Merge

.hgtags

@@ -248,3 +248,4 @@ b32e2219736e42baaf45daf0ad67ed34f6033799 jdk9-b02
 7f655f31f9bcee618cf832f08176ad8c1ed3fdd3 jdk9-b03
 099891b1d86f3719e116ac717ffdafc90d037fb7 jdk9-b04
 dd311791ad6895a3989020dd6c6c46db87972ab8 jdk9-b05
+85dbdc227c5e11429b4fc4a8ba763f50107edd6e jdk9-b06

.hgtags-top-repo

@@ -248,3 +248,4 @@ cd3825b2983045784d6fc6d1729c799b08215752 jdk8-b120
 fd8d51bdf9aadf7ae83e65e8655c53581017c363 jdk9-b03
 cb4c3440bc2748101923e2488506e61009ab1bf5 jdk9-b04
 8c63f0b6ada282f27e3a80125e53c3be603f9af7 jdk9-b05
+d0b525cd31b87abeb6d5b7e3516953eeb13b323c jdk9-b06

corba/.hgtags

@@ -248,3 +248,4 @@ a7d3638deb2f4e33217b1ecf889479e90f9e5b50 jdk9-b00
 d338b892a13db19b093f85cf5f949a4504e4d31f jdk9-b03
 1ed19de263e1e0772da0269118cdd9deeb9fff04 jdk9-b04
 167c39eb44731a5d66770d0f00e231164653a2ff jdk9-b05
+a4bf701ac316946c2e5e83138ad8e687da6a4b30 jdk9-b06

hotspot/.hgtags

@@ -408,3 +408,4 @@ b188446de75bda5fc52d102cddf242c3ef5ecbdf jdk9-b02
 b2fee789d23f3cdabb3db4e51af43038e5692d3a jdk9-b03
 3812c088b9456ee22c933e88aee1ece71f4e783a jdk9-b04
 bdc5311e1db7598589b77015119b821bf8c828bd jdk9-b05
+52377a30a3f87b62d6135706997b8c7a47366e37 jdk9-b06

hotspot/src/cpu/sparc/vm/assembler_sparc.hpp

@@ -630,11 +630,20 @@ class Assembler : public AbstractAssembler  {
   }
 
  protected:
+  // Insert a nop if the previous is cbcond
+  void insert_nop_after_cbcond() {
+    if (UseCBCond && cbcond_before()) {
+      nop();
+    }
+  }
   // Delay slot helpers
   // cti is called when emitting control-transfer instruction,
   // BEFORE doing the emitting.
   // Only effective when assertion-checking is enabled.
   void cti() {
+    // A cbcond instruction immediately followed by a CTI
+    // instruction introduces pipeline stalls, we need to avoid that.
+    no_cbcond_before();
 #ifdef CHECK_DELAY
     assert_not_delayed("cti should not be in delay slot");
 #endif
@@ -658,7 +667,6 @@ class Assembler : public AbstractAssembler  {
   void no_cbcond_before() {
     assert(offset() == 0 || !cbcond_before(), "cbcond should not follow an other cbcond");
   }
-
 public:
 
   bool use_cbcond(Label& L) {

hotspot/src/cpu/sparc/vm/assembler_sparc.inline.hpp

@@ -54,33 +54,33 @@ inline void Assembler::emit_data(int x, RelocationHolder const& rspec) {
 inline void Assembler::add(Register s1, Register s2, Register d ) { emit_int32( op(arith_op) | rd(d) | op3(add_op3) | rs1(s1) | rs2(s2) ); }
 inline void Assembler::add(Register s1, int simm13a, Register d ) { emit_int32( op(arith_op) | rd(d) | op3(add_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
 
-inline void Assembler::bpr( RCondition c, bool a, Predict p, Register s1, address d, relocInfo::relocType rt ) { v9_only();  cti();  emit_data( op(branch_op) | annul(a) | cond(c) | op2(bpr_op2) | wdisp16(intptr_t(d), intptr_t(pc())) | predict(p) | rs1(s1), rt);  has_delay_slot(); }
+inline void Assembler::bpr( RCondition c, bool a, Predict p, Register s1, address d, relocInfo::relocType rt ) { v9_only(); insert_nop_after_cbcond(); cti();  emit_data( op(branch_op) | annul(a) | cond(c) | op2(bpr_op2) | wdisp16(intptr_t(d), intptr_t(pc())) | predict(p) | rs1(s1), rt);  has_delay_slot(); }
-inline void Assembler::bpr( RCondition c, bool a, Predict p, Register s1, Label& L) { bpr( c, a, p, s1, target(L)); }
+inline void Assembler::bpr( RCondition c, bool a, Predict p, Register s1, Label& L) { insert_nop_after_cbcond(); bpr( c, a, p, s1, target(L)); }
 
-inline void Assembler::fb( Condition c, bool a, address d, relocInfo::relocType rt ) { v9_dep();  cti();  emit_data( op(branch_op) | annul(a) | cond(c) | op2(fb_op2) | wdisp(intptr_t(d), intptr_t(pc()), 22), rt);  has_delay_slot(); }
+inline void Assembler::fb( Condition c, bool a, address d, relocInfo::relocType rt ) { v9_dep();  insert_nop_after_cbcond(); cti();  emit_data( op(branch_op) | annul(a) | cond(c) | op2(fb_op2) | wdisp(intptr_t(d), intptr_t(pc()), 22), rt);  has_delay_slot(); }
-inline void Assembler::fb( Condition c, bool a, Label& L ) { fb(c, a, target(L)); }
+inline void Assembler::fb( Condition c, bool a, Label& L ) { insert_nop_after_cbcond(); fb(c, a, target(L)); }
 
-inline void Assembler::fbp( Condition c, bool a, CC cc, Predict p, address d, relocInfo::relocType rt ) { v9_only();  cti();  emit_data( op(branch_op) | annul(a) | cond(c) | op2(fbp_op2) | branchcc(cc) | predict(p) | wdisp(intptr_t(d), intptr_t(pc()), 19), rt);  has_delay_slot(); }
+inline void Assembler::fbp( Condition c, bool a, CC cc, Predict p, address d, relocInfo::relocType rt ) { v9_only(); insert_nop_after_cbcond(); cti();  emit_data( op(branch_op) | annul(a) | cond(c) | op2(fbp_op2) | branchcc(cc) | predict(p) | wdisp(intptr_t(d), intptr_t(pc()), 19), rt);  has_delay_slot(); }
-inline void Assembler::fbp( Condition c, bool a, CC cc, Predict p, Label& L ) { fbp(c, a, cc, p, target(L)); }
+inline void Assembler::fbp( Condition c, bool a, CC cc, Predict p, Label& L ) { insert_nop_after_cbcond(); fbp(c, a, cc, p, target(L)); }
 
-inline void Assembler::br( Condition c, bool a, address d, relocInfo::relocType rt ) { v9_dep();  cti();   emit_data( op(branch_op) | annul(a) | cond(c) | op2(br_op2) | wdisp(intptr_t(d), intptr_t(pc()), 22), rt);  has_delay_slot(); }
+inline void Assembler::br( Condition c, bool a, address d, relocInfo::relocType rt ) { v9_dep(); insert_nop_after_cbcond(); cti();   emit_data( op(branch_op) | annul(a) | cond(c) | op2(br_op2) | wdisp(intptr_t(d), intptr_t(pc()), 22), rt);  has_delay_slot(); }
-inline void Assembler::br( Condition c, bool a, Label& L ) { br(c, a, target(L)); }
+inline void Assembler::br( Condition c, bool a, Label& L ) { insert_nop_after_cbcond(); br(c, a, target(L)); }
 
-inline void Assembler::bp( Condition c, bool a, CC cc, Predict p, address d, relocInfo::relocType rt ) { v9_only();  cti();  emit_data( op(branch_op) | annul(a) | cond(c) | op2(bp_op2) | branchcc(cc) | predict(p) | wdisp(intptr_t(d), intptr_t(pc()), 19), rt);  has_delay_slot(); }
+inline void Assembler::bp( Condition c, bool a, CC cc, Predict p, address d, relocInfo::relocType rt ) { v9_only();  insert_nop_after_cbcond(); cti();  emit_data( op(branch_op) | annul(a) | cond(c) | op2(bp_op2) | branchcc(cc) | predict(p) | wdisp(intptr_t(d), intptr_t(pc()), 19), rt);  has_delay_slot(); }
-inline void Assembler::bp( Condition c, bool a, CC cc, Predict p, Label& L ) { bp(c, a, cc, p, target(L)); }
+inline void Assembler::bp( Condition c, bool a, CC cc, Predict p, Label& L ) { insert_nop_after_cbcond(); bp(c, a, cc, p, target(L)); }
 
 // compare and branch
 inline void Assembler::cbcond(Condition c, CC cc, Register s1, Register s2, Label& L) { cti();  no_cbcond_before();  emit_data(op(branch_op) | cond_cbcond(c) | op2(bpr_op2) | branchcc(cc) | wdisp10(intptr_t(target(L)), intptr_t(pc())) | rs1(s1) | rs2(s2)); }
 inline void Assembler::cbcond(Condition c, CC cc, Register s1, int simm5, Label& L)   { cti();  no_cbcond_before();  emit_data(op(branch_op) | cond_cbcond(c) | op2(bpr_op2) | branchcc(cc) | wdisp10(intptr_t(target(L)), intptr_t(pc())) | rs1(s1) | immed(true) | simm(simm5, 5)); }
 
-inline void Assembler::call( address d,  relocInfo::relocType rt ) { cti();  emit_data( op(call_op) | wdisp(intptr_t(d), intptr_t(pc()), 30), rt);  has_delay_slot(); assert(rt != relocInfo::virtual_call_type, "must use virtual_call_Relocation::spec"); }
+inline void Assembler::call( address d,  relocInfo::relocType rt ) { insert_nop_after_cbcond(); cti();  emit_data( op(call_op) | wdisp(intptr_t(d), intptr_t(pc()), 30), rt);  has_delay_slot(); assert(rt != relocInfo::virtual_call_type, "must use virtual_call_Relocation::spec"); }
-inline void Assembler::call( Label& L,   relocInfo::relocType rt ) { call( target(L), rt); }
+inline void Assembler::call( Label& L,   relocInfo::relocType rt ) { insert_nop_after_cbcond(); call( target(L), rt); }
 
 inline void Assembler::flush( Register s1, Register s2) { emit_int32( op(arith_op) | op3(flush_op3) | rs1(s1) | rs2(s2)); }
 inline void Assembler::flush( Register s1, int simm13a) { emit_data( op(arith_op) | op3(flush_op3) | rs1(s1) | immed(true) | simm(simm13a, 13)); }
 
-inline void Assembler::jmpl( Register s1, Register s2, Register d ) { cti();  emit_int32( op(arith_op) | rd(d) | op3(jmpl_op3) | rs1(s1) | rs2(s2));  has_delay_slot(); }
+inline void Assembler::jmpl( Register s1, Register s2, Register d ) { insert_nop_after_cbcond(); cti();  emit_int32( op(arith_op) | rd(d) | op3(jmpl_op3) | rs1(s1) | rs2(s2));  has_delay_slot(); }
-inline void Assembler::jmpl( Register s1, int simm13a, Register d, RelocationHolder const& rspec ) { cti();  emit_data( op(arith_op) | rd(d) | op3(jmpl_op3) | rs1(s1) | immed(true) | simm(simm13a, 13), rspec);  has_delay_slot(); }
+inline void Assembler::jmpl( Register s1, int simm13a, Register d, RelocationHolder const& rspec ) { insert_nop_after_cbcond(); cti();  emit_data( op(arith_op) | rd(d) | op3(jmpl_op3) | rs1(s1) | immed(true) | simm(simm13a, 13), rspec);  has_delay_slot(); }
 
 inline void Assembler::ldf(FloatRegisterImpl::Width w, Register s1, Register s2, FloatRegister d) { emit_int32( op(ldst_op) | fd(d, w) | alt_op3(ldf_op3, w) | rs1(s1) | rs2(s2) ); }
 inline void Assembler::ldf(FloatRegisterImpl::Width w, Register s1, int simm13a, FloatRegister d, RelocationHolder const& rspec) { emit_data( op(ldst_op) | fd(d, w) | alt_op3(ldf_op3, w) | rs1(s1) | immed(true) | simm(simm13a, 13), rspec); }

hotspot/src/cpu/sparc/vm/macroAssembler_sparc.inline.hpp

@@ -233,6 +233,7 @@ inline void MacroAssembler::br( Condition c, bool a, Predict p, address d, reloc
 }
 
 inline void MacroAssembler::br( Condition c, bool a, Predict p, Label& L ) {
+  insert_nop_after_cbcond();
   br(c, a, p, target(L));
 }
 
@@ -248,6 +249,7 @@ inline void MacroAssembler::brx( Condition c, bool a, Predict p, address d, relo
 }
 
 inline void MacroAssembler::brx( Condition c, bool a, Predict p, Label& L ) {
+  insert_nop_after_cbcond();
   brx(c, a, p, target(L));
 }
 
@@ -269,6 +271,7 @@ inline void MacroAssembler::fb( Condition c, bool a, Predict p, address d, reloc
 }
 
 inline void MacroAssembler::fb( Condition c, bool a, Predict p, Label& L ) {
+  insert_nop_after_cbcond();
   fb(c, a, p, target(L));
 }
 
@@ -318,6 +321,7 @@ inline void MacroAssembler::call( address d, relocInfo::relocType rt ) {
 }
 
 inline void MacroAssembler::call( Label& L,   relocInfo::relocType rt ) {
+  insert_nop_after_cbcond();
   MacroAssembler::call( target(L), rt);
 }
 

hotspot/src/cpu/sparc/vm/sparc.ad

@@ -1277,8 +1277,12 @@ void MachEpilogNode::format( PhaseRegAlloc *ra_, outputStream *st ) const {
 #endif
   }
 
-  if( do_polling() )
+  if(do_polling()) {
+    if (UseCBCond && !ra_->C->is_method_compilation()) {
+      st->print("NOP\n\t");
+    }
     st->print("RET\n\t");
+  }
 
   st->print("RESTORE");
 }
@@ -1300,6 +1304,11 @@ void MachEpilogNode::emit(CodeBuffer &cbuf, PhaseRegAlloc *ra_) const {
 
   // If this is a return, then stuff the restore in the delay slot
   if(do_polling()) {
+    if (UseCBCond && !ra_->C->is_method_compilation()) {
+      // Insert extra padding for the case when the epilogue is preceded by
+      // a cbcond jump, which can't be followed by a CTI instruction
+      __ nop();
+    }
     __ ret();
     __ delayed()->restore();
   } else {
@@ -3330,7 +3339,18 @@ op_attrib op_cost(1);          // Required cost attribute
 //----------Instruction Attributes---------------------------------------------
 ins_attrib ins_cost(DEFAULT_COST); // Required cost attribute
 ins_attrib ins_size(32);           // Required size attribute (in bits)
-ins_attrib ins_avoid_back_to_back(0); // instruction should not be generated back to back
+
+// avoid_back_to_back attribute is an expression that must return
+// one of the following values defined in MachNode:
+// AVOID_NONE   - instruction can be placed anywhere
+// AVOID_BEFORE - instruction cannot be placed after an
+//                instruction with MachNode::AVOID_AFTER
+// AVOID_AFTER  - the next instruction cannot be the one 
+//                with MachNode::AVOID_BEFORE
+// AVOID_BEFORE_AND_AFTER - BEFORE and AFTER attributes at 
+//                          the same time                                
+ins_attrib ins_avoid_back_to_back(MachNode::AVOID_NONE);
+
 ins_attrib ins_short_branch(0);    // Required flag: is this instruction a
                                    // non-matching short branch variant of some
                                                             // long branch?
@@ -6630,6 +6650,7 @@ instruct encodeHeapOop(iRegN dst, iRegP src) %{
   ins_encode %{
     __ encode_heap_oop($src$$Register, $dst$$Register);
   %}
+  ins_avoid_back_to_back(Universe::narrow_oop_base() == NULL ? AVOID_NONE : AVOID_BEFORE);
   ins_pipe(ialu_reg);
 %}
 
@@ -9199,6 +9220,7 @@ instruct branch(label labl) %{
     __ ba(*L);
     __ delayed()->nop();
   %}
+  ins_avoid_back_to_back(AVOID_BEFORE);
   ins_pipe(br);
 %}
 
@@ -9217,7 +9239,7 @@ instruct branch_short(label labl) %{
     __ ba_short(*L);
   %}
   ins_short_branch(1);
-  ins_avoid_back_to_back(1);
+  ins_avoid_back_to_back(AVOID_BEFORE_AND_AFTER);
   ins_pipe(cbcond_reg_imm);
 %}
 
@@ -9231,6 +9253,7 @@ instruct branchCon(cmpOp cmp, flagsReg icc, label labl) %{
   format %{ "BP$cmp   $icc,$labl" %}
   // Prim = bits 24-22, Secnd = bits 31-30
   ins_encode( enc_bp( labl, cmp, icc ) );
+  ins_avoid_back_to_back(AVOID_BEFORE);
   ins_pipe(br_cc);
 %}
 
@@ -9242,6 +9265,7 @@ instruct branchConU(cmpOpU cmp, flagsRegU icc, label labl) %{
   format %{ "BP$cmp  $icc,$labl" %}
   // Prim = bits 24-22, Secnd = bits 31-30
   ins_encode( enc_bp( labl, cmp, icc ) );
+  ins_avoid_back_to_back(AVOID_BEFORE);
   ins_pipe(br_cc);
 %}
 
@@ -9260,6 +9284,7 @@ instruct branchConP(cmpOpP cmp, flagsRegP pcc, label labl) %{
     __ bp( (Assembler::Condition)($cmp$$cmpcode), false, Assembler::ptr_cc, predict_taken, *L);
     __ delayed()->nop();
   %}
+  ins_avoid_back_to_back(AVOID_BEFORE);
   ins_pipe(br_cc);
 %}
 
@@ -9278,6 +9303,7 @@ instruct branchConF(cmpOpF cmp, flagsRegF fcc, label labl) %{
     __ fbp( (Assembler::Condition)($cmp$$cmpcode), false, (Assembler::CC)($fcc$$reg), predict_taken, *L);
     __ delayed()->nop();
   %}
+  ins_avoid_back_to_back(AVOID_BEFORE);
   ins_pipe(br_fcc);
 %}
 
@@ -9290,6 +9316,7 @@ instruct branchLoopEnd(cmpOp cmp, flagsReg icc, label labl) %{
   format %{ "BP$cmp   $icc,$labl\t! Loop end" %}
   // Prim = bits 24-22, Secnd = bits 31-30
   ins_encode( enc_bp( labl, cmp, icc ) );
+  ins_avoid_back_to_back(AVOID_BEFORE);
   ins_pipe(br_cc);
 %}
 
@@ -9302,6 +9329,7 @@ instruct branchLoopEndU(cmpOpU cmp, flagsRegU icc, label labl) %{
   format %{ "BP$cmp  $icc,$labl\t! Loop end" %}
   // Prim = bits 24-22, Secnd = bits 31-30
   ins_encode( enc_bp( labl, cmp, icc ) );
+  ins_avoid_back_to_back(AVOID_BEFORE);
   ins_pipe(br_cc);
 %}
 
@@ -9552,7 +9580,7 @@ instruct cmpI_reg_branch_short(cmpOp cmp, iRegI op1, iRegI op2, label labl, flag
     __ cbcond((Assembler::Condition)($cmp$$cmpcode), Assembler::icc, $op1$$Register, $op2$$Register, *L);
   %}
   ins_short_branch(1);
-  ins_avoid_back_to_back(1);
+  ins_avoid_back_to_back(AVOID_BEFORE_AND_AFTER);
   ins_pipe(cbcond_reg_reg);
 %}
 
@@ -9570,7 +9598,7 @@ instruct cmpI_imm_branch_short(cmpOp cmp, iRegI op1, immI5 op2, label labl, flag
     __ cbcond((Assembler::Condition)($cmp$$cmpcode), Assembler::icc, $op1$$Register, $op2$$constant, *L);
   %}
   ins_short_branch(1);
-  ins_avoid_back_to_back(1);
+  ins_avoid_back_to_back(AVOID_BEFORE_AND_AFTER);
   ins_pipe(cbcond_reg_imm);
 %}
 
@@ -9588,7 +9616,7 @@ instruct cmpU_reg_branch_short(cmpOpU cmp, iRegI op1, iRegI op2, label labl, fla
     __ cbcond((Assembler::Condition)($cmp$$cmpcode), Assembler::icc, $op1$$Register, $op2$$Register, *L);
   %}
   ins_short_branch(1);
-  ins_avoid_back_to_back(1);
+  ins_avoid_back_to_back(AVOID_BEFORE_AND_AFTER);
   ins_pipe(cbcond_reg_reg);
 %}
 
@@ -9606,7 +9634,7 @@ instruct cmpU_imm_branch_short(cmpOpU cmp, iRegI op1, immI5 op2, label labl, fla
     __ cbcond((Assembler::Condition)($cmp$$cmpcode), Assembler::icc, $op1$$Register, $op2$$constant, *L);
   %}
   ins_short_branch(1);
-  ins_avoid_back_to_back(1);
+  ins_avoid_back_to_back(AVOID_BEFORE_AND_AFTER);
   ins_pipe(cbcond_reg_imm);
 %}
 
@@ -9624,7 +9652,7 @@ instruct cmpL_reg_branch_short(cmpOp cmp, iRegL op1, iRegL op2, label labl, flag
     __ cbcond((Assembler::Condition)($cmp$$cmpcode), Assembler::xcc, $op1$$Register, $op2$$Register, *L);
   %}
   ins_short_branch(1);
-  ins_avoid_back_to_back(1);
+  ins_avoid_back_to_back(AVOID_BEFORE_AND_AFTER);
   ins_pipe(cbcond_reg_reg);
 %}
 
@@ -9642,7 +9670,7 @@ instruct cmpL_imm_branch_short(cmpOp cmp, iRegL op1, immL5 op2, label labl, flag
     __ cbcond((Assembler::Condition)($cmp$$cmpcode), Assembler::xcc, $op1$$Register, $op2$$constant, *L);
   %}
   ins_short_branch(1);
-  ins_avoid_back_to_back(1);
+  ins_avoid_back_to_back(AVOID_BEFORE_AND_AFTER);
   ins_pipe(cbcond_reg_imm);
 %}
 
@@ -9665,7 +9693,7 @@ instruct cmpP_reg_branch_short(cmpOpP cmp, iRegP op1, iRegP op2, label labl, fla
     __ cbcond((Assembler::Condition)($cmp$$cmpcode), Assembler::ptr_cc, $op1$$Register, $op2$$Register, *L);
   %}
   ins_short_branch(1);
-  ins_avoid_back_to_back(1);
+  ins_avoid_back_to_back(AVOID_BEFORE_AND_AFTER);
   ins_pipe(cbcond_reg_reg);
 %}
 
@@ -9687,7 +9715,7 @@ instruct cmpP_null_branch_short(cmpOpP cmp, iRegP op1, immP0 null, label labl, f
     __ cbcond((Assembler::Condition)($cmp$$cmpcode), Assembler::ptr_cc, $op1$$Register, G0, *L);
   %}
   ins_short_branch(1);
-  ins_avoid_back_to_back(1);
+  ins_avoid_back_to_back(AVOID_BEFORE_AND_AFTER);
   ins_pipe(cbcond_reg_reg);
 %}
 
@@ -9705,7 +9733,7 @@ instruct cmpN_reg_branch_short(cmpOp cmp, iRegN op1, iRegN op2, label labl, flag
     __ cbcond((Assembler::Condition)($cmp$$cmpcode), Assembler::icc, $op1$$Register, $op2$$Register, *L);
   %}
   ins_short_branch(1);
-  ins_avoid_back_to_back(1);
+  ins_avoid_back_to_back(AVOID_BEFORE_AND_AFTER);
   ins_pipe(cbcond_reg_reg);
 %}
 
@@ -9723,7 +9751,7 @@ instruct cmpN_null_branch_short(cmpOp cmp, iRegN op1, immN0 null, label labl, fl
     __ cbcond((Assembler::Condition)($cmp$$cmpcode), Assembler::icc, $op1$$Register, G0, *L);
   %}
   ins_short_branch(1);
-  ins_avoid_back_to_back(1);
+  ins_avoid_back_to_back(AVOID_BEFORE_AND_AFTER);
   ins_pipe(cbcond_reg_reg);
 %}
 
@@ -9742,7 +9770,7 @@ instruct cmpI_reg_branchLoopEnd_short(cmpOp cmp, iRegI op1, iRegI op2, label lab
     __ cbcond((Assembler::Condition)($cmp$$cmpcode), Assembler::icc, $op1$$Register, $op2$$Register, *L);
   %}
   ins_short_branch(1);
-  ins_avoid_back_to_back(1);
+  ins_avoid_back_to_back(AVOID_BEFORE_AND_AFTER);
   ins_pipe(cbcond_reg_reg);
 %}
 
@@ -9760,7 +9788,7 @@ instruct cmpI_imm_branchLoopEnd_short(cmpOp cmp, iRegI op1, immI5 op2, label lab
     __ cbcond((Assembler::Condition)($cmp$$cmpcode), Assembler::icc, $op1$$Register, $op2$$constant, *L);
   %}
   ins_short_branch(1);
-  ins_avoid_back_to_back(1);
+  ins_avoid_back_to_back(AVOID_BEFORE_AND_AFTER);
   ins_pipe(cbcond_reg_imm);
 %}
 
@@ -9777,6 +9805,7 @@ instruct branchCon_regI(cmpOp_reg cmp, iRegI op1, immI0 zero, label labl) %{
   ins_cost(BRANCH_COST);
   format %{ "BR$cmp   $op1,$labl" %}
   ins_encode( enc_bpr( labl, cmp, op1 ) );
+  ins_avoid_back_to_back(AVOID_BEFORE);
   ins_pipe(br_reg);
 %}
 
@@ -9789,6 +9818,7 @@ instruct branchCon_regP(cmpOp_reg cmp, iRegP op1, immP0 null, label labl) %{
   ins_cost(BRANCH_COST);
   format %{ "BR$cmp   $op1,$labl" %}
   ins_encode( enc_bpr( labl, cmp, op1 ) );
+  ins_avoid_back_to_back(AVOID_BEFORE);
   ins_pipe(br_reg);
 %}
 
@@ -9801,6 +9831,7 @@ instruct branchCon_regL(cmpOp_reg cmp, iRegL op1, immL0 zero, label labl) %{
   ins_cost(BRANCH_COST);
   format %{ "BR$cmp   $op1,$labl" %}
   ins_encode( enc_bpr( labl, cmp, op1 ) );
+  ins_avoid_back_to_back(AVOID_BEFORE);
   ins_pipe(br_reg);
 %}
 
@@ -9841,6 +9872,7 @@ instruct branchCon_long(cmpOp cmp, flagsRegL xcc, label labl) %{
     __ bp( (Assembler::Condition)($cmp$$cmpcode), false, Assembler::xcc, predict_taken, *L);
     __ delayed()->nop();
   %}
+  ins_avoid_back_to_back(AVOID_BEFORE);
   ins_pipe(br_cc);
 %}
 
@@ -9968,6 +10000,7 @@ instruct CallStaticJavaDirect( method meth ) %{
   ins_cost(CALL_COST);
   format %{ "CALL,static  ; NOP ==> " %}
   ins_encode( Java_Static_Call( meth ), call_epilog );
+  ins_avoid_back_to_back(AVOID_BEFORE);
   ins_pipe(simple_call);
 %}
 
@@ -10004,6 +10037,7 @@ instruct CallRuntimeDirect(method meth, l7RegP l7) %{
   format %{ "CALL,runtime" %}
   ins_encode( Java_To_Runtime( meth ),
               call_epilog, adjust_long_from_native_call );
+  ins_avoid_back_to_back(AVOID_BEFORE);
   ins_pipe(simple_call);
 %}
 
@@ -10016,6 +10050,7 @@ instruct CallLeafDirect(method meth, l7RegP l7) %{
   ins_encode( Java_To_Runtime( meth ),
               call_epilog,
               adjust_long_from_native_call );
+  ins_avoid_back_to_back(AVOID_BEFORE);
   ins_pipe(simple_call);
 %}
 
@@ -10028,6 +10063,7 @@ instruct CallLeafNoFPDirect(method meth, l7RegP l7) %{
   ins_encode( Java_To_Runtime( meth ),
               call_epilog,
               adjust_long_from_native_call );
+  ins_avoid_back_to_back(AVOID_BEFORE);
   ins_pipe(simple_call);
 %}
 
@@ -10041,6 +10077,7 @@ instruct TailCalljmpInd(g3RegP jump_target, inline_cache_regP method_oop) %{
   ins_cost(CALL_COST);
   format %{ "Jmp     $jump_target  ; NOP \t! $method_oop holds method oop" %}
   ins_encode(form_jmpl(jump_target));
+  ins_avoid_back_to_back(AVOID_BEFORE);
   ins_pipe(tail_call);
 %}
 
@@ -10072,6 +10109,7 @@ instruct tailjmpInd(g1RegP jump_target, i0RegP ex_oop) %{
   // opcode(Assembler::jmpl_op3, Assembler::arith_op);
   // The hack duplicates the exception oop into G3, so that CreateEx can use it there.
   // ins_encode( form3_rs1_simm13_rd( jump_target, 0x00, R_G0 ), move_return_pc_to_o1() );
+  ins_avoid_back_to_back(AVOID_BEFORE);
   ins_pipe(tail_call);
 %}
 
