changeset 7571:72bbdcc1d987

Merge
author asaha
date Thu, 04 Jun 2015 15:15:08 -0700
parents 89174fe1856f 67f2485a64d4
children ba5d2ce8d8e4
files .hgtags agent/src/share/classes/sun/jvm/hotspot/memory/FreeList.java make/hotspot_version make/linux/makefiles/arm.make make/linux/makefiles/build_vm_def.sh make/linux/makefiles/ppc.make make/linux/platform_arm make/linux/platform_ppc test/gc/g1/TestEagerReclaimHumongousRegions2.java
diffstat 362 files changed, 9566 insertions(+), 4772 deletions(-) [+]
line wrap: on
line diff
--- a/.hgtags	Thu May 07 18:13:43 2015 -0700
+++ b/.hgtags	Thu Jun 04 15:15:08 2015 -0700
@@ -658,6 +658,48 @@
 ce81c4487dd1e9f89d4570a8cd25e349f6bae00d jdk8u51-b09
 928e1994ad43272f808ca22b9cc1b08a7ce2824f jdk8u51-b10
 1a122beb9dc6881850ef1d1250f40a83709b8b72 jdk8u51-b11
+05c80f1060f0c0d5720de9eadd09162af1168eab jdk8u51-b12
+07e103f3f43886a3b47945e5295eb5accad505de jdk8u51-b13
+a4eea4bee2d4fdb05f1a8358d70ec6adb1135526 jdk8u51-b14
+9a70cba6a3c3e44486f9c199d03a16b2b09d0a13 jdk8u51-b15
+d9349fa8822336e0244da0a8448f3e6b2d62741d jdk8u60-b00
+d9349fa8822336e0244da0a8448f3e6b2d62741d hs25.60-b00
+ebf89088c08ab0508b9002b48dd3d68a340259af hs25.60-b01
+5fa73007ceb92a13742fc4a24ec935a6494f8045 hs25.60-b02
+702cc6067686acaa45f7b455b7490edc056c2ae0 jdk8u60-b01
+1f6ba0d2923dadba87aac4ed779dd1ed0161ec2b hs25.60-b03
+38f6080523831ae9a6907c780f2042b82f3213ca jdk8u60-b02
+9d6eb2757167744a17ea71f8b860430d70941eda jdk8u60-b03
+0fb1ac49ae7764c5d7c6dfb9fe046d0e1a4eb5aa hs25.60-b04
+586a449cd30332dd53c0f74bf2ead6f3d4724bfc jdk8u60-b04
+74931e85352be8556eaa511ca0dd7c38fe272ec3 hs25.60-b05
+b13f1890afb8abc31ecb9c21fd2ba95aba3e33f8 jdk8u60-b05
+b17a8a22a0344e3c93e2e4677de20d35f99cf4f5 hs25.60-b06
+7b70923c8e04920b60278f90ad23a63c773cee7b jdk8u60-b06
+d51ef6da82b486e7b2b3c08eef9ca0a186935ded hs25.60-b07
+353e580ce6878d80c7b7cd27f8ad24609b12c58b jdk8u60-b07
+a72a4192a36d6d84766d6135fe6515346c742007 hs25.60-b08
+bf68e15dc8fe73eeb1eb3c656df51fdb1f707a97 jdk8u60-b08
+d937e6a0674841d670232ecf1611f52e1ae998e7 hs25.60-b09
+f1058b5c6294235d8ad032dcc72c8f8bc202cb5a jdk8u60-b09
+57a14c3927eba6372d909ae164fa90bb9b6a6ce4 hs25.60-b10
+8e4518dc2b38957072704ffe4cbf29f046dc9325 jdk8u60-b10
+64a32bc18e88eed6131ed036dc3e10e566ef339b hs25.60-b11
+d8f133adf05d310bd7e1d9adf32cbeb71ff33c37 jdk8u60-b11
+4390345de45c7768c04bfafabf006a401824c5b5 hs25.60-b12
+ccca7162738eee1be74890342c67d3b26540dcf6 jdk8u60-b12
+ced08ed4924fc6581626c7ce2d769fc18d7b23e0 jdk8u60-b13
+30e04eba9e298cc5094793e279306535239187cc hs25.60-b13
+1f0d760ccac1ff82a03a9b7d6bd5c697ef0a7c4a hs25.60-b14
+c9f8b7319d0a5ab07310cf53507642a8fd91589b jdk8u60-b14
+4187dc92e90b16b4097627b8af4f5e6e63f3b497 hs25.60-b15
+b99f1bf208f385277b03a985d35b6614b4095f3e jdk8u60-b15
+f5800068c61d0627c14e99836e9ce5cf0ef00075 hs25.60-b16
+ab2353694ea7fd4907c5c88b8334f8feaafca8c7 jdk8u60-b16
+5efc25c367164b6856554b0d625f3c422fdf9558 hs25.60-b17
+c26d09f1065cd26bd8b926efc5d3938b71e09eb5 jdk8u60-b17
+624f4cc05e7e95dd2103f343c54d7bdea6a81919 hs25.60-b18
+3fa5c654c143fe309e5ddda92adc5fb132365bcf jdk8u60-b18
 0219ab69f00782e5c49687e2fa75138a7ffddea1 jdk8u52-b06
 9b6f44853eed8caba935915c7e710c546b205c8e jdk8u52-b07
 0219ab69f00782e5c49687e2fa75138a7ffddea1 jdk8u65-b00
--- a/agent/src/os/bsd/MacosxDebuggerLocal.m	Thu May 07 18:13:43 2015 -0700
+++ b/agent/src/os/bsd/MacosxDebuggerLocal.m	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2002, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2002, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -26,7 +26,7 @@
 #import <Foundation/Foundation.h>
 #import <JavaNativeFoundation/JavaNativeFoundation.h>
 
