changeset 4545:df0a816d95ef

Merge
author duke
date Wed, 05 Jul 2017 17:05:01 +0200
parents 4b85cf247673 c2c8c97ad733
children 235b2ff63e0a
files jdk/src/share/classes/sun/dyn/util/BytecodeSignature.java
diffstat 336 files changed, 27074 insertions(+), 3844 deletions(-) [+]
line wrap: on
line diff
--- a/.hgtags-top-repo	Thu Dec 17 14:10:29 2009 -0800
+++ b/.hgtags-top-repo	Wed Jul 05 17:05:01 2017 +0200
@@ -52,3 +52,4 @@
 d1516b9f23954b29b8e76e6f4efc467c08c78133 jdk7-b75
 c8b63075403d53a208104a8a6ea5072c1cb66aab jdk7-b76
 1f17ca8353babb13f4908c1f87d11508232518c8 jdk7-b77
+ab4ae8f4514693a9fe17ca2fec0239d8f8450d2c jdk7-b78
--- a/corba/.hgtags	Thu Dec 17 14:10:29 2009 -0800
+++ b/corba/.hgtags	Wed Jul 05 17:05:01 2017 +0200
@@ -52,3 +52,4 @@
 0fb137085952c8e47878e240d1cb40f14de463c4 jdk7-b75
 937144222e2219939101b0129d26a872a7956b13 jdk7-b76
 6881f0383f623394b5ec73f27a5f329ff55d0467 jdk7-b77
+a7f7276b48cd74d8eb1baa83fbf3d1ef4a2603c8 jdk7-b78
--- a/hotspot/.hgtags	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/.hgtags	Wed Jul 05 17:05:01 2017 +0200
@@ -52,3 +52,4 @@
 d8dd291a362acb656026a9c0a9da48501505a1e7 jdk7-b75
 9174bb32e934965288121f75394874eeb1fcb649 jdk7-b76
 455105fc81d941482f8f8056afaa7aa0949c9300 jdk7-b77
+e703499b4b51e3af756ae77c3d5e8b3058a14e4e jdk7-b78
--- a/hotspot/make/hotspot_version	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/make/hotspot_version	Wed Jul 05 17:05:01 2017 +0200
@@ -35,7 +35,7 @@
 
 HS_MAJOR_VER=17
 HS_MINOR_VER=0
-HS_BUILD_NUMBER=05
+HS_BUILD_NUMBER=06
 
 JDK_MAJOR_VER=1
 JDK_MINOR_VER=7
--- a/hotspot/make/linux/makefiles/debug.make	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/make/linux/makefiles/debug.make	Wed Jul 05 17:05:01 2017 +0200
@@ -38,7 +38,7 @@
  "Please use 'make jvmg' to build debug JVM.                            \n" \
  "----------------------------------------------------------------------\n")
 
-G_SUFFIX =
+G_SUFFIX = _g
 VERSION = debug
 SYSDEFS += -DASSERT -DDEBUG
 PICFLAGS = DEFAULT
--- a/hotspot/make/linux/makefiles/fastdebug.make	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/make/linux/makefiles/fastdebug.make	Wed Jul 05 17:05:01 2017 +0200
@@ -58,7 +58,7 @@
 # Linker mapfile
 MAPFILE = $(GAMMADIR)/make/linux/makefiles/mapfile-vers-debug
 
-G_SUFFIX =
+G_SUFFIX = _g
 VERSION = optimized
 SYSDEFS += -DASSERT -DFASTDEBUG
 PICFLAGS = DEFAULT
--- a/hotspot/make/linux/makefiles/jsig.make	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/make/linux/makefiles/jsig.make	Wed Jul 05 17:05:01 2017 +0200
@@ -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 @@
 	@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)"
--- a/hotspot/make/linux/makefiles/jvmg.make	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/make/linux/makefiles/jvmg.make	Wed Jul 05 17:05:01 2017 +0200
@@ -35,7 +35,7 @@
 # Linker mapfile
 MAPFILE = $(GAMMADIR)/make/linux/makefiles/mapfile-vers-debug
 
-G_SUFFIX =
+G_SUFFIX = _g
 VERSION = debug
 SYSDEFS += -DASSERT -DDEBUG
 PICFLAGS = DEFAULT
--- a/hotspot/make/linux/makefiles/launcher.make	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/make/linux/makefiles/launcher.make	Wed Jul 05 17:05:01 2017 +0200
@@ -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 @@
 	    $(LINK_LAUNCHER/PRE_HOOK) \
 	    $(LINK_LAUNCHER) $(LFLAGS_LAUNCHER) -o $@ $(LAUNCHER.o) $(LIBS_LAUNCHER); \
 	    $(LINK_LAUNCHER/POST_HOOK) \
+	    [ -f $(LAUNCHER_G) ] || { ln -s $@ $(LAUNCHER_G); }; \
         }
--- a/hotspot/make/linux/makefiles/saproc.make	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/make/linux/makefiles/saproc.make	Wed Jul 05 17:05:01 2017 +0200
@@ -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 @@
 	           $(SA_DEBUG_CFLAGS)                                   \
 	           -o $@                                                \
 	           -lthread_db
+	$(QUIETLY) [ -f $(LIBSAPROC_G) ] || { ln -s $@ $(LIBSAPROC_G); }
 
 install_saproc: checkAndBuildSA
 	$(QUIETLY) if [ -e $(LIBSAPROC) ] ; then             \
--- a/hotspot/make/linux/makefiles/vm.make	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/make/linux/makefiles/vm.make	Wed Jul 05 17:05:01 2017 +0200
@@ -113,8 +113,9 @@
 #----------------------------------------------------------------------
 # JVM
 
-JVM    = jvm$(G_SUFFIX)
-LIBJVM = lib$(JVM).so
+JVM      = jvm
+LIBJVM   = lib$(JVM).so
+LIBJVM_G = lib$(JVM)$(G_SUFFIX).so
 
 JVM_OBJ_FILES = $(Obj_Files)
 
@@ -201,6 +202,7 @@
 		       $(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					\
--- a/hotspot/make/solaris/makefiles/debug.make	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/make/solaris/makefiles/debug.make	Wed Jul 05 17:05:01 2017 +0200
@@ -54,7 +54,7 @@
  "Please use 'gnumake jvmg' to build debug JVM.                            \n" \
  "-------------------------------------------------------------------------\n")
 
-G_SUFFIX =
+G_SUFFIX = _g
 VERSION = debug
 SYSDEFS += -DASSERT -DDEBUG
 PICFLAGS = DEFAULT
--- a/hotspot/make/solaris/makefiles/dtrace.make	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/make/solaris/makefiles/dtrace.make	Wed Jul 05 17:05:01 2017 +0200
@@ -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
-# but not for CORE configuration
+# We build libjvm_dtrace/libjvm_db/dtrace for COMPILER1 and COMPILER2
+# but not for CORE or KERNEL configurations.
 
 ifneq ("${TYPE}", "CORE")
 ifneq ("${TYPE}", "KERNEL")
@@ -37,12 +37,13 @@
 
 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_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 @@
 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 @@
 	@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
--- a/hotspot/make/solaris/makefiles/fastdebug.make	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/make/solaris/makefiles/fastdebug.make	Wed Jul 05 17:05:01 2017 +0200
@@ -90,7 +90,6 @@
 # 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 @@
 # and mustn't be otherwise.
 MAPFILE_DTRACE = $(GAMMADIR)/make/solaris/makefiles/mapfile-vers-$(TYPE)
 
-
-G_SUFFIX =
+G_SUFFIX = _g
 VERSION = optimized
 SYSDEFS += -DASSERT -DFASTDEBUG -DCHECK_UNHANDLED_OOPS
 PICFLAGS = DEFAULT
--- a/hotspot/make/solaris/makefiles/jsig.make	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/make/solaris/makefiles/jsig.make	Wed Jul 05 17:05:01 2017 +0200
@@ -25,8 +25,11 @@
 # Rules to build signal interposition library, used by vm.make
 
 # libjsig[_g].so: signal interposition library
-JSIG = jsig$(G_SUFFIX)
-LIBJSIG = lib$(JSIG).so
+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
 
@@ -46,6 +49,7 @@
 	@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)"
--- a/hotspot/make/solaris/makefiles/jvmg.make	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/make/solaris/makefiles/jvmg.make	Wed Jul 05 17:05:01 2017 +0200
@@ -51,7 +51,7 @@
 # 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
--- a/hotspot/make/solaris/makefiles/launcher.make	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/make/solaris/makefiles/launcher.make	Wed Jul 05 17:05:01 2017 +0200
@@ -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 @@
 	    $(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
--- a/hotspot/make/solaris/makefiles/saproc.make	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/make/solaris/makefiles/saproc.make	Wed Jul 05 17:05:01 2017 +0200
@@ -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 @@
 	           $(SA_LFLAGS)                                         \
 	           -o $@                                                \
 	           -ldl -ldemangle -lthread -lc
+	[ -f $(LIBSAPROC_G) ] || { ln -s $@ $(LIBSAPROC_G); }
 
 install_saproc: checkAndBuildSA
 	$(QUIETLY) if [ -f $(LIBSAPROC) ] ; then             \
--- a/hotspot/make/solaris/makefiles/sparcWorks.make	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/make/solaris/makefiles/sparcWorks.make	Wed Jul 05 17:05:01 2017 +0200
@@ -281,8 +281,6 @@
 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 @@
 # [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/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
--- a/hotspot/make/solaris/makefiles/vm.make	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/make/solaris/makefiles/vm.make	Wed Jul 05 17:05:01 2017 +0200
@@ -108,11 +108,16 @@
 #   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 @@
 #----------------------------------------------------------------------
 # JVM
 
-JVM    = jvm$(G_SUFFIX)
-LIBJVM = lib$(JVM).so
+JVM      = jvm
+LIBJVM   = lib$(JVM).so
+LIBJVM_G = lib$(JVM)$(G_SUFFIX).so
 
 JVM_OBJ_FILES = $(Obj_Files) $(DTRACE_OBJS)
 
@@ -173,11 +179,12 @@
 	-sbfast|-xsbfast) \
 	    ;; \
 	*) \
-	    echo Linking vm...;                                                  \
-	    $(LINK_LIB.CC/PRE_HOOK)                                              \
-	    $(LINK_VM) $(LFLAGS_VM) -o $@ $(LIBJVM.o) $(LIBS_VM);                \
-	    $(LINK_LIB.CC/POST_HOOK)                                             \
-	    rm -f $@.1; ln -s $@ $@.1;                                           \
+	    echo Linking vm...; \
+	    $(LINK_LIB.CC/PRE_HOOK) \
+	    $(LINK_VM) $(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; }; \
 	    ;; \
 	esac
 
--- a/hotspot/src/cpu/sparc/vm/c1_LIRAssembler_sparc.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/sparc/vm/c1_LIRAssembler_sparc.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -189,14 +189,17 @@
   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 @@
       }
 #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 @@
         } 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 @@
           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,                reg);
+            __ ldf(FloatRegisterImpl::S, base, offset + 4, reg->successor());
+            __ ldf(FloatRegisterImpl::S, base, offset,     reg);
           } else {
             __ ldf(FloatRegisterImpl::D, base, offset, to_reg->as_double_reg());
           }
@@ -2200,6 +2205,7 @@
   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 @@
   }
 
   __ 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 {
--- a/hotspot/src/cpu/sparc/vm/c1_LIRGenerator_sparc.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/sparc/vm/c1_LIRGenerator_sparc.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -144,17 +144,17 @@
   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 @@
     }
   } 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);
-    __ move(LIR_OprFact::intConst(disp), index);
+    index = new_pointer_register();
+    __ move(LIR_OprFact::intptrConst(disp), index);
     disp = 0;
   }
 
--- a/hotspot/src/cpu/sparc/vm/c1_globals_sparc.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/sparc/vm/c1_globals_sparc.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -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(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, MaxPermSize,                 64*M );
-define_pd_global(bool, NeverActAsServerClassMachine, true);
+define_pd_global(uintx,PermSize,                     12*M );
+define_pd_global(uintx,MaxPermSize,                  64*M );
+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    );
--- a/hotspot/src/cpu/sparc/vm/c2_globals_sparc.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/sparc/vm/c2_globals_sparc.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -59,7 +59,6 @@
 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 @@
 // 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, ReservedCodeCacheSize,    48*M);
-define_pd_global(intx, CodeCacheExpansionSize,   64*K);
+define_pd_global(intx, InitialCodeCacheSize,         2048*K); // Integral multiple of CodeCacheExpansionSize
+define_pd_global(intx, ReservedCodeCacheSize,        48*M);
+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, ReservedCodeCacheSize,    32*M);
-define_pd_global(intx, CodeCacheExpansionSize,   32*K);
+define_pd_global(intx, InitialCodeCacheSize,         1536*K); // Integral multiple of CodeCacheExpansionSize
+define_pd_global(intx, ReservedCodeCacheSize,        32*M);
+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, MaxPermSize, ScaleForWordSize(64*M));
+define_pd_global(uintx,PermSize,    ScaleForWordSize(16*M));
+define_pd_global(uintx,MaxPermSize, ScaleForWordSize(64*M));
 
 // Ergonomics related flags
 define_pd_global(bool, NeverActAsServerClassMachine, false);
--- a/hotspot/src/cpu/sparc/vm/globals_sparc.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/sparc/vm/globals_sparc.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -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, 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(uintx, TLABSize,              0);
-define_pd_global(uintx, NewSize, ScaleForWordSize((2048 * K) + (2 * (64 * K))));
-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);
+define_pd_global(intx, CodeEntryAlignment,    32);
+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,  VMThreadStackSize,     1024);
+define_pd_global(intx, ThreadStackSize,       1024);
+define_pd_global(intx, VMThreadStackSize,     1024);
 #else
-define_pd_global(intx,  ThreadStackSize,       512);
-define_pd_global(intx,  VMThreadStackSize,     512);
+define_pd_global(intx, ThreadStackSize,       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);
--- a/hotspot/src/cpu/sparc/vm/templateInterpreter_sparc.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/sparc/vm/templateInterpreter_sparc.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -150,8 +150,7 @@
 }
 
 
-address TemplateInterpreterGenerator::generate_return_entry_for(TosState state, int step, bool unbox) {
-  assert(!unbox, "NYI");//6815692//
+address TemplateInterpreterGenerator::generate_return_entry_for(TosState state, int step) {
   address compiled_entry = __ pc();
   Label cont;
 
--- a/hotspot/src/cpu/x86/vm/assembler_x86.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/x86/vm/assembler_x86.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -2251,6 +2251,7 @@
   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 @@
   emit_byte(0x8F);
   emit_operand(rax, dst);
 }
+#endif
 
 void Assembler::prefetch_prefix(Address src) {
   prefix(src);
@@ -2428,6 +2430,7 @@
   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 @@
   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::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::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 @@
 
 #ifdef ASSERT
   Label L;
-  testl(tmp, tmp);
+  testptr(tmp, tmp);
   jccb(Assembler::notZero, L);
   hlt();
   bind(L);
--- a/hotspot/src/cpu/x86/vm/assembler_x86.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/x86/vm/assembler_x86.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -1244,7 +1244,9 @@
   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 @@
   // 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);
 
--- a/hotspot/src/cpu/x86/vm/c1_LIRAssembler_x86.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/x86/vm/c1_LIRAssembler_x86.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -301,22 +301,25 @@
   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 @@
           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 @@
       __ 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 @@
 
 #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);
 
--- a/hotspot/src/cpu/x86/vm/c1_LIRGenerator_x86.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/x86/vm/c1_LIRGenerator_x86.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -755,8 +755,19 @@
   }
 
   LIR_Opr addr = new_pointer_register();
-  __ move(obj.result(), addr);
-  __ add(addr, offset.result(), addr);
+  LIR_Address* a;
+  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.
--- a/hotspot/src/cpu/x86/vm/c1_globals_x86.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/x86/vm/c1_globals_x86.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -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, 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, MaxPermSize,                 64*M );
-define_pd_global(bool, NeverActAsServerClassMachine, true);
-define_pd_global(uintx, DefaultMaxRAM,               1*G);
+define_pd_global(uintx,PermSize,                     12*M );
+define_pd_global(uintx,MaxPermSize,                  64*M );
+define_pd_global(bool, NeverActAsServerClassMachine, true );
+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, CSEArrayLength,               false);
+define_pd_global(bool, TwoOperandLIRForm,            true );
 
-define_pd_global(bool, LIRFillDelaySlots,            false);
-define_pd_global(bool, OptimizeSinglePrecision,      true);
-define_pd_global(bool, CSEArrayLength,               false);
-define_pd_global(bool, TwoOperandLIRForm,            true);
-
-
-define_pd_global(intx, SafepointPollOffset, 256);
+define_pd_global(intx, SafepointPollOffset,          256  );
--- a/hotspot/src/cpu/x86/vm/c2_globals_x86.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/x86/vm/c2_globals_x86.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -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,             10000);
 #endif // TIERED
 define_pd_global(intx, Tier2CompileThreshold,        10000);
-define_pd_global(intx, Tier3CompileThreshold,        20000 );
-define_pd_global(intx, Tier4CompileThreshold,        40000 );
+define_pd_global(intx, Tier3CompileThreshold,        20000);
+define_pd_global(intx, Tier4CompileThreshold,        40000);
 
 define_pd_global(intx, BackEdgeThreshold,            100000);
 define_pd_global(intx, Tier2BackEdgeThreshold,       100000);
@@ -61,7 +60,6 @@
 #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, 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(uintx,CodeCacheMinBlockLength,      4);
 
 // Heap related flags
-define_pd_global(uintx, PermSize,    ScaleForWordSize(16*M));
-define_pd_global(uintx, MaxPermSize, ScaleForWordSize(64*M));
+define_pd_global(uintx,PermSize,    ScaleForWordSize(16*M));
+define_pd_global(uintx,MaxPermSize, ScaleForWordSize(64*M));
 
 // Ergonomics related flags
 define_pd_global(bool, NeverActAsServerClassMachine, false);
--- a/hotspot/src/cpu/x86/vm/globals_x86.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/x86/vm/globals_x86.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -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,  ShareVtableStubs,         true);
-define_pd_global(bool,  CountInterpCalls,         true);
+define_pd_global(bool, ConvertSleepToYield,      true);
+define_pd_global(bool, ShareVtableStubs,         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, UncommonNullCast,            true);  // Uncommon-trap NULLs past to check cast
+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
 
 // 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 @@
 // 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(uintx, TLABSize,                 0);
+define_pd_global(intx, StackYellowPages, 2);
+define_pd_global(intx, StackRedPages, 1);
 #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, StackRedPages, 1);
+define_pd_global(intx, PreInflateSpin,           10);
 
 define_pd_global(bool, RewriteBytecodes,     true);
 define_pd_global(bool, RewriteFrequentPairs, true);
--- a/hotspot/src/cpu/x86/vm/interp_masm_x86_32.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/x86/vm/interp_masm_x86_32.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -196,6 +196,9 @@
   } 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
   }
--- a/hotspot/src/cpu/x86/vm/interp_masm_x86_64.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/x86/vm/interp_masm_x86_64.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -185,12 +185,30 @@
 }
 
 
+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) {
-  assert(bcp_offset > 0, "bcp is still pointing to start of bytecode");
+                                                           int bcp_offset,
+                                                           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_entry_pointer_at_bcp(Register cache,
                                                                Register tmp,
-                                                               int bcp_offset) {
-  assert(bcp_offset > 0, "bcp is still pointing to start of bytecode");
+                                                               int bcp_offset,
+                                                               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_virtual_call(Register receiver,
                                                      Register mdp,
-                                                     Register reg2) {
+                                                     Register reg2,
+                                                     bool receiver_can_be_null) {
   if (ProfileInterpreter) {
     Label profile_continue;
 
@@ -1246,8 +1265,15 @@
     // 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,
--- a/hotspot/src/cpu/x86/vm/interp_masm_x86_64.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/x86/vm/interp_masm_x86_64.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -95,9 +95,10 @@
 
   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 @@
   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);
--- a/hotspot/src/cpu/x86/vm/interpreter_x86_64.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/x86/vm/interpreter_x86_64.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -277,12 +277,11 @@
   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
-  __ mov(rsp, r13);
+  //  pop return address, reset last_sp to NULL
+  __ 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 @@
   if (!EnableMethodHandles) {
     return generate_abstract_entry();
   }
-  return generate_abstract_entry(); //6815692//
+
+  address entry_point = MethodHandles::generate_method_handle_interpreter_entry(_masm);
+
+  return entry_point;
 }
 
 
--- a/hotspot/src/cpu/x86/vm/methodHandles_x86.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/x86/vm/methodHandles_x86.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -448,7 +448,7 @@
                                 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);
--- a/hotspot/src/cpu/x86/vm/stubGenerator_x86_64.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/x86/vm/stubGenerator_x86_64.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -2935,6 +2935,16 @@
 
     // 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:
--- a/hotspot/src/cpu/x86/vm/templateInterpreter_x86_32.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/x86/vm/templateInterpreter_x86_32.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -155,15 +155,8 @@
 }
 
 
-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 @@
   __ 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 @@
   }
   __ 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 @@
     __ 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;
--- a/hotspot/src/cpu/x86/vm/templateInterpreter_x86_64.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/x86/vm/templateInterpreter_x86_64.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -100,21 +100,26 @@
   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
-  // exception happened
+  __ verify_oop(c_rarg1);
+  __ 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::
-                              throw_WrongMethodTypeException),
+                              InterpreterRuntime::throw_WrongMethodTypeException),
              // pass required type, failing object (or NULL)
              c_rarg1, c_rarg2);
   return entry;
@@ -166,8 +171,7 @@
 
 
 address TemplateInterpreterGenerator::generate_return_entry_for(TosState state,
-                                                                int step, bool unbox) {
-  assert(!unbox, "NYI");//6815692//
+                                                                int step) {
 
   // 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 @@
   __ 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;
 }
 
--- a/hotspot/src/cpu/x86/vm/templateTable_x86_32.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/x86/vm/templateTable_x86_32.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -2890,9 +2890,6 @@
 
 
 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 @@
   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 @@
   // load return address
   {
     address table_addr;
-    if (is_invdyn_bootstrap)
-      table_addr = (address)Interpreter::return_5_unbox_addrs_by_index_table();
-    else if (is_invokeinterface || is_invokedynamic)
+    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 @@
     __ profile_call(rsi);
   }
 
-  Label handle_unlinked_site;
-  __ movptr(rcx, Address(rax, __ delayed_value(sun_dyn_CallSiteImpl::target_offset_in_bytes, rcx)));
-  __ testptr(rcx, rcx);
-  __ jcc(Assembler::zero, handle_unlinked_site);
-
+  __ 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);
-
-  // 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);
 }
 
 //----------------------------------------------------------------------------------------------------
--- a/hotspot/src/cpu/x86/vm/templateTable_x86_64.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/x86/vm/templateTable_x86_64.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -203,18 +203,15 @@
     __ jcc(Assembler::notEqual, fast_patch);
     __ get_method(scratch);
     // Let breakpoint table handling rewrite to quicker bytecode
-    __ call_VM(noreg,
-               CAST_FROM_FN_PTR(address,
-                                InterpreterRuntime::set_original_bytecode_at),
-               scratch, r13, bc);
+    __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::set_original_bytecode_at), scratch, r13, bc);
 #ifndef ASSERT
     __ jmpb(patch_done);
+#else
+    __ jmp(patch_done);
+#endif
     __ bind(fast_patch);
   }
-#else
-    __ jmp(patch_done);
-    __ 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::resolve_cache_and_index(int byte_no,
-                                            Register Rcache,
-                                            Register index) {
+void TemplateTable::resolve_cache_and_index(int byte_no, 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);
-  __ movl(temp, Address(Rcache,
-                        index, Address::times_8,
-                        constantPoolCacheOopDesc::base_offset() +
-                        ConstantPoolCacheEntry::indices_offset()));
-  __ shrl(temp, shift_count);
-  // have we resolved this bytecode?
-  __ andl(temp, 0xFF);
-  __ cmpl(temp, (int) bytecode());
-  __ jcc(Assembler::equal, resolved);
+  __ get_cache_and_index_at_bcp(Rcache, index, 1, is_invokedynamic);
+  if (is_invokedynamic) {
+    // we are resolved if the f1 field contains a non-null CallSite object
+    __ cmpptr(Address(Rcache, index, Address::times_ptr, constantPoolCacheOopDesc::base_offset() + ConstantPoolCacheEntry::f1_offset()), (int32_t) NULL_WORD);
+    __ jcc(Assembler::notEqual, resolved);
+  } else {
+    __ movl(temp, Address(Rcache, index, Address::times_ptr, constantPoolCacheOopDesc::base_offset() + ConstantPoolCacheEntry::indices_offset()));
+    __ shrl(temp, shift_count);
+    // have we resolved this bytecode?
+    __ andl(temp, 0xFF);
+    __ cmpl(temp, (int) bytecode());
+    __ jcc(Assembler::equal, resolved);
+  }
 
   // resolve first time through
   address entry;
@@ -2090,6 +2089,9 @@
   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 @@
   __ 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 @@
   ShouldNotReachHere();
 }
 
-void TemplateTable::prepare_invoke(Register method,
-                                   Register index,
-                                   int byte_no,
-                                   Bytecodes::Code code) {
+void TemplateTable::prepare_invoke(Register method, Register index, int byte_no) {
   // 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 @@
     __ movl(recv, flags);
     __ andl(recv, 0xFF);
     if (TaggedStackInterpreter) __ shll(recv, 1);  // index*2
-    __ movptr(recv, Address(rsp, recv, Address::times_8,
-                                 -Interpreter::expr_offset_in_bytes(1)));
-    __ verify_oop(recv);
+    Address recv_addr(rsp, recv, Address::times_8, -Interpreter::expr_offset_in_bytes(1));
+    if (is_invokedynamic) {
+      __ lea(recv, recv_addr);
+    } else {
+      __ movptr(recv, recv_addr);
+      __ verify_oop(recv);
+    }
   }
 
   // do null check if needed
@@ -2878,10 +2883,14 @@
   ConstantPoolCacheEntry::verify_tosBits();
   // load return address
   {
-    ExternalAddress return_5((address)Interpreter::return_5_addrs_by_index_table());
-    ExternalAddress return_3((address)Interpreter::return_3_addrs_by_index_table());
-    __ lea(rscratch1, (is_invokeinterface ? return_5 : return_3));
-    __ movptr(flags, Address(rscratch1, flags, Address::times_8));
+    address table_addr;
+    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();
+    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(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::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::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::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 @@
     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);
 }
 
 
--- a/hotspot/src/cpu/x86/vm/templateTable_x86_64.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/x86/vm/templateTable_x86_64.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -22,8 +22,7 @@
  *
  */
 
-  static void prepare_invoke(Register method, Register index, int byte_no,
-                             Bytecodes::Code code);
+  static void prepare_invoke(Register method, Register index, int byte_no);
   static void invokevirtual_helper(Register index, Register recv,
                                    Register flags);
   static void volatile_barrier(Assembler::Membar_mask_bits order_constraint);
--- a/hotspot/src/cpu/x86/vm/vm_version_x86.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/x86/vm/vm_version_x86.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -255,6 +255,8 @@
   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
--- a/hotspot/src/cpu/zero/vm/cppInterpreter_zero.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/zero/vm/cppInterpreter_zero.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -204,6 +204,20 @@
     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;
--- a/hotspot/src/cpu/zero/vm/frame_zero.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/zero/vm/frame_zero.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -36,11 +36,8 @@
   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 @@
   return frame(sender_sp(), sp() + 1);
 }
 
-frame frame::sender_for_interpreter_frame(RegisterMap *map) const {
-  return frame(sender_sp(), sp() + 1);
-}
-
-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 {
+frame frame::sender_for_nonentry_frame(RegisterMap *map) const {
+  assert(zeroframe()->is_interpreter_frame() ||
+         zeroframe()->is_shark_frame() ||
+         zeroframe()->is_fake_stub_frame(), "wrong type of frame");
   return frame(sender_sp(), sp() + 1);
 }
 
@@ -69,17 +61,8 @@
 
   if (is_entry_frame())
     return sender_for_entry_frame(map);
-
-  if (is_interpreted_frame())
-    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();
+  else
+    return sender_for_nonentry_frame(map);
 }
 
 #ifdef CC_INTERP
--- a/hotspot/src/cpu/zero/vm/frame_zero.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/zero/vm/frame_zero.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -65,10 +65,7 @@
   }
 
  public:
-  bool is_fake_stub_frame() const;
-
- public:
-  frame sender_for_fake_stub_frame(RegisterMap* map) const;
+  frame sender_for_nonentry_frame(RegisterMap* map) const;
 
  public:
   void zero_print_on_error(int           index,
--- a/hotspot/src/cpu/zero/vm/globals_zero.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/zero/vm/globals_zero.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -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,  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);
--- a/hotspot/src/cpu/zero/vm/sharedRuntime_zero.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/zero/vm/sharedRuntime_zero.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -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 @@
                                                 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,
--- a/hotspot/src/cpu/zero/vm/sharkFrame_zero.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/cpu/zero/vm/sharkFrame_zero.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -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() {
--- a/hotspot/src/os/linux/vm/os_linux.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/os/linux/vm/os_linux.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -223,8 +223,8 @@
                      "environment on Linux when /proc filesystem is not mounted.";
 
 void os::Linux::initialize_system_info() {
-  _processor_count = sysconf(_SC_NPROCESSORS_CONF);
-  if (_processor_count == 1) {
+  set_processor_count(sysconf(_SC_NPROCESSORS_CONF));
+  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 @@
     }
   }
   _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 @@
   // Return immediately if a permit is available.
   if (_counter > 0) {
       _counter = 0 ;
+      OrderAccess::fence();
       return ;
   }
 
@@ -4725,6 +4726,7 @@
     _counter = 0;
     status = pthread_mutex_unlock(_mutex);
     assert (status == 0, "invariant") ;
+    OrderAccess::fence();
     return;
   }
 
