< prev index next >
src/hotspot/share/opto/parse2.cpp
Print this page
@@ -184,14 +184,14 @@
return ptr;
}
// returns IfNode
-IfNode* Parse::jump_if_fork_int(Node* a, Node* b, BoolTest::mask mask) {
- Node *cmp = _gvn.transform( new CmpINode( a, b)); // two cases: shiftcount > 32 and shiftcount <= 32
- Node *tst = _gvn.transform( new BoolNode( cmp, mask));
- IfNode *iff = create_and_map_if( control(), tst, ((mask == BoolTest::eq) ? PROB_STATIC_INFREQUENT : PROB_FAIR), COUNT_UNKNOWN );
+IfNode* Parse::jump_if_fork_int(Node* a, Node* b, BoolTest::mask mask, float prob, float cnt) {
+ Node *cmp = _gvn.transform(new CmpINode(a, b)); // two cases: shiftcount > 32 and shiftcount <= 32
+ Node *tst = _gvn.transform(new BoolNode(cmp, mask));
+ IfNode *iff = create_and_map_if(control(), tst, prob, cnt);
return iff;
}
// return Region node
Node* Parse::jump_if_join(Node* iffalse, Node* iftrue) {
@@ -203,44 +203,74 @@
region = _gvn.transform(region);
set_control (region);
return region;
}
+// sentinel value for the target bci to mark never taken branches
+// (according to profiling)
+static const int never_reached = INT_MAX;
//------------------------------helper for tableswitch-------------------------
-void Parse::jump_if_true_fork(IfNode *iff, int dest_bci_if_true, int prof_table_index) {
+void Parse::jump_if_true_fork(IfNode *iff, int dest_bci_if_true, int prof_table_index, bool unc) {
// True branch, use existing map info
{ PreserveJVMState pjvms(this);
Node *iftrue = _gvn.transform( new IfTrueNode (iff) );
set_control( iftrue );
+ if (unc) {
+ repush_if_args();
+ uncommon_trap(Deoptimization::Reason_unstable_if,
+ Deoptimization::Action_reinterpret,
+ NULL,
+ "taken always");
+ } else {
+ assert(dest_bci_if_true != never_reached, "inconsistent dest");
profile_switch_case(prof_table_index);
merge_new_path(dest_bci_if_true);
}
+ }
// False branch
Node *iffalse = _gvn.transform( new IfFalseNode(iff) );
set_control( iffalse );
}
-void Parse::jump_if_false_fork(IfNode *iff, int dest_bci_if_true, int prof_table_index) {
+void Parse::jump_if_false_fork(IfNode *iff, int dest_bci_if_true, int prof_table_index, bool unc) {
// True branch, use existing map info
{ PreserveJVMState pjvms(this);
Node *iffalse = _gvn.transform( new IfFalseNode (iff) );
set_control( iffalse );
+ if (unc) {
+ repush_if_args();
+ uncommon_trap(Deoptimization::Reason_unstable_if,
+ Deoptimization::Action_reinterpret,
+ NULL,
+ "taken never");
+ } else {
+ assert(dest_bci_if_true != never_reached, "inconsistent dest");
profile_switch_case(prof_table_index);
merge_new_path(dest_bci_if_true);
}
+ }
// False branch
Node *iftrue = _gvn.transform( new IfTrueNode(iff) );
set_control( iftrue );
}
-void Parse::jump_if_always_fork(int dest_bci, int prof_table_index) {
+void Parse::jump_if_always_fork(int dest_bci, int prof_table_index, bool unc) {
// False branch, use existing map and control()
+ if (unc) {
+ repush_if_args();
+ uncommon_trap(Deoptimization::Reason_unstable_if,
+ Deoptimization::Action_reinterpret,
+ NULL,
+ "taken never");
+ } else {
+ assert(dest_bci != never_reached, "inconsistent dest");
profile_switch_case(prof_table_index);
merge_new_path(dest_bci);
+ }
}
extern "C" {
static int jint_cmp(const void *i, const void *j) {
@@ -259,55 +289,123 @@
// a range of integers coupled with a bci destination
jint _lo; // inclusive lower limit
jint _hi; // inclusive upper limit
int _dest;
int _table_index; // index into method data table
+ float _cnt; // how many times this range was hit according to profiling
public:
jint lo() const { return _lo; }
jint hi() const { return _hi; }
int dest() const { return _dest; }
int table_index() const { return _table_index; }
bool is_singleton() const { return _lo == _hi; }
+ float cnt() const { return _cnt; }
