Merge

.hgtags-top-repo

@@ -52,3 +52,4 @@ e1b972ff53cd58f825791f8ed9b2deffd16e768c jdk7-b68
 d1516b9f23954b29b8e76e6f4efc467c08c78133 jdk7-b75
 c8b63075403d53a208104a8a6ea5072c1cb66aab jdk7-b76
 1f17ca8353babb13f4908c1f87d11508232518c8 jdk7-b77
+ab4ae8f4514693a9fe17ca2fec0239d8f8450d2c jdk7-b78

hotspot/.hgtags

@@ -52,3 +52,4 @@ f4b900403d6e4b0af51447bd13bbe23fe3a1dac7 jdk7-b74
 d8dd291a362acb656026a9c0a9da48501505a1e7 jdk7-b75
 9174bb32e934965288121f75394874eeb1fcb649 jdk7-b76
 455105fc81d941482f8f8056afaa7aa0949c9300 jdk7-b77
+e703499b4b51e3af756ae77c3d5e8b3058a14e4e jdk7-b78

hotspot/make/hotspot_version

@@ -35,7 +35,7 @@ HOTSPOT_VM_COPYRIGHT=Copyright 2009
 
 HS_MAJOR_VER=17
 HS_MINOR_VER=0
-HS_BUILD_NUMBER=05
+HS_BUILD_NUMBER=06
 
 JDK_MAJOR_VER=1
 JDK_MINOR_VER=7

hotspot/make/linux/makefiles/debug.make

@@ -38,7 +38,7 @@ _JUNK_ := $(shell echo -e >&2 ""\
  "Please use 'make jvmg' to build debug JVM.                            \n" \
  "----------------------------------------------------------------------\n")
 
-G_SUFFIX =
+G_SUFFIX = _g
 VERSION = debug
 SYSDEFS += -DASSERT -DDEBUG
 PICFLAGS = DEFAULT

hotspot/make/linux/makefiles/fastdebug.make

@@ -58,7 +58,7 @@ CFLAGS$(HOTSPARC_GENERIC) += $(OPT_CFLAGS/BYFILE)
 # Linker mapfile
 MAPFILE = $(GAMMADIR)/make/linux/makefiles/mapfile-vers-debug
 
-G_SUFFIX =
+G_SUFFIX = _g
 VERSION = optimized
 SYSDEFS += -DASSERT -DFASTDEBUG
 PICFLAGS = DEFAULT

hotspot/make/linux/makefiles/jsig.make

@@ -25,9 +25,12 @@
 # Rules to build signal interposition library, used by vm.make
 
 # libjsig[_g].so: signal interposition library
-JSIG = jsig$(G_SUFFIX)
+JSIG = jsig
 LIBJSIG = lib$(JSIG).so
 
+JSIG_G    = $(JSIG)$(G_SUFFIX)
+LIBJSIG_G = lib$(JSIG_G).so
+
 JSIGSRCDIR = $(GAMMADIR)/src/os/$(Platform_os_family)/vm
 
 DEST_JSIG  = $(JDK_LIBDIR)/$(LIBJSIG)
@@ -50,6 +53,7 @@ $(LIBJSIG): $(JSIGSRCDIR)/jsig.c $(LIBJSIG_MAPFILE)
 	@echo Making signal interposition lib...
 	$(QUIETLY) $(CC) $(SYMFLAG) $(ARCHFLAG) $(SHARED_FLAG) $(PICFLAG) \
                          $(LFLAGS_JSIG) $(JSIG_DEBUG_CFLAGS) -o $@ $< -ldl
+	$(QUIETLY) [ -f $(LIBJSIG_G) ] || { ln -s $@ $(LIBJSIG_G); }
 
 install_jsig: $(LIBJSIG)
 	@echo "Copying $(LIBJSIG) to $(DEST_JSIG)"

hotspot/make/linux/makefiles/jvmg.make

@@ -35,7 +35,7 @@ CFLAGS += $(DEBUG_CFLAGS/BYFILE)
 # Linker mapfile
 MAPFILE = $(GAMMADIR)/make/linux/makefiles/mapfile-vers-debug
 
-G_SUFFIX =
+G_SUFFIX = _g
 VERSION = debug
 SYSDEFS += -DASSERT -DDEBUG
 PICFLAGS = DEFAULT

hotspot/make/linux/makefiles/launcher.make

@@ -25,7 +25,9 @@
 # Rules to build gamma launcher, used by vm.make
 
 # gamma[_g]: launcher
-LAUNCHER = gamma$(G_SUFFIX)
+
+LAUNCHER   = gamma
+LAUNCHER_G = $(LAUNCHER)$(G_SUFFIX)
 
 LAUNCHERDIR   = $(GAMMADIR)/src/os/$(Platform_os_family)/launcher
 LAUNCHERFLAGS = $(ARCHFLAG) \
@@ -70,4 +72,5 @@ $(LAUNCHER): $(LAUNCHER.o) $(LIBJVM) $(LAUNCHER_MAPFILE)
 	    $(LINK_LAUNCHER/PRE_HOOK) \
 	    $(LINK_LAUNCHER) $(LFLAGS_LAUNCHER) -o $@ $(LAUNCHER.o) $(LIBS_LAUNCHER); \
 	    $(LINK_LAUNCHER/POST_HOOK) \
+	    [ -f $(LAUNCHER_G) ] || { ln -s $@ $(LAUNCHER_G); }; \
         }

hotspot/make/linux/makefiles/saproc.make

@@ -25,9 +25,13 @@
 # Rules to build serviceability agent library, used by vm.make
 
 # libsaproc[_g].so: serviceability agent
-SAPROC = saproc$(G_SUFFIX)
+
+SAPROC = saproc
 LIBSAPROC = lib$(SAPROC).so
 
+SAPROC_G = $(SAPROC)$(G_SUFFIX)
+LIBSAPROC_G = lib$(SAPROC_G).so
+
 AGENT_DIR = $(GAMMADIR)/agent
 
 SASRCDIR = $(AGENT_DIR)/src/os/$(Platform_os_family)
@@ -75,6 +79,7 @@ $(LIBSAPROC): $(SASRCFILES) $(SAMAPFILE)
 	           $(SA_DEBUG_CFLAGS)                                   \
 	           -o $@                                                \
 	           -lthread_db
+	$(QUIETLY) [ -f $(LIBSAPROC_G) ] || { ln -s $@ $(LIBSAPROC_G); }
 
 install_saproc: checkAndBuildSA
 	$(QUIETLY) if [ -e $(LIBSAPROC) ] ; then             \

hotspot/make/linux/makefiles/vm.make

@@ -113,8 +113,9 @@ include $(MAKEFILES_DIR)/dtrace.make
 #----------------------------------------------------------------------
 # JVM
 
-JVM    = jvm$(G_SUFFIX)
+JVM      = jvm
-LIBJVM = lib$(JVM).so
+LIBJVM   = lib$(JVM).so
+LIBJVM_G = lib$(JVM)$(G_SUFFIX).so
 
 JVM_OBJ_FILES = $(Obj_Files)
 
@@ -201,6 +202,7 @@ $(LIBJVM): $(LIBJVM.o) $(LIBJVM_MAPFILE) $(LD_SCRIPT)
 		       $(LFLAGS_VM) -o $@ $(LIBJVM.o) $(LIBS_VM);       \
 	    $(LINK_LIB.CC/POST_HOOK)                                    \
 	    rm -f $@.1; ln -s $@ $@.1;                                  \
+	    [ -f $(LIBJVM_G) ] || { ln -s $@ $(LIBJVM_G); ln -s $@.1 $(LIBJVM_G).1; }; \
 	    if [ -x /usr/sbin/selinuxenabled ] ; then                   \
 	      /usr/sbin/selinuxenabled;                                 \
               if [ $$? = 0 ] ; then					\

hotspot/make/solaris/makefiles/debug.make

@@ -54,7 +54,7 @@ _JUNK_ := $(shell echo >&2 ""\
  "Please use 'gnumake jvmg' to build debug JVM.                            \n" \
  "-------------------------------------------------------------------------\n")
 
-G_SUFFIX =
+G_SUFFIX = _g
 VERSION = debug
 SYSDEFS += -DASSERT -DDEBUG
 PICFLAGS = DEFAULT

hotspot/make/solaris/makefiles/dtrace.make

@@ -24,8 +24,8 @@
 
 # Rules to build jvm_db/dtrace, used by vm.make
 
-# we build libjvm_dtrace/libjvm_db/dtrace for COMPILER1 and COMPILER2
+# We build libjvm_dtrace/libjvm_db/dtrace for COMPILER1 and COMPILER2
-# but not for CORE configuration
+# but not for CORE or KERNEL configurations.
 
 ifneq ("${TYPE}", "CORE")
 ifneq ("${TYPE}", "KERNEL")
@@ -37,12 +37,13 @@ dtraceCheck:
 
 else
 
-
 JVM_DB = libjvm_db
-LIBJVM_DB = libjvm$(G_SUFFIX)_db.so
+LIBJVM_DB = libjvm_db.so
+LIBJVM_DB_G = libjvm$(G_SUFFIX)_db.so
 
 JVM_DTRACE = jvm_dtrace
-LIBJVM_DTRACE = libjvm$(G_SUFFIX)_dtrace.so
+LIBJVM_DTRACE = libjvm_dtrace.so
+LIBJVM_DTRACE_G = libjvm$(G_SUFFIX)_dtrace.so
 
 JVMOFFS = JvmOffsets
 JVMOFFS.o = $(JVMOFFS).o
@@ -77,7 +78,7 @@ LFLAGS_JVM_DB += -D_REENTRANT $(PICFLAG)
 LFLAGS_JVM_DTRACE += -D_REENTRANT $(PICFLAG)
 else
 LFLAGS_JVM_DB += -mt $(PICFLAG) -xnolib
-LFLAGS_JVM_DTRACE += -mt $(PICFLAG) -xnolib
+LFLAGS_JVM_DTRACE += -mt $(PICFLAG) -xnolib -ldl
 endif
 
 ISA = $(subst i386,i486,$(shell isainfo -n))
@@ -86,18 +87,24 @@ ISA = $(subst i386,i486,$(shell isainfo -n))
 ifneq ("${ISA}","${BUILDARCH}")
 
 XLIBJVM_DB = 64/$(LIBJVM_DB)
+XLIBJVM_DB_G = 64/$(LIBJVM_DB_G)
 XLIBJVM_DTRACE = 64/$(LIBJVM_DTRACE)
+XLIBJVM_DTRACE_G = 64/$(LIBJVM_DTRACE_G)
 
 $(XLIBJVM_DB): $(DTRACE_SRCDIR)/$(JVM_DB).c $(JVMOFFS).h $(LIBJVM_DB_MAPFILE)
 	@echo Making $@
 	$(QUIETLY) mkdir -p 64/ ; \
 	$(CC) $(SYMFLAG) $(ARCHFLAG/$(ISA)) -D$(TYPE) -I. -I$(GENERATED) \
 		$(SHARED_FLAG) $(LFLAGS_JVM_DB) -o $@ $(DTRACE_SRCDIR)/$(JVM_DB).c -lc
+	[ -f $(XLIBJVM_DB_G) ] || { ln -s $(LIBJVM_DB) $(XLIBJVM_DB_G); }
+
 $(XLIBJVM_DTRACE): $(DTRACE_SRCDIR)/$(JVM_DTRACE).c $(DTRACE_SRCDIR)/$(JVM_DTRACE).h $(LIBJVM_DTRACE_MAPFILE)
 	@echo Making $@
 	$(QUIETLY) mkdir -p 64/ ; \
 	$(CC) $(SYMFLAG) $(ARCHFLAG/$(ISA)) -D$(TYPE) -I. \
 		$(SHARED_FLAG) $(LFLAGS_JVM_DTRACE) -o $@ $(DTRACE_SRCDIR)/$(JVM_DTRACE).c -lc -lthread -ldoor
+	[ -f $(XLIBJVM_DTRACE_G) ] || { ln -s $(LIBJVM_DTRACE) $(XLIBJVM_DTRACE_G); }
+
 endif # ifneq ("${ISA}","${BUILDARCH}")
 
 ifdef USE_GCC
@@ -142,11 +149,13 @@ $(LIBJVM_DB): $(DTRACE_SRCDIR)/$(JVM_DB).c $(JVMOFFS.o) $(XLIBJVM_DB) $(LIBJVM_D
 	@echo Making $@
 	$(QUIETLY) $(CC) $(SYMFLAG) $(ARCHFLAG) -D$(TYPE) -I. -I$(GENERATED) \
 		$(SHARED_FLAG) $(LFLAGS_JVM_DB) -o $@ $(DTRACE_SRCDIR)/$(JVM_DB).c -lc
+	[ -f $(LIBJVM_DB_G) ] || { ln -s $@ $(LIBJVM_DB_G); }
 
 $(LIBJVM_DTRACE): $(DTRACE_SRCDIR)/$(JVM_DTRACE).c $(XLIBJVM_DTRACE) $(DTRACE_SRCDIR)/$(JVM_DTRACE).h $(LIBJVM_DTRACE_MAPFILE)
 	@echo Making $@
 	$(QUIETLY) $(CC) $(SYMFLAG) $(ARCHFLAG) -D$(TYPE) -I.  \
 		$(SHARED_FLAG) $(LFLAGS_JVM_DTRACE) -o $@ $(DTRACE_SRCDIR)/$(JVM_DTRACE).c -lc -lthread -ldoor
+	[ -f $(LIBJVM_DTRACE_G) ] || { ln -s $@ $(LIBJVM_DTRACE_G); }
 
 $(DTRACE).d: $(DTRACE_SRCDIR)/hotspot.d $(DTRACE_SRCDIR)/hotspot_jni.d \
              $(DTRACE_SRCDIR)/hs_private.d $(DTRACE_SRCDIR)/jhelper.d

hotspot/make/solaris/makefiles/fastdebug.make

@@ -90,7 +90,6 @@ endif # Platform_compiler == sparcWorks
 # for this method for now. (fix this when dtrace bug 6258412 is fixed)
 OPT_CFLAGS/ciEnv.o = $(OPT_CFLAGS) -xinline=no%__1cFciEnvbFpost_compiled_method_load_event6MpnHnmethod__v_
 
-
 # (OPT_CFLAGS/SLOWER is also available, to alter compilation of buggy files)
 
 # If you set HOTSPARC_GENERIC=yes, you disable all OPT_CFLAGS settings
@@ -115,8 +114,7 @@ MAPFILE = $(GAMMADIR)/make/solaris/makefiles/mapfile-vers \
 # and mustn't be otherwise.
 MAPFILE_DTRACE = $(GAMMADIR)/make/solaris/makefiles/mapfile-vers-$(TYPE)
 
-
+G_SUFFIX = _g
-G_SUFFIX =
 VERSION = optimized
 SYSDEFS += -DASSERT -DFASTDEBUG -DCHECK_UNHANDLED_OOPS
 PICFLAGS = DEFAULT

hotspot/make/solaris/makefiles/jsig.make

@@ -25,8 +25,11 @@
 # Rules to build signal interposition library, used by vm.make
 
 # libjsig[_g].so: signal interposition library
-JSIG = jsig$(G_SUFFIX)
+JSIG      = jsig
-LIBJSIG = lib$(JSIG).so
+LIBJSIG   = lib$(JSIG).so
+
+JSIG_G    = $(JSIG)$(G_SUFFIX)
+LIBJSIG_G = lib$(JSIG_G).so
 
 JSIGSRCDIR = $(GAMMADIR)/src/os/$(Platform_os_family)/vm
 
@@ -46,6 +49,7 @@ $(LIBJSIG): $(JSIGSRCDIR)/jsig.c $(LIBJSIG_MAPFILE)
 	@echo Making signal interposition lib...
 	$(QUIETLY) $(CC) $(SYMFLAG) $(ARCHFLAG) $(SHARED_FLAG) $(PICFLAG) \
                          $(LFLAGS_JSIG) -o $@ $< -ldl
+	[ -f $(LIBJSIG_G) ] || { ln -s $@ $(LIBJSIG_G); }
 
 install_jsig: $(LIBJSIG)
 	@echo "Copying $(LIBJSIG) to $(DEST_JSIG)"

hotspot/make/solaris/makefiles/jvmg.make

@@ -51,7 +51,7 @@ MAPFILE = $(GAMMADIR)/make/solaris/makefiles/mapfile-vers \
 # and mustn't be otherwise.
 MAPFILE_DTRACE = $(GAMMADIR)/make/solaris/makefiles/mapfile-vers-$(TYPE)
 
-G_SUFFIX =
+G_SUFFIX = _g
 VERSION = debug
 SYSDEFS += -DASSERT -DDEBUG
 PICFLAGS = DEFAULT

hotspot/make/solaris/makefiles/launcher.make

@@ -25,7 +25,8 @@
 # Rules to build gamma launcher, used by vm.make
 
 # gamma[_g]: launcher
-LAUNCHER = gamma$(G_SUFFIX)
+LAUNCHER   = gamma
+LAUNCHER_G = $(LAUNCHER)$(G_SUFFIX)
 
 LAUNCHERDIR   = $(GAMMADIR)/src/os/$(Platform_os_family)/launcher
 LAUNCHERFLAGS = $(ARCHFLAG) \
@@ -88,5 +89,6 @@ $(LAUNCHER): $(LAUNCHER.o) $(LIBJVM) $(LAUNCHER_MAPFILE)
 	    $(LINK_LAUNCHER/PRE_HOOK) \
 	    $(LINK_LAUNCHER) $(LFLAGS_LAUNCHER) -o $@ $(LAUNCHER.o) $(LIBS_LAUNCHER); \
 	    $(LINK_LAUNCHER/POST_HOOK) \
+	    [ -f $(LAUNCHER_G) ] || { ln -s $@ $(LAUNCHER_G); }; \
 	    ;; \
 	esac

hotspot/make/solaris/makefiles/saproc.make

@@ -25,9 +25,13 @@
 # Rules to build serviceability agent library, used by vm.make
 
 # libsaproc[_g].so: serviceability agent
-SAPROC = saproc$(G_SUFFIX)
+
+SAPROC = saproc
 LIBSAPROC = lib$(SAPROC).so
 
+SAPROC_G = $(SAPROC)$(G_SUFFIX)
+LIBSAPROC_G = lib$(SAPROC_G).so
+
 AGENT_DIR = $(GAMMADIR)/agent
 
 SASRCDIR = $(AGENT_DIR)/src/os/$(Platform_os_family)/proc
@@ -69,6 +73,7 @@ $(LIBSAPROC): $(SASRCFILES) $(SAMAPFILE)
 	           $(SA_LFLAGS)                                         \
 	           -o $@                                                \
 	           -ldl -ldemangle -lthread -lc
+	[ -f $(LIBSAPROC_G) ] || { ln -s $@ $(LIBSAPROC_G); }
 
 install_saproc: checkAndBuildSA
 	$(QUIETLY) if [ -f $(LIBSAPROC) ] ; then             \

hotspot/make/solaris/makefiles/sparcWorks.make

@@ -281,8 +281,6 @@ else
 OPT_CFLAGS=-xO4 $(EXTRA_OPT_CFLAGS)
 endif
 
-CFLAGS += $(GAMMADIR)/src/os_cpu/solaris_sparc/vm/solaris_sparc.il
-
 endif # sparc
 
 ifeq ("${Platform_arch_model}", "x86_32")
@@ -293,13 +291,14 @@ OPT_CFLAGS=-xtarget=pentium $(EXTRA_OPT_CFLAGS)
 # [phh] Is this still true for 6.1?
 OPT_CFLAGS+=-xO3
 
-CFLAGS += $(GAMMADIR)/src/os_cpu/solaris_x86/vm/solaris_x86_32.il
-
 endif # 32bit x86
 
 # no more exceptions
 CFLAGS/NOEX=-noex
 
+# Inline functions
+CFLAGS += $(GAMMADIR)/src/os_cpu/solaris_${Platform_arch}/vm/solaris_${Platform_arch_model}.il
+
 # Reduce code bloat by reverting back to 5.0 behavior for static initializers
 CFLAGS += -Qoption ccfe -one_static_init
 
@@ -312,6 +311,15 @@ PICFLAG/DEFAULT = $(PICFLAG)
 PICFLAG/BETTER  = $(PICFLAG/DEFAULT)
 PICFLAG/BYFILE  = $(PICFLAG/$@)$(PICFLAG/DEFAULT$(PICFLAG/$@))
 
+# Use $(MAPFLAG:FILENAME=real_file_name) to specify a map file.
+MAPFLAG = -M FILENAME
+
+# Use $(SONAMEFLAG:SONAME=soname) to specify the intrinsic name of a shared obj
+SONAMEFLAG = -h SONAME
+
+# Build shared library
+SHARED_FLAG = -G
+
 # Would be better if these weren't needed, since we link with CC, but
 # at present removing them causes run-time errors
 LFLAGS += -library=Crun

hotspot/make/solaris/makefiles/vm.make

@@ -108,11 +108,16 @@ ifeq ($(shell expr $(COMPILER_REV_NUMERIC) \>= 505), 1)
 #   older libm before libCrun, just to make sure it's found and used first.
 LIBS += -lsocket -lsched -ldl $(LIBM) -lCrun -lthread -ldoor -lc
 else
+ifeq ($(COMPILER_REV_NUMERIC), 502)
+# SC6.1 has it's own libm.so: specifying anything else provokes a name conflict.
+LIBS += -ldl -lthread -lsocket -lm -lsched -ldoor
+else
 LIBS += -ldl -lthread -lsocket $(LIBM) -lsched -ldoor
-endif
+endif # 502
+endif # 505
 else
 LIBS += -lsocket -lsched -ldl $(LIBM) -lthread -lc
-endif
+endif # sparcWorks
 
 # By default, link the *.o into the library, not the executable.
 LINK_INTO$(LINK_INTO) = LIBJVM
@@ -126,8 +131,9 @@ include $(MAKEFILES_DIR)/dtrace.make
 #----------------------------------------------------------------------
 # JVM
 
-JVM    = jvm$(G_SUFFIX)
+JVM      = jvm
-LIBJVM = lib$(JVM).so
+LIBJVM   = lib$(JVM).so
+LIBJVM_G = lib$(JVM)$(G_SUFFIX).so
 
 JVM_OBJ_FILES = $(Obj_Files) $(DTRACE_OBJS)
 
@@ -173,11 +179,12 @@ $(LIBJVM): $(LIBJVM.o) $(LIBJVM_MAPFILE)
 	-sbfast|-xsbfast) \
 	    ;; \
 	*) \
-	    echo Linking vm...;                                                  \
+	    echo Linking vm...; \
-	    $(LINK_LIB.CC/PRE_HOOK)                                              \
+	    $(LINK_LIB.CC/PRE_HOOK) \
-	    $(LINK_VM) $(LFLAGS_VM) -o $@ $(LIBJVM.o) $(LIBS_VM);                \
+	    $(LINK_VM) $(LFLAGS_VM) -o $@ $(LIBJVM.o) $(LIBS_VM); \
-	    $(LINK_LIB.CC/POST_HOOK)                                             \
+	    $(LINK_LIB.CC/POST_HOOK) \
-	    rm -f $@.1; ln -s $@ $@.1;                                           \
+	    rm -f $@.1; ln -s $@ $@.1; \
+	    [ -f $(LIBJVM_G) ] || { ln -s $@ $(LIBJVM_G); ln -s $@.1 $(LIBJVM_G).1; }; \
 	    ;; \
 	esac
 

hotspot/src/cpu/sparc/vm/c1_LIRAssembler_sparc.cpp

@@ -189,14 +189,17 @@ void LIR_Assembler::osr_entry() {
   Register OSR_buf = osrBufferPointer()->as_register();
   { assert(frame::interpreter_frame_monitor_size() == BasicObjectLock::size(), "adjust code below");
     int monitor_offset = BytesPerWord * method()->max_locals() +
-      (BasicObjectLock::size() * BytesPerWord) * (number_of_locks - 1);
+      (2 * BytesPerWord) * (number_of_locks - 1);
+    // SharedRuntime::OSR_migration_begin() packs BasicObjectLocks in
+    // the OSR buffer using 2 word entries: first the lock and then
+    // the oop.
     for (int i = 0; i < number_of_locks; i++) {
-      int slot_offset = monitor_offset - ((i * BasicObjectLock::size()) * BytesPerWord);
+      int slot_offset = monitor_offset - ((i * 2) * BytesPerWord);
 #ifdef ASSERT
       // verify the interpreter's monitor has a non-null object
       {
         Label L;
-        __ ld_ptr(OSR_buf, slot_offset + BasicObjectLock::obj_offset_in_bytes(), O7);
+        __ ld_ptr(OSR_buf, slot_offset + 1*BytesPerWord, O7);
         __ cmp(G0, O7);
         __ br(Assembler::notEqual, false, Assembler::pt, L);
         __ delayed()->nop();
@@ -205,9 +208,9 @@ void LIR_Assembler::osr_entry() {
       }
 #endif // ASSERT
       // Copy the lock field into the compiled activation.
-      __ ld_ptr(OSR_buf, slot_offset + BasicObjectLock::lock_offset_in_bytes(), O7);
+      __ ld_ptr(OSR_buf, slot_offset + 0, O7);
       __ st_ptr(O7, frame_map()->address_for_monitor_lock(i));
-      __ ld_ptr(OSR_buf, slot_offset + BasicObjectLock::obj_offset_in_bytes(), O7);
+      __ ld_ptr(OSR_buf, slot_offset + 1*BytesPerWord, O7);
       __ st_ptr(O7, frame_map()->address_for_monitor_object(i));
     }
   }
@@ -953,9 +956,11 @@ int LIR_Assembler::load(Register base, int offset, LIR_Opr to_reg, BasicType typ
         } else {
 #ifdef _LP64
           assert(base != to_reg->as_register_lo(), "can't handle this");
+          assert(O7 != to_reg->as_register_lo(), "can't handle this");
           __ ld(base, offset + hi_word_offset_in_bytes, to_reg->as_register_lo());
+          __ lduw(base, offset + lo_word_offset_in_bytes, O7); // in case O7 is base or offset, use it last
           __ sllx(to_reg->as_register_lo(), 32, to_reg->as_register_lo());
-          __ ld(base, offset + lo_word_offset_in_bytes, to_reg->as_register_lo());
+          __ or3(to_reg->as_register_lo(), O7, to_reg->as_register_lo());
 #else
           if (base == to_reg->as_register_lo()) {
             __ ld(base, offset + hi_word_offset_in_bytes, to_reg->as_register_hi());
@@ -976,8 +981,8 @@ int LIR_Assembler::load(Register base, int offset, LIR_Opr to_reg, BasicType typ
           FloatRegister reg = to_reg->as_double_reg();
           // split unaligned loads
           if (unaligned || PatchALot) {
-            __ ldf(FloatRegisterImpl::S, base, offset + BytesPerWord, reg->successor());
+            __ ldf(FloatRegisterImpl::S, base, offset + 4, reg->successor());
-            __ ldf(FloatRegisterImpl::S, base, offset,                reg);
+            __ ldf(FloatRegisterImpl::S, base, offset,     reg);
           } else {
             __ ldf(FloatRegisterImpl::D, base, offset, to_reg->as_double_reg());
           }
@@ -2200,6 +2205,7 @@ void LIR_Assembler::emit_arraycopy(LIR_OpArrayCopy* op) {
   Register len     = O2;
 
   __ add(src, arrayOopDesc::base_offset_in_bytes(basic_type), src_ptr);
+  LP64_ONLY(__ sra(src_pos, 0, src_pos);) //higher 32bits must be null
   if (shift == 0) {
     __ add(src_ptr, src_pos, src_ptr);
   } else {
@@ -2208,6 +2214,7 @@ void LIR_Assembler::emit_arraycopy(LIR_OpArrayCopy* op) {
   }
 