@@ -4765,6 +4767,7 @@
     jt->java_suspend_self();
   }
 
+  OrderAccess::fence();
 }
 
 void Parker::unpark() {
--- a/hotspot/src/os/solaris/dtrace/libjvm_db.c	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/os/solaris/dtrace/libjvm_db.c	Wed Jul 05 17:05:01 2017 +0200
@@ -937,54 +937,56 @@
   return err;
 }
 
-static int
-scopeDesc_chain(Nmethod_t *N)
-{
+static int scopeDesc_chain(Nmethod_t *N) {
   int32_t decode_offset = 0;
   int32_t err;
 
-  if (debug > 2)
-      fprintf(stderr, "\t scopeDesc_chain: BEGIN\n");
+  if (debug > 2) {
+    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)
-          fprintf(stderr, "\t scopeDesc_chain: decode_offset: %#x\n", decode_offset);
+    Vframe_t *vf = &N->vframes[N->vf_cnt];
 
-      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 (vf->methodIdx > N->oops_len) {
-          fprintf(stderr, "\t scopeDesc_chain: (methodIdx > oops_len) !\n");
-          return -1;
+      if (debug > 2) {
+        fprintf(stderr, "\t scopeDesc_chain: methodOop: %#8llx, line: %ld\n",
+                vf->methodOop, vf->line);
       }
-      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) {
-              fprintf(stderr, "\t scopeDesc_chain: methodOop: %#8llx, line: %ld\n",
-                              vf->methodOop, vf->line);
-          }
-      }
-      decode_offset = vf->sender_decode_offset;
+    }
+    decode_offset = vf->sender_decode_offset;
   }
-  if (debug > 2)
-      fprintf(stderr, "\t scopeDesc_chain: END \n\n");
+  if (debug > 2) {
+    fprintf(stderr, "\t scopeDesc_chain: END \n\n");
+  }
   return PS_OK;
 
  fail:
-  if (debug)
-      fprintf(stderr, "\t scopeDesc_chain: FAIL \n\n");
+  if (debug) {
+    fprintf(stderr, "\t scopeDesc_chain: FAIL \n\n");
+  }
   return err;
 }
 
--- a/hotspot/src/os/solaris/vm/os_solaris.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/os/solaris/vm/os_solaris.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -457,7 +457,7 @@
 
 
 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 @@
   // Return immediately if a permit is available.
   if (_counter > 0) {
       _counter = 0 ;
+      OrderAccess::fence();
       return ;
   }
 
@@ -5846,6 +5847,7 @@
     _counter = 0;
     status = os::Solaris::mutex_unlock(_mutex);
     assert (status == 0, "invariant") ;
+    OrderAccess::fence();
     return;
   }
 
@@ -5892,6 +5894,7 @@
     jt->java_suspend_self();
   }
 
+  OrderAccess::fence();
 }
 
 void Parker::unpark() {
--- a/hotspot/src/os/windows/vm/os_windows.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/os/windows/vm/os_windows.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -3150,7 +3150,7 @@
   _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);
--- a/hotspot/src/os_cpu/linux_x86/vm/globals_linux_x86.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/os_cpu/linux_x86/vm/globals_linux_x86.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -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 @@
 #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);
--- a/hotspot/src/os_cpu/solaris_x86/vm/globals_solaris_x86.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/os_cpu/solaris_x86/vm/globals_solaris_x86.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -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);
--- a/hotspot/src/os_cpu/windows_x86/vm/globals_windows_x86.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/os_cpu/windows_x86/vm/globals_windows_x86.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -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(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
--- a/hotspot/src/share/vm/c1/c1_GraphBuilder.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/c1/c1_GraphBuilder.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -365,7 +365,7 @@
     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");
   }
 }
 
--- a/hotspot/src/share/vm/c1/c1_LIRGenerator.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/c1/c1_LIRGenerator.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -1855,12 +1855,26 @@
     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 @@
   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);
-    __ move(idx.result(), index_op);
+    index_op = new_pointer_register();
+#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);
--- a/hotspot/src/share/vm/c1/c1_LinearScan.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/c1/c1_LinearScan.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -2464,6 +2464,10 @@
 
     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 @@
         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 @@
   } 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;
--- a/hotspot/src/share/vm/ci/bcEscapeAnalyzer.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/ci/bcEscapeAnalyzer.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -61,9 +61,11 @@
   BCEscapeAnalyzer* _parent;
   int               _level;
 
+ public:
   class  ArgumentMap;
   class  StateInfo;
 
+ private:
   // helper functions
   bool is_argument(int i)    { return i >= 0 && i < _arg_size; }
 
--- a/hotspot/src/share/vm/ci/ciEnv.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/ci/ciEnv.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -46,6 +46,9 @@
 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 @@
   _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 @@
   _ArrayIndexOutOfBoundsException_instance = NULL;
   _ArrayStoreException_instance = NULL;
   _ClassCastException_instance = NULL;
+  _the_null_string = NULL;
+  _the_min_jint_string = NULL;
 }
 
 ciEnv::~ciEnv() {
@@ -248,6 +255,22 @@
   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 @@
   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();
--- a/hotspot/src/share/vm/ci/ciEnv.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/ci/ciEnv.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -82,6 +82,9 @@
   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 @@
   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 @@
   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 @@
   ciInstance* ArrayStoreException_instance();
   ciInstance* ClassCastException_instance();
 
+  ciInstance* the_null_string();
+  ciInstance* the_min_jint_string();
+
   static ciSymbol* unloaded_cisymbol() {
     return _unloaded_cisymbol;
   }
--- a/hotspot/src/share/vm/ci/ciInstanceKlass.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/ci/ciInstanceKlass.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -341,6 +341,20 @@
 }
 
 // ------------------------------------------------------------------
+// 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.
 
 class NonStaticFieldFiller: public FieldClosure {
--- a/hotspot/src/share/vm/ci/ciInstanceKlass.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/ci/ciInstanceKlass.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -148,6 +148,7 @@
 
   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();
 
--- a/hotspot/src/share/vm/ci/ciObjectFactory.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/ci/ciObjectFactory.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -168,6 +168,15 @@
   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();
--- a/hotspot/src/share/vm/classfile/classFileParser.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/classfile/classFileParser.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -2511,23 +2511,12 @@
       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;
-      } else {
-        fac_ptr->nonstatic_double_count += 1;
-      }
-
-      FieldAllocationType atype = (FieldAllocationType) (*fields_ptr)->ushort_at(i+4);
+      fac_ptr->nonstatic_word_count += 1;
+
+      FieldAllocationType atype = (FieldAllocationType) (*fields_ptr)->ushort_at(i + instanceKlass::low_offset);
       assert(atype == NONSTATIC_BYTE, "");
       FieldAllocationType new_atype = NONSTATIC_WORD;
-      if (wordSize > jintSize) {
-        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);
+      (*fields_ptr)->ushort_at_put(i + instanceKlass::low_offset, new_atype);
 
       found_vmentry = true;
       break;
@@ -3085,7 +3074,7 @@
     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 @@
         default:
           ShouldNotReachHere();
       }
-      fields->short_at_put(i+4, 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::low_offset,  extract_low_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::is_supported_version(u2 major, u2 minor) {
-  u2 max_version = JDK_Version::is_gte_jdk17x_version() ?
-    JAVA_MAX_SUPPORTED_VERSION : JAVA_6_VERSION;
+  u2 max_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) ||
--- a/hotspot/src/share/vm/classfile/javaClasses.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/classfile/javaClasses.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -1124,8 +1124,7 @@
     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(),
-                                    _methods->array_size()));
+      bs->write_ref_array((HeapWord*)_methods->base(), _methods->length());
       _dirty = false;
     }
   }
@@ -2430,15 +2429,15 @@
 }
 
 
-// Support for sun_dyn_CallSiteImpl
-
-int sun_dyn_CallSiteImpl::_type_offset;
-int sun_dyn_CallSiteImpl::_target_offset;
-int sun_dyn_CallSiteImpl::_vmmethod_offset;
-
-void sun_dyn_CallSiteImpl::compute_offsets() {
+// Support for java_dyn_CallSite
+
+int java_dyn_CallSite::_type_offset;
+int java_dyn_CallSite::_target_offset;
+int java_dyn_CallSite::_vmmethod_offset;
+
+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 @@
   }
 }
 
-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 @@
     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
--- a/hotspot/src/share/vm/classfile/javaClasses.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/classfile/javaClasses.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -1061,9 +1061,9 @@
 };
 
 
-// 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:
--- a/hotspot/src/share/vm/classfile/systemDictionary.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/classfile/systemDictionary.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -99,6 +99,15 @@
   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 @@
       // 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 @@
 }
 
 // Support parallel classloading
-// Initial implementation for bootstrap classloader
-// For custom class loaders that support parallel classloading,
+// All parallel class loaders, including bootstrap classloader
+// 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 @@
   {
     MutexLocker mu(SystemDictionary_lock, THREAD);
     // First check if class already defined
-    klassOop check = find_class(d_index, d_hash, name_h, class_loader);
-    if (check != NULL) {
-      return(instanceKlassHandle(THREAD, check));
+    if (UnsyncloadClass || (is_parallelDefine(class_loader))) {
+      klassOop check = find_class(d_index, d_hash, name_h, class_loader);
+      if (check != NULL) {
+        return(instanceKlassHandle(THREAD, check));
+      }
     }
 
     // Acquire define token for this class/classloader
@@ -1529,7 +1543,7 @@
     // 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 @@
   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 @@
                                                 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 @@
   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");
-  sun_dyn_CallSiteImpl::set_vmmethod(call_site_oop, mh_invdyn());
+         /*&& java_dyn_CallSite::is_instance(call_site_oop)*/, "must be sane");
+  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 @@
 
   instanceKlassHandle ik(THREAD, caller());
 
-  if (ik->bootstrap_method() != NULL) {
-    return Handle(THREAD, ik->bootstrap_method());
+  oop boot_method_oop = 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 @@
                          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 @@
     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();
   }
 
--- a/hotspot/src/share/vm/classfile/systemDictionary.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/classfile/systemDictionary.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -144,13 +144,13 @@
   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 @@
   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);
 
--- a/hotspot/src/share/vm/classfile/verifier.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/classfile/verifier.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -1903,17 +1903,8 @@
   verify_cp_type(index, cp, types, CHECK_VERIFY(this));
 
   // Get method name and signature
-  symbolHandle method_name;
-  symbolHandle method_sig;
-  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));
-  }
+  symbolHandle method_name(THREAD, cp->name_ref_at(index));
+  symbolHandle method_sig(THREAD, cp->signature_ref_at(index));
 
   if (!SignatureVerifier::is_valid_method_signature(method_sig)) {
     class_format_error(
--- a/hotspot/src/share/vm/classfile/vmSymbols.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/classfile/vmSymbols.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -303,6 +303,11 @@
   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* 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;
--- a/hotspot/src/share/vm/classfile/vmSymbols.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/classfile/vmSymbols.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -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)   \
-  /*    (symbol object_initializer_name defined above) */                                                               \
-                                                                                                                        \
+                                                                                                                          \
+  do_intrinsic(_StringBuilder_void,   java_lang_StringBuilder, object_initializer_name, void_method_signature,     F_R)   \
+  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 @@
   enum Flags {
     // AccessFlags syndromes relevant to intrinsics.
     F_none = 0,
-    F_R,                        // !static        !synchronized (R="regular")
-    F_S,                        //  static        !synchronized
-    F_RN,                       // !static native !synchronized
-    F_SN,                       //  static native !synchronized
-    F_RNY                       // !static native  synchronized
+    F_R,                        // !static ?native !synchronized (R="regular")
+    F_S,                        //  static ?native !synchronized
+    F_Y,                        // !static ?native  synchronized
+    F_RN,                       // !static  native !synchronized
+    F_SN,                       //  static  native !synchronized
+    F_RNY                       // !static  native  synchronized
   };
 
 public:
--- a/hotspot/src/share/vm/code/dependencies.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/code/dependencies.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -1528,19 +1528,23 @@
   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)
-           ++nsup;
-      else tty->print_cr("  context super = %s", instanceKlass::cast(k)->external_name());
+      if (!WizardMode) {
+        ++nsup;
+      } else {
+        tty->print_cr("  context super = %s", instanceKlass::cast(k)->external_name());
+      }
       break;
     case Change_new_impl:
-      if (!WizardMode)
-           ++nint;
-      else tty->print_cr("  context interface = %s", instanceKlass::cast(k)->external_name());
+      if (!WizardMode) {
+        ++nint;
+      } else {
+        tty->print_cr("  context interface = %s", instanceKlass::cast(k)->external_name());
+      }
       break;
     }
   }
--- a/hotspot/src/share/vm/code/dependencies.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/code/dependencies.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -470,7 +470,7 @@
 // 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 @@
     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 ContextStream : public StackObj {
    private:
-    DepChange&       _changes;
+    DepChange&  _changes;
     friend class DepChange;
 
     // iteration variables:
-    ChangeType            _change_type;
-    klassOop              _klass;
-    objArrayOop           _ti_base;    // i.e., transitive_interfaces
-    int                   _ti_index;
-    int                   _ti_limit;
+    ChangeType  _change_type;
+    klassOop    _klass;
+    objArrayOop _ti_base;    // i.e., transitive_interfaces
+    int         _ti_index;
+    int         _ti_limit;
 
     // start at the beginning:
     void start() {
@@ -530,11 +531,11 @@
       _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 @@
 
     bool next();
 
+    ChangeType change_type()     { return _change_type; }
     klassOop   klass()           { return _klass; }
   };
   friend class DepChange::ContextStream;
--- a/hotspot/src/share/vm/code/nmethod.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/code/nmethod.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -414,9 +414,8 @@
 }
 
 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 @@
   }
   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::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 @@
   _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 "'",
-                       (state == zombie ? "zombie='1' " : ""),
-                       os::current_thread_id());
+      if (flags.state == unloaded) {
+        xtty->begin_elem("make_unloaded thread='" UINTX_FORMAT "'",
+                         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) {
-    print_on(tty, state == zombie ? "made zombie " : "made not entrant ");
+  if (PrintCompilation && flags.state != unloaded) {
+    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.
-  // They never become zombie/non-entrant, so the nmethod sweeper will never remove
-  // them. Instead the entry_bci is set to InvalidOSREntryBci, so the osr nmethod
-  // 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 there is nothing to do
+  if (is_zombie()) {
+    return false;
   }
 
-  // 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 @@
 
     // 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 @@
   // 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 @@
 
   // 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 @@
     HandleMark hm;
     method()->clear_code();
   }
+
+  return true;
 }
 
 
--- a/hotspot/src/share/vm/code/nmethod.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/code/nmethod.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -252,7 +252,9 @@
   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 @@
   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.
-  // It is used when an uncommon trap happens.
-  void  make_not_entrant()                        { make_not_entrant_or_zombie(not_entrant); }
-  void  make_zombie()                             { make_not_entrant_or_zombie(zombie); }
+  // Make the nmethod non entrant. The nmethod will continue to be
+  // alive.  It is used when an uncommon trap happens.  Returns true
+  // if this thread changed the state of the nmethod or false if
+  // 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 @@
   // 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);
--- a/hotspot/src/share/vm/gc_implementation/concurrentMarkSweep/concurrentMarkSweepGeneration.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/gc_implementation/concurrentMarkSweep/concurrentMarkSweepGeneration.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -709,7 +709,8 @@
 
   // 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);
--- a/hotspot/src/share/vm/gc_implementation/g1/collectionSetChooser.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/collectionSetChooser.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -351,9 +351,16 @@
 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");
 }
 
--- a/hotspot/src/share/vm/gc_implementation/g1/concurrentG1Refine.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/concurrentG1Refine.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -42,28 +42,49 @@
   _n_periods(0),
   _threads(NULL), _n_threads(0)
 {
-  if (G1ConcRefine) {
-    _n_threads = (int)thread_num();
-    if (_n_threads > 0) {
-      _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;
-      }
-    }
+
+  // Ergomonically select initial concurrent refinement parameters
+  if (FLAG_IS_DEFAULT(G1ConcRefineGreenZone)) {
+    FLAG_SET_DEFAULT(G1ConcRefineGreenZone, MAX2<int>(ParallelGCThreads, 1));
+  }
+  set_green_zone(G1ConcRefineGreenZone);
+
+  if (FLAG_IS_DEFAULT(G1ConcRefineYellowZone)) {
+    FLAG_SET_DEFAULT(G1ConcRefineYellowZone, green_zone() * 3);
+  }
+  set_yellow_zone(MAX2<int>(G1ConcRefineYellowZone, green_zone()));
+
+  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() {
-  if (G1ConcRefine) {
-    return (G1ParallelRSetThreads > 0) ? G1ParallelRSetThreads : ParallelGCThreads;
+void ConcurrentG1Refine::reset_threshold_step() {
+  if (FLAG_IS_DEFAULT(G1ConcRefineThresholdStep)) {
+    _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::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 @@
     st->cr();
   }
 }
-
--- a/hotspot/src/share/vm/gc_implementation/g1/concurrentG1Refine.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/concurrentG1Refine.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -29,6 +29,31 @@
 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 @@
   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 @@
 
   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; }
 };
--- a/hotspot/src/share/vm/gc_implementation/g1/concurrentG1RefineThread.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/concurrentG1RefineThread.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -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 @@
   _worker_id(worker_id),
   _active(false),
   _next(next),
+  _monitor(NULL),
   _cg1r(cg1r),
-  _vtime_accum(0.0),
-  _interval_ms(5.0)
+  _vtime_accum(0.0)
 {
+
+  // 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::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.
-    {
-      MutexLockerEx x(DirtyCardQ_CBL_mon, Mutex::_no_safepoint_check_flag);
-      while (((_worker_id == 0 && !dcqs.process_completed_buffers()) ||
-              (_worker_id > 0 && !is_active())) &&
-             !_should_terminate) {
-         DirtyCardQ_CBL_mon->wait(Mutex::_no_safepoint_check_flag);
-      }
+
+    // Wait for work
+    wait_for_completed_buffers();
+
+    if (_should_terminate) {
+      break;
     }
 
-    if (_should_terminate) {
-      return;
-    }
+    _sts.join();
 
-    // 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);
+    do {
+      int curr_buffer_num = (int)dcqs.completed_buffers_num();
+      // If the number of the buffers falls down into the yellow zone,
+      // that means that the transition period after the evacuation pause has ended.
+      if (dcqs.completed_queue_padding() > 0 && curr_buffer_num <= cg1r()->yellow_zone()) {
+        dcqs.set_completed_queue_padding(0);
+      }
 
-    if (G1SmoothConcRefine) {
-      lower_limit = 0;
-      start_vtime_sec = os::elapsedVTime();
-      prev_buffer_num = (int) dcqs.completed_buffers_num();
-    } else {
-      lower_limit = G1UpdateBufferQueueProcessingThreshold / 4; // For now.
-    }
-    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();
-      } else if (curr_buffer_num < deactivation_threshold) {
+      if (_worker_id > 0 && 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()) {
-        MutexLockerEx x(DirtyCardQ_CBL_mon, Mutex::_no_safepoint_check_flag);
+      if (_next != NULL && !_next->is_active() && curr_buffer_num > _next->_threshold) {
         _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) {
-        prev_buffer_num = curr_buffer_num;
-        _sts.leave();
-        os::sleep(Thread::current(), (jlong) _interval_ms, false);
-        _sts.join();
-        start_vtime_sec = os::elapsedVTime();
-      }
-      n_logs++;
+    // We can exit the loop above while being active if there was a yield request.
+    if (is_active()) {
+      deactivate();
     }
+
     _sts.leave();
 
     if (os::supports_vtime()) {
@@ -172,7 +213,6 @@
     }
   }
   assert(_should_terminate, "just checking");
-
   terminate();
 }
 
@@ -191,8 +231,8 @@
   }
 
   {
-    MutexLockerEx x(DirtyCardQ_CBL_mon, Mutex::_no_safepoint_check_flag);
-    DirtyCardQ_CBL_mon->notify_all();
+    MutexLockerEx x(_monitor, Mutex::_no_safepoint_check_flag);
+    _monitor->notify();
   }
 
   {
--- a/hotspot/src/share/vm/gc_implementation/g1/concurrentG1RefineThread.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/concurrentG1RefineThread.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -40,42 +40,36 @@
   // 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;
- public:
-  virtual void run();
+  ConcurrentG1RefineThread* _next;
+  Monitor* _monitor;
+  ConcurrentG1Refine* _cg1r;
 
-  bool is_active()  { return _active;  }
-  void activate()   { _active = true;  }
-  void deactivate() { _active = false; }
+  int _thread_threshold_step;
+  // This thread activation threshold
+  int _threshold;
+  // This thread deactivation threshold
+  int _deactivation_threshold;
 
- private:
-  ConcurrentG1Refine*              _cg1r;
+  void sample_young_list_rs_lengths();
+  void run_young_rs_sampling();
+  void wait_for_completed_buffers();
 
-  double                           _interval_ms;
-
-  void decreaseInterval(int processing_time_ms) {
-    double min_interval_ms = (double) processing_time_ms;
-    _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();
+  void set_active(bool x) { _active = x; }
+  bool is_active();
+  void activate();
+  void deactivate();
 
   // 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 @@
   // Total virtual time so far.
   double vtime_accum() { return _vtime_accum; }
 
-  ConcurrentG1Refine* cg1r()                     { return _cg1r;     }
-
-  void            sample_young_list_rs_lengths();
+  ConcurrentG1Refine* cg1r() { return _cg1r;     }
 
   // Yield for GC
-  void            yield();
-
+  void yield();
   // shutdown
   void stop();
 };
--- a/hotspot/src/share/vm/gc_implementation/g1/concurrentMark.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/concurrentMark.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -760,7 +760,6 @@
   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
--- a/hotspot/src/share/vm/gc_implementation/g1/dirtyCardQueue.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/dirtyCardQueue.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -61,8 +61,8 @@
 #pragma warning( disable:4355 ) // 'this' : used in base member initializer list
 #endif // _MSC_VER
 
-DirtyCardQueueSet::DirtyCardQueueSet() :
-  PtrQueueSet(true /*notify_when_complete*/),
+DirtyCardQueueSet::DirtyCardQueueSet(bool notify_when_complete) :
+  PtrQueueSet(notify_when_complete),
   _closure(NULL),
   _shared_dirty_card_queue(this, true /*perm*/),
   _free_ids(NULL),
@@ -77,12 +77,12 @@
 }
 
 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 @@
   return b;
 }
 
-DirtyCardQueueSet::CompletedBufferNode*
-DirtyCardQueueSet::get_completed_buffer_lock(int stop_at) {
-  CompletedBufferNode* nd = NULL;
+
+BufferNode*
+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 @@
 
   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,
-                                              nd->index, _sz,
+      DirtyCardQueue::apply_closure_to_buffer(_closure, buf,
+                                              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 @@
 
 bool DirtyCardQueueSet::apply_closure_to_completed_buffer(int worker_i,
                                                           int stop_at,
-                                                          bool with_CAS)
+                                                          bool during_pause)
 {
-  CompletedBufferNode* nd = NULL;
-  if (with_CAS) {
-    guarantee(stop_at == 0, "Precondition");
-    nd = get_completed_buffer_CAS();
-  } else {
-    nd = get_completed_buffer_lock(stop_at);
-  }
+  assert(!during_pause || stop_at == 0, "Should not leave any completed buffers during a pause");
+  BufferNode* nd = get_completed_buffer(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,
-                                              false);
+      DirtyCardQueue::apply_closure_to_buffer(_closure,
+                                              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;
-      _completed_buffers_head = nd->next;
-      nd->next = buffers_to_delete;
+      BufferNode* nd = _completed_buffers_head;
+      _completed_buffers_head = nd->next();
+      nd->set_next(buffers_to_delete);
       buffers_to_delete = nd;
     }
     _n_completed_buffers = 0;
@@ -263,10 +238,9 @@
     debug_only(assert_completed_buffer_list_len_correct_locked());
   }
   while (buffers_to_delete != NULL) {
-    CompletedBufferNode* nd = buffers_to_delete;
-    buffers_to_delete = nd->next;
-    deallocate_buffer(nd->buf);
-    delete nd;
+    BufferNode* nd = buffers_to_delete;
+    buffers_to_delete = nd->next();
+    deallocate_buffer(BufferNode::make_buffer_from_node(nd));
   }
   // Since abandon is done only at safepoints, we can safely manipulate
   // these queues.
--- a/hotspot/src/share/vm/gc_implementation/g1/dirtyCardQueue.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/dirtyCardQueue.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -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 @@
   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,
-                  Mutex* lock = NULL, PtrQueueSet* fl_owner = NULL);
+                  int process_completed_threshold,
+                  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 @@
   // 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);
+                                                BufferNode* nd);
 
-  CompletedBufferNode* get_completed_buffer_CAS();
-  CompletedBufferNode* get_completed_buffer_lock(int stop_at);
+  BufferNode* get_completed_buffer(int stop_at);
+
   // Applies the current closure to all completed buffers,
   // non-consumptively.
   void apply_closure_to_all_completed_buffers();
--- a/hotspot/src/share/vm/gc_implementation/g1/g1CollectedHeap.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1CollectedHeap.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -928,6 +928,8 @@
     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 @@
 
     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 @@
 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 @@
   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 @@
 
   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 @@
   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_locked(GCCause::Cause cause) {
-  // 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();
+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");
+
+  int gc_count_before;
   {
-    MutexUnlocker mu(Heap_lock);  // give up heap lock, execute gets it back
-    VM_G1CollectFull op(gc_count_before, cause);
-    VMThread::execute(&op);
+    MutexLocker ml(Heap_lock);
+    // Read the GC count while holding the Heap_lock
+    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::max_capacity() const {
-  return _g1_committed.byte_size();
+  return g1_reserved_obj_bytes();
 }
 
 jlong G1CollectedHeap::millis_since_last_gc() {
@@ -2638,6 +2650,8 @@
   }
 
   {
+    ResourceMark rm;
+
     char verbose_str[128];
     sprintf(verbose_str, "GC pause ");
     if (g1_policy()->in_young_gc_mode()) {
@@ -2649,8 +2663,6 @@
     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 @@
     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 @@
           _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 @@
 
       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 @@
     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());
 }
 
--- a/hotspot/src/share/vm/gc_implementation/g1/g1CollectedHeap.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1CollectedHeap.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -692,7 +692,7 @@
 
   // 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
--- a/hotspot/src/share/vm/gc_implementation/g1/g1CollectorPolicy.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1CollectorPolicy.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -1516,8 +1516,30 @@
       (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
-    assert(recent_avg_pause_time_ratio() < 1.01, "All GC?");
+    if (recent_avg_pause_time_ratio() < 0.0 ||
+        (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 @@
   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 @@
 
 // <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 @@
   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 @@
   G1CollectorPolicy::record_collection_pause_end(abandoned);
   assert(assertMarkedBytesDataOK(), "Marked regions not OK at pause end.");
 }
-
-// Local Variables: ***
-// c-indentation-style: gnu ***
-// End: ***
--- a/hotspot/src/share/vm/gc_implementation/g1/g1CollectorPolicy.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1CollectorPolicy.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -316,6 +316,10 @@
   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;
--- a/hotspot/src/share/vm/gc_implementation/g1/g1MMUTracker.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1MMUTracker.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -86,12 +86,22 @@
     //   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
-    //     (this mighte allow less GC time than what's allowed)
-    guarantee(0, "array full, currently we can't recover");
+    //   consolidate the two entries with the minimum gap between them
+    //     (this might allow less GC time than what's allowed)
+    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);
 }
 
--- a/hotspot/src/share/vm/gc_implementation/g1/g1MMUTracker.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1MMUTracker.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -99,7 +99,10 @@
   // 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;
--- a/hotspot/src/share/vm/gc_implementation/g1/g1_globals.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1_globals.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -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.")                                        \
                                                                             \
--- a/hotspot/src/share/vm/gc_implementation/g1/ptrQueue.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/ptrQueue.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -64,8 +64,8 @@
   while (_index == 0) {
     handle_zero_index();
   }
+
   assert(_index > 0, "postcondition");
-
   _index -= oopSize;
   _buf[byte_index_to_index((int)_index)] = ptr;
   assert(0 <= _index && _index <= _sz, "Invariant.");
@@ -99,94 +99,110 @@
   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;
-    _fl_owner->_buf_free_list = (void**)_fl_owner->_buf_free_list[0];
+    void** res = BufferNode::make_buffer_from_node(_fl_owner->_buf_free_list);
+    _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;
-  _fl_owner->_buf_free_list = buf;
+  BufferNode *node = BufferNode::make_node_from_buffer(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;
-    _buf_free_list = (void**)_buf_free_list[0];
-    FREE_C_HEAP_ARRAY(void*,head);
+    void* b = BufferNode::make_block_from_node(_buf_free_list);
+    _buf_free_list = _buf_free_list->next();
+    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) {
-  // I use explicit locking here because there's a bailout in the middle.
-  _cbl_mon->lock_without_safepoint_check();
+void PtrQueue::handle_zero_index() {
+  assert(0 == _index, "Precondition.");
+  // This thread records the full buffer and allocates a new one (while
+  // holding the lock if there is one).
+  if (_buf != NULL) {
+    if (_lock) {
+      locking_enqueue_completed_buffer(_buf);
+    } else {
+      if (qset()->process_or_enqueue_complete_buffer(_buf)) {
+        // Recycle the buffer. No allocation.
+        _sz = qset()->buffer_size();
+        _index = _sz;
+        return;
+      }
+    }
+  }
+  // Reallocate the buffer
+  _buf = qset()->allocate_buffer();
+  _sz = qset()->buffer_size();
+  _index = _sz;
+  assert(0 <= _index && _index <= _sz, "Invariant.");
+}
 
-  Thread* thread = Thread::current();
-  assert( ignore_max_completed ||
-          thread->is_Java_thread() ||
-          SafepointSynchronize::is_at_safepoint(),
-          "invariant" );
-  ignore_max_completed = ignore_max_completed || !thread->is_Java_thread();
+bool PtrQueueSet::process_or_enqueue_complete_buffer(void** buf) {
+  if (Thread::current()->is_Java_thread()) {
+    // We don't lock. It is fine to be epsilon-precise here.
+    if (_max_completed_queue == 0 || _max_completed_queue > 0 &&
+        _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;
+}
 
-  if (!ignore_max_completed && _max_completed_queue > 0 &&
-      _n_completed_buffers >= (size_t) _max_completed_queue) {
-    _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();
-  }
-
-  // Here we still hold the _cbl_mon.
-  CompletedBufferNode* cbn = new CompletedBufferNode;
-  cbn->buf = buf;
-  cbn->next = NULL;
-  cbn->index = index;
+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_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 @@
   _sz = sz * oopSize;
 }
 
-void PtrQueueSet::set_process_completed_threshold(size_t sz) {
-  _process_completed_threshold = sz;
-}
-
-// 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
+// Merge lists of buffers. Notify the processing threads.
+// 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 @@
   } 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 @@
   assert(_completed_buffers_head == NULL && _completed_buffers_tail == NULL ||
          _completed_buffers_head != NULL && _completed_buffers_tail != NULL,
          "Sanity");
+}
 
-  if (!_process_completed &&
-      _n_completed_buffers >= _process_completed_threshold) {
+void PtrQueueSet::notify_if_necessary() {
+  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;
-}
--- a/hotspot/src/share/vm/gc_implementation/g1/ptrQueue.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/ptrQueue.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -27,8 +27,10 @@
 // the addresses of modified old-generation objects.  This type supports
 // this operation.
 