- void setRange(jint lo, jint hi, int dest, int table_index) {
+ void setRange(jint lo, jint hi, int dest, int table_index, float cnt) {
assert(lo <= hi, "must be a non-empty range");
- _lo = lo, _hi = hi; _dest = dest; _table_index = table_index;
+ _lo = lo, _hi = hi; _dest = dest; _table_index = table_index; _cnt = cnt;
+ assert(_cnt >= 0, "");
}
- bool adjoinRange(jint lo, jint hi, int dest, int table_index) {
+ bool adjoinRange(jint lo, jint hi, int dest, int table_index, float cnt, bool trim_ranges) {
assert(lo <= hi, "must be a non-empty range");
- if (lo == _hi+1 && dest == _dest && table_index == _table_index) {
+ if (lo == _hi+1 && table_index == _table_index) {
+ // see merge_ranges() comment below
+ if (trim_ranges) {
+ if (cnt == 0) {
+ if (_cnt != 0) {
+ return false;
+ }
+ if (dest != _dest) {
+ _dest = never_reached;
+ }
+ } else {
+ if (_cnt == 0) {
+ return false;
+ }
+ if (dest != _dest) {
+ return false;
+ }
+ }
+ } else {
+ if (dest != _dest) {
+ return false;
+ }
+ }
_hi = hi;
+ _cnt += cnt;
return true;
}
return false;
}
- void set (jint value, int dest, int table_index) {
- setRange(value, value, dest, table_index);
+ void set (jint value, int dest, int table_index, float cnt) {
+ setRange(value, value, dest, table_index, cnt);
+ }
+ bool adjoin(jint value, int dest, int table_index, float cnt, bool trim_ranges) {
+ return adjoinRange(value, value, dest, table_index, cnt, trim_ranges);
}
- bool adjoin(jint value, int dest, int table_index) {
- return adjoinRange(value, value, dest, table_index);
+ bool adjoin(SwitchRange& other) {
+ return adjoinRange(other._lo, other._hi, other._dest, other._table_index, other._cnt, false);
}
void print() {
if (is_singleton())
- tty->print(" {%d}=>%d", lo(), dest());
+ tty->print(" {%d}=>%d (cnt=%f)", lo(), dest(), cnt());
else if (lo() == min_jint)
- tty->print(" {..%d}=>%d", hi(), dest());
+ tty->print(" {..%d}=>%d (cnt=%f)", hi(), dest(), cnt());
else if (hi() == max_jint)
- tty->print(" {%d..}=>%d", lo(), dest());
+ tty->print(" {%d..}=>%d (cnt=%f)", lo(), dest(), cnt());
else
- tty->print(" {%d..%d}=>%d", lo(), hi(), dest());
+ tty->print(" {%d..%d}=>%d (cnt=%f)", lo(), hi(), dest(), cnt());
}
};
+// We try to minimize the number of ranges and the size of the taken
+// ones using profiling data. When ranges are created,
+// SwitchRange::adjoinRange() only allows 2 adjoining ranges to merge
+// if both were never hit or both were hit to build longer unreached
+// ranges. Here, we now merge adjoining ranges with the same
+// destination and finally set destination of unreached ranges to the
+// special value never_reached because it can help minimize the number
+// of tests that are necessary.
+//
+// For instance:
+// [0, 1] to target1 sometimes taken
+// [1, 2] to target1 never taken
+// [2, 3] to target2 never taken
+// would lead to:
+// [0, 1] to target1 sometimes taken
+// [1, 3] never taken
+//
+// (first 2 ranges to target1 are not merged)
+static void merge_ranges(SwitchRange* ranges, int& rp) {
+ if (rp == 0) {
+ return;
+ }
+ int shift = 0;
+ for (int j = 0; j < rp; j++) {
+ SwitchRange& r1 = ranges[j-shift];
+ SwitchRange& r2 = ranges[j+1];
+ if (r1.adjoin(r2)) {
+ shift++;
+ } else if (shift > 0) {
+ ranges[j+1-shift] = r2;
+ }
+ }
+ rp -= shift;
+ for (int j = 0; j <= rp; j++) {
+ SwitchRange& r = ranges[j];
+ if (r.cnt() == 0 && r.dest() != never_reached) {
+ r.setRange(r.lo(), r.hi(), never_reached, r.table_index(), r.cnt());
+ }
+ }
+}
//-------------------------------do_tableswitch--------------------------------
void Parse::do_tableswitch() {
Node* lookup = pop();
-
// Get information about tableswitch
int default_dest = iter().get_dest_table(0);
int lo_index = iter().get_int_table(1);
int hi_index = iter().get_int_table(2);
int len = hi_index - lo_index + 1;
@@ -317,35 +415,62 @@