@@ -10102,6 +10140,7 @@ instruct RethrowException()
   // use the following format syntax
   format %{ "Jmp    rethrow_stub" %}
   ins_encode(enc_rethrow);
+  ins_avoid_back_to_back(AVOID_BEFORE);
   ins_pipe(tail_call);
 %}
 
@@ -10130,6 +10169,7 @@ instruct partialSubtypeCheck( o0RegP index, o1RegP sub, o2RegP super, flagsRegP
   ins_cost(DEFAULT_COST*10);
   format %{ "CALL   PartialSubtypeCheck\n\tNOP" %}
   ins_encode( enc_PartialSubtypeCheck() );
+  ins_avoid_back_to_back(AVOID_BEFORE);
   ins_pipe(partial_subtype_check_pipe);
 %}
 
@@ -10139,6 +10179,7 @@ instruct partialSubtypeCheck_vs_zero( flagsRegP pcc, o1RegP sub, o2RegP super, i
   ins_cost(DEFAULT_COST*10);
   format %{ "CALL   PartialSubtypeCheck\n\tNOP\t# (sets condition codes)" %}
   ins_encode( enc_PartialSubtypeCheck() );
+  ins_avoid_back_to_back(AVOID_BEFORE);
   ins_pipe(partial_subtype_check_pipe);
 %}
 

hotspot/src/cpu/x86/vm/globals_x86.hpp

@@ -162,7 +162,7 @@ define_pd_global(uintx, TypeProfileLevel, 111);
           "Number of milliseconds to wait before start calculating aborts " \
           "for RTM locking")                                                \
                                                                             \
-  experimental(bool, UseRTMXendForLockBusy, false,                          \
+  experimental(bool, UseRTMXendForLockBusy, true,                           \
           "Use RTM Xend instead of Xabort when lock busy")                  \
                                                                             \
   /* assembler */                                                           \

hotspot/src/cpu/x86/vm/macroAssembler_x86.cpp

@@ -1488,11 +1488,10 @@ void MacroAssembler::rtm_stack_locking(Register objReg, Register tmpReg, Registe
     movl(retry_on_abort_count_Reg, RTMRetryCount); // Retry on abort
     bind(L_rtm_retry);
   }
-  if (!UseRTMXendForLockBusy) {
   movptr(tmpReg, Address(objReg, 0));
   testptr(tmpReg, markOopDesc::monitor_value);  // inflated vs stack-locked|neutral|biased
   jcc(Assembler::notZero, IsInflated);
-  }
+
   if (PrintPreciseRTMLockingStatistics || profile_rtm) {
     Label L_noincrement;
     if (RTMTotalCountIncrRate > 1) {
@@ -1512,10 +1511,7 @@ void MacroAssembler::rtm_stack_locking(Register objReg, Register tmpReg, Registe
   Register abort_status_Reg = tmpReg; // status of abort is stored in RAX
   if (UseRTMXendForLockBusy) {
     xend();
-    movptr(tmpReg, Address(objReg, 0));
+    movptr(abort_status_Reg, 0x2);   // Set the abort status to 2 (so we can retry)
-    testptr(tmpReg, markOopDesc::monitor_value);  // inflated vs stack-locked|neutral|biased
-    jcc(Assembler::notZero, IsInflated);
-    movptr(abort_status_Reg, 0x1);                // Set the abort status to 1 (as xabort does)
     jmp(L_decrement_retry);
   }
   else {

hotspot/src/os/aix/vm/os_aix.cpp

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1999, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1999, 2014, Oracle and/or its affiliates. All rights reserved.
  * Copyright 2012, 2014 SAP AG. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
@@ -2811,18 +2811,13 @@ void os::yield() {
 
 os::YieldResult os::NakedYield() { sched_yield(); return os::YIELD_UNKNOWN; }
 
-void os::yield_all(int attempts) {
+void os::yield_all() {
   // Yields to all threads, including threads with lower priorities
   // Threads on Linux are all with same priority. The Solaris style
   // os::yield_all() with nanosleep(1ms) is not necessary.
   sched_yield();
 }
 
-// Called from the tight loops to possibly influence time-sharing heuristics
-void os::loop_breaker(int attempts) {
-  os::yield_all(attempts);
-}
-
 ////////////////////////////////////////////////////////////////////////////////
 // thread priority support
 
@@ -3079,7 +3074,7 @@ static bool do_suspend(OSThread* osthread) {
 
   for (int n = 0; !osthread->sr.is_suspended(); n++) {
     for (int i = 0; i < RANDOMLY_LARGE_INTEGER2 && !osthread->sr.is_suspended(); i++) {
-      os::yield_all(i);
+      os::yield_all();
     }
 
     // timeout, try to cancel the request
@@ -3113,7 +3108,7 @@ static void do_resume(OSThread* osthread) {
     if (sr_notify(osthread) == 0) {
       for (int n = 0; n < RANDOMLY_LARGE_INTEGER && !osthread->sr.is_running(); n++) {
         for (int i = 0; i < 100 && !osthread->sr.is_running(); i++) {
-          os::yield_all(i);
+          os::yield_all();
         }
       }
     } else {

hotspot/src/os/bsd/vm/os_bsd.cpp

@@ -917,9 +917,20 @@ void os::free_thread(OSThread* osthread) {
 //////////////////////////////////////////////////////////////////////////////
 // thread local storage
 
+// Restore the thread pointer if the destructor is called. This is in case
+// someone from JNI code sets up a destructor with pthread_key_create to run
+// detachCurrentThread on thread death. Unless we restore the thread pointer we
+// will hang or crash. When detachCurrentThread is called the key will be set
+// to null and we will not be called again. If detachCurrentThread is never
+// called we could loop forever depending on the pthread implementation.
+static void restore_thread_pointer(void* p) {
+  Thread* thread = (Thread*) p;
+  os::thread_local_storage_at_put(ThreadLocalStorage::thread_index(), thread);
+}
+
 int os::allocate_thread_local_storage() {
   pthread_key_t key;
-  int rslt = pthread_key_create(&key, NULL);
+  int rslt = pthread_key_create(&key, restore_thread_pointer);
   assert(rslt == 0, "cannot allocate thread local storage");
   return (int)key;
 }
@@ -2551,18 +2562,13 @@ void os::yield() {
 
 os::YieldResult os::NakedYield() { sched_yield(); return os::YIELD_UNKNOWN ;}
 
-void os::yield_all(int attempts) {
+void os::yield_all() {
   // Yields to all threads, including threads with lower priorities
   // Threads on Bsd are all with same priority. The Solaris style
   // os::yield_all() with nanosleep(1ms) is not necessary.
   sched_yield();
 }
 
-// Called from the tight loops to possibly influence time-sharing heuristics
-void os::loop_breaker(int attempts) {
-  os::yield_all(attempts);
-}
-
 ////////////////////////////////////////////////////////////////////////////////
 // thread priority support
 

hotspot/src/os/linux/vm/os_linux.cpp

@@ -1032,9 +1032,20 @@ void os::free_thread(OSThread* osthread) {
 //////////////////////////////////////////////////////////////////////////////
 // thread local storage
 
+// Restore the thread pointer if the destructor is called. This is in case
+// someone from JNI code sets up a destructor with pthread_key_create to run
+// detachCurrentThread on thread death. Unless we restore the thread pointer we
+// will hang or crash. When detachCurrentThread is called the key will be set
+// to null and we will not be called again. If detachCurrentThread is never
+// called we could loop forever depending on the pthread implementation.
+static void restore_thread_pointer(void* p) {
+  Thread* thread = (Thread*) p;
+  os::thread_local_storage_at_put(ThreadLocalStorage::thread_index(), thread);
+}
+
 int os::allocate_thread_local_storage() {
   pthread_key_t key;
-  int rslt = pthread_key_create(&key, NULL);
+  int rslt = pthread_key_create(&key, restore_thread_pointer);
   assert(rslt == 0, "cannot allocate thread local storage");
   return (int)key;
 }
@@ -3781,18 +3792,13 @@ void os::yield() {
 
 os::YieldResult os::NakedYield() { sched_yield(); return os::YIELD_UNKNOWN ;}
 
-void os::yield_all(int attempts) {
+void os::yield_all() {
   // Yields to all threads, including threads with lower priorities
   // Threads on Linux are all with same priority. The Solaris style
   // os::yield_all() with nanosleep(1ms) is not necessary.
   sched_yield();
 }
 
-// Called from the tight loops to possibly influence time-sharing heuristics
-void os::loop_breaker(int attempts) {
-  os::yield_all(attempts);
-}
-
 ////////////////////////////////////////////////////////////////////////////////
 // thread priority support
 

hotspot/src/os/solaris/vm/attachListener_solaris.cpp

@@ -29,6 +29,7 @@
 #include "services/dtraceAttacher.hpp"
 
 #include <door.h>
+#include <limits.h>
 #include <string.h>
 #include <signal.h>
 #include <sys/types.h>
@@ -668,11 +669,13 @@ static jint enable_dprobes(AttachOperation* op, outputStream* out) {
     out->print_cr("No probe specified");
     return JNI_ERR;
   } else {
-    int probe_typess = atoi(probe);
+    char *end;
-    if (errno) {
+    long val = strtol(probe, &end, 10);
+    if (end == probe || val < 0 || val > INT_MAX) {
       out->print_cr("invalid probe type");
       return JNI_ERR;
     } else {
+      int probe_typess = (int) val;
       DTrace::enable_dprobes(probe_typess);
       return JNI_OK;
     }
@@ -703,8 +706,9 @@ jint AttachListener::pd_set_flag(AttachOperation* op, outputStream* out) {
   bool flag = true;
   const char* arg1;
   if ((arg1 = op->arg(1)) != NULL) {
-    flag = (atoi(arg1) != 0);
+    char *end;
-    if (errno) {
+    flag = (strtol(arg1, &end, 10) != 0);
+    if (arg1 == end) {
       out->print_cr("flag value has to be an integer");
       return JNI_ERR;
     }

hotspot/src/os/solaris/vm/osThread_solaris.cpp

@@ -49,16 +49,6 @@ void OSThread::pd_destroy() {
 
 // copied from synchronizer.cpp
 
-void OSThread::handle_spinlock_contention(int tries) {
-  if (NoYieldsInMicrolock) return;
-
-  if (tries > 10) {
-    os::yield_all(tries); // Yield to threads of any priority
-  } else if (tries > 5) {
-    os::yield();          // Yield to threads of same or higher priority
-  }
-}
-
 void OSThread::SR_handler(Thread* thread, ucontext_t* uc) {
   os::Solaris::SR_handler(thread, uc);
 }

hotspot/src/os/solaris/vm/osThread_solaris.hpp

@@ -82,8 +82,6 @@
   void set_ucontext(ucontext_t* ptr) { _ucontext = ptr; }
   static void SR_handler(Thread* thread, ucontext_t* uc);
 
-  static void       handle_spinlock_contention(int tries);                      // Used for thread local eden locking
-
   // ***************************************************************
   // Platform dependent initialization and cleanup
   // ***************************************************************

hotspot/src/os/solaris/vm/os_solaris.cpp

@@ -969,9 +969,6 @@ bool os::create_main_thread(JavaThread* thread) {
   return true;
 }
 
-// _T2_libthread is true if we believe we are running with the newer
-// SunSoft lwp/libthread.so (2.8 patch, 2.9 default)
-bool os::Solaris::_T2_libthread = false;
 
 bool os::create_thread(Thread* thread, ThreadType thr_type, size_t stack_size) {
   // Allocate the OSThread object
@@ -1056,71 +1053,10 @@ bool os::create_thread(Thread* thread, ThreadType thr_type, size_t stack_size) {
   thread->set_osthread(osthread);
 
   // Create the Solaris thread
-  // explicit THR_BOUND for T2_libthread case in case
-  // that assumption is not accurate, but our alternate signal stack
-  // handling is based on it which must have bound threads
   thread_t tid = 0;
-  long     flags = (UseDetachedThreads ? THR_DETACHED : 0) | THR_SUSPENDED
+  long     flags = (UseDetachedThreads ? THR_DETACHED : 0) | THR_SUSPENDED;
-                   | ((UseBoundThreads || os::Solaris::T2_libthread() ||
-                       (thr_type == vm_thread) ||
-                       (thr_type == cgc_thread) ||
-                       (thr_type == pgc_thread) ||
-                       (thr_type == compiler_thread && BackgroundCompilation)) ?
-                      THR_BOUND : 0);
   int      status;
 
-  // 4376845 -- libthread/kernel don't provide enough LWPs to utilize all CPUs.
-  //
-  // On multiprocessors systems, libthread sometimes under-provisions our
-  // process with LWPs.  On a 30-way systems, for instance, we could have
-  // 50 user-level threads in ready state and only 2 or 3 LWPs assigned
-  // to our process.  This can result in under utilization of PEs.
-  // I suspect the problem is related to libthread's LWP
-  // pool management and to the kernel's SIGBLOCKING "last LWP parked"
-  // upcall policy.
-  //
-  // The following code is palliative -- it attempts to ensure that our
-  // process has sufficient LWPs to take advantage of multiple PEs.
-  // Proper long-term cures include using user-level threads bound to LWPs
-  // (THR_BOUND) or using LWP-based synchronization.  Note that there is a
-  // slight timing window with respect to sampling _os_thread_count, but
-  // the race is benign.  Also, we should periodically recompute
-  // _processors_online as the min of SC_NPROCESSORS_ONLN and the
-  // the number of PEs in our partition.  You might be tempted to use
-  // THR_NEW_LWP here, but I'd recommend against it as that could
-  // result in undesirable growth of the libthread's LWP pool.
-  // The fix below isn't sufficient; for instance, it doesn't take into count
-  // LWPs parked on IO.  It does, however, help certain CPU-bound benchmarks.
-  //
-  // Some pathologies this scheme doesn't handle:
-  // *  Threads can block, releasing the LWPs.  The LWPs can age out.
-  //    When a large number of threads become ready again there aren't
-  //    enough LWPs available to service them.  This can occur when the
-  //    number of ready threads oscillates.
-  // *  LWPs/Threads park on IO, thus taking the LWP out of circulation.
-  //
-  // Finally, we should call thr_setconcurrency() periodically to refresh
-  // the LWP pool and thwart the LWP age-out mechanism.
-  // The "+3" term provides a little slop -- we want to slightly overprovision.
-
-  if (AdjustConcurrency && os::Solaris::_os_thread_count < (_processors_online+3)) {
-    if (!(flags & THR_BOUND)) {
-      thr_setconcurrency (os::Solaris::_os_thread_count);       // avoid starvation
-    }
-  }
-  // Although this doesn't hurt, we should warn of undefined behavior
-  // when using unbound T1 threads with schedctl().  This should never
-  // happen, as the compiler and VM threads are always created bound
-  DEBUG_ONLY(
-      if ((VMThreadHintNoPreempt || CompilerThreadHintNoPreempt) &&
-          (!os::Solaris::T2_libthread() && (!(flags & THR_BOUND))) &&
-          ((thr_type == vm_thread) || (thr_type == cgc_thread) ||
-           (thr_type == pgc_thread) || (thr_type == compiler_thread && BackgroundCompilation))) {
-         warning("schedctl behavior undefined when Compiler/VM/GC Threads are Unbound");
-      }
-  );
-
-
   // Mark that we don't have an lwp or thread id yet.
   // In case we attempt to set the priority before the thread starts.
   osthread->set_lwp_id(-1);
@@ -1145,13 +1081,6 @@ bool os::create_thread(Thread* thread, ThreadType thr_type, size_t stack_size) {
   // Remember that we created this thread so we can set priority on it
   osthread->set_vm_created();
 
-  // Set the default thread priority.  If using bound threads, setting
-  // lwp priority will be delayed until thread start.
-  set_native_priority(thread,
-                      DefaultThreadPriority == -1 ?
-                        java_to_os_priority[NormPriority] :
-                        DefaultThreadPriority);
-
   // Initial thread state is INITIALIZED, not SUSPENDED
   osthread->set_state(INITIALIZED);
 
@@ -1333,39 +1262,8 @@ void os::initialize_thread(Thread* thr) {
     jt->set_stack_size(stack_size);
   }
 
-   // 5/22/01: Right now alternate signal stacks do not handle
+  // With the T2 libthread (T1 is no longer supported) threads are always bound
-   // throwing stack overflow exceptions, see bug 4463178
+  // and we use stackbanging in all cases.
-   // Until a fix is found for this, T2 will NOT imply alternate signal
-   // stacks.
-   // If using T2 libthread threads, install an alternate signal stack.
-   // Because alternate stacks associate with LWPs on Solaris,
-   // see sigaltstack(2), if using UNBOUND threads, or if UseBoundThreads
-   // we prefer to explicitly stack bang.
-   // If not using T2 libthread, but using UseBoundThreads any threads
-   // (primordial thread, jni_attachCurrentThread) we do not create,
-   // probably are not bound, therefore they can not have an alternate
-   // signal stack. Since our stack banging code is generated and
-   // is shared across threads, all threads must be bound to allow
-   // using alternate signal stacks.  The alternative is to interpose
-   // on _lwp_create to associate an alt sig stack with each LWP,
-   // and this could be a problem when the JVM is embedded.
-   // We would prefer to use alternate signal stacks with T2
-   // Since there is currently no accurate way to detect T2
-   // we do not. Assuming T2 when running T1 causes sig 11s or assertions
-   // on installing alternate signal stacks
-
-
-   // 05/09/03: removed alternate signal stack support for Solaris
-   // The alternate signal stack mechanism is no longer needed to
-   // handle stack overflow. This is now handled by allocating
-   // guard pages (red zone) and stackbanging.
-   // Initially the alternate signal stack mechanism was removed because
-   // it did not work with T1 llibthread. Alternate
-   // signal stacks MUST have all threads bound to lwps. Applications
-   // can create their own threads and attach them without their being
-   // bound under T1. This is frequently the case for the primordial thread.
-   // If we were ever to reenable this mechanism we would need to
-   // use the dynamic check for T2 libthread.
 
   os::Solaris::init_thread_fpu_state();
   std::set_terminate(_handle_uncaught_cxx_exception);
@@ -2092,12 +1990,7 @@ void os::Solaris::print_distro_info(outputStream* st) {
 }
 
 void os::Solaris::print_libversion_info(outputStream* st) {
-  if (os::Solaris::T2_libthread()) {
   st->print("  (T2 libthread)");
-  }
-  else {
-    st->print("  (T1 libthread)");
-  }
   st->cr();
 }
 
@@ -3323,41 +3216,10 @@ void os::yield() {
 
 os::YieldResult os::NakedYield() { thr_yield(); return os::YIELD_UNKNOWN; }
 
-
+void os::yield_all() {
-// On Solaris we found that yield_all doesn't always yield to all other threads.
-// There have been cases where there is a thread ready to execute but it doesn't
-// get an lwp as the VM thread continues to spin with sleeps of 1 millisecond.
-// The 1 millisecond wait doesn't seem long enough for the kernel to issue a
-// SIGWAITING signal which will cause a new lwp to be created. So we count the
-// number of times yield_all is called in the one loop and increase the sleep
-// time after 8 attempts. If this fails too we increase the concurrency level
-// so that the starving thread would get an lwp
-
-void os::yield_all(int attempts) {
   // Yields to all threads, including threads with lower priorities
-  if (attempts == 0) {
   os::sleep(Thread::current(), 1, false);
-  } else {
-    int iterations = attempts % 30;
-    if (iterations == 0 && !os::Solaris::T2_libthread()) {
-      // thr_setconcurrency and _getconcurrency make sense only under T1.
-      int noofLWPS = thr_getconcurrency();
-      if (noofLWPS < (Threads::number_of_threads() + 2)) {
-        thr_setconcurrency(thr_getconcurrency() + 1);
 }
-    } else if (iterations < 25) {
-      os::sleep(Thread::current(), 1, false);
-    } else {
-      os::sleep(Thread::current(), 10, false);
-    }
-  }
-}
-
-// Called from the tight loops to possibly influence time-sharing heuristics
-void os::loop_breaker(int attempts) {
-  os::yield_all(attempts);
-}
-
 
 // Interface for setting lwp priorities.  If we are using T2 libthread,
 // which forces the use of BoundThreads or we manually set UseBoundThreads,
@@ -3365,6 +3227,9 @@ void os::loop_breaker(int attempts) {
 // function is meaningless in this mode so we must adjust the real lwp's priority
 // The routines below implement the getting and setting of lwp priorities.
 //
+// Note: T2 is now the only supported libthread. UseBoundThreads flag is
+//       being deprecated and all threads are now BoundThreads
+//
 // Note: There are three priority scales used on Solaris.  Java priotities
 //       which range from 1 to 10, libthread "thr_setprio" scale which range
 //       from 0 to 127, and the current scheduling class of the process we
@@ -3437,8 +3302,6 @@ static int lwp_priocntl_init () {
 
   if (!UseThreadPriorities) return 0;
 
-  // We are using Bound threads, we need to determine our priority ranges
-  if (os::Solaris::T2_libthread() || UseBoundThreads) {
   // If ThreadPriorityPolicy is 1, switch tables
   if (ThreadPriorityPolicy == 1) {
     for (i = 0 ; i < CriticalPriority+1; i++)
@@ -3453,14 +3316,6 @@ static int lwp_priocntl_init () {
     // Set negative to distinguish from other priorities
     os::java_to_os_priority[MaxPriority] = -criticalPrio;
   }
-  }
-  // Not using Bound Threads, set to ThreadPolicy 1
-  else {
-    for ( i = 0 ; i < CriticalPriority+1; i++ ) {
-      os::java_to_os_priority[i] = prio_policy1[i];
-    }
-    return 0;
-  }
 
   // Get IDs for a set of well-known scheduling classes.
   // TODO-FIXME: GETCLINFO returns the current # of classes in the
@@ -3583,10 +3438,6 @@ int     scale_to_lwp_priority (int rMin, int rMax, int x)
 
 
 // set_lwp_class_and_priority
-//
-// Set the class and priority of the lwp.  This call should only
-// be made when using bound threads (T2 threads are bound by default).
-//
 int set_lwp_class_and_priority(int ThreadID, int lwpid,
                                int newPrio, int new_class, bool scale) {
   int rslt;
@@ -3812,8 +3663,6 @@ OSReturn os::set_native_priority(Thread* thread, int newpri) {
     status = thr_setprio(thread->osthread()->thread_id(), newpri);
   }
 
-  if (os::Solaris::T2_libthread() ||
-      (UseBoundThreads && osthread->is_vm_created())) {
   int lwp_status =
           set_lwp_class_and_priority(osthread->thread_id(),
           osthread->lwp_id(),
@@ -3828,7 +3677,6 @@ OSReturn os::set_native_priority(Thread* thread, int newpri) {
             newpri, myClass, false);
   }
   status |= lwp_status;
-  }
   return (status == 0) ? OS_OK : OS_ERR;
 }
 
@@ -4495,13 +4343,6 @@ const char* os::exception_name(int exception_code, char* buf, size_t size) {
   }
 }
 
-// (Static) wrappers for the new libthread API
-int_fnP_thread_t_iP_uP_stack_tP_gregset_t os::Solaris::_thr_getstate;
-int_fnP_thread_t_i_gregset_t os::Solaris::_thr_setstate;
-int_fnP_thread_t_i os::Solaris::_thr_setmutator;
-int_fnP_thread_t os::Solaris::_thr_suspend_mutator;
-int_fnP_thread_t os::Solaris::_thr_continue_mutator;
-
 // (Static) wrapper for getisax(2) call.
 os::Solaris::getisax_func_t os::Solaris::_getisax = 0;
 
@@ -4536,78 +4377,9 @@ static address resolve_symbol(const char* name) {
   return addr;
 }
 
-
-
-// isT2_libthread()
-//
-// Routine to determine if we are currently using the new T2 libthread.
-//
-// We determine if we are using T2 by reading /proc/self/lstatus and
-// looking for a thread with the ASLWP bit set.  If we find this status
-// bit set, we must assume that we are NOT using T2.  The T2 team
-// has approved this algorithm.
-//
-// We need to determine if we are running with the new T2 libthread
-// since setting native thread priorities is handled differently
-// when using this library.  All threads created using T2 are bound
-// threads. Calling thr_setprio is meaningless in this case.
-//
-bool isT2_libthread() {
-  static prheader_t * lwpArray = NULL;
-  static int lwpSize = 0;
-  static int lwpFile = -1;
-  lwpstatus_t * that;
-  char lwpName [128];
-  bool isT2 = false;
-
-#define ADR(x)  ((uintptr_t)(x))
-#define LWPINDEX(ary,ix)   ((lwpstatus_t *)(((ary)->pr_entsize * (ix)) + (ADR((ary) + 1))))
-
-  lwpFile = ::open("/proc/self/lstatus", O_RDONLY, 0);
-  if (lwpFile < 0) {
-      if (ThreadPriorityVerbose) warning ("Couldn't open /proc/self/lstatus\n");
-      return false;
-  }
-  lwpSize = 16*1024;
-  for (;;) {
-    ::lseek64 (lwpFile, 0, SEEK_SET);
-    lwpArray = (prheader_t *)NEW_C_HEAP_ARRAY(char, lwpSize, mtInternal);
-    if (::read(lwpFile, lwpArray, lwpSize) < 0) {
-      if (ThreadPriorityVerbose) warning("Error reading /proc/self/lstatus\n");
-      break;
-    }
-    if ((lwpArray->pr_nent * lwpArray->pr_entsize) <= lwpSize) {
-       // We got a good snapshot - now iterate over the list.
-      int aslwpcount = 0;
-      for (int i = 0; i < lwpArray->pr_nent; i++ ) {
-        that = LWPINDEX(lwpArray,i);
-        if (that->pr_flags & PR_ASLWP) {
-          aslwpcount++;
-        }
-      }
-      if (aslwpcount == 0) isT2 = true;
-      break;
-    }
-    lwpSize = lwpArray->pr_nent * lwpArray->pr_entsize;
-    FREE_C_HEAP_ARRAY(char, lwpArray, mtInternal);  // retry.
-  }
-
-  FREE_C_HEAP_ARRAY(char, lwpArray, mtInternal);
-  ::close (lwpFile);
-  if (ThreadPriorityVerbose) {
-    if (isT2) tty->print_cr("We are running with a T2 libthread\n");
-    else tty->print_cr("We are not running with a T2 libthread\n");
-  }
-  return isT2;
-}
-
-
 void os::Solaris::libthread_init() {
   address func = (address)dlsym(RTLD_DEFAULT, "_thr_suspend_allmutators");
 
-  // Determine if we are running with the new T2 libthread
-  os::Solaris::set_T2_libthread(isT2_libthread());
-
   lwp_priocntl_init();
 
   // RTLD_DEFAULT was not defined on some early versions of 5.5.1
@@ -4618,22 +4390,6 @@ void os::Solaris::libthread_init() {
     guarantee(func != NULL, "libthread.so is too old.");
   }
 
-  // Initialize the new libthread getstate API wrappers
-  func = resolve_symbol("thr_getstate");
-  os::Solaris::set_thr_getstate(CAST_TO_FN_PTR(int_fnP_thread_t_iP_uP_stack_tP_gregset_t, func));
-
-  func = resolve_symbol("thr_setstate");
-  os::Solaris::set_thr_setstate(CAST_TO_FN_PTR(int_fnP_thread_t_i_gregset_t, func));
-
-  func = resolve_symbol("thr_setmutator");
-  os::Solaris::set_thr_setmutator(CAST_TO_FN_PTR(int_fnP_thread_t_i, func));
-
-  func = resolve_symbol("thr_suspend_mutator");
-  os::Solaris::set_thr_suspend_mutator(CAST_TO_FN_PTR(int_fnP_thread_t, func));
-
-  func = resolve_symbol("thr_continue_mutator");
-  os::Solaris::set_thr_continue_mutator(CAST_TO_FN_PTR(int_fnP_thread_t, func));
-
   int size;
   void (*handler_info_func)(address *, int *);
   handler_info_func = CAST_TO_FN_PTR(void (*)(address *, int *), resolve_symbol("thr_sighndlrinfo"));
@@ -5536,11 +5292,7 @@ void os::thread_cpu_time_info(jvmtiTimerInfo *info_ptr) {
 }
 
 bool os::is_thread_cpu_time_supported() {
-  if ( os::Solaris::T2_libthread() || UseBoundThreads ) {
   return true;
-  } else {
-    return false;
-  }
 }
 
 // System loadavg support.  Returns -1 if load average cannot be obtained.

hotspot/src/os/solaris/vm/os_solaris.hpp

@@ -41,19 +41,6 @@ class Solaris {
   #define TRS_LWPID       2
   #define TRS_INVALID     3
 
-  // _T2_libthread is true if we believe we are running with the newer
-  // SunSoft lib/lwp/libthread: default Solaris 9, available Solaris 8
-  // which is a lightweight libthread that also supports all T1
-  static bool _T2_libthread;
-  // These refer to new libthread interface functions
-  // They get intialized if we dynamically detect new libthread
-  static int_fnP_thread_t_iP_uP_stack_tP_gregset_t _thr_getstate;
-  static int_fnP_thread_t_i_gregset_t _thr_setstate;
-  static int_fnP_thread_t_i _thr_setmutator;
-  static int_fnP_thread_t _thr_suspend_mutator;
-  static int_fnP_thread_t _thr_continue_mutator;
-  // libthread_init sets the above, if the new functionality is detected
-
   // initialized to libthread or lwp synchronization primitives depending on UseLWPSychronization
   static int_fnP_mutex_tP _mutex_lock;
   static int_fnP_mutex_tP _mutex_trylock;
@@ -214,29 +201,6 @@ class Solaris {
   static struct sigaction *get_chained_signal_action(int sig);
   static bool chained_handler(int sig, siginfo_t *siginfo, void *context);
 
