_window+1, ..., end=last_cycle_in_window+1},
+ or {start=last_cycle_in_window, last_cycle_in_window-1, ...,
+ end=first_cycle_in_window-1} if step is -1. */
+ first_cycle_in_window = (step == 1) ? start : end - step;
+ last_cycle_in_window = (step == 1) ? end - step : start;
+
+ sbitmap_zero (must_precede);
+ sbitmap_zero (must_follow);
+
+ if (dump_file)
+ fprintf (dump_file, "\nmust_precede: ");
+
+ /* Instead of checking if:
+ (SMODULO (SCHED_TIME (e->src), ii) == first_row_in_window)
+ && ((SCHED_TIME (e->src) + e->latency - (e->distance * ii)) ==
+ first_cycle_in_window)
+ && e->latency == 0
+ we use the fact that latency is non-negative:
+ SCHED_TIME (e->src) - (e->distance * ii) <=
+ SCHED_TIME (e->src) + e->latency - (e->distance * ii)) <=
+ first_cycle_in_window
+ and check only if
+ SCHED_TIME (e->src) - (e->distance * ii) == first_cycle_in_window */
+ for (e = u_node->in; e != 0; e = e->next_in)
+ if (TEST_BIT (sched_nodes, e->src->cuid)
+ && ((SCHED_TIME (e->src) - (e->distance * ii)) ==
+ first_cycle_in_window))
+ {
+ if (dump_file)
+ fprintf (dump_file, "%d ", e->src->cuid);
+
+ SET_BIT (must_precede, e->src->cuid);
+ }
+
+ if (dump_file)
+ fprintf (dump_file, "\nmust_follow: ");
+
+ /* Instead of checking if:
+ (SMODULO (SCHED_TIME (e->dest), ii) == last_row_in_window)
+ && ((SCHED_TIME (e->dest) - e->latency + (e->distance * ii)) ==
+ last_cycle_in_window)
+ && e->latency == 0
+ we use the fact that latency is non-negative:
+ SCHED_TIME (e->dest) + (e->distance * ii) >=
+ SCHED_TIME (e->dest) - e->latency + (e->distance * ii)) >=
+ last_cycle_in_window
+ and check only if
+ SCHED_TIME (e->dest) + (e->distance * ii) == last_cycle_in_window */
+ for (e = u_node->out; e != 0; e = e->next_out)
+ if (TEST_BIT (sched_nodes, e->dest->cuid)
+ && ((SCHED_TIME (e->dest) + (e->distance * ii)) ==
+ last_cycle_in_window))
+ {
+ if (dump_file)
+ fprintf (dump_file, "%d ", e->dest->cuid);
+
+ SET_BIT (must_follow, e->dest->cuid);
+ }
+
+ if (dump_file)
+ fprintf (dump_file, "\n");
+}
+
+/* Return 1 if U_NODE can be scheduled in CYCLE. Use the following
+ parameters to decide if that's possible:
+ PS - The partial schedule.
+ U - The serial number of U_NODE.
+ NUM_SPLITS - The number of row splits made so far.