maybe_add_safepoint(default_dest);
merge(default_dest);
return;
}
+ ciMethodData* methodData = method()->method_data();
+ ciMultiBranchData* profile = NULL;
+ if (methodData->is_mature() && UseSwitchProfiling) {
+ ciProfileData* data = methodData->bci_to_data(bci());
+ if (data != NULL && data->is_MultiBranchData()) {
+ profile = (ciMultiBranchData*)data;
+ }
+ }
+ bool trim_ranges = !method_data_update() && !C->too_many_traps(method(), bci(), Deoptimization::Reason_unstable_if);
+
// generate decision tree, using trichotomy when possible
int rnum = len+2;
bool makes_backward_branch = false;
SwitchRange* ranges = NEW_RESOURCE_ARRAY(SwitchRange, rnum);
int rp = -1;
if (lo_index != min_jint) {
- ranges[++rp].setRange(min_jint, lo_index-1, default_dest, NullTableIndex);
+ uint cnt = 1;
+ if (profile != NULL) {
+ cnt = profile->default_count() / (hi_index != max_jint ? 2 : 1);
+ }
+ ranges[++rp].setRange(min_jint, lo_index-1, default_dest, NullTableIndex, cnt);
}
for (int j = 0; j < len; j++) {
jint match_int = lo_index+j;
int dest = iter().get_dest_table(j+3);
makes_backward_branch |= (dest <= bci());
int table_index = method_data_update() ? j : NullTableIndex;
- if (rp < 0 || !ranges[rp].adjoin(match_int, dest, table_index)) {
- ranges[++rp].set(match_int, dest, table_index);
+ uint cnt = 1;
+ if (profile != NULL) {
+ cnt = profile->count_at(j);
+ }
+ if (rp < 0 || !ranges[rp].adjoin(match_int, dest, table_index, cnt, trim_ranges)) {
+ ranges[++rp].set(match_int, dest, table_index, cnt);
}
}
jint highest = lo_index+(len-1);
assert(ranges[rp].hi() == highest, "");
- if (highest != max_jint
- && !ranges[rp].adjoinRange(highest+1, max_jint, default_dest, NullTableIndex)) {
- ranges[++rp].setRange(highest+1, max_jint, default_dest, NullTableIndex);
+ if (highest != max_jint) {
+ uint cnt = 1;
+ if (profile != NULL) {
+ cnt = profile->default_count() / (lo_index != min_jint ? 2 : 1);
+ }
+ if (!ranges[rp].adjoinRange(highest+1, max_jint, default_dest, NullTableIndex, cnt, trim_ranges)) {
+ ranges[++rp].setRange(highest+1, max_jint, default_dest, NullTableIndex, cnt);
+ }
}
assert(rp < len+2, "not too many ranges");
+ if (trim_ranges) {
+ merge_ranges(ranges, rp);
+ }
+
// Safepoint in case if backward branch observed
if( makes_backward_branch && UseLoopSafepoints )
add_safepoint();
jump_switch_ranges(lookup, &ranges[0], &ranges[rp]);
@@ -363,52 +488,267 @@
maybe_add_safepoint(default_dest);
merge(default_dest);
return;
}
+ ciMethodData* methodData = method()->method_data();
+ ciMultiBranchData* profile = NULL;
+ if (methodData->is_mature() && UseSwitchProfiling) {
+ ciProfileData* data = methodData->bci_to_data(bci());
+ if (data != NULL && data->is_MultiBranchData()) {
+ profile = (ciMultiBranchData*)data;
+ }
+ }
+ bool trim_ranges = !method_data_update() && !C->too_many_traps(method(), bci(), Deoptimization::Reason_unstable_if);
+
// generate decision tree, using trichotomy when possible
- jint* table = NEW_RESOURCE_ARRAY(jint, len*2);
+ jint* table = NEW_RESOURCE_ARRAY(jint, len*3);
{
- for( int j = 0; j < len; j++ ) {
- table[j+j+0] = iter().get_int_table(2+j+j);
- table[j+j+1] = iter().get_dest_table(2+j+j+1);
+ for (int j = 0; j < len; j++) {
+ table[3*j+0] = iter().get_int_table(2+2*j);
+ table[3*j+1] = iter().get_dest_table(2+2*j+1);
+ table[3*j+2] = profile == NULL ? 1 : profile->count_at(j);
+ }
+ qsort(table, len, 3*sizeof(table[0]), jint_cmp);
}
- qsort( table, len, 2*sizeof(table[0]), jint_cmp );
+
+ float defaults = 0;
+ jint prev = min_jint;
+ for (int j = 0; j < len; j++) {
+ jint match_int = table[3*j+0];
+ if (match_int != prev) {
+ defaults += (float)match_int - prev;
+ }
+ prev = match_int+1;
+ }
+ if (prev-1 != max_jint) {
+ defaults += (float)max_jint - prev + 1;
+ }
+ float default_cnt = 1;