-  // The following allow us to link against both the old and new libthread (2.8)
-  // and exploit the new libthread functionality if available.
-
-  static bool T2_libthread()                                { return _T2_libthread; }
-  static void set_T2_libthread(bool T2_libthread) { _T2_libthread = T2_libthread; }
-
-  static int thr_getstate(thread_t tid, int *flag, unsigned *lwp, stack_t *ss, gregset_t rs)
-    { return _thr_getstate(tid, flag, lwp, ss, rs); }
-  static void set_thr_getstate(int_fnP_thread_t_iP_uP_stack_tP_gregset_t func)
-    { _thr_getstate = func; }
-
-  static int thr_setstate(thread_t tid, int flag, gregset_t rs)   { return _thr_setstate(tid, flag, rs); }
-  static void set_thr_setstate(int_fnP_thread_t_i_gregset_t func) { _thr_setstate = func; }
-
-  static int thr_setmutator(thread_t tid, int enabled)    { return _thr_setmutator(tid, enabled); }
-  static void set_thr_setmutator(int_fnP_thread_t_i func) { _thr_setmutator = func; }
-
-  static int  thr_suspend_mutator(thread_t tid)              { return _thr_suspend_mutator(tid); }
-  static void set_thr_suspend_mutator(int_fnP_thread_t func) { _thr_suspend_mutator = func; }
-
-  static int  thr_continue_mutator(thread_t tid)              { return _thr_continue_mutator(tid); }
-  static void set_thr_continue_mutator(int_fnP_thread_t func) { _thr_continue_mutator = func; }
-
   // Allows us to switch between lwp and thread -based synchronization
   static int mutex_lock(mutex_t *mx)    { return _mutex_lock(mx); }
   static int mutex_trylock(mutex_t *mx) { return _mutex_trylock(mx); }

hotspot/src/os/windows/vm/os_windows.cpp

@@ -3518,7 +3518,7 @@ os::YieldResult os::NakedYield() {
 
 void os::yield() {  os::NakedYield(); }
 
-void os::yield_all(int attempts) {
+void os::yield_all() {
   // Yields to all threads, including threads with lower priorities
   Sleep(1);
 }
@@ -3864,12 +3864,6 @@ void os::init(void) {
   win32::setmode_streams();
   init_page_sizes((size_t) win32::vm_page_size());
 
-  // For better scalability on MP systems (must be called after initialize_system_info)
-#ifndef PRODUCT
-  if (is_MP()) {
-    NoYieldsInMicrolock = true;
-  }
-#endif
   // This may be overridden later when argument processing is done.
   FLAG_SET_ERGO(bool, UseLargePagesIndividualAllocation,
     os::win32::is_windows_2003());

hotspot/src/os/windows/vm/os_windows.inline.hpp

@@ -52,9 +52,6 @@ inline void* os::dll_lookup(void *lib, const char *name) {
   return (void*)::GetProcAddress((HMODULE)lib, name);
 }
 
-// Used to improve time-sharing on some systems
-inline void os::loop_breaker(int attempts) {}
-
 inline bool os::obsolete_option(const JavaVMOption *option) {
   return false;
 }

hotspot/src/os_cpu/solaris_sparc/vm/os_solaris_sparc.cpp

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1999, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1999, 2014, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -270,31 +270,6 @@ frame os::current_frame() {
   }
 }
 
-static int threadgetstate(thread_t tid, int *flags, lwpid_t *lwp, stack_t *ss, gregset_t rs, lwpstatus_t *lwpstatus) {
-  char lwpstatusfile[PROCFILE_LENGTH];
-  int lwpfd, err;
-
-  if (err = os::Solaris::thr_getstate(tid, flags, lwp, ss, rs))
-    return (err);
-  if (*flags == TRS_LWPID) {
-    sprintf(lwpstatusfile, "/proc/%d/lwp/%d/lwpstatus", getpid(),
-            *lwp);
-    if ((lwpfd = ::open(lwpstatusfile, O_RDONLY)) < 0) {
-      perror("thr_mutator_status: open lwpstatus");
-      return (EINVAL);
-    }
-    if (pread(lwpfd, lwpstatus, sizeof (lwpstatus_t), (off_t)0) !=
-        sizeof (lwpstatus_t)) {
-      perror("thr_mutator_status: read lwpstatus");
-      (void) ::close(lwpfd);
-      return (EINVAL);
-    }
-    (void) ::close(lwpfd);
-  }
-  return (0);
-}
-
-
 bool os::is_allocatable(size_t bytes) {
 #ifdef _LP64
    return true;

hotspot/src/os_cpu/solaris_x86/vm/os_solaris_x86.cpp

@@ -256,30 +256,6 @@ frame os::current_frame() {
   }
 }
 
-static int threadgetstate(thread_t tid, int *flags, lwpid_t *lwp, stack_t *ss, gregset_t rs, lwpstatus_t *lwpstatus) {
-  char lwpstatusfile[PROCFILE_LENGTH];
-  int lwpfd, err;
-
-  if (err = os::Solaris::thr_getstate(tid, flags, lwp, ss, rs))
-    return (err);
-  if (*flags == TRS_LWPID) {
-    sprintf(lwpstatusfile, "/proc/%d/lwp/%d/lwpstatus", getpid(),
-            *lwp);
-    if ((lwpfd = open(lwpstatusfile, O_RDONLY)) < 0) {
-      perror("thr_mutator_status: open lwpstatus");
-      return (EINVAL);
-    }
-    if (pread(lwpfd, lwpstatus, sizeof (lwpstatus_t), (off_t)0) !=
-        sizeof (lwpstatus_t)) {
-      perror("thr_mutator_status: read lwpstatus");
-      (void) close(lwpfd);
-      return (EINVAL);
-    }
-    (void) close(lwpfd);
-  }
-  return (0);
-}
-
 #ifndef AMD64
 
 // Detecting SSE support by OS

hotspot/src/share/tools/LogCompilation/src/com/sun/hotspot/tools/compiler/CallSite.java

@@ -25,6 +25,7 @@
 package com.sun.hotspot.tools.compiler;
 
 import java.io.PrintStream;
+import java.util.ArrayDeque;
 import java.util.ArrayList;
 import java.util.List;
 
@@ -40,6 +41,7 @@ public class CallSite {
     private int endNodes;
     private int endLiveNodes;
     private double timeStamp;
+    private long inlineId;
 
     CallSite() {
     }
@@ -94,7 +96,7 @@ public class CallSite {
 
     public void print(PrintStream stream, int indent) {
         emit(stream, indent);
-        String m = getMethod().getHolder().replace('/', '.') + "::" + getMethod().getName();
+        String m = getMethod().getHolder() + "::" + getMethod().getName();
         if (getReason() == null) {
             stream.print("  @ " + getBci() + " " + m + " (" + getMethod().getBytes() + " bytes)");
 
@@ -214,4 +216,45 @@ public class CallSite {
         return timeStamp;
     }
 
+    private boolean matches(CallSite other) {
+        // Every late inline call site has a unique inline id. If the
+        // call site we're looking for has one then use it other rely
+        // on method name and bci.
+        if (other.inlineId != 0) {
+            return inlineId == other.inlineId;
+        }
+        return method.equals(other.method) && bci == other.bci;
+    }
+
+    public CallSite findCallSite(ArrayDeque<CallSite> sites) {
+        // Locate a late inline call site. Multiple chains of
+        // identical call sites with the same method name/bci are
+        // possible so we have to try them all until we find the late
+        // inline call site that has a matching inline id.
+        CallSite site = sites.pop();
+        for (CallSite c : calls) {
+            if (c.matches(site)) {
+                if (!sites.isEmpty()) {
+                    CallSite res = c.findCallSite(sites);
+                    if (res != null) {
+                        sites.push(site);
+                        return res;
+                    }
+                } else {
+                    sites.push(site);
+                    return c;
+                }
+            }
+        }
+        sites.push(site);
+        return null;
+    }
+
+    public long getInlineId() {
+        return inlineId;
+    }
+
+    public void setInlineId(long inlineId) {
+        this.inlineId = inlineId;
+    }
 }

hotspot/src/share/tools/LogCompilation/src/com/sun/hotspot/tools/compiler/LogParser.java

@@ -31,6 +31,7 @@ package com.sun.hotspot.tools.compiler;
 
 import java.io.FileReader;
 import java.io.Reader;
+import java.util.ArrayDeque;
 import java.util.ArrayList;
 import java.util.Collections;
 import java.util.Comparator;
@@ -144,9 +145,12 @@ public class LogParser extends DefaultHandler implements ErrorHandler, Constants
     private Stack<CallSite> scopes = new Stack<CallSite>();
     private Compilation compile;
     private CallSite site;
+    private CallSite methodHandleSite;
     private Stack<Phase> phaseStack = new Stack<Phase>();
     private UncommonTrapEvent currentTrap;
-    private Stack<CallSite> late_inline_scope;
+    private Stack<CallSite> lateInlineScope;
+    private boolean lateInlining;
+
 
     long parseLong(String l) {
         try {
@@ -330,18 +334,61 @@ public class LogParser extends DefaultHandler implements ErrorHandler, Constants
             }
             methods.put(id, m);
         } else if (qname.equals("call")) {
-            site = new CallSite(bci, method(search(atts, "method")));
+            if (methodHandleSite != null) {
+                methodHandleSite = null;
+            }
+            Method m = method(search(atts, "method"));
+            if (lateInlining && scopes.size() == 0) {
+                // re-attempting already seen call site (late inlining for MH invokes)
+                if (m != site.getMethod()) {
+                    if (bci != site.getBci()) {
+                        System.out.println(m + " bci: " + bci);
+                        System.out.println(site.getMethod() +  " bci: " + site.getBci());
+                        throw new InternalError("bci mismatch after late inlining");
+                    }
+                    site.setMethod(m);
+                }
+            } else {
+                site = new CallSite(bci, m);
+            }
             site.setCount(Integer.parseInt(search(atts, "count", "0")));
             String receiver = atts.getValue("receiver");
             if (receiver != null) {
                 site.setReceiver(type(receiver));
                 site.setReceiver_count(Integer.parseInt(search(atts, "receiver_count")));
             }
+            int methodHandle = Integer.parseInt(search(atts, "method_handle_intrinsic", "0"));
+            if (lateInlining && scopes.size() == 0) {
+                // The call was added before this round of late inlining
+            } else if (methodHandle == 0) {
                 scopes.peek().add(site);
+            } else {
+                // method handle call site can be followed by another
+                // call (in case it is inlined). If that happens we
+                // discard the method handle call site. So we keep
+                // track of it but don't add it to the list yet.
+                methodHandleSite = site;
+            }
         } else if (qname.equals("regalloc")) {
             compile.setAttempts(Integer.parseInt(search(atts, "attempts")));
         } else if (qname.equals("inline_fail")) {
+            if (methodHandleSite != null) {
+                scopes.peek().add(methodHandleSite);
+                methodHandleSite = null;
+            }
+            if (lateInlining && scopes.size() == 0) {
+                site.setReason(search(atts, "reason"));
+                lateInlining = false;
+            } else {
                 scopes.peek().last().setReason(search(atts, "reason"));
+            }
+        } else if (qname.equals("inline_success")) {
+            if (methodHandleSite != null) {
+                throw new InternalError("method handle site should have been replaced");
+            }
+            if (lateInlining && scopes.size() == 0) {
+                site.setReason(null);
+            }
         } else if (qname.equals("failure")) {
             failureReason = search(atts, "reason");
         } else if (qname.equals("task_done")) {
@@ -371,22 +418,30 @@ public class LogParser extends DefaultHandler implements ErrorHandler, Constants
                 // ignore for now
             }
         } else if (qname.equals("late_inline")) {
-            late_inline_scope = new Stack<CallSite>();
+            long inlineId = Long.parseLong(search(atts, "inline_id"));
+            lateInlineScope = new Stack<CallSite>();
             site = new CallSite(-999, method(search(atts, "method")));
-            late_inline_scope.push(site);
+            site.setInlineId(inlineId);
+            lateInlineScope.push(site);
         } else if (qname.equals("jvms")) {
             // <jvms bci='4' method='java/io/DataInputStream readChar ()C' bytes='40' count='5815' iicount='20815'/>
             if (currentTrap != null) {
                 currentTrap.addJVMS(atts.getValue("method"), Integer.parseInt(atts.getValue("bci")));
-            } else if (late_inline_scope != null) {
+            } else if (lateInlineScope != null) {
                 bci = Integer.parseInt(search(atts, "bci"));
                 site = new CallSite(bci, method(search(atts, "method")));
-                late_inline_scope.push(site);
+                lateInlineScope.push(site);
             } else {
                 // Ignore <eliminate_allocation type='667'>,
                 //        <eliminate_lock lock='1'>,
                 //        <replace_string_concat arguments='2' string_alloc='0' multiple='0'>
             }
+        } else if (qname.equals("inline_id")) {
+            if (methodHandleSite != null) {
+                throw new InternalError("method handle site should have been replaced");
+            }
+            long id = Long.parseLong(search(atts, "id"));
+            site.setInlineId(id);
         } else if (qname.equals("nmethod")) {
             String id = makeId(atts);
             NMethod nm = new NMethod(Double.parseDouble(search(atts, "stamp")),
@@ -396,8 +451,18 @@ public class LogParser extends DefaultHandler implements ErrorHandler, Constants
             nmethods.put(id, nm);
             events.add(nm);
         } else if (qname.equals("parse")) {
+            if (methodHandleSite != null) {
+                throw new InternalError("method handle site should have been replaced");
+            }
             Method m = method(search(atts, "method"));
-            if (scopes.size() == 0) {
+            if (lateInlining && scopes.size() == 0) {
+                if (site.getMethod() != m) {
+                    System.out.println(site.getMethod());
+                    System.out.println(m);
+                    throw new InternalError("Unexpected method mismatch during late inlining");
+                }
+            }
+            if (scopes.size() == 0 && !lateInlining) {
                 compile.setMethod(m);
                 scopes.push(site);
             } else {
@@ -427,14 +492,19 @@ public class LogParser extends DefaultHandler implements ErrorHandler, Constants
         if (qname.equals("parse")) {
             indent -= 2;
             scopes.pop();
+            if (scopes.size() == 0) {
+                lateInlining = false;
+            }
         } else if (qname.equals("uncommon_trap")) {
             currentTrap = null;
         } else if (qname.equals("late_inline")) {
             // Populate late inlining info.
-
+            if (scopes.size() != 0) {
-            // late_inline scopes are specified in reverse order:
+                throw new InternalError("scopes should be empty for late inline");
+            }
+            // late inline scopes are specified in reverse order:
             // compiled method should be on top of stack.
-            CallSite caller = late_inline_scope.pop();
+            CallSite caller = lateInlineScope.pop();
             Method m = compile.getMethod();
             if (m != caller.getMethod()) {
                 System.out.println(m);
@@ -444,28 +514,42 @@ public class LogParser extends DefaultHandler implements ErrorHandler, Constants
 
             // late_inline contains caller+bci info, convert it
             // to bci+callee info used by LogCompilation.
-            site = compile.getLateInlineCall();
+            CallSite lateInlineSite = compile.getLateInlineCall();
+            ArrayDeque<CallSite> thisCallScopes = new ArrayDeque<CallSite>();
             do {
                 bci = caller.getBci();
                 // Next inlined call.
-                caller = late_inline_scope.pop();
+                caller = lateInlineScope.pop();
                 CallSite callee =  new CallSite(bci, caller.getMethod());
-                site.add(callee);
+                callee.setInlineId(caller.getInlineId());
-                site = callee;
+                thisCallScopes.addLast(callee);
-            } while (!late_inline_scope.empty());
+                lateInlineSite.add(callee);
+                lateInlineSite = callee;
+            } while (!lateInlineScope.empty());
+
+            site = compile.getCall().findCallSite(thisCallScopes);
+            if (site == null) {
+                System.out.println(caller.getMethod() + " bci: " + bci);
+                throw new InternalError("couldn't find call site");
+            }
+            lateInlining = true;
 
             if (caller.getBci() != -999) {
                 System.out.println(caller.getMethod());
                 throw new InternalError("broken late_inline info");
             }
             if (site.getMethod() != caller.getMethod()) {
+                if (site.getInlineId() == caller.getInlineId()) {
+                    site.setMethod(caller.getMethod());
+                } else {
                     System.out.println(site.getMethod());
                     System.out.println(caller.getMethod());
                     throw new InternalError("call site and late_inline info don't match");
                 }
+            }
             // late_inline is followed by parse with scopes.size() == 0,
             // 'site' will be pushed to scopes.
-            late_inline_scope = null;
+            lateInlineScope = null;
         } else if (qname.equals("task")) {
             types.clear();
             methods.clear();

hotspot/src/share/tools/LogCompilation/src/com/sun/hotspot/tools/compiler/Method.java

@@ -51,15 +51,15 @@ public class Method implements Constants {
 
     String format(int osr_bci) {
         if (osr_bci >= 0) {
-            return getHolder().replace('/', '.') + "::" + getName() + " @ " + osr_bci + " (" + getBytes() + " bytes)";
+            return getHolder() + "::" + getName() + " @ " + osr_bci + " (" + getBytes() + " bytes)";
         } else {
-            return getHolder().replace('/', '.') + "::" + getName() + " (" + getBytes() + " bytes)";
+            return getHolder() + "::" + getName() + " (" + getBytes() + " bytes)";
         }
     }
 
     @Override
     public String toString() {
-        return getHolder().replace('/', '.') + "::" + getName() + " (" + getBytes() + " bytes)";
+        return getHolder() + "::" + getName() + " (" + getBytes() + " bytes)";
     }
 
     public String getHolder() {
@@ -117,4 +117,14 @@ public class Method implements Constants {
     public void setFlags(String flags) {
         this.flags = flags;
     }
+
+    @Override
+    public boolean equals(Object o) {
+        if (o instanceof Method) {
+            Method other = (Method)o;
+            return holder.equals(other.holder) && name.equals(other.name) &&
+                arguments.equals(other.arguments) && returnType.equals(other.returnType);
+        }
+        return false;
+    }
 }

hotspot/src/share/vm/adlc/output_h.cpp

@@ -1613,21 +1613,20 @@ void ArchDesc::declareClasses(FILE *fp) {
     // Each instruction attribute results in a virtual call of same name.
     // The ins_cost is not handled here.
     Attribute *attr = instr->_attribs;
-    bool avoid_back_to_back = false;
+    Attribute *avoid_back_to_back_attr = NULL;
     while (attr != NULL) {
-      if (strcmp (attr->_ident, "ins_cost") != 0 &&
+      if (strcmp (attr->_ident, "ins_is_TrapBasedCheckNode") == 0) {
+        fprintf(fp, "  virtual bool           is_TrapBasedCheckNode() const { return %s; }\n", attr->_val);
+      } else if (strcmp (attr->_ident, "ins_cost") != 0 &&
           strncmp(attr->_ident, "ins_field_", 10) != 0 &&
           // Must match function in node.hpp: return type bool, no prefix "ins_".
           strcmp (attr->_ident, "ins_is_TrapBasedCheckNode") != 0 &&
           strcmp (attr->_ident, "ins_short_branch") != 0) {
         fprintf(fp, "  virtual int            %s() const { return %s; }\n", attr->_ident, attr->_val);
       }
-      // Check value for ins_avoid_back_to_back, and if it is true (1), set the flag
+      if (strcmp(attr->_ident, "ins_avoid_back_to_back") == 0) {
-      if (!strcmp(attr->_ident, "ins_avoid_back_to_back") != 0 && attr->int_val(*this) != 0)
+        avoid_back_to_back_attr = attr;
-        avoid_back_to_back = true;
+      }
-      if (strcmp (attr->_ident, "ins_is_TrapBasedCheckNode") == 0)
-        fprintf(fp, "  virtual bool           is_TrapBasedCheckNode() const { return %s; }\n", attr->_val);
-
       attr = (Attribute *)attr->_next;
     }
 
@@ -1799,11 +1798,11 @@ void ArchDesc::declareClasses(FILE *fp) {
     }
 
     // flag: if this instruction should not be generated back to back.
-    if ( avoid_back_to_back ) {
+    if (avoid_back_to_back_attr != NULL) {
       if (node_flags_set) {
-        fprintf(fp," | Flag_avoid_back_to_back");
+        fprintf(fp," | (%s)", avoid_back_to_back_attr->_val);
       } else {
-        fprintf(fp,"init_flags(Flag_avoid_back_to_back");
+        fprintf(fp,"init_flags((%s)", avoid_back_to_back_attr->_val);
         node_flags_set = true;
       }
     }

hotspot/src/share/vm/asm/codeBuffer.cpp

@@ -968,6 +968,7 @@ void CodeBuffer::verify_section_allocation() {
 
 void CodeBuffer::log_section_sizes(const char* name) {
   if (xtty != NULL) {
+    ttyLocker ttyl;
     // log info about buffer usage
     xtty->print_cr("<blob name='%s' size='%d'>", name, _total_size);
     for (int n = (int) CodeBuffer::SECT_FIRST; n < (int) CodeBuffer::SECT_LIMIT; n++) {

hotspot/src/share/vm/ci/ciKlass.cpp

@@ -237,3 +237,9 @@ void ciKlass::print_impl(outputStream* st) {
 void ciKlass::print_name_on(outputStream* st) {
   name()->print_symbol_on(st);
 }
+
+const char* ciKlass::external_name() const {
+  GUARDED_VM_ENTRY(
+    return get_Klass()->external_name();
+  )
+}

hotspot/src/share/vm/ci/ciKlass.hpp

@@ -125,6 +125,8 @@ public:
   virtual ciKlass* exact_klass() = 0;
 
   void print_name_on(outputStream* st);
+
+  const char* external_name() const;
 };
 
 #endif // SHARE_VM_CI_CIKLASS_HPP

hotspot/src/share/vm/ci/ciSymbol.cpp

@@ -123,6 +123,10 @@ void ciSymbol::print_symbol_on(outputStream *st) {
   GUARDED_VM_ENTRY(get_symbol()->print_symbol_on(st);)
 }
 
+const char* ciSymbol::as_klass_external_name() const {
+  GUARDED_VM_ENTRY(return get_symbol()->as_klass_external_name(););
+}
+
 // ------------------------------------------------------------------
 // ciSymbol::make_impl
 //

hotspot/src/share/vm/ci/ciSymbol.hpp

@@ -90,6 +90,7 @@ public:
   void print_symbol() {
     print_symbol_on(tty);
   }
+  const char* as_klass_external_name() const;
 
   // Make a ciSymbol from a C string.
   // Consider adding to vmSymbols.hpp instead of using this constructor.

hotspot/src/share/vm/classfile/classFileParser.cpp

@@ -4180,8 +4180,12 @@ ClassFileParser::~ClassFileParser() {
 
   clear_class_metadata();
 
-  // deallocate the klass if already created.
+  // deallocate the klass if already created.  Don't directly deallocate, but add
-  MetadataFactory::free_metadata(_loader_data, _klass);
+  // to the deallocate list so that the klass is removed from the CLD::_klasses list
+  // at a safepoint.
+  if (_klass != NULL) {
+    _loader_data->add_to_deallocate_list(_klass);
+  }
   _klass = NULL;
 }
 

hotspot/src/share/vm/classfile/javaClasses.cpp

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1997, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1997, 2014, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -464,25 +464,26 @@ bool java_lang_String::equals(oop str1, oop str2) {
 void java_lang_String::print(oop java_string, outputStream* st) {
   assert(java_string->klass() == SystemDictionary::String_klass(), "must be java_string");
   typeArrayOop value  = java_lang_String::value(java_string);
-  int          offset = java_lang_String::offset(java_string);
-  int          length = java_lang_String::length(java_string);
 
-  int end = MIN2(length, 100);
   if (value == NULL) {
     // This can happen if, e.g., printing a String
     // object before its initializer has been called
-    st->print_cr("NULL");
+    st->print("NULL");
-  } else {
+    return;
+  }
+
+  int offset = java_lang_String::offset(java_string);
+  int length = java_lang_String::length(java_string);
+
   st->print("\"");
   for (int index = 0; index < length; index++) {
     st->print("%c", value->char_at(index + offset));
   }
   st->print("\"");
 }
-}
 
-static void initialize_static_field(fieldDescriptor* fd, TRAPS) {
+
-  Handle mirror (THREAD, fd->field_holder()->java_mirror());
+static void initialize_static_field(fieldDescriptor* fd, Handle mirror, TRAPS) {
   assert(mirror.not_null() && fd->is_static(), "just checking");
   if (fd->has_initial_value()) {
     BasicType t = fd->field_type();
@@ -549,21 +550,45 @@ void java_lang_Class::fixup_mirror(KlassHandle k, TRAPS) {
   create_mirror(k, Handle(NULL), CHECK);
 }
 
-oop java_lang_Class::create_mirror(KlassHandle k, Handle protection_domain, TRAPS) {
+void java_lang_Class::initialize_mirror_fields(KlassHandle k,
+                                               Handle mirror,
+                                               Handle protection_domain,
+                                               TRAPS) {
+  // Allocate a simple java object for a lock.
+  // This needs to be a java object because during class initialization
+  // it can be held across a java call.
+  typeArrayOop r = oopFactory::new_typeArray(T_INT, 0, CHECK);
+  set_init_lock(mirror(), r);
+
+  // Set protection domain also
+  set_protection_domain(mirror(), protection_domain());
+
+  // Initialize static fields
+  InstanceKlass::cast(k())->do_local_static_fields(&initialize_static_field, mirror, CHECK);
+}
+
+void java_lang_Class::create_mirror(KlassHandle k, Handle protection_domain, TRAPS) {
   assert(k->java_mirror() == NULL, "should only assign mirror once");
   // Use this moment of initialization to cache modifier_flags also,
   // to support Class.getModifiers().  Instance classes recalculate
   // the cached flags after the class file is parsed, but before the
   // class is put into the system dictionary.
-  int computed_modifiers = k->compute_modifier_flags(CHECK_0);
+  int computed_modifiers = k->compute_modifier_flags(CHECK);
   k->set_modifier_flags(computed_modifiers);
   // Class_klass has to be loaded because it is used to allocate
   // the mirror.
   if (SystemDictionary::Class_klass_loaded()) {
     // Allocate mirror (java.lang.Class instance)
-    Handle mirror = InstanceMirrorKlass::cast(SystemDictionary::Class_klass())->allocate_instance(k, CHECK_0);
+    Handle mirror = InstanceMirrorKlass::cast(SystemDictionary::Class_klass())->allocate_instance(k, CHECK);
+
+    // Setup indirection from mirror->klass
+    if (!k.is_null()) {
+      java_lang_Class::set_klass(mirror(), k());
+    }
 
     InstanceMirrorKlass* mk = InstanceMirrorKlass::cast(mirror->klass());
+    assert(oop_size(mirror()) == mk->instance_size(k), "should have been set");
+
     java_lang_Class::set_static_oop_field_count(mirror(), mk->compute_static_oop_field_count(mirror()));
 
     // It might also have a component mirror.  This mirror must already exist.
@@ -586,19 +611,22 @@ oop java_lang_Class::create_mirror(KlassHandle k, Handle protection_domain, TRAP
     } else {
       assert(k->oop_is_instance(), "Must be");
 
-      // Allocate a simple java object for a lock.
+      initialize_mirror_fields(k, mirror, protection_domain, THREAD);
-      // This needs to be a java object because during class initialization
+      if (HAS_PENDING_EXCEPTION) {
-      // it can be held across a java call.
+        // If any of the fields throws an exception like OOM remove the klass field
-      typeArrayOop r = oopFactory::new_typeArray(T_INT, 0, CHECK_NULL);
+        // from the mirror so GC doesn't follow it after the klass has been deallocated.
-      set_init_lock(mirror(), r);
+        // This mirror looks like a primitive type, which logically it is because it
-
+        // it represents no class.
-      // Set protection domain also
+        java_lang_Class::set_klass(mirror(), NULL);
-      set_protection_domain(mirror(), protection_domain());
+        return;
-
+      }
-      // Initialize static fields
+    }
-      InstanceKlass::cast(k())->do_local_static_fields(&initialize_static_field, CHECK_NULL);
+
+    // Setup indirection from klass->mirror last
+    // after any exceptions can happen during allocations.
+    if (!k.is_null()) {
+      k->set_java_mirror(mirror());
     }
-    return mirror();
   } else {
     if (fixup_mirror_list() == NULL) {
       GrowableArray<Klass*>* list =
@@ -606,12 +634,10 @@ oop java_lang_Class::create_mirror(KlassHandle k, Handle protection_domain, TRAP
       set_fixup_mirror_list(list);
     }
     fixup_mirror_list()->push(k());
-    return NULL;
   }
 }
 
 
-
 int  java_lang_Class::oop_size(oop java_class) {
   assert(_oop_size_offset != 0, "must be set");
   return java_class->int_field(_oop_size_offset);

hotspot/src/share/vm/classfile/javaClasses.hpp

@@ -246,11 +246,12 @@ class java_lang_Class : AllStatic {
 
   static void set_init_lock(oop java_class, oop init_lock);
   static void set_protection_domain(oop java_class, oop protection_domain);
+  static void initialize_mirror_fields(KlassHandle k, Handle mirror, Handle protection_domain, TRAPS);
  public:
   static void compute_offsets();
 