+ MUST_PRECEDE - The nodes that must precede U_NODE. (only valid at
+ the first row of the scheduling window)
+ MUST_FOLLOW - The nodes that must follow U_NODE. (only valid at the
+ last row of the scheduling window) */
+
+static bool
+try_scheduling_node_in_cycle (partial_schedule_ptr ps, ddg_node_ptr u_node,
+ int u, int cycle, sbitmap sched_nodes,
+ int *num_splits, sbitmap must_precede,
+ sbitmap must_follow)
+{
+ ps_insn_ptr psi;
+ bool success = 0;
+
+ verify_partial_schedule (ps, sched_nodes);
+ psi = ps_add_node_check_conflicts (ps, u_node, cycle,
+ must_precede, must_follow);
+ if (psi)
+ {
+ SCHED_TIME (u_node) = cycle;
+ SET_BIT (sched_nodes, u);
+ success = 1;
+ *num_splits = 0;
+ if (dump_file)
+ fprintf (dump_file, "Scheduled w/o split in %d\n", cycle);
+
+ }
+
+ return success;
+}
+
/* This function implements the scheduling algorithm for SMS according to the
above algorithm. */
static partial_schedule_ptr
int i, c, success, num_splits = 0;
int flush_and_start_over = true;
int num_nodes = g->num_nodes;
- ddg_edge_ptr e;
- ps_insn_ptr psi;
int start, end, step; /* Place together into one struct? */
sbitmap sched_nodes = sbitmap_alloc (num_nodes);
sbitmap must_precede = sbitmap_alloc (num_nodes);
ddg_node_ptr u_node = &ps->g->nodes[u];
rtx insn = u_node->insn;
- if (!INSN_P (insn))
+ if (!NONDEBUG_INSN_P (insn))
{
RESET_BIT (tobe_scheduled, u);
continue;
fprintf (dump_file, "\nTrying to schedule node %d \
INSN = %d in (%d .. %d) step %d\n", u, (INSN_UID
(g->nodes[u].insn)), start, end, step);
- /* Use must_follow & must_precede bitmaps to determine order
- of nodes within the cycle. */
-
- /* use must_follow & must_precede bitmaps to determine order
- of nodes within the cycle. */
- sbitmap_zero (must_precede);
- sbitmap_zero (must_follow);
- /* TODO: We can add an insn to the must_precede or must_follow
- bitmaps only if it has tight dependence to U and they
- both scheduled in the same row. The current check is less
- conservative and content with the fact that both U and the
- insn are scheduled in the same row. */
- for (e = u_node->in; e != 0; e = e->next_in)
- if (TEST_BIT (sched_nodes, e->src->cuid)
- && (SMODULO (SCHED_TIME (e->src), ii) ==
- SMODULO (start, ii)))
- SET_BIT (must_precede, e->src->cuid);
-
- for (e = u_node->out; e != 0; e = e->next_out)
- if (TEST_BIT (sched_nodes, e->dest->cuid)
- && (SMODULO (SCHED_TIME (e->dest), ii) ==
- SMODULO ((end - step), ii)))
- SET_BIT (must_follow, e->dest->cuid);
gcc_assert ((step > 0 && start < end)
|| (step < 0 && start > end));
+ calculate_must_precede_follow (u_node, start, end, step, ii,
+ sched_nodes, must_precede,
+ must_follow);
+
for (c = start; c != end; c += step)
{
- verify_partial_schedule (ps, sched_nodes);
+ sbitmap tmp_precede = NULL;
+ sbitmap tmp_follow = NULL;
- psi = ps_add_node_check_conflicts (ps, u_node, c,
- must_precede,
- must_follow);
-
- if (psi)
+ if (c == start)
+ {
+ if (step == 1)
+ tmp_precede = must_precede;
+ else /* step == -1. */