+// 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 @@
     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 @@
 
 };
 
+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;
-  CompletedBufferNode* _completed_buffers_tail;
-  size_t _n_completed_buffers;
-  size_t _process_completed_threshold;
+  BufferNode* _completed_buffers_head;
+  BufferNode* _completed_buffers_tail;
+  int _n_completed_buffers;
+  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 @@
   // 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 @@
   // 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 @@
   void deallocate_buffer(void** buf);
 
   // Declares that "buf" is a complete buffer.
-  void enqueue_complete_buffer(void** buf, size_t index = 0,
-                               bool ignore_max_completed = false);
+  void enqueue_complete_buffer(void** buf, size_t index = 0);
+
+  // 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 @@
   // Get the buffer size.
   size_t buffer_size() { return _sz; }
 
-  // Set the number of completed buffers that triggers log processing.
-  void set_process_completed_threshold(size_t sz);
+  // Get/Set the number of completed buffers that triggers log processing.
+  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();
 };
--- a/hotspot/src/share/vm/gc_implementation/g1/satbQueue.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/satbQueue.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -67,9 +67,9 @@
 {}
 
 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 @@
 
 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 @@
   }
   ObjectClosure* cl = (par ? _par_closures[worker] : _closure);
   if (nd != NULL) {
-    ObjPtrQueue::apply_closure_to_buffer(cl, nd->buf, 0, _sz);
-    deallocate_buffer(nd->buf);
-    delete nd;
+    void **buf = BufferNode::make_buffer_from_node(nd);
+    ObjPtrQueue::apply_closure_to_buffer(cl, buf, 0, _sz);
+    deallocate_buffer(buf);
     return true;
   } else {
     return false;
@@ -145,13 +145,13 @@
 }
 
 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;
-      _completed_buffers_head = nd->next;
-      nd->next = buffers_to_delete;
+      BufferNode* nd = _completed_buffers_head;
+      _completed_buffers_head = nd->next();
+      nd->set_next(buffers_to_delete);
       buffers_to_delete = nd;
     }
     _completed_buffers_tail = NULL;
@@ -159,10 +159,9 @@
     DEBUG_ONLY(assert_completed_buffer_list_len_correct_locked());
   }
   while (buffers_to_delete != NULL) {
-    CompletedBufferNode* nd = buffers_to_delete;
-    buffers_to_delete = nd->next;
-    deallocate_buffer(nd->buf);
-    delete nd;
+    BufferNode* nd = buffers_to_delete;
+    buffers_to_delete = nd->next();
+    deallocate_buffer(BufferNode::make_buffer_from_node(nd));
   }
   assert(SafepointSynchronize::is_at_safepoint(), "Must be at safepoint.");
   // So we can safely manipulate these queues.
--- a/hotspot/src/share/vm/gc_implementation/g1/satbQueue.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/satbQueue.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -60,8 +60,8 @@
   SATBMarkQueueSet();
 
   void initialize(Monitor* cbl_mon, Mutex* fl_lock,
-                  int max_completed_queue = 0,
-                  Mutex* lock = NULL);
+                  int process_completed_threshold,
+                  Mutex* lock);
 
   static void handle_zero_index_for_thread(JavaThread* t);
 
--- a/hotspot/src/share/vm/gc_implementation/g1/vm_operations_g1.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/vm_operations_g1.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -42,7 +42,7 @@
 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();
 }
 
--- a/hotspot/src/share/vm/gc_implementation/g1/vm_operations_g1.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/gc_implementation/g1/vm_operations_g1.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -68,8 +68,9 @@
 
 class VM_G1IncCollectionPause: public VM_GC_Operation {
  public:
-  VM_G1IncCollectionPause(int gc_count_before) :
-    VM_GC_Operation(gc_count_before) {}
+  VM_G1IncCollectionPause(int 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 {
--- a/hotspot/src/share/vm/gc_implementation/includeDB_gc_g1	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/gc_implementation/includeDB_gc_g1	Wed Jul 05 17:05:01 2017 +0200
@@ -109,7 +109,6 @@
 dirtyCardQueue.cpp                      dirtyCardQueue.hpp
 dirtyCardQueue.cpp			heapRegionRemSet.hpp
 dirtyCardQueue.cpp                      mutexLocker.hpp
-dirtyCardQueue.cpp                      ptrQueue.inline.hpp
 dirtyCardQueue.cpp                      safepoint.hpp
 dirtyCardQueue.cpp                      thread.hpp
 dirtyCardQueue.cpp                      thread_<os_family>.inline.hpp
@@ -222,6 +221,15 @@
 g1MarkSweep.hpp                         timer.hpp
 g1MarkSweep.hpp                         universe.hpp
 
+g1MemoryPool.cpp                        heapRegion.hpp
+g1MemoryPool.cpp                        g1CollectedHeap.inline.hpp
+g1MemoryPool.cpp                        g1CollectedHeap.hpp
+g1MemoryPool.cpp                        g1CollectorPolicy.hpp
+g1MemoryPool.cpp                        g1MemoryPool.hpp
+
+g1MemoryPool.hpp                        memoryUsage.hpp
+g1MemoryPool.hpp                        memoryPool.hpp
+
 g1OopClosures.inline.hpp		concurrentMark.hpp
 g1OopClosures.inline.hpp		g1OopClosures.hpp
 g1OopClosures.inline.hpp		g1CollectedHeap.hpp
@@ -303,12 +311,13 @@
 
 klass.hpp				g1OopClosures.hpp
 
+memoryService.cpp                       g1MemoryPool.hpp
+
 ptrQueue.cpp                            allocation.hpp
 ptrQueue.cpp                            allocation.inline.hpp
 ptrQueue.cpp                            mutex.hpp
 ptrQueue.cpp                            mutexLocker.hpp
 ptrQueue.cpp                            ptrQueue.hpp
-ptrQueue.cpp                            ptrQueue.inline.hpp
 ptrQueue.cpp                            thread_<os_family>.inline.hpp
 
 ptrQueue.hpp                            allocation.hpp
@@ -318,7 +327,6 @@
 
 satbQueue.cpp                           allocation.inline.hpp
 satbQueue.cpp                           mutexLocker.hpp
-satbQueue.cpp                           ptrQueue.inline.hpp
 satbQueue.cpp                           satbQueue.hpp
 satbQueue.cpp                           sharedHeap.hpp
 satbQueue.cpp                           thread.hpp
--- a/hotspot/src/share/vm/gc_implementation/parallelScavenge/psGCAdaptivePolicyCounters.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/psGCAdaptivePolicyCounters.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -51,7 +51,7 @@
 
     cname = PerfDataManager::counter_name(name_space(), "oldCapacity");
     _old_capacity = PerfDataManager::create_variable(SUN_GC, cname,
-      PerfData::U_Bytes, (jlong) Arguments::initial_heap_size(), CHECK);
+      PerfData::U_Bytes, (jlong) InitialHeapSize, CHECK);
 
     cname = PerfDataManager::counter_name(name_space(), "boundaryMoved");
     _boundary_moved = PerfDataManager::create_variable(SUN_GC, cname,
--- a/hotspot/src/share/vm/includeDB_compiler2	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/includeDB_compiler2	Wed Jul 05 17:05:01 2017 +0200
@@ -149,6 +149,7 @@
 c2compiler.hpp                          abstractCompiler.hpp
 
 callGenerator.cpp                       addnode.hpp
+callGenerator.cpp                       bcEscapeAnalyzer.hpp
 callGenerator.cpp                       callGenerator.hpp
 callGenerator.cpp                       callnode.hpp
 callGenerator.cpp                       cfgnode.hpp
@@ -321,6 +322,7 @@
 compile.cpp                             rootnode.hpp
 compile.cpp                             runtime.hpp
 compile.cpp                             signature.hpp
+compile.cpp                             stringopts.hpp
 compile.cpp                             stubRoutines.hpp
 compile.cpp                             systemDictionary.hpp
 compile.cpp                             timer.hpp
@@ -476,12 +478,16 @@
 graphKit.cpp                            runtime.hpp
 graphKit.cpp                            sharedRuntime.hpp
 
+graphKit.hpp                            addnode.hpp
 graphKit.hpp                            callnode.hpp
 graphKit.hpp                            cfgnode.hpp
 graphKit.hpp                            ciEnv.hpp
+graphKit.hpp                            divnode.hpp
 graphKit.hpp                            compile.hpp
 graphKit.hpp                            deoptimization.hpp
 graphKit.hpp                            phaseX.hpp
+graphKit.hpp                            mulnode.hpp
+graphKit.hpp                            subnode.hpp
 graphKit.hpp                            type.hpp
 
 idealKit.cpp                            addnode.hpp
@@ -490,7 +496,10 @@
 idealKit.cpp                            idealKit.hpp
 idealKit.cpp				runtime.hpp
 
+idealKit.hpp                            addnode.hpp
+idealKit.hpp                            cfgnode.hpp
 idealKit.hpp                            connode.hpp
+idealKit.hpp                            divnode.hpp
 idealKit.hpp                            mulnode.hpp
 idealKit.hpp                            phaseX.hpp
 idealKit.hpp                            subnode.hpp
@@ -641,6 +650,7 @@
 macro.cpp                               callnode.hpp
 macro.cpp                               cfgnode.hpp
 macro.cpp                               compile.hpp
+macro.cpp                              compileLog.hpp
 macro.cpp                               connode.hpp
 macro.cpp                               locknode.hpp
 macro.cpp                               loopnode.hpp
@@ -993,6 +1003,21 @@
 split_if.cpp                            connode.hpp
 split_if.cpp                            loopnode.hpp
 
+stringopts.hpp                          phaseX.hpp
+stringopts.hpp                          node.hpp
+
+stringopts.cpp                          addnode.hpp
+stringopts.cpp                          callnode.hpp
+stringopts.cpp                          callGenerator.hpp
+stringopts.cpp                          compileLog.hpp
+stringopts.cpp                          divnode.hpp
+stringopts.cpp                          idealKit.hpp
+stringopts.cpp                          graphKit.hpp
+stringopts.cpp                          rootnode.hpp
+stringopts.cpp                          runtime.hpp
+stringopts.cpp                          subnode.hpp
+stringopts.cpp                          stringopts.hpp
+
 stubGenerator_<arch_model>.cpp          runtime.hpp
 
 stubRoutines.cpp                        runtime.hpp
--- a/hotspot/src/share/vm/includeDB_core	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/includeDB_core	Wed Jul 05 17:05:01 2017 +0200
@@ -289,7 +289,7 @@
 attachListener.hpp                      debug.hpp
 attachListener.hpp                      ostream.hpp
 
-barrierSet.cpp				barrierSet.hpp
+barrierSet.cpp				barrierSet.inline.hpp
 barrierSet.cpp			        collectedHeap.hpp
 barrierSet.cpp				universe.hpp
 
@@ -570,6 +570,7 @@
 ciEnv.hpp                               dependencies.hpp
 ciEnv.hpp                               exceptionHandlerTable.hpp
 ciEnv.hpp                               oopMap.hpp
+ciEnv.hpp                               systemDictionary.hpp
 ciEnv.hpp                               thread.hpp
 
 ciExceptionHandler.cpp                  ciExceptionHandler.hpp
@@ -1291,6 +1292,7 @@
 cpCacheOop.cpp                          markSweep.inline.hpp
 cpCacheOop.cpp                          objArrayOop.hpp
 cpCacheOop.cpp                          oop.inline.hpp
+cpCacheOop.cpp                          rewriter.hpp
 cpCacheOop.cpp                          universe.inline.hpp
 
 cpCacheOop.hpp                          allocation.hpp
--- a/hotspot/src/share/vm/interpreter/bytecodeInterpreter.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/interpreter/bytecodeInterpreter.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -281,7 +281,7 @@
 
 #define DO_BACKEDGE_CHECKS(skip, branch_pc)                                                         \
     if ((skip) <= 0) {                                                                              \
-      if (UseCompiler && UseLoopCounter) {                                                          \
+      if (UseLoopCounter) {                                                                         \
         bool do_OSR = UseOnStackReplacement;                                                        \
         BACKEDGE_COUNT->increment();                                                                \
         if (do_OSR) do_OSR = BACKEDGE_COUNT->reached_InvocationLimit();                             \
@@ -289,16 +289,12 @@
           nmethod*  osr_nmethod;                                                                    \
           OSR_REQUEST(osr_nmethod, branch_pc);                                                      \
           if (osr_nmethod != NULL && osr_nmethod->osr_entry_bci() != InvalidOSREntryBci) {          \
-            intptr_t* buf;                                                                          \
-            CALL_VM(buf=SharedRuntime::OSR_migration_begin(THREAD), handle_exception);              \
+            intptr_t* buf = SharedRuntime::OSR_migration_begin(THREAD);                             \
             istate->set_msg(do_osr);                                                                \
             istate->set_osr_buf((address)buf);                                                      \
             istate->set_osr_entry(osr_nmethod->osr_entry());                                        \
             return;                                                                                 \
           }                                                                                         \
-        } else {                                                                                    \
-          INCR_INVOCATION_COUNT;                                                                    \
-          SAFEPOINT;                                                                                \
         }                                                                                           \
       }  /* UseCompiler ... */                                                                      \
       INCR_INVOCATION_COUNT;                                                                        \
@@ -1281,12 +1277,7 @@
           jfloat f;
           jdouble r;
           f = STACK_FLOAT(-1);
-#ifdef IA64
-          // IA64 gcc bug
-          r = ( f == 0.0f ) ? (jdouble) f : (jdouble) f + ia64_double_zero;
-#else
           r = (jdouble) f;
-#endif
           MORE_STACK(-1); // POP
           SET_STACK_DOUBLE(r, 1);
           UPDATE_PC_AND_TOS_AND_CONTINUE(1, 2);
--- a/hotspot/src/share/vm/interpreter/bytecodeTracer.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/interpreter/bytecodeTracer.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -282,18 +282,21 @@
   constantPoolOop constants = method()->constants();
   constantTag tag = constants->tag_at(i);
 
+  int nt_index = -1;
+
   switch (tag.value()) {
   case JVM_CONSTANT_InterfaceMethodref:
   case JVM_CONSTANT_Methodref:
   case JVM_CONSTANT_Fieldref:
+  case JVM_CONSTANT_NameAndType:
     break;
   default:
     st->print_cr(" bad tag=%d at %d", tag.value(), i);
     return;
   }
 
-  symbolOop name = constants->name_ref_at(orig_i);
-  symbolOop signature = constants->signature_ref_at(orig_i);
+  symbolOop name = constants->uncached_name_ref_at(i);
+  symbolOop signature = constants->uncached_signature_ref_at(i);
   st->print_cr(" %d <%s> <%s> ", i, name->as_C_string(), signature->as_C_string());
 }
 
--- a/hotspot/src/share/vm/interpreter/interpreter.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/interpreter/interpreter.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -314,6 +314,20 @@
       break;
     }
 
+   case Bytecodes::_invokedynamic: {
+      Thread *thread = Thread::current();
+      ResourceMark rm(thread);
+      methodHandle mh(thread, method);
+      type = Bytecode_invoke_at(mh, bci)->result_type(thread);
+      // since the cache entry might not be initialized:
+      // (NOT needed for the old calling convension)
+      if (!is_top_frame) {
+        int index = Bytes::get_native_u4(bcp+1);
+        method->constants()->cache()->entry_at(index)->set_parameter_size(callee_parameters);
+      }
+      break;
+    }
+
     case Bytecodes::_ldc   :
       type = constant_pool_type( method, *(bcp+1) );
       break;
--- a/hotspot/src/share/vm/interpreter/interpreterRuntime.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/interpreter/interpreterRuntime.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -681,7 +681,7 @@
 IRT_END
 
 
-// First time execution:  Resolve symbols, create a permanent CallSiteImpl object.
+// First time execution:  Resolve symbols, create a permanent CallSite object.
 IRT_ENTRY(void, InterpreterRuntime::resolve_invokedynamic(JavaThread* thread)) {
   ResourceMark rm(thread);
 
@@ -708,21 +708,16 @@
   constantPoolHandle pool(thread, caller_method->constants());
   pool->set_invokedynamic();    // mark header to flag active call sites
 
-  int raw_index = four_byte_index(thread);
-  assert(constantPoolCacheOopDesc::is_secondary_index(raw_index), "invokedynamic indexes marked specially");
-
-  // there are two CPC entries that are of interest:
-  int site_index = constantPoolCacheOopDesc::decode_secondary_index(raw_index);
-  int main_index = pool->cache()->entry_at(site_index)->main_entry_index();
-  // and there is one CP entry, a NameAndType:
-  int nt_index = pool->map_instruction_operand_to_index(raw_index);
+  int site_index = four_byte_index(thread);
+  // there is a second CPC entries that is of interest; it caches signature info:
+  int main_index = pool->cache()->secondary_entry_at(site_index)->main_entry_index();
 
   // first resolve the signature to a MH.invoke methodOop
   if (!pool->cache()->entry_at(main_index)->is_resolved(bytecode)) {
     JvmtiHideSingleStepping jhss(thread);
     CallInfo info;
     LinkResolver::resolve_invoke(info, Handle(), pool,
-                                 raw_index, bytecode, CHECK);
+                                 site_index, bytecode, CHECK);
     // The main entry corresponds to a JVM_CONSTANT_NameAndType, and serves
     // as a common reference point for all invokedynamic call sites with
     // that exact call descriptor.  We will link it in the CP cache exactly
@@ -741,7 +736,7 @@
   assert(mh_invdyn.not_null() && mh_invdyn->is_method() && mh_invdyn->is_method_handle_invoke(),
          "correct result from LinkResolver::resolve_invokedynamic");
 
-  symbolHandle call_site_name(THREAD, pool->nt_name_ref_at(nt_index));
+  symbolHandle call_site_name(THREAD, pool->name_ref_at(site_index));
   Handle call_site
     = SystemDictionary::make_dynamic_call_site(caller_method->method_holder(),
                                                caller_method->method_idnum(),
@@ -753,61 +748,11 @@
   // In the secondary entry, the f1 field is the call site, and the f2 (index)
   // field is some data about the invoke site.
   int extra_data = 0;
-  pool->cache()->entry_at(site_index)->set_dynamic_call(call_site(), extra_data);
+  pool->cache()->secondary_entry_at(site_index)->set_dynamic_call(call_site(), extra_data);
 }
 IRT_END
 
 
-// Called on first time execution, and also whenever the CallSite.target is null.
-// FIXME:  Do more of this in Java code.
-IRT_ENTRY(void, InterpreterRuntime::bootstrap_invokedynamic(JavaThread* thread, oopDesc* call_site)) {
-  methodHandle   mh_invdyn(thread, (methodOop) sun_dyn_CallSiteImpl::vmmethod(call_site));
-  Handle         mh_type(thread,   mh_invdyn->method_handle_type());
-  objArrayHandle mh_ptypes(thread, java_dyn_MethodType::ptypes(mh_type()));
-
-  // squish the arguments down to a single array
-  int nargs = mh_ptypes->length();
-  objArrayHandle arg_array;
-  {
-    objArrayOop aaoop = oopFactory::new_objArray(SystemDictionary::object_klass(), nargs, CHECK);
-    arg_array = objArrayHandle(thread, aaoop);
-  }
-  frame fr = thread->last_frame();
-  assert(fr.interpreter_frame_bcp() != NULL, "sanity");
-  int tos_offset = 0;
-  for (int i = nargs; --i >= 0; ) {
-    intptr_t* slot_addr = fr.interpreter_frame_tos_at(tos_offset++);
-    oop ptype = mh_ptypes->obj_at(i);
-    oop arg = NULL;
-    if (!java_lang_Class::is_primitive(ptype)) {
-      arg = *(oop*) slot_addr;
-    } else {
-      BasicType bt = java_lang_Class::primitive_type(ptype);
-      assert(frame::interpreter_frame_expression_stack_direction() < 0, "else reconsider this code");
-      jvalue value;
-      Interpreter::get_jvalue_in_slot(slot_addr, bt, &value);
-      tos_offset += type2size[bt]-1;
-      arg = java_lang_boxing_object::create(bt, &value, CHECK);
-      // FIXME:  These boxing objects are not canonicalized under
-      // the Java autoboxing rules.  They should be...
-      // The best approach would be to push the arglist creation into Java.
-      // The JVM should use a lower-level interface to communicate argument lists.
-    }
-    arg_array->obj_at_put(i, arg);
-  }
-
-  // now find the bootstrap method
-  oop bootstrap_mh_oop = instanceKlass::cast(fr.interpreter_frame_method()->method_holder())->bootstrap_method();
-  assert(bootstrap_mh_oop != NULL, "resolve_invokedynamic ensures a BSM");
-
-  // return the bootstrap method and argument array via vm_result/_2
-  thread->set_vm_result(bootstrap_mh_oop);
-  thread->set_vm_result_2(arg_array());
-}
-IRT_END
-
-
-
 //------------------------------------------------------------------------------------------------------------------------
 // Miscellaneous
 
--- a/hotspot/src/share/vm/interpreter/interpreterRuntime.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/interpreter/interpreterRuntime.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -91,7 +91,6 @@
   // Calls
   static void    resolve_invoke       (JavaThread* thread, Bytecodes::Code bytecode);
   static void    resolve_invokedynamic(JavaThread* thread);
-  static void  bootstrap_invokedynamic(JavaThread* thread, oopDesc* call_site);
 
   // Breakpoints
   static void _breakpoint(JavaThread* thread, methodOopDesc* method, address bcp);
--- a/hotspot/src/share/vm/interpreter/linkResolver.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/interpreter/linkResolver.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -1015,11 +1015,8 @@
 
   // This guy is reached from InterpreterRuntime::resolve_invokedynamic.
 
-  assert(constantPoolCacheOopDesc::is_secondary_index(raw_index), "must be secondary index");
-  int nt_index = pool->map_instruction_operand_to_index(raw_index);
-
   // At this point, we only need the signature, and can ignore the name.
-  symbolHandle method_signature(THREAD, pool->nt_signature_ref_at(nt_index));
+  symbolHandle method_signature(THREAD, pool->signature_ref_at(raw_index));  // raw_index works directly
   symbolHandle method_name = vmSymbolHandles::invoke_name();
   KlassHandle resolved_klass = SystemDictionaryHandles::MethodHandle_klass();
 
--- a/hotspot/src/share/vm/interpreter/rewriter.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/interpreter/rewriter.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -48,16 +48,6 @@
 }
 
 
-int Rewriter::add_extra_cp_cache_entry(int main_entry) {
-  // Hack: We put it on the map as an encoded value.
-  // The only place that consumes this is ConstantPoolCacheEntry::set_initial_state
-  int encoded = constantPoolCacheOopDesc::encode_secondary_index(main_entry);
-  int plain_secondary_index = _cp_cache_map.append(encoded);
-  return constantPoolCacheOopDesc::encode_secondary_index(plain_secondary_index);
-}
-
-
-
 // Creates a constant pool cache given a CPC map
 // This creates the constant pool cache initially in a state
 // that is unsafe for concurrent GC processing but sets it to
@@ -127,7 +117,7 @@
   assert(p[-1] == Bytecodes::_invokedynamic, "");
   int cp_index = Bytes::get_Java_u2(p);
   int cpc  = maybe_add_cp_cache_entry(cp_index);  // add lazily
-  int cpc2 = add_extra_cp_cache_entry(cpc);
+  int cpc2 = add_secondary_cp_cache_entry(cpc);
 
   // Replace the trailing four bytes with a CPC index for the dynamic
   // call site.  Unlike other CPC entries, there is one per bytecode,
@@ -137,7 +127,7 @@
   // all these entries.  That is the main reason invokedynamic
   // must have a five-byte instruction format.  (Of course, other JVM
   // implementations can use the bytes for other purposes.)
-  Bytes::put_native_u4(p, cpc2);
+  Bytes::put_native_u4(p, constantPoolCacheOopDesc::encode_secondary_index(cpc2));
   // Note: We use native_u4 format exclusively for 4-byte indexes.
 }
 
--- a/hotspot/src/share/vm/interpreter/rewriter.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/interpreter/rewriter.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -43,13 +43,18 @@
   bool has_cp_cache(int i) { return (uint)i < (uint)_cp_map.length() && _cp_map[i] >= 0; }
   int maybe_add_cp_cache_entry(int i) { return has_cp_cache(i) ? _cp_map[i] : add_cp_cache_entry(i); }
   int add_cp_cache_entry(int cp_index) {
+    assert((cp_index & _secondary_entry_tag) == 0, "bad tag");
     assert(_cp_map[cp_index] == -1, "not twice on same cp_index");
     int cache_index = _cp_cache_map.append(cp_index);
     _cp_map.at_put(cp_index, cache_index);
     assert(cp_entry_to_cp_cache(cp_index) == cache_index, "");
     return cache_index;
   }
-  int add_extra_cp_cache_entry(int main_entry);
+  int add_secondary_cp_cache_entry(int main_cpc_entry) {
+    assert(main_cpc_entry < _cp_cache_map.length(), "must be earlier CP cache entry");
+    int cache_index = _cp_cache_map.append(main_cpc_entry | _secondary_entry_tag);
+    return cache_index;
+  }
 
   // All the work goes in here:
   Rewriter(instanceKlassHandle klass, TRAPS);
@@ -65,4 +70,8 @@
  public:
   // Driver routine:
   static void rewrite(instanceKlassHandle klass, TRAPS);
+
+  enum {
+    _secondary_entry_tag = nth_bit(30)
+  };
 };
--- a/hotspot/src/share/vm/interpreter/templateInterpreter.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/interpreter/templateInterpreter.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -178,14 +178,12 @@
 #endif // !PRODUCT
 EntryPoint TemplateInterpreter::_return_entry[TemplateInterpreter::number_of_return_entries];
 EntryPoint TemplateInterpreter::_earlyret_entry;
-EntryPoint TemplateInterpreter::_return_unbox_entry;
 EntryPoint TemplateInterpreter::_deopt_entry [TemplateInterpreter::number_of_deopt_entries ];
 EntryPoint TemplateInterpreter::_continuation_entry;
 EntryPoint TemplateInterpreter::_safept_entry;
 
 address    TemplateInterpreter::_return_3_addrs_by_index[TemplateInterpreter::number_of_return_addrs];
 address    TemplateInterpreter::_return_5_addrs_by_index[TemplateInterpreter::number_of_return_addrs];
-address    TemplateInterpreter::_return_5_unbox_addrs_by_index[TemplateInterpreter::number_of_return_addrs];
 
 DispatchTable TemplateInterpreter::_active_table;
 DispatchTable TemplateInterpreter::_normal_table;
@@ -253,22 +251,6 @@
     }
   }
 
-  if (EnableInvokeDynamic) {
-    CodeletMark cm(_masm, "unboxing return entry points");
-    Interpreter::_return_unbox_entry =
-      EntryPoint(
-        generate_return_unbox_entry_for(btos, 5),
-        generate_return_unbox_entry_for(ctos, 5),
-        generate_return_unbox_entry_for(stos, 5),
-        generate_return_unbox_entry_for(atos, 5), // cast conversion
-        generate_return_unbox_entry_for(itos, 5),
-        generate_return_unbox_entry_for(ltos, 5),
-        generate_return_unbox_entry_for(ftos, 5),
-        generate_return_unbox_entry_for(dtos, 5),
-        Interpreter::_return_entry[5].entry(vtos) // no unboxing for void
-      );
-  }
-
   { CodeletMark cm(_masm, "earlyret entry points");
     Interpreter::_earlyret_entry =
       EntryPoint(
@@ -319,8 +301,6 @@
     int index = Interpreter::TosState_as_index(states[j]);
     Interpreter::_return_3_addrs_by_index[index] = Interpreter::return_entry(states[j], 3);
     Interpreter::_return_5_addrs_by_index[index] = Interpreter::return_entry(states[j], 5);
-    if (EnableInvokeDynamic)
-      Interpreter::_return_5_unbox_addrs_by_index[index] = Interpreter::return_unbox_entry(states[j], 5);
   }
 