   // Instance creation
-  static oop  create_mirror(KlassHandle k, Handle protection_domain, TRAPS);
+  static void create_mirror(KlassHandle k, Handle protection_domain, TRAPS);
   static void fixup_mirror(KlassHandle k, TRAPS);
   static oop  create_basic_type_mirror(const char* basic_type_name, BasicType type, TRAPS);
   // Conversion

hotspot/src/share/vm/classfile/symbolTable.cpp

@@ -810,11 +810,11 @@ void StringTable::buckets_oops_do(OopClosure* f, int start_idx, int end_idx) {
   const int limit = the_table()->table_size();
 
   assert(0 <= start_idx && start_idx <= limit,
-         err_msg("start_idx (" INT32_FORMAT ") is out of bounds", start_idx));
+         err_msg("start_idx (%d) is out of bounds", start_idx));
   assert(0 <= end_idx && end_idx <= limit,
-         err_msg("end_idx (" INT32_FORMAT ") is out of bounds", end_idx));
+         err_msg("end_idx (%d) is out of bounds", end_idx));
   assert(start_idx <= end_idx,
-         err_msg("Index ordering: start_idx=" INT32_FORMAT", end_idx=" INT32_FORMAT,
+         err_msg("Index ordering: start_idx=%d, end_idx=%d",
                  start_idx, end_idx));
 
   for (int i = start_idx; i < end_idx; i += 1) {
@@ -833,11 +833,11 @@ void StringTable::buckets_unlink_or_oops_do(BoolObjectClosure* is_alive, OopClos
   const int limit = the_table()->table_size();
 
   assert(0 <= start_idx && start_idx <= limit,
-         err_msg("start_idx (" INT32_FORMAT ") is out of bounds", start_idx));
+         err_msg("start_idx (%d) is out of bounds", start_idx));
   assert(0 <= end_idx && end_idx <= limit,
-         err_msg("end_idx (" INT32_FORMAT ") is out of bounds", end_idx));
+         err_msg("end_idx (%d) is out of bounds", end_idx));
   assert(start_idx <= end_idx,
-         err_msg("Index ordering: start_idx=" INT32_FORMAT", end_idx=" INT32_FORMAT,
+         err_msg("Index ordering: start_idx=%d, end_idx=%d",
                  start_idx, end_idx));
 
   for (int i = start_idx; i < end_idx; ++i) {

hotspot/src/share/vm/classfile/systemDictionary.cpp

@@ -826,47 +826,6 @@ Klass* SystemDictionary::resolve_instance_class_or_null(Symbol* name,
       }
     } // load_instance_class loop
 
-    if (HAS_PENDING_EXCEPTION) {
-      // An exception, such as OOM could have happened at various places inside
-      // load_instance_class. We might have partially initialized a shared class
-      // and need to clean it up.
-      if (class_loader.is_null()) {
-        // In some cases k may be null. Let's find the shared class again.
-        instanceKlassHandle ik(THREAD, find_shared_class(name));
-        if (ik.not_null()) {
-          if (ik->class_loader_data() == NULL) {
-            // We didn't go as far as Klass::restore_unshareable_info(),
-            // so nothing to clean up.
-          } else {
-            Klass *kk;
-            {
-              MutexLocker mu(SystemDictionary_lock, THREAD);
-              kk = find_class(d_index, d_hash, name, ik->class_loader_data());
-            }
-            if (kk != NULL) {
-              // No clean up is needed if the shared class has been entered
-              // into system dictionary, as load_shared_class() won't be called
-              // again.
-            } else {
-              // This must be done outside of the SystemDictionary_lock to
-              // avoid deadlock.
-              //
-              // Note that Klass::restore_unshareable_info (called via
-              // load_instance_class above) is also called outside
-              // of SystemDictionary_lock. Other threads are blocked from
-              // loading this class because they are waiting on the
-              // SystemDictionary_lock until this thread removes
-              // the placeholder below.
-              //
-              // This need to be re-thought when parallel-capable non-boot
-              // classloaders are supported by CDS (today they're not).
-              clean_up_shared_class(ik, class_loader, THREAD);
-            }
-          }
-        }
-      }
-    }
-
     if (load_instance_added == true) {
       // clean up placeholder entries for LOAD_INSTANCE success or error
       // This brackets the SystemDictionary updates for both defining
@@ -1272,19 +1231,6 @@ instanceKlassHandle SystemDictionary::load_shared_class(
   return ik;
 }
 
-void SystemDictionary::clean_up_shared_class(instanceKlassHandle ik, Handle class_loader, TRAPS) {
-  // Updating methods must be done under a lock so multiple
-  // threads don't update these in parallel
-  // Shared classes are all currently loaded by the bootstrap
-  // classloader, so this will never cause a deadlock on
-  // a custom class loader lock.
-  {
-    Handle lockObject = compute_loader_lock_object(class_loader, THREAD);
-    check_loader_lock_contention(lockObject, THREAD);
-    ObjectLocker ol(lockObject, THREAD, true);
-    ik->remove_unshareable_info();
-  }
-}
 
 instanceKlassHandle SystemDictionary::load_instance_class(Symbol* class_name, Handle class_loader, TRAPS) {
   instanceKlassHandle nh = instanceKlassHandle(); // null Handle

hotspot/src/share/vm/classfile/systemDictionary.hpp

@@ -617,7 +617,6 @@ private:
                                                Handle class_loader, TRAPS);
   static instanceKlassHandle load_shared_class(instanceKlassHandle ik,
                                                Handle class_loader, TRAPS);
-  static void clean_up_shared_class(instanceKlassHandle ik, Handle class_loader, TRAPS);
   static instanceKlassHandle load_instance_class(Symbol* class_name, Handle class_loader, TRAPS);
   static Handle compute_loader_lock_object(Handle class_loader, TRAPS);
   static void check_loader_lock_contention(Handle loader_lock, TRAPS);

hotspot/src/share/vm/compiler/compileLog.cpp

@@ -106,7 +106,7 @@ int CompileLog::identify(ciBaseObject* obj) {
     if (mobj->is_klass()) {
       ciKlass* klass = mobj->as_klass();
       begin_elem("klass id='%d'", id);
-      name(klass->name());
+      name(klass);
       if (!klass->is_loaded()) {
         print(" unloaded='1'");
       } else {
@@ -171,6 +171,15 @@ void CompileLog::name(ciSymbol* name) {
   print("'");
 }
 
+void CompileLog::name(ciKlass* k) {
+  print(" name='");
+  if (!k->is_loaded()) {
+    text()->print(k->name()->as_klass_external_name());
+  } else {
+    text()->print(k->external_name());
+  }
+  print("'");
+}
 
 // ------------------------------------------------------------------
 // CompileLog::clear_identities

hotspot/src/share/vm/compiler/compileLog.hpp

@@ -28,6 +28,7 @@
 #include "utilities/xmlstream.hpp"
 
 class ciBaseObject;
+class ciKlass;
 class ciObject;
 class ciMetadata;
 class ciSymbol;
@@ -72,6 +73,7 @@ class CompileLog : public xmlStream {
 
   void          name(ciSymbol* s);               // name='s'
   void          name(Symbol* s)                  { xmlStream::name(s); }
+  void          name(ciKlass* k);
 
   // Output an object description, return obj->ident().
   int           identify(ciBaseObject* obj);

hotspot/src/share/vm/gc_implementation/g1/concurrentG1Refine.cpp

@@ -57,10 +57,10 @@ ConcurrentG1Refine::ConcurrentG1Refine(G1CollectedHeap* g1h) :
 
   _threads = NEW_C_HEAP_ARRAY(ConcurrentG1RefineThread*, _n_threads, mtGC);
 
-  int worker_id_offset = (int)DirtyCardQueueSet::num_par_ids();
+  uint worker_id_offset = DirtyCardQueueSet::num_par_ids();
 
   ConcurrentG1RefineThread *next = NULL;
-  for (int i = _n_threads - 1; i >= 0; i--) {
+  for (uint i = _n_threads - 1; i != UINT_MAX; i--) {
     ConcurrentG1RefineThread* t = new ConcurrentG1RefineThread(this, next, worker_id_offset, i);
     assert(t != NULL, "Conc refine should have been created");
     if (t->osthread() == NULL) {
@@ -87,7 +87,7 @@ void ConcurrentG1Refine::init() {
 
 void ConcurrentG1Refine::stop() {
   if (_threads != NULL) {
-    for (int i = 0; i < _n_threads; i++) {
+    for (uint i = 0; i < _n_threads; i++) {
       _threads[i]->stop();
     }
   }
@@ -96,7 +96,7 @@ void ConcurrentG1Refine::stop() {
 void ConcurrentG1Refine::reinitialize_threads() {
   reset_threshold_step();
   if (_threads != NULL) {
-    for (int i = 0; i < _n_threads; i++) {
+    for (uint i = 0; i < _n_threads; i++) {
       _threads[i]->initialize();
     }
   }
@@ -104,7 +104,7 @@ void ConcurrentG1Refine::reinitialize_threads() {
 
 ConcurrentG1Refine::~ConcurrentG1Refine() {
   if (_threads != NULL) {
-    for (int i = 0; i < _n_threads; i++) {
+    for (uint i = 0; i < _n_threads; i++) {
       delete _threads[i];
     }
     FREE_C_HEAP_ARRAY(ConcurrentG1RefineThread*, _threads, mtGC);
@@ -113,7 +113,7 @@ ConcurrentG1Refine::~ConcurrentG1Refine() {
 
 void ConcurrentG1Refine::threads_do(ThreadClosure *tc) {
   if (_threads != NULL) {
-    for (int i = 0; i < _n_threads; i++) {
+    for (uint i = 0; i < _n_threads; i++) {
       tc->do_thread(_threads[i]);
     }
   }
@@ -121,20 +121,20 @@ void ConcurrentG1Refine::threads_do(ThreadClosure *tc) {
 
 void ConcurrentG1Refine::worker_threads_do(ThreadClosure * tc) {
   if (_threads != NULL) {
-    for (int i = 0; i < worker_thread_num(); i++) {
+    for (uint i = 0; i < worker_thread_num(); i++) {
       tc->do_thread(_threads[i]);
     }
   }
 }
 
-int ConcurrentG1Refine::thread_num() {
+uint ConcurrentG1Refine::thread_num() {
-  int n_threads = (G1ConcRefinementThreads > 0) ? G1ConcRefinementThreads
+  uint n_threads = (G1ConcRefinementThreads > 0) ? G1ConcRefinementThreads
                                                 : ParallelGCThreads;
-  return MAX2<int>(n_threads, 1);
+  return MAX2<uint>(n_threads, 1);
 }
 
 void ConcurrentG1Refine::print_worker_threads_on(outputStream* st) const {
-  for (int i = 0; i < _n_threads; ++i) {
+  for (uint i = 0; i < _n_threads; ++i) {
     _threads[i]->print_on(st);
     st->cr();
   }

hotspot/src/share/vm/gc_implementation/g1/concurrentG1Refine.hpp

@@ -39,8 +39,8 @@ class DirtyCardQueue;
 
 class ConcurrentG1Refine: public CHeapObj<mtGC> {
   ConcurrentG1RefineThread** _threads;
-  int _n_threads;
+  uint _n_threads;
-  int _n_worker_threads;
+  uint _n_worker_threads;
  /*
   * The value of the update buffer queue length falls into one of 3 zones:
   * green, yellow, red. If the value is in [0, green) nothing is
@@ -88,7 +88,7 @@ class ConcurrentG1Refine: public CHeapObj<mtGC> {
   // The RS sampling thread
   ConcurrentG1RefineThread * sampling_thread() const;
 
-  static int thread_num();
+  static uint thread_num();
 
   void print_worker_threads_on(outputStream* st) const;
 
@@ -100,8 +100,8 @@ class ConcurrentG1Refine: public CHeapObj<mtGC> {
   int yellow_zone() const     { return _yellow_zone; }
   int red_zone() const        { return _red_zone;    }
 
-  int total_thread_num() const  { return _n_threads;        }
+  uint total_thread_num() const  { return _n_threads;        }
-  int worker_thread_num() const { return _n_worker_threads; }
+  uint worker_thread_num() const { return _n_worker_threads; }
 
   int thread_threshold_step() const { return _thread_threshold_step; }
 

hotspot/src/share/vm/gc_implementation/g1/concurrentG1RefineThread.cpp

@@ -33,7 +33,7 @@
 
 ConcurrentG1RefineThread::
 ConcurrentG1RefineThread(ConcurrentG1Refine* cg1r, ConcurrentG1RefineThread *next,
-                         int worker_id_offset, int worker_id) :
+                         uint worker_id_offset, uint worker_id) :
   ConcurrentGCThread(),
   _worker_id_offset(worker_id_offset),
   _worker_id(worker_id),

hotspot/src/share/vm/gc_implementation/g1/concurrentG1RefineThread.hpp

@@ -38,8 +38,8 @@ class ConcurrentG1RefineThread: public ConcurrentGCThread {
 
   double _vtime_start;  // Initial virtual time.
   double _vtime_accum;  // Initial virtual time.
-  int _worker_id;
+  uint _worker_id;
-  int _worker_id_offset;
+  uint _worker_id_offset;
 
   // The refinement threads collection is linked list. A predecessor can activate a successor
   // when the number of the rset update buffer crosses a certain threshold. A successor
@@ -71,7 +71,7 @@ public:
   virtual void run();
   // Constructor
   ConcurrentG1RefineThread(ConcurrentG1Refine* cg1r, ConcurrentG1RefineThread* next,
-                           int worker_id_offset, int worker_id);
+                           uint worker_id_offset, uint worker_id);
 
   void initialize();
 

hotspot/src/share/vm/gc_implementation/g1/concurrentMark.cpp

@@ -567,8 +567,8 @@ ConcurrentMark::ConcurrentMark(G1CollectedHeap* g1h, ReservedSpace heap_rs) :
   _root_regions.init(_g1h, this);
 
   if (ConcGCThreads > ParallelGCThreads) {
-    warning("Can't have more ConcGCThreads (" UINT32_FORMAT ") "
+    warning("Can't have more ConcGCThreads (" UINTX_FORMAT ") "
-            "than ParallelGCThreads (" UINT32_FORMAT ").",
+            "than ParallelGCThreads (" UINTX_FORMAT ").",
             ConcGCThreads, ParallelGCThreads);
     return;
   }
@@ -1804,7 +1804,6 @@ class G1ParNoteEndTask;
 
 class G1NoteEndOfConcMarkClosure : public HeapRegionClosure {
   G1CollectedHeap* _g1;
-  int _worker_num;
   size_t _max_live_bytes;
   uint _regions_claimed;
   size_t _freed_bytes;
@@ -1817,10 +1816,9 @@ class G1NoteEndOfConcMarkClosure : public HeapRegionClosure {
 
 public:
   G1NoteEndOfConcMarkClosure(G1CollectedHeap* g1,
-                             int worker_num,
                              FreeRegionList* local_cleanup_list,
                              HRRSCleanupTask* hrrs_cleanup_task) :
-    _g1(g1), _worker_num(worker_num),
+    _g1(g1),
     _max_live_bytes(0), _regions_claimed(0),
     _freed_bytes(0),
     _claimed_region_time(0.0), _max_region_time(0.0),
@@ -1893,7 +1891,7 @@ public:
     double start = os::elapsedTime();
     FreeRegionList local_cleanup_list("Local Cleanup List");
     HRRSCleanupTask hrrs_cleanup_task;
-    G1NoteEndOfConcMarkClosure g1_note_end(_g1h, worker_id, &local_cleanup_list,
+    G1NoteEndOfConcMarkClosure g1_note_end(_g1h, &local_cleanup_list,
                                            &hrrs_cleanup_task);
     if (G1CollectedHeap::use_parallel_gc_threads()) {
       _g1h->heap_region_par_iterate_chunked(&g1_note_end, worker_id,
@@ -2145,7 +2143,7 @@ void ConcurrentMark::completeCleanup() {
 
   G1CollectedHeap* g1h = G1CollectedHeap::heap();
 
-  _cleanup_list.verify_list();
+  _cleanup_list.verify_optional();
   FreeRegionList tmp_free_list("Tmp Free List");
 
   if (G1ConcRegionFreeingVerbose) {

hotspot/src/share/vm/gc_implementation/g1/dirtyCardQueue.cpp

@@ -34,12 +34,12 @@
 
 bool DirtyCardQueue::apply_closure(CardTableEntryClosure* cl,
                                    bool consume,
-                                   size_t worker_i) {
+                                   uint worker_i) {
   bool res = true;
   if (_buf != NULL) {
     res = apply_closure_to_buffer(cl, _buf, _index, _sz,
                                   consume,
-                                  (int) worker_i);
+                                  worker_i);
     if (res && consume) _index = _sz;
   }
   return res;
@@ -49,7 +49,7 @@ bool DirtyCardQueue::apply_closure_to_buffer(CardTableEntryClosure* cl,
                                              void** buf,
                                              size_t index, size_t sz,
                                              bool consume,
-                                             int worker_i) {
+                                             uint worker_i) {
   if (cl == NULL) return true;
   for (size_t i = index; i < sz; i += oopSize) {
     int ind = byte_index_to_index((int)i);
@@ -79,8 +79,8 @@ DirtyCardQueueSet::DirtyCardQueueSet(bool notify_when_complete) :
 }
 
 // Determines how many mutator threads can process the buffers in parallel.
-size_t DirtyCardQueueSet::num_par_ids() {
+uint DirtyCardQueueSet::num_par_ids() {
-  return os::processor_count();
+  return (uint)os::processor_count();
 }
 
 void DirtyCardQueueSet::initialize(Monitor* cbl_mon, Mutex* fl_lock,
@@ -103,7 +103,7 @@ void DirtyCardQueueSet::set_closure(CardTableEntryClosure* closure) {
 }
 
 void DirtyCardQueueSet::iterate_closure_all_threads(bool consume,
-                                                    size_t worker_i) {
+                                                    uint worker_i) {
   assert(SafepointSynchronize::is_at_safepoint(), "Must be at safepoint.");
   for(JavaThread* t = Threads::first(); t; t = t->next()) {
     bool b = t->dirty_card_queue().apply_closure(_closure, consume);
@@ -126,11 +126,11 @@ bool DirtyCardQueueSet::mut_process_buffer(void** buf) {
 
   // We get the the number of any par_id that this thread
   // might have already claimed.
-  int worker_i = thread->get_claimed_par_id();
+  uint worker_i = thread->get_claimed_par_id();
 
-  // If worker_i is not -1 then the thread has already claimed
+  // If worker_i is not UINT_MAX then the thread has already claimed
   // a par_id. We make note of it using the already_claimed value
-  if (worker_i != -1) {
+  if (worker_i != UINT_MAX) {
     already_claimed = true;
   } else {
 
@@ -142,7 +142,7 @@ bool DirtyCardQueueSet::mut_process_buffer(void** buf) {
   }
 
   bool b = false;
-  if (worker_i != -1) {
+  if (worker_i != UINT_MAX) {
     b = DirtyCardQueue::apply_closure_to_buffer(_closure, buf, 0,
                                                 _sz, true, worker_i);
     if (b) Atomic::inc(&_processed_buffers_mut);
@@ -154,8 +154,8 @@ bool DirtyCardQueueSet::mut_process_buffer(void** buf) {
       // we release the id
       _free_ids->release_par_id(worker_i);
 
-      // and set the claimed_id in the thread to -1
+      // and set the claimed_id in the thread to UINT_MAX
-      thread->set_claimed_par_id(-1);
+      thread->set_claimed_par_id(UINT_MAX);
     }
   }
   return b;
@@ -186,7 +186,7 @@ DirtyCardQueueSet::get_completed_buffer(int stop_at) {
 
 bool DirtyCardQueueSet::
 apply_closure_to_completed_buffer_helper(CardTableEntryClosure* cl,
-                                         int worker_i,
+                                         uint worker_i,
                                          BufferNode* nd) {
   if (nd != NULL) {
     void **buf = BufferNode::make_buffer_from_node(nd);
@@ -208,7 +208,7 @@ apply_closure_to_completed_buffer_helper(CardTableEntryClosure* cl,
 }
 
 bool DirtyCardQueueSet::apply_closure_to_completed_buffer(CardTableEntryClosure* cl,
-                                                          int worker_i,
+                                                          uint worker_i,
                                                           int stop_at,
                                                           bool during_pause) {
   assert(!during_pause || stop_at == 0, "Should not leave any completed buffers during a pause");
@@ -218,7 +218,7 @@ bool DirtyCardQueueSet::apply_closure_to_completed_buffer(CardTableEntryClosure*
   return res;
 }
 
-bool DirtyCardQueueSet::apply_closure_to_completed_buffer(int worker_i,
+bool DirtyCardQueueSet::apply_closure_to_completed_buffer(uint worker_i,
                                                           int stop_at,
                                                           bool during_pause) {
   return apply_closure_to_completed_buffer(_closure, worker_i,

hotspot/src/share/vm/gc_implementation/g1/dirtyCardQueue.hpp

@@ -36,7 +36,7 @@ class CardTableEntryClosure: public CHeapObj<mtGC> {
 public:
   // Process the card whose card table entry is "card_ptr".  If returns
   // "false", terminate the iteration early.
-  virtual bool do_card_ptr(jbyte* card_ptr, int worker_i = 0) = 0;
+  virtual bool do_card_ptr(jbyte* card_ptr, uint worker_i = 0) = 0;
 };
 
 // A ptrQueue whose elements are "oops", pointers to object heads.
@@ -53,7 +53,7 @@ public:
   // deletes processed entries from logs.
   bool apply_closure(CardTableEntryClosure* cl,
                      bool consume = true,
-                     size_t worker_i = 0);
+                     uint worker_i = 0);
 
   // Apply the closure to all elements of "buf", down to "index"
   // (inclusive.)  If returns "false", then a closure application returned
@@ -63,7 +63,7 @@ public:
   static bool apply_closure_to_buffer(CardTableEntryClosure* cl,
                                       void** buf, size_t index, size_t sz,
                                       bool consume = true,
-                                      int worker_i = 0);
+                                      uint worker_i = 0);
   void **get_buf() { return _buf;}
   void set_buf(void **buf) {_buf = buf;}
   size_t get_index() { return _index;}
@@ -98,7 +98,7 @@ public:
 
   // The number of parallel ids that can be claimed to allow collector or
   // mutator threads to do card-processing work.
-  static size_t num_par_ids();
+  static uint num_par_ids();
 
   static void handle_zero_index_for_thread(JavaThread* t);
 
@@ -115,7 +115,7 @@ public:
   // change in the future.)  If "consume" is true, processed entries are
   // discarded.
   void iterate_closure_all_threads(bool consume = true,
-                                   size_t worker_i = 0);
+                                   uint worker_i = 0);
 
   // If there exists some completed buffer, pop it, then apply the
   // registered closure to all its elements, nulling out those elements
@@ -124,7 +124,7 @@ public:
   // but is only partially completed before a "yield" happens, the
   // partially completed buffer (with its processed elements set to NULL)
   // is returned to the completed buffer set, and this call returns false.
-  bool apply_closure_to_completed_buffer(int worker_i = 0,
+  bool apply_closure_to_completed_buffer(uint worker_i = 0,
                                          int stop_at = 0,
                                          bool during_pause = false);
 
@@ -136,13 +136,13 @@ public:
   // partially completed buffer (with its processed elements set to NULL)
   // is returned to the completed buffer set, and this call returns false.
   bool apply_closure_to_completed_buffer(CardTableEntryClosure* cl,
-                                         int worker_i = 0,
+                                         uint worker_i = 0,
                                          int stop_at = 0,
                                          bool during_pause = false);
 
   // Helper routine for the above.
   bool apply_closure_to_completed_buffer_helper(CardTableEntryClosure* cl,
-                                                int worker_i,
+                                                uint worker_i,
                                                 BufferNode* nd);
 
   BufferNode* get_completed_buffer(int stop_at);

hotspot/src/share/vm/gc_implementation/g1/g1BlockOffsetTable.cpp

@@ -304,26 +304,26 @@ void G1BlockOffsetArray::check_all_cards(size_t start_card, size_t end_card) con
     if (c - start_card > BlockOffsetArray::power_to_cards_back(1)) {
       guarantee(entry > N_words,
                 err_msg("Should be in logarithmic region - "
-                        "entry: " UINT32_FORMAT ", "
+                        "entry: %u, "
-                        "_array->offset_array(c): " UINT32_FORMAT ", "
+                        "_array->offset_array(c): %u, "
-                        "N_words: " UINT32_FORMAT,
+                        "N_words: %u",
-                        entry, _array->offset_array(c), N_words));
+                        (uint)entry, (uint)_array->offset_array(c), (uint)N_words));
     }
     size_t backskip = BlockOffsetArray::entry_to_cards_back(entry);
     size_t landing_card = c - backskip;
     guarantee(landing_card >= (start_card - 1), "Inv");
     if (landing_card >= start_card) {
       guarantee(_array->offset_array(landing_card) <= entry,
-                err_msg("Monotonicity - landing_card offset: " UINT32_FORMAT ", "
+                err_msg("Monotonicity - landing_card offset: %u, "
-                        "entry: " UINT32_FORMAT,
+                        "entry: %u",
-                        _array->offset_array(landing_card), entry));
+                        (uint)_array->offset_array(landing_card), (uint)entry));
     } else {
       guarantee(landing_card == start_card - 1, "Tautology");
       // Note that N_words is the maximum offset value
       guarantee(_array->offset_array(landing_card) <= N_words,
-                err_msg("landing card offset: " UINT32_FORMAT ", "
+                err_msg("landing card offset: %u, "
-                        "N_words: " UINT32_FORMAT,
+                        "N_words: %u",
-                        _array->offset_array(landing_card), N_words));
+                        (uint)_array->offset_array(landing_card), (uint)N_words));
     }
   }
 }
@@ -554,21 +554,20 @@ void G1BlockOffsetArray::alloc_block_work2(HeapWord** threshold_, size_t* index_
           (_array->offset_array(orig_index) > 0 &&
          _array->offset_array(orig_index) <= N_words),
          err_msg("offset array should have been set - "
-                  "orig_index offset: " UINT32_FORMAT ", "
+                  "orig_index offset: %u, "
                   "blk_start: " PTR_FORMAT ", "
                   "boundary: " PTR_FORMAT,
-                  _array->offset_array(orig_index),
+                  (uint)_array->offset_array(orig_index),
                   blk_start, boundary));
   for (size_t j = orig_index + 1; j <= end_index; j++) {
     assert(_array->offset_array(j) > 0 &&
            _array->offset_array(j) <=
              (u_char) (N_words+BlockOffsetArray::N_powers-1),
            err_msg("offset array should have been set - "
-                   UINT32_FORMAT " not > 0 OR "
+                   "%u not > 0 OR %u not <= %u",
-                   UINT32_FORMAT " not <= " UINT32_FORMAT,
+                   (uint) _array->offset_array(j),
-                   _array->offset_array(j),
+                   (uint) _array->offset_array(j),
-                   _array->offset_array(j),
+                   (uint) (N_words+BlockOffsetArray::N_powers-1)));
-                   (u_char) (N_words+BlockOffsetArray::N_powers-1)));
   }
 #endif
 }

hotspot/src/share/vm/gc_implementation/g1/g1BlockOffsetTable.hpp

@@ -146,8 +146,8 @@ private:
   void check_offset(size_t offset, const char* msg) const {
     assert(offset <= N_words,
            err_msg("%s - "
-                   "offset: " UINT32_FORMAT", N_words: " UINT32_FORMAT,
+                   "offset: " SIZE_FORMAT", N_words: %u",
-                   msg, offset, N_words));
+                   msg, offset, (uint)N_words));
   }
 
   // Bounds checking accessors:

hotspot/src/share/vm/gc_implementation/g1/g1CollectedHeap.cpp

@@ -102,7 +102,7 @@ public:
                               ConcurrentG1Refine* cg1r) :
     _sts(sts), _g1rs(g1rs), _cg1r(cg1r), _concurrent(true)
   {}
-  bool do_card_ptr(jbyte* card_ptr, int worker_i) {
+  bool do_card_ptr(jbyte* card_ptr, uint worker_i) {
     bool oops_into_cset = _g1rs->refine_card(card_ptr, worker_i, false);
     // This path is executed by the concurrent refine or mutator threads,
     // concurrently, and so we do not care if card_ptr contains references
@@ -131,7 +131,7 @@ public:
   {
     for (int i = 0; i < 256; i++) _histo[i] = 0;
   }
-  bool do_card_ptr(jbyte* card_ptr, int worker_i) {
+  bool do_card_ptr(jbyte* card_ptr, uint worker_i) {
     if (_g1h->is_in_reserved(_ctbs->addr_for(card_ptr))) {
       _calls++;
       unsigned char* ujb = (unsigned char*)card_ptr;
@@ -160,7 +160,7 @@ public:
   RedirtyLoggedCardTableEntryClosure() :
     _calls(0), _g1h(G1CollectedHeap::heap()), _ctbs(_g1h->g1_barrier_set()) {}
 
-  bool do_card_ptr(jbyte* card_ptr, int worker_i) {
+  bool do_card_ptr(jbyte* card_ptr, uint worker_i) {
     if (_g1h->is_in_reserved(_ctbs->addr_for(card_ptr))) {
       _calls++;
       *card_ptr = 0;
@@ -1288,7 +1288,7 @@ bool G1CollectedHeap::do_collection(bool explicit_gc,
   print_heap_before_gc();
   trace_heap_before_gc(gc_tracer);
 
-  size_t metadata_prev_used = MetaspaceAux::allocated_used_bytes();
+  size_t metadata_prev_used = MetaspaceAux::used_bytes();
 
   verify_region_sets_optional();
 