+ tmp_follow = must_follow;
+ }
+ if (c == end - step)
{
- SCHED_TIME (u_node) = c;
- SET_BIT (sched_nodes, u);
- success = 1;
- num_splits = 0;
- if (dump_file)
- fprintf (dump_file, "Scheduled w/o split in %d\n", c);
-
- break;
+ if (step == 1)
+ tmp_follow = must_follow;
+ else /* step == -1. */
+ tmp_precede = must_precede;
}
+
+ success =
+ try_scheduling_node_in_cycle (ps, u_node, u, c,
+ sched_nodes,
+ &num_splits, tmp_precede,
+ tmp_follow);
+ if (success)
+ break;
}
+
verify_partial_schedule (ps, sched_nodes);
}
if (!success)
}
num_splits++;
+ /* The scheduling window is exclusive of 'end'
+ whereas compute_split_window() expects an inclusive,
+ ordered range. */
if (step == 1)
- split_row = compute_split_row (sched_nodes, start, end,
+ split_row = compute_split_row (sched_nodes, start, end - 1,
ps->ii, u_node);
else
- split_row = compute_split_row (sched_nodes, end, start,
+ split_row = compute_split_row (sched_nodes, end + 1, start,
ps->ii, u_node);
ps_insert_empty_row (ps, split_row, sched_nodes);
/* Given U_NODE which is the node that failed to be scheduled; LOW and
UP which are the boundaries of it's scheduling window; compute using
- SCHED_NODES and II a row in the partial schedule that can be splitted
+ SCHED_NODES and II a row in the partial schedule that can be split
which will separate a critical predecessor from a critical successor
thereby expanding the window, and return it. */
static int
static void order_nodes_of_sccs (ddg_all_sccs_ptr, int * result);
static int order_nodes_in_scc (ddg_ptr, sbitmap, sbitmap, int*, int);
-static nopa calculate_order_params (ddg_ptr, int mii);
+static nopa calculate_order_params (ddg_ptr, int, int *);
static int find_max_asap (ddg_ptr, sbitmap);
static int find_max_hv_min_mob (ddg_ptr, sbitmap);
static int find_max_dv_min_mob (ddg_ptr, sbitmap);
sbitmap_zero (tmp);
+ if (dump_file)
+ fprintf (dump_file, "SMS final nodes order: \n");
+
for (i = 0; i < num_nodes; i++)
{
int u = node_order[i];
+ if (dump_file)
+ fprintf (dump_file, "%d ", u);
gcc_assert (u < num_nodes && u >= 0 && !TEST_BIT (tmp, u));
SET_BIT (tmp, u);
}
+ if (dump_file)
+ fprintf (dump_file, "\n");
+
sbitmap_free (tmp);
}
/* Order the nodes of G for scheduling and pass the result in
NODE_ORDER. Also set aux.count of each node to ASAP.
- Return the recMII for the given DDG. */
+ Put maximal ASAP to PMAX_ASAP. Return the recMII for the given DDG. */
static int
-sms_order_nodes (ddg_ptr g, int mii, int * node_order)
+sms_order_nodes (ddg_ptr g, int mii, int * node_order, int *pmax_asap)
{
int i;
int rec_mii = 0;
ddg_all_sccs_ptr sccs = create_ddg_all_sccs (g);
- nopa nops = calculate_order_params (g, mii);
+ nopa nops = calculate_order_params (g, mii, pmax_asap);
if (dump_file)
print_sccs (dump_file, sccs, g);
sbitmap_zero (prev_sccs);
sbitmap_ones (ones);
- /* Perfrom the node ordering starting from the SCC with the highest recMII.
+ /* Perform the node ordering starting from the SCC with the highest recMII.