   __ add(dst, arrayOopDesc::base_offset_in_bytes(basic_type), dst_ptr);
+  LP64_ONLY(__ sra(dst_pos, 0, dst_pos);) //higher 32bits must be null
   if (shift == 0) {
     __ add(dst_ptr, dst_pos, dst_ptr);
   } else {

hotspot/src/cpu/sparc/vm/c1_LIRGenerator_sparc.cpp

@@ -144,17 +144,17 @@ LIR_Address* LIRGenerator::generate_address(LIR_Opr base, LIR_Opr index,
   if (index->is_register()) {
     // apply the shift and accumulate the displacement
     if (shift > 0) {
-      LIR_Opr tmp = new_register(T_INT);
+      LIR_Opr tmp = new_pointer_register();
       __ shift_left(index, shift, tmp);
       index = tmp;
     }
     if (disp != 0) {
-      LIR_Opr tmp = new_register(T_INT);
+      LIR_Opr tmp = new_pointer_register();
       if (Assembler::is_simm13(disp)) {
-        __ add(tmp, LIR_OprFact::intConst(disp), tmp);
+        __ add(tmp, LIR_OprFact::intptrConst(disp), tmp);
         index = tmp;
       } else {
-        __ move(LIR_OprFact::intConst(disp), tmp);
+        __ move(LIR_OprFact::intptrConst(disp), tmp);
         __ add(tmp, index, tmp);
         index = tmp;
       }
@@ -162,8 +162,8 @@ LIR_Address* LIRGenerator::generate_address(LIR_Opr base, LIR_Opr index,
     }
   } else if (disp != 0 && !Assembler::is_simm13(disp)) {
     // index is illegal so replace it with the displacement loaded into a register
-    index = new_register(T_INT);
+    index = new_pointer_register();
-    __ move(LIR_OprFact::intConst(disp), index);
+    __ move(LIR_OprFact::intptrConst(disp), index);
     disp = 0;
   }
 

hotspot/src/cpu/sparc/vm/c1_globals_sparc.hpp

@@ -22,10 +22,9 @@
  *
  */
 
-//
 // Sets the default values for platform dependent flags used by the client compiler.
 // (see c1_globals.hpp)
-//
+
 #ifndef TIERED
 define_pd_global(bool, BackgroundCompilation,        true );
 define_pd_global(bool, CICompileOSR,                 true );
@@ -48,27 +47,24 @@ define_pd_global(intx, OnStackReplacePercentage,     1400 );
 define_pd_global(bool, UseTLAB,                      true );
 define_pd_global(bool, ProfileInterpreter,           false);
 define_pd_global(intx, FreqInlineSize,               325  );
-define_pd_global(intx, NewRatio,                     8    ); // Design center runs on 1.3.1
 define_pd_global(bool, ResizeTLAB,                   true );
 define_pd_global(intx, ReservedCodeCacheSize,        32*M );
 define_pd_global(intx, CodeCacheExpansionSize,       32*K );
 define_pd_global(uintx,CodeCacheMinBlockLength,      1);
-define_pd_global(uintx, PermSize,                    12*M );
+define_pd_global(uintx,PermSize,                     12*M );
-define_pd_global(uintx, MaxPermSize,                 64*M );
+define_pd_global(uintx,MaxPermSize,                  64*M );
-define_pd_global(bool, NeverActAsServerClassMachine, true);
+define_pd_global(bool, NeverActAsServerClassMachine, true );
 define_pd_global(intx, NewSizeThreadIncrease,        16*K );
-define_pd_global(uintx, DefaultMaxRAM,               1*G);
+define_pd_global(uint64_t,MaxRAM,                    1ULL*G);
 define_pd_global(intx, InitialCodeCacheSize,         160*K);
-#endif // TIERED
+#endif // !TIERED
 
 define_pd_global(bool, UseTypeProfile,               false);
 define_pd_global(bool, RoundFPResults,               false);
 
-
+define_pd_global(bool, LIRFillDelaySlots,            true );
-define_pd_global(bool, LIRFillDelaySlots,            true);
 define_pd_global(bool, OptimizeSinglePrecision,      false);
-define_pd_global(bool, CSEArrayLength,               true);
+define_pd_global(bool, CSEArrayLength,               true );
 define_pd_global(bool, TwoOperandLIRForm,            false);
 
-
+define_pd_global(intx, SafepointPollOffset,          0    );
-define_pd_global(intx, SafepointPollOffset, 0);

hotspot/src/cpu/sparc/vm/c2_globals_sparc.hpp

@@ -59,7 +59,6 @@ define_pd_global(intx, FLOATPRESSURE,                52);  // C2 on V9 gets to u
 define_pd_global(intx, FreqInlineSize,               175);
 define_pd_global(intx, INTPRESSURE,                  48);  // large register set
 define_pd_global(intx, InteriorEntryAlignment,       16);  // = CodeEntryAlignment
-define_pd_global(intx, NewRatio,                     2);
 define_pd_global(intx, NewSizeThreadIncrease, ScaleForWordSize(4*K));
 // The default setting 16/16 seems to work best.
 // (For _228_jack 16/16 is 2% better than 4/4, 16/4, 32/32, 32/16, or 16/32.)
@@ -83,25 +82,25 @@ define_pd_global(bool, OptoScheduling,               true);
 // sequence of instructions to load a 64 bit pointer.
 //
 // InitialCodeCacheSize derived from specjbb2000 run.
-define_pd_global(intx, InitialCodeCacheSize,     2048*K); // Integral multiple of CodeCacheExpansionSize
+define_pd_global(intx, InitialCodeCacheSize,         2048*K); // Integral multiple of CodeCacheExpansionSize
-define_pd_global(intx, ReservedCodeCacheSize,    48*M);
+define_pd_global(intx, ReservedCodeCacheSize,        48*M);
-define_pd_global(intx, CodeCacheExpansionSize,   64*K);
+define_pd_global(intx, CodeCacheExpansionSize,       64*K);
 
 // Ergonomics related flags
-define_pd_global(uintx, DefaultMaxRAM,           32*G);
+define_pd_global(uint64_t,MaxRAM,                    128ULL*G);
 #else
 // InitialCodeCacheSize derived from specjbb2000 run.
-define_pd_global(intx, InitialCodeCacheSize,     1536*K); // Integral multiple of CodeCacheExpansionSize
+define_pd_global(intx, InitialCodeCacheSize,         1536*K); // Integral multiple of CodeCacheExpansionSize
-define_pd_global(intx, ReservedCodeCacheSize,    32*M);
+define_pd_global(intx, ReservedCodeCacheSize,        32*M);
-define_pd_global(intx, CodeCacheExpansionSize,   32*K);
+define_pd_global(intx, CodeCacheExpansionSize,       32*K);
 // Ergonomics related flags
-define_pd_global(uintx, DefaultMaxRAM,           1*G);
+define_pd_global(uint64_t,MaxRAM,                    4ULL*G);
 #endif
-define_pd_global(uintx,CodeCacheMinBlockLength,  4);
+define_pd_global(uintx,CodeCacheMinBlockLength,      4);
 
 // Heap related flags
-define_pd_global(uintx, PermSize,    ScaleForWordSize(16*M));
+define_pd_global(uintx,PermSize,    ScaleForWordSize(16*M));
-define_pd_global(uintx, MaxPermSize, ScaleForWordSize(64*M));
+define_pd_global(uintx,MaxPermSize, ScaleForWordSize(64*M));
 
 // Ergonomics related flags
 define_pd_global(bool, NeverActAsServerClassMachine, false);

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

@@ -22,10 +22,8 @@
  *
  */
 
-//
 // Sets the default values for platform dependent flags used by the runtime system.
 // (see globals.hpp)
-//
 
 // For sparc we do not do call backs when a thread is in the interpreter, because the
 // interpreter dispatch needs at least two instructions - first to load the dispatch address
@@ -41,26 +39,23 @@ define_pd_global(bool, NeedsDeoptSuspend,           true); // register window ma
 define_pd_global(bool, ImplicitNullChecks,          true);  // Generate code for implicit null checks
 define_pd_global(bool, UncommonNullCast,            true);  // Uncommon-trap NULLs past to check cast
 
-define_pd_global(intx,  CodeEntryAlignment,    32);
+define_pd_global(intx, CodeEntryAlignment,    32);
-define_pd_global(uintx, TLABSize,              0);
+define_pd_global(intx, InlineFrequencyCount,  50);  // we can use more inlining on the SPARC
-define_pd_global(uintx, NewSize, ScaleForWordSize((2048 * K) + (2 * (64 * K))));
+define_pd_global(intx, InlineSmallCode,       1500);
-define_pd_global(intx,  SurvivorRatio,         8);
-define_pd_global(intx,  InlineFrequencyCount,  50);  // we can use more inlining on the SPARC
-define_pd_global(intx,  InlineSmallCode,       1500);
 #ifdef _LP64
 // Stack slots are 2X larger in LP64 than in the 32 bit VM.
-define_pd_global(intx,  ThreadStackSize,       1024);
+define_pd_global(intx, ThreadStackSize,       1024);
-define_pd_global(intx,  VMThreadStackSize,     1024);
+define_pd_global(intx, VMThreadStackSize,     1024);
 #else
-define_pd_global(intx,  ThreadStackSize,       512);
+define_pd_global(intx, ThreadStackSize,       512);
-define_pd_global(intx,  VMThreadStackSize,     512);
+define_pd_global(intx, VMThreadStackSize,     512);
 #endif
 
 define_pd_global(intx, StackYellowPages, 2);
 define_pd_global(intx, StackRedPages, 1);
 define_pd_global(intx, StackShadowPages, 3 DEBUG_ONLY(+1));
 
-define_pd_global(intx,  PreInflateSpin,        40);  // Determined by running design center
+define_pd_global(intx, PreInflateSpin,       40);  // Determined by running design center
 
 define_pd_global(bool, RewriteBytecodes,     true);
 define_pd_global(bool, RewriteFrequentPairs, true);

hotspot/src/cpu/sparc/vm/templateInterpreter_sparc.cpp

@@ -150,8 +150,7 @@ address TemplateInterpreterGenerator::generate_StackOverflowError_handler() {
 }
 
 
-address TemplateInterpreterGenerator::generate_return_entry_for(TosState state, int step, bool unbox) {
+address TemplateInterpreterGenerator::generate_return_entry_for(TosState state, int step) {
-  assert(!unbox, "NYI");//6815692//
   address compiled_entry = __ pc();
   Label cont;
 

hotspot/src/cpu/x86/vm/assembler_x86.cpp

@@ -2251,6 +2251,7 @@ void Assembler::popf() {
   emit_byte(0x9D);
 }
 
+#ifndef _LP64 // no 32bit push/pop on amd64
 void Assembler::popl(Address dst) {
   // NOTE: this will adjust stack by 8byte on 64bits
   InstructionMark im(this);
@@ -2258,6 +2259,7 @@ void Assembler::popl(Address dst) {
   emit_byte(0x8F);
   emit_operand(rax, dst);
 }
+#endif
 
 void Assembler::prefetch_prefix(Address src) {
   prefix(src);
@@ -2428,6 +2430,7 @@ void Assembler::pushf() {
   emit_byte(0x9C);
 }
 
+#ifndef _LP64 // no 32bit push/pop on amd64
 void Assembler::pushl(Address src) {
   // Note this will push 64bit on 64bit
   InstructionMark im(this);
@@ -2435,6 +2438,7 @@ void Assembler::pushl(Address src) {
   emit_byte(0xFF);
   emit_operand(rsi, src);
 }
+#endif
 
 void Assembler::pxor(XMMRegister dst, Address src) {
   NOT_LP64(assert(VM_Version::supports_sse2(), ""));
@@ -5591,7 +5595,12 @@ void MacroAssembler::align(int modulus) {
 }
 
 void MacroAssembler::andpd(XMMRegister dst, AddressLiteral src) {
-  andpd(dst, as_Address(src));
+  if (reachable(src)) {
+    andpd(dst, as_Address(src));
+  } else {
+    lea(rscratch1, src);
+    andpd(dst, Address(rscratch1, 0));
+  }
 }
 
 void MacroAssembler::andptr(Register dst, int32_t imm32) {
@@ -6078,11 +6087,21 @@ void MacroAssembler::cmpxchgptr(Register reg, Address adr) {
 }
 
 void MacroAssembler::comisd(XMMRegister dst, AddressLiteral src) {
-  comisd(dst, as_Address(src));
+  if (reachable(src)) {
+    comisd(dst, as_Address(src));
+  } else {
+    lea(rscratch1, src);
+    comisd(dst, Address(rscratch1, 0));
+  }
 }
 
 void MacroAssembler::comiss(XMMRegister dst, AddressLiteral src) {
-  comiss(dst, as_Address(src));
+  if (reachable(src)) {
+    comiss(dst, as_Address(src));
+  } else {
+    lea(rscratch1, src);
+    comiss(dst, Address(rscratch1, 0));
+  }
 }
 
 
@@ -7647,7 +7666,7 @@ RegisterOrConstant MacroAssembler::delayed_value_impl(intptr_t* delayed_value_ad
 
 #ifdef ASSERT
   Label L;
-  testl(tmp, tmp);
+  testptr(tmp, tmp);
   jccb(Assembler::notZero, L);
   hlt();
   bind(L);

hotspot/src/cpu/x86/vm/assembler_x86.hpp

@@ -1244,7 +1244,9 @@ private:
   void pcmpestri(XMMRegister xmm1, XMMRegister xmm2, int imm8);
   void pcmpestri(XMMRegister xmm1, Address src, int imm8);
 
+#ifndef _LP64 // no 32bit push/pop on amd64
   void popl(Address dst);
+#endif
 
 #ifdef _LP64
   void popq(Address dst);
@@ -1285,7 +1287,9 @@ private:
   // Interleave Low Bytes
   void punpcklbw(XMMRegister dst, XMMRegister src);
 
+#ifndef _LP64 // no 32bit push/pop on amd64
   void pushl(Address src);
+#endif
 
   void pushq(Address src);
 

hotspot/src/cpu/x86/vm/c1_LIRAssembler_x86.cpp

@@ -301,22 +301,25 @@ void LIR_Assembler::osr_entry() {
   Register OSR_buf = osrBufferPointer()->as_pointer_register();
   { assert(frame::interpreter_frame_monitor_size() == BasicObjectLock::size(), "adjust code below");
     int monitor_offset = BytesPerWord * method()->max_locals() +
-      (BasicObjectLock::size() * BytesPerWord) * (number_of_locks - 1);
+      (2 * BytesPerWord) * (number_of_locks - 1);
+    // SharedRuntime::OSR_migration_begin() packs BasicObjectLocks in
+    // the OSR buffer using 2 word entries: first the lock and then
+    // the oop.
     for (int i = 0; i < number_of_locks; i++) {
-      int slot_offset = monitor_offset - ((i * BasicObjectLock::size()) * BytesPerWord);
+      int slot_offset = monitor_offset - ((i * 2) * BytesPerWord);
 #ifdef ASSERT
       // verify the interpreter's monitor has a non-null object
       {
         Label L;
-        __ cmpptr(Address(OSR_buf, slot_offset + BasicObjectLock::obj_offset_in_bytes()), (int32_t)NULL_WORD);
+        __ cmpptr(Address(OSR_buf, slot_offset + 1*BytesPerWord), (int32_t)NULL_WORD);
         __ jcc(Assembler::notZero, L);
         __ stop("locked object is NULL");
         __ bind(L);
       }
 #endif
-      __ movptr(rbx, Address(OSR_buf, slot_offset + BasicObjectLock::lock_offset_in_bytes()));
+      __ movptr(rbx, Address(OSR_buf, slot_offset + 0));
       __ movptr(frame_map()->address_for_monitor_lock(i), rbx);
-      __ movptr(rbx, Address(OSR_buf, slot_offset + BasicObjectLock::obj_offset_in_bytes()));
+      __ movptr(rbx, Address(OSR_buf, slot_offset + 1*BytesPerWord));
       __ movptr(frame_map()->address_for_monitor_object(i), rbx);
     }
   }
@@ -785,7 +788,13 @@ void LIR_Assembler::const2mem(LIR_Opr src, LIR_Opr dest, BasicType type, CodeEmi
           ShouldNotReachHere();
           __ movoop(as_Address(addr, noreg), c->as_jobject());
         } else {
+#ifdef _LP64
+          __ movoop(rscratch1, c->as_jobject());
+          null_check_here = code_offset();
+          __ movptr(as_Address_lo(addr), rscratch1);
+#else
           __ movoop(as_Address(addr), c->as_jobject());
+#endif
         }
       }
       break;
@@ -1118,8 +1127,14 @@ void LIR_Assembler::stack2stack(LIR_Opr src, LIR_Opr dest, BasicType type) {
       __ pushptr(frame_map()->address_for_slot(src ->single_stack_ix()));
       __ popptr (frame_map()->address_for_slot(dest->single_stack_ix()));
     } else {
+#ifndef _LP64
       __ pushl(frame_map()->address_for_slot(src ->single_stack_ix()));
       __ popl (frame_map()->address_for_slot(dest->single_stack_ix()));
+#else
+      //no pushl on 64bits
+      __ movl(rscratch1, frame_map()->address_for_slot(src ->single_stack_ix()));
+      __ movl(frame_map()->address_for_slot(dest->single_stack_ix()), rscratch1);
+#endif
     }
 
   } else if (src->is_double_stack()) {
@@ -3136,8 +3151,10 @@ void LIR_Assembler::emit_arraycopy(LIR_OpArrayCopy* op) {
 
 #ifdef _LP64
   assert_different_registers(c_rarg0, dst, dst_pos, length);
+  __ movl2ptr(src_pos, src_pos); //higher 32bits must be null
   __ lea(c_rarg0, Address(src, src_pos, scale, arrayOopDesc::base_offset_in_bytes(basic_type)));
   assert_different_registers(c_rarg1, length);
+  __ movl2ptr(dst_pos, dst_pos); //higher 32bits must be null
   __ lea(c_rarg1, Address(dst, dst_pos, scale, arrayOopDesc::base_offset_in_bytes(basic_type)));
   __ mov(c_rarg2, length);
 

hotspot/src/cpu/x86/vm/c1_LIRGenerator_x86.cpp

@@ -755,8 +755,19 @@ void LIRGenerator::do_CompareAndSwap(Intrinsic* x, ValueType* type) {
   }
 
   LIR_Opr addr = new_pointer_register();
-  __ move(obj.result(), addr);
+  LIR_Address* a;
-  __ add(addr, offset.result(), addr);
+  if(offset.result()->is_constant()) {
+    a = new LIR_Address(obj.result(),
+                        NOT_LP64(offset.result()->as_constant_ptr()->as_jint()) LP64_ONLY((int)offset.result()->as_constant_ptr()->as_jlong()),
+                        as_BasicType(type));
+  } else {
+    a = new LIR_Address(obj.result(),
+                        offset.result(),
+                        LIR_Address::times_1,
+                        0,
+                        as_BasicType(type));
+  }
+  __ leal(LIR_OprFact::address(a), addr);
 
   if (type == objectType) {  // Write-barrier needed for Object fields.
     // Do the pre-write barrier, if any.

hotspot/src/cpu/x86/vm/c1_globals_x86.hpp

@@ -22,10 +22,8 @@
  *
  */
 
-//
 // Sets the default values for platform dependent flags used by the client compiler.
 // (see c1_globals.hpp)
-//
 
 #ifndef TIERED
 define_pd_global(bool, BackgroundCompilation,        true );
@@ -48,27 +46,24 @@ define_pd_global(intx, Tier4BackEdgeThreshold,       100000);
 
 define_pd_global(intx, OnStackReplacePercentage,     933  );
 define_pd_global(intx, FreqInlineSize,               325  );
-define_pd_global(intx, NewRatio,                     12   );
 define_pd_global(intx, NewSizeThreadIncrease,        4*K  );
 define_pd_global(intx, InitialCodeCacheSize,         160*K);
 define_pd_global(intx, ReservedCodeCacheSize,        32*M );
 define_pd_global(bool, ProfileInterpreter,           false);
 define_pd_global(intx, CodeCacheExpansionSize,       32*K );
 define_pd_global(uintx,CodeCacheMinBlockLength,      1);
-define_pd_global(uintx, PermSize,                    12*M );
+define_pd_global(uintx,PermSize,                     12*M );
-define_pd_global(uintx, MaxPermSize,                 64*M );
+define_pd_global(uintx,MaxPermSize,                  64*M );
-define_pd_global(bool, NeverActAsServerClassMachine, true);
+define_pd_global(bool, NeverActAsServerClassMachine, true );
-define_pd_global(uintx, DefaultMaxRAM,               1*G);
+define_pd_global(uint64_t,MaxRAM,                    1ULL*G);
 define_pd_global(bool, CICompileOSR,                 true );
-#endif // TIERED
+#endif // !TIERED
 define_pd_global(bool, UseTypeProfile,               false);
 define_pd_global(bool, RoundFPResults,               true );
 
-
 define_pd_global(bool, LIRFillDelaySlots,            false);
-define_pd_global(bool, OptimizeSinglePrecision,      true);
+define_pd_global(bool, OptimizeSinglePrecision,      true );
 define_pd_global(bool, CSEArrayLength,               false);
-define_pd_global(bool, TwoOperandLIRForm,            true);
+define_pd_global(bool, TwoOperandLIRForm,            true );
 
-
+define_pd_global(intx, SafepointPollOffset,          256  );
-define_pd_global(intx, SafepointPollOffset, 256);

hotspot/src/cpu/x86/vm/c2_globals_x86.hpp

@@ -22,7 +22,6 @@
  *
  */
 
-//
 // Sets the default values for platform dependent flags used by the server compiler.
 // (see c2_globals.hpp).  Alpha-sorted.
 
@@ -46,8 +45,8 @@ define_pd_global(intx, CompileThreshold,             1000);
 define_pd_global(intx, CompileThreshold,             10000);
 #endif // TIERED
 define_pd_global(intx, Tier2CompileThreshold,        10000);
-define_pd_global(intx, Tier3CompileThreshold,        20000 );
+define_pd_global(intx, Tier3CompileThreshold,        20000);
-define_pd_global(intx, Tier4CompileThreshold,        40000 );
+define_pd_global(intx, Tier4CompileThreshold,        40000);
 
 define_pd_global(intx, BackEdgeThreshold,            100000);
 define_pd_global(intx, Tier2BackEdgeThreshold,       100000);
@@ -61,7 +60,6 @@ define_pd_global(intx, FreqInlineSize,               325);
 #ifdef AMD64
 define_pd_global(intx, INTPRESSURE,                  13);
 define_pd_global(intx, InteriorEntryAlignment,       16);
-define_pd_global(intx, NewRatio,                     2);
 define_pd_global(intx, NewSizeThreadIncrease, ScaleForWordSize(4*K));
 define_pd_global(intx, LoopUnrollLimit,              60);
 // InitialCodeCacheSize derived from specjbb2000 run.
@@ -69,19 +67,18 @@ define_pd_global(intx, InitialCodeCacheSize,         2496*K); // Integral multip
 define_pd_global(intx, CodeCacheExpansionSize,       64*K);
 
 // Ergonomics related flags
-define_pd_global(uintx, DefaultMaxRAM, 32*G);
+define_pd_global(uint64_t,MaxRAM,                    128ULL*G);
 #else
 define_pd_global(intx, INTPRESSURE,                  6);
 define_pd_global(intx, InteriorEntryAlignment,       4);
-define_pd_global(intx, NewRatio,                     8); // Design center runs on 1.3.1
 define_pd_global(intx, NewSizeThreadIncrease,        4*K);
-define_pd_global(intx, LoopUnrollLimit,              50); // Design center runs on 1.3.1
+define_pd_global(intx, LoopUnrollLimit,              50);     // Design center runs on 1.3.1
 // InitialCodeCacheSize derived from specjbb2000 run.
 define_pd_global(intx, InitialCodeCacheSize,         2304*K); // Integral multiple of CodeCacheExpansionSize
 define_pd_global(intx, CodeCacheExpansionSize,       32*K);
 
 // Ergonomics related flags
-define_pd_global(uintx, DefaultMaxRAM, 1*G);
+define_pd_global(uint64_t,MaxRAM,                    4ULL*G);
 #endif // AMD64
 define_pd_global(intx, OptoLoopAlignment,            16);
 define_pd_global(intx, RegisterCostAreaRatio,        16000);
@@ -97,8 +94,8 @@ define_pd_global(intx, ReservedCodeCacheSize,        48*M);
 define_pd_global(uintx,CodeCacheMinBlockLength,      4);
 
 // Heap related flags
-define_pd_global(uintx, PermSize,    ScaleForWordSize(16*M));
+define_pd_global(uintx,PermSize,    ScaleForWordSize(16*M));
-define_pd_global(uintx, MaxPermSize, ScaleForWordSize(64*M));
+define_pd_global(uintx,MaxPermSize, ScaleForWordSize(64*M));
 
 // Ergonomics related flags
 define_pd_global(bool, NeverActAsServerClassMachine, false);

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

@@ -22,17 +22,16 @@
  *
  */
 
-//
 // Sets the default values for platform dependent flags used by the runtime system.
 // (see globals.hpp)
-//
 
-define_pd_global(bool,  ConvertSleepToYield,      true);
+define_pd_global(bool, ConvertSleepToYield,      true);
-define_pd_global(bool,  ShareVtableStubs,         true);
+define_pd_global(bool, ShareVtableStubs,         true);
-define_pd_global(bool,  CountInterpCalls,         true);
+define_pd_global(bool, CountInterpCalls,         true);
+define_pd_global(bool, NeedsDeoptSuspend,        false); // only register window machines need this
 
-define_pd_global(bool, ImplicitNullChecks,          true);  // Generate code for implicit null checks
+define_pd_global(bool, ImplicitNullChecks,       true);  // Generate code for implicit null checks
-define_pd_global(bool, UncommonNullCast,            true);  // Uncommon-trap NULLs past to check cast
+define_pd_global(bool, UncommonNullCast,         true);  // Uncommon-trap NULLs past to check cast
 
 // See 4827828 for this change. There is no globals_core_i486.hpp. I can't
 // assign a different value for C2 without touching a number of files. Use
@@ -42,29 +41,24 @@ define_pd_global(bool, UncommonNullCast,            true);  // Uncommon-trap NUL
 // the uep and the vep doesn't get real alignment but just slops on by
 // only assured that the entry instruction meets the 5 byte size requirement.
 #ifdef COMPILER2
-define_pd_global(intx,  CodeEntryAlignment,       32);
+define_pd_global(intx, CodeEntryAlignment,       32);
 #else
-define_pd_global(intx,  CodeEntryAlignment,       16);
+define_pd_global(intx, CodeEntryAlignment,       16);
 #endif // COMPILER2
+define_pd_global(intx, InlineFrequencyCount,     100);
+define_pd_global(intx, InlineSmallCode,          1000);
 
-define_pd_global(bool, NeedsDeoptSuspend,           false); // only register window machines need this
+define_pd_global(intx, StackYellowPages, 2);
-
+define_pd_global(intx, StackRedPages, 1);
-define_pd_global(uintx, TLABSize,                 0);
 #ifdef AMD64
-define_pd_global(uintx, NewSize, ScaleForWordSize(2048 * K));
 // Very large C++ stack frames using solaris-amd64 optimized builds
 // due to lack of optimization caused by C++ compiler bugs
 define_pd_global(intx, StackShadowPages, SOLARIS_ONLY(20) NOT_SOLARIS(6) DEBUG_ONLY(+2));
 #else
-define_pd_global(uintx, NewSize,                  1024 * K);
 define_pd_global(intx, StackShadowPages, 3 DEBUG_ONLY(+1));
 #endif // AMD64
-define_pd_global(intx,  InlineFrequencyCount,     100);
-define_pd_global(intx,  InlineSmallCode,          1000);
-define_pd_global(intx,  PreInflateSpin,           10);
 
-define_pd_global(intx, StackYellowPages, 2);
+define_pd_global(intx, PreInflateSpin,           10);
-define_pd_global(intx, StackRedPages, 1);
 
 define_pd_global(bool, RewriteBytecodes,     true);
 define_pd_global(bool, RewriteFrequentPairs, true);

hotspot/src/cpu/x86/vm/interp_masm_x86_32.cpp

@@ -196,6 +196,9 @@ void InterpreterMacroAssembler::get_cache_index_at_bcp(Register reg, int bcp_off
   } else {
     assert(EnableInvokeDynamic, "giant index used only for EnableInvokeDynamic");
     movl(reg, Address(rsi, bcp_offset));
+    // Check if the secondary index definition is still ~x, otherwise
+    // we have to change the following assembler code to calculate the
+    // plain index.
     assert(constantPoolCacheOopDesc::decode_secondary_index(~123) == 123, "else change next line");
     notl(reg);  // convert to plain index
   }

hotspot/src/cpu/x86/vm/interp_masm_x86_64.cpp

@@ -185,12 +185,30 @@ void InterpreterMacroAssembler::get_unsigned_2_byte_index_at_bcp(
 }
 