@@ -2314,7 +2314,7 @@ void G1CollectedHeap::check_gc_time_stamps() {
 void G1CollectedHeap::iterate_dirty_card_closure(CardTableEntryClosure* cl,
                                                  DirtyCardQueue* into_cset_dcq,
                                                  bool concurrent,
-                                                 int worker_i) {
+                                                 uint worker_i) {
   // Clean cards in the hot card cache
   G1HotCardCache* hot_card_cache = _cg1r->hot_card_cache();
   hot_card_cache->drain(worker_i, g1_rem_set(), into_cset_dcq);
@@ -2843,7 +2843,7 @@ void G1CollectedHeap::clear_cset_start_regions() {
 
 // Given the id of a worker, obtain or calculate a suitable
 // starting region for iterating over the current collection set.
-HeapRegion* G1CollectedHeap::start_cset_region_for_worker(int worker_i) {
+HeapRegion* G1CollectedHeap::start_cset_region_for_worker(uint worker_i) {
   assert(get_gc_time_stamp() > 0, "should have been updated by now");
 
   HeapRegion* result = NULL;
@@ -5103,7 +5103,7 @@ g1_process_strong_roots(bool is_scavenging,
                         OopClosure* scan_non_heap_roots,
                         OopsInHeapRegionClosure* scan_rs,
                         G1KlassScanClosure* scan_klasses,
-                        int worker_i) {
+                        uint worker_i) {
 
   // First scan the strong roots
   double ext_roots_start = os::elapsedTime();
@@ -5207,10 +5207,10 @@ public:
 
   ~G1StringSymbolTableUnlinkTask() {
     guarantee(!_process_strings || !_do_in_parallel || StringTable::parallel_claimed_index() >= _initial_string_table_size,
-              err_msg("claim value "INT32_FORMAT" after unlink less than initial string table size "INT32_FORMAT,
+              err_msg("claim value %d after unlink less than initial string table size %d",
                       StringTable::parallel_claimed_index(), _initial_string_table_size));
     guarantee(!_process_symbols || !_do_in_parallel || SymbolTable::parallel_claimed_index() >= _initial_symbol_table_size,
-              err_msg("claim value "INT32_FORMAT" after unlink less than initial symbol table size "INT32_FORMAT,
+              err_msg("claim value %d after unlink less than initial symbol table size %d",
                       SymbolTable::parallel_claimed_index(), _initial_symbol_table_size));
   }
 
@@ -5275,7 +5275,7 @@ void G1CollectedHeap::unlink_string_and_symbol_table(BoolObjectClosure* is_alive
 
 class RedirtyLoggedCardTableEntryFastClosure : public CardTableEntryClosure {
 public:
-  bool do_card_ptr(jbyte* card_ptr, int worker_i) {
+  bool do_card_ptr(jbyte* card_ptr, uint worker_i) {
     *card_ptr = CardTableModRefBS::dirty_card_val();
     return true;
   }

hotspot/src/share/vm/gc_implementation/g1/g1CollectedHeap.hpp

@@ -845,7 +845,7 @@ protected:
                                OopClosure* scan_non_heap_roots,
                                OopsInHeapRegionClosure* scan_rs,
                                G1KlassScanClosure* scan_klasses,
-                               int worker_i);
+                               uint worker_i);
 
   // Notifies all the necessary spaces that the committed space has
   // been updated (either expanded or shrunk). It should be called
@@ -1139,7 +1139,7 @@ public:
 
   void iterate_dirty_card_closure(CardTableEntryClosure* cl,
                                   DirtyCardQueue* into_cset_dcq,
-                                  bool concurrent, int worker_i);
+                                  bool concurrent, uint worker_i);
 
   // The shared block offset table array.
   G1BlockOffsetSharedArray* bot_shared() const { return _bot_shared; }
@@ -1370,7 +1370,7 @@ public:
 
   // Given the id of a worker, obtain or calculate a suitable
   // starting region for iterating over the current collection set.
-  HeapRegion* start_cset_region_for_worker(int worker_i);
+  HeapRegion* start_cset_region_for_worker(uint worker_i);
 
   // This is a convenience method that is used by the
   // HeapRegionIterator classes to calculate the starting region for

hotspot/src/share/vm/gc_implementation/g1/g1CollectorPolicy.cpp

@@ -1204,7 +1204,7 @@ void G1CollectorPolicy::record_heap_size_info_at_start(bool full) {
          (_young_list_target_length * HeapRegion::GrainBytes) - _survivor_used_bytes_before_gc;
 
   if (full) {
-    _metaspace_used_bytes_before_gc = MetaspaceAux::allocated_used_bytes();
+    _metaspace_used_bytes_before_gc = MetaspaceAux::used_bytes();
   }
 }
 

hotspot/src/share/vm/gc_implementation/g1/g1GCPhaseTimes.cpp

@@ -147,7 +147,7 @@ template <class T>
 void WorkerDataArray<T>::verify() {
   for (uint i = 0; i < _length; i++) {
     assert(_data[i] != _uninitialized,
-        err_msg("Invalid data for worker " UINT32_FORMAT ", data: %lf, uninitialized: %lf",
+        err_msg("Invalid data for worker %u, data: %lf, uninitialized: %lf",
             i, (double)_data[i], (double)_uninitialized));
   }
 }
@@ -246,8 +246,8 @@ void G1GCPhaseTimes::print_stats(int level, const char* str, double value) {
   LineBuffer(level).append_and_print_cr("[%s: %.1lf ms]", str, value);
 }
 
-void G1GCPhaseTimes::print_stats(int level, const char* str, double value, int workers) {
+void G1GCPhaseTimes::print_stats(int level, const char* str, double value, uint workers) {
-  LineBuffer(level).append_and_print_cr("[%s: %.1lf ms, GC Workers: %d]", str, value, workers);
+  LineBuffer(level).append_and_print_cr("[%s: %.1lf ms, GC Workers: %u]", str, value, workers);
 }
 
 double G1GCPhaseTimes::accounted_time_ms() {

hotspot/src/share/vm/gc_implementation/g1/g1GCPhaseTimes.hpp

@@ -161,7 +161,7 @@ class G1GCPhaseTimes : public CHeapObj<mtGC> {
 
   // Helper methods for detailed logging
   void print_stats(int level, const char* str, double value);
-  void print_stats(int level, const char* str, double value, int workers);
+  void print_stats(int level, const char* str, double value, uint workers);
 
  public:
   G1GCPhaseTimes(uint max_gc_threads);

hotspot/src/share/vm/gc_implementation/g1/g1HotCardCache.cpp

@@ -44,9 +44,9 @@ void G1HotCardCache::initialize() {
     _hot_cache_idx = 0;
 
     // For refining the cards in the hot cache in parallel
-    int n_workers = (ParallelGCThreads > 0 ?
+    uint n_workers = (ParallelGCThreads > 0 ?
                         _g1h->workers()->total_workers() : 1);
-    _hot_cache_par_chunk_size = MAX2(1, _hot_cache_size / n_workers);
+    _hot_cache_par_chunk_size = MAX2(1, _hot_cache_size / (int)n_workers);
     _hot_cache_par_claimed_idx = 0;
 
     _card_counts.initialize();
@@ -89,7 +89,7 @@ jbyte* G1HotCardCache::insert(jbyte* card_ptr) {
   return res;
 }
 
-void G1HotCardCache::drain(int worker_i,
+void G1HotCardCache::drain(uint worker_i,
                            G1RemSet* g1rs,
                            DirtyCardQueue* into_cset_dcq) {
   if (!default_use_cache()) {
@@ -122,8 +122,8 @@ void G1HotCardCache::drain(int worker_i,
             // RSet updating while within an evacuation pause.
             // In this case worker_i should be the id of a GC worker thread
             assert(SafepointSynchronize::is_at_safepoint(), "Should be at a safepoint");
-            assert(worker_i < (int) (ParallelGCThreads == 0 ? 1 : ParallelGCThreads),
+            assert(worker_i < (ParallelGCThreads == 0 ? 1 : ParallelGCThreads),
-                   err_msg("incorrect worker id: "INT32_FORMAT, worker_i));
+                   err_msg("incorrect worker id: %u", worker_i));
 
             into_cset_dcq->enqueue(card_ptr);
           }

hotspot/src/share/vm/gc_implementation/g1/g1HotCardCache.hpp

@@ -99,7 +99,7 @@ class G1HotCardCache: public CHeapObj<mtGC> {
 
   // Refine the cards that have delayed as a result of
   // being in the cache.
-  void drain(int worker_i, G1RemSet* g1rs, DirtyCardQueue* into_cset_dcq);
+  void drain(uint worker_i, G1RemSet* g1rs, DirtyCardQueue* into_cset_dcq);
 
   // Set up for parallel processing of the cards in the hot cache
   void reset_hot_cache_claimed_index() {

hotspot/src/share/vm/gc_implementation/g1/g1OopClosures.hpp

@@ -234,14 +234,14 @@ class G1UpdateRSOrPushRefOopClosure: public ExtendedOopClosure {
   HeapRegion* _from;
   OopsInHeapRegionClosure* _push_ref_cl;
   bool _record_refs_into_cset;
-  int _worker_i;
+  uint _worker_i;
 
 public:
   G1UpdateRSOrPushRefOopClosure(G1CollectedHeap* g1h,
                                 G1RemSet* rs,
                                 OopsInHeapRegionClosure* push_ref_cl,
                                 bool record_refs_into_cset,
-                                int worker_i = 0);
+                                uint worker_i = 0);
 
   void set_from(HeapRegion* from) {
     assert(from != NULL, "from region must be non-NULL");

hotspot/src/share/vm/gc_implementation/g1/g1RemSet.cpp

@@ -113,14 +113,14 @@ class ScanRSClosure : public HeapRegionClosure {
   G1SATBCardTableModRefBS *_ct_bs;
 
   double _strong_code_root_scan_time_sec;
-  int    _worker_i;
+  uint   _worker_i;
   int    _block_size;
   bool   _try_claimed;
 
 public:
   ScanRSClosure(OopsInHeapRegionClosure* oc,
                 CodeBlobToOopClosure* code_root_cl,
-                int worker_i) :
+                uint worker_i) :
     _oc(oc),
     _code_root_cl(code_root_cl),
     _strong_code_root_scan_time_sec(0.0),
@@ -162,7 +162,7 @@ public:
 
   void printCard(HeapRegion* card_region, size_t card_index,
                  HeapWord* card_start) {
-    gclog_or_tty->print_cr("T %d Region [" PTR_FORMAT ", " PTR_FORMAT ") "
+    gclog_or_tty->print_cr("T %u Region [" PTR_FORMAT ", " PTR_FORMAT ") "
                            "RS names card %p: "
                            "[" PTR_FORMAT ", " PTR_FORMAT ")",
                            _worker_i,
@@ -241,7 +241,7 @@ public:
 
 void G1RemSet::scanRS(OopsInHeapRegionClosure* oc,
                       CodeBlobToOopClosure* code_root_cl,
-                      int worker_i) {
+                      uint worker_i) {
   double rs_time_start = os::elapsedTime();
   HeapRegion *startRegion = _g1->start_cset_region_for_worker(worker_i);
 
@@ -274,13 +274,13 @@ public:
                                               DirtyCardQueue* into_cset_dcq) :
     _g1rs(g1h->g1_rem_set()), _into_cset_dcq(into_cset_dcq)
   {}
-  bool do_card_ptr(jbyte* card_ptr, int worker_i) {
+  bool do_card_ptr(jbyte* card_ptr, uint worker_i) {
     // The only time we care about recording cards that
     // contain references that point into the collection set
     // is during RSet updating within an evacuation pause.
     // In this case worker_i should be the id of a GC worker thread.
     assert(SafepointSynchronize::is_at_safepoint(), "not during an evacuation pause");
-    assert(worker_i < (int) (ParallelGCThreads == 0 ? 1 : ParallelGCThreads), "should be a GC worker");
+    assert(worker_i < (ParallelGCThreads == 0 ? 1 : ParallelGCThreads), "should be a GC worker");
 
     if (_g1rs->refine_card(card_ptr, worker_i, true)) {
       // 'card_ptr' contains references that point into the collection
@@ -295,7 +295,7 @@ public:
   }
 };
 
-void G1RemSet::updateRS(DirtyCardQueue* into_cset_dcq, int worker_i) {
+void G1RemSet::updateRS(DirtyCardQueue* into_cset_dcq, uint worker_i) {
   double start = os::elapsedTime();
   // Apply the given closure to all remaining log entries.
   RefineRecordRefsIntoCSCardTableEntryClosure into_cset_update_rs_cl(_g1, into_cset_dcq);
@@ -320,14 +320,14 @@ void G1RemSet::cleanupHRRS() {
 
 void G1RemSet::oops_into_collection_set_do(OopsInHeapRegionClosure* oc,
                                            CodeBlobToOopClosure* code_root_cl,
-                                           int worker_i) {
+                                           uint worker_i) {
 #if CARD_REPEAT_HISTO
   ct_freq_update_histo_and_reset();
 #endif
 
   // We cache the value of 'oc' closure into the appropriate slot in the
   // _cset_rs_update_cl for this worker
-  assert(worker_i < (int)n_workers(), "sanity");
+  assert(worker_i < n_workers(), "sanity");
   _cset_rs_update_cl[worker_i] = oc;
 
   // A DirtyCardQueue that is used to hold cards containing references
@@ -399,7 +399,7 @@ public:
     _g1(g1), _ct_bs(bs)
   { }
 
-  bool do_card_ptr(jbyte* card_ptr, int worker_i) {
+  bool do_card_ptr(jbyte* card_ptr, uint worker_i) {
     // Construct the region representing the card.
     HeapWord* start = _ct_bs->addr_for(card_ptr);
     // And find the region containing it.
@@ -543,7 +543,7 @@ G1UpdateRSOrPushRefOopClosure(G1CollectedHeap* g1h,
                               G1RemSet* rs,
                               OopsInHeapRegionClosure* push_ref_cl,
                               bool record_refs_into_cset,
-                              int worker_i) :
+                              uint worker_i) :
   _g1(g1h), _g1_rem_set(rs), _from(NULL),
   _record_refs_into_cset(record_refs_into_cset),
   _push_ref_cl(push_ref_cl), _worker_i(worker_i) { }
@@ -552,7 +552,7 @@ G1UpdateRSOrPushRefOopClosure(G1CollectedHeap* g1h,
 // into the collection set, if we're checking for such references;
 // false otherwise.
 
-bool G1RemSet::refine_card(jbyte* card_ptr, int worker_i,
+bool G1RemSet::refine_card(jbyte* card_ptr, uint worker_i,
                            bool check_for_refs_into_cset) {
 
   // If the card is no longer dirty, nothing to do.

hotspot/src/share/vm/gc_implementation/g1/g1RemSet.hpp

@@ -97,7 +97,7 @@ public:
   // In the sequential case this param will be ignored.
   void oops_into_collection_set_do(OopsInHeapRegionClosure* blk,
                                    CodeBlobToOopClosure* code_root_cl,
-                                   int worker_i);
+                                   uint worker_i);
 
   // Prepare for and cleanup after an oops_into_collection_set_do
   // call.  Must call each of these once before and after (in sequential
@@ -109,9 +109,9 @@ public:
 
   void scanRS(OopsInHeapRegionClosure* oc,
               CodeBlobToOopClosure* code_root_cl,
-              int worker_i);
+              uint worker_i);
 
-  void updateRS(DirtyCardQueue* into_cset_dcq, int worker_i);
+  void updateRS(DirtyCardQueue* into_cset_dcq, uint worker_i);
 
   CardTableModRefBS* ct_bs() { return _ct_bs; }
   size_t cardsScanned() { return _total_cards_scanned; }
@@ -138,7 +138,7 @@ public:
   // if the given card contains oops that have references into the
   // current collection set.
   virtual bool refine_card(jbyte* card_ptr,
-                           int worker_i,
+                           uint worker_i,
                            bool check_for_refs_into_cset);
 
   // Print accumulated summary info from the start of the VM.
@@ -171,12 +171,12 @@ public:
 class UpdateRSOopClosure: public ExtendedOopClosure {
   HeapRegion* _from;
   G1RemSet* _rs;
-  int _worker_i;
+  uint _worker_i;
 
   template <class T> void do_oop_work(T* p);
 
 public:
-  UpdateRSOopClosure(G1RemSet* rs, int worker_i = 0) :
+  UpdateRSOopClosure(G1RemSet* rs, uint worker_i = 0) :
     _from(NULL), _rs(rs), _worker_i(worker_i)
   {}
 

hotspot/src/share/vm/gc_implementation/g1/heapRegionRemSet.cpp

@@ -390,7 +390,7 @@ void FromCardCache::shrink(uint new_num_regions) {
 void FromCardCache::print(outputStream* out) {
   for (uint i = 0; i < HeapRegionRemSet::num_par_rem_sets(); i++) {
     for (uint j = 0; j < _max_regions; j++) {
-      out->print_cr("_from_card_cache["UINT32_FORMAT"]["UINT32_FORMAT"] = "INT32_FORMAT".",
+      out->print_cr("_from_card_cache[%u][%u] = %d.",
                     i, j, at(i, j));
     }
   }
@@ -430,7 +430,7 @@ void OtherRegionsTable::add_reference(OopOrNarrowOopStar from, int tid) {
   int from_card = (int)(uintptr_t(from) >> CardTableModRefBS::card_shift);
 
   if (G1TraceHeapRegionRememberedSet) {
-    gclog_or_tty->print_cr("Table for [" PTR_FORMAT "...): card %d (cache = "INT32_FORMAT")",
+    gclog_or_tty->print_cr("Table for [" PTR_FORMAT "...): card %d (cache = %d)",
                   hr()->bottom(), from_card,
                   FromCardCache::at((uint)tid, cur_hrs_ind));
   }
@@ -853,13 +853,13 @@ OtherRegionsTable::do_cleanup_work(HRRSCleanupTask* hrrs_cleanup_task) {
 // This can be done by either mutator threads together with the
 // concurrent refinement threads or GC threads.
 uint HeapRegionRemSet::num_par_rem_sets() {
-  return (uint)MAX2(DirtyCardQueueSet::num_par_ids() + ConcurrentG1Refine::thread_num(), ParallelGCThreads);
+  return MAX2(DirtyCardQueueSet::num_par_ids() + ConcurrentG1Refine::thread_num(), (uint)ParallelGCThreads);
 }
 
 HeapRegionRemSet::HeapRegionRemSet(G1BlockOffsetSharedArray* bosa,
                                    HeapRegion* hr)
   : _bosa(bosa),
-    _m(Mutex::leaf, FormatBuffer<128>("HeapRegionRemSet lock #"UINT32_FORMAT, hr->hrs_index()), true),
+    _m(Mutex::leaf, FormatBuffer<128>("HeapRegionRemSet lock #%u", hr->hrs_index()), true),
     _code_roots(), _other_regions(hr, &_m) {
   reset_for_par_iteration();
 }

hotspot/src/share/vm/gc_implementation/g1/heapRegionSet.inline.hpp

@@ -30,7 +30,7 @@
 inline void HeapRegionSetBase::add(HeapRegion* hr) {
   check_mt_safety();
   assert(hr->containing_set() == NULL, hrs_ext_msg(this, "should not already have a containing set %u"));
-  assert(hr->next() == NULL, hrs_ext_msg(this, "should not already be linked"));
+  assert(hr->next() == NULL && hr->prev() == NULL, hrs_ext_msg(this, "should not already be linked"));
 
   _count.increment(1u, hr->capacity());
   hr->set_containing_set(this);
@@ -40,7 +40,7 @@ inline void HeapRegionSetBase::add(HeapRegion* hr) {
 inline void HeapRegionSetBase::remove(HeapRegion* hr) {
   check_mt_safety();
   verify_region(hr);
-  assert(hr->next() == NULL, hrs_ext_msg(this, "should already be unlinked"));
+  assert(hr->next() == NULL && hr->prev() == NULL, hrs_ext_msg(this, "should already be unlinked"));
 
   hr->set_containing_set(NULL);
   assert(_count.length() > 0, hrs_ext_msg(this, "pre-condition"));

hotspot/src/share/vm/gc_implementation/g1/satbQueue.cpp

@@ -290,7 +290,7 @@ void SATBMarkQueueSet::iterate_closure_all_threads() {
   shared_satb_queue()->apply_closure_and_empty(_closure);
 }
 
-void SATBMarkQueueSet::par_iterate_closure_all_threads(int worker) {
+void SATBMarkQueueSet::par_iterate_closure_all_threads(uint worker) {
   SharedHeap* sh = SharedHeap::heap();
   int parity = sh->strong_roots_parity();
 
@@ -315,7 +315,7 @@ void SATBMarkQueueSet::par_iterate_closure_all_threads(int worker) {
 }
 
 bool SATBMarkQueueSet::apply_closure_to_completed_buffer_work(bool par,
-                                                              int worker) {
+                                                              uint worker) {
   BufferNode* nd = NULL;
   {
     MutexLockerEx x(_cbl_mon, Mutex::_no_safepoint_check_flag);

hotspot/src/share/vm/gc_implementation/g1/satbQueue.hpp

@@ -84,7 +84,7 @@ class SATBMarkQueueSet: public PtrQueueSet {
   // Utility function to support sequential and parallel versions.  If
   // "par" is true, then "worker" is the par thread id; if "false", worker
   // is ignored.
-  bool apply_closure_to_completed_buffer_work(bool par, int worker);
+  bool apply_closure_to_completed_buffer_work(bool par, uint worker);
 
 #ifdef ASSERT
   void dump_active_states(bool expected_active);
@@ -124,7 +124,7 @@ public:
   // be called serially and at a safepoint.
   void iterate_closure_all_threads();
   // Parallel version of the above.
-  void par_iterate_closure_all_threads(int worker);
+  void par_iterate_closure_all_threads(uint worker);
 
   // If there exists some completed buffer, pop it, then apply the
   // registered closure to all its elements, and return true.  If no
@@ -133,7 +133,7 @@ public:
     return apply_closure_to_completed_buffer_work(false, 0);
   }
   // Parallel version of the above.
-  bool par_apply_closure_to_completed_buffer(int worker) {
+  bool par_apply_closure_to_completed_buffer(uint worker) {
     return apply_closure_to_completed_buffer_work(true, worker);
   }
 

hotspot/src/share/vm/gc_implementation/parallelScavenge/psMarkSweep.cpp

@@ -184,7 +184,7 @@ bool PSMarkSweep::invoke_no_policy(bool clear_all_softrefs) {
     size_t prev_used = heap->used();
 
     // Capture metadata size before collection for sizing.
-    size_t metadata_prev_used = MetaspaceAux::allocated_used_bytes();
+    size_t metadata_prev_used = MetaspaceAux::used_bytes();
 
     // For PrintGCDetails
     size_t old_gen_prev_used = old_gen->used_in_bytes();

hotspot/src/share/vm/gc_implementation/parallelScavenge/psParallelCompact.cpp

@@ -928,7 +928,7 @@ public:
     _heap_used      = heap->used();
     _young_gen_used = heap->young_gen()->used_in_bytes();
     _old_gen_used   = heap->old_gen()->used_in_bytes();
-    _metadata_used  = MetaspaceAux::allocated_used_bytes();
+    _metadata_used  = MetaspaceAux::used_bytes();
   };
 
   size_t heap_used() const      { return _heap_used; }

hotspot/src/share/vm/gc_implementation/shared/gcHeapSummary.hpp

@@ -62,16 +62,16 @@ public:
 };
 
 class MetaspaceSizes : public StackObj {
-  size_t _capacity;
+  size_t _committed;
   size_t _used;
   size_t _reserved;
 
  public:
-  MetaspaceSizes() : _capacity(0), _used(0), _reserved(0) {}
+  MetaspaceSizes() : _committed(0), _used(0), _reserved(0) {}
-  MetaspaceSizes(size_t capacity, size_t used, size_t reserved) :
+  MetaspaceSizes(size_t committed, size_t used, size_t reserved) :
-    _capacity(capacity), _used(used), _reserved(reserved) {}
+    _committed(committed), _used(used), _reserved(reserved) {}
 
-  size_t capacity() const { return _capacity; }
+  size_t committed() const { return _committed; }
   size_t used() const { return _used; }
   size_t reserved() const { return _reserved; }
 };

hotspot/src/share/vm/gc_implementation/shared/gcTraceSend.cpp

@@ -258,7 +258,7 @@ void GCTracer::send_gc_heap_summary_event(GCWhen::Type when, const GCHeapSummary
 static TraceStructMetaspaceSizes to_trace_struct(const MetaspaceSizes& sizes) {
   TraceStructMetaspaceSizes meta_sizes;
 
-  meta_sizes.set_capacity(sizes.capacity());
+  meta_sizes.set_committed(sizes.committed());
   meta_sizes.set_used(sizes.used());
   meta_sizes.set_reserved(sizes.reserved());
 

hotspot/src/share/vm/gc_interface/collectedHeap.cpp

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2001, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2001, 2014, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -85,16 +85,16 @@ GCHeapSummary CollectedHeap::create_heap_summary() {
 
 MetaspaceSummary CollectedHeap::create_metaspace_summary() {
   const MetaspaceSizes meta_space(
-      MetaspaceAux::allocated_capacity_bytes(),
+      MetaspaceAux::committed_bytes(),
-      MetaspaceAux::allocated_used_bytes(),
+      MetaspaceAux::used_bytes(),
       MetaspaceAux::reserved_bytes());
   const MetaspaceSizes data_space(
-      MetaspaceAux::allocated_capacity_bytes(Metaspace::NonClassType),
+      MetaspaceAux::committed_bytes(Metaspace::NonClassType),
-      MetaspaceAux::allocated_used_bytes(Metaspace::NonClassType),
+      MetaspaceAux::used_bytes(Metaspace::NonClassType),
       MetaspaceAux::reserved_bytes(Metaspace::NonClassType));
   const MetaspaceSizes class_space(
-      MetaspaceAux::allocated_capacity_bytes(Metaspace::ClassType),
+      MetaspaceAux::committed_bytes(Metaspace::ClassType),
-      MetaspaceAux::allocated_used_bytes(Metaspace::ClassType),
+      MetaspaceAux::used_bytes(Metaspace::ClassType),
       MetaspaceAux::reserved_bytes(Metaspace::ClassType));
 
   const MetaspaceChunkFreeListSummary& ms_chunk_free_list_summary =
@@ -582,36 +582,6 @@ void CollectedHeap::post_full_gc_dump(GCTimer* timer) {
   }
 }
 
-oop CollectedHeap::Class_obj_allocate(KlassHandle klass, int size, KlassHandle real_klass, TRAPS) {
-  debug_only(check_for_valid_allocation_state());
-  assert(!Universe::heap()->is_gc_active(), "Allocation during gc not allowed");
-  assert(size >= 0, "int won't convert to size_t");
-  HeapWord* obj;
-    assert(ScavengeRootsInCode > 0, "must be");
-    obj = common_mem_allocate_init(real_klass, size, CHECK_NULL);
-  post_allocation_setup_common(klass, obj);
-  assert(Universe::is_bootstrapping() ||
-         !((oop)obj)->is_array(), "must not be an array");
-  NOT_PRODUCT(Universe::heap()->check_for_bad_heap_word_value(obj, size));
-  oop mirror = (oop)obj;
-
-  java_lang_Class::set_oop_size(mirror, size);
-
-  // Setup indirections
-  if (!real_klass.is_null()) {
-    java_lang_Class::set_klass(mirror, real_klass());
-    real_klass->set_java_mirror(mirror);
-  }
-
-  InstanceMirrorKlass* mk = InstanceMirrorKlass::cast(mirror->klass());
-  assert(size == mk->instance_size(real_klass), "should have been set");
-
-  // notify jvmti and dtrace
-  post_allocation_notify(klass, (oop)obj);
-
-  return mirror;
-}
-
 /////////////// Unit tests ///////////////
 
 #ifndef PRODUCT

hotspot/src/share/vm/gc_interface/collectedHeap.hpp

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2001, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2001, 2014, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -312,9 +312,6 @@ class CollectedHeap : public CHeapObj<mtInternal> {
   // May be overridden to set additional parallelism.
   virtual void set_par_threads(uint t) { _n_par_threads = t; };
 
-  // Allocate and initialize instances of Class
-  static oop Class_obj_allocate(KlassHandle klass, int size, KlassHandle real_klass, TRAPS);
-
   // General obj/array allocation facilities.
   inline static oop obj_allocate(KlassHandle klass, int size, TRAPS);
   inline static oop array_allocate(KlassHandle klass, int size, int length, TRAPS);

hotspot/src/share/vm/memory/collectorPolicy.cpp

@@ -257,6 +257,12 @@ void GenCollectorPolicy::assert_size_info() {
   assert(_min_gen0_size % _gen_alignment == 0, "_min_gen0_size alignment");
   assert(_initial_gen0_size % _gen_alignment == 0, "_initial_gen0_size alignment");
   assert(_max_gen0_size % _gen_alignment == 0, "_max_gen0_size alignment");
+  assert(_min_gen0_size <= bound_minus_alignment(_min_gen0_size, _min_heap_byte_size),
+      "Ergonomics made minimum young generation larger than minimum heap");
+  assert(_initial_gen0_size <=  bound_minus_alignment(_initial_gen0_size, _initial_heap_byte_size),
+      "Ergonomics made initial young generation larger than initial heap");
+  assert(_max_gen0_size <= bound_minus_alignment(_max_gen0_size, _max_heap_byte_size),
+      "Ergonomics made maximum young generation lager than maximum heap");
 }
 
 void TwoGenerationCollectorPolicy::assert_size_info() {
@@ -267,6 +273,9 @@ void TwoGenerationCollectorPolicy::assert_size_info() {
   assert(_max_gen1_size % _gen_alignment == 0, "_max_gen1_size alignment");
   assert(_initial_gen1_size % _gen_alignment == 0, "_initial_gen1_size alignment");
   assert(_max_heap_byte_size <= (_max_gen0_size + _max_gen1_size), "Total maximum heap sizes must be sum of generation maximum sizes");
+  assert(_min_gen0_size + _min_gen1_size <= _min_heap_byte_size, "Minimum generation sizes exceed minimum heap size");
+  assert(_initial_gen0_size + _initial_gen1_size == _initial_heap_byte_size, "Initial generation sizes should match initial heap size");
+  assert(_max_gen0_size + _max_gen1_size == _max_heap_byte_size, "Maximum generation sizes should match maximum heap size");
 }
 #endif // ASSERT
 