-#include <JavaVM/jni.h>
+#include <jni.h>
 
 #import <mach/mach.h>
 #import <mach/mach_types.h>
--- a/agent/src/os/bsd/Makefile	Thu May 07 18:13:43 2015 -0700
+++ b/agent/src/os/bsd/Makefile	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 #
-# Copyright (c) 2002, 2013, Oracle and/or its affiliates. All rights reserved.
+# Copyright (c) 2002, 2015, Oracle and/or its affiliates. All rights reserved.
 # DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 #
 # This code is free software; you can redistribute it and/or modify it
@@ -50,9 +50,9 @@
 	ps_core.c
 OBJS    = $(SOURCES:.c=.o)
 OBJSPLUS = MacosxDebuggerLocal.o sadis.o $(OBJS)
-EXTINCLUDE = -I/System/Library/Frameworks/JavaVM.framework/Headers -I.
+EXTINCLUDE = -I.
 EXTCFLAGS = -m64 -D__APPLE__ -framework JavaNativeFoundation
-FOUNDATIONFLAGS = -framework Foundation -F/System/Library/Frameworks/JavaVM.framework/Frameworks -framework JavaNativeFoundation -framework Security -framework CoreFoundation
+FOUNDATIONFLAGS = -framework Foundation -framework JavaNativeFoundation -framework Security -framework CoreFoundation
 LIBSA = $(ARCH)/libsaproc.dylib
 endif   # Darwin
 
--- a/agent/src/os/linux/libproc.h	Thu May 07 18:13:43 2015 -0700
+++ b/agent/src/os/linux/libproc.h	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2003, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2003, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -30,7 +30,7 @@
 #include <stdint.h>
 #include "proc_service.h"
 
-#if defined(arm) || defined(ppc)
+#ifdef ALT_SASRCDIR
 #include "libproc_md.h"
 #endif
 
--- a/agent/src/os/linux/ps_proc.c	Thu May 07 18:13:43 2015 -0700
+++ b/agent/src/os/linux/ps_proc.c	Thu Jun 04 15:15:08 2015 -0700
@@ -27,9 +27,11 @@
 #include <string.h>
 #include <signal.h>
 #include <errno.h>
+#include <elf.h>
 #include <sys/types.h>
 #include <sys/wait.h>
 #include <sys/ptrace.h>
+#include <sys/uio.h>
 #include "libproc_impl.h"
 