 
+void InterpreterMacroAssembler::get_cache_index_at_bcp(Register index,
+                                                       int bcp_offset,
+                                                       bool giant_index) {
+  assert(bcp_offset > 0, "bcp is still pointing to start of bytecode");
+  if (!giant_index) {
+    load_unsigned_short(index, Address(r13, bcp_offset));
+  } else {
+    assert(EnableInvokeDynamic, "giant index used only for EnableInvokeDynamic");
+    movl(index, Address(r13, bcp_offset));
+    // Check if the secondary index definition is still ~x, otherwise
+    // we have to change the following assembler code to calculate the
+    // plain index.
+    assert(constantPoolCacheOopDesc::decode_secondary_index(~123) == 123, "else change next line");
+    notl(index);  // convert to plain index
+  }
+}
+
+
 void InterpreterMacroAssembler::get_cache_and_index_at_bcp(Register cache,
                                                            Register index,
-                                                           int bcp_offset) {
+                                                           int bcp_offset,
-  assert(bcp_offset > 0, "bcp is still pointing to start of bytecode");
+                                                           bool giant_index) {
   assert(cache != index, "must use different registers");
-  load_unsigned_short(index, Address(r13, bcp_offset));
+  get_cache_index_at_bcp(index, bcp_offset, giant_index);
   movptr(cache, Address(rbp, frame::interpreter_frame_cache_offset * wordSize));
   assert(sizeof(ConstantPoolCacheEntry) == 4 * wordSize, "adjust code below");
   // convert from field index to ConstantPoolCacheEntry index
@@ -200,10 +218,10 @@ void InterpreterMacroAssembler::get_cache_and_index_at_bcp(Register cache,
 
 void InterpreterMacroAssembler::get_cache_entry_pointer_at_bcp(Register cache,
                                                                Register tmp,
-                                                               int bcp_offset) {
+                                                               int bcp_offset,
-  assert(bcp_offset > 0, "bcp is still pointing to start of bytecode");
+                                                               bool giant_index) {
   assert(cache != tmp, "must use different register");
-  load_unsigned_short(tmp, Address(r13, bcp_offset));
+  get_cache_index_at_bcp(tmp, bcp_offset, giant_index);
   assert(sizeof(ConstantPoolCacheEntry) == 4 * wordSize, "adjust code below");
   // convert from field index to ConstantPoolCacheEntry index
   // and from word offset to byte offset
@@ -1236,7 +1254,8 @@ void InterpreterMacroAssembler::profile_final_call(Register mdp) {
 
 void InterpreterMacroAssembler::profile_virtual_call(Register receiver,
                                                      Register mdp,
-                                                     Register reg2) {
+                                                     Register reg2,
+                                                     bool receiver_can_be_null) {
   if (ProfileInterpreter) {
     Label profile_continue;
 
@@ -1246,8 +1265,15 @@ void InterpreterMacroAssembler::profile_virtual_call(Register receiver,
     // We are making a call.  Increment the count.
     increment_mdp_data_at(mdp, in_bytes(CounterData::count_offset()));
 
+    Label skip_receiver_profile;
+    if (receiver_can_be_null) {
+      testptr(receiver, receiver);
+      jcc(Assembler::zero, skip_receiver_profile);
+    }
+
     // Record the receiver type.
     record_klass_in_profile(receiver, mdp, reg2);
+    bind(skip_receiver_profile);
 
     // The method data pointer needs to be updated to reflect the new target.
     update_mdp_by_constant(mdp,

hotspot/src/cpu/x86/vm/interp_masm_x86_64.hpp

@@ -95,9 +95,10 @@ class InterpreterMacroAssembler: public MacroAssembler {
 
   void get_unsigned_2_byte_index_at_bcp(Register reg, int bcp_offset);
   void get_cache_and_index_at_bcp(Register cache, Register index,
-                                  int bcp_offset);
+                                  int bcp_offset, bool giant_index = false);
   void get_cache_entry_pointer_at_bcp(Register cache, Register tmp,
-                                      int bcp_offset);
+                                      int bcp_offset, bool giant_index = false);
+  void get_cache_index_at_bcp(Register index, int bcp_offset, bool giant_index = false);
 
 
   void pop_ptr(Register r = rax);
@@ -236,7 +237,8 @@ class InterpreterMacroAssembler: public MacroAssembler {
   void profile_call(Register mdp);
   void profile_final_call(Register mdp);
   void profile_virtual_call(Register receiver, Register mdp,
-                            Register scratch2);
+                            Register scratch2,
+                            bool receiver_can_be_null = false);
   void profile_ret(Register return_bci, Register mdp);
   void profile_null_seen(Register mdp);
   void profile_typecheck(Register mdp, Register klass, Register scratch);

hotspot/src/cpu/x86/vm/interpreter_x86_64.cpp

@@ -277,12 +277,11 @@ address InterpreterGenerator::generate_abstract_entry(void) {
   address entry_point = __ pc();
 
   // abstract method entry
-  // remove return address. Not really needed, since exception
-  // handling throws away expression stack
-  __ pop(rbx);
 
-  // adjust stack to what a normal return would do
+  //  pop return address, reset last_sp to NULL
-  __ mov(rsp, r13);
+  __ empty_expression_stack();
+  __ restore_bcp();      // rsi must be correct for exception handler   (was destroyed)
+  __ restore_locals();   // make sure locals pointer is correct as well (was destroyed)
 
   // throw exception
   __ call_VM(noreg, CAST_FROM_FN_PTR(address,
@@ -300,7 +299,10 @@ address InterpreterGenerator::generate_method_handle_entry(void) {
   if (!EnableMethodHandles) {
     return generate_abstract_entry();
   }
-  return generate_abstract_entry(); //6815692//
+
+  address entry_point = MethodHandles::generate_method_handle_interpreter_entry(_masm);
+
+  return entry_point;
 }
 
 

hotspot/src/cpu/x86/vm/methodHandles_x86.cpp

@@ -448,7 +448,7 @@ void MethodHandles::generate_method_handle_stub(MacroAssembler* _masm, MethodHan
                                 rbx_index, Address::times_ptr,
                                 base + vtableEntry::method_offset_in_bytes());
       Register rbx_method = rbx_temp;
-      __ movl(rbx_method, vtable_entry_addr);
+      __ movptr(rbx_method, vtable_entry_addr);
 
       __ verify_oop(rbx_method);
       __ jmp(rbx_method_fie);

hotspot/src/cpu/x86/vm/stubGenerator_x86_64.cpp

@@ -2935,6 +2935,16 @@ class StubGenerator: public StubCodeGenerator {
 
     // arraycopy stubs used by compilers
     generate_arraycopy_stubs();
+
+    // generic method handle stubs
+    if (EnableMethodHandles && SystemDictionary::MethodHandle_klass() != NULL) {
+      for (MethodHandles::EntryKind ek = MethodHandles::_EK_FIRST;
+           ek < MethodHandles::_EK_LIMIT;
+           ek = MethodHandles::EntryKind(1 + (int)ek)) {
+        StubCodeMark mark(this, "MethodHandle", MethodHandles::entry_name(ek));
+        MethodHandles::generate_method_handle_stub(_masm, ek);
+      }
+    }
   }
 
  public:

hotspot/src/cpu/x86/vm/templateInterpreter_x86_32.cpp

@@ -155,15 +155,8 @@ address TemplateInterpreterGenerator::generate_continuation_for(TosState state)
 }
 
 
-address TemplateInterpreterGenerator::generate_return_entry_for(TosState state, int step, bool unbox) {
+address TemplateInterpreterGenerator::generate_return_entry_for(TosState state, int step) {
   TosState incoming_state = state;
-  if (EnableInvokeDynamic) {
-    if (unbox) {
-      incoming_state = atos;
-    }
-  } else {
-    assert(!unbox, "old behavior");
-  }
 
   Label interpreter_entry;
   address compiled_entry = __ pc();
@@ -216,46 +209,6 @@ address TemplateInterpreterGenerator::generate_return_entry_for(TosState state,
   __ restore_bcp();
   __ restore_locals();
 
-  Label L_fail;
-
-  if (unbox && state != atos) {
-    // cast and unbox
-    BasicType type = as_BasicType(state);
-    if (type == T_BYTE)  type = T_BOOLEAN; // FIXME
-    KlassHandle boxk = SystemDictionaryHandles::box_klass(type);
-    __ mov32(rbx, ExternalAddress((address) boxk.raw_value()));
-    __ testl(rax, rax);
-    Label L_got_value, L_get_value;
-    // convert nulls to zeroes (avoid NPEs here)
-    if (!(type == T_FLOAT || type == T_DOUBLE)) {
-      // if rax already contains zero bits, forge ahead
-      __ jcc(Assembler::zero, L_got_value);
-    } else {
-      __ jcc(Assembler::notZero, L_get_value);
-      __ fldz();
-      __ jmp(L_got_value);
-    }
-    __ bind(L_get_value);
-    __ cmp32(rbx, Address(rax, oopDesc::klass_offset_in_bytes()));
-    __ jcc(Assembler::notEqual, L_fail);
-    int offset = java_lang_boxing_object::value_offset_in_bytes(type);
-    // Cf. TemplateTable::getfield_or_static
-    switch (type) {
-      case T_BYTE:     // fall through:
-      case T_BOOLEAN:  __ load_signed_byte(rax, Address(rax, offset));    break;
-      case T_CHAR:     __ load_unsigned_short(rax, Address(rax, offset)); break;
-      case T_SHORT:    __ load_signed_short(rax, Address(rax, offset));   break;
-      case T_INT:      __ movl(rax, Address(rax, offset));                break;
-      case T_FLOAT:    __ fld_s(Address(rax, offset));                    break;
-      case T_DOUBLE:   __ fld_d(Address(rax, offset));                    break;
-      // Access to java.lang.Double.value does not need to be atomic:
-      case T_LONG:   { __ movl(rdx, Address(rax, offset + 4));
-                       __ movl(rax, Address(rax, offset + 0));  }         break;
-      default: ShouldNotReachHere();
-    }
-    __ bind(L_got_value);
-  }
-
   Label L_got_cache, L_giant_index;
   if (EnableInvokeDynamic) {
     __ cmpb(Address(rsi, 0), Bytecodes::_invokedynamic);
@@ -263,32 +216,6 @@ address TemplateInterpreterGenerator::generate_return_entry_for(TosState state,
   }
   __ get_cache_and_index_at_bcp(rbx, rcx, 1, false);
   __ bind(L_got_cache);
-  if (unbox && state == atos) {
-    // insert a casting conversion, to keep verifier sane
-    Label L_ok, L_ok_pops;
-    __ testl(rax, rax);
-    __ jcc(Assembler::zero, L_ok);
-    __ push(rax);               // save the object to check
-    __ push(rbx);               // save CP cache reference
-    __ movl(rdx, Address(rax, oopDesc::klass_offset_in_bytes()));
-    __ movl(rbx, Address(rbx, rcx,
-                      Address::times_4, constantPoolCacheOopDesc::base_offset() +
-                      ConstantPoolCacheEntry::f1_offset()));
-    __ movl(rbx, Address(rbx, __ delayed_value(sun_dyn_CallSiteImpl::type_offset_in_bytes, rcx)));
-    __ movl(rbx, Address(rbx, __ delayed_value(java_dyn_MethodType::rtype_offset_in_bytes, rcx)));
-    __ movl(rax, Address(rbx, __ delayed_value(java_lang_Class::klass_offset_in_bytes, rcx)));
-    __ check_klass_subtype(rdx, rax, rbx, L_ok_pops);
-    __ pop(rcx);                // pop and discard CP cache
-    __ mov(rbx, rax);           // target supertype into rbx for L_fail
-    __ pop(rax);                // failed object into rax for L_fail
-    __ jmp(L_fail);
-
-    __ bind(L_ok_pops);
-    // restore pushed temp regs:
-    __ pop(rbx);
-    __ pop(rax);
-    __ bind(L_ok);
-  }
   __ movl(rbx, Address(rbx, rcx,
                     Address::times_ptr, constantPoolCacheOopDesc::base_offset() +
                     ConstantPoolCacheEntry::flags_offset()));
@@ -301,14 +228,6 @@ address TemplateInterpreterGenerator::generate_return_entry_for(TosState state,
     __ bind(L_giant_index);
     __ get_cache_and_index_at_bcp(rbx, rcx, 1, true);
     __ jmp(L_got_cache);
-
-    if (unbox) {
-      __ bind(L_fail);
-      __ push(rbx);             // missed klass (required)
-      __ push(rax);             // bad object (actual)
-      __ movptr(rdx, ExternalAddress((address) &Interpreter::_throw_WrongMethodType_entry));
-      __ call(rdx);
-    }
   }
 
   return entry;

hotspot/src/cpu/x86/vm/templateInterpreter_x86_64.cpp

@@ -100,21 +100,26 @@ address TemplateInterpreterGenerator::generate_ClassCastException_handler() {
   return entry;
 }
 
-// Arguments are: required type in rarg1, failing object (or NULL) in rarg2
+// Arguments are: required type at TOS+8, failing object (or NULL) at TOS+4.
 address TemplateInterpreterGenerator::generate_WrongMethodType_handler() {
   address entry = __ pc();
 
   __ pop(c_rarg2);              // failing object is at TOS
   __ pop(c_rarg1);              // required type is at TOS+8
 
-  // expression stack must be empty before entering the VM if an
+  __ verify_oop(c_rarg1);
-  // exception happened
+  __ verify_oop(c_rarg2);
+
+  // Various method handle types use interpreter registers as temps.
+  __ restore_bcp();
+  __ restore_locals();
+
+  // Expression stack must be empty before entering the VM for an exception.
   __ empty_expression_stack();
 
   __ call_VM(noreg,
              CAST_FROM_FN_PTR(address,
-                              InterpreterRuntime::
+                              InterpreterRuntime::throw_WrongMethodTypeException),
-                              throw_WrongMethodTypeException),
              // pass required type, failing object (or NULL)
              c_rarg1, c_rarg2);
   return entry;
@@ -166,8 +171,7 @@ address TemplateInterpreterGenerator::generate_continuation_for(TosState state)
 
 
 address TemplateInterpreterGenerator::generate_return_entry_for(TosState state,
-                                                                int step, bool unbox) {
+                                                                int step) {
-  assert(!unbox, "NYI");//6815692//
 
   // amd64 doesn't need to do anything special about compiled returns
   // to the interpreter so the code that exists on x86 to place a sentinel
@@ -183,15 +187,29 @@ address TemplateInterpreterGenerator::generate_return_entry_for(TosState state,
   __ restore_bcp();
   __ restore_locals();
 
-  __ get_cache_and_index_at_bcp(rbx, rcx, 1);
+  Label L_got_cache, L_giant_index;
+  if (EnableInvokeDynamic) {
+    __ cmpb(Address(r13, 0), Bytecodes::_invokedynamic);
+    __ jcc(Assembler::equal, L_giant_index);
+  }
+  __ get_cache_and_index_at_bcp(rbx, rcx, 1, false);
+  __ bind(L_got_cache);
   __ movl(rbx, Address(rbx, rcx,
-                       Address::times_8,
+                       Address::times_ptr,
                        in_bytes(constantPoolCacheOopDesc::base_offset()) +
                        3 * wordSize));
   __ andl(rbx, 0xFF);
   if (TaggedStackInterpreter) __ shll(rbx, 1); // 2 slots per parameter.
   __ lea(rsp, Address(rsp, rbx, Address::times_8));
   __ dispatch_next(state, step);
+
+  // out of the main line of code...
+  if (EnableInvokeDynamic) {
+    __ bind(L_giant_index);
+    __ get_cache_and_index_at_bcp(rbx, rcx, 1, true);
+    __ jmp(L_got_cache);
+  }
+
   return entry;
 }
 

hotspot/src/cpu/x86/vm/templateTable_x86_32.cpp

@@ -2890,9 +2890,6 @@ void TemplateTable::count_calls(Register method, Register temp) {
 
 
 void TemplateTable::prepare_invoke(Register method, Register index, int byte_no) {
-  bool is_invdyn_bootstrap = (byte_no < 0);
-  if (is_invdyn_bootstrap)  byte_no = -byte_no;
-
   // determine flags
   Bytecodes::Code code = bytecode();
   const bool is_invokeinterface  = code == Bytecodes::_invokeinterface;
@@ -2907,8 +2904,6 @@ void TemplateTable::prepare_invoke(Register method, Register index, int byte_no)
   const Register flags  = rdx;
   assert_different_registers(method, index, recv, flags);
 
-  assert(!is_invdyn_bootstrap || is_invokedynamic, "byte_no<0 hack only for invdyn");
-
   // save 'interpreter return address'
   __ save_bcp();
 
@@ -2944,9 +2939,7 @@ void TemplateTable::prepare_invoke(Register method, Register index, int byte_no)
   // load return address
   {
     address table_addr;
-    if (is_invdyn_bootstrap)
+    if (is_invokeinterface || is_invokedynamic)
-      table_addr = (address)Interpreter::return_5_unbox_addrs_by_index_table();
-    else if (is_invokeinterface || is_invokedynamic)
       table_addr = (address)Interpreter::return_5_addrs_by_index_table();
     else
       table_addr = (address)Interpreter::return_3_addrs_by_index_table();
@@ -3153,54 +3146,10 @@ void TemplateTable::invokedynamic(int byte_no) {
     __ profile_call(rsi);
   }
 
-  Label handle_unlinked_site;
+  __ movptr(rcx, Address(rax, __ delayed_value(java_dyn_CallSite::target_offset_in_bytes, rcx)));
-  __ movptr(rcx, Address(rax, __ delayed_value(sun_dyn_CallSiteImpl::target_offset_in_bytes, rcx)));
+  __ null_check(rcx);
-  __ testptr(rcx, rcx);
-  __ jcc(Assembler::zero, handle_unlinked_site);
-
   __ prepare_to_jump_from_interpreted();
   __ jump_to_method_handle_entry(rcx, rdx);
-
-  // Initial calls come here...
-  __ bind(handle_unlinked_site);
-  __ pop(rcx);                 // remove return address pushed by prepare_invoke
-
-  // box stacked arguments into an array for the bootstrap method
-  address entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::bootstrap_invokedynamic);
-  __ restore_bcp();      // rsi must be correct for call_VM
-  __ call_VM(rax, entry, rax);
-  __ movl(rdi, rax);            // protect bootstrap MH from prepare_invoke
-
-  // recompute return address
-  __ restore_bcp();      // rsi must be correct for prepare_invoke
-  prepare_invoke(rax, rbx, -byte_no);  // smashes rcx, rdx
-  // rax: CallSite object (f1)
-  // rbx: unused (f2)
-  // rdi: bootstrap MH
-  // rdx: flags
-
-  // now load up the arglist, which has been neatly boxed
-  __ get_thread(rcx);
-  __ movptr(rdx, Address(rcx, JavaThread::vm_result_2_offset()));
-  __ movptr(Address(rcx, JavaThread::vm_result_2_offset()), NULL_WORD);
-  __ verify_oop(rdx);
-  // rdx = arglist
-
-  // save SP now, before we add the bootstrap call to the stack
-  // We must preserve a fiction that the original arguments are outgoing,
-  // because the return sequence will reset the stack to this point
-  // and then pop all those arguments.  It seems error-prone to use
-  // a different argument list size just for bootstrapping.
-  __ prepare_to_jump_from_interpreted();
-
-  // Now let's play adapter, pushing the real arguments on the stack.
-  __ pop(rbx);                  // return PC
-  __ push(rdi);                 // boot MH
-  __ push(rax);                 // call site
-  __ push(rdx);                 // arglist
-  __ push(rbx);                 // return PC, again
-  __ mov(rcx, rdi);
-  __ jump_to_method_handle_entry(rcx, rdx);
 }
 
 //----------------------------------------------------------------------------------------------------

hotspot/src/cpu/x86/vm/templateTable_x86_64.cpp

@@ -203,18 +203,15 @@ void TemplateTable::patch_bytecode(Bytecodes::Code bytecode, Register bc,
     __ jcc(Assembler::notEqual, fast_patch);
     __ get_method(scratch);
     // Let breakpoint table handling rewrite to quicker bytecode
-    __ call_VM(noreg,
+    __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::set_original_bytecode_at), scratch, r13, bc);
-               CAST_FROM_FN_PTR(address,
-                                InterpreterRuntime::set_original_bytecode_at),
-               scratch, r13, bc);
 #ifndef ASSERT
     __ jmpb(patch_done);
-    __ bind(fast_patch);
-  }
 #else
     __ jmp(patch_done);
+#endif
     __ bind(fast_patch);
   }
+#ifdef ASSERT
   Label okay;
   __ load_unsigned_byte(scratch, at_bcp(0));
   __ cmpl(scratch, (int) Bytecodes::java_code(bytecode));
@@ -2054,26 +2051,28 @@ void TemplateTable::volatile_barrier(Assembler::Membar_mask_bits
   }
 }
 
-void TemplateTable::resolve_cache_and_index(int byte_no,
+void TemplateTable::resolve_cache_and_index(int byte_no, Register Rcache, Register index) {
-                                            Register Rcache,
-                                            Register index) {
   assert(byte_no == 1 || byte_no == 2, "byte_no out of range");
+  bool is_invokedynamic = (bytecode() == Bytecodes::_invokedynamic);
 
   const Register temp = rbx;
   assert_different_registers(Rcache, index, temp);
 
   const int shift_count = (1 + byte_no) * BitsPerByte;
   Label resolved;
-  __ get_cache_and_index_at_bcp(Rcache, index, 1);
+  __ get_cache_and_index_at_bcp(Rcache, index, 1, is_invokedynamic);
-  __ movl(temp, Address(Rcache,
+  if (is_invokedynamic) {
-                        index, Address::times_8,
+    // we are resolved if the f1 field contains a non-null CallSite object
-                        constantPoolCacheOopDesc::base_offset() +
+    __ cmpptr(Address(Rcache, index, Address::times_ptr, constantPoolCacheOopDesc::base_offset() + ConstantPoolCacheEntry::f1_offset()), (int32_t) NULL_WORD);
-                        ConstantPoolCacheEntry::indices_offset()));
+    __ jcc(Assembler::notEqual, resolved);
-  __ shrl(temp, shift_count);
+  } else {
-  // have we resolved this bytecode?
+    __ movl(temp, Address(Rcache, index, Address::times_ptr, constantPoolCacheOopDesc::base_offset() + ConstantPoolCacheEntry::indices_offset()));
-  __ andl(temp, 0xFF);
+    __ shrl(temp, shift_count);
-  __ cmpl(temp, (int) bytecode());
+    // have we resolved this bytecode?
-  __ jcc(Assembler::equal, resolved);
+    __ andl(temp, 0xFF);
+    __ cmpl(temp, (int) bytecode());
+    __ jcc(Assembler::equal, resolved);
+  }
 
   // resolve first time through
   address entry;
@@ -2090,6 +2089,9 @@ void TemplateTable::resolve_cache_and_index(int byte_no,
   case Bytecodes::_invokeinterface:
     entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_invoke);
     break;
+  case Bytecodes::_invokedynamic:
+    entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_invokedynamic);
+    break;
   default:
     ShouldNotReachHere();
     break;
@@ -2098,7 +2100,7 @@ void TemplateTable::resolve_cache_and_index(int byte_no,
   __ call_VM(noreg, entry, temp);
 
   // Update registers with resolved info
-  __ get_cache_and_index_at_bcp(Rcache, index, 1);
+  __ get_cache_and_index_at_bcp(Rcache, index, 1, is_invokedynamic);
   __ bind(resolved);
 }
 
@@ -2832,15 +2834,14 @@ void TemplateTable::count_calls(Register method, Register temp) {
   ShouldNotReachHere();
 }
 
-void TemplateTable::prepare_invoke(Register method,
+void TemplateTable::prepare_invoke(Register method, Register index, int byte_no) {
-                                   Register index,
-                                   int byte_no,
-                                   Bytecodes::Code code) {
   // determine flags
+  Bytecodes::Code code = bytecode();
   const bool is_invokeinterface  = code == Bytecodes::_invokeinterface;
+  const bool is_invokedynamic    = code == Bytecodes::_invokedynamic;
   const bool is_invokevirtual    = code == Bytecodes::_invokevirtual;
   const bool is_invokespecial    = code == Bytecodes::_invokespecial;
-  const bool load_receiver       = code != Bytecodes::_invokestatic;
+  const bool load_receiver      = (code != Bytecodes::_invokestatic && code != Bytecodes::_invokedynamic);
   const bool receiver_null_check = is_invokespecial;
   const bool save_flags = is_invokeinterface || is_invokevirtual;
   // setup registers & access constant pool cache
@@ -2858,9 +2859,13 @@ void TemplateTable::prepare_invoke(Register method,
     __ movl(recv, flags);
     __ andl(recv, 0xFF);
     if (TaggedStackInterpreter) __ shll(recv, 1);  // index*2
-    __ movptr(recv, Address(rsp, recv, Address::times_8,
+    Address recv_addr(rsp, recv, Address::times_8, -Interpreter::expr_offset_in_bytes(1));
-                                 -Interpreter::expr_offset_in_bytes(1)));
+    if (is_invokedynamic) {
-    __ verify_oop(recv);
+      __ lea(recv, recv_addr);
+    } else {
+      __ movptr(recv, recv_addr);
+      __ verify_oop(recv);
+    }
   }
 
   // do null check if needed
@@ -2878,10 +2883,14 @@ void TemplateTable::prepare_invoke(Register method,
   ConstantPoolCacheEntry::verify_tosBits();
   // load return address
   {
-    ExternalAddress return_5((address)Interpreter::return_5_addrs_by_index_table());
+    address table_addr;
-    ExternalAddress return_3((address)Interpreter::return_3_addrs_by_index_table());
+    if (is_invokeinterface || is_invokedynamic)
-    __ lea(rscratch1, (is_invokeinterface ? return_5 : return_3));
+      table_addr = (address)Interpreter::return_5_addrs_by_index_table();
-    __ movptr(flags, Address(rscratch1, flags, Address::times_8));
+    else
+      table_addr = (address)Interpreter::return_3_addrs_by_index_table();
+    ExternalAddress table(table_addr);
+    __ lea(rscratch1, table);
+    __ movptr(flags, Address(rscratch1, flags, Address::times_ptr));
   }
 
   // push return address
@@ -2947,7 +2956,7 @@ void TemplateTable::invokevirtual_helper(Register index,
 
 void TemplateTable::invokevirtual(int byte_no) {
   transition(vtos, vtos);
-  prepare_invoke(rbx, noreg, byte_no, bytecode());
+  prepare_invoke(rbx, noreg, byte_no);
 
   // rbx: index
   // rcx: receiver
@@ -2959,7 +2968,7 @@ void TemplateTable::invokevirtual(int byte_no) {
 
 void TemplateTable::invokespecial(int byte_no) {
   transition(vtos, vtos);
-  prepare_invoke(rbx, noreg, byte_no, bytecode());
+  prepare_invoke(rbx, noreg, byte_no);
   // do the call
   __ verify_oop(rbx);
   __ profile_call(rax);
@@ -2969,7 +2978,7 @@ void TemplateTable::invokespecial(int byte_no) {
 
 void TemplateTable::invokestatic(int byte_no) {
   transition(vtos, vtos);
-  prepare_invoke(rbx, noreg, byte_no, bytecode());
+  prepare_invoke(rbx, noreg, byte_no);
   // do the call
   __ verify_oop(rbx);
   __ profile_call(rax);
@@ -2983,7 +2992,7 @@ void TemplateTable::fast_invokevfinal(int byte_no) {
 
 void TemplateTable::invokeinterface(int byte_no) {
   transition(vtos, vtos);
-  prepare_invoke(rax, rbx, byte_no, bytecode());
+  prepare_invoke(rax, rbx, byte_no);
 
   // rax: Interface
   // rbx: index
@@ -3072,7 +3081,24 @@ void TemplateTable::invokedynamic(int byte_no) {
     return;
   }
 
-  __ stop("invokedynamic NYI");//6815692//
+  prepare_invoke(rax, rbx, byte_no);
+
+  // rax: CallSite object (f1)
+  // rbx: unused (f2)
+  // rcx: receiver address
+  // rdx: flags (unused)
+
+  if (ProfileInterpreter) {
+    Label L;
+    // %%% should make a type profile for any invokedynamic that takes a ref argument
+    // profile this call
+    __ profile_call(r13);
+  }
+
+  __ movptr(rcx, Address(rax, __ delayed_value(java_dyn_CallSite::target_offset_in_bytes, rcx)));
+  __ null_check(rcx);
+  __ prepare_to_jump_from_interpreted();
+  __ jump_to_method_handle_entry(rcx, rdx);
 }
 
 

hotspot/src/cpu/x86/vm/templateTable_x86_64.hpp

@@ -22,8 +22,7 @@
  *
  */
 
-  static void prepare_invoke(Register method, Register index, int byte_no,
+  static void prepare_invoke(Register method, Register index, int byte_no);
-                             Bytecodes::Code code);
   static void invokevirtual_helper(Register index, Register recv,
                                    Register flags);
   static void volatile_barrier(Assembler::Membar_mask_bits order_constraint);

hotspot/src/cpu/x86/vm/vm_version_x86.cpp

@@ -255,6 +255,8 @@ void VM_Version::get_processor_features() {
   if (!VM_Version::supports_sse2()) {
     vm_exit_during_initialization("Unknown x64 processor: SSE2 not supported");
   }
+  // in 64 bit the use of SSE2 is the minimum
+  if (UseSSE < 2) UseSSE = 2;
 #endif
 