@@ -303,20 +312,26 @@ void GenCollectorPolicy::initialize_flags() {
     }
   }
 
+  // Make sure NewSize allows an old generation to fit even if set on the command line
+  if (FLAG_IS_CMDLINE(NewSize) && NewSize >= _initial_heap_byte_size) {
+    warning("NewSize was set larger than initial heap size, will use initial heap size.");
+    NewSize = bound_minus_alignment(NewSize, _initial_heap_byte_size);
+  }
+
   // Now take the actual NewSize into account. We will silently increase NewSize
   // if the user specified a smaller or unaligned value.
-  smallest_new_size = MAX2(smallest_new_size, (uintx)align_size_down(NewSize, _gen_alignment));
+  uintx bounded_new_size = bound_minus_alignment(NewSize, MaxHeapSize);
-  if (smallest_new_size != NewSize) {
+  bounded_new_size = MAX2(smallest_new_size, (uintx)align_size_down(bounded_new_size, _gen_alignment));
+  if (bounded_new_size != NewSize) {
     // Do not use FLAG_SET_ERGO to update NewSize here, since this will override
     // if NewSize was set on the command line or not. This information is needed
     // later when setting the initial and minimum young generation size.
-    NewSize = smallest_new_size;
+    NewSize = bounded_new_size;
   }
+  _min_gen0_size = smallest_new_size;
   _initial_gen0_size = NewSize;
 
   if (!FLAG_IS_DEFAULT(MaxNewSize)) {
-    uintx min_new_size = MAX2(_gen_alignment, _min_gen0_size);
-
     if (MaxNewSize >= MaxHeapSize) {
       // Make sure there is room for an old generation
       uintx smaller_max_new_size = MaxHeapSize - _gen_alignment;
@@ -330,8 +345,8 @@ void GenCollectorPolicy::initialize_flags() {
         FLAG_SET_ERGO(uintx, NewSize, MaxNewSize);
         _initial_gen0_size = NewSize;
       }
-    } else if (MaxNewSize < min_new_size) {
+    } else if (MaxNewSize < _initial_gen0_size) {
-      FLAG_SET_ERGO(uintx, MaxNewSize, min_new_size);
+      FLAG_SET_ERGO(uintx, MaxNewSize, _initial_gen0_size);
     } else if (!is_size_aligned(MaxNewSize, _gen_alignment)) {
       FLAG_SET_ERGO(uintx, MaxNewSize, align_size_down(MaxNewSize, _gen_alignment));
     }
@@ -361,7 +376,9 @@ void TwoGenerationCollectorPolicy::initialize_flags() {
   GenCollectorPolicy::initialize_flags();
 
   if (!is_size_aligned(OldSize, _gen_alignment)) {
-    FLAG_SET_ERGO(uintx, OldSize, align_size_down(OldSize, _gen_alignment));
+    // Setting OldSize directly to preserve information about the possible
+    // setting of OldSize on the command line.
+    OldSize = align_size_down(OldSize, _gen_alignment);
   }
 
   if (FLAG_IS_CMDLINE(OldSize) && FLAG_IS_DEFAULT(MaxHeapSize)) {
@@ -400,6 +417,20 @@ void TwoGenerationCollectorPolicy::initialize_flags() {
     }
   }
 
+  // Update NewSize, if possible, to avoid sizing gen0 to small when only
+  // OldSize is set on the command line.
+  if (FLAG_IS_CMDLINE(OldSize) && !FLAG_IS_CMDLINE(NewSize)) {
+    if (OldSize < _initial_heap_byte_size) {
+      size_t new_size = _initial_heap_byte_size - OldSize;
+      // Need to compare against the flag value for max since _max_gen0_size
+      // might not have been set yet.
+      if (new_size >= _min_gen0_size && new_size <= MaxNewSize) {
+        FLAG_SET_ERGO(uintx, NewSize, new_size);
+        _initial_gen0_size = NewSize;
+      }
+    }
+  }
+
   always_do_update_barrier = UseConcMarkSweepGC;
 
   DEBUG_ONLY(TwoGenerationCollectorPolicy::assert_flags();)
@@ -441,57 +472,37 @@ void GenCollectorPolicy::initialize_size_info() {
   // Given the maximum gen0 size, determine the initial and
   // minimum gen0 sizes.
 
+  if (_max_heap_byte_size == _initial_heap_byte_size) {
+    // The maxium and initial heap sizes are the same so the generation's
+    // initial size must be the same as it maximum size. Use NewSize as the
+    // size if set on command line.
+    size_t fixed_young_size = FLAG_IS_CMDLINE(NewSize) ? NewSize : max_new_size;
+
+    _initial_gen0_size = fixed_young_size;
+    _max_gen0_size = fixed_young_size;
+
+    // Also update the minimum size if min == initial == max.
     if (_max_heap_byte_size == _min_heap_byte_size) {
-    // The maximum and minimum heap sizes are the same so the generations
+      _min_gen0_size = fixed_young_size;
-    // minimum and initial must be the same as its maximum.
+    }
-    _min_gen0_size = max_new_size;
-    _initial_gen0_size = max_new_size;
-    _max_gen0_size = max_new_size;
   } else {
     size_t desired_new_size = 0;
     if (FLAG_IS_CMDLINE(NewSize)) {
-      // If NewSize is set on the command line, we must use it as
+      // If NewSize is set on the command line, we should use it as
-      // the initial size and it also makes sense to use it as the
+      // the initial size, but make sure it is within the heap bounds.
-      // lower limit.
-      _min_gen0_size = NewSize;
-      desired_new_size = NewSize;
-      max_new_size = MAX2(max_new_size, NewSize);
-    } else if (FLAG_IS_ERGO(NewSize)) {
-      // If NewSize is set ergonomically, we should use it as a lower
-      // limit, but use NewRatio to calculate the initial size.
-      _min_gen0_size = NewSize;
       desired_new_size =
-        MAX2(scale_by_NewRatio_aligned(_initial_heap_byte_size), NewSize);
+        MIN2(max_new_size, bound_minus_alignment(NewSize, _initial_heap_byte_size));
-      max_new_size = MAX2(max_new_size, NewSize);
+      _min_gen0_size = bound_minus_alignment(desired_new_size, _min_heap_byte_size);
     } else {
-      // For the case where NewSize is the default, use NewRatio
+      // For the case where NewSize is not set on the command line, use
-      // to size the minimum and initial generation sizes.
+      // NewRatio to size the initial generation size. Use the current
-      // Use the default NewSize as the floor for these values.  If
+      // NewSize as the floor, because if NewRatio is overly large, the resulting
-      // NewRatio is overly large, the resulting sizes can be too small.
+      // size can be too small.
-      _min_gen0_size = MAX2(scale_by_NewRatio_aligned(_min_heap_byte_size), NewSize);
       desired_new_size =
-        MAX2(scale_by_NewRatio_aligned(_initial_heap_byte_size), NewSize);
+        MIN2(max_new_size, MAX2(scale_by_NewRatio_aligned(_initial_heap_byte_size), NewSize));
     }
-
-    assert(_min_gen0_size > 0, "Sanity check");
     _initial_gen0_size = desired_new_size;
     _max_gen0_size = max_new_size;
-
-    // At this point the desirable initial and minimum sizes have been
-    // determined without regard to the maximum sizes.
-
-    // Bound the sizes by the corresponding overall heap sizes.
-    _min_gen0_size = bound_minus_alignment(_min_gen0_size, _min_heap_byte_size);
-    _initial_gen0_size = bound_minus_alignment(_initial_gen0_size, _initial_heap_byte_size);
-    _max_gen0_size = bound_minus_alignment(_max_gen0_size, _max_heap_byte_size);
-
-    // At this point all three sizes have been checked against the
-    // maximum sizes but have not been checked for consistency among the three.
-
-    // Final check min <= initial <= max
-    _min_gen0_size = MIN2(_min_gen0_size, _max_gen0_size);
-    _initial_gen0_size = MAX2(MIN2(_initial_gen0_size, _max_gen0_size), _min_gen0_size);
-    _min_gen0_size = MIN2(_min_gen0_size, _initial_gen0_size);
   }
 
   // Write back to flags if necessary.
@@ -512,33 +523,6 @@ void GenCollectorPolicy::initialize_size_info() {
   DEBUG_ONLY(GenCollectorPolicy::assert_size_info();)
 }
 
-// Call this method during the sizing of the gen1 to make
-// adjustments to gen0 because of gen1 sizing policy.  gen0 initially has
-// the most freedom in sizing because it is done before the
-// policy for gen1 is applied.  Once gen1 policies have been applied,
-// there may be conflicts in the shape of the heap and this method
-// is used to make the needed adjustments.  The application of the
-// policies could be more sophisticated (iterative for example) but
-// keeping it simple also seems a worthwhile goal.
-bool TwoGenerationCollectorPolicy::adjust_gen0_sizes(size_t* gen0_size_ptr,
-                                                     size_t* gen1_size_ptr,
-                                                     const size_t heap_size) {
-  bool result = false;
-
-  if ((*gen0_size_ptr + *gen1_size_ptr) > heap_size) {
-    uintx smallest_new_size = young_gen_size_lower_bound();
-    if ((heap_size < (*gen0_size_ptr + _min_gen1_size)) &&
-        (heap_size >= _min_gen1_size + smallest_new_size)) {
-      // Adjust gen0 down to accommodate _min_gen1_size
-      *gen0_size_ptr = align_size_down_bounded(heap_size - _min_gen1_size, _gen_alignment);
-      result = true;
-    } else {
-      *gen1_size_ptr = align_size_down_bounded(heap_size - *gen0_size_ptr, _gen_alignment);
-    }
-  }
-  return result;
-}
-
 // Minimum sizes of the generations may be different than
 // the initial sizes.  An inconsistency is permitted here
 // in the total size that can be specified explicitly by
@@ -564,56 +548,63 @@ void TwoGenerationCollectorPolicy::initialize_size_info() {
     // with the overall heap size).  In either case make
     // the minimum, maximum and initial sizes consistent
     // with the gen0 sizes and the overall heap sizes.
-    _min_gen1_size = MAX2(_min_heap_byte_size - _min_gen0_size, _gen_alignment);
+    _min_gen1_size = _gen_alignment;
-    _initial_gen1_size = MAX2(_initial_heap_byte_size - _initial_gen0_size, _gen_alignment);
+    _initial_gen1_size = MIN2(_max_gen1_size, MAX2(_initial_heap_byte_size - _initial_gen0_size, _min_gen1_size));
     // _max_gen1_size has already been made consistent above
     FLAG_SET_ERGO(uintx, OldSize, _initial_gen1_size);
   } else {
-    // OldSize has been explicitly set on the command line. Use the
+    // OldSize has been explicitly set on the command line. Use it
-    // OldSize and then determine the consequences.
+    // for the initial size but make sure the minimum allow a young
-    _min_gen1_size = MIN2(OldSize, _min_heap_byte_size - _min_gen0_size);
+    // generation to fit as well.
-    _initial_gen1_size = OldSize;
-
     // If the user has explicitly set an OldSize that is inconsistent
     // with other command line flags, issue a warning.
     // The generation minimums and the overall heap minimum should
     // be within one generation alignment.
-    if ((_min_gen1_size + _min_gen0_size + _gen_alignment) < _min_heap_byte_size) {
-      warning("Inconsistency between minimum heap size and minimum "
-              "generation sizes: using minimum heap = " SIZE_FORMAT,
-              _min_heap_byte_size);
-    }
     if (OldSize > _max_gen1_size) {
       warning("Inconsistency between maximum heap size and maximum "
           "generation sizes: using maximum heap = " SIZE_FORMAT
           " -XX:OldSize flag is being ignored",
           _max_heap_byte_size);
+      FLAG_SET_ERGO(uintx, OldSize, _max_gen1_size);
     }
-    // If there is an inconsistency between the OldSize and the minimum and/or
+
-    // initial size of gen0, since OldSize was explicitly set, OldSize wins.
+    _min_gen1_size = MIN2(OldSize, _min_heap_byte_size - _min_gen0_size);
-    if (adjust_gen0_sizes(&_min_gen0_size, &_min_gen1_size, _min_heap_byte_size)) {
+    _initial_gen1_size = OldSize;
+  }
+
+  // The initial generation sizes should match the initial heap size,
+  // if not issue a warning and resize the generations. This behavior
+  // differs from JDK8 where the generation sizes have higher priority
+  // than the initial heap size.
+  if ((_initial_gen1_size + _initial_gen0_size) != _initial_heap_byte_size) {
+    warning("Inconsistency between generation sizes and heap size, resizing "
+            "the generations to fit the heap.");
+
+    size_t desired_gen0_size = _initial_heap_byte_size - _initial_gen1_size;
+    if (_initial_heap_byte_size < _initial_gen1_size) {
+      // Old want all memory, use minimum for young and rest for old
+      _initial_gen0_size = _min_gen0_size;
+      _initial_gen1_size = _initial_heap_byte_size - _min_gen0_size;
+    } else if (desired_gen0_size > _max_gen0_size) {
+      // Need to increase both young and old generation
+      _initial_gen0_size = _max_gen0_size;
+      _initial_gen1_size = _initial_heap_byte_size - _max_gen0_size;
+    } else if (desired_gen0_size < _min_gen0_size) {
+      // Need to decrease both young and old generation
+      _initial_gen0_size = _min_gen0_size;
+      _initial_gen1_size = _initial_heap_byte_size - _min_gen0_size;
+    } else {
+      // The young generation boundaries allow us to only update the
+      // young generation.
+      _initial_gen0_size = desired_gen0_size;
+    }
+
     if (PrintGCDetails && Verbose) {
       gclog_or_tty->print_cr("2: Minimum gen0 " SIZE_FORMAT "  Initial gen0 "
         SIZE_FORMAT "  Maximum gen0 " SIZE_FORMAT,
         _min_gen0_size, _initial_gen0_size, _max_gen0_size);
     }
   }
-    // The same as above for the old gen initial size.
-    if (adjust_gen0_sizes(&_initial_gen0_size, &_initial_gen1_size,
-                          _initial_heap_byte_size)) {
-      if (PrintGCDetails && Verbose) {
-        gclog_or_tty->print_cr("3: Minimum gen0 " SIZE_FORMAT "  Initial gen0 "
-          SIZE_FORMAT "  Maximum gen0 " SIZE_FORMAT,
-          _min_gen0_size, _initial_gen0_size, _max_gen0_size);
-      }
-    }
-  }
-
-  _min_gen1_size = MIN2(_min_gen1_size, _max_gen1_size);
-
-  // Make sure that min gen1 <= initial gen1 <= max gen1.
-  _initial_gen1_size = MAX2(_initial_gen1_size, _min_gen1_size);
-  _initial_gen1_size = MIN2(_initial_gen1_size, _max_gen1_size);
 
   // Write back to flags if necessary
   if (NewSize != _initial_gen0_size) {
@@ -994,56 +985,88 @@ void MarkSweepPolicy::initialize_gc_policy_counters() {
 // verify that there are some basic rules for NewSize honored by the policies.
 class TestGenCollectorPolicy {
 public:
-  static void test() {
+  static void test_new_size() {
     size_t flag_value;
 
     save_flags();
 
-    // Set some limits that makes the math simple.
-    FLAG_SET_ERGO(uintx, MaxHeapSize, 180 * M);
-    FLAG_SET_ERGO(uintx, InitialHeapSize, 120 * M);
-    Arguments::set_min_heap_size(40 * M);
-
     // If NewSize is set on the command line, it should be used
     // for both min and initial young size if less than min heap.
     flag_value = 20 * M;
+    set_basic_flag_values();
     FLAG_SET_CMDLINE(uintx, NewSize, flag_value);
-    verify_min(flag_value);
+    verify_gen0_min(flag_value);
-    verify_initial(flag_value);
+
+    set_basic_flag_values();
+    FLAG_SET_CMDLINE(uintx, NewSize, flag_value);
+    verify_gen0_initial(flag_value);
 
     // If NewSize is set on command line, but is larger than the min
     // heap size, it should only be used for initial young size.
     flag_value = 80 * M;
+    set_basic_flag_values();
     FLAG_SET_CMDLINE(uintx, NewSize, flag_value);
-    verify_initial(flag_value);
+    verify_gen0_initial(flag_value);
 
     // If NewSize has been ergonomically set, the collector policy
     // should use it for min but calculate the initial young size
     // using NewRatio.
     flag_value = 20 * M;
+    set_basic_flag_values();
     FLAG_SET_ERGO(uintx, NewSize, flag_value);
-    verify_min(flag_value);
+    verify_gen0_min(flag_value);
-    verify_scaled_initial(InitialHeapSize);
+
+    set_basic_flag_values();
+    FLAG_SET_ERGO(uintx, NewSize, flag_value);
+    verify_scaled_gen0_initial(InitialHeapSize);
 
     restore_flags();
+  }
+
+  static void test_old_size() {
+      size_t flag_value;
+
+      save_flags();
+
+      // If OldSize is set on the command line, it should be used
+      // for both min and initial old size if less than min heap.
+      flag_value = 20 * M;
+      set_basic_flag_values();
+      FLAG_SET_CMDLINE(uintx, OldSize, flag_value);
+      verify_gen1_min(flag_value);
+
+      set_basic_flag_values();
+      FLAG_SET_CMDLINE(uintx, OldSize, flag_value);
+      verify_gen1_initial(flag_value);
+
+      // If MaxNewSize is large, the maximum OldSize will be less than
+      // what's requested on the command line and it should be reset
+      // ergonomically.
+      flag_value = 30 * M;
+      set_basic_flag_values();
+      FLAG_SET_CMDLINE(uintx, OldSize, flag_value);
+      FLAG_SET_CMDLINE(uintx, MaxNewSize, 170*M);
+      // Calculate what we expect the flag to be.
+      flag_value = MaxHeapSize - MaxNewSize;
+      verify_gen1_initial(flag_value);
 
   }
 
-  static void verify_min(size_t expected) {
+  static void verify_gen0_min(size_t expected) {
     MarkSweepPolicy msp;
     msp.initialize_all();
 
     assert(msp.min_gen0_size() <= expected, err_msg("%zu  > %zu", msp.min_gen0_size(), expected));
   }
 
-  static void verify_initial(size_t expected) {
+  static void verify_gen0_initial(size_t expected) {
     MarkSweepPolicy msp;
     msp.initialize_all();
 
     assert(msp.initial_gen0_size() == expected, err_msg("%zu != %zu", msp.initial_gen0_size(), expected));
   }
 
-  static void verify_scaled_initial(size_t initial_heap_size) {
+  static void verify_scaled_gen0_initial(size_t initial_heap_size) {
     MarkSweepPolicy msp;
     msp.initialize_all();
 
@@ -1053,6 +1076,21 @@ public:
         err_msg("NewSize should have been set ergonomically to %zu, but was %zu", expected, NewSize));
   }
 
+  static void verify_gen1_min(size_t expected) {
+    MarkSweepPolicy msp;
+    msp.initialize_all();
+
+    assert(msp.min_gen1_size() <= expected, err_msg("%zu  > %zu", msp.min_gen1_size(), expected));
+  }
+
+  static void verify_gen1_initial(size_t expected) {
+    MarkSweepPolicy msp;
+    msp.initialize_all();
+
+    assert(msp.initial_gen1_size() == expected, err_msg("%zu != %zu", msp.initial_gen1_size(), expected));
+  }
+
+
 private:
   static size_t original_InitialHeapSize;
   static size_t original_MaxHeapSize;
@@ -1061,6 +1099,15 @@ private:
   static size_t original_NewSize;
   static size_t original_OldSize;
 
+  static void set_basic_flag_values() {
+    FLAG_SET_ERGO(uintx, MaxHeapSize, 180 * M);
+    FLAG_SET_ERGO(uintx, InitialHeapSize, 100 * M);
+    FLAG_SET_ERGO(uintx, OldSize, 4 * M);
+    FLAG_SET_ERGO(uintx, NewSize, 1 * M);
+    FLAG_SET_ERGO(uintx, MaxNewSize, 80 * M);
+    Arguments::set_min_heap_size(40 * M);
+  }
+
   static void save_flags() {
     original_InitialHeapSize   = InitialHeapSize;
     original_MaxHeapSize       = MaxHeapSize;
@@ -1088,7 +1135,11 @@ size_t TestGenCollectorPolicy::original_NewSize           = 0;
 size_t TestGenCollectorPolicy::original_OldSize           = 0;
 
 void TestNewSize_test() {
-  TestGenCollectorPolicy::test();
+  TestGenCollectorPolicy::test_new_size();
+}
+
+void TestOldSize_test() {
+  TestGenCollectorPolicy::test_old_size();
 }
 
 #endif

hotspot/src/share/vm/memory/collectorPolicy.hpp

@@ -335,10 +335,6 @@ class TwoGenerationCollectorPolicy : public GenCollectorPolicy {
   virtual CollectorPolicy::Name kind() {
     return CollectorPolicy::TwoGenerationCollectorPolicyKind;
   }
-
-  // Returns true if gen0 sizes were adjusted
-  bool adjust_gen0_sizes(size_t* gen0_size_ptr, size_t* gen1_size_ptr,
-                         const size_t heap_size);
 };
 
 class MarkSweepPolicy : public TwoGenerationCollectorPolicy {

hotspot/src/share/vm/memory/genCollectedHeap.cpp

@@ -374,7 +374,7 @@ void GenCollectedHeap::do_collection(bool  full,
 
   ClearedAllSoftRefs casr(do_clear_all_soft_refs, collector_policy());
 
-  const size_t metadata_prev_used = MetaspaceAux::allocated_used_bytes();
+  const size_t metadata_prev_used = MetaspaceAux::used_bytes();
 
   print_heap_before_gc();
 

hotspot/src/share/vm/memory/metaspace.cpp

@@ -1447,7 +1447,7 @@ void MetaspaceGC::compute_new_size() {
   uint current_shrink_factor = _shrink_factor;
   _shrink_factor = 0;
 
-  const size_t used_after_gc = MetaspaceAux::allocated_capacity_bytes();
+  const size_t used_after_gc = MetaspaceAux::capacity_bytes();
   const size_t capacity_until_GC = MetaspaceGC::capacity_until_GC();
 
   const double minimum_free_percentage = MinMetaspaceFreeRatio / 100.0;
@@ -2538,8 +2538,8 @@ void SpaceManager::mangle_freed_chunks() {
 // MetaspaceAux
 
 
-size_t MetaspaceAux::_allocated_capacity_words[] = {0, 0};
+size_t MetaspaceAux::_capacity_words[] = {0, 0};
-size_t MetaspaceAux::_allocated_used_words[] = {0, 0};
+size_t MetaspaceAux::_used_words[] = {0, 0};
 
 size_t MetaspaceAux::free_bytes(Metaspace::MetadataType mdtype) {
   VirtualSpaceList* list = Metaspace::get_space_list(mdtype);
@@ -2552,38 +2552,38 @@ size_t MetaspaceAux::free_bytes() {
 
 void MetaspaceAux::dec_capacity(Metaspace::MetadataType mdtype, size_t words) {
   assert_lock_strong(SpaceManager::expand_lock());
-  assert(words <= allocated_capacity_words(mdtype),
+  assert(words <= capacity_words(mdtype),
     err_msg("About to decrement below 0: words " SIZE_FORMAT
-            " is greater than _allocated_capacity_words[%u] " SIZE_FORMAT,
+            " is greater than _capacity_words[%u] " SIZE_FORMAT,
-            words, mdtype, allocated_capacity_words(mdtype)));
+            words, mdtype, capacity_words(mdtype)));
-  _allocated_capacity_words[mdtype] -= words;
+  _capacity_words[mdtype] -= words;
 }
 
 void MetaspaceAux::inc_capacity(Metaspace::MetadataType mdtype, size_t words) {
   assert_lock_strong(SpaceManager::expand_lock());
   // Needs to be atomic
-  _allocated_capacity_words[mdtype] += words;
+  _capacity_words[mdtype] += words;
 }
 
 void MetaspaceAux::dec_used(Metaspace::MetadataType mdtype, size_t words) {
-  assert(words <= allocated_used_words(mdtype),
+  assert(words <= used_words(mdtype),
     err_msg("About to decrement below 0: words " SIZE_FORMAT
-            " is greater than _allocated_used_words[%u] " SIZE_FORMAT,
+            " is greater than _used_words[%u] " SIZE_FORMAT,
-            words, mdtype, allocated_used_words(mdtype)));
+            words, mdtype, used_words(mdtype)));
   // For CMS deallocation of the Metaspaces occurs during the
   // sweep which is a concurrent phase.  Protection by the expand_lock()
   // is not enough since allocation is on a per Metaspace basis
   // and protected by the Metaspace lock.
   jlong minus_words = (jlong) - (jlong) words;
-  Atomic::add_ptr(minus_words, &_allocated_used_words[mdtype]);
+  Atomic::add_ptr(minus_words, &_used_words[mdtype]);
 }
 
 void MetaspaceAux::inc_used(Metaspace::MetadataType mdtype, size_t words) {
-  // _allocated_used_words tracks allocations for
+  // _used_words tracks allocations for
   // each piece of metadata.  Those allocations are
   // generally done concurrently by different application
   // threads so must be done atomically.
-  Atomic::add_ptr(words, &_allocated_used_words[mdtype]);
+  Atomic::add_ptr(words, &_used_words[mdtype]);
 }
 
 size_t MetaspaceAux::used_bytes_slow(Metaspace::MetadataType mdtype) {
@@ -2630,16 +2630,16 @@ size_t MetaspaceAux::capacity_bytes_slow(Metaspace::MetadataType mdtype) {
 
 size_t MetaspaceAux::capacity_bytes_slow() {
 #ifdef PRODUCT
-  // Use allocated_capacity_bytes() in PRODUCT instead of this function.
+  // Use capacity_bytes() in PRODUCT instead of this function.
   guarantee(false, "Should not call capacity_bytes_slow() in the PRODUCT");
 #endif
   size_t class_capacity = capacity_bytes_slow(Metaspace::ClassType);
   size_t non_class_capacity = capacity_bytes_slow(Metaspace::NonClassType);
-  assert(allocated_capacity_bytes() == class_capacity + non_class_capacity,
+  assert(capacity_bytes() == class_capacity + non_class_capacity,
-      err_msg("bad accounting: allocated_capacity_bytes() " SIZE_FORMAT
+      err_msg("bad accounting: capacity_bytes() " SIZE_FORMAT
         " class_capacity + non_class_capacity " SIZE_FORMAT
         " class_capacity " SIZE_FORMAT " non_class_capacity " SIZE_FORMAT,
-        allocated_capacity_bytes(), class_capacity + non_class_capacity,
+        capacity_bytes(), class_capacity + non_class_capacity,
         class_capacity, non_class_capacity));
 
   return class_capacity + non_class_capacity;
@@ -2699,14 +2699,14 @@ void MetaspaceAux::print_metaspace_change(size_t prev_metadata_used) {
                         "->" SIZE_FORMAT
                         "("  SIZE_FORMAT ")",
                         prev_metadata_used,
-                        allocated_used_bytes(),
+                        used_bytes(),
                         reserved_bytes());
   } else {
     gclog_or_tty->print(" "  SIZE_FORMAT "K"
                         "->" SIZE_FORMAT "K"
                         "("  SIZE_FORMAT "K)",
                         prev_metadata_used/K,
-                        allocated_used_bytes()/K,
+                        used_bytes()/K,
                         reserved_bytes()/K);
   }
 
@@ -2722,8 +2722,8 @@ void MetaspaceAux::print_on(outputStream* out) {
                 "capacity "  SIZE_FORMAT "K, "
                 "committed " SIZE_FORMAT "K, "
                 "reserved "  SIZE_FORMAT "K",
-                allocated_used_bytes()/K,
+                used_bytes()/K,
-                allocated_capacity_bytes()/K,
+                capacity_bytes()/K,
                 committed_bytes()/K,
                 reserved_bytes()/K);
 