 #if defined(x86_64) && !defined(amd64)
@@ -138,6 +140,15 @@
    return false;
  }
  return true;
+#elif defined(PTRACE_GETREGSET)
+ struct iovec iov;
+ iov.iov_base = user;
+ iov.iov_len = sizeof(*user);
+ if (ptrace(PTRACE_GETREGSET, pid, NT_PRSTATUS, (void*) &iov) < 0) {
+   print_debug("ptrace(PTRACE_GETREGSET, ...) failed for lwp %d\n", pid);
+   return false;
+ }
+ return true;
 #else
  print_debug("ptrace(PTRACE_GETREGS, ...) not supported\n");
  return false;
--- a/agent/src/os/solaris/proc/saproc.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/agent/src/os/solaris/proc/saproc.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -314,7 +314,7 @@
     handle = dlopen(name, mode);
   }
   if (_libsaproc_debug) {
-    printf("libsaproc DEBUG: pathmap_dlopen %s return 0x%x\n", name, handle);
+    printf("libsaproc DEBUG: pathmap_dlopen %s return 0x%lx\n", name, (unsigned long) handle);
   }
   return handle;
 }
@@ -661,30 +661,30 @@
   // read FileMapHeader
   size_t n = read(fd, pheader, sizeof(struct FileMapHeader));
   if (n != sizeof(struct FileMapHeader)) {
-    free(pheader);
-    close(fd);
     char errMsg[ERR_MSG_SIZE];
     sprintf(errMsg, "unable to read shared archive file map header from %s", classes_jsa);
+    close(fd);
+    free(pheader);
     THROW_NEW_DEBUGGER_EXCEPTION_(errMsg, 1);
   }
 
   // check file magic
   if (pheader->_magic != 0xf00baba2) {
-    free(pheader);
-    close(fd);
     char errMsg[ERR_MSG_SIZE];
     sprintf(errMsg, "%s has bad shared archive magic 0x%x, expecting 0xf00baba2",
                    classes_jsa, pheader->_magic);
+    close(fd);
+    free(pheader);
     THROW_NEW_DEBUGGER_EXCEPTION_(errMsg, 1);
   }
 
   // check version
   if (pheader->_version != CURRENT_ARCHIVE_VERSION) {
-    free(pheader);
-    close(fd);
     char errMsg[ERR_MSG_SIZE];
     sprintf(errMsg, "%s has wrong shared archive version %d, expecting %d",
                    classes_jsa, pheader->_version, CURRENT_ARCHIVE_VERSION);
+    close(fd);
+    free(pheader);
     THROW_NEW_DEBUGGER_EXCEPTION_(errMsg, 1);
   }
 
--- a/agent/src/share/classes/com/sun/java/swing/action/ActionManager.java	Thu May 07 18:13:43 2015 -0700
+++ b/agent/src/share/classes/com/sun/java/swing/action/ActionManager.java	Thu Jun 04 15:15:08 2015 -0700
@@ -46,6 +46,11 @@
         return manager;
     }
 
+    protected static void setInstance(ActionManager m)
+    {
+        manager = m;
+    }
+
     protected abstract void addActions();
 
     protected void addAction(String cmdname, Action action)
@@ -90,6 +95,6 @@
 
     private HashMap actions;
     private static ActionUtilities utilities = new ActionUtilities();
-    protected static ActionManager manager;
+    private static ActionManager manager;
 
 }
--- a/agent/src/share/classes/com/sun/java/swing/ui/CommonToolBar.java	Thu May 07 18:13:43 2015 -0700
+++ b/agent/src/share/classes/com/sun/java/swing/ui/CommonToolBar.java	Thu Jun 04 15:15:08 2015 -0700
@@ -46,7 +46,7 @@
     {
         this.manager = manager;
         statusBar = status;
-        buttonSize = new Dimension(CommonUI.buttconPrefSize);
+        buttonSize = new Dimension(CommonUI.getButtconPrefSize());
         buttonInsets = new Insets(0, 0, 0, 0);
         addComponents();
     }
--- a/agent/src/share/classes/com/sun/java/swing/ui/CommonUI.java	Thu May 07 18:13:43 2015 -0700
+++ b/agent/src/share/classes/com/sun/java/swing/ui/CommonUI.java	Thu Jun 04 15:15:08 2015 -0700
@@ -373,20 +373,25 @@
         comp.setCursor(Cursor.getPredefinedCursor(0));
     }
 