+ if (profile != NULL) {
+ default_cnt = profile->default_count()/defaults;
}
int rnum = len*2+1;
bool makes_backward_branch = false;
SwitchRange* ranges = NEW_RESOURCE_ARRAY(SwitchRange, rnum);
int rp = -1;
- for( int j = 0; j < len; j++ ) {
- jint match_int = table[j+j+0];
- int dest = table[j+j+1];
+ for (int j = 0; j < len; j++) {
+ jint match_int = table[3*j+0];
+ int dest = table[3*j+1];
+ int cnt = table[3*j+2];
int next_lo = rp < 0 ? min_jint : ranges[rp].hi()+1;
int table_index = method_data_update() ? j : NullTableIndex;
makes_backward_branch |= (dest <= bci());
- if( match_int != next_lo ) {
- ranges[++rp].setRange(next_lo, match_int-1, default_dest, NullTableIndex);
+ float c = default_cnt * ((float)match_int - next_lo);
+ if (match_int != next_lo && (rp < 0 || !ranges[rp].adjoinRange(next_lo, match_int-1, default_dest, NullTableIndex, c, trim_ranges))) {
+ assert(default_dest != never_reached, "sentinel value for dead destinations");
+ ranges[++rp].setRange(next_lo, match_int-1, default_dest, NullTableIndex, c);
}
- if( rp < 0 || !ranges[rp].adjoin(match_int, dest, table_index) ) {
- ranges[++rp].set(match_int, dest, table_index);
+ if (rp < 0 || !ranges[rp].adjoin(match_int, dest, table_index, cnt, trim_ranges)) {
+ assert(dest != never_reached, "sentinel value for dead destinations");
+ ranges[++rp].set(match_int, dest, table_index, cnt);
}
}
- jint highest = table[2*(len-1)];
+ jint highest = table[3*(len-1)];
assert(ranges[rp].hi() == highest, "");
- if( highest != max_jint
- && !ranges[rp].adjoinRange(highest+1, max_jint, default_dest, NullTableIndex) ) {
- ranges[++rp].setRange(highest+1, max_jint, default_dest, NullTableIndex);
+ if (highest != max_jint &&
+ !ranges[rp].adjoinRange(highest+1, max_jint, default_dest, NullTableIndex, default_cnt * ((float)max_jint - highest), trim_ranges)) {
+ ranges[++rp].setRange(highest+1, max_jint, default_dest, NullTableIndex, default_cnt * ((float)max_jint - highest));
}
assert(rp < rnum, "not too many ranges");
+ if (trim_ranges) {
+ merge_ranges(ranges, rp);
+ }
+
// Safepoint in case backward branch observed
- if( makes_backward_branch && UseLoopSafepoints )
+ if (makes_backward_branch && UseLoopSafepoints)
add_safepoint();
jump_switch_ranges(lookup, &ranges[0], &ranges[rp]);
}
+static float if_prob(float taken_cnt, float total_cnt) {
+ assert(taken_cnt <= total_cnt, "");
+ if (total_cnt == 0) {
+ return PROB_FAIR;
+ }
+ float p = taken_cnt / total_cnt;
+ return MIN2(MAX2(p, PROB_MIN), PROB_MAX);
+}
+
+static float if_cnt(float cnt) {
+ if (cnt == 0) {
+ return COUNT_UNKNOWN;
+ }
+ return cnt;
+}
+
+static float sum_of_cnts(SwitchRange *lo, SwitchRange *hi) {
+ float total_cnt = 0;
+ for (SwitchRange* sr = lo; sr <= hi; sr++) {
+ total_cnt += sr->cnt();
+ }
+ return total_cnt;
+}
+
+class SwitchRanges : public ResourceObj {
+public:
+ SwitchRange* _lo;
+ SwitchRange* _hi;
+ SwitchRange* _mid;
+ float _cost;
+
+ enum {
+ Start,
+ LeftDone,
+ RightDone,
+ Done
+ } _state;
+
+ SwitchRanges(SwitchRange *lo, SwitchRange *hi)
+ : _lo(lo), _hi(hi), _mid(NULL),
+ _cost(0), _state(Start) {
+ }
+
+ SwitchRanges()
+ : _lo(NULL), _hi(NULL), _mid(NULL),
+ _cost(0), _state(Start) {}
+};
+
+// Estimate cost of performing a binary search on lo..hi
+static float compute_tree_cost(SwitchRange *lo, SwitchRange *hi, float total_cnt) {
+ GrowableArray<SwitchRanges> tree;
+ SwitchRanges root(lo, hi);
+ tree.push(root);
+
+ float cost = 0;
+ do {
+ SwitchRanges& r = *tree.adr_at(tree.length()-1);
+ if (r._hi != r._lo) {
+ if (r._mid == NULL) {
+ float r_cnt = sum_of_cnts(r._lo, r._hi);
+
+ if (r_cnt == 0) {
+ tree.pop();
+ cost = 0;
+ continue;
+ }
+
+ SwitchRange* mid = NULL;
+ mid = r._lo;
+ for (float cnt = 0; ; ) {
+ assert(mid <= r._hi, "out of bounds");
+ cnt += mid->cnt();
+ if (cnt > r_cnt / 2) {