   // If the OS doesn't support SSE, we can't use this feature even if the HW does

hotspot/src/cpu/zero/vm/cppInterpreter_zero.cpp

@@ -204,6 +204,20 @@ void CppInterpreter::native_entry(methodOop method, intptr_t UNUSED, TRAPS) {
     goto unwind_and_return;
   }
 
+  // Update the invocation counter
+  if ((UseCompiler || CountCompiledCalls) && !method->is_synchronized()) {
+    thread->set_do_not_unlock();
+    InvocationCounter *counter = method->invocation_counter();
+    counter->increment();
+    if (counter->reached_InvocationLimit()) {
+      CALL_VM_NOCHECK(
+        InterpreterRuntime::frequency_counter_overflow(thread, NULL));
+      if (HAS_PENDING_EXCEPTION)
+        goto unwind_and_return;
+    }
+    thread->clr_do_not_unlock();
+  }
+
   // Lock if necessary
   BasicObjectLock *monitor;
   monitor = NULL;

hotspot/src/cpu/zero/vm/frame_zero.cpp

@@ -36,11 +36,8 @@ bool frame::is_interpreted_frame() const {
   return zeroframe()->is_interpreter_frame();
 }
 
-bool frame::is_fake_stub_frame() const {
-  return zeroframe()->is_fake_stub_frame();
-}
-
 frame frame::sender_for_entry_frame(RegisterMap *map) const {
+  assert(zeroframe()->is_entry_frame(), "wrong type of frame");
   assert(map != NULL, "map must be set");
   assert(!entry_frame_is_first(), "next Java fp must be non zero");
   assert(entry_frame_call_wrapper()->anchor()->last_Java_sp() == sender_sp(),
@@ -50,15 +47,10 @@ frame frame::sender_for_entry_frame(RegisterMap *map) const {
   return frame(sender_sp(), sp() + 1);
 }
 
-frame frame::sender_for_interpreter_frame(RegisterMap *map) const {
+frame frame::sender_for_nonentry_frame(RegisterMap *map) const {
-  return frame(sender_sp(), sp() + 1);
+  assert(zeroframe()->is_interpreter_frame() ||
-}
+         zeroframe()->is_shark_frame() ||
-
+         zeroframe()->is_fake_stub_frame(), "wrong type of frame");
-frame frame::sender_for_compiled_frame(RegisterMap *map) const {
-  return frame(sender_sp(), sp() + 1);
-}
-
-frame frame::sender_for_fake_stub_frame(RegisterMap *map) const {
   return frame(sender_sp(), sp() + 1);
 }
 
@@ -69,17 +61,8 @@ frame frame::sender(RegisterMap* map) const {
 
   if (is_entry_frame())
     return sender_for_entry_frame(map);
-
+  else
-  if (is_interpreted_frame())
+    return sender_for_nonentry_frame(map);
-    return sender_for_interpreter_frame(map);
-
-  if (is_compiled_frame())
-    return sender_for_compiled_frame(map);
-
-  if (is_fake_stub_frame())
-    return sender_for_fake_stub_frame(map);
-
-  ShouldNotReachHere();
 }
 
 #ifdef CC_INTERP

hotspot/src/cpu/zero/vm/frame_zero.hpp

@@ -65,10 +65,7 @@
   }
 
  public:
-  bool is_fake_stub_frame() const;
+  frame sender_for_nonentry_frame(RegisterMap* map) const;
-
- public:
-  frame sender_for_fake_stub_frame(RegisterMap* map) const;
 
  public:
   void zero_print_on_error(int           index,

hotspot/src/cpu/zero/vm/globals_zero.hpp

@@ -23,10 +23,8 @@
  *
  */
 
-//
 // Set the default values for platform dependent flags used by the
 // runtime system.  See globals.hpp for details of what they do.
-//
 
 define_pd_global(bool,  ConvertSleepToYield,  true);
 define_pd_global(bool,  ShareVtableStubs,     true);
@@ -37,14 +35,7 @@ define_pd_global(bool,  ImplicitNullChecks,   true);
 define_pd_global(bool,  UncommonNullCast,     true);
 
 define_pd_global(intx,  CodeEntryAlignment,   32);
-define_pd_global(uintx, TLABSize,             0);
-#ifdef _LP64
-define_pd_global(uintx, NewSize,              ScaleForWordSize(2048 * K));
-#else
-define_pd_global(uintx, NewSize,              ScaleForWordSize(1024 * K));
-#endif // _LP64
 define_pd_global(intx,  InlineFrequencyCount, 100);
-define_pd_global(intx,  InlineSmallCode,      1000);
 define_pd_global(intx,  PreInflateSpin,       10);
 
 define_pd_global(intx,  StackYellowPages,     2);

hotspot/src/cpu/zero/vm/sharedRuntime_zero.cpp

@@ -1,6 +1,6 @@
 /*
  * Copyright 2003-2007 Sun Microsystems, Inc.  All Rights Reserved.
- * Copyright 2007, 2008 Red Hat, Inc.
+ * Copyright 2007, 2008, 2009 Red Hat, Inc.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -61,7 +61,14 @@ nmethod *SharedRuntime::generate_native_wrapper(MacroAssembler *masm,
                                                 BasicType *in_sig_bt,
                                                 VMRegPair *in_regs,
                                                 BasicType ret_type) {
+#ifdef SHARK
+  return SharkCompiler::compiler()->generate_native_wrapper(masm,
+                                                            method,
+                                                            in_sig_bt,
+                                                            ret_type);
+#else
   ShouldNotCallThis();
+#endif // SHARK
 }
 
 int Deoptimization::last_frame_adjust(int callee_parameters,

hotspot/src/cpu/zero/vm/sharkFrame_zero.hpp

@@ -1,6 +1,6 @@
 /*
  * Copyright 2003-2007 Sun Microsystems, Inc.  All Rights Reserved.
- * Copyright 2008 Red Hat, Inc.
+ * Copyright 2008, 2009 Red Hat, Inc.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -41,7 +41,7 @@
 // |  ...               |
 
 class SharkFrame : public ZeroFrame {
-  friend class SharkFunction;
+  friend class SharkStack;
 
  private:
   SharkFrame() : ZeroFrame() {

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

@@ -223,8 +223,8 @@ static const char *unstable_chroot_error = "/proc file system not found.\n"
                      "environment on Linux when /proc filesystem is not mounted.";
 
 void os::Linux::initialize_system_info() {
-  _processor_count = sysconf(_SC_NPROCESSORS_CONF);
+  set_processor_count(sysconf(_SC_NPROCESSORS_CONF));
-  if (_processor_count == 1) {
+  if (processor_count() == 1) {
     pid_t pid = os::Linux::gettid();
     char fname[32];
     jio_snprintf(fname, sizeof(fname), "/proc/%d", pid);
@@ -236,7 +236,7 @@ void os::Linux::initialize_system_info() {
     }
   }
   _physical_memory = (julong)sysconf(_SC_PHYS_PAGES) * (julong)sysconf(_SC_PAGESIZE);
-  assert(_processor_count > 0, "linux error");
+  assert(processor_count() > 0, "linux error");
 }
 
 void os::init_system_properties_values() {
@@ -4683,6 +4683,7 @@ void Parker::park(bool isAbsolute, jlong time) {
   // Return immediately if a permit is available.
   if (_counter > 0) {
       _counter = 0 ;
+      OrderAccess::fence();
       return ;
   }
 
@@ -4725,6 +4726,7 @@ void Parker::park(bool isAbsolute, jlong time) {
     _counter = 0;
     status = pthread_mutex_unlock(_mutex);
     assert (status == 0, "invariant") ;
+    OrderAccess::fence();
     return;
   }
 
@@ -4765,6 +4767,7 @@ void Parker::park(bool isAbsolute, jlong time) {
     jt->java_suspend_self();
   }
 
+  OrderAccess::fence();
 }
 
 void Parker::unpark() {

hotspot/src/os/solaris/dtrace/libjvm_db.c

@@ -937,54 +937,56 @@ scope_desc_at(Nmethod_t *N, int32_t decode_offset, Vframe_t *vf)
   return err;
 }
 
-static int
+static int scopeDesc_chain(Nmethod_t *N) {
-scopeDesc_chain(Nmethod_t *N)
-{
   int32_t decode_offset = 0;
   int32_t err;
 
-  if (debug > 2)
+  if (debug > 2) {
-      fprintf(stderr, "\t scopeDesc_chain: BEGIN\n");
+    fprintf(stderr, "\t scopeDesc_chain: BEGIN\n");
+  }
 
   err = ps_pread(N->J->P, N->pc_desc + OFFSET_PcDesc_scope_decode_offset,
                  &decode_offset, SZ32);
   CHECK_FAIL(err);
 
   while (decode_offset > 0) {
-      if (debug > 2)
+    Vframe_t *vf = &N->vframes[N->vf_cnt];
-          fprintf(stderr, "\t scopeDesc_chain: decode_offset: %#x\n", decode_offset);
 
-      Vframe_t *vf = &N->vframes[N->vf_cnt];
+    if (debug > 2) {
+      fprintf(stderr, "\t scopeDesc_chain: decode_offset: %#x\n", decode_offset);
+    }
 
-      err = scope_desc_at(N, decode_offset, vf);
+    err = scope_desc_at(N, decode_offset, vf);
+    CHECK_FAIL(err);
+
+    if (vf->methodIdx > N->oops_len) {
+      fprintf(stderr, "\t scopeDesc_chain: (methodIdx > oops_len) !\n");
+      return -1;
+    }
+    err = read_pointer(N->J, N->nm + N->oops_beg + (vf->methodIdx-1)*POINTER_SIZE,
+                       &vf->methodOop);
+    CHECK_FAIL(err);
+
+    if (vf->methodOop) {
+      N->vf_cnt++;
+      err = line_number_from_bci(N->J, vf);
       CHECK_FAIL(err);
-
+      if (debug > 2) {
-      if (vf->methodIdx > N->oops_len) {
+        fprintf(stderr, "\t scopeDesc_chain: methodOop: %#8llx, line: %ld\n",
-          fprintf(stderr, "\t scopeDesc_chain: (methodIdx > oops_len) !\n");
+                vf->methodOop, vf->line);
-          return -1;
       }
-      err = read_pointer(N->J, N->nm + N->oops_beg + (vf->methodIdx-1)*POINTER_SIZE,
+    }
-                               &vf->methodOop);
+    decode_offset = vf->sender_decode_offset;
-      CHECK_FAIL(err);
+  }
-
+  if (debug > 2) {
-      if (vf->methodOop) {
+    fprintf(stderr, "\t scopeDesc_chain: END \n\n");
-          N->vf_cnt++;
-          err = line_number_from_bci(N->J, vf);
-          CHECK_FAIL(err);
-          if (debug > 2) {
-              fprintf(stderr, "\t scopeDesc_chain: methodOop: %#8llx, line: %ld\n",
-                              vf->methodOop, vf->line);
-          }
-      }
-      decode_offset = vf->sender_decode_offset;
   }
-  if (debug > 2)
-      fprintf(stderr, "\t scopeDesc_chain: END \n\n");
   return PS_OK;
 
  fail:
-  if (debug)
+  if (debug) {
-      fprintf(stderr, "\t scopeDesc_chain: FAIL \n\n");
+    fprintf(stderr, "\t scopeDesc_chain: FAIL \n\n");
+  }
   return err;
 }
 

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

@@ -457,7 +457,7 @@ static volatile int max_hrtime_lock = LOCK_FREE;     // Update counter with LSB
 
 
 void os::Solaris::initialize_system_info() {
-  _processor_count = sysconf(_SC_NPROCESSORS_CONF);
+  set_processor_count(sysconf(_SC_NPROCESSORS_CONF));
   _processors_online = sysconf (_SC_NPROCESSORS_ONLN);
   _physical_memory = (julong)sysconf(_SC_PHYS_PAGES) * (julong)sysconf(_SC_PAGESIZE);
 }
@@ -5803,6 +5803,7 @@ void Parker::park(bool isAbsolute, jlong time) {
   // Return immediately if a permit is available.
   if (_counter > 0) {
       _counter = 0 ;
+      OrderAccess::fence();
       return ;
   }
 
@@ -5846,6 +5847,7 @@ void Parker::park(bool isAbsolute, jlong time) {
     _counter = 0;
     status = os::Solaris::mutex_unlock(_mutex);
     assert (status == 0, "invariant") ;
+    OrderAccess::fence();
     return;
   }
 
@@ -5892,6 +5894,7 @@ void Parker::park(bool isAbsolute, jlong time) {
     jt->java_suspend_self();
   }
 
+  OrderAccess::fence();
 }
 
 void Parker::unpark() {

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

@@ -3150,7 +3150,7 @@ void os::win32::initialize_system_info() {
   _vm_allocation_granularity = si.dwAllocationGranularity;
   _processor_type  = si.dwProcessorType;
   _processor_level = si.wProcessorLevel;
-  _processor_count = si.dwNumberOfProcessors;
+  set_processor_count(si.dwNumberOfProcessors);
 
   MEMORYSTATUSEX ms;
   ms.dwLength = sizeof(ms);

hotspot/src/os_cpu/linux_x86/vm/globals_linux_x86.hpp

@@ -22,10 +22,9 @@
  *
  */
 
-//
 // Sets the default values for platform dependent flags used by the runtime system.
 // (see globals.hpp)
-//
+
 define_pd_global(bool, DontYieldALot,            false);
 #ifdef AMD64
 define_pd_global(intx, ThreadStackSize,          1024); // 0 => use system default
@@ -39,11 +38,10 @@ define_pd_global(intx, VMThreadStackSize,        512);
 #endif // AMD64
 
 define_pd_global(intx, CompilerThreadStackSize,  0);
-define_pd_global(intx, SurvivorRatio,            8);
 
-define_pd_global(uintx, JVMInvokeMethodSlack,    8192);
+define_pd_global(uintx,JVMInvokeMethodSlack,     8192);
 
 // Only used on 64 bit platforms
-define_pd_global(uintx, HeapBaseMinAddress,      2*G);
+define_pd_global(uintx,HeapBaseMinAddress,       2*G);
 // Only used on 64 bit Windows platforms
 define_pd_global(bool, UseVectoredExceptions,    false);

hotspot/src/os_cpu/solaris_x86/vm/globals_solaris_x86.hpp

@@ -22,31 +22,25 @@
  *
  */
 
-//
 // Sets the default values for platform dependent flags used by the runtime system.
 // (see globals.hpp)
-//
+
 define_pd_global(bool, DontYieldALot,            true); // Determined in the design center
 #ifdef AMD64
 define_pd_global(intx, ThreadStackSize,          1024); // 0 => use system default
 define_pd_global(intx, VMThreadStackSize,        1024);
-define_pd_global(intx, SurvivorRatio,            6);
+define_pd_global(uintx,JVMInvokeMethodSlack,     8*K);
-define_pd_global(uintx, JVMInvokeMethodSlack,    8*K);
 #else
-//  UseStackBanging is not pd
-// define_pd_global(bool, UseStackBanging,          true);
-
 // ThreadStackSize 320 allows TaggedStackInterpreter and a couple of test cases
 // to run while keeping the number of threads that can be created high.
 define_pd_global(intx, ThreadStackSize,          320);
 define_pd_global(intx, VMThreadStackSize,        512);
-define_pd_global(intx, SurvivorRatio,            8);
+define_pd_global(uintx,JVMInvokeMethodSlack,     10*K);
-define_pd_global(uintx, JVMInvokeMethodSlack,    10*K);
 #endif // AMD64
 
 define_pd_global(intx, CompilerThreadStackSize,  0);
 
 // Only used on 64 bit platforms
-define_pd_global(uintx, HeapBaseMinAddress,      256*M);
+define_pd_global(uintx,HeapBaseMinAddress,       256*M);
 // Only used on 64 bit Windows platforms
 define_pd_global(bool, UseVectoredExceptions,    false);

hotspot/src/os_cpu/windows_x86/vm/globals_windows_x86.hpp

@@ -22,10 +22,9 @@
  *
  */
 
-//
 // Sets the default values for platform dependent flags used by the runtime system.
 // (see globals.hpp)
-//
+
 define_pd_global(bool, DontYieldALot,            false);
 
 // Default stack size on Windows is determined by the executable (java.exe
@@ -35,8 +34,6 @@ define_pd_global(bool, DontYieldALot,            false);
 define_pd_global(intx, ThreadStackSize,          0); // 0 => use system default
 define_pd_global(intx, VMThreadStackSize,        0); // 0 => use system default
 
-define_pd_global(intx, SurvivorRatio,            8);
-
 #ifdef ASSERT
 define_pd_global(intx, CompilerThreadStackSize,  1024);
 #else

hotspot/src/share/vm/c1/c1_GraphBuilder.cpp

@@ -365,7 +365,7 @@ void BlockListBuilder::make_loop_header(BlockBegin* block) {
     if (_next_loop_index < 31) _next_loop_index++;
   } else {
     // block already marked as loop header
-    assert(is_power_of_2(_loop_map.at(block->block_id())), "exactly one bit must be set");
+    assert(is_power_of_2((unsigned int)_loop_map.at(block->block_id())), "exactly one bit must be set");
   }
 }
 

hotspot/src/share/vm/c1/c1_LIRGenerator.cpp

@@ -1855,12 +1855,26 @@ void LIRGenerator::do_UnsafeGetRaw(UnsafeGetRaw* x) {
     addr = new LIR_Address(base_op, index_op->as_jint(), dst_type);
   } else {
 #ifdef X86
+#ifdef _LP64
+    if (!index_op->is_illegal() && index_op->type() == T_INT) {
+      LIR_Opr tmp = new_pointer_register();
+      __ convert(Bytecodes::_i2l, index_op, tmp);
+      index_op = tmp;
+    }
+#endif
     addr = new LIR_Address(base_op, index_op, LIR_Address::Scale(log2_scale), 0, dst_type);
 #else
     if (index_op->is_illegal() || log2_scale == 0) {
+#ifdef _LP64
+      if (!index_op->is_illegal() && index_op->type() == T_INT) {
+        LIR_Opr tmp = new_pointer_register();
+        __ convert(Bytecodes::_i2l, index_op, tmp);
+        index_op = tmp;
+      }
+#endif
       addr = new LIR_Address(base_op, index_op, dst_type);
     } else {
-      LIR_Opr tmp = new_register(T_INT);
+      LIR_Opr tmp = new_pointer_register();
       __ shift_left(index_op, log2_scale, tmp);
       addr = new LIR_Address(base_op, tmp, dst_type);
     }
@@ -1915,10 +1929,25 @@ void LIRGenerator::do_UnsafePutRaw(UnsafePutRaw* x) {
   LIR_Opr index_op = idx.result();
   if (log2_scale != 0) {
     // temporary fix (platform dependent code without shift on Intel would be better)
-    index_op = new_register(T_INT);
+    index_op = new_pointer_register();
-    __ move(idx.result(), index_op);
+#ifdef _LP64
+    if(idx.result()->type() == T_INT) {
+      __ convert(Bytecodes::_i2l, idx.result(), index_op);
+    } else {
+#endif
+      __ move(idx.result(), index_op);
+#ifdef _LP64
+    }
+#endif
     __ shift_left(index_op, log2_scale, index_op);
   }
+#ifdef _LP64
+  else if(!index_op->is_illegal() && index_op->type() == T_INT) {
+    LIR_Opr tmp = new_pointer_register();
+    __ convert(Bytecodes::_i2l, index_op, tmp);
+    index_op = tmp;
+  }
+#endif
 
   LIR_Address* addr = new LIR_Address(base_op, index_op, x->basic_type());
   __ move(value.result(), addr);

hotspot/src/share/vm/c1/c1_LinearScan.cpp

@@ -2464,6 +2464,10 @@ int LinearScan::append_scope_value_for_constant(LIR_Opr opr, GrowableArray<Scope
 
     case T_LONG: // fall through
     case T_DOUBLE: {
+#ifdef _LP64
+      scope_values->append(&_int_0_scope_value);
+      scope_values->append(new ConstantLongValue(c->as_jlong_bits()));
+#else
       if (hi_word_offset_in_bytes > lo_word_offset_in_bytes) {
         scope_values->append(new ConstantIntValue(c->as_jint_hi_bits()));
         scope_values->append(new ConstantIntValue(c->as_jint_lo_bits()));
@@ -2471,7 +2475,7 @@ int LinearScan::append_scope_value_for_constant(LIR_Opr opr, GrowableArray<Scope
         scope_values->append(new ConstantIntValue(c->as_jint_lo_bits()));
         scope_values->append(new ConstantIntValue(c->as_jint_hi_bits()));
       }
-
+#endif
       return 2;
     }
 
@@ -2503,17 +2507,18 @@ int LinearScan::append_scope_value_for_operand(LIR_Opr opr, GrowableArray<ScopeV
   } else if (opr->is_single_cpu()) {
     bool is_oop = opr->is_oop_register();
     int cache_idx = opr->cpu_regnr() * 2 + (is_oop ? 1 : 0);
+    Location::Type int_loc_type = NOT_LP64(Location::normal) LP64_ONLY(Location::int_in_long);
 
     ScopeValue* sv = _scope_value_cache.at(cache_idx);
     if (sv == NULL) {
-      Location::Type loc_type = is_oop ? Location::oop : Location::normal;
+      Location::Type loc_type = is_oop ? Location::oop : int_loc_type;
       VMReg rname = frame_map()->regname(opr);
       sv = new LocationValue(Location::new_reg_loc(loc_type, rname));
       _scope_value_cache.at_put(cache_idx, sv);
     }
 
     // check if cached value is correct
-    DEBUG_ONLY(assert_equal(sv, new LocationValue(Location::new_reg_loc(is_oop ? Location::oop : Location::normal, frame_map()->regname(opr)))));
+    DEBUG_ONLY(assert_equal(sv, new LocationValue(Location::new_reg_loc(is_oop ? Location::oop : int_loc_type, frame_map()->regname(opr)))));
 
     scope_values->append(sv);
     return 1;

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

@@ -61,9 +61,11 @@ class BCEscapeAnalyzer : public ResourceObj {
   BCEscapeAnalyzer* _parent;
   int               _level;
 
+ public:
   class  ArgumentMap;
   class  StateInfo;
 
+ private:
   // helper functions
   bool is_argument(int i)    { return i >= 0 && i < _arg_size; }
 

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

@@ -46,6 +46,9 @@ ciInstanceKlass* ciEnv::_Throwable;
 ciInstanceKlass* ciEnv::_Thread;
 ciInstanceKlass* ciEnv::_OutOfMemoryError;
 ciInstanceKlass* ciEnv::_String;
+ciInstanceKlass* ciEnv::_StringBuffer;
+ciInstanceKlass* ciEnv::_StringBuilder;
+ciInstanceKlass* ciEnv::_Integer;
 
 ciSymbol*        ciEnv::_unloaded_cisymbol = NULL;
 ciInstanceKlass* ciEnv::_unloaded_ciinstance_klass = NULL;
@@ -110,6 +113,8 @@ ciEnv::ciEnv(CompileTask* task, int system_dictionary_modification_counter) {
   _ArrayIndexOutOfBoundsException_instance = NULL;
   _ArrayStoreException_instance = NULL;
   _ClassCastException_instance = NULL;
+  _the_null_string = NULL;
+  _the_min_jint_string = NULL;
 }
 
 ciEnv::ciEnv(Arena* arena) {
@@ -163,6 +168,8 @@ ciEnv::ciEnv(Arena* arena) {
   _ArrayIndexOutOfBoundsException_instance = NULL;
   _ArrayStoreException_instance = NULL;
   _ClassCastException_instance = NULL;
+  _the_null_string = NULL;
+  _the_min_jint_string = NULL;
 }
 
 ciEnv::~ciEnv() {
@@ -248,6 +255,22 @@ ciInstance* ciEnv::ClassCastException_instance() {
   return _ClassCastException_instance;
 }
 
+ciInstance* ciEnv::the_null_string() {
+  if (_the_null_string == NULL) {
+    VM_ENTRY_MARK;
+    _the_null_string = get_object(Universe::the_null_string())->as_instance();
+  }
+  return _the_null_string;
+}
+
+ciInstance* ciEnv::the_min_jint_string() {
+  if (_the_min_jint_string == NULL) {
+    VM_ENTRY_MARK;
+    _the_min_jint_string = get_object(Universe::the_min_jint_string())->as_instance();
+  }
+  return _the_min_jint_string;
+}
+
 // ------------------------------------------------------------------
 // ciEnv::get_method_from_handle
 ciMethod* ciEnv::get_method_from_handle(jobject method) {
@@ -690,10 +713,8 @@ ciMethod* ciEnv::get_method_by_index_impl(ciInstanceKlass* accessor,
   ciInstanceKlass* declared_holder = get_instance_klass_for_declared_method_holder(holder);
 
   // Get the method's name and signature.
-  int nt_index = cpool->name_and_type_ref_index_at(index);
-  int sig_index = cpool->signature_ref_index_at(nt_index);
   symbolOop name_sym = cpool->name_ref_at(index);
-  symbolOop sig_sym = cpool->symbol_at(sig_index);
+  symbolOop sig_sym  = cpool->signature_ref_at(index);
 
   if (holder_is_accessible) { // Our declared holder is loaded.
     instanceKlass* lookup = declared_holder->get_instanceKlass();

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

@@ -82,6 +82,9 @@ private:
   static ciInstanceKlass* _Thread;
   static ciInstanceKlass* _OutOfMemoryError;
   static ciInstanceKlass* _String;
+  static ciInstanceKlass* _StringBuffer;
+  static ciInstanceKlass* _StringBuilder;
+  static ciInstanceKlass* _Integer;
 
   static ciSymbol*        _unloaded_cisymbol;
   static ciInstanceKlass* _unloaded_ciinstance_klass;
@@ -97,6 +100,9 @@ private:
   ciInstance* _ArrayStoreException_instance;
   ciInstance* _ClassCastException_instance;
 
+  ciInstance* _the_null_string;      // The Java string "null"
+  ciInstance* _the_min_jint_string; // The Java string "-2147483648"
+
   // Look up a klass by name from a particular class loader (the accessor's).
   // If require_local, result must be defined in that class loader, or NULL.
   // If !require_local, a result from remote class loader may be reported,
@@ -310,6 +316,15 @@ public:
   ciInstanceKlass* String_klass() {
     return _String;
   }
+  ciInstanceKlass* StringBuilder_klass() {
+    return _StringBuilder;
+  }
+  ciInstanceKlass* StringBuffer_klass() {
+    return _StringBuffer;
+  }
+  ciInstanceKlass* Integer_klass() {
+    return _Integer;
+  }
   ciInstance* NullPointerException_instance() {
     assert(_NullPointerException_instance != NULL, "initialization problem");
     return _NullPointerException_instance;
@@ -324,6 +339,9 @@ public:
   ciInstance* ArrayStoreException_instance();
   ciInstance* ClassCastException_instance();
 
+  ciInstance* the_null_string();
+  ciInstance* the_min_jint_string();
+
   static ciSymbol* unloaded_cisymbol() {
     return _unloaded_cisymbol;
   }

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

@@ -340,6 +340,20 @@ ciField* ciInstanceKlass::get_field_by_offset(int field_offset, bool is_static)
   return field;
 }
 
+// ------------------------------------------------------------------
+// ciInstanceKlass::get_field_by_name
+ciField* ciInstanceKlass::get_field_by_name(ciSymbol* name, ciSymbol* signature, bool is_static) {
+  VM_ENTRY_MARK;
+  instanceKlass* k = get_instanceKlass();
+  fieldDescriptor fd;
+  klassOop def = k->find_field(name->get_symbolOop(), signature->get_symbolOop(), is_static, &fd);
+  if (def == NULL) {
+    return NULL;
+  }
+  ciField* field = new (CURRENT_THREAD_ENV->arena()) ciField(&fd);
+  return field;
+}
+
 // ------------------------------------------------------------------
 // ciInstanceKlass::non_static_fields.
 