   { CodeletMark cm(_masm, "continuation entry points");
@@ -485,9 +465,11 @@
 void TemplateInterpreterGenerator::set_short_entry_points(Template* t, address& bep, address& cep, address& sep, address& aep, address& iep, address& lep, address& fep, address& dep, address& vep) {
   assert(t->is_valid(), "template must exist");
   switch (t->tos_in()) {
-    case btos: vep = __ pc(); __ pop(btos); bep = __ pc(); generate_and_dispatch(t); break;
-    case ctos: vep = __ pc(); __ pop(ctos); sep = __ pc(); generate_and_dispatch(t); break;
-    case stos: vep = __ pc(); __ pop(stos); sep = __ pc(); generate_and_dispatch(t); break;
+    case btos:
+    case ctos:
+    case stos:
+      ShouldNotReachHere();  // btos/ctos/stos should use itos.
+      break;
     case atos: vep = __ pc(); __ pop(atos); aep = __ pc(); generate_and_dispatch(t); break;
     case itos: vep = __ pc(); __ pop(itos); iep = __ pc(); generate_and_dispatch(t); break;
     case ltos: vep = __ pc(); __ pop(ltos); lep = __ pc(); generate_and_dispatch(t); break;
@@ -547,18 +529,6 @@
 }
 
 
-address TemplateInterpreter::return_unbox_entry(TosState state, int length) {
-  assert(EnableInvokeDynamic, "");
-  if (state == vtos) {
-    // no unboxing to do, actually
-    return return_entry(state, length);
-  } else {
-    assert(length == 5, "unboxing entries generated for invokedynamic only");
-    return _return_unbox_entry.entry(state);
-  }
-}
-
-
 address TemplateInterpreter::deopt_entry(TosState state, int length) {
   guarantee(0 <= length && length < Interpreter::number_of_deopt_entries, "illegal length");
   return _deopt_entry[length].entry(state);
--- a/hotspot/src/share/vm/interpreter/templateInterpreter.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/interpreter/templateInterpreter.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -110,14 +110,12 @@
 #endif // !PRODUCT
   static EntryPoint _return_entry[number_of_return_entries];    // entry points to return to from a call
   static EntryPoint _earlyret_entry;                            // entry point to return early from a call
-  static EntryPoint _return_unbox_entry;                        // entry point to unbox a return value from a call
   static EntryPoint _deopt_entry[number_of_deopt_entries];      // entry points to return to from a deoptimization
   static EntryPoint _continuation_entry;
   static EntryPoint _safept_entry;
 
   static address    _return_3_addrs_by_index[number_of_return_addrs];     // for invokevirtual   return entries
   static address    _return_5_addrs_by_index[number_of_return_addrs];     // for invokeinterface return entries
-  static address    _return_5_unbox_addrs_by_index[number_of_return_addrs]; // for invokedynamic bootstrap methods
 
   static DispatchTable _active_table;                           // the active    dispatch table (used by the interpreter for dispatch)
   static DispatchTable _normal_table;                           // the normal    dispatch table (used to set the active table in normal mode)
@@ -159,12 +157,10 @@
   // Support for invokes
   static address*   return_3_addrs_by_index_table()             { return _return_3_addrs_by_index; }
   static address*   return_5_addrs_by_index_table()             { return _return_5_addrs_by_index; }
-  static address*   return_5_unbox_addrs_by_index_table()       { return _return_5_unbox_addrs_by_index; }
   static int        TosState_as_index(TosState state);          // computes index into return_3_entry_by_index table
 
   static address    return_entry  (TosState state, int length);
   static address    deopt_entry   (TosState state, int length);
-  static address    return_unbox_entry(TosState state, int length);
 
   // Safepoint support
   static void       notice_safepoints();                        // stops the thread when reaching a safepoint
--- a/hotspot/src/share/vm/interpreter/templateInterpreterGenerator.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/interpreter/templateInterpreterGenerator.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -51,10 +51,7 @@
   address generate_WrongMethodType_handler();
   address generate_ArrayIndexOutOfBounds_handler(const char* name);
   address generate_continuation_for(TosState state);
-  address generate_return_entry_for(TosState state, int step, bool unbox = false);
-  address generate_return_unbox_entry_for(TosState state, int step) {
-    return generate_return_entry_for(state, step, true);
-  }
+  address generate_return_entry_for(TosState state, int step);
   address generate_earlyret_entry_for(TosState state);
   address generate_deopt_entry_for(TosState state, int step);
   address generate_safept_entry_for(TosState state, address runtime_entry);
--- a/hotspot/src/share/vm/memory/barrierSet.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/memory/barrierSet.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -41,11 +41,6 @@
 
 // count is number of array elements being written
 void BarrierSet::static_write_ref_array_post(HeapWord* start, size_t count) {
-  assert(count <= (size_t)max_intx, "count too large");
-  HeapWord* end = start + objArrayOopDesc::array_size((int)count);
-#if 0
-  warning("Post:\t" INTPTR_FORMAT "[" SIZE_FORMAT "] : [" INTPTR_FORMAT","INTPTR_FORMAT")\t",
-                   start,            count,              start,          end);
-#endif
-  Universe::heap()->barrier_set()->write_ref_array_work(MemRegion(start, end));
+  // simply delegate to instance method
+  Universe::heap()->barrier_set()->write_ref_array(start, count);
 }
--- a/hotspot/src/share/vm/memory/barrierSet.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/memory/barrierSet.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -121,17 +121,20 @@
   virtual void read_ref_array(MemRegion mr) = 0;
   virtual void read_prim_array(MemRegion mr) = 0;
 
+  // Below length is the # array elements being written
   virtual void write_ref_array_pre(      oop* dst, int length) {}
   virtual void write_ref_array_pre(narrowOop* dst, int length) {}
+  // Below MemRegion mr is expected to be HeapWord-aligned
   inline void write_ref_array(MemRegion mr);
+  // Below count is the # array elements being written, starting
+  // at the address "start", which may not necessarily be HeapWord-aligned
+  inline void write_ref_array(HeapWord* start, size_t count);
 
-  // Static versions, suitable for calling from generated code.
+  // Static versions, suitable for calling from generated code;
+  // count is # array elements being written, starting with "start",
+  // which may not necessarily be HeapWord-aligned.
   static void static_write_ref_array_pre(HeapWord* start, size_t count);
   static void static_write_ref_array_post(HeapWord* start, size_t count);
-  // Narrow oop versions of the above; count is # of array elements being written,
-  // starting with "start", which is HeapWord-aligned.
-  static void static_write_ref_array_pre_narrow(HeapWord* start, size_t count);
-  static void static_write_ref_array_post_narrow(HeapWord* start, size_t count);
 
 protected:
   virtual void write_ref_array_work(MemRegion mr) = 0;
--- a/hotspot/src/share/vm/memory/barrierSet.inline.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/memory/barrierSet.inline.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -43,6 +43,8 @@
 }
 
 void BarrierSet::write_ref_array(MemRegion mr) {
+  assert((HeapWord*)align_size_down((uintptr_t)mr.start(), HeapWordSize) == mr.start() , "Unaligned start");
+  assert((HeapWord*)align_size_up  ((uintptr_t)mr.end(),   HeapWordSize) == mr.end(),    "Unaligned end"  );
   if (kind() == CardTableModRef) {
     ((CardTableModRefBS*)this)->inline_write_ref_array(mr);
   } else {
@@ -50,6 +52,34 @@
   }
 }
 
+// count is number of array elements being written
+void BarrierSet::write_ref_array(HeapWord* start, size_t count) {
+  assert(count <= (size_t)max_intx, "count too large");
+  HeapWord* end = (HeapWord*)((char*)start + (count*heapOopSize));
+  // In the case of compressed oops, start and end may potentially be misaligned;
+  // so we need to conservatively align the first downward (this is not
+  // strictly necessary for current uses, but a case of good hygiene and,
+  // if you will, aesthetics) and the second upward (this is essential for
+  // current uses) to a HeapWord boundary, so we mark all cards overlapping
+  // this write. In the event that this evolves in the future to calling a
+  // logging barrier of narrow oop granularity, like the pre-barrier for G1
+  // (mentioned here merely by way of example), we will need to change this
+  // interface, much like the pre-barrier one above, so it is "exactly precise"
+  // (if i may be allowed the adverbial redundancy for emphasis) and does not
+  // include narrow oop slots not included in the original write interval.
+  HeapWord* aligned_start = (HeapWord*)align_size_down((uintptr_t)start, HeapWordSize);
+  HeapWord* aligned_end   = (HeapWord*)align_size_up  ((uintptr_t)end,   HeapWordSize);
+  // If compressed oops were not being used, these should already be aligned
+  assert(UseCompressedOops || (aligned_start == start && aligned_end == end),
+         "Expected heap word alignment of start and end");
+#if 0
+  warning("Post:\t" INTPTR_FORMAT "[" SIZE_FORMAT "] : [" INTPTR_FORMAT","INTPTR_FORMAT")\t",
+                   start,            count,              aligned_start,   aligned_end);
+#endif
+  write_ref_array_work(MemRegion(aligned_start, aligned_end));
+}
+
+
 void BarrierSet::write_region(MemRegion mr) {
   if (kind() == CardTableModRef) {
     ((CardTableModRefBS*)this)->inline_write_region(mr);
--- a/hotspot/src/share/vm/memory/cardTableModRefBS.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/memory/cardTableModRefBS.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -511,6 +511,8 @@
 }
 
 void CardTableModRefBS::dirty_MemRegion(MemRegion mr) {
+  assert((HeapWord*)align_size_down((uintptr_t)mr.start(), HeapWordSize) == mr.start(), "Unaligned start");
+  assert((HeapWord*)align_size_up  ((uintptr_t)mr.end(),   HeapWordSize) == mr.end(),   "Unaligned end"  );
   jbyte* cur  = byte_for(mr.start());
   jbyte* last = byte_after(mr.last());
   while (cur < last) {
@@ -520,6 +522,8 @@
 }
 
 void CardTableModRefBS::invalidate(MemRegion mr, bool whole_heap) {
+  assert((HeapWord*)align_size_down((uintptr_t)mr.start(), HeapWordSize) == mr.start(), "Unaligned start");
+  assert((HeapWord*)align_size_up  ((uintptr_t)mr.end(),   HeapWordSize) == mr.end(),   "Unaligned end"  );
   for (int i = 0; i < _cur_covered_regions; i++) {
     MemRegion mri = mr.intersection(_covered[i]);
     if (!mri.is_empty()) dirty_MemRegion(mri);
--- a/hotspot/src/share/vm/memory/collectorPolicy.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/memory/collectorPolicy.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -55,7 +55,7 @@
 
 void CollectorPolicy::initialize_size_info() {
   // User inputs from -mx and ms are aligned
-  set_initial_heap_byte_size(Arguments::initial_heap_size());
+  set_initial_heap_byte_size(InitialHeapSize);
   if (initial_heap_byte_size() == 0) {
     set_initial_heap_byte_size(NewSize + OldSize);
   }
--- a/hotspot/src/share/vm/memory/heap.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/memory/heap.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -464,7 +464,7 @@
   }
 
   // Verify that freelist contains the right amount of free space
-  guarantee(len == _free_segments, "wrong freelist");
+  //  guarantee(len == _free_segments, "wrong freelist");
 
   // Verify that the number of free blocks is not out of hand.
   static int free_block_threshold = 10000;
@@ -479,5 +479,5 @@
   for(HeapBlock *h = first_block(); h != NULL; h = next_block(h)) {
     if (h->free()) count--;
   }
-  guarantee(count == 0, "missing free blocks");
+  //  guarantee(count == 0, "missing free blocks");
 }
--- a/hotspot/src/share/vm/memory/referenceProcessor.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/memory/referenceProcessor.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -299,8 +299,8 @@
 
 
 template <class T>
-static bool enqueue_discovered_ref_helper(ReferenceProcessor* ref,
-                                          AbstractRefProcTaskExecutor* task_executor) {
+bool enqueue_discovered_ref_helper(ReferenceProcessor* ref,
+                                   AbstractRefProcTaskExecutor* task_executor) {
 
   // Remember old value of pending references list
   T* pending_list_addr = (T*)java_lang_ref_Reference::pending_list_addr();
--- a/hotspot/src/share/vm/memory/sharedHeap.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/memory/sharedHeap.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -224,10 +224,6 @@
                           CodeBlobClosure* code_roots,
                           OopClosure* non_root_closure);
 
-
-  // Like CollectedHeap::collect, but assume that the caller holds the Heap_lock.
-  virtual void collect_locked(GCCause::Cause cause) = 0;
-
   // The functions below are helper functions that a subclass of
   // "SharedHeap" can use in the implementation of its virtual
   // functions.
--- a/hotspot/src/share/vm/memory/universe.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/memory/universe.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -67,6 +67,8 @@
 objArrayOop Universe::_the_empty_system_obj_array     = NULL;
 objArrayOop Universe::_the_empty_class_klass_array    = NULL;
 objArrayOop Universe::_the_array_interfaces_array     = NULL;
+oop Universe::_the_null_string                        = NULL;
+oop Universe::_the_min_jint_string                   = NULL;
 LatestMethodOopCache* Universe::_finalizer_register_cache = NULL;
 LatestMethodOopCache* Universe::_loader_addClass_cache    = NULL;
 ActiveMethodOopsCache* Universe::_reflect_invoke_cache    = NULL;
@@ -187,6 +189,8 @@
   f->do_oop((oop*)&_the_empty_system_obj_array);
   f->do_oop((oop*)&_the_empty_class_klass_array);
   f->do_oop((oop*)&_the_array_interfaces_array);
+  f->do_oop((oop*)&_the_null_string);
+  f->do_oop((oop*)&_the_min_jint_string);
   _finalizer_register_cache->oops_do(f);
   _loader_addClass_cache->oops_do(f);
   _reflect_invoke_cache->oops_do(f);
@@ -289,6 +293,9 @@
 
     klassOop ok = SystemDictionary::object_klass();
 
+    _the_null_string            = StringTable::intern("null", CHECK);
+    _the_min_jint_string       = StringTable::intern("-2147483648", CHECK);
+
     if (UseSharedSpaces) {
       // Verify shared interfaces array.
       assert(_the_array_interfaces_array->obj_at(0) ==
@@ -744,22 +751,22 @@
 static const uint64_t OopEncodingHeapMax = NarrowOopHeapMax << LogMinObjAlignmentInBytes;
 
 char* Universe::preferred_heap_base(size_t heap_size, NARROW_OOP_MODE mode) {
+  size_t base = 0;
 #ifdef _LP64
   if (UseCompressedOops) {
     assert(mode == UnscaledNarrowOop  ||
            mode == ZeroBasedNarrowOop ||
            mode == HeapBasedNarrowOop, "mode is invalid");
+    const size_t total_size = heap_size + HeapBaseMinAddress;
     // Return specified base for the first request.
     if (!FLAG_IS_DEFAULT(HeapBaseMinAddress) && (mode == UnscaledNarrowOop)) {
-      return (char*)HeapBaseMinAddress;
-    }
-    const size_t total_size = heap_size + HeapBaseMinAddress;
-    if (total_size <= OopEncodingHeapMax && (mode != HeapBasedNarrowOop)) {
+      base = HeapBaseMinAddress;
+    } else if (total_size <= OopEncodingHeapMax && (mode != HeapBasedNarrowOop)) {
       if (total_size <= NarrowOopHeapMax && (mode == UnscaledNarrowOop) &&
           (Universe::narrow_oop_shift() == 0)) {
         // Use 32-bits oops without encoding and
         // place heap's top on the 4Gb boundary
-        return (char*)(NarrowOopHeapMax - heap_size);
+        base = (NarrowOopHeapMax - heap_size);
       } else {
         // Can't reserve with NarrowOopShift == 0
         Universe::set_narrow_oop_shift(LogMinObjAlignmentInBytes);
@@ -768,16 +775,38 @@
           // Use zero based compressed oops with encoding and
           // place heap's top on the 32Gb boundary in case
           // total_size > 4Gb or failed to reserve below 4Gb.
-          return (char*)(OopEncodingHeapMax - heap_size);
+          base = (OopEncodingHeapMax - heap_size);
         }
       }
     } else {
       // Can't reserve below 32Gb.
       Universe::set_narrow_oop_shift(LogMinObjAlignmentInBytes);
     }
+    // Set narrow_oop_base and narrow_oop_use_implicit_null_checks
+    // used in ReservedHeapSpace() constructors.
+    // The final values will be set in initialize_heap() below.
+    if (base != 0 && (base + heap_size) <= OopEncodingHeapMax) {
+      // Use zero based compressed oops
+      Universe::set_narrow_oop_base(NULL);
+      // Don't need guard page for implicit checks in indexed
+      // addressing mode with zero based Compressed Oops.
+      Universe::set_narrow_oop_use_implicit_null_checks(true);
+    } else {
+      // Set to a non-NULL value so the ReservedSpace ctor computes
+      // the correct no-access prefix.
+      // The final value will be set in initialize_heap() below.
+      Universe::set_narrow_oop_base((address)NarrowOopHeapMax);
+#ifdef _WIN64
+      if (UseLargePages) {
+        // Cannot allocate guard pages for implicit checks in indexed
+        // addressing mode when large pages are specified on windows.
+        Universe::set_narrow_oop_use_implicit_null_checks(false);
+      }
+#endif //  _WIN64
+    }
   }
 #endif
-  return NULL; // also return NULL (don't care) for 32-bit VM
+  return (char*)base; // also return NULL (don't care) for 32-bit VM
 }
 
 jint Universe::initialize_heap() {
--- a/hotspot/src/share/vm/memory/universe.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/memory/universe.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -169,6 +169,8 @@
   static objArrayOop  _the_empty_system_obj_array;    // Canonicalized system obj array
   static objArrayOop  _the_empty_class_klass_array;   // Canonicalized obj array of type java.lang.Class
   static objArrayOop  _the_array_interfaces_array;    // Canonicalized 2-array of cloneable & serializable klasses
+  static oop          _the_null_string;               // A cache of "null" as a Java string
+  static oop          _the_min_jint_string;          // A cache of "-2147483648" as a Java string
   static LatestMethodOopCache* _finalizer_register_cache; // static method for registering finalizable objects
   static LatestMethodOopCache* _loader_addClass_cache;    // method for registering loaded classes in class loader vector
   static ActiveMethodOopsCache* _reflect_invoke_cache;    // method for security checks
@@ -310,6 +312,8 @@
   static objArrayOop  the_empty_system_obj_array ()   { return _the_empty_system_obj_array;    }
   static objArrayOop  the_empty_class_klass_array ()  { return _the_empty_class_klass_array;   }
   static objArrayOop  the_array_interfaces_array()    { return _the_array_interfaces_array;    }
+  static oop          the_null_string()               { return _the_null_string;               }
+  static oop          the_min_jint_string()          { return _the_min_jint_string;          }
   static methodOop    finalizer_register_method()     { return _finalizer_register_cache->get_methodOop(); }
   static methodOop    loader_addClass_method()        { return _loader_addClass_cache->get_methodOop(); }
   static ActiveMethodOopsCache* reflect_invoke_cache() { return _reflect_invoke_cache; }
--- a/hotspot/src/share/vm/oops/constantPoolOop.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/oops/constantPoolOop.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -262,25 +262,48 @@
 
 
 int constantPoolOopDesc::impl_name_and_type_ref_index_at(int which, bool uncached) {
-  jint ref_index = field_or_method_at(which, uncached);
+  int i = which;
+  if (!uncached && cache() != NULL) {
+    if (constantPoolCacheOopDesc::is_secondary_index(which))
+      // Invokedynamic indexes are always processed in native order
+      // so there is no question of reading a native u2 in Java order here.
+      return cache()->main_entry_at(which)->constant_pool_index();
+    // change byte-ordering and go via cache
+    i = remap_instruction_operand_from_cache(which);
+  } else {
+    if (tag_at(which).is_name_and_type())
+      // invokedynamic index is a simple name-and-type
+      return which;
+  }
+  assert(tag_at(i).is_field_or_method(), "Corrupted constant pool");
+  jint ref_index = *int_at_addr(i);
   return extract_high_short_from_int(ref_index);
 }
 
 
 int constantPoolOopDesc::impl_klass_ref_index_at(int which, bool uncached) {
-  jint ref_index = field_or_method_at(which, uncached);
+  guarantee(!constantPoolCacheOopDesc::is_secondary_index(which),
+            "an invokedynamic instruction does not have a klass");
+  int i = which;
+  if (!uncached && cache() != NULL) {
+    // change byte-ordering and go via cache
+    i = remap_instruction_operand_from_cache(which);
+  }
+  assert(tag_at(i).is_field_or_method(), "Corrupted constant pool");
+  jint ref_index = *int_at_addr(i);
   return extract_low_short_from_int(ref_index);
 }
 
 
 
-int constantPoolOopDesc::map_instruction_operand_to_index(int operand) {
-  if (constantPoolCacheOopDesc::is_secondary_index(operand)) {
-    return cache()->main_entry_at(operand)->constant_pool_index();
-  }
+int constantPoolOopDesc::remap_instruction_operand_from_cache(int operand) {
+  // Operand was fetched by a stream using get_Java_u2, yet was stored
+  // by Rewriter::rewrite_member_reference in native order.
+  // So now we have to fix the damage by swapping back to native order.
   assert((int)(u2)operand == operand, "clean u2");
-  int index = Bytes::swap_u2(operand);
-  return cache()->entry_at(index)->constant_pool_index();
+  int cpc_index = Bytes::swap_u2(operand);
+  int member_index = cache()->entry_at(cpc_index)->constant_pool_index();
+  return member_index;
 }
 
 
--- a/hotspot/src/share/vm/oops/constantPoolOop.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/oops/constantPoolOop.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -342,12 +342,14 @@
   }
 
   // The following methods (name/signature/klass_ref_at, klass_ref_at_noresolve,
-  // name_and_type_ref_index_at) all expect constant pool indices
-  // from the bytecodes to be passed in, which are actually potentially byte-swapped
-  // or rewritten constant pool cache indices.  They all call map_instruction_operand_to_index.
-  int map_instruction_operand_to_index(int operand);
+  // name_and_type_ref_index_at) all expect to be passed indices obtained
+  // directly from the bytecode, and extracted according to java byte order.
+  // If the indices are meant to refer to fields or methods, they are
+  // actually potentially byte-swapped, rewritten constant pool cache indices.
+  // The routine remap_instruction_operand_from_cache manages the adjustment
+  // of these values back to constant pool indices.
 
-  // There are also "uncached" versions which do not map the operand index; see below.
+  // There are also "uncached" versions which do not adjust the operand index; see below.
 
   // Lookup for entries consisting of (klass_index, name_and_type index)
   klassOop klass_ref_at(int which, TRAPS);
@@ -361,8 +363,6 @@
   // Lookup for entries consisting of (name_index, signature_index)
   int name_ref_index_at(int which_nt);            // ==  low-order jshort of name_and_type_at(which_nt)
   int signature_ref_index_at(int which_nt);       // == high-order jshort of name_and_type_at(which_nt)
-  symbolOop nt_name_ref_at(int which_nt)          { return symbol_at(name_ref_index_at(which_nt)); }
-  symbolOop nt_signature_ref_at(int which_nt)     { return symbol_at(signature_ref_index_at(which_nt)); }
 
   BasicType basic_type_for_signature_at(int which);
 
@@ -425,18 +425,7 @@
   int       impl_klass_ref_index_at(int which, bool uncached);
   int       impl_name_and_type_ref_index_at(int which, bool uncached);
 
-  // Takes either a constant pool cache index in possibly byte-swapped
-  // byte order (which comes from the bytecodes after rewriting) or,
-  // if "uncached" is true, a vanilla constant pool index
-  jint field_or_method_at(int which, bool uncached) {
-    int i = which;
-    if (!uncached && cache() != NULL) {
-      // change byte-ordering and go via cache
-      i = map_instruction_operand_to_index(which);
-    }
-    assert(tag_at(i).is_field_or_method(), "Corrupted constant pool");
-    return *int_at_addr(i);
-  }
+  int remap_instruction_operand_from_cache(int operand);
 
   // Used while constructing constant pool (only by ClassFileParser)
   jint klass_index_at(int which) {
--- a/hotspot/src/share/vm/oops/cpCacheOop.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/oops/cpCacheOop.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -28,21 +28,17 @@
 
 // Implememtation of ConstantPoolCacheEntry
 
-void ConstantPoolCacheEntry::set_initial_state(int index) {
-  if (constantPoolCacheOopDesc::is_secondary_index(index)) {
-    // Hack:  The rewriter is trying to say that this entry itself
-    // will be a secondary entry.
-    int main_index = constantPoolCacheOopDesc::decode_secondary_index(index);
-    assert(0 <= main_index && main_index < 0x10000, "sanity check");
-    _indices = (main_index << 16);
-    assert(main_entry_index() == main_index, "");
-    return;
-  }
+void ConstantPoolCacheEntry::initialize_entry(int index) {
   assert(0 < index && index < 0x10000, "sanity check");
   _indices = index;
   assert(constant_pool_index() == index, "");
 }
 
+void ConstantPoolCacheEntry::initialize_secondary_entry(int main_index) {
+  assert(0 <= main_index && main_index < 0x10000, "sanity check");
+  _indices = (main_index << 16);
+  assert(main_entry_index() == main_index, "");
+}
 
 int ConstantPoolCacheEntry::as_flags(TosState state, bool is_final,
                     bool is_vfinal, bool is_volatile,
@@ -223,10 +219,10 @@
 
 
 void ConstantPoolCacheEntry::set_dynamic_call(Handle call_site, int extra_data) {
-  methodOop method = (methodOop) sun_dyn_CallSiteImpl::vmmethod(call_site());
+  methodOop method = (methodOop) java_dyn_CallSite::vmmethod(call_site());
   assert(method->is_method(), "must be initialized properly");
   int param_size = method->size_of_parameters();
-  assert(param_size > 1, "method argument size must include MH.this & initial dynamic receiver");
+  assert(param_size >= 1, "method argument size must include MH.this");
   param_size -= 1;              // do not count MH.this; it is not stacked for invokedynamic
   if (Atomic::cmpxchg_ptr(call_site(), &_f1, NULL) == NULL) {
     // racing threads might be trying to install their own favorites
@@ -439,7 +435,18 @@
 
 void constantPoolCacheOopDesc::initialize(intArray& inverse_index_map) {
   assert(inverse_index_map.length() == length(), "inverse index map must have same length as cache");
-  for (int i = 0; i < length(); i++) entry_at(i)->set_initial_state(inverse_index_map[i]);
+  for (int i = 0; i < length(); i++) {
+    ConstantPoolCacheEntry* e = entry_at(i);
+    int original_index = inverse_index_map[i];
+    if ((original_index & Rewriter::_secondary_entry_tag) != 0) {
+      int main_index = (original_index - Rewriter::_secondary_entry_tag);
+      assert(!entry_at(main_index)->is_secondary_entry(), "valid main index");
+      e->initialize_secondary_entry(main_index);
+    } else {
+      e->initialize_entry(original_index);
+    }
+    assert(entry_at(i) == e, "sanity");
+  }
 }
 
 // RedefineClasses() API support:
--- a/hotspot/src/share/vm/oops/cpCacheOop.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/oops/cpCacheOop.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -154,7 +154,8 @@
   };
 
   // Initialization
-  void set_initial_state(int index);             // sets entry to initial state
+  void initialize_entry(int original_index);     // initialize primary entry
+  void initialize_secondary_entry(int main_index); // initialize secondary entry
 
   void set_field(                                // sets entry to resolved field state
     Bytecodes::Code get_code,                    // the bytecode used for reading the field
@@ -251,6 +252,7 @@
 
   // Code generation support
   static WordSize size()                         { return in_WordSize(sizeof(ConstantPoolCacheEntry) / HeapWordSize); }
+  static ByteSize size_in_bytes()                { return in_ByteSize(sizeof(ConstantPoolCacheEntry)); }
   static ByteSize indices_offset()               { return byte_offset_of(ConstantPoolCacheEntry, _indices); }
   static ByteSize f1_offset()                    { return byte_offset_of(ConstantPoolCacheEntry, _f1); }
   static ByteSize f2_offset()                    { return byte_offset_of(ConstantPoolCacheEntry, _f2); }
@@ -321,6 +323,7 @@
   ConstantPoolCacheEntry* base() const           { return (ConstantPoolCacheEntry*)((address)this + in_bytes(base_offset())); }
 
   friend class constantPoolCacheKlass;
+  friend class ConstantPoolCacheEntry;
 
  public:
   // Initialization
@@ -329,7 +332,8 @@
   // Secondary indexes.
   // They must look completely different from normal indexes.
   // The main reason is that byte swapping is sometimes done on normal indexes.
-  // Also, it is helpful for debugging to tell the two apart.
+  // Also, some of the CP accessors do different things for secondary indexes.
+  // Finally, it is helpful for debugging to tell the two apart.
   static bool is_secondary_index(int i) { return (i < 0); }
   static int  decode_secondary_index(int i) { assert(is_secondary_index(i),  ""); return ~i; }
   static int  encode_secondary_index(int i) { assert(!is_secondary_index(i), ""); return ~i; }
@@ -337,18 +341,35 @@
   // Accessors
   void set_constant_pool(constantPoolOop pool)   { oop_store_without_check((oop*)&_constant_pool, (oop)pool); }
   constantPoolOop constant_pool() const          { return _constant_pool; }
-  ConstantPoolCacheEntry* entry_at(int i) const  { assert(0 <= i && i < length(), "index out of bounds"); return base() + i; }
+  // Fetches the entry at the given index.
+  // The entry may be either primary or secondary.
+  // In either case the index must not be encoded or byte-swapped in any way.
+  ConstantPoolCacheEntry* entry_at(int i) const {
+    assert(0 <= i && i < length(), "index out of bounds");
+    return base() + i;
+  }
+  // Fetches the secondary entry referred to by index.
+  // The index may be a secondary index, and must not be byte-swapped.
+  ConstantPoolCacheEntry* secondary_entry_at(int i) const {
+    int raw_index = i;
+    if (is_secondary_index(i)) {  // correct these on the fly
+      raw_index = decode_secondary_index(i);
+    }
+    assert(entry_at(raw_index)->is_secondary_entry(), "not a secondary entry");
+    return entry_at(raw_index);
+  }
+  // Given a primary or secondary index, fetch the corresponding primary entry.
+  // Indirect through the secondary entry, if the index is encoded as a secondary index.
+  // The index must not be byte-swapped.
   ConstantPoolCacheEntry* main_entry_at(int i) const {
-    ConstantPoolCacheEntry* e;
+    int primary_index = i;
     if (is_secondary_index(i)) {
       // run through an extra level of indirection:
-      i = decode_secondary_index(i);
-      e = entry_at(i);
-      i = e->main_entry_index();
+      int raw_index = decode_secondary_index(i);
+      primary_index = entry_at(raw_index)->main_entry_index();
     }
-    e = entry_at(i);
-    assert(!e->is_secondary_entry(), "only one level of indirection");
-    return e;
+    assert(!entry_at(primary_index)->is_secondary_entry(), "only one level of indirection");
+    return entry_at(primary_index);
   }
 