+ break;
+ }
+ mid++;
+ }
+ assert(mid <= r._hi, "out of bounds");
+ r._mid = mid;
+ r._cost = r_cnt / total_cnt;
+ }
+ r._cost += cost;
+ if (r._state < SwitchRanges::LeftDone && r._mid > r._lo) {
+ cost = 0;
+ r._state = SwitchRanges::LeftDone;
+ tree.push(SwitchRanges(r._lo, r._mid-1));
+ } else if (r._state < SwitchRanges::RightDone) {
+ cost = 0;
+ r._state = SwitchRanges::RightDone;
+ tree.push(SwitchRanges(r._mid == r._lo ? r._mid+1 : r._mid, r._hi));
+ } else {
+ tree.pop();
+ cost = r._cost;
+ }
+ } else {
+ tree.pop();
+ cost = r._cost;
+ }
+ } while (tree.length() > 0);
+
+
+ return cost;
+}
+
+// It sometimes pays off to test most common ranges before the binary search
+void Parse::linear_search_switch_ranges(Node* key_val, SwitchRange*& lo, SwitchRange*& hi) {
+ uint nr = hi - lo + 1;
+ float total_cnt = sum_of_cnts(lo, hi);
+
+ float min = compute_tree_cost(lo, hi, total_cnt);
+ float extra = 1;
+ float sub = 0;
+
+ SwitchRange* array1 = lo;
+ SwitchRange* array2 = NEW_RESOURCE_ARRAY(SwitchRange, nr);
+
+ SwitchRange* ranges = NULL;
+
+ while (nr >= 2) {
+ assert(lo == array1 || lo == array2, "one the 2 already allocated arrays");
+ ranges = (lo == array1) ? array2 : array1;
+
+ // Find highest frequency range
+ SwitchRange* candidate = lo;
+ for (SwitchRange* sr = lo+1; sr <= hi; sr++) {
+ if (sr->cnt() > candidate->cnt()) {
+ candidate = sr;
+ }
+ }
+ SwitchRange most_freq = *candidate;
+ if (most_freq.cnt() == 0) {
+ break;
+ }
+
+ // Copy remaining ranges into another array
+ int shift = 0;
+ for (uint i = 0; i < nr; i++) {
+ SwitchRange* sr = &lo[i];
+ if (sr != candidate) {
+ ranges[i-shift] = *sr;
+ } else {
+ shift++;
+ if (i > 0 && i < nr-1) {
+ SwitchRange prev = lo[i-1];
+ prev.setRange(prev.lo(), sr->hi(), prev.dest(), prev.table_index(), prev.cnt());
+ if (prev.adjoin(lo[i+1])) {
+ shift++;
+ i++;
+ }
+ ranges[i-shift] = prev;
+ }
+ }
+ }
+ nr -= shift;
+
+ // Evaluate cost of testing the most common range and performing a
+ // binary search on the other ranges
+ float cost = extra + compute_tree_cost(&ranges[0], &ranges[nr-1], total_cnt);
+ if (cost >= min) {
+ break;
+ }
+ // swap arrays
+ lo = &ranges[0];
+ hi = &ranges[nr-1];
+
+ // It pays off: emit the test for the most common range
+ assert(most_freq.cnt() > 0, "must be taken");
+ Node* val = _gvn.transform(new SubINode(key_val, _gvn.intcon(most_freq.lo())));
+ Node* cmp = _gvn.transform(new CmpUNode(val, _gvn.intcon(most_freq.hi() - most_freq.lo())));
+ Node* tst = _gvn.transform(new BoolNode(cmp, BoolTest::le));
+ IfNode* iff = create_and_map_if(control(), tst, if_prob(most_freq.cnt(), total_cnt), if_cnt(most_freq.cnt()));
+ jump_if_true_fork(iff, most_freq.dest(), most_freq.table_index(), false);
+
+ sub += most_freq.cnt() / total_cnt;
+ extra += 1 - sub;
+ min = cost;
+ }
+}
+
//----------------------------create_jump_tables-------------------------------
bool Parse::create_jump_tables(Node* key_val, SwitchRange* lo, SwitchRange* hi) {
// Are jumptables enabled
if (!UseJumpTables) return false;
@@ -416,10 +756,12 @@
if (!Matcher::has_match_rule(Op_Jump)) return false;
// Don't make jump table if profiling
if (method_data_update()) return false;
+ bool trim_ranges = !C->too_many_traps(method(), bci(), Deoptimization::Reason_unstable_if);
+
// Decide if a guard is needed to lop off big ranges at either (or
// both) end(s) of the input set. We'll call this the default target
// even though we can't be sure that it is the true "default".
bool needs_guard = false;
@@ -437,25 +779,50 @@
} else {
total_outlier_size = hi_size;
default_dest = hi->dest();
}
+ float total = sum_of_cnts(lo, hi);
+ float cost = compute_tree_cost(lo, hi, total);
+
// If a guard test will eliminate very sparse end ranges, then
// it is worth the cost of an extra jump.