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

@@ -148,6 +148,7 @@ public:
 
   ciInstanceKlass* get_canonical_holder(int offset);
   ciField* get_field_by_offset(int field_offset, bool is_static);
+  ciField* get_field_by_name(ciSymbol* name, ciSymbol* signature, bool is_static);
 
   GrowableArray<ciField*>* non_static_fields();
 

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

@@ -168,6 +168,15 @@ void ciObjectFactory::init_shared_objects() {
   ciEnv::_String =
     get(SystemDictionary::string_klass())
       ->as_instance_klass();
+  ciEnv::_StringBuffer =
+    get(SystemDictionary::stringBuffer_klass())
+      ->as_instance_klass();
+  ciEnv::_StringBuilder =
+    get(SystemDictionary::StringBuilder_klass())
+      ->as_instance_klass();
+  ciEnv::_Integer =
+    get(SystemDictionary::int_klass())
+      ->as_instance_klass();
 
   for (int len = -1; len != _ci_objects->length(); ) {
     len = _ci_objects->length();

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

@@ -2511,23 +2511,12 @@ void ClassFileParser::java_dyn_MethodHandle_fix_pre(constantPoolHandle cp,
       fac_ptr->nonstatic_byte_count -= 1;
       (*fields_ptr)->ushort_at_put(i + instanceKlass::signature_index_offset,
                                    word_sig_index);
-      if (wordSize == jintSize) {
+      fac_ptr->nonstatic_word_count += 1;
-        fac_ptr->nonstatic_word_count += 1;
-      } else {
-        fac_ptr->nonstatic_double_count += 1;
-      }
 
-      FieldAllocationType atype = (FieldAllocationType) (*fields_ptr)->ushort_at(i+4);
+      FieldAllocationType atype = (FieldAllocationType) (*fields_ptr)->ushort_at(i + instanceKlass::low_offset);
       assert(atype == NONSTATIC_BYTE, "");
       FieldAllocationType new_atype = NONSTATIC_WORD;
-      if (wordSize > jintSize) {
+      (*fields_ptr)->ushort_at_put(i + instanceKlass::low_offset, new_atype);
-        if (Universe::field_type_should_be_aligned(T_LONG)) {
-          atype = NONSTATIC_ALIGNED_DOUBLE;
-        } else {
-          atype = NONSTATIC_DOUBLE;
-        }
-      }
-      (*fields_ptr)->ushort_at_put(i+4, new_atype);
 
       found_vmentry = true;
       break;
@@ -3085,7 +3074,7 @@ instanceKlassHandle ClassFileParser::parseClassFile(symbolHandle name,
     int len = fields->length();
     for (int i = 0; i < len; i += instanceKlass::next_offset) {
       int real_offset;
-      FieldAllocationType atype = (FieldAllocationType) fields->ushort_at(i+4);
+      FieldAllocationType atype = (FieldAllocationType) fields->ushort_at(i + instanceKlass::low_offset);
       switch (atype) {
         case STATIC_OOP:
           real_offset = next_static_oop_offset;
@@ -3173,8 +3162,8 @@ instanceKlassHandle ClassFileParser::parseClassFile(symbolHandle name,
         default:
           ShouldNotReachHere();
       }
-      fields->short_at_put(i+4, extract_low_short_from_int(real_offset) );
+      fields->short_at_put(i + instanceKlass::low_offset,  extract_low_short_from_int(real_offset));
-      fields->short_at_put(i+5, extract_high_short_from_int(real_offset) );
+      fields->short_at_put(i + instanceKlass::high_offset, extract_high_short_from_int(real_offset));
     }
 
     // Size of instances
@@ -3766,8 +3755,9 @@ bool ClassFileParser::has_illegal_visibility(jint flags) {
 }
 
 bool ClassFileParser::is_supported_version(u2 major, u2 minor) {
-  u2 max_version = JDK_Version::is_gte_jdk17x_version() ?
+  u2 max_version =
-    JAVA_MAX_SUPPORTED_VERSION : JAVA_6_VERSION;
+    JDK_Version::is_gte_jdk17x_version() ? JAVA_MAX_SUPPORTED_VERSION :
+    (JDK_Version::is_gte_jdk16x_version() ? JAVA_6_VERSION : JAVA_1_5_VERSION);
   return (major >= JAVA_MIN_SUPPORTED_VERSION) &&
          (major <= max_version) &&
          ((major != max_version) ||

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

@@ -1124,8 +1124,7 @@ class BacktraceBuilder: public StackObj {
     if (_dirty && _methods != NULL) {
       BarrierSet* bs = Universe::heap()->barrier_set();
       assert(bs->has_write_ref_array_opt(), "Barrier set must have ref array opt");
-      bs->write_ref_array(MemRegion((HeapWord*)_methods->base(),
+      bs->write_ref_array((HeapWord*)_methods->base(), _methods->length());
-                                    _methods->array_size()));
       _dirty = false;
     }
   }
@@ -2430,15 +2429,15 @@ oop java_dyn_MethodTypeForm::erasedType(oop mtform) {
 }
 
 
-// Support for sun_dyn_CallSiteImpl
+// Support for java_dyn_CallSite
 
-int sun_dyn_CallSiteImpl::_type_offset;
+int java_dyn_CallSite::_type_offset;
-int sun_dyn_CallSiteImpl::_target_offset;
+int java_dyn_CallSite::_target_offset;
-int sun_dyn_CallSiteImpl::_vmmethod_offset;
+int java_dyn_CallSite::_vmmethod_offset;
 
-void sun_dyn_CallSiteImpl::compute_offsets() {
+void java_dyn_CallSite::compute_offsets() {
   if (!EnableInvokeDynamic)  return;
-  klassOop k = SystemDictionary::CallSiteImpl_klass();
+  klassOop k = SystemDictionary::CallSite_klass();
   if (k != NULL) {
     compute_offset(_type_offset,   k, vmSymbols::type_name(),   vmSymbols::java_dyn_MethodType_signature(), true);
     compute_offset(_target_offset, k, vmSymbols::target_name(), vmSymbols::java_dyn_MethodHandle_signature(), true);
@@ -2446,23 +2445,23 @@ void sun_dyn_CallSiteImpl::compute_offsets() {
   }
 }
 
-oop sun_dyn_CallSiteImpl::type(oop site) {
+oop java_dyn_CallSite::type(oop site) {
   return site->obj_field(_type_offset);
 }
 
-oop sun_dyn_CallSiteImpl::target(oop site) {
+oop java_dyn_CallSite::target(oop site) {
   return site->obj_field(_target_offset);
 }
 
-void sun_dyn_CallSiteImpl::set_target(oop site, oop target) {
+void java_dyn_CallSite::set_target(oop site, oop target) {
   site->obj_field_put(_target_offset, target);
 }
 
-oop sun_dyn_CallSiteImpl::vmmethod(oop site) {
+oop java_dyn_CallSite::vmmethod(oop site) {
   return site->obj_field(_vmmethod_offset);
 }
 
-void sun_dyn_CallSiteImpl::set_vmmethod(oop site, oop ref) {
+void java_dyn_CallSite::set_vmmethod(oop site, oop ref) {
   site->obj_field_put(_vmmethod_offset, ref);
 }
 
@@ -2811,7 +2810,7 @@ void JavaClasses::compute_offsets() {
     java_dyn_MethodTypeForm::compute_offsets();
   }
   if (EnableInvokeDynamic) {
-    sun_dyn_CallSiteImpl::compute_offsets();
+    java_dyn_CallSite::compute_offsets();
   }
   java_security_AccessControlContext::compute_offsets();
   // Initialize reflection classes. The layouts of these classes

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

@@ -1061,9 +1061,9 @@ class java_dyn_MethodTypeForm: AllStatic {
 };
 
 
-// Interface to sun.dyn.CallSiteImpl objects
+// Interface to java.dyn.CallSite objects
 
-class sun_dyn_CallSiteImpl: AllStatic {
+class java_dyn_CallSite: AllStatic {
   friend class JavaClasses;
 
 private:

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

@@ -99,6 +99,15 @@ bool SystemDictionary::is_parallelCapable(Handle class_loader) {
   return java_lang_Class::parallelCapable(class_loader());
 }
 // ----------------------------------------------------------------------------
+// ParallelDefineClass flag does not apply to bootclass loader
+bool SystemDictionary::is_parallelDefine(Handle class_loader) {
+   if (class_loader.is_null()) return false;
+   if (AllowParallelDefineClass && java_lang_Class::parallelCapable(class_loader())) {
+     return true;
+   }
+   return false;
+}
+// ----------------------------------------------------------------------------
 // Resolving of classes
 
 // Forwards to resolve_or_null
@@ -724,13 +733,13 @@ klassOop SystemDictionary::resolve_instance_class_or_null(symbolHandle class_nam
       // Do actual loading
       k = load_instance_class(name, class_loader, THREAD);
 
-      // For UnsyncloadClass and AllowParallelDefineClass only:
+      // For UnsyncloadClass only
       // If they got a linkageError, check if a parallel class load succeeded.
       // If it did, then for bytecode resolution the specification requires
       // that we return the same result we did for the other thread, i.e. the
       // successfully loaded instanceKlass
       // Should not get here for classloaders that support parallelism
-      // with the new cleaner mechanism
+      // with the new cleaner mechanism, even with AllowParallelDefineClass
       // Bootstrap goes through here to allow for an extra guarantee check
       if (UnsyncloadClass || (class_loader.is_null())) {
         if (k.is_null() && HAS_PENDING_EXCEPTION
@@ -1483,14 +1492,17 @@ void SystemDictionary::define_instance_class(instanceKlassHandle k, TRAPS) {
 }
 
 // Support parallel classloading
-// Initial implementation for bootstrap classloader
+// All parallel class loaders, including bootstrap classloader
-// For custom class loaders that support parallel classloading,
+// lock a placeholder entry for this class/class_loader pair
+// to allow parallel defines of different classes for this class loader
 // With AllowParallelDefine flag==true, in case they do not synchronize around
 // FindLoadedClass/DefineClass, calls, we check for parallel
 // loading for them, wait if a defineClass is in progress
 // and return the initial requestor's results
+// This flag does not apply to the bootstrap classloader.
 // With AllowParallelDefine flag==false, call through to define_instance_class
 // which will throw LinkageError: duplicate class definition.
+// False is the requested default.
 // For better performance, the class loaders should synchronize
 // findClass(), i.e. FindLoadedClass/DefineClassIfAbsent or they
 // potentially waste time reading and parsing the bytestream.
@@ -1511,9 +1523,11 @@ instanceKlassHandle SystemDictionary::find_or_define_instance_class(symbolHandle
   {
     MutexLocker mu(SystemDictionary_lock, THREAD);
     // First check if class already defined
-    klassOop check = find_class(d_index, d_hash, name_h, class_loader);
+    if (UnsyncloadClass || (is_parallelDefine(class_loader))) {
-    if (check != NULL) {
+      klassOop check = find_class(d_index, d_hash, name_h, class_loader);
-      return(instanceKlassHandle(THREAD, check));
+      if (check != NULL) {
+        return(instanceKlassHandle(THREAD, check));
+      }
     }
 
     // Acquire define token for this class/classloader
@@ -1529,7 +1543,7 @@ instanceKlassHandle SystemDictionary::find_or_define_instance_class(symbolHandle
     // Only special cases allow parallel defines and can use other thread's results
     // Other cases fall through, and may run into duplicate defines
     // caught by finding an entry in the SystemDictionary
-    if ((UnsyncloadClass || AllowParallelDefineClass) && (probe->instanceKlass() != NULL)) {
+    if ((UnsyncloadClass || is_parallelDefine(class_loader)) && (probe->instanceKlass() != NULL)) {
         probe->remove_seen_thread(THREAD, PlaceholderTable::DEFINE_CLASS);
         placeholders()->find_and_remove(p_index, p_hash, name_h, class_loader, THREAD);
         SystemDictionary_lock->notify_all();
@@ -1973,7 +1987,7 @@ void SystemDictionary::initialize_preloaded_classes(TRAPS) {
   WKID indy_group_end   = WK_KLASS_ENUM_NAME(Dynamic_klass);
   initialize_wk_klasses_until(indy_group_start, scan, CHECK);
   if (EnableInvokeDynamic) {
-    initialize_wk_klasses_through(indy_group_start, scan, CHECK);
+    initialize_wk_klasses_through(indy_group_end, scan, CHECK);
   }
   if (_well_known_klasses[indy_group_start] == NULL) {
     // Skip the rest of the dynamic typing classes, if Linkage is not loaded.
@@ -2404,7 +2418,7 @@ Handle SystemDictionary::make_dynamic_call_site(KlassHandle caller,
                                                 methodHandle mh_invdyn,
                                                 TRAPS) {
   Handle empty;
-  // call sun.dyn.CallSiteImpl::makeSite(caller, name, mtype, cmid, cbci)
+  // call java.dyn.CallSite::makeSite(caller, name, mtype, cmid, cbci)
   oop name_str_oop = StringTable::intern(name(), CHECK_(empty)); // not a handle!
   JavaCallArguments args(Handle(THREAD, caller->java_mirror()));
   args.push_oop(name_str_oop);
@@ -2413,17 +2427,19 @@ Handle SystemDictionary::make_dynamic_call_site(KlassHandle caller,
   args.push_int(caller_bci);
   JavaValue result(T_OBJECT);
   JavaCalls::call_static(&result,
-                         SystemDictionary::CallSiteImpl_klass(),
+                         SystemDictionary::CallSite_klass(),
                          vmSymbols::makeSite_name(), vmSymbols::makeSite_signature(),
                          &args, CHECK_(empty));
   oop call_site_oop = (oop) result.get_jobject();
   assert(call_site_oop->is_oop()
-         /*&& sun_dyn_CallSiteImpl::is_instance(call_site_oop)*/, "must be sane");
+         /*&& java_dyn_CallSite::is_instance(call_site_oop)*/, "must be sane");
-  sun_dyn_CallSiteImpl::set_vmmethod(call_site_oop, mh_invdyn());
+  java_dyn_CallSite::set_vmmethod(call_site_oop, mh_invdyn());
   if (TraceMethodHandles) {
+#ifndef PRODUCT
     tty->print_cr("Linked invokedynamic bci=%d site="INTPTR_FORMAT":", caller_bci, call_site_oop);
     call_site_oop->print();
     tty->cr();
+#endif //PRODUCT
   }
   return call_site_oop;
 }
@@ -2436,9 +2452,17 @@ Handle SystemDictionary::find_bootstrap_method(KlassHandle caller,
 
   instanceKlassHandle ik(THREAD, caller());
 
-  if (ik->bootstrap_method() != NULL) {
+  oop boot_method_oop = ik->bootstrap_method();
-    return Handle(THREAD, ik->bootstrap_method());
+  if (boot_method_oop != NULL) {
+    if (TraceMethodHandles) {
+      tty->print_cr("bootstrap method for "PTR_FORMAT" cached as "PTR_FORMAT":", ik(), boot_method_oop);
+    }
+    NOT_PRODUCT(if (!boot_method_oop->is_oop()) { tty->print_cr("*** boot MH of "PTR_FORMAT" = "PTR_FORMAT, ik(), boot_method_oop); ik()->print(); });
+    assert(boot_method_oop->is_oop()
+           && java_dyn_MethodHandle::is_instance(boot_method_oop), "must be sane");
+    return Handle(THREAD, boot_method_oop);
   }
+  boot_method_oop = NULL;  // GC safety
 
   // call java.dyn.Linkage::findBootstrapMethod(caller, sbk)
   JavaCallArguments args(Handle(THREAD, ik->java_mirror()));
@@ -2452,9 +2476,18 @@ Handle SystemDictionary::find_bootstrap_method(KlassHandle caller,
                          vmSymbols::findBootstrapMethod_name(),
                          vmSymbols::findBootstrapMethod_signature(),
                          &args, CHECK_(empty));
-  oop boot_method_oop = (oop) result.get_jobject();
+  boot_method_oop = (oop) result.get_jobject();
 
   if (boot_method_oop != NULL) {
+    if (TraceMethodHandles) {
+#ifndef PRODUCT
+      tty->print_cr("--------");
+      tty->print_cr("bootstrap method for "PTR_FORMAT" computed as "PTR_FORMAT":", ik(), boot_method_oop);
+      ik()->print();
+      boot_method_oop->print();
+      tty->print_cr("========");
+#endif //PRODUCT
+    }
     assert(boot_method_oop->is_oop()
            && java_dyn_MethodHandle::is_instance(boot_method_oop), "must be sane");
     // probably no race conditions, but let's be careful:
@@ -2463,6 +2496,14 @@ Handle SystemDictionary::find_bootstrap_method(KlassHandle caller,
     else
       boot_method_oop = ik->bootstrap_method();
   } else {
+    if (TraceMethodHandles) {
+#ifndef PRODUCT
+      tty->print_cr("--------");
+      tty->print_cr("bootstrap method for "PTR_FORMAT" computed as NULL:", ik());
+      ik()->print();
+      tty->print_cr("========");
+#endif //PRODUCT
+    }
     boot_method_oop = ik->bootstrap_method();
   }
 

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

@@ -144,13 +144,13 @@ class SymbolPropertyTable;
   template(WrongMethodTypeException_klass, java_dyn_WrongMethodTypeException, Opt) \
   template(Linkage_klass,                java_dyn_Linkage,               Opt) \
   template(CallSite_klass,               java_dyn_CallSite,              Opt) \
-  template(CallSiteImpl_klass,           sun_dyn_CallSiteImpl,     Opt) \
   template(Dynamic_klass,                java_dyn_Dynamic,               Opt) \
   /* Note: MethodHandle must be first, and Dynamic last in group */           \
                                                                               \
   template(vector_klass,                 java_util_Vector,               Pre) \
   template(hashtable_klass,              java_util_Hashtable,            Pre) \
   template(stringBuffer_klass,           java_lang_StringBuffer,         Pre) \
+  template(StringBuilder_klass,          java_lang_StringBuilder,        Pre) \
                                                                               \
   /* It's NULL in non-1.4 JDKs. */                                            \
   template(stackTraceElement_klass,      java_lang_StackTraceElement,    Opt) \
@@ -578,6 +578,7 @@ private:
   static Handle compute_loader_lock_object(Handle class_loader, TRAPS);
   static void check_loader_lock_contention(Handle loader_lock, TRAPS);
   static bool is_parallelCapable(Handle class_loader);
+  static bool is_parallelDefine(Handle class_loader);
 
   static klassOop find_shared_class(symbolHandle class_name);
 

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

@@ -1903,17 +1903,8 @@ void ClassVerifier::verify_invoke_instructions(
   verify_cp_type(index, cp, types, CHECK_VERIFY(this));
 
   // Get method name and signature
-  symbolHandle method_name;
+  symbolHandle method_name(THREAD, cp->name_ref_at(index));
-  symbolHandle method_sig;
+  symbolHandle method_sig(THREAD, cp->signature_ref_at(index));
-  if (opcode == Bytecodes::_invokedynamic) {
-    int name_index = cp->name_ref_index_at(index);
-    int sig_index  = cp->signature_ref_index_at(index);
-    method_name = symbolHandle(THREAD, cp->symbol_at(name_index));
-    method_sig  = symbolHandle(THREAD, cp->symbol_at(sig_index));
-  } else {
-    method_name = symbolHandle(THREAD, cp->name_ref_at(index));
-    method_sig  = symbolHandle(THREAD, cp->signature_ref_at(index));
-  }
 
   if (!SignatureVerifier::is_valid_method_signature(method_sig)) {
     class_format_error(

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

@@ -303,6 +303,11 @@ inline bool match_F_R(jshort flags) {
   const int neg = JVM_ACC_STATIC | JVM_ACC_SYNCHRONIZED;
   return (flags & (req | neg)) == req;
 }
+inline bool match_F_Y(jshort flags) {
+  const int req = JVM_ACC_SYNCHRONIZED;
+  const int neg = JVM_ACC_STATIC;
+  return (flags & (req | neg)) == req;
+}
 inline bool match_F_RN(jshort flags) {
   const int req = JVM_ACC_NATIVE;
   const int neg = JVM_ACC_STATIC | JVM_ACC_SYNCHRONIZED;
@@ -361,6 +366,7 @@ const char* vmIntrinsics::short_name_as_C_string(vmIntrinsics::ID id, char* buf,
   const char* sname = vmSymbols::name_for(signature_for(id));
   const char* fname = "";
   switch (flags_for(id)) {
+  case F_Y:  fname = "synchronized ";  break;
   case F_RN: fname = "native ";        break;
   case F_SN: fname = "native static "; break;
   case F_S:  fname = "static ";        break;

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

@@ -84,6 +84,7 @@
   template(java_lang_reflect_Field,                   "java/lang/reflect/Field")                  \
   template(java_lang_reflect_Array,                   "java/lang/reflect/Array")                  \
   template(java_lang_StringBuffer,                    "java/lang/StringBuffer")                   \
+  template(java_lang_StringBuilder,                   "java/lang/StringBuilder")                  \
   template(java_lang_CharSequence,                    "java/lang/CharSequence")                   \
   template(java_security_AccessControlContext,        "java/security/AccessControlContext")       \
   template(java_security_ProtectionDomain,            "java/security/ProtectionDomain")           \
@@ -104,6 +105,7 @@
   template(java_lang_AssertionStatusDirectives,       "java/lang/AssertionStatusDirectives")      \
   template(sun_jkernel_DownloadManager,               "sun/jkernel/DownloadManager")              \
   template(getBootClassPathEntryForClass_name,        "getBootClassPathEntryForClass")            \
+  template(setBootClassLoaderHook_name,               "setBootClassLoaderHook")                   \
                                                                                                   \
   /* class file format tags */                                                                    \
   template(tag_source_file,                           "SourceFile")                               \
@@ -233,10 +235,9 @@
   template(sun_dyn_AdapterMethodHandle,               "sun/dyn/AdapterMethodHandle")              \
   template(sun_dyn_BoundMethodHandle,                 "sun/dyn/BoundMethodHandle")                \
   template(sun_dyn_DirectMethodHandle,                "sun/dyn/DirectMethodHandle")               \
-  template(sun_dyn_CallSiteImpl,                      "sun/dyn/CallSiteImpl")                     \
   template(makeImpl_name,                             "makeImpl") /*MethodType::makeImpl*/        \
   template(makeImpl_signature,    "(Ljava/lang/Class;[Ljava/lang/Class;ZZ)Ljava/dyn/MethodType;") \
-  template(makeSite_name,                             "makeSite") /*CallSiteImpl::makeImpl*/       \
+  template(makeSite_name,                             "makeSite") /*CallSite::makeSite*/          \
   template(makeSite_signature,    "(Ljava/lang/Class;Ljava/lang/String;Ljava/dyn/MethodType;II)Ljava/dyn/CallSite;") \
   template(findBootstrapMethod_name,                  "findBootstrapMethod")                      \
   template(findBootstrapMethod_signature, "(Ljava/lang/Class;Ljava/lang/Class;)Ljava/dyn/MethodHandle;") \
@@ -335,6 +336,7 @@
   template(ptypes_name,                               "ptypes")                                   \
   template(form_name,                                 "form")                                     \
   template(erasedType_name,                           "erasedType")                               \
+  template(append_name,                               "append")                                   \
                                                                                                   \
   /* non-intrinsic name/signature pairs: */                                                       \
   template(register_method_name,                      "register")                                 \
@@ -416,6 +418,13 @@
   template(string_signature,                          "Ljava/lang/String;")                                       \
   template(reference_signature,                       "Ljava/lang/ref/Reference;")                                \
   template(concurrenthashmap_signature,               "Ljava/util/concurrent/ConcurrentHashMap;")                 \
+  template(String_StringBuilder_signature,            "(Ljava/lang/String;)Ljava/lang/StringBuilder;")            \
+  template(int_StringBuilder_signature,               "(I)Ljava/lang/StringBuilder;")                             \
+  template(char_StringBuilder_signature,              "(C)Ljava/lang/StringBuilder;")                             \
+  template(String_StringBuffer_signature,             "(Ljava/lang/String;)Ljava/lang/StringBuffer;")             \
+  template(int_StringBuffer_signature,                "(I)Ljava/lang/StringBuffer;")                              \
+  template(char_StringBuffer_signature,               "(C)Ljava/lang/StringBuffer;")                              \
+  template(int_String_signature,                      "(I)Ljava/lang/String;")                                    \
   /* signature symbols needed by intrinsics */                                                                    \
   VM_INTRINSICS_DO(VM_INTRINSIC_IGNORE, VM_SYMBOL_IGNORE, VM_SYMBOL_IGNORE, template, VM_ALIAS_IGNORE)            \
                                                                                                                   \
@@ -815,10 +824,34 @@
     /*the compiler does have special inlining code for these; bytecode inline is just fine */                           \
                                                                                                                         \
   do_intrinsic(_fillInStackTrace,         java_lang_Throwable, fillInStackTrace_name, void_throwable_signature,  F_RNY) \
-                                                                                                                        \
+                                                                                                                          \
-  do_intrinsic(_Object_init,              java_lang_Object, object_initializer_name, void_method_signature,      F_R)   \
+  do_intrinsic(_StringBuilder_void,   java_lang_StringBuilder, object_initializer_name, void_method_signature,     F_R)   \
-  /*    (symbol object_initializer_name defined above) */                                                               \
+  do_intrinsic(_StringBuilder_int,    java_lang_StringBuilder, object_initializer_name, int_void_signature,        F_R)   \
-                                                                                                                        \
+  do_intrinsic(_StringBuilder_String, java_lang_StringBuilder, object_initializer_name, string_void_signature,     F_R)   \
+                                                                                                                          \
+  do_intrinsic(_StringBuilder_append_char,   java_lang_StringBuilder, append_name, char_StringBuilder_signature,   F_R)   \
+  do_intrinsic(_StringBuilder_append_int,    java_lang_StringBuilder, append_name, int_StringBuilder_signature,    F_R)   \
+  do_intrinsic(_StringBuilder_append_String, java_lang_StringBuilder, append_name, String_StringBuilder_signature, F_R)   \
+                                                                                                                          \
+  do_intrinsic(_StringBuilder_toString, java_lang_StringBuilder, toString_name, void_string_signature,             F_R)   \
+                                                                                                                          \
+  do_intrinsic(_StringBuffer_void,   java_lang_StringBuffer, object_initializer_name, void_method_signature,       F_R)   \
+  do_intrinsic(_StringBuffer_int,    java_lang_StringBuffer, object_initializer_name, int_void_signature,          F_R)   \
+  do_intrinsic(_StringBuffer_String, java_lang_StringBuffer, object_initializer_name, string_void_signature,       F_R)   \
+                                                                                                                          \
+  do_intrinsic(_StringBuffer_append_char,   java_lang_StringBuffer, append_name, char_StringBuffer_signature,      F_Y)   \
+  do_intrinsic(_StringBuffer_append_int,    java_lang_StringBuffer, append_name, int_StringBuffer_signature,       F_Y)   \
+  do_intrinsic(_StringBuffer_append_String, java_lang_StringBuffer, append_name, String_StringBuffer_signature,    F_Y)   \
+                                                                                                                          \
+  do_intrinsic(_StringBuffer_toString,  java_lang_StringBuffer, toString_name, void_string_signature,              F_Y)   \
+                                                                                                                          \
+  do_intrinsic(_Integer_toString,      java_lang_Integer, toString_name, int_String_signature,                     F_S)   \
+                                                                                                                          \
+  do_intrinsic(_String_String, java_lang_String, object_initializer_name, string_void_signature,                   F_R)   \
+                                                                                                                          \
+  do_intrinsic(_Object_init,              java_lang_Object, object_initializer_name, void_method_signature,        F_R)   \
+  /*    (symbol object_initializer_name defined above) */                                                                 \
+                                                                                                                          \
   do_intrinsic(_invoke,                   java_lang_reflect_Method, invoke_name, object_array_object_object_signature, F_R) \
   /*   (symbols invoke_name and invoke_signature defined above) */                                                      \
                                                                                                                         \
@@ -946,11 +979,12 @@ class vmIntrinsics: AllStatic {
   enum Flags {
     // AccessFlags syndromes relevant to intrinsics.
     F_none = 0,
-    F_R,                        // !static        !synchronized (R="regular")
+    F_R,                        // !static ?native !synchronized (R="regular")
-    F_S,                        //  static        !synchronized
+    F_S,                        //  static ?native !synchronized
-    F_RN,                       // !static native !synchronized
+    F_Y,                        // !static ?native  synchronized
-    F_SN,                       //  static native !synchronized
+    F_RN,                       // !static  native !synchronized
-    F_RNY                       // !static native  synchronized
+    F_SN,                       //  static  native !synchronized
+    F_RNY                       // !static  native  synchronized
   };
 
 public:

hotspot/src/share/vm/code/dependencies.cpp

@@ -1528,19 +1528,23 @@ void DepChange::print() {
   int nsup = 0, nint = 0;
   for (ContextStream str(*this); str.next(); ) {
     klassOop k = str.klass();
-    switch (str._change_type) {
+    switch (str.change_type()) {
     case Change_new_type:
       tty->print_cr("  dependee = %s", instanceKlass::cast(k)->external_name());
       break;
     case Change_new_sub:
-      if (!WizardMode)
+      if (!WizardMode) {
-           ++nsup;
+        ++nsup;
-      else tty->print_cr("  context super = %s", instanceKlass::cast(k)->external_name());
+      } else {
+        tty->print_cr("  context super = %s", instanceKlass::cast(k)->external_name());
+      }
       break;
     case Change_new_impl:
-      if (!WizardMode)
+      if (!WizardMode) {
-           ++nint;
+        ++nint;
-      else tty->print_cr("  context interface = %s", instanceKlass::cast(k)->external_name());
+      } else {
+        tty->print_cr("  context interface = %s", instanceKlass::cast(k)->external_name());
+      }
       break;
     }
   }

hotspot/src/share/vm/code/dependencies.hpp

@@ -470,7 +470,7 @@ class Dependencies: public ResourceObj {
 // super types can be context types for a relevant dependency, which the
 // new type could invalidate.
 class DepChange : public StackObj {
- private:
+ public:
   enum ChangeType {
     NO_CHANGE = 0,              // an uninvolved klass
     Change_new_type,            // a newly loaded type
@@ -480,6 +480,7 @@ class DepChange : public StackObj {
     Start_Klass = CHANGE_LIMIT  // internal indicator for ContextStream
   };
 