   // GC support
@@ -359,6 +380,12 @@
 
   // Code generation
   static ByteSize base_offset()                  { return in_ByteSize(sizeof(constantPoolCacheOopDesc)); }
+  static ByteSize entry_offset(int raw_index) {
+    int index = raw_index;
+    if (is_secondary_index(raw_index))
+      index = decode_secondary_index(raw_index);
+    return (base_offset() + ConstantPoolCacheEntry::size_in_bytes() * index);
+  }
 
   // RedefineClasses() API support:
   // If any entry of this constantPoolCache points to any of
--- a/hotspot/src/share/vm/oops/generateOopMap.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/oops/generateOopMap.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -1556,13 +1556,13 @@
     case Bytecodes::_getfield:          do_field(true,  false, itr->get_index_big(), itr->bci()); break;
     case Bytecodes::_putfield:          do_field(false, false, itr->get_index_big(), itr->bci()); break;
 
-    case Bytecodes::_invokevirtual:
-    case Bytecodes::_invokespecial:     do_method(false, false, itr->get_index_big(), itr->bci()); break;
-    case Bytecodes::_invokestatic:      do_method(true,  false, itr->get_index_big(), itr->bci()); break;
-    case Bytecodes::_invokedynamic:     do_method(false, true,  itr->get_index_int(), itr->bci()); break;
-    case Bytecodes::_invokeinterface:   do_method(false, true,  itr->get_index_big(), itr->bci()); break;
-    case Bytecodes::_newarray:
-    case Bytecodes::_anewarray:         pp_new_ref(vCTS, itr->bci()); break;
+   case Bytecodes::_invokevirtual:
+   case Bytecodes::_invokespecial:     do_method(false, false, itr->get_index_big(), itr->bci()); break;
+   case Bytecodes::_invokestatic:      do_method(true,  false, itr->get_index_big(), itr->bci()); break;
+    case Bytecodes::_invokedynamic:     do_method(true,  false, itr->get_index_int(), itr->bci()); break;
+   case Bytecodes::_invokeinterface:   do_method(false, true,  itr->get_index_big(), itr->bci()); break;
+   case Bytecodes::_newarray:
+   case Bytecodes::_anewarray:         pp_new_ref(vCTS, itr->bci()); break;
     case Bytecodes::_checkcast:         do_checkcast(); break;
     case Bytecodes::_arraylength:
     case Bytecodes::_instanceof:        pp(rCTS, vCTS); break;
@@ -1900,11 +1900,9 @@
 }
 
 void GenerateOopMap::do_method(int is_static, int is_interface, int idx, int bci) {
-  // Dig up signature for field in constant pool
-  constantPoolOop cp    = _method->constants();
-  int nameAndTypeIdx    = cp->name_and_type_ref_index_at(idx);
-  int signatureIdx      = cp->signature_ref_index_at(nameAndTypeIdx);  // @@@@@
-  symbolOop signature   = cp->symbol_at(signatureIdx);
+ // Dig up signature for field in constant pool
+  constantPoolOop cp  = _method->constants();
+  symbolOop signature = cp->signature_ref_at(idx);
 
   // Parse method signature
   CellTypeState out[4];
--- a/hotspot/src/share/vm/oops/instanceKlass.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/oops/instanceKlass.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -2045,8 +2045,9 @@
     // As we walk along, look for equalities between outer1 and class2.
     // Eventually, the walks will terminate as outer1 stops
     // at the top-level class around the original class.
-    symbolOop ignore_name;
-    klassOop next = outer1->compute_enclosing_class(ignore_name, CHECK_false);
+    bool ignore_inner_is_member;
+    klassOop next = outer1->compute_enclosing_class(&ignore_inner_is_member,
+                                                    CHECK_false);
     if (next == NULL)  break;
     if (next == class2())  return true;
     outer1 = instanceKlassHandle(THREAD, next);
@@ -2055,8 +2056,9 @@
   // Now do the same for class2.
   instanceKlassHandle outer2 = class2;
   for (;;) {
-    symbolOop ignore_name;
-    klassOop next = outer2->compute_enclosing_class(ignore_name, CHECK_false);
+    bool ignore_inner_is_member;
+    klassOop next = outer2->compute_enclosing_class(&ignore_inner_is_member,
+                                                    CHECK_false);
     if (next == NULL)  break;
     // Might as well check the new outer against all available values.
     if (next == class1())  return true;
--- a/hotspot/src/share/vm/oops/instanceKlass.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/oops/instanceKlass.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -337,12 +337,12 @@
   static bool is_same_class_package(oop class_loader1, symbolOop class_name1, oop class_loader2, symbolOop class_name2);
 
   // find an enclosing class (defined where original code was, in jvm.cpp!)
-  klassOop compute_enclosing_class(symbolOop& simple_name_result, TRAPS) {
+  klassOop compute_enclosing_class(bool* inner_is_member, TRAPS) {
     instanceKlassHandle self(THREAD, this->as_klassOop());
-    return compute_enclosing_class_impl(self, simple_name_result, THREAD);
+    return compute_enclosing_class_impl(self, inner_is_member, THREAD);
   }
   static klassOop compute_enclosing_class_impl(instanceKlassHandle self,
-                                               symbolOop& simple_name_result, TRAPS);
+                                               bool* inner_is_member, TRAPS);
 
   // tell if two classes have the same enclosing class (at package level)
   bool is_same_package_member(klassOop class2, TRAPS) {
--- a/hotspot/src/share/vm/oops/instanceKlassKlass.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/oops/instanceKlassKlass.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -317,6 +317,11 @@
     pm->claim_or_forward_breadth(sg_addr);
   }
 
+  oop* bsm_addr = ik->adr_bootstrap_method();
+  if (PSScavenge::should_scavenge(bsm_addr)) {
+    pm->claim_or_forward_breadth(bsm_addr);
+  }
+
   klassKlass::oop_copy_contents(pm, obj);
 }
 
@@ -345,6 +350,11 @@
     pm->claim_or_forward_depth(sg_addr);
   }
 
+  oop* bsm_addr = ik->adr_bootstrap_method();
+  if (PSScavenge::should_scavenge(bsm_addr)) {
+    pm->claim_or_forward_depth(bsm_addr);
+  }
+
   klassKlass::oop_copy_contents(pm, obj);
 }
 
--- a/hotspot/src/share/vm/oops/instanceRefKlass.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/oops/instanceRefKlass.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -78,9 +78,9 @@
 
 #ifndef SERIALGC
 template <class T>
-static void specialized_oop_follow_contents(instanceRefKlass* ref,
-                                            ParCompactionManager* cm,
-                                            oop obj) {
+void specialized_oop_follow_contents(instanceRefKlass* ref,
+                                     ParCompactionManager* cm,
+                                     oop obj) {
   T* referent_addr = (T*)java_lang_ref_Reference::referent_addr(obj);
   T heap_oop = oopDesc::load_heap_oop(referent_addr);
   debug_only(
--- a/hotspot/src/share/vm/oops/objArrayKlass.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/oops/objArrayKlass.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -127,16 +127,14 @@
           // pointer delta is scaled to number of elements (length field in
           // objArrayOop) which we assume is 32 bit.
           assert(pd == (size_t)(int)pd, "length field overflow");
-          const size_t done_word_len = objArrayOopDesc::array_size((int)pd);
-          bs->write_ref_array(MemRegion((HeapWord*)dst, done_word_len));
+          bs->write_ref_array((HeapWord*)dst, pd);
           THROW(vmSymbols::java_lang_ArrayStoreException());
           return;
         }
       }
     }
   }
-  const size_t word_len = objArrayOopDesc::array_size(length);
-  bs->write_ref_array(MemRegion((HeapWord*)dst, word_len));
+  bs->write_ref_array((HeapWord*)dst, length);
 }
 
 void objArrayKlass::copy_array(arrayOop s, int src_pos, arrayOop d,
--- a/hotspot/src/share/vm/oops/objArrayOop.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/oops/objArrayOop.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -37,6 +37,32 @@
     return &((T*)base())[index];
   }
 
+private:
+  // Give size of objArrayOop in HeapWords minus the header
+  static int array_size(int length) {
+    const int OopsPerHeapWord = HeapWordSize/heapOopSize;
+    assert(OopsPerHeapWord >= 1 && (HeapWordSize % heapOopSize == 0),
+           "Else the following (new) computation would be in error");
+#ifdef ASSERT
+    // The old code is left in for sanity-checking; it'll
+    // go away pretty soon. XXX
+    // Without UseCompressedOops, this is simply:
+    // oop->length() * HeapWordsPerOop;
+    // With narrowOops, HeapWordsPerOop is 1/2 or equal 0 as an integer.
+    // The oop elements are aligned up to wordSize
+    const int HeapWordsPerOop = heapOopSize/HeapWordSize;
+    int old_res;
+    if (HeapWordsPerOop > 0) {
+      old_res = length * HeapWordsPerOop;
+    } else {
+      old_res = align_size_up(length, OopsPerHeapWord)/OopsPerHeapWord;
+    }
+#endif  // ASSERT
+    int res = ((uint)length + OopsPerHeapWord - 1)/OopsPerHeapWord;
+    assert(res == old_res, "Inconsistency between old and new.");
+    return res;
+  }
+
  public:
   // Returns the offset of the first element.
   static int base_offset_in_bytes() {
@@ -67,27 +93,14 @@
   // Sizing
   static int header_size()    { return arrayOopDesc::header_size(T_OBJECT); }
   int object_size()           { return object_size(length()); }
-  int array_size()            { return array_size(length()); }
 
   static int object_size(int length) {
     // This returns the object size in HeapWords.
-    return align_object_size(header_size() + array_size(length));
-  }
-
-  // Give size of objArrayOop in HeapWords minus the header
-  static int array_size(int length) {
-    // Without UseCompressedOops, this is simply:
-    // oop->length() * HeapWordsPerOop;
-    // With narrowOops, HeapWordsPerOop is 1/2 or equal 0 as an integer.
-    // The oop elements are aligned up to wordSize
-    const int HeapWordsPerOop = heapOopSize/HeapWordSize;
-    if (HeapWordsPerOop > 0) {
-      return length * HeapWordsPerOop;
-    } else {
-      const int OopsPerHeapWord = HeapWordSize/heapOopSize;
-      int word_len = align_size_up(length, OopsPerHeapWord)/OopsPerHeapWord;
-      return word_len;
-    }
+    uint asz = array_size(length);
+    uint osz = align_object_size(header_size() + asz);
+    assert(osz >= asz,   "no overflow");
+    assert((int)osz > 0, "no overflow");
+    return (int)osz;
   }
 
   // special iterators for index ranges, returns size of object
--- a/hotspot/src/share/vm/oops/oop.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/oops/oop.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -30,13 +30,12 @@
 // no virtual functions allowed
 
 // store into oop with store check
-template <class T> void oop_store(T* p, oop v);
-template <class T> void oop_store(volatile T* p, oop v);
+template <class T> inline void oop_store(T* p, oop v);
+template <class T> inline void oop_store(volatile T* p, oop v);
 
 // store into oop without store check
-template <class T> void oop_store_without_check(T* p, oop v);
-template <class T> void oop_store_without_check(volatile T* p, oop v);
-
+template <class T> inline void oop_store_without_check(T* p, oop v);
+template <class T> inline void oop_store_without_check(volatile T* p, oop v);
 
 extern bool always_do_update_barrier;
 
--- a/hotspot/src/share/vm/opto/c2_globals.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/opto/c2_globals.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -25,4 +25,4 @@
 # include "incls/_precompiled.incl"
 # include "incls/_c2_globals.cpp.incl"
 
-C2_FLAGS(MATERIALIZE_DEVELOPER_FLAG, MATERIALIZE_PD_DEVELOPER_FLAG, MATERIALIZE_PRODUCT_FLAG, MATERIALIZE_PD_PRODUCT_FLAG, MATERIALIZE_DIAGNOSTIC_FLAG, MATERIALIZE_NOTPRODUCT_FLAG)
+C2_FLAGS(MATERIALIZE_DEVELOPER_FLAG, MATERIALIZE_PD_DEVELOPER_FLAG, MATERIALIZE_PRODUCT_FLAG, MATERIALIZE_PD_PRODUCT_FLAG, MATERIALIZE_DIAGNOSTIC_FLAG, MATERIALIZE_EXPERIMENTAL_FLAG, MATERIALIZE_NOTPRODUCT_FLAG)
--- a/hotspot/src/share/vm/opto/c2_globals.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/opto/c2_globals.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -26,7 +26,7 @@
 // Defines all globals flags used by the server compiler.
 //
 
-#define C2_FLAGS(develop, develop_pd, product, product_pd, diagnostic, notproduct) \
+#define C2_FLAGS(develop, develop_pd, product, product_pd, diagnostic, experimental, notproduct) \
                                                                             \
   notproduct(intx, CompileZapFirst, 0,                                      \
           "If +ZapDeadCompiledLocals, "                                     \
@@ -394,6 +394,12 @@
   product(bool, UseOptoBiasInlining, true,                                  \
           "Generate biased locking code in C2 ideal graph")                 \
                                                                             \
+  product(bool, OptimizeStringConcat, false,                                \
+          "Optimize the construction of Strings by StringBuilder")          \
+                                                                            \
+  notproduct(bool, PrintOptimizeStringConcat, false,                        \
+          "Print information about transformations performed on Strings")   \
+                                                                            \
   product(intx, ValueSearchLimit, 1000,                                     \
           "Recursion limit in PhaseMacroExpand::value_from_mem_phi")        \
                                                                             \
@@ -413,4 +419,4 @@
   product(bool, BlockLayoutRotateLoops, true,                               \
           "Allow back branches to be fall throughs in the block layour")    \
 
-C2_FLAGS(DECLARE_DEVELOPER_FLAG, DECLARE_PD_DEVELOPER_FLAG, DECLARE_PRODUCT_FLAG, DECLARE_PD_PRODUCT_FLAG, DECLARE_DIAGNOSTIC_FLAG, DECLARE_NOTPRODUCT_FLAG)
+C2_FLAGS(DECLARE_DEVELOPER_FLAG, DECLARE_PD_DEVELOPER_FLAG, DECLARE_PRODUCT_FLAG, DECLARE_PD_PRODUCT_FLAG, DECLARE_DIAGNOSTIC_FLAG, DECLARE_EXPERIMENTAL_FLAG, DECLARE_NOTPRODUCT_FLAG)
--- a/hotspot/src/share/vm/opto/callGenerator.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/opto/callGenerator.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -98,12 +98,21 @@
 //---------------------------DirectCallGenerator------------------------------
 // Internal class which handles all out-of-line calls w/o receiver type checks.
 class DirectCallGenerator : public CallGenerator {
-public:
-  DirectCallGenerator(ciMethod* method)
-    : CallGenerator(method)
+ private:
+  CallStaticJavaNode* _call_node;
+  // Force separate memory and I/O projections for the exceptional
+  // paths to facilitate late inlinig.
+  bool                _separate_io_proj;
+
+ public:
+  DirectCallGenerator(ciMethod* method, bool separate_io_proj)
+    : CallGenerator(method),
+      _separate_io_proj(separate_io_proj)
   {
   }
   virtual JVMState* generate(JVMState* jvms);
+
+  CallStaticJavaNode* call_node() const { return _call_node; }
 };
 
 JVMState* DirectCallGenerator::generate(JVMState* jvms) {
@@ -129,9 +138,10 @@
     call->set_optimized_virtual(true);
   }
   kit.set_arguments_for_java_call(call);
-  kit.set_edges_for_java_call(call);
-  Node* ret = kit.set_results_for_java_call(call);
+  kit.set_edges_for_java_call(call, false, _separate_io_proj);
+  Node* ret = kit.set_results_for_java_call(call, _separate_io_proj);
   kit.push_node(method()->return_type()->basic_type(), ret);
+  _call_node = call;  // Save the call node in case we need it later
   return kit.transfer_exceptions_into_jvms();
 }
 
@@ -238,9 +248,9 @@
   return new ParseGenerator(m, expected_uses, true);
 }
 
-CallGenerator* CallGenerator::for_direct_call(ciMethod* m) {
+CallGenerator* CallGenerator::for_direct_call(ciMethod* m, bool separate_io_proj) {
   assert(!m->is_abstract(), "for_direct_call mismatch");
-  return new DirectCallGenerator(m);
+  return new DirectCallGenerator(m, separate_io_proj);
 }
 
 CallGenerator* CallGenerator::for_virtual_call(ciMethod* m, int vtable_index) {
@@ -248,6 +258,108 @@
   return new VirtualCallGenerator(m, vtable_index);
 }
 
+// Allow inlining decisions to be delayed
+class LateInlineCallGenerator : public DirectCallGenerator {
+  CallGenerator* _inline_cg;
+
+ public:
+  LateInlineCallGenerator(ciMethod* method, CallGenerator* inline_cg) :
+    DirectCallGenerator(method, true), _inline_cg(inline_cg) {}
+
+  virtual bool      is_late_inline() const { return true; }
+
+  // Convert the CallStaticJava into an inline
+  virtual void do_late_inline();
+
+  JVMState* generate(JVMState* jvms) {
+    // Record that this call site should be revisited once the main
+    // parse is finished.
+    Compile::current()->add_late_inline(this);
+
+    // Emit the CallStaticJava and request separate projections so
+    // that the late inlining logic can distinguish between fall
+    // through and exceptional uses of the memory and io projections
+    // as is done for allocations and macro expansion.
+    return DirectCallGenerator::generate(jvms);
+  }
+
+};
+
+
+void LateInlineCallGenerator::do_late_inline() {
+  // Can't inline it
+  if (call_node() == NULL || call_node()->outcnt() == 0 ||
+      call_node()->in(0) == NULL || call_node()->in(0)->is_top())
+    return;
+
+  CallStaticJavaNode* call = call_node();
+
+  // Make a clone of the JVMState that appropriate to use for driving a parse
+  Compile* C = Compile::current();
+  JVMState* jvms     = call->jvms()->clone_shallow(C);
+  uint size = call->req();
+  SafePointNode* map = new (C, size) SafePointNode(size, jvms);
+  for (uint i1 = 0; i1 < size; i1++) {
+    map->init_req(i1, call->in(i1));
+  }
+
+  // Make sure the state is a MergeMem for parsing.
+  if (!map->in(TypeFunc::Memory)->is_MergeMem()) {
+    map->set_req(TypeFunc::Memory, MergeMemNode::make(C, map->in(TypeFunc::Memory)));
+  }
+
+  // Make enough space for the expression stack and transfer the incoming arguments
+  int nargs    = method()->arg_size();
+  jvms->set_map(map);
+  map->ensure_stack(jvms, jvms->method()->max_stack());
+  if (nargs > 0) {
+    for (int i1 = 0; i1 < nargs; i1++) {
+      map->set_req(i1 + jvms->argoff(), call->in(TypeFunc::Parms + i1));
+    }
+  }
+
+  CompileLog* log = C->log();
+  if (log != NULL) {
+    log->head("late_inline method='%d'", log->identify(method()));
+    JVMState* p = jvms;
+    while (p != NULL) {
+      log->elem("jvms bci='%d' method='%d'", p->bci(), log->identify(p->method()));
+      p = p->caller();
+    }
+    log->tail("late_inline");
+  }
+
+  // Setup default node notes to be picked up by the inlining
+  Node_Notes* old_nn = C->default_node_notes();
+  if (old_nn != NULL) {
+    Node_Notes* entry_nn = old_nn->clone(C);
+    entry_nn->set_jvms(jvms);
+    C->set_default_node_notes(entry_nn);
+  }
+
+  // Now perform the inling using the synthesized JVMState
+  JVMState* new_jvms = _inline_cg->generate(jvms);
+  if (new_jvms == NULL)  return;  // no change
+  if (C->failing())      return;
+
+  // Capture any exceptional control flow
+  GraphKit kit(new_jvms);
+
+  // Find the result object
+  Node* result = C->top();
+  int   result_size = method()->return_type()->size();
+  if (result_size != 0 && !kit.stopped()) {
+    result = (result_size == 1) ? kit.pop() : kit.pop_pair();
+  }
+
+  kit.replace_call(call, result);
+}
+
+
+CallGenerator* CallGenerator::for_late_inline(ciMethod* method, CallGenerator* inline_cg) {
+  return new LateInlineCallGenerator(method, inline_cg);
+}
+
 
 //---------------------------WarmCallGenerator--------------------------------
 // Internal class which handles initial deferral of inlining decisions.
@@ -315,70 +427,7 @@
 }
 
 void WarmCallInfo::make_hot() {
-  Compile* C = Compile::current();
-  // Replace the callnode with something better.
-  CallJavaNode* call = this->call()->as_CallJava();
-  ciMethod* method   = call->method();
-  int       nargs    = method->arg_size();
-  JVMState* jvms     = call->jvms()->clone_shallow(C);
-  uint size = TypeFunc::Parms + MAX2(2, nargs);
-  SafePointNode* map = new (C, size) SafePointNode(size, jvms);
-  for (uint i1 = 0; i1 < (uint)(TypeFunc::Parms + nargs); i1++) {
-    map->init_req(i1, call->in(i1));
-  }
-  jvms->set_map(map);
-  jvms->set_offsets(map->req());
-  jvms->set_locoff(TypeFunc::Parms);
-  jvms->set_stkoff(TypeFunc::Parms);
-  GraphKit kit(jvms);
-
-  JVMState* new_jvms = _hot_cg->generate(kit.jvms());
-  if (new_jvms == NULL)  return;  // no change
-  if (C->failing())      return;
-
-  kit.set_jvms(new_jvms);
-  Node* res = C->top();
-  int   res_size = method->return_type()->size();
-  if (res_size != 0) {
-    kit.inc_sp(-res_size);
-    res = kit.argument(0);
-  }
-  GraphKit ekit(kit.combine_and_pop_all_exception_states()->jvms());
-
-  // Replace the call:
-  for (DUIterator i = call->outs(); call->has_out(i); i++) {
-    Node* n = call->out(i);
-    Node* nn = NULL;  // replacement
-    if (n->is_Proj()) {
-      ProjNode* nproj = n->as_Proj();
-      assert(nproj->_con < (uint)(TypeFunc::Parms + (res_size ? 1 : 0)), "sane proj");
-      if (nproj->_con == TypeFunc::Parms) {
-        nn = res;
-      } else {
-        nn = kit.map()->in(nproj->_con);
-      }
-      if (nproj->_con == TypeFunc::I_O) {
-        for (DUIterator j = nproj->outs(); nproj->has_out(j); j++) {
-          Node* e = nproj->out(j);
-          if (e->Opcode() == Op_CreateEx) {
-            e->replace_by(ekit.argument(0));
-          } else if (e->Opcode() == Op_Catch) {
-            for (DUIterator k = e->outs(); e->has_out(k); k++) {
-              CatchProjNode* p = e->out(j)->as_CatchProj();
-              if (p->is_handler_proj()) {
-                p->replace_by(ekit.control());
-              } else {
-                p->replace_by(kit.control());
-              }
-            }
-          }
-        }
-      }
-    }
-    NOT_PRODUCT(if (!nn)  n->dump(2));
-    assert(nn != NULL, "don't know what to do with this user");
-    n->replace_by(nn);
-  }
+  Unimplemented();
 }
 
 void WarmCallInfo::make_cold() {
--- a/hotspot/src/share/vm/opto/callGenerator.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/opto/callGenerator.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -57,6 +57,13 @@
   // is_trap: Does not return to the caller.  (E.g., uncommon trap.)
   virtual bool      is_trap() const             { return false; }
 
+  // is_late_inline: supports conversion of call into an inline
+  virtual bool      is_late_inline() const      { return false; }
+  // Replace the call with an inline version of the code
+  virtual void do_late_inline() { ShouldNotReachHere(); }
+
+  virtual CallStaticJavaNode* call_node() const { ShouldNotReachHere(); return NULL; }
+
   // Note:  It is possible for a CG to be both inline and virtual.
   // (The hashCode intrinsic does a vtable check and an inlined fast path.)
 
@@ -92,9 +99,12 @@
   static CallGenerator* for_osr(ciMethod* m, int osr_bci);
 
   // How to generate vanilla out-of-line call sites:
-  static CallGenerator* for_direct_call(ciMethod* m);   // static, special
+  static CallGenerator* for_direct_call(ciMethod* m, bool separate_io_projs = false);   // static, special
   static CallGenerator* for_virtual_call(ciMethod* m, int vtable_index);  // virtual, interface
 
+  // How to generate a replace a direct call with an inline version
+  static CallGenerator* for_late_inline(ciMethod* m, CallGenerator* inline_cg);
+
   // How to make a call but defer the decision whether to inline or not.
   static CallGenerator* for_warm_call(WarmCallInfo* ci,
                                       CallGenerator* if_cold,
--- a/hotspot/src/share/vm/opto/callnode.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/opto/callnode.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -693,6 +693,84 @@
 }
 
 
+void CallNode::extract_projections(CallProjections* projs, bool separate_io_proj) {
+  projs->fallthrough_proj      = NULL;
+  projs->fallthrough_catchproj = NULL;
+  projs->fallthrough_ioproj    = NULL;
+  projs->catchall_ioproj       = NULL;
+  projs->catchall_catchproj    = NULL;
+  projs->fallthrough_memproj   = NULL;
+  projs->catchall_memproj      = NULL;
+  projs->resproj               = NULL;
+  projs->exobj                 = NULL;
+
+  for (DUIterator_Fast imax, i = fast_outs(imax); i < imax; i++) {
+    ProjNode *pn = fast_out(i)->as_Proj();
+    if (pn->outcnt() == 0) continue;
+    switch (pn->_con) {
+    case TypeFunc::Control:
+      {
+        // For Control (fallthrough) and I_O (catch_all_index) we have CatchProj -> Catch -> Proj
+        projs->fallthrough_proj = pn;
+        DUIterator_Fast jmax, j = pn->fast_outs(jmax);
+        const Node *cn = pn->fast_out(j);
+        if (cn->is_Catch()) {
+          ProjNode *cpn = NULL;
+          for (DUIterator_Fast kmax, k = cn->fast_outs(kmax); k < kmax; k++) {
+            cpn = cn->fast_out(k)->as_Proj();
+            assert(cpn->is_CatchProj(), "must be a CatchProjNode");
+            if (cpn->_con == CatchProjNode::fall_through_index)
+              projs->fallthrough_catchproj = cpn;
+            else {
+              assert(cpn->_con == CatchProjNode::catch_all_index, "must be correct index.");
+              projs->catchall_catchproj = cpn;
+            }
+          }
+        }
+        break;
+      }
+    case TypeFunc::I_O:
+      if (pn->_is_io_use)
+        projs->catchall_ioproj = pn;
+      else
+        projs->fallthrough_ioproj = pn;
+      for (DUIterator j = pn->outs(); pn->has_out(j); j++) {
+        Node* e = pn->out(j);
+        if (e->Opcode() == Op_CreateEx && e->in(0)->is_CatchProj()) {
+          assert(projs->exobj == NULL, "only one");
+          projs->exobj = e;
+        }
+      }
+      break;
+    case TypeFunc::Memory:
+      if (pn->_is_io_use)
+        projs->catchall_memproj = pn;
+      else
+        projs->fallthrough_memproj = pn;
+      break;
+    case TypeFunc::Parms:
+      projs->resproj = pn;
+      break;
+    default:
+      assert(false, "unexpected projection from allocation node.");
+    }
+  }
+
+  // The resproj may not exist because the result couuld be ignored
+  // and the exception object may not exist if an exception handler
+  // swallows the exception but all the other must exist and be found.
+  assert(projs->fallthrough_proj      != NULL, "must be found");
+  assert(projs->fallthrough_catchproj != NULL, "must be found");
+  assert(projs->fallthrough_memproj   != NULL, "must be found");
+  assert(projs->fallthrough_ioproj    != NULL, "must be found");
+  assert(projs->catchall_catchproj    != NULL, "must be found");
+  if (separate_io_proj) {
+    assert(projs->catchall_memproj      != NULL, "must be found");
+    assert(projs->catchall_ioproj       != NULL, "must be found");
+  }
+}
+
+
 //=============================================================================
 uint CallJavaNode::size_of() const { return sizeof(*this); }
 uint CallJavaNode::cmp( const Node &n ) const {
--- a/hotspot/src/share/vm/opto/callnode.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/opto/callnode.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -470,6 +470,23 @@
 #endif
 };
 
+
+// Simple container for the outgoing projections of a call.  Useful
+// for serious surgery on calls.
+class CallProjections : public StackObj {
+public:
+  Node* fallthrough_proj;
+  Node* fallthrough_catchproj;
+  Node* fallthrough_memproj;
+  Node* fallthrough_ioproj;
+  Node* catchall_catchproj;
+  Node* catchall_memproj;
+  Node* catchall_ioproj;
+  Node* resproj;
+  Node* exobj;
+};
+
+
 //------------------------------CallNode---------------------------------------
 // Call nodes now subsume the function of debug nodes at callsites, so they
 // contain the functionality of a full scope chain of debug nodes.
@@ -521,6 +538,11 @@
   // or returns NULL if there is no one.
   Node *result_cast();
 