@@ -2734,8 +2734,8 @@ void MetaspaceAux::print_on(outputStream* out) {
                   "capacity "  SIZE_FORMAT "K, "
                   "committed " SIZE_FORMAT "K, "
                   "reserved "  SIZE_FORMAT "K",
-                  allocated_used_bytes(ct)/K,
+                  used_bytes(ct)/K,
-                  allocated_capacity_bytes(ct)/K,
+                  capacity_bytes(ct)/K,
                   committed_bytes(ct)/K,
                   reserved_bytes(ct)/K);
   }
@@ -2837,42 +2837,42 @@ void MetaspaceAux::verify_free_chunks() {
 
 void MetaspaceAux::verify_capacity() {
 #ifdef ASSERT
-  size_t running_sum_capacity_bytes = allocated_capacity_bytes();
+  size_t running_sum_capacity_bytes = capacity_bytes();
   // For purposes of the running sum of capacity, verify against capacity
   size_t capacity_in_use_bytes = capacity_bytes_slow();
   assert(running_sum_capacity_bytes == capacity_in_use_bytes,
-    err_msg("allocated_capacity_words() * BytesPerWord " SIZE_FORMAT
+    err_msg("capacity_words() * BytesPerWord " SIZE_FORMAT
             " capacity_bytes_slow()" SIZE_FORMAT,
             running_sum_capacity_bytes, capacity_in_use_bytes));
   for (Metaspace::MetadataType i = Metaspace::ClassType;
        i < Metaspace:: MetadataTypeCount;
        i = (Metaspace::MetadataType)(i + 1)) {
     size_t capacity_in_use_bytes = capacity_bytes_slow(i);
-    assert(allocated_capacity_bytes(i) == capacity_in_use_bytes,
+    assert(capacity_bytes(i) == capacity_in_use_bytes,
-      err_msg("allocated_capacity_bytes(%u) " SIZE_FORMAT
+      err_msg("capacity_bytes(%u) " SIZE_FORMAT
               " capacity_bytes_slow(%u)" SIZE_FORMAT,
-              i, allocated_capacity_bytes(i), i, capacity_in_use_bytes));
+              i, capacity_bytes(i), i, capacity_in_use_bytes));
   }
 #endif
 }
 
 void MetaspaceAux::verify_used() {
 #ifdef ASSERT
-  size_t running_sum_used_bytes = allocated_used_bytes();
+  size_t running_sum_used_bytes = used_bytes();
   // For purposes of the running sum of used, verify against used
   size_t used_in_use_bytes = used_bytes_slow();
-  assert(allocated_used_bytes() == used_in_use_bytes,
+  assert(used_bytes() == used_in_use_bytes,
-    err_msg("allocated_used_bytes() " SIZE_FORMAT
+    err_msg("used_bytes() " SIZE_FORMAT
             " used_bytes_slow()" SIZE_FORMAT,
-            allocated_used_bytes(), used_in_use_bytes));
+            used_bytes(), used_in_use_bytes));
   for (Metaspace::MetadataType i = Metaspace::ClassType;
        i < Metaspace:: MetadataTypeCount;
        i = (Metaspace::MetadataType)(i + 1)) {
     size_t used_in_use_bytes = used_bytes_slow(i);
-    assert(allocated_used_bytes(i) == used_in_use_bytes,
+    assert(used_bytes(i) == used_in_use_bytes,
-      err_msg("allocated_used_bytes(%u) " SIZE_FORMAT
+      err_msg("used_bytes(%u) " SIZE_FORMAT
               " used_bytes_slow(%u)" SIZE_FORMAT,
-              i, allocated_used_bytes(i), i, used_in_use_bytes));
+              i, used_bytes(i), i, used_in_use_bytes));
   }
 #endif
 }

hotspot/src/share/vm/memory/metaspace.hpp

@@ -280,11 +280,11 @@ class MetaspaceAux : AllStatic {
   // allocated to a Metaspace.  This is used instead of
   // iterating over all the classloaders. One for each
   // type of Metadata
-  static size_t _allocated_capacity_words[Metaspace:: MetadataTypeCount];
+  static size_t _capacity_words[Metaspace:: MetadataTypeCount];
-  // Running sum of space in all Metachunks that have
+  // Running sum of space in all Metachunks that
   // are being used for metadata. One for each
   // type of Metadata.
-  static size_t _allocated_used_words[Metaspace:: MetadataTypeCount];
+  static size_t _used_words[Metaspace:: MetadataTypeCount];
 
  public:
   // Decrement and increment _allocated_capacity_words
@@ -308,32 +308,32 @@ class MetaspaceAux : AllStatic {
   static size_t free_chunks_total_bytes();
   static size_t free_chunks_total_bytes(Metaspace::MetadataType mdtype);
 
-  static size_t allocated_capacity_words(Metaspace::MetadataType mdtype) {
+  static size_t capacity_words(Metaspace::MetadataType mdtype) {
-    return _allocated_capacity_words[mdtype];
+    return _capacity_words[mdtype];
   }
-  static size_t allocated_capacity_words() {
+  static size_t capacity_words() {
-    return allocated_capacity_words(Metaspace::NonClassType) +
+    return capacity_words(Metaspace::NonClassType) +
-           allocated_capacity_words(Metaspace::ClassType);
+           capacity_words(Metaspace::ClassType);
   }
-  static size_t allocated_capacity_bytes(Metaspace::MetadataType mdtype) {
+  static size_t capacity_bytes(Metaspace::MetadataType mdtype) {
-    return allocated_capacity_words(mdtype) * BytesPerWord;
+    return capacity_words(mdtype) * BytesPerWord;
   }
-  static size_t allocated_capacity_bytes() {
+  static size_t capacity_bytes() {
-    return allocated_capacity_words() * BytesPerWord;
+    return capacity_words() * BytesPerWord;
   }
 
-  static size_t allocated_used_words(Metaspace::MetadataType mdtype) {
+  static size_t used_words(Metaspace::MetadataType mdtype) {
-    return _allocated_used_words[mdtype];
+    return _used_words[mdtype];
   }
-  static size_t allocated_used_words() {
+  static size_t used_words() {
-    return allocated_used_words(Metaspace::NonClassType) +
+    return used_words(Metaspace::NonClassType) +
-           allocated_used_words(Metaspace::ClassType);
+           used_words(Metaspace::ClassType);
   }
-  static size_t allocated_used_bytes(Metaspace::MetadataType mdtype) {
+  static size_t used_bytes(Metaspace::MetadataType mdtype) {
-    return allocated_used_words(mdtype) * BytesPerWord;
+    return used_words(mdtype) * BytesPerWord;
   }
-  static size_t allocated_used_bytes() {
+  static size_t used_bytes() {
-    return allocated_used_words() * BytesPerWord;
+    return used_words() * BytesPerWord;
   }
 
   static size_t free_bytes();

hotspot/src/share/vm/memory/metaspaceCounters.cpp

@@ -66,7 +66,7 @@ class MetaspacePerfCounters: public CHeapObj<mtInternal> {
 MetaspacePerfCounters* MetaspaceCounters::_perf_counters = NULL;
 
 size_t MetaspaceCounters::used() {
-  return MetaspaceAux::allocated_used_bytes();
+  return MetaspaceAux::used_bytes();
 }
 
 size_t MetaspaceCounters::capacity() {
@@ -98,7 +98,7 @@ void MetaspaceCounters::update_performance_counters() {
 MetaspacePerfCounters* CompressedClassSpaceCounters::_perf_counters = NULL;
 
 size_t CompressedClassSpaceCounters::used() {
-  return MetaspaceAux::allocated_used_bytes(Metaspace::ClassType);
+  return MetaspaceAux::used_bytes(Metaspace::ClassType);
 }
 
 size_t CompressedClassSpaceCounters::capacity() {

hotspot/src/share/vm/oops/constantPool.cpp

@@ -144,6 +144,10 @@ void ConstantPool::initialize_resolved_references(ClassLoaderData* loader_data,
 // CDS support. Create a new resolved_references array.
 void ConstantPool::restore_unshareable_info(TRAPS) {
 
+  // Only create the new resolved references array and lock if it hasn't been
+  // attempted before
+  if (resolved_references() != NULL) return;
+
   // restore the C++ vtable from the shared archive
   restore_vtable();
 

hotspot/src/share/vm/oops/instanceKlass.cpp

@@ -1289,17 +1289,18 @@ void InstanceKlass::do_local_static_fields(FieldClosure* cl) {
 }
 
 
-void InstanceKlass::do_local_static_fields(void f(fieldDescriptor*, TRAPS), TRAPS) {
+void InstanceKlass::do_local_static_fields(void f(fieldDescriptor*, Handle, TRAPS), Handle mirror, TRAPS) {
   instanceKlassHandle h_this(THREAD, this);
-  do_local_static_fields_impl(h_this, f, CHECK);
+  do_local_static_fields_impl(h_this, f, mirror, CHECK);
 }
 
 
-void InstanceKlass::do_local_static_fields_impl(instanceKlassHandle this_k, void f(fieldDescriptor* fd, TRAPS), TRAPS) {
+void InstanceKlass::do_local_static_fields_impl(instanceKlassHandle this_k,
+                             void f(fieldDescriptor* fd, Handle, TRAPS), Handle mirror, TRAPS) {
   for (JavaFieldStream fs(this_k()); !fs.done(); fs.next()) {
     if (fs.access_flags().is_static()) {
       fieldDescriptor& fd = fs.field_descriptor();
-      f(&fd, CHECK);
+      f(&fd, mirror, CHECK);
     }
   }
 }
@@ -2240,9 +2241,7 @@ void InstanceKlass::restore_unshareable_info(TRAPS) {
   int num_methods = methods->length();
   for (int index2 = 0; index2 < num_methods; ++index2) {
     methodHandle m(THREAD, methods->at(index2));
-    m()->link_method(m, CHECK);
+    m->restore_unshareable_info(CHECK);
-    // restore method's vtable by calling a virtual function
-    m->restore_vtable();
   }
   if (JvmtiExport::has_redefined_a_class()) {
     // Reinitialize vtable because RedefineClasses may have changed some
@@ -3409,6 +3408,10 @@ static void purge_previous_versions_internal(InstanceKlass* ik, int emcp_method_
               ("purge: %s(%s): prev method @%d in version @%d is alive",
               method->name()->as_C_string(),
               method->signature()->as_C_string(), j, i));
+            if (method->method_data() != NULL) {
+              // Clean out any weak method links
+              method->method_data()->clean_weak_method_links();
+            }
           }
         }
       }
@@ -3418,6 +3421,14 @@ static void purge_previous_versions_internal(InstanceKlass* ik, int emcp_method_
       ("purge: previous version stats: live=%d, deleted=%d", live_count,
       deleted_count));
   }
+
+  Array<Method*>* methods = ik->methods();
+  int num_methods = methods->length();
+  for (int index2 = 0; index2 < num_methods; ++index2) {
+    if (methods->at(index2)->method_data() != NULL) {
+      methods->at(index2)->method_data()->clean_weak_method_links();
+    }
+  }
 }
 
 // External interface for use during class unloading.

hotspot/src/share/vm/oops/instanceKlass.hpp

@@ -802,7 +802,7 @@ class InstanceKlass: public Klass {
   // Iterators
   void do_local_static_fields(FieldClosure* cl);
   void do_nonstatic_fields(FieldClosure* cl); // including inherited fields
-  void do_local_static_fields(void f(fieldDescriptor*, TRAPS), TRAPS);
+  void do_local_static_fields(void f(fieldDescriptor*, Handle, TRAPS), Handle, TRAPS);
 
   void methods_do(void f(Method* method));
   void array_klasses_do(void f(Klass* k));
@@ -1010,7 +1010,7 @@ private:
   static void set_initialization_state_and_notify_impl  (instanceKlassHandle this_k, ClassState state, TRAPS);
   static void call_class_initializer_impl               (instanceKlassHandle this_k, TRAPS);
   static Klass* array_klass_impl                        (instanceKlassHandle this_k, bool or_null, int n, TRAPS);
-  static void do_local_static_fields_impl               (instanceKlassHandle this_k, void f(fieldDescriptor* fd, TRAPS), TRAPS);
+  static void do_local_static_fields_impl               (instanceKlassHandle this_k, void f(fieldDescriptor* fd, Handle, TRAPS), Handle, TRAPS);
   /* jni_id_for_impl for jfieldID only */
   static JNIid* jni_id_for_impl                         (instanceKlassHandle this_k, int offset);
 

hotspot/src/share/vm/oops/instanceMirrorKlass.cpp

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2011, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2011, 2014, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -367,7 +367,12 @@ instanceOop InstanceMirrorKlass::allocate_instance(KlassHandle k, TRAPS) {
   // Query before forming handle.
   int size = instance_size(k);
   KlassHandle h_k(THREAD, this);
-  instanceOop i = (instanceOop) CollectedHeap::Class_obj_allocate(h_k, size, k, CHECK_NULL);
+  instanceOop i = (instanceOop)CollectedHeap::obj_allocate(h_k, size, CHECK_NULL);
+
+  // Since mirrors can be variable sized because of the static fields, store
+  // the size in the mirror itself.
+  java_lang_Class::set_oop_size(i, size);
+
   return i;
 }
 

hotspot/src/share/vm/oops/klass.cpp

@@ -475,12 +475,8 @@ void Klass::oops_do(OopClosure* cl) {
 }
 
 void Klass::remove_unshareable_info() {
-  if (!DumpSharedSpaces) {
+  assert (DumpSharedSpaces, "only called for DumpSharedSpaces");
-    // Clean up after OOM during class loading
+
-    if (class_loader_data() != NULL) {
-      class_loader_data()->remove_class(this);
-    }
-  }
   set_subklass(NULL);
   set_next_sibling(NULL);
   // Clear the java mirror
@@ -492,6 +488,10 @@ void Klass::remove_unshareable_info() {
 }
 
 void Klass::restore_unshareable_info(TRAPS) {
+  // If an exception happened during CDS restore, some of these fields may already be
+  // set.  We leave the class on the CLD list, even if incomplete so that we don't
+  // modify the CLD list outside a safepoint.
+  if (class_loader_data() == NULL) {
     ClassLoaderData* loader_data = ClassLoaderData::the_null_class_loader_data();
     // Restore class_loader_data to the null class loader data
     set_class_loader_data(loader_data);
@@ -499,11 +499,16 @@ void Klass::restore_unshareable_info(TRAPS) {
     // Add to null class loader list first before creating the mirror
     // (same order as class file parsing)
     loader_data->add_class(this);
+  }
 
   // Recreate the class mirror.  The protection_domain is always null for
   // boot loader, for now.
+  // Only recreate it if not present.  A previous attempt to restore may have
+  // gotten an OOM later but keep the mirror if it was created.
+  if (java_mirror() == NULL) {
     java_lang_Class::create_mirror(this, Handle(NULL), CHECK);
   }
+}
 
 Klass* Klass::array_klass_or_null(int rank) {
   EXCEPTION_MARK;

hotspot/src/share/vm/oops/method.cpp

@@ -903,6 +903,19 @@ address Method::make_adapters(methodHandle mh, TRAPS) {
   return adapter->get_c2i_entry();
 }
 
+void Method::restore_unshareable_info(TRAPS) {
+  // Since restore_unshareable_info can be called more than once for a method, don't
+  // redo any work.   If this field is restored, there is nothing to do.
+  if (_from_compiled_entry == NULL) {
+    // restore method's vtable by calling a virtual function
+    restore_vtable();
+
+    methodHandle mh(THREAD, this);
+    link_method(mh, CHECK);
+  }
+}
+
+
 // The verified_code_entry() must be called when a invoke is resolved
 // on this method.
 

hotspot/src/share/vm/oops/method.hpp

@@ -123,6 +123,8 @@ class Method : public Metadata {
   void restore_vtable() { guarantee(is_method(), "vtable restored by this call"); }
   bool is_method() const volatile { return true; }
 
+  void restore_unshareable_info(TRAPS);
+
   // accessors for instance variables
 
   ConstMethod* constMethod() const             { return _constMethod; }

hotspot/src/share/vm/oops/methodData.cpp

@@ -1531,9 +1531,35 @@ void MethodData::clean_extra_data_helper(DataLayout* dp, int shift, bool reset)
   }
 }
 
-// Remove SpeculativeTrapData entries that reference an unloaded
+class CleanExtraDataClosure : public StackObj {
-// method
+public:
-void MethodData::clean_extra_data(BoolObjectClosure* is_alive) {
+  virtual bool is_live(Method* m) = 0;
+};
+
+// Check for entries that reference an unloaded method
+class CleanExtraDataKlassClosure : public CleanExtraDataClosure {
+private:
+  BoolObjectClosure* _is_alive;
+public:
+  CleanExtraDataKlassClosure(BoolObjectClosure* is_alive) : _is_alive(is_alive) {}
+  bool is_live(Method* m) {
+    return m->method_holder()->is_loader_alive(_is_alive);
+  }
+};
+
+// Check for entries that reference a redefined method
+class CleanExtraDataMethodClosure : public CleanExtraDataClosure {
+public:
+  CleanExtraDataMethodClosure() {}
+  bool is_live(Method* m) {
+    return m->on_stack();
+  }
+};
+
+
+// Remove SpeculativeTrapData entries that reference an unloaded or
+// redefined method
+void MethodData::clean_extra_data(CleanExtraDataClosure* cl) {
   DataLayout* dp  = extra_data_base();
   DataLayout* end = extra_data_limit();
 
@@ -1544,7 +1570,7 @@ void MethodData::clean_extra_data(BoolObjectClosure* is_alive) {
       SpeculativeTrapData* data = new SpeculativeTrapData(dp);
       Method* m = data->method();
       assert(m != NULL, "should have a method");
-      if (!m->method_holder()->is_loader_alive(is_alive)) {
+      if (!cl->is_live(m)) {
         // "shift" accumulates the number of cells for dead
         // SpeculativeTrapData entries that have been seen so
         // far. Following entries must be shifted left by that many
@@ -1575,9 +1601,9 @@ void MethodData::clean_extra_data(BoolObjectClosure* is_alive) {
   }
 }
 
-// Verify there's no unloaded method referenced by a
+// Verify there's no unloaded or redefined method referenced by a
 // SpeculativeTrapData entry
-void MethodData::verify_extra_data_clean(BoolObjectClosure* is_alive) {
+void MethodData::verify_extra_data_clean(CleanExtraDataClosure* cl) {
 #ifdef ASSERT
   DataLayout* dp  = extra_data_base();
   DataLayout* end = extra_data_limit();
@@ -1587,7 +1613,7 @@ void MethodData::verify_extra_data_clean(BoolObjectClosure* is_alive) {
     case DataLayout::speculative_trap_data_tag: {
       SpeculativeTrapData* data = new SpeculativeTrapData(dp);
       Method* m = data->method();
-      assert(m != NULL && m->method_holder()->is_loader_alive(is_alive), "Method should exist");
+      assert(m != NULL && cl->is_live(m), "Method should exist");
       break;
     }
     case DataLayout::bit_data_tag:
@@ -1613,6 +1639,19 @@ void MethodData::clean_method_data(BoolObjectClosure* is_alive) {
     parameters->clean_weak_klass_links(is_alive);
   }
 
-  clean_extra_data(is_alive);
+  CleanExtraDataKlassClosure cl(is_alive);
-  verify_extra_data_clean(is_alive);
+  clean_extra_data(&cl);
+  verify_extra_data_clean(&cl);
+}
+
+void MethodData::clean_weak_method_links() {
+  for (ProfileData* data = first_data();
+       is_valid(data);
+       data = next_data(data)) {
+    data->clean_weak_method_links();
+  }
+
+  CleanExtraDataMethodClosure cl;
+  clean_extra_data(&cl);
+  verify_extra_data_clean(&cl);
 }

hotspot/src/share/vm/oops/methodData.hpp

@@ -251,6 +251,9 @@ public:
 
   // GC support
   void clean_weak_klass_links(BoolObjectClosure* cl);
+
+  // Redefinition support
+  void clean_weak_method_links();
 };
 
 
@@ -506,6 +509,9 @@ public:
   // GC support
   virtual void clean_weak_klass_links(BoolObjectClosure* is_alive_closure) {}
 
+  // Redefinition support
+  virtual void clean_weak_method_links() {}
+
   // CI translation: ProfileData can represent both MethodDataOop data
   // as well as CIMethodData data. This function is provided for translating
   // an oop in a ProfileData to the ci equivalent. Generally speaking,
@@ -1989,6 +1995,7 @@ public:
 //
 
 CC_INTERP_ONLY(class BytecodeInterpreter;)
+class CleanExtraDataClosure;
 
 class MethodData : public Metadata {
   friend class VMStructs;
@@ -2146,9 +2153,9 @@ private:
   static bool profile_parameters_jsr292_only();
   static bool profile_all_parameters();
 
-  void clean_extra_data(BoolObjectClosure* is_alive);
+  void clean_extra_data(CleanExtraDataClosure* cl);
   void clean_extra_data_helper(DataLayout* dp, int shift, bool reset = false);
-  void verify_extra_data_clean(BoolObjectClosure* is_alive);
+  void verify_extra_data_clean(CleanExtraDataClosure* cl);
 
 public:
   static int header_size() {
@@ -2440,6 +2447,8 @@ public:
   static bool profile_return_jsr292_only();
 
   void clean_method_data(BoolObjectClosure* is_alive);
+
+  void clean_weak_method_links();
 };
 
 #endif // SHARE_VM_OOPS_METHODDATAOOP_HPP

hotspot/src/share/vm/opto/callGenerator.cpp

@@ -70,6 +70,7 @@ public:
 
 JVMState* ParseGenerator::generate(JVMState* jvms, Parse* parent_parser) {
   Compile* C = Compile::current();
+  C->print_inlining_update(this);
 
   if (is_osr()) {
     // The JVMS for a OSR has a single argument (see its TypeFunc).
@@ -126,6 +127,7 @@ class DirectCallGenerator : public CallGenerator {
 
 JVMState* DirectCallGenerator::generate(JVMState* jvms, Parse* parent_parser) {
   GraphKit kit(jvms);
+  kit.C->print_inlining_update(this);
   bool is_static = method()->is_static();
   address target = is_static ? SharedRuntime::get_resolve_static_call_stub()
                              : SharedRuntime::get_resolve_opt_virtual_call_stub();
@@ -178,6 +180,8 @@ JVMState* VirtualCallGenerator::generate(JVMState* jvms, Parse* parent_parser) {
   GraphKit kit(jvms);
   Node* receiver = kit.argument(0);
 
+  kit.C->print_inlining_update(this);
+
   if (kit.C->log() != NULL) {
     kit.C->log()->elem("virtual_call bci='%d'", jvms->bci());
   }
@@ -262,14 +266,17 @@ CallGenerator* CallGenerator::for_virtual_call(ciMethod* m, int vtable_index) {
 
 // Allow inlining decisions to be delayed
 class LateInlineCallGenerator : public DirectCallGenerator {
+ private:
+  // unique id for log compilation
+  jlong _unique_id;
+
  protected:
   CallGenerator* _inline_cg;
-
   virtual bool do_late_inline_check(JVMState* jvms) { return true; }
 
  public:
   LateInlineCallGenerator(ciMethod* method, CallGenerator* inline_cg) :
-    DirectCallGenerator(method, true), _inline_cg(inline_cg) {}
+    DirectCallGenerator(method, true), _inline_cg(inline_cg), _unique_id(0) {}
 
   virtual bool is_late_inline() const { return true; }
 
@@ -278,7 +285,8 @@ class LateInlineCallGenerator : public DirectCallGenerator {
 
   virtual JVMState* generate(JVMState* jvms, Parse* parent_parser) {
     Compile *C = Compile::current();
-    C->print_inlining_skip(this);
+
+    C->log_inline_id(this);
 
     // Record that this call site should be revisited once the main
     // parse is finished.
@@ -296,10 +304,19 @@ class LateInlineCallGenerator : public DirectCallGenerator {
   virtual void print_inlining_late(const char* msg) {
     CallNode* call = call_node();
     Compile* C = Compile::current();
-    C->print_inlining_insert(this);
+    C->print_inlining_assert_ready();
     C->print_inlining(method(), call->jvms()->depth()-1, call->jvms()->bci(), msg);
+    C->print_inlining_move_to(this);
+    C->print_inlining_update_delayed(this);
   }
 
+  virtual void set_unique_id(jlong id) {
+    _unique_id = id;
+  }
+
+  virtual jlong unique_id() const {
+    return _unique_id;
+  }
 };
 
 void LateInlineCallGenerator::do_late_inline() {
@@ -360,6 +377,12 @@ void LateInlineCallGenerator::do_late_inline() {
     map->set_argument(jvms, i1, call->in(TypeFunc::Parms + i1));
   }
 
+  C->print_inlining_assert_ready();
+
+  C->print_inlining_move_to(this);
+
+  C->log_late_inline(this);
+
   // This check is done here because for_method_handle_inline() method
   // needs jvms for inlined state.
   if (!do_late_inline_check(jvms)) {
@@ -367,19 +390,6 @@ void LateInlineCallGenerator::do_late_inline() {
     return;
   }
 
-  C->print_inlining_insert(this);
-
-  CompileLog* log = C->log();
-  if (log != NULL) {
-    log->head("late_inline method='%d'", log->identify(method()));
-    JVMState* p = jvms;
-    while (p != NULL) {
-      log->elem("jvms bci='%d' method='%d'", p->bci(), log->identify(p->method()));
-      p = p->caller();
-    }
-    log->tail("late_inline");
-  }
-
   // Setup default node notes to be picked up by the inlining
   Node_Notes* old_nn = C->default_node_notes();
   if (old_nn != NULL) {
@@ -388,7 +398,7 @@ void LateInlineCallGenerator::do_late_inline() {
     C->set_default_node_notes(entry_nn);
   }
 
-  // Now perform the inling using the synthesized JVMState
+  // Now perform the inlining using the synthesized JVMState
   JVMState* new_jvms = _inline_cg->generate(jvms, NULL);
   if (new_jvms == NULL)  return;  // no change
   if (C->failing())      return;
@@ -431,25 +441,24 @@ class LateInlineMHCallGenerator : public LateInlineCallGenerator {
 
   virtual JVMState* generate(JVMState* jvms, Parse* parent_parser) {
     JVMState* new_jvms = LateInlineCallGenerator::generate(jvms, parent_parser);
+
+    Compile* C = Compile::current();
     if (_input_not_const) {
       // inlining won't be possible so no need to enqueue right now.
       call_node()->set_generator(this);
     } else {
-      Compile::current()->add_late_inline(this);
+      C->add_late_inline(this);
     }
     return new_jvms;
   }
-
-  virtual void print_inlining_late(const char* msg) {
-    if (!_input_not_const) return;
-    LateInlineCallGenerator::print_inlining_late(msg);
-  }
 };
 
 bool LateInlineMHCallGenerator::do_late_inline_check(JVMState* jvms) {
 
   CallGenerator* cg = for_method_handle_inline(jvms, _caller, method(), _input_not_const);
 
+  Compile::current()->print_inlining_update_delayed(this);
+
   if (!_input_not_const) {
     _attempt++;
   }
@@ -479,7 +488,8 @@ class LateInlineStringCallGenerator : public LateInlineCallGenerator {
 
   virtual JVMState* generate(JVMState* jvms, Parse* parent_parser) {
     Compile *C = Compile::current();
-    C->print_inlining_skip(this);
+
+    C->log_inline_id(this);
 
     C->add_string_late_inline(this);
 
@@ -502,7 +512,8 @@ class LateInlineBoxingCallGenerator : public LateInlineCallGenerator {
 
   virtual JVMState* generate(JVMState* jvms, Parse* parent_parser) {
     Compile *C = Compile::current();
-    C->print_inlining_skip(this);
+
+    C->log_inline_id(this);
 
     C->add_boxing_late_inline(this);
 
@@ -554,6 +565,8 @@ CallGenerator* CallGenerator::for_warm_call(WarmCallInfo* ci,
 
 JVMState* WarmCallGenerator::generate(JVMState* jvms, Parse* parent_parser) {
   Compile* C = Compile::current();
+  C->print_inlining_update(this);
+
   if (C->log() != NULL) {
     C->log()->elem("warm_call bci='%d'", jvms->bci());
   }
@@ -632,6 +645,7 @@ CallGenerator* CallGenerator::for_predicted_call(ciKlass* predicted_receiver,
 
 JVMState* PredictedCallGenerator::generate(JVMState* jvms, Parse* parent_parser) {
   GraphKit kit(jvms);
+  kit.C->print_inlining_update(this);
   PhaseGVN& gvn = kit.gvn();
   // We need an explicit receiver null_check before checking its type.
   // We share a map with the caller, so his JVMS gets adjusted.
@@ -779,6 +793,10 @@ CallGenerator* CallGenerator::for_method_handle_inline(JVMState* jvms, ciMethod*
         assert(cg == NULL || !cg->is_late_inline() || cg->is_mh_late_inline(), "no late inline here");
         if (cg != NULL && cg->is_inline())
           return cg;
+      } else {
+        const char* msg = "receiver not constant";
+        if (PrintInlining)  C->print_inlining(callee, jvms->depth() - 1, jvms->bci(), msg);
+        C->log_inline_failure(msg);
       }
     }
     break;
@@ -844,11 +862,14 @@ CallGenerator* CallGenerator::for_method_handle_inline(JVMState* jvms, ciMethod*
           // provide us with a type
           speculative_receiver_type = receiver_type->speculative_type();
         }
-
         CallGenerator* cg = C->call_generator(target, vtable_index, call_does_dispatch, jvms, true, PROB_ALWAYS, speculative_receiver_type, true, true);
         assert(cg == NULL || !cg->is_late_inline() || cg->is_mh_late_inline(), "no late inline here");
         if (cg != NULL && cg->is_inline())
           return cg;
+      } else {
+        const char* msg = "member_name not constant";
+        if (PrintInlining)  C->print_inlining(callee, jvms->depth() - 1, jvms->bci(), msg);
+        C->log_inline_failure(msg);
       }
     }
     break;
@@ -904,6 +925,7 @@ JVMState* PredictedIntrinsicGenerator::generate(JVMState* jvms, Parse* parent_pa
   if (kit.failing())
     return NULL;  // might happen because of NodeCountInliningCutoff
 
+  kit.C->print_inlining_update(this);
   SafePointNode* slow_map = NULL;
   JVMState* slow_jvms;
   if (slow_ctl != NULL) {
@@ -1017,6 +1039,7 @@ CallGenerator::for_uncommon_trap(ciMethod* m,
 