+ private:
   // each change set is rooted in exactly one new type (at present):
   KlassHandle _new_type;
 
@@ -510,15 +511,15 @@ class DepChange : public StackObj {
   // }
   class ContextStream : public StackObj {
    private:
-    DepChange&       _changes;
+    DepChange&  _changes;
     friend class DepChange;
 
     // iteration variables:
-    ChangeType            _change_type;
+    ChangeType  _change_type;
-    klassOop              _klass;
+    klassOop    _klass;
-    objArrayOop           _ti_base;    // i.e., transitive_interfaces
+    objArrayOop _ti_base;    // i.e., transitive_interfaces
-    int                   _ti_index;
+    int         _ti_index;
-    int                   _ti_limit;
+    int         _ti_limit;
 
     // start at the beginning:
     void start() {
@@ -530,11 +531,11 @@ class DepChange : public StackObj {
       _ti_limit = 0;
     }
 
+   public:
     ContextStream(DepChange& changes)
       : _changes(changes)
     { start(); }
 
-   public:
     ContextStream(DepChange& changes, No_Safepoint_Verifier& nsv)
       : _changes(changes)
       // the nsv argument makes it safe to hold oops like _klass
@@ -542,6 +543,7 @@ class DepChange : public StackObj {
 
     bool next();
 
+    ChangeType change_type()     { return _change_type; }
     klassOop   klass()           { return _klass; }
   };
   friend class DepChange::ContextStream;

hotspot/src/share/vm/code/nmethod.cpp

@@ -414,9 +414,8 @@ int nmethod::total_size() const {
 }
 
 const char* nmethod::compile_kind() const {
-  if (method() == NULL)    return "unloaded";
-  if (is_native_method())  return "c2n";
   if (is_osr_method())     return "osr";
+  if (method() != NULL && is_native_method())  return "c2n";
   return NULL;
 }
 
@@ -1127,6 +1126,9 @@ void nmethod::make_unloaded(BoolObjectClosure* is_alive, oop cause) {
   }
   flags.state = unloaded;
 
+  // Log the unloading.
+  log_state_change();
+
   // The methodOop is gone at this point
   assert(_method == NULL, "Tautology");
 
@@ -1137,8 +1139,6 @@ void nmethod::make_unloaded(BoolObjectClosure* is_alive, oop cause) {
 
 void nmethod::invalidate_osr_method() {
   assert(_entry_bci != InvocationEntryBci, "wrong kind of nmethod");
-  if (_entry_bci != InvalidOSREntryBci)
-    inc_decompile_count();
   // Remove from list of active nmethods
   if (method() != NULL)
     instanceKlass::cast(method()->method_holder())->remove_osr_nmethod(this);
@@ -1146,59 +1146,63 @@ void nmethod::invalidate_osr_method() {
   _entry_bci = InvalidOSREntryBci;
 }
 
-void nmethod::log_state_change(int state) const {
+void nmethod::log_state_change() const {
   if (LogCompilation) {
     if (xtty != NULL) {
       ttyLocker ttyl;  // keep the following output all in one block
-      xtty->begin_elem("make_not_entrant %sthread='" UINTX_FORMAT "'",
+      if (flags.state == unloaded) {
-                       (state == zombie ? "zombie='1' " : ""),
+        xtty->begin_elem("make_unloaded thread='" UINTX_FORMAT "'",
-                       os::current_thread_id());
+                         os::current_thread_id());
+      } else {
+        xtty->begin_elem("make_not_entrant thread='" UINTX_FORMAT "'%s",
+                         os::current_thread_id(),
+                         (flags.state == zombie ? " zombie='1'" : ""));
+      }
       log_identity(xtty);
       xtty->stamp();
       xtty->end_elem();
     }
   }
-  if (PrintCompilation) {
+  if (PrintCompilation && flags.state != unloaded) {
-    print_on(tty, state == zombie ? "made zombie " : "made not entrant ");
+    print_on(tty, flags.state == zombie ? "made zombie " : "made not entrant ");
     tty->cr();
   }
 }
 
 // Common functionality for both make_not_entrant and make_zombie
-void nmethod::make_not_entrant_or_zombie(int state) {
+bool nmethod::make_not_entrant_or_zombie(int state) {
   assert(state == zombie || state == not_entrant, "must be zombie or not_entrant");
 
-  // Code for an on-stack-replacement nmethod is removed when a class gets unloaded.
+  // If the method is already zombie there is nothing to do
-  // They never become zombie/non-entrant, so the nmethod sweeper will never remove
+  if (is_zombie()) {
-  // them. Instead the entry_bci is set to InvalidOSREntryBci, so the osr nmethod
+    return false;
-  // will never be used anymore. That the nmethods only gets removed when class unloading
-  // happens, make life much simpler, since the nmethods are not just going to disappear
-  // out of the blue.
-  if (is_osr_method()) {
-    if (osr_entry_bci() != InvalidOSREntryBci) {
-      // only log this once
-      log_state_change(state);
-    }
-    invalidate_osr_method();
-    return;
   }
 
-  // If the method is already zombie or set to the state we want, nothing to do
-  if (is_zombie() || (state == not_entrant && is_not_entrant())) {
-    return;
-  }
-
-  log_state_change(state);
-
   // Make sure the nmethod is not flushed in case of a safepoint in code below.
   nmethodLocker nml(this);
 
   {
+    // invalidate osr nmethod before acquiring the patching lock since
+    // they both acquire leaf locks and we don't want a deadlock.
+    // This logic is equivalent to the logic below for patching the
+    // verified entry point of regular methods.
+    if (is_osr_method()) {
+      // this effectively makes the osr nmethod not entrant
+      invalidate_osr_method();
+    }
+
     // Enter critical section.  Does not block for safepoint.
     MutexLockerEx pl(Patching_lock, Mutex::_no_safepoint_check_flag);
+
+    if (flags.state == state) {
+      // another thread already performed this transition so nothing
+      // to do, but return false to indicate this.
+      return false;
+    }
+
     // The caller can be calling the method statically or through an inline
     // cache call.
-    if (!is_not_entrant()) {
+    if (!is_osr_method() && !is_not_entrant()) {
       NativeJump::patch_verified_entry(entry_point(), verified_entry_point(),
                   SharedRuntime::get_handle_wrong_method_stub());
       assert (NativeJump::instruction_size == nmethod::_zombie_instruction_size, "");
@@ -1217,6 +1221,10 @@ void nmethod::make_not_entrant_or_zombie(int state) {
 
     // Change state
     flags.state = state;
+
+    // Log the transition once
+    log_state_change();
+
   } // leave critical region under Patching_lock
 
   if (state == not_entrant) {
@@ -1240,7 +1248,6 @@ void nmethod::make_not_entrant_or_zombie(int state) {
   // It's a true state change, so mark the method as decompiled.
   inc_decompile_count();
 
-
   // zombie only - if a JVMTI agent has enabled the CompiledMethodUnload event
   // and it hasn't already been reported for this nmethod then report it now.
   // (the event may have been reported earilier if the GC marked it for unloading).
@@ -1268,7 +1275,7 @@ void nmethod::make_not_entrant_or_zombie(int state) {
 
   // Check whether method got unloaded at a safepoint before this,
   // if so we can skip the flushing steps below
-  if (method() == NULL) return;
+  if (method() == NULL) return true;
 
   // Remove nmethod from method.
   // We need to check if both the _code and _from_compiled_code_entry_point
@@ -1282,6 +1289,8 @@ void nmethod::make_not_entrant_or_zombie(int state) {
     HandleMark hm;
     method()->clear_code();
   }
+
+  return true;
 }
 
 

hotspot/src/share/vm/code/nmethod.hpp

@@ -252,7 +252,9 @@ class nmethod : public CodeBlob {
   void* operator new(size_t size, int nmethod_size);
 
   const char* reloc_string_for(u_char* begin, u_char* end);
-  void make_not_entrant_or_zombie(int state);
+  // Returns true if this thread changed the state of the nmethod or
+  // false if another thread performed the transition.
+  bool make_not_entrant_or_zombie(int state);
   void inc_decompile_count();
 
   // used to check that writes to nmFlags are done consistently.
@@ -375,10 +377,12 @@ class nmethod : public CodeBlob {
   bool  is_zombie() const                         { return flags.state == zombie; }
   bool  is_unloaded() const                       { return flags.state == unloaded;   }
 
-  // Make the nmethod non entrant. The nmethod will continue to be alive.
+  // Make the nmethod non entrant. The nmethod will continue to be
-  // It is used when an uncommon trap happens.
+  // alive.  It is used when an uncommon trap happens.  Returns true
-  void  make_not_entrant()                        { make_not_entrant_or_zombie(not_entrant); }
+  // if this thread changed the state of the nmethod or false if
-  void  make_zombie()                             { make_not_entrant_or_zombie(zombie); }
+  // another thread performed the transition.
+  bool  make_not_entrant()                        { return make_not_entrant_or_zombie(not_entrant); }
+  bool  make_zombie()                             { return make_not_entrant_or_zombie(zombie); }
 
   // used by jvmti to track if the unload event has been reported
   bool  unload_reported()                         { return _unload_reported; }
@@ -563,7 +567,7 @@ class nmethod : public CodeBlob {
   // Logging
   void log_identity(xmlStream* log) const;
   void log_new_nmethod() const;
-  void log_state_change(int state) const;
+  void log_state_change() const;
 
   // Prints a comment for one native instruction (reloc info, pc desc)
   void print_code_comment_on(outputStream* st, int column, address begin, address end);

hotspot/src/share/vm/gc_implementation/concurrentMarkSweep/concurrentMarkSweepGeneration.cpp

@@ -709,7 +709,8 @@ CMSCollector::CMSCollector(ConcurrentMarkSweepGeneration* cmsGen,
 
   // Support for parallelizing survivor space rescan
   if (CMSParallelRemarkEnabled && CMSParallelSurvivorRemarkEnabled) {
-    size_t max_plab_samples = MaxNewSize/((SurvivorRatio+2)*MinTLABSize);
+    size_t max_plab_samples = cp->max_gen0_size()/
+                                ((SurvivorRatio+2)*MinTLABSize);
     _survivor_plab_array  = NEW_C_HEAP_ARRAY(ChunkArray, ParallelGCThreads);
     _survivor_chunk_array = NEW_C_HEAP_ARRAY(HeapWord*, 2*max_plab_samples);
     _cursor               = NEW_C_HEAP_ARRAY(size_t, ParallelGCThreads);

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

@@ -351,9 +351,16 @@ void
 CollectionSetChooser::printSortedHeapRegions() {
   gclog_or_tty->print_cr("Printing %d Heap Regions sorted by amount of known garbage",
                 _numMarkedRegions);
+
+  DEBUG_ONLY(int marked_count = 0;)
   for (int i = 0; i < _markedRegions.length(); i++) {
-    printHeapRegion(_markedRegions.at(i));
+    HeapRegion* r = _markedRegions.at(i);
+    if (r != NULL) {
+      printHeapRegion(r);
+      DEBUG_ONLY(marked_count++;)
+    }
   }
+  assert(marked_count == _numMarkedRegions, "must be");
   gclog_or_tty->print_cr("Done sorted heap region print");
 }
 

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

@@ -42,28 +42,49 @@ ConcurrentG1Refine::ConcurrentG1Refine() :
   _n_periods(0),
   _threads(NULL), _n_threads(0)
 {
-  if (G1ConcRefine) {
+
-    _n_threads = (int)thread_num();
+  // Ergomonically select initial concurrent refinement parameters
-    if (_n_threads > 0) {
+  if (FLAG_IS_DEFAULT(G1ConcRefineGreenZone)) {
-      _threads = NEW_C_HEAP_ARRAY(ConcurrentG1RefineThread*, _n_threads);
+    FLAG_SET_DEFAULT(G1ConcRefineGreenZone, MAX2<int>(ParallelGCThreads, 1));
-      int worker_id_offset = (int)DirtyCardQueueSet::num_par_ids();
+  }
-      ConcurrentG1RefineThread *next = NULL;
+  set_green_zone(G1ConcRefineGreenZone);
-      for (int i = _n_threads - 1; i >= 0; i--) {
+
-        ConcurrentG1RefineThread* t = new ConcurrentG1RefineThread(this, next, worker_id_offset, i);
+  if (FLAG_IS_DEFAULT(G1ConcRefineYellowZone)) {
-        assert(t != NULL, "Conc refine should have been created");
+    FLAG_SET_DEFAULT(G1ConcRefineYellowZone, green_zone() * 3);
-        assert(t->cg1r() == this, "Conc refine thread should refer to this");
+  }
-        _threads[i] = t;
+  set_yellow_zone(MAX2<int>(G1ConcRefineYellowZone, green_zone()));
-        next = t;
+
-      }
+  if (FLAG_IS_DEFAULT(G1ConcRefineRedZone)) {
-    }
+    FLAG_SET_DEFAULT(G1ConcRefineRedZone, yellow_zone() * 2);
+  }
+  set_red_zone(MAX2<int>(G1ConcRefineRedZone, yellow_zone()));
+  _n_worker_threads = thread_num();
+  // We need one extra thread to do the young gen rset size sampling.
+  _n_threads = _n_worker_threads + 1;
+  reset_threshold_step();
+
+  _threads = NEW_C_HEAP_ARRAY(ConcurrentG1RefineThread*, _n_threads);
+  int worker_id_offset = (int)DirtyCardQueueSet::num_par_ids();
+  ConcurrentG1RefineThread *next = NULL;
+  for (int i = _n_threads - 1; i >= 0; i--) {
+    ConcurrentG1RefineThread* t = new ConcurrentG1RefineThread(this, next, worker_id_offset, i);
+    assert(t != NULL, "Conc refine should have been created");
+    assert(t->cg1r() == this, "Conc refine thread should refer to this");
+    _threads[i] = t;
+    next = t;
   }
 }
 
-size_t ConcurrentG1Refine::thread_num() {
+void ConcurrentG1Refine::reset_threshold_step() {
-  if (G1ConcRefine) {
+  if (FLAG_IS_DEFAULT(G1ConcRefineThresholdStep)) {
-    return (G1ParallelRSetThreads > 0) ? G1ParallelRSetThreads : ParallelGCThreads;
+    _thread_threshold_step = (yellow_zone() - green_zone()) / (worker_thread_num() + 1);
+  } else {
+    _thread_threshold_step = G1ConcRefineThresholdStep;
   }
-  return 0;
+}
+
+int ConcurrentG1Refine::thread_num() {
+  return MAX2<int>((G1ParallelRSetThreads > 0) ? G1ParallelRSetThreads : ParallelGCThreads, 1);
 }
 
 void ConcurrentG1Refine::init() {
@@ -123,6 +144,15 @@ void ConcurrentG1Refine::stop() {
   }
 }
 
+void ConcurrentG1Refine::reinitialize_threads() {
+  reset_threshold_step();
+  if (_threads != NULL) {
+    for (int i = 0; i < _n_threads; i++) {
+      _threads[i]->initialize();
+    }
+  }
+}
+
 ConcurrentG1Refine::~ConcurrentG1Refine() {
   if (G1ConcRSLogCacheSize > 0) {
     assert(_card_counts != NULL, "Logic");
@@ -384,4 +414,3 @@ void ConcurrentG1Refine::print_worker_threads_on(outputStream* st) const {
     st->cr();
   }
 }
-

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

@@ -29,6 +29,31 @@ class G1RemSet;
 class ConcurrentG1Refine: public CHeapObj {
   ConcurrentG1RefineThread** _threads;
   int _n_threads;
+  int _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
+  * done, the buffers are left unprocessed to enable the caching effect of the
+  * dirtied cards. In the yellow zone [green, yellow) the concurrent refinement
+  * threads are gradually activated. In [yellow, red) all threads are
+  * running. If the length becomes red (max queue length) the mutators start
+  * processing the buffers.
+  *
+  * There are some interesting cases (with G1AdaptiveConcRefine turned off):
+  * 1) green = yellow = red = 0. In this case the mutator will process all
+  *    buffers. Except for those that are created by the deferred updates
+  *    machinery during a collection.
+  * 2) green = 0. Means no caching. Can be a good way to minimize the
+  *    amount of time spent updating rsets during a collection.
+  */
+  int _green_zone;
+  int _yellow_zone;
+  int _red_zone;
+
+  int _thread_threshold_step;
+
+  // Reset the threshold step value based of the current zone boundaries.
+  void reset_threshold_step();
 
   // The cache for card refinement.
   bool   _use_cache;
@@ -147,6 +172,8 @@ class ConcurrentG1Refine: public CHeapObj {
   void init(); // Accomplish some initialization that has to wait.
   void stop();
 
+  void reinitialize_threads();
+
   // Iterate over the conc refine threads
   void threads_do(ThreadClosure *tc);
 
@@ -178,7 +205,20 @@ class ConcurrentG1Refine: public CHeapObj {
 
   void clear_and_record_card_counts();
 
-  static size_t thread_num();
+  static int thread_num();
 
   void print_worker_threads_on(outputStream* st) const;
+
+  void set_green_zone(int x)  { _green_zone = x;  }
+  void set_yellow_zone(int x) { _yellow_zone = x; }
+  void set_red_zone(int x)    { _red_zone = x;    }
+
+  int green_zone() const      { return _green_zone;  }
+  int yellow_zone() const     { return _yellow_zone; }
+  int red_zone() const        { return _red_zone;    }
+
+  int total_thread_num() const  { return _n_threads;        }
+  int 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

@@ -25,10 +25,6 @@
 #include "incls/_precompiled.incl"
 #include "incls/_concurrentG1RefineThread.cpp.incl"
 
-// ======= Concurrent Mark Thread ========
-
-// The CM thread is created when the G1 garbage collector is used
-
 ConcurrentG1RefineThread::
 ConcurrentG1RefineThread(ConcurrentG1Refine* cg1r, ConcurrentG1RefineThread *next,
                          int worker_id_offset, int worker_id) :
@@ -37,19 +33,42 @@ ConcurrentG1RefineThread(ConcurrentG1Refine* cg1r, ConcurrentG1RefineThread *nex
   _worker_id(worker_id),
   _active(false),
   _next(next),
+  _monitor(NULL),
   _cg1r(cg1r),
-  _vtime_accum(0.0),
+  _vtime_accum(0.0)
-  _interval_ms(5.0)
 {
+
+  // Each thread has its own monitor. The i-th thread is responsible for signalling
+  // to thread i+1 if the number of buffers in the queue exceeds a threashold for this
+  // thread. Monitors are also used to wake up the threads during termination.
+  // The 0th worker in notified by mutator threads and has a special monitor.
+  // The last worker is used for young gen rset size sampling.
+  if (worker_id > 0) {
+    _monitor = new Monitor(Mutex::nonleaf, "Refinement monitor", true);
+  } else {
+    _monitor = DirtyCardQ_CBL_mon;
+  }
+  initialize();
   create_and_start();
 }
 
+void ConcurrentG1RefineThread::initialize() {
+  if (_worker_id < cg1r()->worker_thread_num()) {
+    // Current thread activation threshold
+    _threshold = MIN2<int>(cg1r()->thread_threshold_step() * (_worker_id + 1) + cg1r()->green_zone(),
+                           cg1r()->yellow_zone());
+    // A thread deactivates once the number of buffer reached a deactivation threshold
+    _deactivation_threshold = MAX2<int>(_threshold - cg1r()->thread_threshold_step(), cg1r()->green_zone());
+  } else {
+    set_active(true);
+  }
+}
+
 void ConcurrentG1RefineThread::sample_young_list_rs_lengths() {
   G1CollectedHeap* g1h = G1CollectedHeap::heap();
   G1CollectorPolicy* g1p = g1h->g1_policy();
   if (g1p->adaptive_young_list_length()) {
     int regions_visited = 0;
-
     g1h->young_list_rs_length_sampling_init();
     while (g1h->young_list_rs_length_sampling_more()) {
       g1h->young_list_rs_length_sampling_next();
@@ -70,99 +89,121 @@ void ConcurrentG1RefineThread::sample_young_list_rs_lengths() {
   }
 }
 
+void ConcurrentG1RefineThread::run_young_rs_sampling() {
+  DirtyCardQueueSet& dcqs = JavaThread::dirty_card_queue_set();
+  _vtime_start = os::elapsedVTime();
+  while(!_should_terminate) {
+    _sts.join();
+    sample_young_list_rs_lengths();
+    _sts.leave();
+
+    if (os::supports_vtime()) {
+      _vtime_accum = (os::elapsedVTime() - _vtime_start);
+    } else {
+      _vtime_accum = 0.0;
+    }
+
+    MutexLockerEx x(_monitor, Mutex::_no_safepoint_check_flag);
+    if (_should_terminate) {
+      break;
+    }
+    _monitor->wait(Mutex::_no_safepoint_check_flag, G1ConcRefineServiceInterval);
+  }
+}
+
+void ConcurrentG1RefineThread::wait_for_completed_buffers() {
+  DirtyCardQueueSet& dcqs = JavaThread::dirty_card_queue_set();
+  MutexLockerEx x(_monitor, Mutex::_no_safepoint_check_flag);
+  while (!_should_terminate && !is_active()) {
+    _monitor->wait(Mutex::_no_safepoint_check_flag);
+  }
+}
+
+bool ConcurrentG1RefineThread::is_active() {
+  DirtyCardQueueSet& dcqs = JavaThread::dirty_card_queue_set();
+  return _worker_id > 0 ? _active : dcqs.process_completed_buffers();
+}
+
+void ConcurrentG1RefineThread::activate() {
+  MutexLockerEx x(_monitor, Mutex::_no_safepoint_check_flag);
+  if (_worker_id > 0) {
+    if (G1TraceConcurrentRefinement) {
+      DirtyCardQueueSet& dcqs = JavaThread::dirty_card_queue_set();
+      gclog_or_tty->print_cr("G1-Refine-activated worker %d, on threshold %d, current %d",
+                             _worker_id, _threshold, (int)dcqs.completed_buffers_num());
+    }
+    set_active(true);
+  } else {
+    DirtyCardQueueSet& dcqs = JavaThread::dirty_card_queue_set();
+    dcqs.set_process_completed(true);
+  }
+  _monitor->notify();
+}
+
+void ConcurrentG1RefineThread::deactivate() {
+  MutexLockerEx x(_monitor, Mutex::_no_safepoint_check_flag);
+  if (_worker_id > 0) {
+    if (G1TraceConcurrentRefinement) {
+      DirtyCardQueueSet& dcqs = JavaThread::dirty_card_queue_set();
+      gclog_or_tty->print_cr("G1-Refine-deactivated worker %d, off threshold %d, current %d",
+                             _worker_id, _deactivation_threshold, (int)dcqs.completed_buffers_num());
+    }
+    set_active(false);
+  } else {
+    DirtyCardQueueSet& dcqs = JavaThread::dirty_card_queue_set();
+    dcqs.set_process_completed(false);
+  }
+}
+
 void ConcurrentG1RefineThread::run() {
   initialize_in_thread();
-  _vtime_start = os::elapsedVTime();
   wait_for_universe_init();
 
+  if (_worker_id >= cg1r()->worker_thread_num()) {
+    run_young_rs_sampling();
+    terminate();
+  }
+
+  _vtime_start = os::elapsedVTime();
   while (!_should_terminate) {
     DirtyCardQueueSet& dcqs = JavaThread::dirty_card_queue_set();
-    // Wait for completed log buffers to exist.
+
-    {
+    // Wait for work
-      MutexLockerEx x(DirtyCardQ_CBL_mon, Mutex::_no_safepoint_check_flag);
+    wait_for_completed_buffers();
-      while (((_worker_id == 0 && !dcqs.process_completed_buffers()) ||
-              (_worker_id > 0 && !is_active())) &&
-             !_should_terminate) {
-         DirtyCardQ_CBL_mon->wait(Mutex::_no_safepoint_check_flag);
-      }
-    }
 
     if (_should_terminate) {
-      return;
+      break;
     }
 
-    // Now we take them off (this doesn't hold locks while it applies
-    // closures.)  (If we did a full collection, then we'll do a full
-    // traversal.
     _sts.join();
-    int n_logs = 0;
-    int lower_limit = 0;
-    double start_vtime_sec; // only used when G1SmoothConcRefine is on
-    int prev_buffer_num; // only used when G1SmoothConcRefine is on
-    // This thread activation threshold
-    int threshold = G1UpdateBufferQueueProcessingThreshold * _worker_id;
-    // Next thread activation threshold
-    int next_threshold = threshold + G1UpdateBufferQueueProcessingThreshold;
-    int deactivation_threshold = MAX2<int>(threshold - G1UpdateBufferQueueProcessingThreshold / 2, 0);
 
-    if (G1SmoothConcRefine) {
+    do {
-      lower_limit = 0;
+      int curr_buffer_num = (int)dcqs.completed_buffers_num();
-      start_vtime_sec = os::elapsedVTime();
+      // If the number of the buffers falls down into the yellow zone,
-      prev_buffer_num = (int) dcqs.completed_buffers_num();
+      // that means that the transition period after the evacuation pause has ended.
-    } else {
+      if (dcqs.completed_queue_padding() > 0 && curr_buffer_num <= cg1r()->yellow_zone()) {
-      lower_limit = G1UpdateBufferQueueProcessingThreshold / 4; // For now.
+        dcqs.set_completed_queue_padding(0);
-    }
-    while (dcqs.apply_closure_to_completed_buffer(_worker_id + _worker_id_offset, lower_limit)) {
-      double end_vtime_sec;
-      double elapsed_vtime_sec;
-      int elapsed_vtime_ms;
-      int curr_buffer_num = (int) dcqs.completed_buffers_num();
-
-      if (G1SmoothConcRefine) {
-        end_vtime_sec = os::elapsedVTime();
-        elapsed_vtime_sec = end_vtime_sec - start_vtime_sec;
-        elapsed_vtime_ms = (int) (elapsed_vtime_sec * 1000.0);
-
-        if (curr_buffer_num > prev_buffer_num ||
-            curr_buffer_num > next_threshold) {
-          decreaseInterval(elapsed_vtime_ms);
-        } else if (curr_buffer_num < prev_buffer_num) {
-          increaseInterval(elapsed_vtime_ms);
-        }
       }
-      if (_worker_id == 0) {
+
-        sample_young_list_rs_lengths();
+      if (_worker_id > 0 && curr_buffer_num <= _deactivation_threshold) {
-      } else if (curr_buffer_num < deactivation_threshold) {
         // If the number of the buffer has fallen below our threshold
         // we should deactivate. The predecessor will reactivate this
         // thread should the number of the buffers cross the threshold again.
-        MutexLockerEx x(DirtyCardQ_CBL_mon, Mutex::_no_safepoint_check_flag);
         deactivate();
-        if (G1TraceConcurrentRefinement) {
-          gclog_or_tty->print_cr("G1-Refine-deactivated worker %d", _worker_id);
-        }
         break;
       }
 
       // Check if we need to activate the next thread.
-      if (curr_buffer_num > next_threshold && _next != NULL && !_next->is_active()) {
+      if (_next != NULL && !_next->is_active() && curr_buffer_num > _next->_threshold) {
-        MutexLockerEx x(DirtyCardQ_CBL_mon, Mutex::_no_safepoint_check_flag);
         _next->activate();
-        DirtyCardQ_CBL_mon->notify_all();
-        if (G1TraceConcurrentRefinement) {
-          gclog_or_tty->print_cr("G1-Refine-activated worker %d", _next->_worker_id);
-        }
       }
+    } while (dcqs.apply_closure_to_completed_buffer(_worker_id + _worker_id_offset, cg1r()->green_zone()));
 
-      if (G1SmoothConcRefine) {
+    // We can exit the loop above while being active if there was a yield request.
-        prev_buffer_num = curr_buffer_num;
+    if (is_active()) {
-        _sts.leave();
+      deactivate();
-        os::sleep(Thread::current(), (jlong) _interval_ms, false);
-        _sts.join();
-        start_vtime_sec = os::elapsedVTime();
-      }
-      n_logs++;
     }
+
     _sts.leave();
 
     if (os::supports_vtime()) {
@@ -172,7 +213,6 @@ void ConcurrentG1RefineThread::run() {
     }
   }
   assert(_should_terminate, "just checking");
-
   terminate();
 }
 