+  // Collect all the interesting edges from a call for use in
+  // replacing the call by something else.  Used by macro expansion
+  // and the late inlining support.
+  void extract_projections(CallProjections* projs, bool separate_io_proj);
+
   virtual uint match_edge(uint idx) const;
 
 #ifndef PRODUCT
@@ -529,6 +551,7 @@
 #endif
 };
 
+
 //------------------------------CallJavaNode-----------------------------------
 // Make a static or dynamic subroutine call node using Java calling
 // convention.  (The "Java" calling convention is the compiler's calling
--- a/hotspot/src/share/vm/opto/compile.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/opto/compile.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -224,6 +224,32 @@
 }
 
 
+void Compile::gvn_replace_by(Node* n, Node* nn) {
+  for (DUIterator_Last imin, i = n->last_outs(imin); i >= imin; ) {
+    Node* use = n->last_out(i);
+    bool is_in_table = initial_gvn()->hash_delete(use);
+    uint uses_found = 0;
+    for (uint j = 0; j < use->len(); j++) {
+      if (use->in(j) == n) {
+        if (j < use->req())
+          use->set_req(j, nn);
+        else
+          use->set_prec(j, nn);
+        uses_found++;
+      }
+    }
+    if (is_in_table) {
+      // reinsert into table
+      initial_gvn()->hash_find_insert(use);
+    }
+    record_for_igvn(use);
+    i -= uses_found;    // we deleted 1 or more copies of this edge
+  }
+}
+
+
+
+
 // Identify all nodes that are reachable from below, useful.
 // Use breadth-first pass that records state in a Unique_Node_List,
 // recursive traversal is slower.
@@ -554,6 +580,28 @@
       rethrow_exceptions(kit.transfer_exceptions_into_jvms());
     }
 
+    if (!failing() && has_stringbuilder()) {
+      {
+        // remove useless nodes to make the usage analysis simpler
+        ResourceMark rm;
+        PhaseRemoveUseless pru(initial_gvn(), &for_igvn);
+      }
+
+      {
+        ResourceMark rm;
+        print_method("Before StringOpts", 3);
+        PhaseStringOpts pso(initial_gvn(), &for_igvn);
+        print_method("After StringOpts", 3);
+      }
+
+      // now inline anything that we skipped the first time around
+      while (_late_inlines.length() > 0) {
+        CallGenerator* cg = _late_inlines.pop();
+        cg->do_late_inline();
+      }
+    }
+    assert(_late_inlines.length() == 0, "should have been processed");
+
     print_method("Before RemoveUseless", 3);
 
     // Remove clutter produced by parsing.
@@ -820,6 +868,7 @@
   _fixed_slots = 0;
   set_has_split_ifs(false);
   set_has_loops(has_method() && method()->has_loops()); // first approximation
+  set_has_stringbuilder(false);
   _deopt_happens = true;  // start out assuming the worst
   _trap_can_recompile = false;  // no traps emitted yet
   _major_progress = true; // start out assuming good things will happen
@@ -1803,6 +1852,7 @@
           !n->is_Phi() &&       // a few noisely useless nodes
           !n->is_Proj() &&
           !n->is_MachTemp() &&
+          !n->is_SafePointScalarObject() &&
           !n->is_Catch() &&     // Would be nice to print exception table targets
           !n->is_MergeMem() &&  // Not very interesting
           !n->is_top() &&       // Debug info table constants
@@ -2240,6 +2290,30 @@
     break;
   }
 
+  case Op_Proj: {
+    if (OptimizeStringConcat) {
+      ProjNode* p = n->as_Proj();
+      if (p->_is_io_use) {
+        // Separate projections were used for the exception path which
+        // are normally removed by a late inline.  If it wasn't inlined
+        // then they will hang around and should just be replaced with
+        // the original one.
+        Node* proj = NULL;
+        // Replace with just one
+        for (SimpleDUIterator i(p->in(0)); i.has_next(); i.next()) {
+          Node *use = i.get();
+          if (use->is_Proj() && p != use && use->as_Proj()->_con == p->_con) {
+            proj = use;
+            break;
+          }
+        }
+        assert(p != NULL, "must be found");
+        p->subsume_by(proj);
+      }
+    }
+    break;
+  }
+
   case Op_Phi:
     if (n->as_Phi()->bottom_type()->isa_narrowoop()) {
       // The EncodeP optimization may create Phi with the same edges
--- a/hotspot/src/share/vm/opto/compile.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/opto/compile.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -149,6 +149,7 @@
   bool                  _has_loops;             // True if the method _may_ have some loops
   bool                  _has_split_ifs;         // True if the method _may_ have some split-if
   bool                  _has_unsafe_access;     // True if the method _may_ produce faults in unsafe loads or stores.
+  bool                  _has_stringbuilder;     // True StringBuffers or StringBuilders are allocated
   uint                  _trap_hist[trapHistLength];  // Cumulative traps
   bool                  _trap_can_recompile;    // Have we emitted a recompiling trap?
   uint                  _decompile_count;       // Cumulative decompilation counts.
@@ -219,6 +220,9 @@
   Unique_Node_List*     _for_igvn;              // Initial work-list for next round of Iterative GVN
   WarmCallInfo*         _warm_calls;            // Sorted work-list for heat-based inlining.
 
+  GrowableArray<CallGenerator*> _late_inlines;  // List of CallGenerators to be revisited after
+                                                // main parsing has finished.
+
   // Matching, CFG layout, allocation, code generation
   PhaseCFG*             _cfg;                   // Results of CFG finding
   bool                  _select_24_bit_instr;   // We selected an instruction with a 24-bit result
@@ -298,6 +302,8 @@
   void          set_has_split_ifs(bool z)       { _has_split_ifs = z; }
   bool              has_unsafe_access() const   { return _has_unsafe_access; }
   void          set_has_unsafe_access(bool z)   { _has_unsafe_access = z; }
+  bool              has_stringbuilder() const   { return _has_stringbuilder; }
+  void          set_has_stringbuilder(bool z)   { _has_stringbuilder = z; }
   void          set_trap_count(uint r, uint c)  { assert(r < trapHistLength, "oob");        _trap_hist[r] = c; }
   uint              trap_count(uint r) const    { assert(r < trapHistLength, "oob"); return _trap_hist[r]; }
   bool              trap_can_recompile() const  { return _trap_can_recompile; }
@@ -475,6 +481,7 @@
   // Decide how to build a call.
   // The profile factor is a discount to apply to this site's interp. profile.
   CallGenerator*    call_generator(ciMethod* call_method, int vtable_index, bool call_is_virtual, JVMState* jvms, bool allow_inline, float profile_factor);
+  bool should_delay_inlining(ciMethod* call_method, JVMState* jvms);
 
   // Report if there were too many traps at a current method and bci.
   // Report if a trap was recorded, and/or PerMethodTrapLimit was exceeded.
@@ -495,6 +502,11 @@
   void          set_initial_gvn(PhaseGVN *gvn)           { _initial_gvn = gvn; }
   void          set_for_igvn(Unique_Node_List *for_igvn) { _for_igvn = for_igvn; }
 
+  // Replace n by nn using initial_gvn, calling hash_delete and
+  // record_for_igvn as needed.
+  void gvn_replace_by(Node* n, Node* nn);
+
+
   void              identify_useful_nodes(Unique_Node_List &useful);
   void              remove_useless_nodes  (Unique_Node_List &useful);
 
@@ -502,6 +514,9 @@
   void          set_warm_calls(WarmCallInfo* l) { _warm_calls = l; }
   WarmCallInfo* pop_warm_call();
 
+  // Record this CallGenerator for inlining at the end of parsing.
+  void              add_late_inline(CallGenerator* cg) { _late_inlines.push(cg); }
+
   // Matching, CFG layout, allocation, code generation
   PhaseCFG*         cfg()                       { return _cfg; }
   bool              select_24_bit_instr() const { return _select_24_bit_instr; }
--- a/hotspot/src/share/vm/opto/doCall.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/opto/doCall.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -128,6 +128,12 @@
 
       if (allow_inline) {
         CallGenerator* cg = CallGenerator::for_inline(call_method, expected_uses);
+        if (require_inline && cg != NULL && should_delay_inlining(call_method, jvms)) {
+          // Delay the inlining of this method to give us the
+          // opportunity to perform some high level optimizations
+          // first.
+          return CallGenerator::for_late_inline(call_method, cg);
+        }
         if (cg == NULL) {
           // Fall through.
         } else if (require_inline || !InlineWarmCalls) {
@@ -225,10 +231,63 @@
   } else {
     // Class Hierarchy Analysis or Type Profile reveals a unique target,
     // or it is a static or special call.
-    return CallGenerator::for_direct_call(call_method);
+    return CallGenerator::for_direct_call(call_method, should_delay_inlining(call_method, jvms));
   }
 }
 
+// Return true for methods that shouldn't be inlined early so that
+// they are easier to analyze and optimize as intrinsics.
+bool Compile::should_delay_inlining(ciMethod* call_method, JVMState* jvms) {
+  if (has_stringbuilder()) {
+
+    if ((call_method->holder() == C->env()->StringBuilder_klass() ||
+         call_method->holder() == C->env()->StringBuffer_klass()) &&
+        (jvms->method()->holder() == C->env()->StringBuilder_klass() ||
+         jvms->method()->holder() == C->env()->StringBuffer_klass())) {
+      // Delay SB calls only when called from non-SB code
+      return false;
+    }
+
+    switch (call_method->intrinsic_id()) {
+      case vmIntrinsics::_StringBuilder_void:
+      case vmIntrinsics::_StringBuilder_int:
+      case vmIntrinsics::_StringBuilder_String:
+      case vmIntrinsics::_StringBuilder_append_char:
+      case vmIntrinsics::_StringBuilder_append_int:
+      case vmIntrinsics::_StringBuilder_append_String:
+      case vmIntrinsics::_StringBuilder_toString:
+      case vmIntrinsics::_StringBuffer_void:
+      case vmIntrinsics::_StringBuffer_int:
+      case vmIntrinsics::_StringBuffer_String:
+      case vmIntrinsics::_StringBuffer_append_char:
+      case vmIntrinsics::_StringBuffer_append_int:
+      case vmIntrinsics::_StringBuffer_append_String:
+      case vmIntrinsics::_StringBuffer_toString:
+      case vmIntrinsics::_Integer_toString:
+        return true;
+
+      case vmIntrinsics::_String_String:
+        {
+          Node* receiver = jvms->map()->in(jvms->argoff() + 1);
+          if (receiver->is_Proj() && receiver->in(0)->is_CallStaticJava()) {
+            CallStaticJavaNode* csj = receiver->in(0)->as_CallStaticJava();
+            ciMethod* m = csj->method();
+            if (m != NULL &&
+                (m->intrinsic_id() == vmIntrinsics::_StringBuffer_toString ||
+                 m->intrinsic_id() == vmIntrinsics::_StringBuilder_toString))
+              // Delay String.<init>(new SB())
+              return true;
+          }
+          return false;
+        }
+
+      default:
+        return false;
+    }
+  }
+  return false;
+}
+
 
 // uncommon-trap call-sites where callee is unloaded, uninitialized or will not link
 bool Parse::can_not_compile_call_site(ciMethod *dest_method, ciInstanceKlass* klass) {
--- a/hotspot/src/share/vm/opto/escape.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/opto/escape.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -537,11 +537,13 @@
   }
 
   const TypeOopPtr *tinst = base_t->add_offset(t->offset())->is_oopptr();
-  // Do NOT remove the next call: ensure an new alias index is allocated
-  // for the instance type
+  // Do NOT remove the next line: ensure a new alias index is allocated
+  // for the instance type. Note: C++ will not remove it since the call
+  // has side effect.
   int alias_idx = _compile->get_alias_index(tinst);
   igvn->set_type(addp, tinst);
   // record the allocation in the node map
+  assert(ptnode_adr(addp->_idx)->_node != NULL, "should be registered");
   set_map(addp->_idx, get_map(base->_idx));
 
   // Set addp's Base and Address to 'base'.
@@ -617,9 +619,14 @@
   const TypePtr *atype = C->get_adr_type(alias_idx);
   result = PhiNode::make(orig_phi->in(0), NULL, Type::MEMORY, atype);
   C->copy_node_notes_to(result, orig_phi);
-  set_map_phi(orig_phi->_idx, result);
   igvn->set_type(result, result->bottom_type());
   record_for_optimizer(result);
+
+  debug_only(Node* pn = ptnode_adr(orig_phi->_idx)->_node;)
+  assert(pn == NULL || pn == orig_phi, "wrong node");
+  set_map(orig_phi->_idx, result);
+  ptnode_adr(orig_phi->_idx)->_node = orig_phi;
+
   new_created = true;
   return result;
 }
@@ -710,6 +717,81 @@
 }
 
 //
+// Move memory users to their memory slices.
+//
+void ConnectionGraph::move_inst_mem(Node* n, GrowableArray<PhiNode *>  &orig_phis, PhaseGVN *igvn) {
+  Compile* C = _compile;
+
+  const TypePtr* tp = igvn->type(n->in(MemNode::Address))->isa_ptr();
+  assert(tp != NULL, "ptr type");
+  int alias_idx = C->get_alias_index(tp);
+  int general_idx = C->get_general_index(alias_idx);
+
+  // Move users first
+  for (DUIterator_Fast imax, i = n->fast_outs(imax); i < imax; i++) {
+    Node* use = n->fast_out(i);
+    if (use->is_MergeMem()) {
+      MergeMemNode* mmem = use->as_MergeMem();
+      assert(n == mmem->memory_at(alias_idx), "should be on instance memory slice");
+      if (n != mmem->memory_at(general_idx) || alias_idx == general_idx) {
+        continue; // Nothing to do
+      }
+      // Replace previous general reference to mem node.
+      uint orig_uniq = C->unique();
+      Node* m = find_inst_mem(n, general_idx, orig_phis, igvn);
+      assert(orig_uniq == C->unique(), "no new nodes");
+      mmem->set_memory_at(general_idx, m);
+      --imax;
+      --i;
+    } else if (use->is_MemBar()) {
+      assert(!use->is_Initialize(), "initializing stores should not be moved");
+      if (use->req() > MemBarNode::Precedent &&
+          use->in(MemBarNode::Precedent) == n) {
+        // Don't move related membars.
+        record_for_optimizer(use);
+        continue;
+      }
+      tp = use->as_MemBar()->adr_type()->isa_ptr();
+      if (tp != NULL && C->get_alias_index(tp) == alias_idx ||
+          alias_idx == general_idx) {
+        continue; // Nothing to do
+      }
+      // Move to general memory slice.
+      uint orig_uniq = C->unique();
+      Node* m = find_inst_mem(n, general_idx, orig_phis, igvn);
+      assert(orig_uniq == C->unique(), "no new nodes");
+      igvn->hash_delete(use);
+      imax -= use->replace_edge(n, m);
+      igvn->hash_insert(use);
+      record_for_optimizer(use);
+      --i;
+#ifdef ASSERT
+    } else if (use->is_Mem()) {
+      if (use->Opcode() == Op_StoreCM && use->in(MemNode::OopStore) == n) {
+        // Don't move related cardmark.
+        continue;
+      }
+      // Memory nodes should have new memory input.
+      tp = igvn->type(use->in(MemNode::Address))->isa_ptr();
+      assert(tp != NULL, "ptr type");
+      int idx = C->get_alias_index(tp);
+      assert(get_map(use->_idx) != NULL || idx == alias_idx,
+             "Following memory nodes should have new memory input or be on the same memory slice");
+    } else if (use->is_Phi()) {
+      // Phi nodes should be split and moved already.
+      tp = use->as_Phi()->adr_type()->isa_ptr();
+      assert(tp != NULL, "ptr type");
+      int idx = C->get_alias_index(tp);
+      assert(idx == alias_idx, "Following Phi nodes should be on the same memory slice");
+    } else {
+      use->dump();
+      assert(false, "should not be here");
+#endif
+    }
+  }
+}
+
+//
 // Search memory chain of "mem" to find a MemNode whose address
 // is the specified alias index.
 //
@@ -774,10 +856,18 @@
                C->get_alias_index(result->as_Phi()->adr_type()) != alias_idx) {
       Node *un = result->as_Phi()->unique_input(phase);
       if (un != NULL) {
+        orig_phis.append_if_missing(result->as_Phi());
         result = un;
       } else {
         break;
       }
+    } else if (result->is_ClearArray()) {
+      if (!ClearArrayNode::step_through(&result, (uint)tinst->instance_id(), phase)) {
+        // Can not bypass initialization of the instance
+        // we are looking for.
+        break;
+      }
+      // Otherwise skip it (the call updated 'result' value).
     } else if (result->Opcode() == Op_SCMemProj) {
       assert(result->in(0)->is_LoadStore(), "sanity");
       const Type *at = phase->type(result->in(0)->in(MemNode::Address));
@@ -807,7 +897,6 @@
   return result;
 }
 
-
 //
 //  Convert the types of unescaped object to instance types where possible,
 //  propagate the new type information through the graph, and update memory
@@ -899,12 +988,13 @@
 //
 void ConnectionGraph::split_unique_types(GrowableArray<Node *>  &alloc_worklist) {
   GrowableArray<Node *>  memnode_worklist;
-  GrowableArray<Node *>  mergemem_worklist;
   GrowableArray<PhiNode *>  orig_phis;
+
   PhaseGVN  *igvn = _compile->initial_gvn();
   uint new_index_start = (uint) _compile->num_alias_types();
-  VectorSet visited(Thread::current()->resource_area());
-  VectorSet ptset(Thread::current()->resource_area());
+  Arena* arena = Thread::current()->resource_area();
+  VectorSet visited(arena);
+  VectorSet ptset(arena);
 
 
   //  Phase 1:  Process possible allocations from alloc_worklist.
@@ -980,6 +1070,8 @@
       //   - non-escaping
       //   - eligible to be a unique type
       //   - not determined to be ineligible by escape analysis
+      assert(ptnode_adr(alloc->_idx)->_node != NULL &&
+             ptnode_adr(n->_idx)->_node != NULL, "should be registered");
       set_map(alloc->_idx, n);
       set_map(n->_idx, alloc);
       const TypeOopPtr *t = igvn->type(n)->isa_oopptr();
@@ -1024,7 +1116,7 @@
               alloc_worklist.append_if_missing(addp2);
             }
             alloc_worklist.append_if_missing(use);
-          } else if (use->is_Initialize()) {
+          } else if (use->is_MemBar()) {
             memnode_worklist.append_if_missing(use);
           }
         }
@@ -1034,10 +1126,12 @@
       PointsTo(ptset, get_addp_base(n), igvn);
       assert(ptset.Size() == 1, "AddP address is unique");
       uint elem = ptset.getelem(); // Allocation node's index
-      if (elem == _phantom_object)
+      if (elem == _phantom_object) {
+        assert(false, "escaped allocation");
         continue; // Assume the value was set outside this method.
+      }
       Node *base = get_map(elem);  // CheckCastPP node
-      if (!split_AddP(n, base, igvn)) continue; // wrong type
+      if (!split_AddP(n, base, igvn)) continue; // wrong type from dead path
       tinst = igvn->type(base)->isa_oopptr();
     } else if (n->is_Phi() ||
                n->is_CheckCastPP() ||
@@ -1052,8 +1146,10 @@
       PointsTo(ptset, n, igvn);
       if (ptset.Size() == 1) {
         uint elem = ptset.getelem(); // Allocation node's index
-        if (elem == _phantom_object)
+        if (elem == _phantom_object) {
+          assert(false, "escaped allocation");
           continue; // Assume the value was set outside this method.
+        }
         Node *val = get_map(elem);   // CheckCastPP node
         TypeNode *tn = n->as_Type();
         tinst = igvn->type(val)->isa_oopptr();
@@ -1068,8 +1164,7 @@
           tn_t = tn_type->isa_oopptr();
         }
 
-        if (tn_t != NULL &&
-            tinst->cast_to_instance_id(TypeOopPtr::InstanceBot)->higher_equal(tn_t)) {
+        if (tn_t != NULL && tinst->klass()->is_subtype_of(tn_t->klass())) {
           if (tn_type->isa_narrowoop()) {
             tn_type = tinst->make_narrowoop();
           } else {
@@ -1081,33 +1176,25 @@
           igvn->hash_insert(tn);
           record_for_optimizer(n);
         } else {
-          continue; // wrong type
+          assert(tn_type == TypePtr::NULL_PTR ||
+                 tn_t != NULL && !tinst->klass()->is_subtype_of(tn_t->klass()),
+                 "unexpected type");
+          continue; // Skip dead path with different type
         }
       }
     } else {
+      debug_only(n->dump();)
+      assert(false, "EA: unexpected node");
       continue;
     }
-    // push users on appropriate worklist
+    // push allocation's users on appropriate worklist
     for (DUIterator_Fast imax, i = n->fast_outs(imax); i < imax; i++) {
       Node *use = n->fast_out(i);
       if(use->is_Mem() && use->in(MemNode::Address) == n) {
+        // Load/store to instance's field
         memnode_worklist.append_if_missing(use);
-      } else if (use->is_Initialize()) {
+      } else if (use->is_MemBar()) {
         memnode_worklist.append_if_missing(use);
-      } else if (use->is_MergeMem()) {
-        mergemem_worklist.append_if_missing(use);
-      } else if (use->is_SafePoint() && tinst != NULL) {
-        // Look for MergeMem nodes for calls which reference unique allocation
-        // (through CheckCastPP nodes) even for debug info.
-        Node* m = use->in(TypeFunc::Memory);
-        uint iid = tinst->instance_id();
-        while (m->is_Proj() && m->in(0)->is_SafePoint() &&
-               m->in(0) != use && !m->in(0)->_idx != iid) {
-          m = m->in(0)->in(TypeFunc::Memory);
-        }
-        if (m->is_MergeMem()) {
-          mergemem_worklist.append_if_missing(m);
-        }
       } else if (use->is_AddP() && use->outcnt() > 0) { // No dead nodes
         Node* addp2 = find_second_addp(use, n);
         if (addp2 != NULL) {
@@ -1120,6 +1207,29 @@
                  use->is_DecodeN() ||
                  (use->is_ConstraintCast() && use->Opcode() == Op_CastPP)) {
         alloc_worklist.append_if_missing(use);
+#ifdef ASSERT
+      } else if (use->is_Mem()) {
+        assert(use->in(MemNode::Address) != n, "EA: missing allocation reference path");
+      } else if (use->is_MergeMem()) {
+        assert(_mergemem_worklist.contains(use->as_MergeMem()), "EA: missing MergeMem node in the worklist");
+      } else if (use->is_SafePoint()) {
+        // Look for MergeMem nodes for calls which reference unique allocation
+        // (through CheckCastPP nodes) even for debug info.
+        Node* m = use->in(TypeFunc::Memory);
+        if (m->is_MergeMem()) {
+          assert(_mergemem_worklist.contains(m->as_MergeMem()), "EA: missing MergeMem node in the worklist");
+        }
+      } else {
+        uint op = use->Opcode();
+        if (!(op == Op_CmpP || op == Op_Conv2B ||
+              op == Op_CastP2X || op == Op_StoreCM ||
+              op == Op_FastLock || op == Op_AryEq || op == Op_StrComp ||
+              op == Op_StrEquals || op == Op_StrIndexOf)) {
+          n->dump();
+          use->dump();
+          assert(false, "EA: missing allocation reference path");
+        }
+#endif
       }
     }
 
@@ -1137,19 +1247,16 @@
     Node *n = memnode_worklist.pop();
     if (visited.test_set(n->_idx))
       continue;
-    if (n->is_Phi()) {
-      assert(n->as_Phi()->adr_type() != TypePtr::BOTTOM, "narrow memory slice required");
-      // we don't need to do anything, but the users must be pushed if we haven't processed
-      // this Phi before
-    } else if (n->is_Initialize()) {
-      // we don't need to do anything, but the users of the memory projection must be pushed
-      n = n->as_Initialize()->proj_out(TypeFunc::Memory);
+    if (n->is_Phi() || n->is_ClearArray()) {
+      // we don't need to do anything, but the users must be pushed
+    } else if (n->is_MemBar()) { // Initialize, MemBar nodes
+      // we don't need to do anything, but the users must be pushed
+      n = n->as_MemBar()->proj_out(TypeFunc::Memory);
       if (n == NULL)
         continue;
     } else {
       assert(n->is_Mem(), "memory node required.");
       Node *addr = n->in(MemNode::Address);
-      assert(addr->is_AddP(), "AddP required");
       const Type *addr_t = igvn->type(addr);
       if (addr_t == Type::TOP)
         continue;
@@ -1161,6 +1268,10 @@
         return;
       }
       if (mem != n->in(MemNode::Memory)) {
+        // We delay the memory edge update since we need old one in
+        // MergeMem code below when instances memory slices are separated.
+        debug_only(Node* pn = ptnode_adr(n->_idx)->_node;)
+        assert(pn == NULL || pn == n, "wrong node");
         set_map(n->_idx, mem);
         ptnode_adr(n->_idx)->_node = n;
       }
@@ -1181,36 +1292,55 @@
     // push user on appropriate worklist
     for (DUIterator_Fast imax, i = n->fast_outs(imax); i < imax; i++) {
       Node *use = n->fast_out(i);
-      if (use->is_Phi()) {
+      if (use->is_Phi() || use->is_ClearArray()) {
         memnode_worklist.append_if_missing(use);
       } else if(use->is_Mem() && use->in(MemNode::Memory) == n) {
+        if (use->Opcode() == Op_StoreCM) // Ignore cardmark stores
+          continue;
         memnode_worklist.append_if_missing(use);
-      } else if (use->is_Initialize()) {
+      } else if (use->is_MemBar()) {
         memnode_worklist.append_if_missing(use);
+#ifdef ASSERT
+      } else if(use->is_Mem()) {
+        assert(use->in(MemNode::Memory) != n, "EA: missing memory path");
       } else if (use->is_MergeMem()) {
-        mergemem_worklist.append_if_missing(use);
+        assert(_mergemem_worklist.contains(use->as_MergeMem()), "EA: missing MergeMem node in the worklist");
+      } else {
+        uint op = use->Opcode();
+        if (!(op == Op_StoreCM ||
+              (op == Op_CallLeaf && use->as_CallLeaf()->_name != NULL &&
+               strcmp(use->as_CallLeaf()->_name, "g1_wb_pre") == 0) ||
+              op == Op_AryEq || op == Op_StrComp ||
+              op == Op_StrEquals || op == Op_StrIndexOf)) {
+          n->dump();
+          use->dump();
+          assert(false, "EA: missing memory path");
+        }
+#endif
       }
     }
   }
 
   //  Phase 3:  Process MergeMem nodes from mergemem_worklist.
-  //            Walk each memory moving the first node encountered of each
+  //            Walk each memory slice moving the first node encountered of each
   //            instance type to the the input corresponding to its alias index.
-  while (mergemem_worklist.length() != 0) {
-    Node *n = mergemem_worklist.pop();
-    assert(n->is_MergeMem(), "MergeMem node required.");
-    if (visited.test_set(n->_idx))
-      continue;
-    MergeMemNode *nmm = n->as_MergeMem();
+  uint length = _mergemem_worklist.length();
+  for( uint next = 0; next < length; ++next ) {
+    MergeMemNode* nmm = _mergemem_worklist.at(next);
+    assert(!visited.test_set(nmm->_idx), "should not be visited before");
     // Note: we don't want to use MergeMemStream here because we only want to
-    //  scan inputs which exist at the start, not ones we add during processing.
+    // scan inputs which exist at the start, not ones we add during processing.
+    // Note 2: MergeMem may already contains instance memory slices added
+    // during find_inst_mem() call when memory nodes were processed above.
+    igvn->hash_delete(nmm);
     uint nslices = nmm->req();
-    igvn->hash_delete(nmm);
     for (uint i = Compile::AliasIdxRaw+1; i < nslices; i++) {
       Node* mem = nmm->in(i);
       Node* cur = NULL;
       if (mem == NULL || mem->is_top())
         continue;
+      // First, update mergemem by moving memory nodes to corresponding slices
+      // if their type became more precise since this mergemem was created.
       while (mem->is_Mem()) {
         const Type *at = igvn->type(mem->in(MemNode::Address));
         if (at != Type::TOP) {
@@ -1229,7 +1359,7 @@
       }
       nmm->set_memory_at(i, (cur != NULL) ? cur : mem);
       // Find any instance of the current type if we haven't encountered
-      // a value of the instance along the chain.
+      // already a memory slice of the instance along the memory chain.
       for (uint ni = new_index_start; ni < new_index_end; ni++) {
         if((uint)_compile->get_general_index(ni) == i) {
           Node *m = (ni >= nmm->req()) ? nmm->empty_memory() : nmm->in(ni);
@@ -1245,11 +1375,11 @@
     }
     // Find the rest of instances values
     for (uint ni = new_index_start; ni < new_index_end; ni++) {
-      const TypeOopPtr *tinst = igvn->C->get_adr_type(ni)->isa_oopptr();
+      const TypeOopPtr *tinst = _compile->get_adr_type(ni)->isa_oopptr();
       Node* result = step_through_mergemem(nmm, ni, tinst);
       if (result == nmm->base_memory()) {
         // Didn't find instance memory, search through general slice recursively.
-        result = nmm->memory_at(igvn->C->get_general_index(ni));
+        result = nmm->memory_at(_compile->get_general_index(ni));
         result = find_inst_mem(result, ni, orig_phis, igvn);
         if (_compile->failing()) {
           return;
@@ -1259,41 +1389,6 @@
     }
     igvn->hash_insert(nmm);
     record_for_optimizer(nmm);
-
-    // Propagate new memory slices to following MergeMem nodes.
-    for (DUIterator_Fast imax, i = n->fast_outs(imax); i < imax; i++) {
-      Node *use = n->fast_out(i);
-      if (use->is_Call()) {
-        CallNode* in = use->as_Call();
-        if (in->proj_out(TypeFunc::Memory) != NULL) {
-          Node* m = in->proj_out(TypeFunc::Memory);
-          for (DUIterator_Fast jmax, j = m->fast_outs(jmax); j < jmax; j++) {
-            Node* mm = m->fast_out(j);
-            if (mm->is_MergeMem()) {
-              mergemem_worklist.append_if_missing(mm);
-            }
-          }
-        }
-        if (use->is_Allocate()) {
-          use = use->as_Allocate()->initialization();
-          if (use == NULL) {
-            continue;
-          }
-        }
-      }
-      if (use->is_Initialize()) {
-        InitializeNode* in = use->as_Initialize();
-        if (in->proj_out(TypeFunc::Memory) != NULL) {
-          Node* m = in->proj_out(TypeFunc::Memory);
-          for (DUIterator_Fast jmax, j = m->fast_outs(jmax); j < jmax; j++) {
-            Node* mm = m->fast_out(j);
-            if (mm->is_MergeMem()) {
-              mergemem_worklist.append_if_missing(mm);
-            }
-          }
-        }
-      }
-    }
   }
 