 JVMState* UncommonTrapCallGenerator::generate(JVMState* jvms, Parse* parent_parser) {
   GraphKit kit(jvms);
+  kit.C->print_inlining_update(this);
   // Take the trap with arguments pushed on the stack.  (Cf. null_check_receiver).
   int nargs = method()->arg_size();
   kit.inc_sp(nargs);

hotspot/src/share/vm/opto/callGenerator.hpp

@@ -84,6 +84,9 @@ class CallGenerator : public ResourceObj {
 
   virtual CallStaticJavaNode* call_node() const { ShouldNotReachHere(); return NULL; }
 
+  virtual void set_unique_id(jlong id)          { fatal("unique id only for late inlines"); };
+  virtual jlong unique_id() const               { fatal("unique id only for late inlines"); return 0; };
+
   // Note:  It is possible for a CG to be both inline and virtual.
   // (The hashCode intrinsic does a vtable check and an inlined fast path.)
 

hotspot/src/share/vm/opto/compile.cpp

@@ -662,6 +662,7 @@ Compile::Compile( ciEnv* ci_env, C2Compiler* compiler, ciMethod* target, int osr
                   _inlining_progress(false),
                   _inlining_incrementally(false),
                   _print_inlining_list(NULL),
+                  _print_inlining_stream(NULL),
                   _print_inlining_idx(0),
                   _preserve_jvm_state(0) {
   C = this;
@@ -723,9 +724,7 @@ Compile::Compile( ciEnv* ci_env, C2Compiler* compiler, ciMethod* target, int osr
   PhaseGVN gvn(node_arena(), estimated_size);
   set_initial_gvn(&gvn);
 
-  if (print_inlining() || print_intrinsics()) {
+  print_inlining_init();
-    _print_inlining_list = new (comp_arena())GrowableArray<PrintInliningBuffer>(comp_arena(), 1, 1, PrintInliningBuffer());
-  }
   { // Scope for timing the parser
     TracePhase t3("parse", &_t_parser, true);
 
@@ -967,6 +966,7 @@ Compile::Compile( ciEnv* ci_env,
     _inlining_progress(false),
     _inlining_incrementally(false),
     _print_inlining_list(NULL),
+    _print_inlining_stream(NULL),
     _print_inlining_idx(0),
     _preserve_jvm_state(0),
     _allowed_reasons(0) {
@@ -2023,6 +2023,8 @@ void Compile::Optimize() {
   ResourceMark rm;
   int          loop_opts_cnt;
 
+  print_inlining_reinit();
+
   NOT_PRODUCT( verify_graph_edges(); )
 
   print_method(PHASE_AFTER_PARSING);
@@ -3755,36 +3757,163 @@ void Compile::ConstantTable::fill_jump_table(CodeBuffer& cb, MachConstantNode* n
   }
 }
 
-void Compile::dump_inlining() {
+// The message about the current inlining is accumulated in
+// _print_inlining_stream and transfered into the _print_inlining_list
+// once we know whether inlining succeeds or not. For regular
+// inlining, messages are appended to the buffer pointed by
+// _print_inlining_idx in the _print_inlining_list. For late inlining,
+// a new buffer is added after _print_inlining_idx in the list. This
+// way we can update the inlining message for late inlining call site
+// when the inlining is attempted again.
+void Compile::print_inlining_init() {
   if (print_inlining() || print_intrinsics()) {
+    _print_inlining_stream = new stringStream();
+    _print_inlining_list = new (comp_arena())GrowableArray<PrintInliningBuffer>(comp_arena(), 1, 1, PrintInliningBuffer());
+  }
+}
+
+void Compile::print_inlining_reinit() {
+  if (print_inlining() || print_intrinsics()) {
+    // Re allocate buffer when we change ResourceMark
+    _print_inlining_stream = new stringStream();
+  }
+}
+
+void Compile::print_inlining_reset() {
+  _print_inlining_stream->reset();
+}
+
+void Compile::print_inlining_commit() {
+  assert(print_inlining() || print_intrinsics(), "PrintInlining off?");
+  // Transfer the message from _print_inlining_stream to the current
+  // _print_inlining_list buffer and clear _print_inlining_stream.
+  _print_inlining_list->at(_print_inlining_idx).ss()->write(_print_inlining_stream->as_string(), _print_inlining_stream->size());
+  print_inlining_reset();
+}
+
+void Compile::print_inlining_push() {
+  // Add new buffer to the _print_inlining_list at current position
+  _print_inlining_idx++;
+  _print_inlining_list->insert_before(_print_inlining_idx, PrintInliningBuffer());
+}
+
+Compile::PrintInliningBuffer& Compile::print_inlining_current() {
+  return _print_inlining_list->at(_print_inlining_idx);
+}
+
+void Compile::print_inlining_update(CallGenerator* cg) {
+  if (print_inlining() || print_intrinsics()) {
+    if (!cg->is_late_inline()) {
+      if (print_inlining_current().cg() != NULL) {
+        print_inlining_push();
+      }
+      print_inlining_commit();
+    } else {
+      if (print_inlining_current().cg() != cg &&
+          (print_inlining_current().cg() != NULL ||
+           print_inlining_current().ss()->size() != 0)) {
+        print_inlining_push();
+      }
+      print_inlining_commit();
+      print_inlining_current().set_cg(cg);
+    }
+  }
+}
+
+void Compile::print_inlining_move_to(CallGenerator* cg) {
+  // We resume inlining at a late inlining call site. Locate the
+  // corresponding inlining buffer so that we can update it.
+  if (print_inlining()) {
+    for (int i = 0; i < _print_inlining_list->length(); i++) {
+      if (_print_inlining_list->adr_at(i)->cg() == cg) {
+        _print_inlining_idx = i;
+        return;
+      }
+    }
+    ShouldNotReachHere();
+  }
+}
+
+void Compile::print_inlining_update_delayed(CallGenerator* cg) {
+  if (print_inlining()) {
+    assert(_print_inlining_stream->size() > 0, "missing inlining msg");
+    assert(print_inlining_current().cg() == cg, "wrong entry");
+    // replace message with new message
+    _print_inlining_list->at_put(_print_inlining_idx, PrintInliningBuffer());
+    print_inlining_commit();
+    print_inlining_current().set_cg(cg);
+  }
+}
+
+void Compile::print_inlining_assert_ready() {
+  assert(!_print_inlining || _print_inlining_stream->size() == 0, "loosing data");
+}
+
+void Compile::dump_inlining() {
+  bool do_print_inlining = print_inlining() || print_intrinsics();
+  if (do_print_inlining || log() != NULL) {
     // Print inlining message for candidates that we couldn't inline
-    // for lack of space or non constant receiver
+    // for lack of space
     for (int i = 0; i < _late_inlines.length(); i++) {
       CallGenerator* cg = _late_inlines.at(i);
-      cg->print_inlining_late("live nodes > LiveNodeCountInliningCutoff");
+      if (!cg->is_mh_late_inline()) {
+        const char* msg = "live nodes > LiveNodeCountInliningCutoff";
+        if (do_print_inlining) {
+          cg->print_inlining_late(msg);
         }
-    Unique_Node_List useful;
+        log_late_inline_failure(cg, msg);
-    useful.push(root());
-    for (uint next = 0; next < useful.size(); ++next) {
-      Node* n  = useful.at(next);
-      if (n->is_Call() && n->as_Call()->generator() != NULL && n->as_Call()->generator()->call_node() == n) {
-        CallNode* call = n->as_Call();
-        CallGenerator* cg = call->generator();
-        cg->print_inlining_late("receiver not constant");
-      }
-      uint max = n->len();
-      for ( uint i = 0; i < max; ++i ) {
-        Node *m = n->in(i);
-        if ( m == NULL ) continue;
-        useful.push(m);
       }
     }
+  }
+  if (do_print_inlining) {
     for (int i = 0; i < _print_inlining_list->length(); i++) {
       tty->print(_print_inlining_list->adr_at(i)->ss()->as_string());
     }
   }
 }
 
+void Compile::log_late_inline(CallGenerator* cg) {
+  if (log() != NULL) {
+    log()->head("late_inline method='%d'  inline_id='" JLONG_FORMAT "'", log()->identify(cg->method()),
+                cg->unique_id());
+    JVMState* p = cg->call_node()->jvms();
+    while (p != NULL) {
+      log()->elem("jvms bci='%d' method='%d'", p->bci(), log()->identify(p->method()));
+      p = p->caller();
+    }
+    log()->tail("late_inline");
+  }
+}
+
+void Compile::log_late_inline_failure(CallGenerator* cg, const char* msg) {
+  log_late_inline(cg);
+  if (log() != NULL) {
+    log()->inline_fail(msg);
+  }
+}
+
+void Compile::log_inline_id(CallGenerator* cg) {
+  if (log() != NULL) {
+    // The LogCompilation tool needs a unique way to identify late
+    // inline call sites. This id must be unique for this call site in
+    // this compilation. Try to have it unique across compilations as
+    // well because it can be convenient when grepping through the log
+    // file.
+    // Distinguish OSR compilations from others in case CICountOSR is
+    // on.
+    jlong id = ((jlong)unique()) + (((jlong)compile_id()) << 33) + (CICountOSR && is_osr_compilation() ? ((jlong)1) << 32 : 0);
+    cg->set_unique_id(id);
+    log()->elem("inline_id id='" JLONG_FORMAT "'", id);
+  }
+}
+
+void Compile::log_inline_failure(const char* msg) {
+  if (C->log() != NULL) {
+    C->log()->inline_fail(msg);
+  }
+}
+
+
 // Dump inlining replay data to the stream.
 // Don't change thread state and acquire any locks.
 void Compile::dump_inline_data(outputStream* out) {
@@ -3962,8 +4091,8 @@ void Compile::remove_speculative_types(PhaseIterGVN &igvn) {
     worklist.push(root());
     for (uint next = 0; next < worklist.size(); ++next) {
       Node *n  = worklist.at(next);
-      const Type* t = igvn.type(n);
+      const Type* t = igvn.type_or_null(n);
-      assert(t == t->remove_speculative(), "no more speculative types");
+      assert((t == NULL) || (t == t->remove_speculative()), "no more speculative types");
       if (n->is_Type()) {
         t = n->as_Type()->type();
         assert(t == t->remove_speculative(), "no more speculative types");

hotspot/src/share/vm/opto/compile.hpp

@@ -416,6 +416,7 @@ class Compile : public Phase {
     void set_cg(CallGenerator* cg) { _cg = cg; }
   };
 
+  stringStream* _print_inlining_stream;
   GrowableArray<PrintInliningBuffer>* _print_inlining_list;
   int _print_inlining_idx;
 
@@ -433,33 +434,26 @@ class Compile : public Phase {
 
   void* _replay_inline_data; // Pointer to data loaded from file
 
+  void print_inlining_init();
+  void print_inlining_reinit();
+  void print_inlining_commit();
+  void print_inlining_push();
+  PrintInliningBuffer& print_inlining_current();
+
+  void log_late_inline_failure(CallGenerator* cg, const char* msg);
+
  public:
 
   outputStream* print_inlining_stream() const {
-    return _print_inlining_list->adr_at(_print_inlining_idx)->ss();
+    assert(print_inlining() || print_intrinsics(), "PrintInlining off?");
+    return _print_inlining_stream;
   }
 
-  void print_inlining_skip(CallGenerator* cg) {
+  void print_inlining_update(CallGenerator* cg);
-    if (_print_inlining) {
+  void print_inlining_update_delayed(CallGenerator* cg);
-      _print_inlining_list->adr_at(_print_inlining_idx)->set_cg(cg);
+  void print_inlining_move_to(CallGenerator* cg);
-      _print_inlining_idx++;
+  void print_inlining_assert_ready();
-      _print_inlining_list->insert_before(_print_inlining_idx, PrintInliningBuffer());
+  void print_inlining_reset();
-    }
-  }
-
-  void print_inlining_insert(CallGenerator* cg) {
-    if (_print_inlining) {
-      for (int i = 0; i < _print_inlining_list->length(); i++) {
-        if (_print_inlining_list->adr_at(i)->cg() == cg) {
-          _print_inlining_list->insert_before(i+1, PrintInliningBuffer());
-          _print_inlining_idx = i+1;
-          _print_inlining_list->adr_at(i)->set_cg(NULL);
-          return;
-        }
-      }
-      ShouldNotReachHere();
-    }
-  }
 
   void print_inlining(ciMethod* method, int inline_level, int bci, const char* msg = NULL) {
     stringStream ss;
@@ -467,6 +461,10 @@ class Compile : public Phase {
     print_inlining_stream()->print(ss.as_string());
   }
 
+  void log_late_inline(CallGenerator* cg);
+  void log_inline_id(CallGenerator* cg);
+  void log_inline_failure(const char* msg);
+
   void* replay_inline_data() const { return _replay_inline_data; }
 
   // Dump inlining replay data to the stream.

hotspot/src/share/vm/opto/doCall.cpp

@@ -104,6 +104,9 @@ CallGenerator* Compile::call_generator(ciMethod* callee, int vtable_index, bool
         log->print(" receiver2='%d' receiver2_count='%d'", r2id, profile.receiver_count(1));
       }
     }
+    if (callee->is_method_handle_intrinsic()) {
+      log->print(" method_handle_intrinsic='1'");
+    }
     log->end_elem();
   }
 
@@ -294,6 +297,9 @@ CallGenerator* Compile::call_generator(ciMethod* callee, int vtable_index, bool
   // There was no special inlining tactic, or it bailed out.
   // Use a more generic tactic, like a simple call.
   if (call_does_dispatch) {
+    const char* msg = "virtual call";
+    if (PrintInlining) print_inlining(callee, jvms->depth() - 1, jvms->bci(), msg);
+    C->log_inline_failure(msg);
     return CallGenerator::for_virtual_call(callee, vtable_index);
   } else {
     // Class Hierarchy Analysis or Type Profile reveals a unique target,
@@ -396,6 +402,8 @@ void Parse::do_call() {
   // our contribution to it is cleaned up right here.
   kill_dead_locals();
 
+  C->print_inlining_assert_ready();
+
   // Set frequently used booleans
   const bool is_virtual = bc() == Bytecodes::_invokevirtual;
   const bool is_virtual_or_interface = is_virtual || bc() == Bytecodes::_invokeinterface;
@@ -531,7 +539,8 @@ void Parse::do_call() {
     // intrinsic was expecting to optimize. Should always be possible to
     // get a normal java call that may inline in that case
     cg = C->call_generator(cg->method(), vtable_index, call_does_dispatch, jvms, try_inline, prof_factor(), speculative_receiver_type, /* allow_intrinsics= */ false);
-    if ((new_jvms = cg->generate(jvms, this)) == NULL) {
+    new_jvms = cg->generate(jvms, this);
+    if (new_jvms == NULL) {
       guarantee(failing(), "call failed to generate:  calls should work");
       return;
     }

hotspot/src/share/vm/opto/library_call.cpp

@@ -620,6 +620,7 @@ JVMState* LibraryIntrinsic::generate(JVMState* jvms, Parse* parent_parser) {
     }
     // Push the result from the inlined method onto the stack.
     kit.push_result();
+    C->print_inlining_update(this);
     return kit.transfer_exceptions_into_jvms();
   }
 
@@ -637,6 +638,7 @@ JVMState* LibraryIntrinsic::generate(JVMState* jvms, Parse* parent_parser) {
     }
   }
   C->gather_intrinsic_statistics(intrinsic_id(), is_virtual(), Compile::_intrinsic_failed);
+  C->print_inlining_update(this);
   return NULL;
 }
 

hotspot/src/share/vm/opto/machnode.hpp

@@ -210,7 +210,14 @@ public:
   bool may_be_short_branch() const { return (flags() & Flag_may_be_short_branch) != 0; }
 
   // Avoid back to back some instructions on some CPUs.
-  bool avoid_back_to_back() const { return (flags() & Flag_avoid_back_to_back) != 0; }
+  enum AvoidBackToBackFlag { AVOID_NONE = 0,
+                             AVOID_BEFORE = Flag_avoid_back_to_back_before,
+                             AVOID_AFTER = Flag_avoid_back_to_back_after,
+                             AVOID_BEFORE_AND_AFTER = AVOID_BEFORE | AVOID_AFTER };
+
+  bool avoid_back_to_back(AvoidBackToBackFlag flag_value) const {
+    return (flags() & flag_value) == flag_value;
+  }
 
   // instruction implemented with a call
   bool has_call() const { return (flags() & Flag_has_call) != 0; }

hotspot/src/share/vm/opto/node.hpp

@@ -653,8 +653,9 @@ public:
     Flag_is_cisc_alternate           = Flag_is_Con << 1,
     Flag_is_dead_loop_safe           = Flag_is_cisc_alternate << 1,
     Flag_may_be_short_branch         = Flag_is_dead_loop_safe << 1,
-    Flag_avoid_back_to_back  = Flag_may_be_short_branch << 1,
+    Flag_avoid_back_to_back_before   = Flag_may_be_short_branch << 1,
-    Flag_has_call            = Flag_avoid_back_to_back << 1,
+    Flag_avoid_back_to_back_after    = Flag_avoid_back_to_back_before << 1,
+    Flag_has_call                    = Flag_avoid_back_to_back_after << 1,
     Flag_is_expensive                = Flag_has_call << 1,
     _max_flags = (Flag_is_expensive << 1) - 1 // allow flags combination
   };

hotspot/src/share/vm/opto/output.cpp

@@ -411,7 +411,7 @@ void Compile::shorten_branches(uint* blk_starts, int& code_size, int& reloc_size
             blk_size += nop_size;
           }
         }
-        if (mach->avoid_back_to_back()) {
+        if (mach->avoid_back_to_back(MachNode::AVOID_BEFORE)) {
           // Nop is inserted between "avoid back to back" instructions.
           // ScheduleAndBundle() can rearrange nodes in a block,
           // check for all offsets inside this block.
@@ -439,7 +439,7 @@ void Compile::shorten_branches(uint* blk_starts, int& code_size, int& reloc_size
         last_call_adr = blk_starts[i]+blk_size;
       }
       // Remember end of avoid_back_to_back offset
-      if (nj->is_Mach() && nj->as_Mach()->avoid_back_to_back()) {
+      if (nj->is_Mach() && nj->as_Mach()->avoid_back_to_back(MachNode::AVOID_AFTER)) {
         last_avoid_back_to_back_adr = blk_starts[i]+blk_size;
       }
     }
@@ -525,11 +525,11 @@ void Compile::shorten_branches(uint* blk_starts, int& code_size, int& reloc_size
           int new_size = replacement->size(_regalloc);
           int diff     = br_size - new_size;
           assert(diff >= (int)nop_size, "short_branch size should be smaller");
-          // Conservatively take into accound padding between
+          // Conservatively take into account padding between
           // avoid_back_to_back branches. Previous branch could be
           // converted into avoid_back_to_back branch during next
           // rounds.
-          if (needs_padding && replacement->avoid_back_to_back()) {
+          if (needs_padding && replacement->avoid_back_to_back(MachNode::AVOID_BEFORE)) {
             jmp_offset[i] += nop_size;
             diff -= nop_size;
           }
@@ -548,7 +548,7 @@ void Compile::shorten_branches(uint* blk_starts, int& code_size, int& reloc_size
         }
       } // (mach->may_be_short_branch())
       if (mach != NULL && (mach->may_be_short_branch() ||
-                           mach->avoid_back_to_back())) {
+                           mach->avoid_back_to_back(MachNode::AVOID_AFTER))) {
         last_may_be_short_branch_adr = blk_starts[i] + jmp_offset[i] + jmp_size[i];
       }
       blk_starts[i+1] -= adjust_block_start;
@@ -1313,7 +1313,7 @@ void Compile::fill_buffer(CodeBuffer* cb, uint* blk_starts) {
         if (is_sfn && !is_mcall && padding == 0 && current_offset == last_call_offset) {
           padding = nop_size;
         }
-        if (padding == 0 && mach->avoid_back_to_back() &&
+        if (padding == 0 && mach->avoid_back_to_back(MachNode::AVOID_BEFORE) &&
             current_offset == last_avoid_back_to_back_offset) {
           // Avoid back to back some instructions.
           padding = nop_size;
@@ -1407,7 +1407,7 @@ void Compile::fill_buffer(CodeBuffer* cb, uint* blk_starts) {
               int new_size = replacement->size(_regalloc);
               assert((br_size - new_size) >= (int)nop_size, "short_branch size should be smaller");
               // Insert padding between avoid_back_to_back branches.
-              if (needs_padding && replacement->avoid_back_to_back()) {
+              if (needs_padding && replacement->avoid_back_to_back(MachNode::AVOID_BEFORE)) {
                 MachNode *nop = new (this) MachNopNode();
                 block->insert_node(nop, j++);
                 _cfg->map_node_to_block(nop, block);
@@ -1515,7 +1515,7 @@ void Compile::fill_buffer(CodeBuffer* cb, uint* blk_starts) {
         last_call_offset = current_offset;
       }
 
-      if (n->is_Mach() && n->as_Mach()->avoid_back_to_back()) {
+      if (n->is_Mach() && n->as_Mach()->avoid_back_to_back(MachNode::AVOID_AFTER)) {
         // Avoid back to back some instructions.
         last_avoid_back_to_back_offset = current_offset;
       }

hotspot/src/share/vm/opto/phaseX.cpp

@@ -1529,6 +1529,21 @@ void PhaseCCP::do_transform() {
   C->set_root( transform(C->root())->as_Root() );
   assert( C->top(),  "missing TOP node" );
   assert( C->root(), "missing root" );
+
+  // Eagerly remove castPP nodes here. CastPP nodes might not be
+  // removed in the subsequent IGVN phase if a node that changes
+  // in(1) of a castPP is processed prior to the castPP node.
+  for (uint i = 0; i < _worklist.size(); i++) {
+    Node* n = _worklist.at(i);
+
+    if (n->is_ConstraintCast()) {
+      Node* nn = n->Identity(this);
+      if (nn != n) {
+        replace_node(n, nn);
+        --i;
+      }
+    }
+  }
 }
 
 //------------------------------transform--------------------------------------

hotspot/src/share/vm/prims/jni.cpp

@@ -3877,6 +3877,7 @@ void TestMetaspaceAux_test();
 void TestMetachunk_test();
 void TestVirtualSpaceNode_test();
 void TestNewSize_test();
+void TestOldSize_test();
 void TestKlass_test();
 void TestBitMap_test();
 #if INCLUDE_ALL_GCS
@@ -3903,6 +3904,7 @@ void execute_internal_vm_tests() {
     run_unit_test(AltHashing::test_alt_hash());
     run_unit_test(test_loggc_filename());
     run_unit_test(TestNewSize_test());
+    run_unit_test(TestOldSize_test());
     run_unit_test(TestKlass_test());
     run_unit_test(TestBitMap_test());
 #if INCLUDE_VM_STRUCTS

hotspot/src/share/vm/prims/whitebox.cpp

@@ -438,6 +438,30 @@ WB_ENTRY(jboolean, WB_EnqueueMethodForCompilation(JNIEnv* env, jobject o, jobjec
   return (mh->queued_for_compilation() || nm != NULL);
 WB_END
 
+class VM_WhiteBoxOperation : public VM_Operation {
+ public:
+  VM_WhiteBoxOperation()                         { }
+  VMOp_Type type()                  const        { return VMOp_WhiteBoxOperation; }
+  bool allow_nested_vm_operations() const        { return true; }
+};
+
+class AlwaysFalseClosure : public BoolObjectClosure {
+ public:
+  bool do_object_b(oop p) { return false; }
+};
+
+static AlwaysFalseClosure always_false;
+
+class VM_WhiteBoxCleanMethodData : public VM_WhiteBoxOperation {
+ public:
+  VM_WhiteBoxCleanMethodData(MethodData* mdo) : _mdo(mdo) { }
+  void doit() {
+    _mdo->clean_method_data(&always_false);
+  }
+ private:
+  MethodData* _mdo;
+};
+
 WB_ENTRY(void, WB_ClearMethodState(JNIEnv* env, jobject o, jobject method))
   jmethodID jmid = reflected_method_to_jmid(thread, env, method);
   CHECK_JNI_EXCEPTION(env);
@@ -453,6 +477,8 @@ WB_ENTRY(void, WB_ClearMethodState(JNIEnv* env, jobject o, jobject method))
     for (int i = 0; i < arg_count; i++) {
       mdo->set_arg_modified(i, 0);
     }
+    VM_WhiteBoxCleanMethodData op(mdo);
+    VMThread::execute(&op);
   }
 
   mh->clear_not_c1_compilable();

hotspot/src/share/vm/runtime/advancedThresholdPolicy.cpp

@@ -53,7 +53,8 @@ void AdvancedThresholdPolicy::initialize() {
   }
 
   set_c1_count(MAX2(count / 3, 1));
-  set_c2_count(MAX2(count - count / 3, 1));
+  set_c2_count(MAX2(count - c1_count(), 1));
+  FLAG_SET_ERGO(intx, CICompilerCount, c1_count() + c2_count());
 
   // Some inlining tuning
 #ifdef X86

hotspot/src/share/vm/runtime/arguments.cpp

@@ -307,6 +307,9 @@ static ObsoleteFlag obsolete_jvm_flags[] = {
                            JDK_Version::jdk_update(7, 2), JDK_Version::jdk(8) },
 #endif // PRODUCT
   { "UseVMInterruptibleIO",          JDK_Version::jdk(8), JDK_Version::jdk(9) },
+  { "UseBoundThreads",               JDK_Version::jdk(9), JDK_Version::jdk(10) },
+  { "DefaultThreadPriority",         JDK_Version::jdk(9), JDK_Version::jdk(10) },
+  { "NoYieldsInMicrolock",           JDK_Version::jdk(9), JDK_Version::jdk(10) },
   { NULL, JDK_Version(0), JDK_Version(0) }
 };
 
@@ -2078,17 +2081,6 @@ bool Arguments::check_vm_args_consistency() {
   // Note: Needs platform-dependent factoring.
   bool status = true;
 
-  // Allow both -XX:-UseStackBanging and -XX:-UseBoundThreads in non-product
-  // builds so the cost of stack banging can be measured.
-#if (defined(PRODUCT) && defined(SOLARIS))
-  if (!UseBoundThreads && !UseStackBanging) {
-    jio_fprintf(defaultStream::error_stream(),
-                "-UseStackBanging conflicts with -UseBoundThreads\n");
-
-     status = false;
-  }
-#endif
-
   if (TLABRefillWasteFraction == 0) {
     jio_fprintf(defaultStream::error_stream(),
                 "TLABRefillWasteFraction should be a denominator, "
@@ -2410,6 +2402,10 @@ bool Arguments::check_vm_args_consistency() {
   const int num_min_compiler_threads = (TieredCompilation && (TieredStopAtLevel >= CompLevel_full_optimization)) ? 2 : 1;
   status &=verify_min_value(CICompilerCount, num_min_compiler_threads, "CICompilerCount");
 
+  if (!FLAG_IS_DEFAULT(CICompilerCount) && !FLAG_IS_DEFAULT(CICompilerCountPerCPU) && CICompilerCountPerCPU) {
+    warning("The VM option CICompilerCountPerCPU overrides CICompilerCount.");
+  }
+
   return status;
 }
 

hotspot/src/share/vm/runtime/compilationPolicy.cpp

@@ -182,6 +182,7 @@ void NonTieredCompPolicy::initialize() {
     // max(log2(8)-1,1) = 2 compiler threads on an 8-way machine.
     // May help big-app startup time.
     _compiler_count = MAX2(log2_intptr(os::active_processor_count())-1,1);
+    FLAG_SET_ERGO(intx, CICompilerCount, _compiler_count);
   } else {
     _compiler_count = CICompilerCount;
   }

hotspot/src/share/vm/runtime/os.cpp

@@ -929,6 +929,10 @@ void os::print_cpu_info(outputStream* st) {
 }
 
 void os::print_date_and_time(outputStream *st) {
+  const int secs_per_day  = 86400;
+  const int secs_per_hour = 3600;
+  const int secs_per_min  = 60;
+
   time_t tloc;
   (void)time(&tloc);
   st->print("time: %s", ctime(&tloc));  // ctime adds newline.
@@ -937,7 +941,17 @@ void os::print_date_and_time(outputStream *st) {
   // NOTE: It tends to crash after a SEGV if we want to printf("%f",...) in
   //       Linux. Must be a bug in glibc ? Workaround is to round "t" to int
   //       before printf. We lost some precision, but who cares?
-  st->print_cr("elapsed time: %d seconds", (int)t);
+  int eltime = (int)t;  // elapsed time in seconds
+
+  // print elapsed time in a human-readable format:
+  int eldays = eltime / secs_per_day;
+  int day_secs = eldays * secs_per_day;
+  int elhours = (eltime - day_secs) / secs_per_hour;
+  int hour_secs = elhours * secs_per_hour;
+  int elmins = (eltime - day_secs - hour_secs) / secs_per_min;
+  int minute_secs = elmins * secs_per_min;
+  int elsecs = (eltime - day_secs - hour_secs - minute_secs);
+  st->print_cr("elapsed time: %d seconds (%dd %dh %dm %ds)", eltime, eldays, elhours, elmins, elsecs);
 }
 
 // moved from debug.cpp (used to be find()) but still called from there

hotspot/src/share/vm/runtime/os.hpp

@@ -450,8 +450,8 @@ class os: AllStatic {
     // yield that can be used in lieu of blocking.
   } ;
   static YieldResult NakedYield () ;
-  static void yield_all(int attempts = 0); // Yields to all other threads including lower priority
+  static void yield_all(); // Yields to all other threads including lower priority
-  static void loop_breaker(int attempts);  // called from within tight loops to possibly influence time-sharing
+                           // (for the default scheduling policy)
   static OSReturn set_priority(Thread* thread, ThreadPriority priority);
   static OSReturn get_priority(const Thread* const thread, ThreadPriority& priority);