For each SCC order the nodes according to their ASAP/ALAP/HEIGHT etc. */
for (i = 0; i < all_sccs->num_sccs; i++)
{
/* MII is needed if we consider backarcs (that do not close recursive cycles). */
static struct node_order_params *
-calculate_order_params (ddg_ptr g, int mii ATTRIBUTE_UNUSED)
+calculate_order_params (ddg_ptr g, int mii ATTRIBUTE_UNUSED, int *pmax_asap)
{
int u;
int max_asap;
HEIGHT (e->dest) + e->latency);
}
}
+ if (dump_file)
+ {
+ fprintf (dump_file, "\nOrder params\n");
+ for (u = 0; u < num_nodes; u++)
+ {
+ ddg_node_ptr u_node = &g->nodes[u];
+
+ fprintf (dump_file, "node %d, ASAP: %d, ALAP: %d, HEIGHT: %d\n", u,
+ ASAP (u_node), ALAP (u_node), HEIGHT (u_node));
+ }
+ }
+ *pmax_asap = max_asap;
return node_order_params_arr;
}
{
ps_insn_ptr ps_i = ps->rows[i];
- fprintf (dump, "\n[CYCLE %d ]: ", i);
+ fprintf (dump, "\n[ROW %d ]: ", i);
while (ps_i)
{
fprintf (dump, "%d, ",
next_ps_i;
next_ps_i = next_ps_i->next_in_row)
{
- if (TEST_BIT (must_follow, next_ps_i->node->cuid)
+ if (must_follow && TEST_BIT (must_follow, next_ps_i->node->cuid)
&& ! first_must_follow)
first_must_follow = next_ps_i;
- if (TEST_BIT (must_precede, next_ps_i->node->cuid))
+ if (must_precede && TEST_BIT (must_precede, next_ps_i->node->cuid))
{
/* If we have already met a node that must follow, then
there is no possible column. */
}
/* Advances the PS_INSN one column in its current row; returns false
- in failure and true in success. Bit N is set in MUST_FOLLOW if
- the node with cuid N must be come after the node pointed to by
+ in failure and true in success. Bit N is set in MUST_FOLLOW if
+ the node with cuid N must be come after the node pointed to by
PS_I when scheduled in the same cycle. */
static int
ps_insn_advance_column (partial_schedule_ptr ps, ps_insn_ptr ps_i,
/* Check if next_in_row is dependent on ps_i, both having same sched
times (typically ANTI_DEP). If so, ps_i cannot skip over it. */
- if (TEST_BIT (must_follow, next_node->cuid))
+ if (must_follow && TEST_BIT (must_follow, next_node->cuid))
return false;
/* Advance PS_I over its next_in_row in the doubly linked list. */
}
/* Inserts a DDG_NODE to the given partial schedule at the given cycle.
- Returns 0 if this is not possible and a PS_INSN otherwise. Bit N is
- set in MUST_PRECEDE/MUST_FOLLOW if the node with cuid N must be come
- before/after (respectively) the node pointed to by PS_I when scheduled
+ Returns 0 if this is not possible and a PS_INSN otherwise. Bit N is
+ set in MUST_PRECEDE/MUST_FOLLOW if the node with cuid N must be come
+ before/after (respectively) the node pointed to by PS_I when scheduled
in the same cycle. */
static ps_insn_ptr
add_node_to_ps (partial_schedule_ptr ps, ddg_node_ptr node, int cycle,
{
rtx insn = crr_insn->node->insn;
- if (!INSN_P (insn))
+ if (!NONDEBUG_INSN_P (insn))
continue;
/* Check if there is room for the current insn. */
return true;
/* Update the DFA state and return with failure if the DFA found
- recource conflicts. */
+ resource conflicts. */
if (state_transition (curr_state, insn) >= 0)
return true;
/* Checks if the given node causes resource conflicts when added to PS at
cycle C. If not the node is added to PS and returned; otherwise zero
- is returned. Bit N is set in MUST_PRECEDE/MUST_FOLLOW if the node with
- cuid N must be come before/after (respectively) the node pointed to by
+ is returned. Bit N is set in MUST_PRECEDE/MUST_FOLLOW if the node with
+ cuid N must be come before/after (respectively) the node pointed to by
PS_I when scheduled in the same cycle. */
ps_insn_ptr
ps_add_node_check_conflicts (partial_schedule_ptr ps, ddg_node_ptr n,
return 0;
}
-struct tree_opt_pass pass_sms =
+struct rtl_opt_pass pass_sms =
{
+ {
+ RTL_PASS,
"sms", /* name */
gate_handle_sms, /* gate */
rest_of_handle_sms, /* execute */
TODO_dump_func, /* todo_flags_start */
TODO_df_finish | TODO_verify_rtl_sharing |
TODO_dump_func |
- TODO_ggc_collect, /* todo_flags_finish */
- 'm' /* letter */
+ TODO_ggc_collect /* todo_flags_finish */
+ }
};
OSDN Git Service