   //  Phase 4:  Update the inputs of non-instance memory Phis and
@@ -1322,19 +1417,48 @@
   }
 
   // Update the memory inputs of MemNodes with the value we computed
-  // in Phase 2.
+  // in Phase 2 and move stores memory users to corresponding memory slices.
+#ifdef ASSERT
+  visited.Clear();
+  Node_Stack old_mems(arena, _compile->unique() >> 2);
+#endif
   for (uint i = 0; i < nodes_size(); i++) {
     Node *nmem = get_map(i);
     if (nmem != NULL) {
       Node *n = ptnode_adr(i)->_node;
-      if (n != NULL && n->is_Mem()) {
+      assert(n != NULL, "sanity");
+      if (n->is_Mem()) {
+#ifdef ASSERT
+        Node* old_mem = n->in(MemNode::Memory);
+        if (!visited.test_set(old_mem->_idx)) {
+          old_mems.push(old_mem, old_mem->outcnt());
+        }
+#endif
+        assert(n->in(MemNode::Memory) != nmem, "sanity");
+        if (!n->is_Load()) {
+          // Move memory users of a store first.
+          move_inst_mem(n, orig_phis, igvn);
+        }
+        // Now update memory input
         igvn->hash_delete(n);
         n->set_req(MemNode::Memory, nmem);
         igvn->hash_insert(n);
         record_for_optimizer(n);
+      } else {
+        assert(n->is_Allocate() || n->is_CheckCastPP() ||
+               n->is_AddP() || n->is_Phi(), "unknown node used for set_map()");
       }
     }
   }
+#ifdef ASSERT
+  // Verify that memory was split correctly
+  while (old_mems.is_nonempty()) {
+    Node* old_mem = old_mems.node();
+    uint  old_cnt = old_mems.index();
+    old_mems.pop();
+    assert(old_cnt = old_mem->outcnt(), "old mem could be lost");
+  }
+#endif
 }
 
 bool ConnectionGraph::has_candidates(Compile *C) {
@@ -1381,8 +1505,20 @@
         ptnode_adr(n->_idx)->node_type() == PointsToNode::JavaObject) {
       has_allocations = true;
     }
-    if(n->is_AddP())
-      cg_worklist.append(n->_idx);
+    if(n->is_AddP()) {
+      // Collect address nodes which directly reference an allocation.
+      // Use them during stage 3 below to build initial connection graph
+      // field edges. Other field edges could be added after StoreP/LoadP
+      // nodes are processed during stage 4 below.
+      Node* base = get_addp_base(n);
+      if(base->is_Proj() && base->in(0)->is_Allocate()) {
+        cg_worklist.append(n->_idx);
+      }
+    } else if (n->is_MergeMem()) {
+      // Collect all MergeMem nodes to add memory slices for
+      // scalar replaceable objects in split_unique_types().
+      _mergemem_worklist.append(n->as_MergeMem());
+    }
     for (DUIterator_Fast imax, i = n->fast_outs(imax); i < imax; i++) {
       Node* m = n->fast_out(i);   // Get user
       worklist_init.push(m);
@@ -1423,12 +1559,13 @@
     }
   }
 
-  VectorSet ptset(Thread::current()->resource_area());
+  Arena* arena = Thread::current()->resource_area();
+  VectorSet ptset(arena);
   GrowableArray<uint>  deferred_edges;
-  VectorSet visited(Thread::current()->resource_area());
+  VectorSet visited(arena);
 
-  // 5. Remove deferred edges from the graph and collect
-  //    information needed for type splitting.
+  // 5. Remove deferred edges from the graph and adjust
+  //    escape state of nonescaping objects.
   cg_length = cg_worklist.length();
   for( uint next = 0; next < cg_length; ++next ) {
     int ni = cg_worklist.at(next);
@@ -1438,98 +1575,9 @@
       remove_deferred(ni, &deferred_edges, &visited);
       Node *n = ptn->_node;
       if (n->is_AddP()) {
-        // Search for objects which are not scalar replaceable.
-        // Mark their escape state as ArgEscape to propagate the state
-        // to referenced objects.
-        // Note: currently there are no difference in compiler optimizations
-        // for ArgEscape objects and NoEscape objects which are not
-        // scalar replaceable.
-
-        int offset = ptn->offset();
-        Node *base = get_addp_base(n);
-        ptset.Clear();
-        PointsTo(ptset, base, igvn);
-        int ptset_size = ptset.Size();
-
-        // Check if a field's initializing value is recorded and add
-        // a corresponding NULL field's value if it is not recorded.
-        // Connection Graph does not record a default initialization by NULL
-        // captured by Initialize node.
-        //
-        // Note: it will disable scalar replacement in some cases:
-        //
-        //    Point p[] = new Point[1];
-        //    p[0] = new Point(); // Will be not scalar replaced
-        //
-        // but it will save us from incorrect optimizations in next cases:
-        //
-        //    Point p[] = new Point[1];
-        //    if ( x ) p[0] = new Point(); // Will be not scalar replaced
-        //
-        // Without a control flow analysis we can't distinguish above cases.
-        //
-        if (offset != Type::OffsetBot && ptset_size == 1) {
-          uint elem = ptset.getelem(); // Allocation node's index
-          // It does not matter if it is not Allocation node since
-          // only non-escaping allocations are scalar replaced.
-          if (ptnode_adr(elem)->_node->is_Allocate() &&
-              ptnode_adr(elem)->escape_state() == PointsToNode::NoEscape) {
-            AllocateNode* alloc = ptnode_adr(elem)->_node->as_Allocate();
-            InitializeNode* ini = alloc->initialization();
-            Node* value = NULL;
-            if (ini != NULL) {
-              BasicType ft = UseCompressedOops ? T_NARROWOOP : T_OBJECT;
-              Node* store = ini->find_captured_store(offset, type2aelembytes(ft), igvn);
-              if (store != NULL && store->is_Store())
-                value = store->in(MemNode::ValueIn);
-            }
-            if (value == NULL || value != ptnode_adr(value->_idx)->_node) {
-              // A field's initializing value was not recorded. Add NULL.
-              uint null_idx = UseCompressedOops ? _noop_null : _oop_null;
-              add_pointsto_edge(ni, null_idx);
-            }
-          }
-        }
-
-        // An object is not scalar replaceable if the field which may point
-        // to it has unknown offset (unknown element of an array of objects).
-        //
-        if (offset == Type::OffsetBot) {
-          uint e_cnt = ptn->edge_count();
-          for (uint ei = 0; ei < e_cnt; ei++) {
-            uint npi = ptn->edge_target(ei);
-            set_escape_state(npi, PointsToNode::ArgEscape);
-            ptnode_adr(npi)->_scalar_replaceable = false;
-          }
-        }
-
-        // Currently an object is not scalar replaceable if a LoadStore node
-        // access its field since the field value is unknown after it.
-        //
-        bool has_LoadStore = false;
-        for (DUIterator_Fast imax, i = n->fast_outs(imax); i < imax; i++) {
-          Node *use = n->fast_out(i);
-          if (use->is_LoadStore()) {
-            has_LoadStore = true;
-            break;
-          }
-        }
-        // An object is not scalar replaceable if the address points
-        // to unknown field (unknown element for arrays, offset is OffsetBot).
-        //
-        // Or the address may point to more then one object. This may produce
-        // the false positive result (set scalar_replaceable to false)
-        // since the flow-insensitive escape analysis can't separate
-        // the case when stores overwrite the field's value from the case
-        // when stores happened on different control branches.
-        //
-        if (ptset_size > 1 || ptset_size != 0 &&
-            (has_LoadStore || offset == Type::OffsetBot)) {
-          for( VectorSetI j(&ptset); j.test(); ++j ) {
-            set_escape_state(j.elem, PointsToNode::ArgEscape);
-            ptnode_adr(j.elem)->_scalar_replaceable = false;
-          }
-        }
+        // Search for objects which are not scalar replaceable
+        // and adjust their escape state.
+        verify_escape_state(ni, ptset, igvn);
       }
     }
   }
@@ -1646,6 +1694,150 @@
   return has_non_escaping_obj;
 }
 
+// Search for objects which are not scalar replaceable.
+void ConnectionGraph::verify_escape_state(int nidx, VectorSet& ptset, PhaseTransform* phase) {
+  PointsToNode* ptn = ptnode_adr(nidx);
+  Node* n = ptn->_node;
+  assert(n->is_AddP(), "Should be called for AddP nodes only");
+  // Search for objects which are not scalar replaceable.
+  // Mark their escape state as ArgEscape to propagate the state
+  // to referenced objects.
+  // Note: currently there are no difference in compiler optimizations
+  // for ArgEscape objects and NoEscape objects which are not
+  // scalar replaceable.
+
+  Compile* C = _compile;
+
+  int offset = ptn->offset();
+  Node* base = get_addp_base(n);
+  ptset.Clear();
+  PointsTo(ptset, base, phase);
+  int ptset_size = ptset.Size();
+
+  // Check if a oop field's initializing value is recorded and add
+  // a corresponding NULL field's value if it is not recorded.
+  // Connection Graph does not record a default initialization by NULL
+  // captured by Initialize node.
+  //
+  // Note: it will disable scalar replacement in some cases:
+  //
+  //    Point p[] = new Point[1];
+  //    p[0] = new Point(); // Will be not scalar replaced
+  //
+  // but it will save us from incorrect optimizations in next cases:
+  //
+  //    Point p[] = new Point[1];
+  //    if ( x ) p[0] = new Point(); // Will be not scalar replaced
+  //
+  // Do a simple control flow analysis to distinguish above cases.
+  //
+  if (offset != Type::OffsetBot && ptset_size == 1) {
+    uint elem = ptset.getelem(); // Allocation node's index
+    // It does not matter if it is not Allocation node since
+    // only non-escaping allocations are scalar replaced.
+    if (ptnode_adr(elem)->_node->is_Allocate() &&
+        ptnode_adr(elem)->escape_state() == PointsToNode::NoEscape) {
+      AllocateNode* alloc = ptnode_adr(elem)->_node->as_Allocate();
+      InitializeNode* ini = alloc->initialization();
+
+      // Check only oop fields.
+      const Type* adr_type = n->as_AddP()->bottom_type();
+      BasicType basic_field_type = T_INT;
+      if (adr_type->isa_instptr()) {
+        ciField* field = C->alias_type(adr_type->isa_instptr())->field();
+        if (field != NULL) {
+          basic_field_type = field->layout_type();
+        } else {
+          // Ignore non field load (for example, klass load)
+        }
+      } else if (adr_type->isa_aryptr()) {
+        const Type* elemtype = adr_type->isa_aryptr()->elem();
+        basic_field_type = elemtype->array_element_basic_type();
+      } else {
+        // Raw pointers are used for initializing stores so skip it.
+        assert(adr_type->isa_rawptr() && base->is_Proj() &&
+               (base->in(0) == alloc),"unexpected pointer type");
+      }
+      if (basic_field_type == T_OBJECT ||
+          basic_field_type == T_NARROWOOP ||
+          basic_field_type == T_ARRAY) {
+        Node* value = NULL;
+        if (ini != NULL) {
+          BasicType ft = UseCompressedOops ? T_NARROWOOP : T_OBJECT;
+          Node* store = ini->find_captured_store(offset, type2aelembytes(ft), phase);
+          if (store != NULL && store->is_Store()) {
+            value = store->in(MemNode::ValueIn);
+          } else if (ptn->edge_count() > 0) { // Are there oop stores?
+            // Check for a store which follows allocation without branches.
+            // For example, a volatile field store is not collected
+            // by Initialize node. TODO: it would be nice to use idom() here.
+            for (DUIterator_Fast imax, i = n->fast_outs(imax); i < imax; i++) {
+              store = n->fast_out(i);
+              if (store->is_Store() && store->in(0) != NULL) {
+                Node* ctrl = store->in(0);
+                while(!(ctrl == ini || ctrl == alloc || ctrl == NULL ||
+                        ctrl == C->root() || ctrl == C->top() || ctrl->is_Region() ||
+                        ctrl->is_IfTrue() || ctrl->is_IfFalse())) {
+                   ctrl = ctrl->in(0);
+                }
+                if (ctrl == ini || ctrl == alloc) {
+                  value = store->in(MemNode::ValueIn);
+                  break;
+                }
+              }
+            }
+          }
+        }
+        if (value == NULL || value != ptnode_adr(value->_idx)->_node) {
+          // A field's initializing value was not recorded. Add NULL.
+          uint null_idx = UseCompressedOops ? _noop_null : _oop_null;
+          add_pointsto_edge(nidx, null_idx);
+        }
+      }
+    }
+  }
+
+  // An object is not scalar replaceable if the field which may point
+  // to it has unknown offset (unknown element of an array of objects).
+  //
+  if (offset == Type::OffsetBot) {
+    uint e_cnt = ptn->edge_count();
+    for (uint ei = 0; ei < e_cnt; ei++) {
+      uint npi = ptn->edge_target(ei);
+      set_escape_state(npi, PointsToNode::ArgEscape);
+      ptnode_adr(npi)->_scalar_replaceable = false;
+    }
+  }
+
+  // Currently an object is not scalar replaceable if a LoadStore node
+  // access its field since the field value is unknown after it.
+  //
+  bool has_LoadStore = false;
+  for (DUIterator_Fast imax, i = n->fast_outs(imax); i < imax; i++) {
+    Node *use = n->fast_out(i);
+    if (use->is_LoadStore()) {
+      has_LoadStore = true;
+      break;
+    }
+  }
+  // An object is not scalar replaceable if the address points
+  // to unknown field (unknown element for arrays, offset is OffsetBot).
+  //
+  // Or the address may point to more then one object. This may produce
+  // the false positive result (set scalar_replaceable to false)
+  // since the flow-insensitive escape analysis can't separate
+  // the case when stores overwrite the field's value from the case
+  // when stores happened on different control branches.
+  //
+  if (ptset_size > 1 || ptset_size != 0 &&
+      (has_LoadStore || offset == Type::OffsetBot)) {
+    for( VectorSetI j(&ptset); j.test(); ++j ) {
+      set_escape_state(j.elem, PointsToNode::ArgEscape);
+      ptnode_adr(j.elem)->_scalar_replaceable = false;
+    }
+  }
+}
+
 void ConnectionGraph::process_call_arguments(CallNode *call, PhaseTransform *phase) {
 
     switch (call->Opcode()) {
@@ -1657,6 +1849,7 @@
       assert(false, "should be done already");
       break;
 #endif
+    case Op_CallLeaf:
     case Op_CallLeafNoFP:
     {
       // Stub calls, objects do not escape but they are not scale replaceable.
@@ -1667,9 +1860,23 @@
         const Type* at = d->field_at(i);
         Node *arg = call->in(i)->uncast();
         const Type *aat = phase->type(arg);
-        if (!arg->is_top() && at->isa_ptr() && aat->isa_ptr()) {
+        if (!arg->is_top() && at->isa_ptr() && aat->isa_ptr() &&
+            ptnode_adr(arg->_idx)->escape_state() < PointsToNode::ArgEscape) {
+
           assert(aat == Type::TOP || aat == TypePtr::NULL_PTR ||
                  aat->isa_ptr() != NULL, "expecting an Ptr");
+#ifdef ASSERT
+          if (!(call->Opcode() == Op_CallLeafNoFP &&
+                call->as_CallLeaf()->_name != NULL &&
+                (strstr(call->as_CallLeaf()->_name, "arraycopy")  != 0) ||
+                call->as_CallLeaf()->_name != NULL &&
+                (strcmp(call->as_CallLeaf()->_name, "g1_wb_pre")  == 0 ||
+                 strcmp(call->as_CallLeaf()->_name, "g1_wb_post") == 0 ))
+          ) {
+            call->dump();
+            assert(false, "EA: unexpected CallLeaf");
+          }
+#endif
           set_escape_state(arg->_idx, PointsToNode::ArgEscape);
           if (arg->is_AddP()) {
             //
@@ -1706,9 +1913,10 @@
         for (uint i = TypeFunc::Parms; i < d->cnt(); i++) {
           const Type* at = d->field_at(i);
           int k = i - TypeFunc::Parms;
+          Node *arg = call->in(i)->uncast();
 
-          if (at->isa_oopptr() != NULL) {
-            Node *arg = call->in(i)->uncast();
+          if (at->isa_oopptr() != NULL &&
+              ptnode_adr(arg->_idx)->escape_state() < PointsToNode::ArgEscape) {
 
             bool global_escapes = false;
             bool fields_escapes = false;
@@ -1942,20 +2150,23 @@
       record_for_optimizer(n);
       _processed.set(n->_idx);
     } else {
-      // Have to process call's arguments first.
+      // Don't mark as processed since call's arguments have to be processed.
       PointsToNode::NodeType nt = PointsToNode::UnknownType;
+      PointsToNode::EscapeState es = PointsToNode::UnknownEscape;
 
       // Check if a call returns an object.
       const TypeTuple *r = n->as_Call()->tf()->range();
-      if (n->is_CallStaticJava() && r->cnt() > TypeFunc::Parms &&
+      if (r->cnt() > TypeFunc::Parms &&
+          r->field_at(TypeFunc::Parms)->isa_ptr() &&
           n->as_Call()->proj_out(TypeFunc::Parms) != NULL) {
-        // Note:  use isa_ptr() instead of isa_oopptr() here because
-        //        the _multianewarray functions return a TypeRawPtr.
-        if (r->field_at(TypeFunc::Parms)->isa_ptr() != NULL) {
-          nt = PointsToNode::JavaObject;
+        nt = PointsToNode::JavaObject;
+        if (!n->is_CallStaticJava()) {
+          // Since the called mathod is statically unknown assume
+          // the worst case that the returned value globally escapes.
+          es = PointsToNode::GlobalEscape;
         }
       }
-      add_node(n, nt, PointsToNode::UnknownEscape, false);
+      add_node(n, nt, es, false);
     }
     return;
   }
@@ -2088,18 +2299,27 @@
     }
     case Op_Proj:
     {
-      // we are only interested in the result projection from a call
+      // we are only interested in the oop result projection from a call
       if (n->as_Proj()->_con == TypeFunc::Parms && n->in(0)->is_Call() ) {
-        add_node(n, PointsToNode::LocalVar, PointsToNode::UnknownEscape, false);
-        process_call_result(n->as_Proj(), phase);
-        if (!_processed.test(n->_idx)) {
-          // The call's result may need to be processed later if the call
-          // returns it's argument and the argument is not processed yet.
-          _delayed_worklist.push(n);
+        const TypeTuple *r = n->in(0)->as_Call()->tf()->range();
+        assert(r->cnt() > TypeFunc::Parms, "sanity");
+        if (r->field_at(TypeFunc::Parms)->isa_ptr() != NULL) {
+          add_node(n, PointsToNode::LocalVar, PointsToNode::UnknownEscape, false);
+          int ti = n->in(0)->_idx;
+          // The call may not be registered yet (since not all its inputs are registered)
+          // if this is the projection from backbranch edge of Phi.
+          if (ptnode_adr(ti)->node_type() != PointsToNode::UnknownType) {
+            process_call_result(n->as_Proj(), phase);
+          }
+          if (!_processed.test(n->_idx)) {
+            // The call's result may need to be processed later if the call
+            // returns it's argument and the argument is not processed yet.
+            _delayed_worklist.push(n);
+          }
+          break;
         }
-      } else {
-        _processed.set(n->_idx);
       }
+      _processed.set(n->_idx);
       break;
     }
     case Op_Return:
@@ -2160,6 +2380,15 @@
       }
       break;
     }
+    case Op_AryEq:
+    case Op_StrComp:
+    case Op_StrEquals:
+    case Op_StrIndexOf:
+    {
+      // char[] arrays passed to string intrinsics are not scalar replaceable.
+      add_node(n, PointsToNode::UnknownType, PointsToNode::UnknownEscape, false);
+      break;
+    }
     case Op_ThreadLocal:
     {
       add_node(n, PointsToNode::JavaObject, PointsToNode::ArgEscape, true);
@@ -2174,6 +2403,7 @@
 
 void ConnectionGraph::build_connection_graph(Node *n, PhaseTransform *phase) {
   uint n_idx = n->_idx;
+  assert(ptnode_adr(n_idx)->_node != NULL, "node should be registered");
 
   // Don't set processed bit for AddP, LoadP, StoreP since
   // they may need more then one pass to process.
@@ -2211,6 +2441,7 @@
     case Op_DecodeN:
     {
       int ti = n->in(1)->_idx;
+      assert(ptnode_adr(ti)->node_type() != PointsToNode::UnknownType, "all nodes should be registered");
       if (ptnode_adr(ti)->node_type() == PointsToNode::JavaObject) {
         add_pointsto_edge(n_idx, ti);
       } else {
@@ -2250,7 +2481,6 @@
 #endif
 
       Node* adr = n->in(MemNode::Address)->uncast();
-      const Type *adr_type = phase->type(adr);
       Node* adr_base;
       if (adr->is_AddP()) {
         adr_base = get_addp_base(adr);
@@ -2302,13 +2532,19 @@
     }
     case Op_Proj:
     {
-      // we are only interested in the result projection from a call
+      // we are only interested in the oop result projection from a call
       if (n->as_Proj()->_con == TypeFunc::Parms && n->in(0)->is_Call() ) {
-        process_call_result(n->as_Proj(), phase);
-        assert(_processed.test(n_idx), "all call results should be processed");
-      } else {
-        assert(false, "Op_Proj");
+        assert(ptnode_adr(n->in(0)->_idx)->node_type() != PointsToNode::UnknownType,
+               "all nodes should be registered");
+        const TypeTuple *r = n->in(0)->as_Call()->tf()->range();
+        assert(r->cnt() > TypeFunc::Parms, "sanity");
+        if (r->field_at(TypeFunc::Parms)->isa_ptr() != NULL) {
+          process_call_result(n->as_Proj(), phase);
+          assert(_processed.test(n_idx), "all call results should be processed");
+          break;
+        }
       }
+      assert(false, "Op_Proj");
       break;
     }
     case Op_Return:
@@ -2320,6 +2556,7 @@
       }
 #endif
       int ti = n->in(TypeFunc::Parms)->_idx;
+      assert(ptnode_adr(ti)->node_type() != PointsToNode::UnknownType, "node should be registered");
       if (ptnode_adr(ti)->node_type() == PointsToNode::JavaObject) {
         add_pointsto_edge(n_idx, ti);
       } else {
@@ -2354,14 +2591,38 @@
       }
       break;
     }
+    case Op_AryEq:
+    case Op_StrComp:
+    case Op_StrEquals:
+    case Op_StrIndexOf:
+    {
+      // char[] arrays passed to string intrinsic do not escape but
+      // they are not scalar replaceable. Adjust escape state for them.
+      // Start from in(2) edge since in(1) is memory edge.
+      for (uint i = 2; i < n->req(); i++) {
+        Node* adr = n->in(i)->uncast();
+        const Type *at = phase->type(adr);
+        if (!adr->is_top() && at->isa_ptr()) {
+          assert(at == Type::TOP || at == TypePtr::NULL_PTR ||
+                 at->isa_ptr() != NULL, "expecting an Ptr");
+          if (adr->is_AddP()) {
+            adr = get_addp_base(adr);
+          }
+          // Mark as ArgEscape everything "adr" could point to.
+          set_escape_state(adr->_idx, PointsToNode::ArgEscape);
+        }
+      }
+      _processed.set(n_idx);
+      break;
+    }
     case Op_ThreadLocal:
     {
       assert(false, "Op_ThreadLocal");
       break;
     }
     default:
-      ;
-      // nothing to do
+      // This method should be called only for EA specific nodes.
+      ShouldNotReachHere();
   }
 }
 
--- a/hotspot/src/share/vm/opto/escape.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/opto/escape.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -210,6 +210,8 @@
   Unique_Node_List  _delayed_worklist; // Nodes to be processed before
                                        // the call build_connection_graph().
 
+  GrowableArray<MergeMemNode *>  _mergemem_worklist; // List of all MergeMem nodes
+
   VectorSet                _processed; // Records which nodes have been
                                        // processed.
 
@@ -289,7 +291,7 @@
   bool split_AddP(Node *addp, Node *base,  PhaseGVN  *igvn);
   PhiNode *create_split_phi(PhiNode *orig_phi, int alias_idx, GrowableArray<PhiNode *>  &orig_phi_worklist, PhaseGVN  *igvn, bool &new_created);
   PhiNode *split_memory_phi(PhiNode *orig_phi, int alias_idx, GrowableArray<PhiNode *>  &orig_phi_worklist, PhaseGVN  *igvn);
-  Node *find_mem(Node *mem, int alias_idx, PhaseGVN  *igvn);
+  void  move_inst_mem(Node* n, GrowableArray<PhiNode *>  &orig_phis, PhaseGVN *igvn);
   Node *find_inst_mem(Node *mem, int alias_idx,GrowableArray<PhiNode *>  &orig_phi_worklist,  PhaseGVN  *igvn);
 
   // Propagate unique types created for unescaped allocated objects
@@ -298,7 +300,6 @@
 
   // manage entries in _node_map
   void  set_map(int idx, Node *n)        { _node_map.map(idx, n); }
-  void  set_map_phi(int idx, PhiNode *p) { _node_map.map(idx, (Node *) p); }
   Node *get_map(int idx)                 { return _node_map[idx]; }
   PhiNode *get_map_phi(int idx) {
     Node *phi = _node_map[idx];
@@ -315,6 +316,9 @@
   // Set the escape state of a node
   void set_escape_state(uint ni, PointsToNode::EscapeState es);
 
+  // Search for objects which are not scalar replaceable.
+  void verify_escape_state(int nidx, VectorSet& ptset, PhaseTransform* phase);
+
 public:
   ConnectionGraph(Compile *C);
 
--- a/hotspot/src/share/vm/opto/graphKit.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/opto/graphKit.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -1351,8 +1351,8 @@
 }
 
 //------------------------------set_all_memory_call----------------------------
-void GraphKit::set_all_memory_call(Node* call) {
-  Node* newmem = _gvn.transform( new (C, 1) ProjNode(call, TypeFunc::Memory) );
+void GraphKit::set_all_memory_call(Node* call, bool separate_io_proj) {
+  Node* newmem = _gvn.transform( new (C, 1) ProjNode(call, TypeFunc::Memory, separate_io_proj) );
   set_all_memory(newmem);
 }
 
@@ -1573,7 +1573,7 @@
 //---------------------------set_edges_for_java_call---------------------------
 // Connect a newly created call into the current JVMS.
 // A return value node (if any) is returned from set_edges_for_java_call.
-void GraphKit::set_edges_for_java_call(CallJavaNode* call, bool must_throw) {
+void GraphKit::set_edges_for_java_call(CallJavaNode* call, bool must_throw, bool separate_io_proj) {
 
   // Add the predefined inputs:
   call->init_req( TypeFunc::Control, control() );
@@ -1595,13 +1595,13 @@
   // Re-use the current map to produce the result.
 
   set_control(_gvn.transform(new (C, 1) ProjNode(call, TypeFunc::Control)));
-  set_i_o(    _gvn.transform(new (C, 1) ProjNode(call, TypeFunc::I_O    )));
-  set_all_memory_call(xcall);
+  set_i_o(    _gvn.transform(new (C, 1) ProjNode(call, TypeFunc::I_O    , separate_io_proj)));
+  set_all_memory_call(xcall, separate_io_proj);
 
   //return xcall;   // no need, caller already has it
 }
 
-Node* GraphKit::set_results_for_java_call(CallJavaNode* call) {
+Node* GraphKit::set_results_for_java_call(CallJavaNode* call, bool separate_io_proj) {
   if (stopped())  return top();  // maybe the call folded up?
 
   // Capture the return value, if any.
@@ -1614,8 +1614,15 @@
   // Note:  Since any out-of-line call can produce an exception,
   // we always insert an I_O projection from the call into the result.
 