@@ -191,8 +231,8 @@ void ConcurrentG1RefineThread::stop() {
   }
 
   {
-    MutexLockerEx x(DirtyCardQ_CBL_mon, Mutex::_no_safepoint_check_flag);
+    MutexLockerEx x(_monitor, Mutex::_no_safepoint_check_flag);
-    DirtyCardQ_CBL_mon->notify_all();
+    _monitor->notify();
   }
 
   {

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

@@ -40,42 +40,36 @@ class ConcurrentG1RefineThread: public ConcurrentGCThread {
   // when the number of the rset update buffer crosses a certain threshold. A successor
   // would self-deactivate when the number of the buffers falls below the threshold.
   bool _active;
-  ConcurrentG1RefineThread *       _next;
+  ConcurrentG1RefineThread* _next;
- public:
+  Monitor* _monitor;
-  virtual void run();
+  ConcurrentG1Refine* _cg1r;
 
-  bool is_active()  { return _active;  }
+  int _thread_threshold_step;
-  void activate()   { _active = true;  }
+  // This thread activation threshold
-  void deactivate() { _active = false; }
+  int _threshold;
+  // This thread deactivation threshold
+  int _deactivation_threshold;
 
- private:
+  void sample_young_list_rs_lengths();
-  ConcurrentG1Refine*              _cg1r;
+  void run_young_rs_sampling();
+  void wait_for_completed_buffers();
 
-  double                           _interval_ms;
+  void set_active(bool x) { _active = x; }
-
+  bool is_active();
-  void decreaseInterval(int processing_time_ms) {
+  void activate();
-    double min_interval_ms = (double) processing_time_ms;
+  void deactivate();
-    _interval_ms = 0.8 * _interval_ms;
-    if (_interval_ms < min_interval_ms)
-      _interval_ms = min_interval_ms;
-  }
-  void increaseInterval(int processing_time_ms) {
-    double max_interval_ms = 9.0 * (double) processing_time_ms;
-    _interval_ms = 1.1 * _interval_ms;
-    if (max_interval_ms > 0 && _interval_ms > max_interval_ms)
-      _interval_ms = max_interval_ms;
-  }
-
-  void sleepBeforeNextCycle();
 
   // For use by G1CollectedHeap, which is a friend.
   static SuspendibleThreadSet* sts() { return &_sts; }
 
- public:
+public:
+  virtual void run();
   // Constructor
   ConcurrentG1RefineThread(ConcurrentG1Refine* cg1r, ConcurrentG1RefineThread* next,
                            int worker_id_offset, int worker_id);
 
+  void initialize();
+
   // Printing
   void print() const;
   void print_on(outputStream* st) const;
@@ -83,13 +77,10 @@ class ConcurrentG1RefineThread: public ConcurrentGCThread {
   // Total virtual time so far.
   double vtime_accum() { return _vtime_accum; }
 
-  ConcurrentG1Refine* cg1r()                     { return _cg1r;     }
+  ConcurrentG1Refine* cg1r() { return _cg1r;     }
-
-  void            sample_young_list_rs_lengths();
 
   // Yield for GC
-  void            yield();
+  void yield();
-
   // shutdown
   void stop();
 };

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

@@ -760,7 +760,6 @@ void ConcurrentMark::checkpointRootsInitialPost() {
   rp->setup_policy(false); // snapshot the soft ref policy to be used in this cycle
 
   SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
-  satb_mq_set.set_process_completed_threshold(G1SATBProcessCompletedThreshold);
   satb_mq_set.set_active_all_threads(true);
 
   // update_g1_committed() will be called at the end of an evac pause

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

@@ -61,8 +61,8 @@ bool DirtyCardQueue::apply_closure_to_buffer(CardTableEntryClosure* cl,
 #pragma warning( disable:4355 ) // 'this' : used in base member initializer list
 #endif // _MSC_VER
 
-DirtyCardQueueSet::DirtyCardQueueSet() :
+DirtyCardQueueSet::DirtyCardQueueSet(bool notify_when_complete) :
-  PtrQueueSet(true /*notify_when_complete*/),
+  PtrQueueSet(notify_when_complete),
   _closure(NULL),
   _shared_dirty_card_queue(this, true /*perm*/),
   _free_ids(NULL),
@@ -77,12 +77,12 @@ size_t DirtyCardQueueSet::num_par_ids() {
 }
 
 void DirtyCardQueueSet::initialize(Monitor* cbl_mon, Mutex* fl_lock,
+                                   int process_completed_threshold,
                                    int max_completed_queue,
                                    Mutex* lock, PtrQueueSet* fl_owner) {
-  PtrQueueSet::initialize(cbl_mon, fl_lock, max_completed_queue, fl_owner);
+  PtrQueueSet::initialize(cbl_mon, fl_lock, process_completed_threshold,
+                          max_completed_queue, fl_owner);
   set_buffer_size(G1UpdateBufferSize);
-  set_process_completed_threshold(G1UpdateBufferQueueProcessingThreshold);
-
   _shared_dirty_card_queue.set_lock(lock);
   _free_ids = new FreeIdSet((int) num_par_ids(), _cbl_mon);
 }
@@ -154,9 +154,10 @@ bool DirtyCardQueueSet::mut_process_buffer(void** buf) {
   return b;
 }
 
-DirtyCardQueueSet::CompletedBufferNode*
+
-DirtyCardQueueSet::get_completed_buffer_lock(int stop_at) {
+BufferNode*
-  CompletedBufferNode* nd = NULL;
+DirtyCardQueueSet::get_completed_buffer(int stop_at) {
+  BufferNode* nd = NULL;
   MutexLockerEx x(_cbl_mon, Mutex::_no_safepoint_check_flag);
 
   if ((int)_n_completed_buffers <= stop_at) {
@@ -166,53 +167,31 @@ DirtyCardQueueSet::get_completed_buffer_lock(int stop_at) {
 
   if (_completed_buffers_head != NULL) {
     nd = _completed_buffers_head;
-    _completed_buffers_head = nd->next;
+    _completed_buffers_head = nd->next();
     if (_completed_buffers_head == NULL)
       _completed_buffers_tail = NULL;
     _n_completed_buffers--;
+    assert(_n_completed_buffers >= 0, "Invariant");
   }
   debug_only(assert_completed_buffer_list_len_correct_locked());
   return nd;
 }
 
-// We only do this in contexts where there is no concurrent enqueueing.
-DirtyCardQueueSet::CompletedBufferNode*
-DirtyCardQueueSet::get_completed_buffer_CAS() {
-  CompletedBufferNode* nd = _completed_buffers_head;
-
-  while (nd != NULL) {
-    CompletedBufferNode* next = nd->next;
-    CompletedBufferNode* result =
-      (CompletedBufferNode*)Atomic::cmpxchg_ptr(next,
-                                                &_completed_buffers_head,
-                                                nd);
-    if (result == nd) {
-      return result;
-    } else {
-      nd = _completed_buffers_head;
-    }
-  }
-  assert(_completed_buffers_head == NULL, "Loop post");
-  _completed_buffers_tail = NULL;
-  return NULL;
-}
-
 bool DirtyCardQueueSet::
 apply_closure_to_completed_buffer_helper(int worker_i,
-                                         CompletedBufferNode* nd) {
+                                         BufferNode* nd) {
   if (nd != NULL) {
+    void **buf = BufferNode::make_buffer_from_node(nd);
+    size_t index = nd->index();
     bool b =
-      DirtyCardQueue::apply_closure_to_buffer(_closure, nd->buf,
+      DirtyCardQueue::apply_closure_to_buffer(_closure, buf,
-                                              nd->index, _sz,
+                                              index, _sz,
                                               true, worker_i);
-    void** buf = nd->buf;
-    size_t index = nd->index;
-    delete nd;
     if (b) {
       deallocate_buffer(buf);
       return true;  // In normal case, go on to next buffer.
     } else {
-      enqueue_complete_buffer(buf, index, true);
+      enqueue_complete_buffer(buf, index);
       return false;
     }
   } else {
@@ -222,40 +201,36 @@ apply_closure_to_completed_buffer_helper(int worker_i,
 
 bool DirtyCardQueueSet::apply_closure_to_completed_buffer(int worker_i,
                                                           int stop_at,
-                                                          bool with_CAS)
+                                                          bool during_pause)
 {
-  CompletedBufferNode* nd = NULL;
+  assert(!during_pause || stop_at == 0, "Should not leave any completed buffers during a pause");
-  if (with_CAS) {
+  BufferNode* nd = get_completed_buffer(stop_at);
-    guarantee(stop_at == 0, "Precondition");
-    nd = get_completed_buffer_CAS();
-  } else {
-    nd = get_completed_buffer_lock(stop_at);
-  }
   bool res = apply_closure_to_completed_buffer_helper(worker_i, nd);
   if (res) Atomic::inc(&_processed_buffers_rs_thread);
   return res;
 }
 
 void DirtyCardQueueSet::apply_closure_to_all_completed_buffers() {
-  CompletedBufferNode* nd = _completed_buffers_head;
+  BufferNode* nd = _completed_buffers_head;
   while (nd != NULL) {
     bool b =
-      DirtyCardQueue::apply_closure_to_buffer(_closure, nd->buf, 0, _sz,
+      DirtyCardQueue::apply_closure_to_buffer(_closure,
-                                              false);
+                                              BufferNode::make_buffer_from_node(nd),
+                                              0, _sz, false);
     guarantee(b, "Should not stop early.");
-    nd = nd->next;
+    nd = nd->next();
   }
 }
 
 void DirtyCardQueueSet::abandon_logs() {
   assert(SafepointSynchronize::is_at_safepoint(), "Must be at safepoint.");
-  CompletedBufferNode* buffers_to_delete = NULL;
+  BufferNode* buffers_to_delete = NULL;
   {
     MutexLockerEx x(_cbl_mon, Mutex::_no_safepoint_check_flag);
     while (_completed_buffers_head != NULL) {
-      CompletedBufferNode* nd = _completed_buffers_head;
+      BufferNode* nd = _completed_buffers_head;
-      _completed_buffers_head = nd->next;
+      _completed_buffers_head = nd->next();
-      nd->next = buffers_to_delete;
+      nd->set_next(buffers_to_delete);
       buffers_to_delete = nd;
     }
     _n_completed_buffers = 0;
@@ -263,10 +238,9 @@ void DirtyCardQueueSet::abandon_logs() {
     debug_only(assert_completed_buffer_list_len_correct_locked());
   }
   while (buffers_to_delete != NULL) {
-    CompletedBufferNode* nd = buffers_to_delete;
+    BufferNode* nd = buffers_to_delete;
-    buffers_to_delete = nd->next;
+    buffers_to_delete = nd->next();
-    deallocate_buffer(nd->buf);
+    deallocate_buffer(BufferNode::make_buffer_from_node(nd));
-    delete nd;
   }
   // Since abandon is done only at safepoints, we can safely manipulate
   // these queues.

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

@@ -1,5 +1,5 @@
 /*
- * Copyright 2001-2007 Sun Microsystems, Inc.  All Rights Reserved.
+ * Copyright 2001-2009 Sun Microsystems, Inc.  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
@@ -84,11 +84,12 @@ class DirtyCardQueueSet: public PtrQueueSet {
   jint _processed_buffers_rs_thread;
 
 public:
-  DirtyCardQueueSet();
+  DirtyCardQueueSet(bool notify_when_complete = true);
 
   void initialize(Monitor* cbl_mon, Mutex* fl_lock,
-                  int max_completed_queue = 0,
+                  int process_completed_threshold,
-                  Mutex* lock = NULL, PtrQueueSet* fl_owner = NULL);
+                  int max_completed_queue,
+                  Mutex* lock, PtrQueueSet* fl_owner = NULL);
 
   // The number of parallel ids that can be claimed to allow collector or
   // mutator threads to do card-processing work.
@@ -120,12 +121,13 @@ public:
   // is returned to the completed buffer set, and this call returns false.
   bool apply_closure_to_completed_buffer(int worker_i = 0,
                                          int stop_at = 0,
-                                         bool with_CAS = false);
+                                         bool during_pause = false);
-  bool apply_closure_to_completed_buffer_helper(int worker_i,
+
-                                                CompletedBufferNode* nd);
+  bool apply_closure_to_completed_buffer_helper(int worker_i,
+                                                BufferNode* nd);
+
+  BufferNode* get_completed_buffer(int stop_at);
 
-  CompletedBufferNode* get_completed_buffer_CAS();
-  CompletedBufferNode* get_completed_buffer_lock(int stop_at);
   // Applies the current closure to all completed buffers,
   // non-consumptively.
   void apply_closure_to_all_completed_buffers();

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

@@ -928,6 +928,8 @@ void G1CollectedHeap::do_collection(bool full, bool clear_all_soft_refs,
     TraceCPUTime tcpu(PrintGCDetails, true, gclog_or_tty);
     TraceTime t(full ? "Full GC (System.gc())" : "Full GC", PrintGC, true, gclog_or_tty);
 
+    TraceMemoryManagerStats tms(true /* fullGC */);
+
     double start = os::elapsedTime();
     g1_policy()->record_full_collection_start();
 
@@ -1001,6 +1003,8 @@ void G1CollectedHeap::do_collection(bool full, bool clear_all_soft_refs,
 
     COMPILER2_PRESENT(DerivedPointerTable::update_pointers());
 
+    MemoryService::track_memory_usage();
+
     if (VerifyAfterGC && total_collections() >= VerifyGCStartAt) {
       HandleMark hm;  // Discard invalid handles created during verification
       gclog_or_tty->print(" VerifyAfterGC:");
@@ -1371,6 +1375,7 @@ void G1CollectedHeap::shrink(size_t shrink_bytes) {
 G1CollectedHeap::G1CollectedHeap(G1CollectorPolicy* policy_) :
   SharedHeap(policy_),
   _g1_policy(policy_),
+  _dirty_card_queue_set(false),
   _ref_processor(NULL),
   _process_strong_tasks(new SubTasksDone(G1H_PS_NumElements)),
   _bot_shared(NULL),
@@ -1456,8 +1461,6 @@ jint G1CollectedHeap::initialize() {
   Universe::check_alignment(init_byte_size, HeapRegion::GrainBytes, "g1 heap");
   Universe::check_alignment(max_byte_size, HeapRegion::GrainBytes, "g1 heap");
 
-  // We allocate this in any case, but only do no work if the command line
-  // param is off.
   _cg1r = new ConcurrentG1Refine();
 
   // Reserve the maximum.
@@ -1590,18 +1593,20 @@ jint G1CollectedHeap::initialize() {
 
   JavaThread::satb_mark_queue_set().initialize(SATB_Q_CBL_mon,
                                                SATB_Q_FL_lock,
-                                               0,
+                                               G1SATBProcessCompletedThreshold,
                                                Shared_SATB_Q_lock);
 
   JavaThread::dirty_card_queue_set().initialize(DirtyCardQ_CBL_mon,
                                                 DirtyCardQ_FL_lock,
-                                                G1UpdateBufferQueueMaxLength,
+                                                concurrent_g1_refine()->yellow_zone(),
+                                                concurrent_g1_refine()->red_zone(),
                                                 Shared_DirtyCardQ_lock);
 
   if (G1DeferredRSUpdate) {
     dirty_card_queue_set().initialize(DirtyCardQ_CBL_mon,
                                       DirtyCardQ_FL_lock,
-                                      0,
+                                      -1, // never trigger processing
+                                      -1, // no limit on length
                                       Shared_DirtyCardQ_lock,
                                       &JavaThread::dirty_card_queue_set());
   }
@@ -1732,13 +1737,6 @@ size_t G1CollectedHeap::unsafe_max_alloc() {
   return car->free();
 }
 
-void G1CollectedHeap::collect(GCCause::Cause cause) {
-  // The caller doesn't have the Heap_lock
-  assert(!Heap_lock->owned_by_self(), "this thread should not own the Heap_lock");
-  MutexLocker ml(Heap_lock);
-  collect_locked(cause);
-}
-
 void G1CollectedHeap::collect_as_vm_thread(GCCause::Cause cause) {
   assert(Thread::current()->is_VM_thread(), "Precondition#1");
   assert(Heap_lock->is_locked(), "Precondition#2");
@@ -1755,17 +1753,31 @@ void G1CollectedHeap::collect_as_vm_thread(GCCause::Cause cause) {
   }
 }
 
+void G1CollectedHeap::collect(GCCause::Cause cause) {
+  // The caller doesn't have the Heap_lock
+  assert(!Heap_lock->owned_by_self(), "this thread should not own the Heap_lock");
 
-void G1CollectedHeap::collect_locked(GCCause::Cause cause) {
+  int gc_count_before;
-  // Don't want to do a GC until cleanup is completed.
-  wait_for_cleanup_complete();
-
-  // Read the GC count while holding the Heap_lock
-  int gc_count_before = SharedHeap::heap()->total_collections();
   {
-    MutexUnlocker mu(Heap_lock);  // give up heap lock, execute gets it back
+    MutexLocker ml(Heap_lock);
-    VM_G1CollectFull op(gc_count_before, cause);
+    // Read the GC count while holding the Heap_lock
-    VMThread::execute(&op);
+    gc_count_before = SharedHeap::heap()->total_collections();
+
+    // Don't want to do a GC until cleanup is completed.
+    wait_for_cleanup_complete();
+  } // We give up heap lock; VMThread::execute gets it back below
+  switch (cause) {
+    case GCCause::_scavenge_alot: {
+      // Do an incremental pause, which might sometimes be abandoned.
+      VM_G1IncCollectionPause op(gc_count_before, cause);
+      VMThread::execute(&op);
+      break;
+    }
+    default: {
+      // In all other cases, we currently do a full gc.
+      VM_G1CollectFull op(gc_count_before, cause);
+      VMThread::execute(&op);
+    }
   }
 }
 
@@ -2119,7 +2131,7 @@ size_t G1CollectedHeap::large_typearray_limit() {
 }
 
 size_t G1CollectedHeap::max_capacity() const {
-  return _g1_committed.byte_size();
+  return g1_reserved_obj_bytes();
 }
 
 jlong G1CollectedHeap::millis_since_last_gc() {
@@ -2638,6 +2650,8 @@ G1CollectedHeap::do_collection_pause_at_safepoint() {
   }
 
   {
+    ResourceMark rm;
+
     char verbose_str[128];
     sprintf(verbose_str, "GC pause ");
     if (g1_policy()->in_young_gc_mode()) {
@@ -2649,8 +2663,6 @@ G1CollectedHeap::do_collection_pause_at_safepoint() {
     if (g1_policy()->should_initiate_conc_mark())
       strcat(verbose_str, " (initial-mark)");
 
-    GCCauseSetter x(this, GCCause::_g1_inc_collection_pause);
-
     // if PrintGCDetails is on, we'll print long statistics information
     // in the collector policy code, so let's not print this as the output
     // is messy if we do.
@@ -2658,7 +2670,8 @@ G1CollectedHeap::do_collection_pause_at_safepoint() {
     TraceCPUTime tcpu(PrintGCDetails, true, gclog_or_tty);
     TraceTime t(verbose_str, PrintGC && !PrintGCDetails, true, gclog_or_tty);
 
-    ResourceMark rm;
+    TraceMemoryManagerStats tms(false /* fullGC */);
+
     assert(SafepointSynchronize::is_at_safepoint(), "should be at safepoint");
     assert(Thread::current() == VMThread::vm_thread(), "should be in vm thread");
     guarantee(!is_gc_active(), "collection is not reentrant");
@@ -2802,6 +2815,22 @@ G1CollectedHeap::do_collection_pause_at_safepoint() {
           _young_list->reset_auxilary_lists();
         }
       } else {
+        if (_in_cset_fast_test != NULL) {
+          assert(_in_cset_fast_test_base != NULL, "Since _in_cset_fast_test isn't");
+          FREE_C_HEAP_ARRAY(bool, _in_cset_fast_test_base);
+          //  this is more for peace of mind; we're nulling them here and
+          // we're expecting them to be null at the beginning of the next GC
+          _in_cset_fast_test = NULL;
+          _in_cset_fast_test_base = NULL;
+        }
+        // This looks confusing, because the DPT should really be empty
+        // at this point -- since we have not done any collection work,
+        // there should not be any derived pointers in the table to update;
+        // however, there is some additional state in the DPT which is
+        // reset at the end of the (null) "gc" here via the following call.
+        // A better approach might be to split off that state resetting work
+        // into a separate method that asserts that the DPT is empty and call
+        // that here. That is deferred for now.
         COMPILER2_PRESENT(DerivedPointerTable::update_pointers());
       }
 
@@ -2838,6 +2867,8 @@ G1CollectedHeap::do_collection_pause_at_safepoint() {
 
       assert(regions_accounted_for(), "Region leakage.");
 
+      MemoryService::track_memory_usage();
+
       if (VerifyAfterGC && total_collections() >= VerifyGCStartAt) {
         HandleMark hm;  // Discard invalid handles created during verification
         gclog_or_tty->print(" VerifyAfterGC:");
@@ -4209,10 +4240,11 @@ void G1CollectedHeap::evacuate_collection_set() {
     RedirtyLoggedCardTableEntryFastClosure redirty;
     dirty_card_queue_set().set_closure(&redirty);
     dirty_card_queue_set().apply_closure_to_all_completed_buffers();
-    JavaThread::dirty_card_queue_set().merge_bufferlists(&dirty_card_queue_set());
+
+    DirtyCardQueueSet& dcq = JavaThread::dirty_card_queue_set();
+    dcq.merge_bufferlists(&dirty_card_queue_set());
     assert(dirty_card_queue_set().completed_buffers_num() == 0, "All should be consumed");
   }
-
   COMPILER2_PRESENT(DerivedPointerTable::update_pointers());
 }
 

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

@@ -692,7 +692,7 @@ public:
 
   // Reserved (g1 only; super method includes perm), capacity and the used
   // portion in bytes.
-  size_t g1_reserved_obj_bytes() { return _g1_reserved.byte_size(); }
+  size_t g1_reserved_obj_bytes() const { return _g1_reserved.byte_size(); }
   virtual size_t capacity() const;
   virtual size_t used() const;
   // This should be called when we're not holding the heap lock. The

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

@@ -1516,8 +1516,30 @@ void G1CollectorPolicy::record_collection_pause_end(bool abandoned) {
       (end_time_sec - _recent_prev_end_times_for_all_gcs_sec->oldest()) * 1000.0;
     update_recent_gc_times(end_time_sec, elapsed_ms);
     _recent_avg_pause_time_ratio = _recent_gc_times_ms->sum()/interval_ms;
-    // using 1.01 to account for floating point inaccuracies
+    if (recent_avg_pause_time_ratio() < 0.0 ||
-    assert(recent_avg_pause_time_ratio() < 1.01, "All GC?");
+        (recent_avg_pause_time_ratio() - 1.0 > 0.0)) {
+#ifndef PRODUCT
+      // Dump info to allow post-facto debugging
+      gclog_or_tty->print_cr("recent_avg_pause_time_ratio() out of bounds");
+      gclog_or_tty->print_cr("-------------------------------------------");
+      gclog_or_tty->print_cr("Recent GC Times (ms):");
+      _recent_gc_times_ms->dump();
+      gclog_or_tty->print_cr("(End Time=%3.3f) Recent GC End Times (s):", end_time_sec);
+      _recent_prev_end_times_for_all_gcs_sec->dump();
+      gclog_or_tty->print_cr("GC = %3.3f, Interval = %3.3f, Ratio = %3.3f",
+                             _recent_gc_times_ms->sum(), interval_ms, recent_avg_pause_time_ratio());
+      // In debug mode, terminate the JVM if the user wants to debug at this point.
+      assert(!G1FailOnFPError, "Debugging data for CR 6898948 has been dumped above");
+#endif  // !PRODUCT
+      // Clip ratio between 0.0 and 1.0, and continue. This will be fixed in
+      // CR 6902692 by redoing the manner in which the ratio is incrementally computed.
+      if (_recent_avg_pause_time_ratio < 0.0) {
+        _recent_avg_pause_time_ratio = 0.0;
+      } else {
+        assert(_recent_avg_pause_time_ratio - 1.0 > 0.0, "Ctl-point invariant");
+        _recent_avg_pause_time_ratio = 1.0;
+      }
+    }
   }
 
   if (G1PolicyVerbose > 1) {
@@ -1892,6 +1914,10 @@ void G1CollectorPolicy::record_collection_pause_end(bool abandoned) {
   calculate_young_list_min_length();
   calculate_young_list_target_config();
 
+  // Note that _mmu_tracker->max_gc_time() returns the time in seconds.
+  double update_rs_time_goal_ms = _mmu_tracker->max_gc_time() * MILLIUNITS * G1RSUpdatePauseFractionPercent / 100.0;
+  adjust_concurrent_refinement(update_rs_time, update_rs_processed_buffers, update_rs_time_goal_ms);
+
   // </NEW PREDICTION>
 
   _target_pause_time_ms = -1.0;
@@ -1899,6 +1925,47 @@ void G1CollectorPolicy::record_collection_pause_end(bool abandoned) {
 
 // <NEW PREDICTION>
 
+void G1CollectorPolicy::adjust_concurrent_refinement(double update_rs_time,
+                                                     double update_rs_processed_buffers,
+                                                     double goal_ms) {
+  DirtyCardQueueSet& dcqs = JavaThread::dirty_card_queue_set();
+  ConcurrentG1Refine *cg1r = G1CollectedHeap::heap()->concurrent_g1_refine();
+
+  if (G1AdaptiveConcRefine) {
+    const int k_gy = 3, k_gr = 6;
+    const double inc_k = 1.1, dec_k = 0.9;
+
+    int g = cg1r->green_zone();
+    if (update_rs_time > goal_ms) {
+      g = (int)(g * dec_k);  // Can become 0, that's OK. That would mean a mutator-only processing.
+    } else {
+      if (update_rs_time < goal_ms && update_rs_processed_buffers > g) {
+        g = (int)MAX2(g * inc_k, g + 1.0);
+      }
+    }
+    // Change the refinement threads params
+    cg1r->set_green_zone(g);
+    cg1r->set_yellow_zone(g * k_gy);
+    cg1r->set_red_zone(g * k_gr);
+    cg1r->reinitialize_threads();
+
+    int processing_threshold_delta = MAX2((int)(cg1r->green_zone() * sigma()), 1);
+    int processing_threshold = MIN2(cg1r->green_zone() + processing_threshold_delta,
+                                    cg1r->yellow_zone());
+    // Change the barrier params
+    dcqs.set_process_completed_threshold(processing_threshold);
+    dcqs.set_max_completed_queue(cg1r->red_zone());
+  }
+
+  int curr_queue_size = dcqs.completed_buffers_num();
+  if (curr_queue_size >= cg1r->yellow_zone()) {
+    dcqs.set_completed_queue_padding(curr_queue_size);
+  } else {
+    dcqs.set_completed_queue_padding(0);
+  }
+  dcqs.notify_if_necessary();
+}
+
 double
 G1CollectorPolicy::
 predict_young_collection_elapsed_time_ms(size_t adjustment) {
@@ -2825,8 +2892,15 @@ choose_collection_set() {
   double non_young_start_time_sec;
   start_recording_regions();
 
-  guarantee(_target_pause_time_ms > -1.0,
+  guarantee(_target_pause_time_ms > -1.0
+            NOT_PRODUCT(|| Universe::heap()->gc_cause() == GCCause::_scavenge_alot),
             "_target_pause_time_ms should have been set!");
+#ifndef PRODUCT
+  if (_target_pause_time_ms <= -1.0) {
+    assert(ScavengeALot && Universe::heap()->gc_cause() == GCCause::_scavenge_alot, "Error");
+    _target_pause_time_ms = _mmu_tracker->max_gc_time() * 1000.0;
+  }
+#endif
   assert(_collection_set == NULL, "Precondition");
 
   double base_time_ms = predict_base_elapsed_time_ms(_pending_cards);
@@ -2972,7 +3046,3 @@ record_collection_pause_end(bool abandoned) {
   G1CollectorPolicy::record_collection_pause_end(abandoned);
   assert(assertMarkedBytesDataOK(), "Marked regions not OK at pause end.");
 }
-
-// Local Variables: ***
-// c-indentation-style: gnu ***
-// End: ***

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

@@ -316,6 +316,10 @@ private:
   bool verify_young_ages(HeapRegion* head, SurvRateGroup *surv_rate_group);
 #endif // PRODUCT
 
+  void adjust_concurrent_refinement(double update_rs_time,
+                                    double update_rs_processed_buffers,
+                                    double goal_ms);
+
 protected:
   double _pause_time_target_ms;
   double _recorded_young_cset_choice_time_ms;

hotspot/src/share/vm/gc_implementation/g1/g1MMUTracker.cpp

@@ -86,12 +86,22 @@ void G1MMUTrackerQueue::add_pause(double start, double end, bool gc_thread) {
     //   increase the array size (:-)
     //   remove the oldest entry (this might allow more GC time for
     //     the time slice than what's allowed)
-    //   concolidate the two entries with the minimum gap between them
+    //   consolidate the two entries with the minimum gap between them
-    //     (this mighte allow less GC time than what's allowed)
+    //     (this might allow less GC time than what's allowed)
-    guarantee(0, "array full, currently we can't recover");
+    guarantee(NOT_PRODUCT(ScavengeALot ||) G1ForgetfulMMUTracker,
+              "array full, currently we can't recover unless +G1ForgetfulMMUTracker");
+    // In the case where ScavengeALot is true, such overflow is not
+    // uncommon; in such cases, we can, without much loss of precision
+    // or performance (we are GC'ing most of the time anyway!),
+    // simply overwrite the oldest entry in the tracker: this
+    // is also the behaviour when G1ForgetfulMMUTracker is enabled.
+    _head_index = trim_index(_head_index + 1);
+    assert(_head_index == _tail_index, "Because we have a full circular buffer");
+    _tail_index = trim_index(_tail_index + 1);
+  } else {
+    _head_index = trim_index(_head_index + 1);
+    ++_no_entries;
   }
-  _head_index = trim_index(_head_index + 1);
-  ++_no_entries;
   _array[_head_index] = G1MMUTrackerQueueElem(start, end);
 }
 

hotspot/src/share/vm/gc_implementation/g1/g1MMUTracker.hpp

@@ -99,7 +99,10 @@ private:
   // The array is of fixed size and I don't think we'll need more than
   // two or three entries with the current behaviour of G1 pauses.
   // If the array is full, an easy fix is to look for the pauses with
-  // the shortest gap between them and concolidate them.
+  // the shortest gap between them and consolidate them.
+  // For now, we have taken the expedient alternative of forgetting
+  // the oldest entry in the event that +G1ForgetfulMMUTracker, thus
+  // potentially violating MMU specs for some time thereafter.
 