+ float trimmed_cnt = 0;
if (total_outlier_size > (MaxJumpTableSparseness * 4)) {
needs_guard = true;
- if (default_dest == lo->dest()) lo++;
- if (default_dest == hi->dest()) hi--;
+ if (default_dest == lo->dest()) {
+ trimmed_cnt += lo->cnt();
+ lo++;
+ }
+ if (default_dest == hi->dest()) {
+ trimmed_cnt += hi->cnt();
+ hi--;
+ }
}
// Find the total number of cases and ranges
int64_t num_cases = ((int64_t)hi->hi()) - ((int64_t)lo->lo()) + 1;
int num_range = hi - lo + 1;
// Don't create table if: too large, too small, or too sparse.
- if (num_cases < MinJumpTableSize || num_cases > MaxJumpTableSize)
+ if (num_cases > MaxJumpTableSize)
+ return false;
+ if (UseSwitchProfiling) {
+ // MinJumpTableSize is set so with a well balanced binary tree,
+ // when the number of ranges is MinJumpTableSize, it's cheaper to
+ // go through a JumpNode that a tree of IfNodes. Average cost of a
+ // tree of IfNodes with MinJumpTableSize is
+ // log2f(MinJumpTableSize) comparisons. So if the cost computed
+ // from profile data is less than log2f(MinJumpTableSize) then
+ // going with the binary search is cheaper.
+ if (cost < log2f(MinJumpTableSize)) {
return false;
+ }
+ } else {
+ if (num_cases < MinJumpTableSize)
+ return false;
+ }
if (num_cases > (MaxJumpTableSparseness * num_range))
return false;
// Normalize table lookups to zero
int lowval = lo->lo();
@@ -463,14 +830,16 @@
// Generate a guard to protect against input keyvals that aren't
// in the switch domain.
if (needs_guard) {
Node* size = _gvn.intcon(num_cases);
- Node* cmp = _gvn.transform( new CmpUNode(key_val, size) );
- Node* tst = _gvn.transform( new BoolNode(cmp, BoolTest::ge) );
- IfNode* iff = create_and_map_if( control(), tst, PROB_FAIR, COUNT_UNKNOWN);
- jump_if_true_fork(iff, default_dest, NullTableIndex);
+ Node* cmp = _gvn.transform(new CmpUNode(key_val, size));
+ Node* tst = _gvn.transform(new BoolNode(cmp, BoolTest::ge));
+ IfNode* iff = create_and_map_if(control(), tst, if_prob(trimmed_cnt, total), if_cnt(trimmed_cnt));
+ jump_if_true_fork(iff, default_dest, NullTableIndex, trim_ranges && trimmed_cnt == 0);
+
+ total -= trimmed_cnt;
}
// Create an ideal node JumpTable that has projections
// of all possible ranges for a switch statement
// The key_val input must be converted to a pointer offset and scaled.
@@ -487,21 +856,48 @@
// than a switch value
Node *shiftWord = _gvn.MakeConX(wordSize);
key_val = _gvn.transform( new MulXNode( key_val, shiftWord));
// Create the JumpNode
- Node* jtn = _gvn.transform( new JumpNode(control(), key_val, num_cases) );
+ Arena* arena = C->comp_arena();
+ float* probs = (float*)arena->Amalloc(sizeof(float)*num_cases);
+ int i = 0;
+ if (total == 0) {
+ for (SwitchRange* r = lo; r <= hi; r++) {
+ for (int64_t j = r->lo(); j <= r->hi(); j++, i++) {
+ probs[i] = 1.0F / num_cases;
+ }
+ }
+ } else {
+ for (SwitchRange* r = lo; r <= hi; r++) {
+ float prob = r->cnt()/total;
+ for (int64_t j = r->lo(); j <= r->hi(); j++, i++) {
+ probs[i] = prob / (r->hi() - r->lo() + 1);
+ }
+ }
+ }
+
+ ciMethodData* methodData = method()->method_data();
+ ciMultiBranchData* profile = NULL;
+ if (methodData->is_mature()) {
+ ciProfileData* data = methodData->bci_to_data(bci());
+ if (data != NULL && data->is_MultiBranchData()) {
+ profile = (ciMultiBranchData*)data;
+ }
+ }
+
+ Node* jtn = _gvn.transform(new JumpNode(control(), key_val, num_cases, probs, profile == NULL ? COUNT_UNKNOWN : total));
// These are the switch destinations hanging off the jumpnode
- int i = 0;
+ i = 0;
for (SwitchRange* r = lo; r <= hi; r++) {
for (int64_t j = r->lo(); j <= r->hi(); j++, i++) {
Node* input = _gvn.transform(new JumpProjNode(jtn, i, r->dest(), (int)(j - lowval)));
{
PreserveJVMState pjvms(this);
set_control(input);
- jump_if_always_fork(r->dest(), r->table_index());
+ jump_if_always_fork(r->dest(), r->table_index(), trim_ranges && r->cnt() == 0);
}
}
}
assert(i == num_cases, "miscount of cases");
stop_and_kill_map(); // no more uses for this JVMS
@@ -509,18 +905,20 @@
}
//----------------------------jump_switch_ranges-------------------------------
void Parse::jump_switch_ranges(Node* key_val, SwitchRange *lo, SwitchRange *hi, int switch_depth) {
Block* switch_block = block();
+ bool trim_ranges = !method_data_update() && !C->too_many_traps(method(), bci(), Deoptimization::Reason_unstable_if);
if (switch_depth == 0) {
// Do special processing for the top-level call.