-  make_slow_call_ex(call, env()->Throwable_klass(), false);
-
+  make_slow_call_ex(call, env()->Throwable_klass(), separate_io_proj);
+
+  if (separate_io_proj) {
+    // The caller requested separate projections be used by the fall
+    // through and exceptional paths, so replace the projections for
+    // the fall through path.
+    set_i_o(_gvn.transform( new (C, 1) ProjNode(call, TypeFunc::I_O) ));
+    set_all_memory(_gvn.transform( new (C, 1) ProjNode(call, TypeFunc::Memory) ));
+  }
   return ret;
 }
 
@@ -1678,6 +1685,64 @@
   }
 }
 
+
+// Replace the call with the current state of the kit.
+void GraphKit::replace_call(CallNode* call, Node* result) {
+  JVMState* ejvms = NULL;
+  if (has_exceptions()) {
+    ejvms = transfer_exceptions_into_jvms();
+  }
+
+  SafePointNode* final_state = stop();
+
+  // Find all the needed outputs of this call
+  CallProjections callprojs;
+  call->extract_projections(&callprojs, true);
+
+  // Replace all the old call edges with the edges from the inlining result
+  C->gvn_replace_by(callprojs.fallthrough_catchproj, final_state->in(TypeFunc::Control));
+  C->gvn_replace_by(callprojs.fallthrough_memproj,   final_state->in(TypeFunc::Memory));
+  C->gvn_replace_by(callprojs.fallthrough_ioproj,    final_state->in(TypeFunc::I_O));
+
+  // Replace the result with the new result if it exists and is used
+  if (callprojs.resproj != NULL && result != NULL) {
+    C->gvn_replace_by(callprojs.resproj, result);
+  }
+
+  if (ejvms == NULL) {
+    // No exception edges to simply kill off those paths
+    C->gvn_replace_by(callprojs.catchall_catchproj, C->top());
+    C->gvn_replace_by(callprojs.catchall_memproj,   C->top());
+    C->gvn_replace_by(callprojs.catchall_ioproj,    C->top());
+
+    // Replace the old exception object with top
+    if (callprojs.exobj != NULL) {
+      C->gvn_replace_by(callprojs.exobj, C->top());
+    }
+  } else {
+    GraphKit ekit(ejvms);
+
+    // Load my combined exception state into the kit, with all phis transformed:
+    SafePointNode* ex_map = ekit.combine_and_pop_all_exception_states();
+
+    Node* ex_oop = ekit.use_exception_state(ex_map);
+
+    C->gvn_replace_by(callprojs.catchall_catchproj, ekit.control());
+    C->gvn_replace_by(callprojs.catchall_memproj,   ekit.reset_memory());
+    C->gvn_replace_by(callprojs.catchall_ioproj,    ekit.i_o());
+
+    // Replace the old exception object with the newly created one
+    if (callprojs.exobj != NULL) {
+      C->gvn_replace_by(callprojs.exobj, ex_oop);
+    }
+  }
+
+  // Disconnect the call from the graph
+  call->disconnect_inputs(NULL);
+  C->gvn_replace_by(call, C->top());
+}
+
+
 //------------------------------increment_counter------------------------------
 // for statistics: increment a VM counter by 1
 
@@ -3459,4 +3524,3 @@
   sync_kit(ideal);
 }
 #undef __
-
--- a/hotspot/src/share/vm/opto/graphKit.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/opto/graphKit.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -279,6 +279,34 @@
   }
   Node* basic_plus_adr(Node* base, Node* ptr, Node* offset);
 
+
+  // Some convenient shortcuts for common nodes
+  Node* IfTrue(IfNode* iff)                   { return _gvn.transform(new (C,1) IfTrueNode(iff));      }
+  Node* IfFalse(IfNode* iff)                  { return _gvn.transform(new (C,1) IfFalseNode(iff));     }
+
+  Node* AddI(Node* l, Node* r)                { return _gvn.transform(new (C,3) AddINode(l, r));       }
+  Node* SubI(Node* l, Node* r)                { return _gvn.transform(new (C,3) SubINode(l, r));       }
+  Node* MulI(Node* l, Node* r)                { return _gvn.transform(new (C,3) MulINode(l, r));       }
+  Node* DivI(Node* ctl, Node* l, Node* r)     { return _gvn.transform(new (C,3) DivINode(ctl, l, r));  }
+
+  Node* AndI(Node* l, Node* r)                { return _gvn.transform(new (C,3) AndINode(l, r));       }
+  Node* OrI(Node* l, Node* r)                 { return _gvn.transform(new (C,3) OrINode(l, r));        }
+  Node* XorI(Node* l, Node* r)                { return _gvn.transform(new (C,3) XorINode(l, r));       }
+
+  Node* MaxI(Node* l, Node* r)                { return _gvn.transform(new (C,3) MaxINode(l, r));       }
+  Node* MinI(Node* l, Node* r)                { return _gvn.transform(new (C,3) MinINode(l, r));       }
+
+  Node* LShiftI(Node* l, Node* r)             { return _gvn.transform(new (C,3) LShiftINode(l, r));    }
+  Node* RShiftI(Node* l, Node* r)             { return _gvn.transform(new (C,3) RShiftINode(l, r));    }
+  Node* URShiftI(Node* l, Node* r)            { return _gvn.transform(new (C,3) URShiftINode(l, r));   }
+
+  Node* CmpI(Node* l, Node* r)                { return _gvn.transform(new (C,3) CmpINode(l, r));       }
+  Node* CmpL(Node* l, Node* r)                { return _gvn.transform(new (C,3) CmpLNode(l, r));       }
+  Node* CmpP(Node* l, Node* r)                { return _gvn.transform(new (C,3) CmpPNode(l, r));       }
+  Node* Bool(Node* cmp, BoolTest::mask relop) { return _gvn.transform(new (C,2) BoolNode(cmp, relop)); }
+
+  Node* AddP(Node* b, Node* a, Node* o)       { return _gvn.transform(new (C,4) AddPNode(b, a, o));    }
+
   // Convert between int and long, and size_t.
   // (See macros ConvI2X, etc., in type.hpp for ConvI2X, etc.)
   Node* ConvI2L(Node* offset);
@@ -400,7 +428,7 @@
   void set_all_memory(Node* newmem);
 
   // Create a memory projection from the call, then set_all_memory.
-  void set_all_memory_call(Node* call);
+  void set_all_memory_call(Node* call, bool separate_io_proj = false);
 
   // Create a LoadNode, reading from the parser's memory state.
   // (Note:  require_atomic_access is useful only with T_LONG.)
@@ -543,12 +571,12 @@
   // Transform the call, and update the basics: control, i_o, memory.
   // (The next step is usually to call set_results_for_java_call.)
   void set_edges_for_java_call(CallJavaNode* call,
-                               bool must_throw = false);
+                               bool must_throw = false, bool separate_io_proj = false);
 
   // Finish up a java call that was started by set_edges_for_java_call.
   // Call add_exception on any throw arising from the call.
   // Return the call result (transformed).
-  Node* set_results_for_java_call(CallJavaNode* call);
+  Node* set_results_for_java_call(CallJavaNode* call, bool separate_io_proj = false);
 
   // Similar to set_edges_for_java_call, but simplified for runtime calls.
   void  set_predefined_output_for_runtime_call(Node* call) {
@@ -559,6 +587,11 @@
                                                const TypePtr* hook_mem);
   Node* set_predefined_input_for_runtime_call(SafePointNode* call);
 
+  // Replace the call with the current state of the kit.  Requires
+  // that the call was generated with separate io_projs so that
+  // exceptional control flow can be handled properly.
+  void replace_call(CallNode* call, Node* result);
+
   // helper functions for statistics
   void increment_counter(address counter_addr);   // increment a debug counter
   void increment_counter(Node*   counter_addr);   // increment a debug counter
--- a/hotspot/src/share/vm/opto/lcm.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/opto/lcm.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -616,8 +616,9 @@
           assert(cfg->_bbs[oop_store->_idx]->_dom_depth <= this->_dom_depth, "oop_store must dominate card-mark");
         }
       }
-      if( n->is_Mach() && n->as_Mach()->ideal_Opcode() == Op_MemBarAcquire &&
-          n->req() > TypeFunc::Parms ) {
+      if( n->is_Mach() && n->req() > TypeFunc::Parms &&
+          (n->as_Mach()->ideal_Opcode() == Op_MemBarAcquire ||
+           n->as_Mach()->ideal_Opcode() == Op_MemBarVolatile) ) {
         // MemBarAcquire could be created without Precedent edge.
         // del_req() replaces the specified edge with the last input edge
         // and then removes the last edge. If the specified edge > number of
--- a/hotspot/src/share/vm/opto/loopnode.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/opto/loopnode.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -1279,7 +1279,8 @@
     // Visit all children, looking for Phis
     for (DUIterator i = cl->outs(); cl->has_out(i); i++) {
       Node *out = cl->out(i);
-      if (!out->is_Phi() || out == phi)  continue; // Looking for other phis
+      // Look for other phis (secondary IVs). Skip dead ones
+      if (!out->is_Phi() || out == phi || !phase->has_node(out)) continue;
       PhiNode* phi2 = out->as_Phi();
       Node *incr2 = phi2->in( LoopNode::LoopBackControl );
       // Look for induction variables of the form:  X += constant
--- a/hotspot/src/share/vm/opto/macro.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/opto/macro.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -316,6 +316,21 @@
         assert(adr_idx == Compile::AliasIdxRaw, "address must match or be raw");
       }
       mem = mem->in(MemNode::Memory);
+    } else if (mem->is_ClearArray()) {
+      if (!ClearArrayNode::step_through(&mem, alloc->_idx, phase)) {
+        // Can not bypass initialization of the instance
+        // we are looking.
+        debug_only(intptr_t offset;)
+        assert(alloc == AllocateNode::Ideal_allocation(mem->in(3), phase, offset), "sanity");
+        InitializeNode* init = alloc->as_Allocate()->initialization();
+        // We are looking for stored value, return Initialize node
+        // or memory edge from Allocate node.
+        if (init != NULL)
+          return init;
+        else
+          return alloc->in(TypeFunc::Memory); // It will produce zero value (see callers).
+      }
+      // Otherwise skip it (the call updated 'mem' value).
     } else if (mem->Opcode() == Op_SCMemProj) {
       assert(mem->in(0)->is_LoadStore(), "sanity");
       const TypePtr* atype = mem->in(0)->in(MemNode::Address)->bottom_type()->is_ptr();
@@ -823,6 +838,18 @@
           Node *n = use->last_out(k);
           uint oc2 = use->outcnt();
           if (n->is_Store()) {
+#ifdef ASSERT
+            // Verify that there is no dependent MemBarVolatile nodes,
+            // they should be removed during IGVN, see MemBarNode::Ideal().
+            for (DUIterator_Fast pmax, p = n->fast_outs(pmax);
+                                       p < pmax; p++) {
+              Node* mb = n->fast_out(p);
+              assert(mb->is_Initialize() || !mb->is_MemBar() ||
+                     mb->req() <= MemBarNode::Precedent ||
+                     mb->in(MemBarNode::Precedent) != n,
+                     "MemBarVolatile should be eliminated for non-escaping object");
+            }
+#endif
             _igvn.replace_node(n, n->in(MemNode::Memory));
           } else {
             eliminate_card_mark(n);
@@ -912,15 +939,29 @@
     return false;
   }
 
+  CompileLog* log = C->log();
+  if (log != NULL) {
+    Node* klass = alloc->in(AllocateNode::KlassNode);
+    const TypeKlassPtr* tklass = _igvn.type(klass)->is_klassptr();
+    log->head("eliminate_allocation type='%d'",
+              log->identify(tklass->klass()));
+    JVMState* p = alloc->jvms();
+    while (p != NULL) {
+      log->elem("jvms bci='%d' method='%d'", p->bci(), log->identify(p->method()));
+      p = p->caller();
+    }
+    log->tail("eliminate_allocation");
+  }
+
   process_users_of_allocation(alloc);
 
 #ifndef PRODUCT
-if (PrintEliminateAllocations) {
-  if (alloc->is_AllocateArray())
-    tty->print_cr("++++ Eliminated: %d AllocateArray", alloc->_idx);
-  else
-    tty->print_cr("++++ Eliminated: %d Allocate", alloc->_idx);
-}
+  if (PrintEliminateAllocations) {
+    if (alloc->is_AllocateArray())
+      tty->print_cr("++++ Eliminated: %d AllocateArray", alloc->_idx);
+    else
+      tty->print_cr("++++ Eliminated: %d Allocate", alloc->_idx);
+  }
 #endif
 
   return true;
@@ -1639,6 +1680,18 @@
       } // if (!oldbox->is_eliminated())
   } // if (alock->is_Lock() && !lock->is_coarsened())
 
+  CompileLog* log = C->log();
+  if (log != NULL) {
+    log->head("eliminate_lock lock='%d'",
+              alock->is_Lock());
+    JVMState* p = alock->jvms();
+    while (p != NULL) {
+      log->elem("jvms bci='%d' method='%d'", p->bci(), log->identify(p->method()));
+      p = p->caller();
+    }
+    log->tail("eliminate_lock");
+  }
+
   #ifndef PRODUCT
   if (PrintEliminateLocks) {
     if (alock->is_Lock()) {
--- a/hotspot/src/share/vm/opto/matcher.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/opto/matcher.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -1832,67 +1832,23 @@
       case Op_Binary:         // These are introduced in the Post_Visit state.
         ShouldNotReachHere();
         break;
-      case Op_StoreB:         // Do match these, despite no ideal reg
-      case Op_StoreC:
-      case Op_StoreCM:
-      case Op_StoreD:
-      case Op_StoreF:
-      case Op_StoreI:
-      case Op_StoreL:
-      case Op_StoreP:
-      case Op_StoreN:
-      case Op_Store16B:
-      case Op_Store8B:
-      case Op_Store4B:
-      case Op_Store8C:
-      case Op_Store4C:
-      case Op_Store2C:
-      case Op_Store4I:
-      case Op_Store2I:
-      case Op_Store2L:
-      case Op_Store4F:
-      case Op_Store2F:
-      case Op_Store2D:
       case Op_ClearArray:
       case Op_SafePoint:
         mem_op = true;
         break;
-      case Op_LoadB:
-      case Op_LoadUS:
-      case Op_LoadD:
-      case Op_LoadF:
-      case Op_LoadI:
-      case Op_LoadKlass:
-      case Op_LoadNKlass:
-      case Op_LoadL:
-      case Op_LoadS:
-      case Op_LoadP:
-      case Op_LoadN:
-      case Op_LoadRange:
-      case Op_LoadD_unaligned:
-      case Op_LoadL_unaligned:
-      case Op_Load16B:
-      case Op_Load8B:
-      case Op_Load4B:
-      case Op_Load4C:
-      case Op_Load2C:
-      case Op_Load8C:
-      case Op_Load8S:
-      case Op_Load4S:
-      case Op_Load2S:
-      case Op_Load4I:
-      case Op_Load2I:
-      case Op_Load2L:
-      case Op_Load4F:
-      case Op_Load2F:
-      case Op_Load2D:
-        mem_op = true;
-        // Must be root of match tree due to prior load conflict
-        if( C->subsume_loads() == false ) {
-          set_shared(n);
+      default:
+        if( n->is_Store() ) {
+          // Do match stores, despite no ideal reg
+          mem_op = true;
+          break;
+        }
+        if( n->is_Mem() ) { // Loads and LoadStores
+          mem_op = true;
+          // Loads must be root of match tree due to prior load conflict
+          if( C->subsume_loads() == false )
+            set_shared(n);
         }
         // Fall into default case
-      default:
         if( !n->ideal_reg() )
           set_dontcare(n);  // Unmatchable Nodes
       } // end_switch
@@ -1913,15 +1869,15 @@
           continue; // for(int i = ...)
         }
 
-        // Clone addressing expressions as they are "free" in most instructions
+        if( mop == Op_AddP && m->in(AddPNode::Base)->Opcode() == Op_DecodeN ) {
+          // Bases used in addresses must be shared but since
+          // they are shared through a DecodeN they may appear
+          // to have a single use so force sharing here.
+          set_shared(m->in(AddPNode::Base)->in(1));
+        }
+
+        // Clone addressing expressions as they are "free" in memory access instructions
         if( mem_op && i == MemNode::Address && mop == Op_AddP ) {
-          if (m->in(AddPNode::Base)->Opcode() == Op_DecodeN) {
-            // Bases used in addresses must be shared but since
-            // they are shared through a DecodeN they may appear
-            // to have a single use so force sharing here.
-            set_shared(m->in(AddPNode::Base)->in(1));
-          }
-
           // Some inputs for address expression are not put on stack
           // to avoid marking them as shared and forcing them into register
           // if they are used only in address expressions.
--- a/hotspot/src/share/vm/opto/memnode.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/opto/memnode.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -123,6 +123,13 @@
       } else {
         assert(false, "unexpected projection");
       }
+    } else if (result->is_ClearArray()) {
+      if (!ClearArrayNode::step_through(&result, instance_id, phase)) {
+        // Can not bypass initialization of the instance
+        // we are looking for.
+        break;
+      }
+      // Otherwise skip it (the call updated 'result' value).
     } else if (result->is_MergeMem()) {
       result = step_through_mergemem(phase, result->as_MergeMem(), t_adr, NULL, tty);
     }
@@ -255,6 +262,13 @@
     return NodeSentinel; // caller will return NULL
   }
 
+  // Do NOT remove or optimize the next lines: ensure a new alias index
+  // is allocated for an oop pointer type before Escape Analysis.
+  // Note: C++ will not remove it since the call has side effect.
+  if ( t_adr->isa_oopptr() ) {
+    int alias_idx = phase->C->get_alias_index(t_adr->is_ptr());
+  }
+
 #ifdef ASSERT
   Node* base = NULL;
   if (address->is_AddP())
@@ -530,6 +544,15 @@
       } else if (mem->is_Proj() && mem->in(0)->is_MemBar()) {
         mem = mem->in(0)->in(TypeFunc::Memory);
         continue;           // (a) advance through independent MemBar memory
+      } else if (mem->is_ClearArray()) {
+        if (ClearArrayNode::step_through(&mem, (uint)addr_t->instance_id(), phase)) {
+          // (the call updated 'mem' value)
+          continue;         // (a) advance through independent allocation memory
+        } else {
+          // Can not bypass initialization of the instance
+          // we are looking for.
+          return mem;
+        }
       } else if (mem->is_MergeMem()) {
         int alias_idx = phase->C->get_alias_index(adr_type());
         mem = mem->as_MergeMem()->memory_at(alias_idx);
@@ -1496,6 +1519,8 @@
       }
     }
   } else if (tp->base() == Type::InstPtr) {
+    const TypeInstPtr* tinst = tp->is_instptr();
+    ciKlass* klass = tinst->klass();
     assert( off != Type::OffsetBot ||
             // arrays can be cast to Objects
             tp->is_oopptr()->klass()->is_java_lang_Object() ||
@@ -1503,6 +1528,25 @@
             phase->C->has_unsafe_access(),
             "Field accesses must be precise" );
     // For oop loads, we expect the _type to be precise
+    if (OptimizeStringConcat && klass == phase->C->env()->String_klass() &&
+        adr->is_AddP() && off != Type::OffsetBot) {
+      // For constant Strings treat the fields as compile time constants.
+      Node* base = adr->in(AddPNode::Base);
+      if (base->Opcode() == Op_ConP) {
+        const TypeOopPtr* t = phase->type(base)->isa_oopptr();
+        ciObject* string = t->const_oop();
+        ciConstant constant = string->as_instance()->field_value_by_offset(off);
+        if (constant.basic_type() == T_INT) {
+          return TypeInt::make(constant.as_int());
+        } else if (constant.basic_type() == T_ARRAY) {
+          if (adr->bottom_type()->is_ptr_to_narrowoop()) {
+            return TypeNarrowOop::make_from_constant(constant.as_object());
+          } else {
+            return TypeOopPtr::make_from_constant(constant.as_object());
+          }
+        }
+      }
+    }
   } else if (tp->base() == Type::KlassPtr) {
     assert( off != Type::OffsetBot ||
             // arrays can be cast to Objects
@@ -2426,6 +2470,31 @@
   return mem;
 }
 
+//----------------------------step_through----------------------------------
+// Return allocation input memory edge if it is different instance
+// or itself if it is the one we are looking for.
+bool ClearArrayNode::step_through(Node** np, uint instance_id, PhaseTransform* phase) {
+  Node* n = *np;
+  assert(n->is_ClearArray(), "sanity");
+  intptr_t offset;
+  AllocateNode* alloc = AllocateNode::Ideal_allocation(n->in(3), phase, offset);
+  // This method is called only before Allocate nodes are expanded during
+  // macro nodes expansion. Before that ClearArray nodes are only generated
+  // in LibraryCallKit::generate_arraycopy() which follows allocations.
+  assert(alloc != NULL, "should have allocation");
+  if (alloc->_idx == instance_id) {
+    // Can not bypass initialization of the instance we are looking for.
+    return false;
+  }
+  // Otherwise skip it.
+  InitializeNode* init = alloc->initialization();
+  if (init != NULL)
+    *np = init->in(TypeFunc::Memory);
+  else
+    *np = alloc->in(TypeFunc::Memory);
+  return true;
+}
+
 //----------------------------clear_memory-------------------------------------
 // Generate code to initialize object storage to zero.
 Node* ClearArrayNode::clear_memory(Node* ctl, Node* mem, Node* dest,
@@ -2599,7 +2668,30 @@
 // Return a node which is more "ideal" than the current node.  Strip out
 // control copies
 Node *MemBarNode::Ideal(PhaseGVN *phase, bool can_reshape) {
-  return remove_dead_region(phase, can_reshape) ? this : NULL;
+  if (remove_dead_region(phase, can_reshape)) return this;
+
+  // Eliminate volatile MemBars for scalar replaced objects.
+  if (can_reshape && req() == (Precedent+1) &&
+      (Opcode() == Op_MemBarAcquire || Opcode() == Op_MemBarVolatile)) {
+    // Volatile field loads and stores.
+    Node* my_mem = in(MemBarNode::Precedent);
+    if (my_mem != NULL && my_mem->is_Mem()) {
+      const TypeOopPtr* t_oop = my_mem->in(MemNode::Address)->bottom_type()->isa_oopptr();
+      // Check for scalar replaced object reference.
+      if( t_oop != NULL && t_oop->is_known_instance_field() &&
+          t_oop->offset() != Type::OffsetBot &&
+          t_oop->offset() != Type::OffsetTop) {
+        // Replace MemBar projections by its inputs.
+        PhaseIterGVN* igvn = phase->is_IterGVN();
+        igvn->replace_node(proj_out(TypeFunc::Memory), in(TypeFunc::Memory));
+        igvn->replace_node(proj_out(TypeFunc::Control), in(TypeFunc::Control));
+        // Must return either the original node (now dead) or a new node
+        // (Do not return a top here, since that would break the uniqueness of top.)
+        return new (phase->C, 1) ConINode(TypeInt::ZERO);
+      }
+    }
+  }
+  return NULL;
 }
 
 //------------------------------Value------------------------------------------
--- a/hotspot/src/share/vm/opto/memnode.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/opto/memnode.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -717,7 +717,10 @@
 //------------------------------ClearArray-------------------------------------
 class ClearArrayNode: public Node {
 public:
-  ClearArrayNode( Node *ctrl, Node *arymem, Node *word_cnt, Node *base ) : Node(ctrl,arymem,word_cnt,base) {}
+  ClearArrayNode( Node *ctrl, Node *arymem, Node *word_cnt, Node *base )
+    : Node(ctrl,arymem,word_cnt,base) {
+    init_class_id(Class_ClearArray);
+  }
   virtual int         Opcode() const;
   virtual const Type *bottom_type() const { return Type::MEMORY; }
   // ClearArray modifies array elements, and so affects only the
@@ -743,6 +746,9 @@
                             Node* start_offset,
                             Node* end_offset,
                             PhaseGVN* phase);
+  // Return allocation input memory edge if it is different instance
+  // or itself if it is the one we are looking for.
+  static bool step_through(Node** np, uint instance_id, PhaseTransform* phase);
 };
 
 //------------------------------StrComp-------------------------------------
--- a/hotspot/src/share/vm/opto/node.hpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/opto/node.hpp	Wed Jul 05 17:05:01 2017 +0200
@@ -47,6 +47,7 @@
 class CatchNode;
 class CatchProjNode;
 class CheckCastPPNode;
+class ClearArrayNode;
 class CmpNode;
 class CodeBuffer;
 class ConstraintCastNode;
@@ -599,8 +600,9 @@
     DEFINE_CLASS_ID(BoxLock,  Node, 10)
     DEFINE_CLASS_ID(Add,      Node, 11)
     DEFINE_CLASS_ID(Mul,      Node, 12)
+    DEFINE_CLASS_ID(ClearArray, Node, 13)
 
-    _max_classes  = ClassMask_Mul
+    _max_classes  = ClassMask_ClearArray
   };
   #undef DEFINE_CLASS_ID
 
@@ -661,18 +663,25 @@
     return (_flags & Flag_is_Call) != 0;
   }
 
+  CallNode* isa_Call() const {
+    return is_Call() ? as_Call() : NULL;
+  }
+
   CallNode *as_Call() const { // Only for CallNode (not for MachCallNode)
     assert((_class_id & ClassMask_Call) == Class_Call, "invalid node class");
     return (CallNode*)this;
   }
 
-  #define DEFINE_CLASS_QUERY(type) \
-  bool is_##type() const { \
+  #define DEFINE_CLASS_QUERY(type)                           \
+  bool is_##type() const {                                   \
     return ((_class_id & ClassMask_##type) == Class_##type); \
-  } \
-  type##Node *as_##type() const { \
-    assert(is_##type(), "invalid node class"); \
-    return (type##Node*)this; \
+  }                                                          \
+  type##Node *as_##type() const {                            \
+    assert(is_##type(), "invalid node class");               \
+    return (type##Node*)this;                                \
+  }                                                          \
+  type##Node* isa_##type() const {                           \
+    return (is_##type()) ? as_##type() : NULL;               \
   }
 
   DEFINE_CLASS_QUERY(AbstractLock)
@@ -691,6 +700,7 @@
   DEFINE_CLASS_QUERY(CatchProj)
   DEFINE_CLASS_QUERY(CheckCastPP)
   DEFINE_CLASS_QUERY(ConstraintCast)
+  DEFINE_CLASS_QUERY(ClearArray)
   DEFINE_CLASS_QUERY(CMove)
   DEFINE_CLASS_QUERY(Cmp)
   DEFINE_CLASS_QUERY(CountedLoop)
@@ -1249,6 +1259,24 @@
 #undef I_VDUI_ONLY
 #undef VDUI_ONLY
 
+// An Iterator that truly follows the iterator pattern.  Doesn't
+// support deletion but could be made to.
+//
+//   for (SimpleDUIterator i(n); i.has_next(); i.next()) {
+//     Node* m = i.get();
+//
+class SimpleDUIterator : public StackObj {
+ private:
+  Node* node;
+  DUIterator_Fast i;
+  DUIterator_Fast imax;
+ public:
+  SimpleDUIterator(Node* n): node(n), i(n->fast_outs(imax)) {}
+  bool has_next() { return i < imax; }
+  void next() { i++; }
+  Node* get() { return node->fast_out(i); }
+};
+
 
 //-----------------------------------------------------------------------------
 // Map dense integer indices to Nodes.  Uses classic doubling-array trick.
@@ -1290,6 +1318,12 @@
 public:
   Node_List() : Node_Array(Thread::current()->resource_area()), _cnt(0) {}
   Node_List(Arena *a) : Node_Array(a), _cnt(0) {}
+  bool contains(Node* n) {
+    for (uint e = 0; e < size(); e++) {
+      if (at(e) == n) return true;
+    }
+    return false;
+  }
   void insert( uint i, Node *n ) { Node_Array::insert(i,n); _cnt++; }
   void remove( uint i ) { Node_Array::remove(i); _cnt--; }
   void push( Node *b ) { map(_cnt++,b); }
--- a/hotspot/src/share/vm/opto/parse1.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/opto/parse1.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -231,12 +231,13 @@
 
   // Use the raw liveness computation to make sure that unexpected
   // values don't propagate into the OSR frame.
-  MethodLivenessResult live_locals = method()->raw_liveness_at_bci(osr_bci());
+  MethodLivenessResult live_locals = method()->liveness_at_bci(osr_bci());
   if (!live_locals.is_valid()) {
     // Degenerate or breakpointed method.
     C->record_method_not_compilable("OSR in empty or breakpointed method");
     return;
   }
+  MethodLivenessResult raw_live_locals = method()->raw_liveness_at_bci(osr_bci());
 
   // Extract the needed locals from the interpreter frame.
   Node *locals_addr = basic_plus_adr(osr_buf, osr_buf, (max_locals-1)*wordSize);
@@ -316,6 +317,10 @@
         continue;
       }
     }
+    if (type->basic_type() == T_ADDRESS && !raw_live_locals.at(index)) {
+      // Skip type check for dead address locals
+      continue;
+    }
     set_local(index, check_interpreter_type(l, type, bad_type_exit));
   }
 
--- a/hotspot/src/share/vm/opto/parse3.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/opto/parse3.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -240,19 +240,19 @@
     // membar is dependent on the store, keeping any other membars generated
     // below from floating up past the store.
     int adr_idx = C->get_alias_index(adr_type);
-    insert_mem_bar_volatile(Op_MemBarVolatile, adr_idx);
+    insert_mem_bar_volatile(Op_MemBarVolatile, adr_idx, store);
 
     // Now place a membar for AliasIdxBot for the unknown yet-to-be-parsed
     // volatile alias indices. Skip this if the membar is redundant.
     if (adr_idx != Compile::AliasIdxBot) {
-      insert_mem_bar_volatile(Op_MemBarVolatile, Compile::AliasIdxBot);
+      insert_mem_bar_volatile(Op_MemBarVolatile, Compile::AliasIdxBot, store);
     }
 
     // Finally, place alias-index-specific membars for each volatile index
     // that isn't the adr_idx membar. Typically there's only 1 or 2.
     for( int i = Compile::AliasIdxRaw; i < C->num_alias_types(); i++ ) {
       if (i != adr_idx && C->alias_type(i)->is_volatile()) {
-        insert_mem_bar_volatile(Op_MemBarVolatile, i);
+        insert_mem_bar_volatile(Op_MemBarVolatile, i, store);
       }
     }
   }
--- a/hotspot/src/share/vm/opto/parseHelper.cpp	Thu Dec 17 14:10:29 2009 -0800
+++ b/hotspot/src/share/vm/opto/parseHelper.cpp	Wed Jul 05 17:05:01 2017 +0200
@@ -221,6 +221,14 @@
 
   // Push resultant oop onto stack
   push(obj);
+
+