   G1MMUTrackerQueueElem _array[QueueLength];
   int                   _head_index;

hotspot/src/share/vm/gc_implementation/g1/g1_globals.hpp

@@ -85,7 +85,7 @@
   diagnostic(bool, G1SummarizeZFStats, false,                               \
           "Summarize zero-filling info")                                    \
                                                                             \
-  develop(bool, G1TraceConcurrentRefinement, false,                         \
+  diagnostic(bool, G1TraceConcurrentRefinement, false,                      \
           "Trace G1 concurrent refinement")                                 \
                                                                             \
   product(intx, G1MarkStackSize, 2 * 1024 * 1024,                           \
@@ -94,19 +94,6 @@
   product(intx, G1MarkRegionStackSize, 1024 * 1024,                         \
           "Size of the region stack for concurrent marking.")               \
                                                                             \
-  develop(bool, G1ConcRefine, true,                                         \
-          "If true, run concurrent rem set refinement for G1")              \
-                                                                            \
-  develop(intx, G1ConcRefineTargTraversals, 4,                              \
-          "Number of concurrent refinement we try to achieve")              \
-                                                                            \
-  develop(intx, G1ConcRefineInitialDelta, 4,                                \
-          "Number of heap regions of alloc ahead of starting collection "   \
-          "pause to start concurrent refinement (initially)")               \
-                                                                            \
-  develop(bool, G1SmoothConcRefine, true,                                   \
-          "Attempts to smooth out the overhead of concurrent refinement")   \
-                                                                            \
   develop(bool, G1ConcZeroFill, true,                                       \
           "If true, run concurrent zero-filling thread")                    \
                                                                             \
@@ -178,13 +165,38 @@
   product(intx, G1UpdateBufferSize, 256,                                    \
           "Size of an update buffer")                                       \
                                                                             \
-  product(intx, G1UpdateBufferQueueProcessingThreshold, 5,                  \
+  product(intx, G1ConcRefineYellowZone, 0,                                  \
           "Number of enqueued update buffers that will "                    \
-          "trigger concurrent processing")                                  \
+          "trigger concurrent processing. Will be selected ergonomically "  \
+          "by default.")                                                    \
                                                                             \
-  product(intx, G1UpdateBufferQueueMaxLength, 30,                           \
+  product(intx, G1ConcRefineRedZone, 0,                                     \
           "Maximum number of enqueued update buffers before mutator "       \
-          "threads start processing new ones instead of enqueueing them")   \
+          "threads start processing new ones instead of enqueueing them. "  \
+          "Will be selected ergonomically by default. Zero will disable "   \
+          "concurrent processing.")                                         \
+                                                                            \
+  product(intx, G1ConcRefineGreenZone, 0,                                   \
+          "The number of update buffers that are left in the queue by the " \
+          "concurrent processing threads. Will be selected ergonomically "  \
+          "by default.")                                                    \
+                                                                            \
+  product(intx, G1ConcRefineServiceInterval, 300,                           \
+          "The last concurrent refinement thread wakes up every "           \
+          "specified number of milliseconds to do miscellaneous work.")     \
+                                                                            \
+  product(intx, G1ConcRefineThresholdStep, 0,                               \
+          "Each time the rset update queue increases by this amount "       \
+          "activate the next refinement thread if available. "              \
+          "Will be selected ergonomically by default.")                     \
+                                                                            \
+  product(intx, G1RSUpdatePauseFractionPercent, 10,                         \
+          "A target percentage of time that is allowed to be spend on "     \
+          "process RS update buffers during the collection pause.")         \
+                                                                            \
+  product(bool, G1AdaptiveConcRefine, true,                                 \
+          "Select green, yellow and red zones adaptively to meet the "      \
+          "the pause requirements.")                                        \
                                                                             \
   develop(intx, G1ConcRSLogCacheSize, 10,                                   \
           "Log base 2 of the length of conc RS hot-card cache.")            \
@@ -242,6 +254,10 @@
   product(bool, G1UseSurvivorSpaces, true,                                  \
           "When true, use survivor space.")                                 \
                                                                             \
+  develop(bool, G1FailOnFPError, false,                                     \
+          "When set, G1 will fail when it encounters an FP 'error', "       \
+          "so as to allow debugging")                                       \
+                                                                            \
   develop(bool, G1FixedTenuringThreshold, false,                            \
           "When set, G1 will not adjust the tenuring threshold")            \
                                                                             \
@@ -252,6 +268,9 @@
           "If non-0 is the size of the G1 survivor space, "                 \
           "otherwise SurvivorRatio is used to determine the size")          \
                                                                             \
+  product(bool, G1ForgetfulMMUTracker, false,                               \
+          "If the MMU tracker's memory is full, forget the oldest entry")   \
+                                                                            \
   product(uintx, G1HeapRegionSize, 0,                                       \
           "Size of the G1 regions.")                                        \
                                                                             \

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

@@ -64,8 +64,8 @@ void PtrQueue::enqueue_known_active(void* ptr) {
   while (_index == 0) {
     handle_zero_index();
   }
-  assert(_index > 0, "postcondition");
 
+  assert(_index > 0, "postcondition");
   _index -= oopSize;
   _buf[byte_index_to_index((int)_index)] = ptr;
   assert(0 <= _index && _index <= _sz, "Invariant.");
@@ -99,94 +99,110 @@ void** PtrQueueSet::allocate_buffer() {
   assert(_sz > 0, "Didn't set a buffer size.");
   MutexLockerEx x(_fl_owner->_fl_lock, Mutex::_no_safepoint_check_flag);
   if (_fl_owner->_buf_free_list != NULL) {
-    void** res = _fl_owner->_buf_free_list;
+    void** res = BufferNode::make_buffer_from_node(_fl_owner->_buf_free_list);
-    _fl_owner->_buf_free_list = (void**)_fl_owner->_buf_free_list[0];
+    _fl_owner->_buf_free_list = _fl_owner->_buf_free_list->next();
     _fl_owner->_buf_free_list_sz--;
-    // Just override the next pointer with NULL, just in case we scan this part
-    // of the buffer.
-    res[0] = NULL;
     return res;
   } else {
-    return NEW_C_HEAP_ARRAY(void*, _sz);
+    // Allocate space for the BufferNode in front of the buffer.
+    char *b =  NEW_C_HEAP_ARRAY(char, _sz + BufferNode::aligned_size());
+    return BufferNode::make_buffer_from_block(b);
   }
 }
 
 void PtrQueueSet::deallocate_buffer(void** buf) {
   assert(_sz > 0, "Didn't set a buffer size.");
   MutexLockerEx x(_fl_owner->_fl_lock, Mutex::_no_safepoint_check_flag);
-  buf[0] = (void*)_fl_owner->_buf_free_list;
+  BufferNode *node = BufferNode::make_node_from_buffer(buf);
-  _fl_owner->_buf_free_list = buf;
+  node->set_next(_fl_owner->_buf_free_list);
+  _fl_owner->_buf_free_list = node;
   _fl_owner->_buf_free_list_sz++;
 }
 
 void PtrQueueSet::reduce_free_list() {
+  assert(_fl_owner == this, "Free list reduction is allowed only for the owner");
   // For now we'll adopt the strategy of deleting half.
   MutexLockerEx x(_fl_lock, Mutex::_no_safepoint_check_flag);
   size_t n = _buf_free_list_sz / 2;
   while (n > 0) {
     assert(_buf_free_list != NULL, "_buf_free_list_sz must be wrong.");
-    void** head = _buf_free_list;
+    void* b = BufferNode::make_block_from_node(_buf_free_list);
-    _buf_free_list = (void**)_buf_free_list[0];
+    _buf_free_list = _buf_free_list->next();
-    FREE_C_HEAP_ARRAY(void*,head);
+    FREE_C_HEAP_ARRAY(char, b);
+    _buf_free_list_sz --;
     n--;
   }
 }
 
-void PtrQueueSet::enqueue_complete_buffer(void** buf, size_t index, bool ignore_max_completed) {
+void PtrQueue::handle_zero_index() {
-  // I use explicit locking here because there's a bailout in the middle.
+  assert(0 == _index, "Precondition.");
-  _cbl_mon->lock_without_safepoint_check();
+  // This thread records the full buffer and allocates a new one (while
-
+  // holding the lock if there is one).
-  Thread* thread = Thread::current();
+  if (_buf != NULL) {
-  assert( ignore_max_completed ||
+    if (_lock) {
-          thread->is_Java_thread() ||
+      locking_enqueue_completed_buffer(_buf);
-          SafepointSynchronize::is_at_safepoint(),
+    } else {
-          "invariant" );
+      if (qset()->process_or_enqueue_complete_buffer(_buf)) {
-  ignore_max_completed = ignore_max_completed || !thread->is_Java_thread();
+        // Recycle the buffer. No allocation.
-
+        _sz = qset()->buffer_size();
-  if (!ignore_max_completed && _max_completed_queue > 0 &&
+        _index = _sz;
-      _n_completed_buffers >= (size_t) _max_completed_queue) {
+        return;
-    _cbl_mon->unlock();
+      }
-    bool b = mut_process_buffer(buf);
-    if (b) {
-      deallocate_buffer(buf);
-      return;
     }
-
-    // Otherwise, go ahead and enqueue the buffer.  Must reaquire the lock.
-    _cbl_mon->lock_without_safepoint_check();
   }
+  // Reallocate the buffer
+  _buf = qset()->allocate_buffer();
+  _sz = qset()->buffer_size();
+  _index = _sz;
+  assert(0 <= _index && _index <= _sz, "Invariant.");
+}
 
-  // Here we still hold the _cbl_mon.
+bool PtrQueueSet::process_or_enqueue_complete_buffer(void** buf) {
-  CompletedBufferNode* cbn = new CompletedBufferNode;
+  if (Thread::current()->is_Java_thread()) {
-  cbn->buf = buf;
+    // We don't lock. It is fine to be epsilon-precise here.
-  cbn->next = NULL;
+    if (_max_completed_queue == 0 || _max_completed_queue > 0 &&
-  cbn->index = index;
+        _n_completed_buffers >= _max_completed_queue + _completed_queue_padding) {
+      bool b = mut_process_buffer(buf);
+      if (b) {
+        // True here means that the buffer hasn't been deallocated and the caller may reuse it.
+        return true;
+      }
+    }
+  }
+  // The buffer will be enqueued. The caller will have to get a new one.
+  enqueue_complete_buffer(buf);
+  return false;
+}
+
+void PtrQueueSet::enqueue_complete_buffer(void** buf, size_t index) {
+  MutexLockerEx x(_cbl_mon, Mutex::_no_safepoint_check_flag);
+  BufferNode* cbn = BufferNode::new_from_buffer(buf);
+  cbn->set_index(index);
   if (_completed_buffers_tail == NULL) {
     assert(_completed_buffers_head == NULL, "Well-formedness");
     _completed_buffers_head = cbn;
     _completed_buffers_tail = cbn;
   } else {
-    _completed_buffers_tail->next = cbn;
+    _completed_buffers_tail->set_next(cbn);
     _completed_buffers_tail = cbn;
   }
   _n_completed_buffers++;
 
-  if (!_process_completed &&
+  if (!_process_completed && _process_completed_threshold >= 0 &&
       _n_completed_buffers >= _process_completed_threshold) {
     _process_completed = true;
     if (_notify_when_complete)
-      _cbl_mon->notify_all();
+      _cbl_mon->notify();
   }
   debug_only(assert_completed_buffer_list_len_correct_locked());
-  _cbl_mon->unlock();
 }
 
 int PtrQueueSet::completed_buffers_list_length() {
   int n = 0;
-  CompletedBufferNode* cbn = _completed_buffers_head;
+  BufferNode* cbn = _completed_buffers_head;
   while (cbn != NULL) {
     n++;
-    cbn = cbn->next;
+    cbn = cbn->next();
   }
   return n;
 }
@@ -197,7 +213,7 @@ void PtrQueueSet::assert_completed_buffer_list_len_correct() {
 }
 
 void PtrQueueSet::assert_completed_buffer_list_len_correct_locked() {
-  guarantee((size_t)completed_buffers_list_length() ==  _n_completed_buffers,
+  guarantee(completed_buffers_list_length() ==  _n_completed_buffers,
             "Completed buffer length is wrong.");
 }
 
@@ -206,12 +222,8 @@ void PtrQueueSet::set_buffer_size(size_t sz) {
   _sz = sz * oopSize;
 }
 
-void PtrQueueSet::set_process_completed_threshold(size_t sz) {
+// Merge lists of buffers. Notify the processing threads.
-  _process_completed_threshold = sz;
+// The source queue is emptied as a result. The queues
-}
-
-// Merge lists of buffers. Notify waiting threads if the length of the list
-// exceeds threshold. The source queue is emptied as a result. The queues
 // must share the monitor.
 void PtrQueueSet::merge_bufferlists(PtrQueueSet *src) {
   assert(_cbl_mon == src->_cbl_mon, "Should share the same lock");
@@ -223,7 +235,7 @@ void PtrQueueSet::merge_bufferlists(PtrQueueSet *src) {
   } else {
     assert(_completed_buffers_head != NULL, "Well formedness");
     if (src->_completed_buffers_head != NULL) {
-      _completed_buffers_tail->next = src->_completed_buffers_head;
+      _completed_buffers_tail->set_next(src->_completed_buffers_head);
       _completed_buffers_tail = src->_completed_buffers_tail;
     }
   }
@@ -236,31 +248,13 @@ void PtrQueueSet::merge_bufferlists(PtrQueueSet *src) {
   assert(_completed_buffers_head == NULL && _completed_buffers_tail == NULL ||
          _completed_buffers_head != NULL && _completed_buffers_tail != NULL,
          "Sanity");
+}
 
-  if (!_process_completed &&
+void PtrQueueSet::notify_if_necessary() {
-      _n_completed_buffers >= _process_completed_threshold) {
+  MutexLockerEx x(_cbl_mon, Mutex::_no_safepoint_check_flag);
+  if (_n_completed_buffers >= _process_completed_threshold || _max_completed_queue == 0) {
     _process_completed = true;
     if (_notify_when_complete)
-      _cbl_mon->notify_all();
+      _cbl_mon->notify();
   }
 }
-
-// Merge free lists of the two queues. The free list of the source
-// queue is emptied as a result. The queues must share the same
-// mutex that guards free lists.
-void PtrQueueSet::merge_freelists(PtrQueueSet* src) {
-  assert(_fl_lock == src->_fl_lock, "Should share the same lock");
-  MutexLockerEx x(_fl_lock, Mutex::_no_safepoint_check_flag);
-  if (_buf_free_list != NULL) {
-    void **p = _buf_free_list;
-    while (*p != NULL) {
-      p = (void**)*p;
-    }
-    *p = src->_buf_free_list;
-  } else {
-    _buf_free_list = src->_buf_free_list;
-  }
-  _buf_free_list_sz += src->_buf_free_list_sz;
-  src->_buf_free_list = NULL;
-  src->_buf_free_list_sz = 0;
-}

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

@@ -27,8 +27,10 @@
 // the addresses of modified old-generation objects.  This type supports
 // this operation.
 
-class PtrQueueSet;
+// The definition of placement operator new(size_t, void*) in the <new>.
+#include <new>
 
+class PtrQueueSet;
 class PtrQueue VALUE_OBJ_CLASS_SPEC {
 
 protected:
@@ -77,7 +79,7 @@ public:
     else enqueue_known_active(ptr);
   }
 
-  inline void handle_zero_index();
+  void handle_zero_index();
   void locking_enqueue_completed_buffer(void** buf);
 
   void enqueue_known_active(void* ptr);
@@ -126,34 +128,65 @@ public:
 
 };
 
+class BufferNode {
+  size_t _index;
+  BufferNode* _next;
+public:
+  BufferNode() : _index(0), _next(NULL) { }
+  BufferNode* next() const     { return _next;  }
+  void set_next(BufferNode* n) { _next = n;     }
+  size_t index() const         { return _index; }
+  void set_index(size_t i)     { _index = i;    }
+
+  // Align the size of the structure to the size of the pointer
+  static size_t aligned_size() {
+    static const size_t alignment = round_to(sizeof(BufferNode), sizeof(void*));
+    return alignment;
+  }
+
+  // BufferNode is allocated before the buffer.
+  // The chunk of memory that holds both of them is a block.
+
+  // Produce a new BufferNode given a buffer.
+  static BufferNode* new_from_buffer(void** buf) {
+    return new (make_block_from_buffer(buf)) BufferNode;
+  }
+
+  // The following are the required conversion routines:
+  static BufferNode* make_node_from_buffer(void** buf) {
+    return (BufferNode*)make_block_from_buffer(buf);
+  }
+  static void** make_buffer_from_node(BufferNode *node) {
+    return make_buffer_from_block(node);
+  }
+  static void* make_block_from_node(BufferNode *node) {
+    return (void*)node;
+  }
+  static void** make_buffer_from_block(void* p) {
+    return (void**)((char*)p + aligned_size());
+  }
+  static void* make_block_from_buffer(void** p) {
+    return (void*)((char*)p - aligned_size());
+  }
+};
+
 // A PtrQueueSet represents resources common to a set of pointer queues.
 // In particular, the individual queues allocate buffers from this shared
 // set, and return completed buffers to the set.
 // All these variables are are protected by the TLOQ_CBL_mon. XXX ???
 class PtrQueueSet VALUE_OBJ_CLASS_SPEC {
-
 protected:
-
-  class CompletedBufferNode: public CHeapObj {
-  public:
-    void** buf;
-    size_t index;
-    CompletedBufferNode* next;
-    CompletedBufferNode() : buf(NULL),
-      index(0), next(NULL){ }
-  };
-
   Monitor* _cbl_mon;  // Protects the fields below.
-  CompletedBufferNode* _completed_buffers_head;
+  BufferNode* _completed_buffers_head;
-  CompletedBufferNode* _completed_buffers_tail;
+  BufferNode* _completed_buffers_tail;
-  size_t _n_completed_buffers;
+  int _n_completed_buffers;
-  size_t _process_completed_threshold;
+  int _process_completed_threshold;
   volatile bool _process_completed;
 
   // This (and the interpretation of the first element as a "next"
   // pointer) are protected by the TLOQ_FL_lock.
   Mutex* _fl_lock;
-  void** _buf_free_list;
+  BufferNode* _buf_free_list;
   size_t _buf_free_list_sz;
   // Queue set can share a freelist. The _fl_owner variable
   // specifies the owner. It is set to "this" by default.
@@ -170,6 +203,7 @@ protected:
   // Maximum number of elements allowed on completed queue: after that,
   // enqueuer does the work itself.  Zero indicates no maximum.
   int _max_completed_queue;
+  int _completed_queue_padding;
 
   int completed_buffers_list_length();
   void assert_completed_buffer_list_len_correct_locked();
@@ -191,9 +225,12 @@ public:
   // Because of init-order concerns, we can't pass these as constructor
   // arguments.
   void initialize(Monitor* cbl_mon, Mutex* fl_lock,
-                  int max_completed_queue = 0,
+                  int process_completed_threshold,
+                  int max_completed_queue,
                   PtrQueueSet *fl_owner = NULL) {
     _max_completed_queue = max_completed_queue;
+    _process_completed_threshold = process_completed_threshold;
+    _completed_queue_padding = 0;
     assert(cbl_mon != NULL && fl_lock != NULL, "Init order issue?");
     _cbl_mon = cbl_mon;
     _fl_lock = fl_lock;
@@ -208,14 +245,17 @@ public:
   void deallocate_buffer(void** buf);
 
   // Declares that "buf" is a complete buffer.
-  void enqueue_complete_buffer(void** buf, size_t index = 0,
+  void enqueue_complete_buffer(void** buf, size_t index = 0);
-                               bool ignore_max_completed = false);
+
+  // To be invoked by the mutator.
+  bool process_or_enqueue_complete_buffer(void** buf);
 
   bool completed_buffers_exist_dirty() {
     return _n_completed_buffers > 0;
   }
 
   bool process_completed_buffers() { return _process_completed; }
+  void set_process_completed(bool x) { _process_completed = x; }
 
   bool active() { return _all_active; }
 
@@ -226,15 +266,24 @@ public:
   // Get the buffer size.
   size_t buffer_size() { return _sz; }
 
-  // Set the number of completed buffers that triggers log processing.
+  // Get/Set the number of completed buffers that triggers log processing.
-  void set_process_completed_threshold(size_t sz);
+  void set_process_completed_threshold(int sz) { _process_completed_threshold = sz; }
+  int process_completed_threshold() const { return _process_completed_threshold; }
 
   // Must only be called at a safe point.  Indicates that the buffer free
   // list size may be reduced, if that is deemed desirable.
   void reduce_free_list();
 
-  size_t completed_buffers_num() { return _n_completed_buffers; }
+  int completed_buffers_num() { return _n_completed_buffers; }
 
   void merge_bufferlists(PtrQueueSet* src);
-  void merge_freelists(PtrQueueSet* src);
+
+  void set_max_completed_queue(int m) { _max_completed_queue = m; }
+  int max_completed_queue() { return _max_completed_queue; }
+
+  void set_completed_queue_padding(int padding) { _completed_queue_padding = padding; }
+  int completed_queue_padding() { return _completed_queue_padding; }
+
+  // Notify the consumer if the number of buffers crossed the threshold
+  void notify_if_necessary();
 };

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

@@ -67,9 +67,9 @@ SATBMarkQueueSet::SATBMarkQueueSet() :
 {}
 
 void SATBMarkQueueSet::initialize(Monitor* cbl_mon, Mutex* fl_lock,
-                                  int max_completed_queue,
+                                  int process_completed_threshold,
                                   Mutex* lock) {
-  PtrQueueSet::initialize(cbl_mon, fl_lock, max_completed_queue);
+  PtrQueueSet::initialize(cbl_mon, fl_lock, process_completed_threshold, -1);
   _shared_satb_queue.set_lock(lock);
   if (ParallelGCThreads > 0) {
     _par_closures = NEW_C_HEAP_ARRAY(ObjectClosure*, ParallelGCThreads);
@@ -122,12 +122,12 @@ void SATBMarkQueueSet::par_iterate_closure_all_threads(int worker) {
 
 bool SATBMarkQueueSet::apply_closure_to_completed_buffer_work(bool par,
                                                               int worker) {
-  CompletedBufferNode* nd = NULL;
+  BufferNode* nd = NULL;
   {
     MutexLockerEx x(_cbl_mon, Mutex::_no_safepoint_check_flag);
     if (_completed_buffers_head != NULL) {
       nd = _completed_buffers_head;
-      _completed_buffers_head = nd->next;
+      _completed_buffers_head = nd->next();
       if (_completed_buffers_head == NULL) _completed_buffers_tail = NULL;
       _n_completed_buffers--;
       if (_n_completed_buffers == 0) _process_completed = false;
@@ -135,9 +135,9 @@ bool SATBMarkQueueSet::apply_closure_to_completed_buffer_work(bool par,
   }
   ObjectClosure* cl = (par ? _par_closures[worker] : _closure);
   if (nd != NULL) {
-    ObjPtrQueue::apply_closure_to_buffer(cl, nd->buf, 0, _sz);
+    void **buf = BufferNode::make_buffer_from_node(nd);
-    deallocate_buffer(nd->buf);
+    ObjPtrQueue::apply_closure_to_buffer(cl, buf, 0, _sz);
-    delete nd;
+    deallocate_buffer(buf);
     return true;
   } else {
     return false;
@@ -145,13 +145,13 @@ bool SATBMarkQueueSet::apply_closure_to_completed_buffer_work(bool par,
 }
 
 void SATBMarkQueueSet::abandon_partial_marking() {
-  CompletedBufferNode* buffers_to_delete = NULL;
+  BufferNode* buffers_to_delete = NULL;
   {
     MutexLockerEx x(_cbl_mon, Mutex::_no_safepoint_check_flag);
     while (_completed_buffers_head != NULL) {
-      CompletedBufferNode* nd = _completed_buffers_head;
+      BufferNode* nd = _completed_buffers_head;
-      _completed_buffers_head = nd->next;
+      _completed_buffers_head = nd->next();
-      nd->next = buffers_to_delete;
+      nd->set_next(buffers_to_delete);
       buffers_to_delete = nd;
     }
     _completed_buffers_tail = NULL;
@@ -159,10 +159,9 @@ void SATBMarkQueueSet::abandon_partial_marking() {
     DEBUG_ONLY(assert_completed_buffer_list_len_correct_locked());
   }
   while (buffers_to_delete != NULL) {
-    CompletedBufferNode* nd = buffers_to_delete;
+    BufferNode* nd = buffers_to_delete;
-    buffers_to_delete = nd->next;
+    buffers_to_delete = nd->next();
-    deallocate_buffer(nd->buf);
+    deallocate_buffer(BufferNode::make_buffer_from_node(nd));
-    delete nd;
   }
   assert(SafepointSynchronize::is_at_safepoint(), "Must be at safepoint.");
   // So we can safely manipulate these queues.

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

@@ -60,8 +60,8 @@ public:
   SATBMarkQueueSet();
 
   void initialize(Monitor* cbl_mon, Mutex* fl_lock,
-                  int max_completed_queue = 0,
+                  int process_completed_threshold,
-                  Mutex* lock = NULL);
+                  Mutex* lock);
 
   static void handle_zero_index_for_thread(JavaThread* t);
 

hotspot/src/share/vm/gc_implementation/g1/vm_operations_g1.cpp

@@ -42,7 +42,7 @@ void VM_G1CollectFull::doit() {
 void VM_G1IncCollectionPause::doit() {
   JvmtiGCForAllocationMarker jgcm;
   G1CollectedHeap* g1h = G1CollectedHeap::heap();
-  GCCauseSetter x(g1h, GCCause::_g1_inc_collection_pause);
+  GCCauseSetter x(g1h, _gc_cause);
   g1h->do_collection_pause_at_safepoint();
 }
 

hotspot/src/share/vm/gc_implementation/g1/vm_operations_g1.hpp

@@ -68,8 +68,9 @@ class VM_G1CollectForAllocation: public VM_GC_Operation {
 
 class VM_G1IncCollectionPause: public VM_GC_Operation {
  public:
-  VM_G1IncCollectionPause(int gc_count_before) :
+  VM_G1IncCollectionPause(int gc_count_before,
-    VM_GC_Operation(gc_count_before) {}
+                          GCCause::Cause gc_cause = GCCause::_g1_inc_collection_pause) :
+    VM_GC_Operation(gc_count_before) { _gc_cause = gc_cause; }
   virtual VMOp_Type type() const { return VMOp_G1IncCollectionPause; }
   virtual void doit();
   virtual const char* name() const {