-    public static final int BUTTON_WIDTH = 100;
-    public static final int BUTTON_HEIGHT = 26;
-    public static final int BUTTCON_WIDTH = 28;
-    public static final int BUTTCON_HEIGHT = 28;
-    public static final int SM_BUTTON_WIDTH = 72;
-    public static final int SM_BUTTON_HEIGHT = 26;
-    public static final int LABEL_WIDTH = 100;
-    public static final int LABEL_HEIGHT = 20;
-    public static final int TEXT_WIDTH = 150;
-    public static final int TEXT_HEIGHT = 20;
-    public static Dimension buttonPrefSize = new Dimension(100, 26);
-    public static Dimension buttconPrefSize = new Dimension(28, 28);
-    public static Dimension smbuttonPrefSize = new Dimension(72, 26);
-    public static Dimension labelPrefSize = new Dimension(100, 20);
-    public static Dimension textPrefSize = new Dimension(150, 20);
+    public static Dimension getButtconPrefSize()
+    {
+        return buttconPrefSize;
+    }
+
+    private static final int BUTTON_WIDTH = 100;
+    private static final int BUTTON_HEIGHT = 26;
+    private static final int BUTTCON_WIDTH = 28;
+    private static final int BUTTCON_HEIGHT = 28;
+    private static final int SM_BUTTON_WIDTH = 72;
+    private static final int SM_BUTTON_HEIGHT = 26;
+    private static final int LABEL_WIDTH = 100;
+    private static final int LABEL_HEIGHT = 20;
+    private static final int TEXT_WIDTH = 150;
+    private static final int TEXT_HEIGHT = 20;
+    private static final Dimension buttonPrefSize = new Dimension(100, 26);
+    private static final Dimension buttconPrefSize = new Dimension(28, 28);
+    private static final Dimension smbuttonPrefSize = new Dimension(72, 26);
+    private static final Dimension labelPrefSize = new Dimension(100, 20);
+    private static final Dimension textPrefSize = new Dimension(150, 20);
 
 }
--- a/agent/src/share/classes/sun/jvm/hotspot/HotSpotTypeDataBase.java	Thu May 07 18:13:43 2015 -0700
+++ b/agent/src/share/classes/sun/jvm/hotspot/HotSpotTypeDataBase.java	Thu Jun 04 15:15:08 2015 -0700
@@ -51,6 +51,9 @@
   private static final int C_INT32_SIZE = 4;
   private static final int C_INT64_SIZE = 8;
   private static int pointerSize = UNINITIALIZED_SIZE;
+  // Counter to ensure read loops terminate:
+  private static final int MAX_DUPLICATE_DEFINITIONS = 100;
+  private int duplicateDefCount = 0;
 
   private static final boolean DEBUG;
   static {
@@ -166,6 +169,10 @@
     typeEntrySizeOffset           = getLongValueFromProcess("gHotSpotVMTypeEntrySizeOffset");
     typeEntryArrayStride          = getLongValueFromProcess("gHotSpotVMTypeEntryArrayStride");
 
+    if (typeEntryArrayStride == 0L) {
+      throw new RuntimeException("zero stride: cannot read types.");
+    }
+
     // Start iterating down it until we find an entry with no name
     Address typeNameAddr = null;
     do {
@@ -192,7 +199,11 @@
       }
 
       entryAddr = entryAddr.addOffsetTo(typeEntryArrayStride);
-    } while (typeNameAddr != null);
+    } while (typeNameAddr != null && duplicateDefCount < MAX_DUPLICATE_DEFINITIONS);
+
+    if (duplicateDefCount >= MAX_DUPLICATE_DEFINITIONS) {
+      throw new RuntimeException("too many duplicate definitions");
+    }
   }
 
   private void initializePrimitiveTypes() {
@@ -395,6 +406,10 @@
     structEntryAddressOffset      = getLongValueFromProcess("gHotSpotVMStructEntryAddressOffset");
     structEntryArrayStride        = getLongValueFromProcess("gHotSpotVMStructEntryArrayStride");
 
+    if (structEntryArrayStride == 0L) {
+      throw new RuntimeException("zero stride: cannot read types.");
+    }
+
     // Fetch the address of the VMStructEntry*
     Address entryAddr = lookupInProcess("gHotSpotVMStructs");
     // Dereference this once to get the pointer to the first VMStructEntry
@@ -472,6 +487,11 @@
     intConstantEntryValueOffset = getLongValueFromProcess("gHotSpotVMIntConstantEntryValueOffset");
     intConstantEntryArrayStride = getLongValueFromProcess("gHotSpotVMIntConstantEntryArrayStride");
 