assert(lo->lo() == min_jint, "initial range must exhaust Type::INT");
assert(hi->hi() == max_jint, "initial range must exhaust Type::INT");
// Decrement pred-numbers for the unique set of nodes.
#ifdef ASSERT
+ if (!trim_ranges) {
// Ensure that the block's successors are a (duplicate-free) set.
int successors_counted = 0; // block occurrences in [hi..lo]
int unique_successors = switch_block->num_successors();
for (int i = 0; i < unique_successors; i++) {
Block* target = switch_block->successor_at(i);
@@ -532,10 +930,11 @@
}
assert(successors_found > 0, "successor must be known");
successors_counted += successors_found;
}
assert(successors_counted == (hi-lo)+1, "no unexpected successors");
+ }
#endif
// Maybe prune the inputs, based on the type of key_val.
jint min_val = min_jint;
jint max_val = max_jint;
@@ -543,14 +942,24 @@
if (ti != NULL) {
min_val = ti->_lo;
max_val = ti->_hi;
assert(min_val <= max_val, "invalid int type");
}
- while (lo->hi() < min_val) lo++;
- if (lo->lo() < min_val) lo->setRange(min_val, lo->hi(), lo->dest(), lo->table_index());
- while (hi->lo() > max_val) hi--;
- if (hi->hi() > max_val) hi->setRange(hi->lo(), max_val, hi->dest(), hi->table_index());
+ while (lo->hi() < min_val) {
+ lo++;
+ }
+ if (lo->lo() < min_val) {
+ lo->setRange(min_val, lo->hi(), lo->dest(), lo->table_index(), lo->cnt());
+ }
+ while (hi->lo() > max_val) {
+ hi--;
+ }
+ if (hi->hi() > max_val) {
+ hi->setRange(hi->lo(), max_val, hi->dest(), hi->table_index(), hi->cnt());
+ }
+
+ linear_search_switch_ranges(key_val, lo, hi);
}
#ifndef PRODUCT
if (switch_depth == 0) {
_max_switch_depth = 0;
@@ -558,46 +967,61 @@
}
#endif
assert(lo <= hi, "must be a non-empty set of ranges");
if (lo == hi) {
- jump_if_always_fork(lo->dest(), lo->table_index());
+ jump_if_always_fork(lo->dest(), lo->table_index(), trim_ranges && lo->cnt() == 0);
} else {
assert(lo->hi() == (lo+1)->lo()-1, "contiguous ranges");
assert(hi->lo() == (hi-1)->hi()+1, "contiguous ranges");
if (create_jump_tables(key_val, lo, hi)) return;
+ SwitchRange* mid = NULL;
+ float total_cnt = sum_of_cnts(lo, hi);
+
int nr = hi - lo + 1;
+ if (UseSwitchProfiling) {
+ // Don't keep the binary search tree balanced: pick up mid point
+ // that split frequencies in half.
+ float cnt = 0;
+ for (SwitchRange* sr = lo; sr <= hi; sr++) {
+ cnt += sr->cnt();
+ if (cnt >= total_cnt / 2) {
+ mid = sr;
+ break;
+ }
+ }
+ } else {
+ mid = lo + nr/2;
- SwitchRange* mid = lo + nr/2;
// if there is an easy choice, pivot at a singleton:
if (nr > 3 && !mid->is_singleton() && (mid-1)->is_singleton()) mid--;
assert(lo < mid && mid <= hi, "good pivot choice");
assert(nr != 2 || mid == hi, "should pick higher of 2");
assert(nr != 3 || mid == hi-1, "should pick middle of 3");
+ }
- Node *test_val = _gvn.intcon(mid->lo());
+
+ Node *test_val = _gvn.intcon(mid == lo ? mid->hi() : mid->lo());
if (mid->is_singleton()) {
- IfNode *iff_ne = jump_if_fork_int(key_val, test_val, BoolTest::ne);
- jump_if_false_fork(iff_ne, mid->dest(), mid->table_index());
+ IfNode *iff_ne = jump_if_fork_int(key_val, test_val, BoolTest::ne, 1-if_prob(mid->cnt(), total_cnt), if_cnt(mid->cnt()));
+ jump_if_false_fork(iff_ne, mid->dest(), mid->table_index(), trim_ranges && mid->cnt() == 0);
// Special Case: If there are exactly three ranges, and the high
// and low range each go to the same place, omit the "gt" test,
// since it will not discriminate anything.
- bool eq_test_only = (hi == lo+2 && hi->dest() == lo->dest());
- if (eq_test_only) {
- assert(mid == hi-1, "");
- }
+ bool eq_test_only = (hi == lo+2 && hi->dest() == lo->dest() && mid == hi-1) || mid == lo;
// if there is a higher range, test for it and process it:
if (mid < hi && !eq_test_only) {
// two comparisons of same values--should enable 1 test for 2 branches
// Use BoolTest::le instead of BoolTest::gt
- IfNode *iff_le = jump_if_fork_int(key_val, test_val, BoolTest::le);
+ float cnt = sum_of_cnts(lo, mid-1);
+ IfNode *iff_le = jump_if_fork_int(key_val, test_val, BoolTest::le, if_prob(cnt, total_cnt), if_cnt(cnt));
Node *iftrue = _gvn.transform( new IfTrueNode(iff_le) );
Node *iffalse = _gvn.transform( new IfFalseNode(iff_le) );
{ PreserveJVMState pjvms(this);
set_control(iffalse);
jump_switch_ranges(key_val, mid+1, hi, switch_depth+1);
@@ -605,29 +1029,38 @@
set_control(iftrue);
}
} else {
// mid is a range, not a singleton, so treat mid..hi as a unit
- IfNode *iff_ge = jump_if_fork_int(key_val, test_val, BoolTest::ge);
+ float cnt = sum_of_cnts(mid == lo ? mid+1 : mid, hi);
+ IfNode *iff_ge = jump_if_fork_int(key_val, test_val, mid == lo ? BoolTest::gt : BoolTest::ge, if_prob(cnt, total_cnt), if_cnt(cnt));
// if there is a higher range, test for it and process it:
if (mid == hi) {
- jump_if_true_fork(iff_ge, mid->dest(), mid->table_index());
+ jump_if_true_fork(iff_ge, mid->dest(), mid->table_index(), trim_ranges && cnt == 0);
} else {
Node *iftrue = _gvn.transform( new IfTrueNode(iff_ge) );
Node *iffalse = _gvn.transform( new IfFalseNode(iff_ge) );
{ PreserveJVMState pjvms(this);
set_control(iftrue);
- jump_switch_ranges(key_val, mid, hi, switch_depth+1);
+ jump_switch_ranges(key_val, mid == lo ? mid+1 : mid, hi, switch_depth+1);
}
set_control(iffalse);
}
}
// in any case, process the lower range
+ if (mid == lo) {
+ if (mid->is_singleton()) {
+ jump_switch_ranges(key_val, lo+1, hi, switch_depth+1);
+ } else {
+ jump_if_always_fork(lo->dest(), lo->table_index(), trim_ranges && lo->cnt() == 0);
+ }
+ } else {
jump_switch_ranges(key_val, lo, mid-1, switch_depth+1);
}
+ }
// Decrease pred_count for each successor after all is done.
if (switch_depth == 0) {
int unique_successors = switch_block->num_successors();
for (int i = 0; i < unique_successors; i++) {
@@ -722,11 +1155,11 @@
(divisor & ~(divisor-1)) == divisor) {
// yes !
Node *mask = _gvn.intcon((divisor - 1));
// Sigh, must handle negative dividends
Node *zero = _gvn.intcon(0);
- IfNode *ifff = jump_if_fork_int(a, zero, BoolTest::lt);
+ IfNode *ifff = jump_if_fork_int(a, zero, BoolTest::lt, PROB_FAIR, COUNT_UNKNOWN);
Node *iff = _gvn.transform( new IfFalseNode(ifff) );
Node *ift = _gvn.transform( new IfTrueNode (ifff) );
Node *reg = jump_if_join(ift, iff);
Node *phi = PhiNode::make(reg, NULL, TypeInt::INT);
// Negative path; negate/and/negate
< prev index next >