+    if (intConstantEntryArrayStride == 0L) {
+      throw new RuntimeException("zero stride: cannot read types.");
+    }
+
+
     // Fetch the address of the VMIntConstantEntry*
     Address entryAddr = lookupInProcess("gHotSpotVMIntConstants");
     // Dereference this once to get the pointer to the first VMIntConstantEntry
@@ -501,12 +521,17 @@
           } else {
             System.err.println("Warning: the int constant \"" + name + "\" (declared in the remote VM in VMStructs::localHotSpotVMIntConstants) " +
                                "had its value declared as " + value + " twice. Continuing.");
+            duplicateDefCount++;
           }
         }
       }
 
       entryAddr = entryAddr.addOffsetTo(intConstantEntryArrayStride);
-    } while (nameAddr != null);
+    } while (nameAddr != null && duplicateDefCount < MAX_DUPLICATE_DEFINITIONS);
+
+    if (duplicateDefCount >= MAX_DUPLICATE_DEFINITIONS) {
+      throw new RuntimeException("too many duplicate definitions");
+    }
   }
 
   private void readVMLongConstants() {
@@ -519,6 +544,10 @@
     longConstantEntryValueOffset = getLongValueFromProcess("gHotSpotVMLongConstantEntryValueOffset");
     longConstantEntryArrayStride = getLongValueFromProcess("gHotSpotVMLongConstantEntryArrayStride");
 
+    if (longConstantEntryArrayStride == 0L) {
+      throw new RuntimeException("zero stride: cannot read types.");
+    }
+
     // Fetch the address of the VMLongConstantEntry*
     Address entryAddr = lookupInProcess("gHotSpotVMLongConstants");
     // Dereference this once to get the pointer to the first VMLongConstantEntry
@@ -548,12 +577,17 @@
           } else {
             System.err.println("Warning: the long constant \"" + name + "\" (declared in the remote VM in VMStructs::localHotSpotVMLongConstants) " +
                                "had its value declared as " + value + " twice. Continuing.");
+            duplicateDefCount++;
           }
         }
       }
 
       entryAddr = entryAddr.addOffsetTo(longConstantEntryArrayStride);
-    } while (nameAddr != null);
+    } while (nameAddr != null && duplicateDefCount < MAX_DUPLICATE_DEFINITIONS);
+
+    if (duplicateDefCount >= MAX_DUPLICATE_DEFINITIONS) {
+      throw new RuntimeException("too many duplicate definitions.");
+    }
   }
 
   private BasicType lookupOrFail(String typeName) {
@@ -740,9 +774,10 @@
             }
 
             if (!typeNameIsPointerType(typeName)) {
-            System.err.println("Warning: the type \"" + typeName + "\" (declared in the remote VM in VMStructs::localHotSpotVMTypes) " +
-                               "had its size declared as " + size + " twice. Continuing.");
-        }
+                System.err.println("Warning: the type \"" + typeName + "\" (declared in the remote VM in VMStructs::localHotSpotVMTypes) " +
+                                   "had its size declared as " + size + " twice. Continuing.");
+                duplicateDefCount++;
+            }
         }
 
     }
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/agent/src/share/classes/sun/jvm/hotspot/gc_interface/G1YCType.java	Thu Jun 04 15:15:08 2015 -0700
@@ -0,0 +1,45 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+package sun.jvm.hotspot.gc_interface;
+
+//These definitions should be kept in sync with the definitions in the HotSpot
+//code.
+
+public enum G1YCType {
+  Normal ("Normal"),
+  InitialMark ("Initial Mark"),
+  DuringMark ("During Mark"),
+  Mixed ("Mixed"),
+  G1YCTypeEndSentinel ("Unknown");
+
+  private final String value;
+
+  G1YCType(String val) {
+    this.value = val;
+  }
+  public String value() {
+    return value;
+  }
+}
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/agent/src/share/classes/sun/jvm/hotspot/gc_interface/GCCause.java	Thu Jun 04 15:15:08 2015 -0700
@@ -0,0 +1,69 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+package sun.jvm.hotspot.gc_interface;
+
+//These definitions should be kept in sync with the definitions in the HotSpot code.
+
+public enum GCCause {
+  _java_lang_system_gc ("System.gc()"),
+  _full_gc_alot ("FullGCAlot"),
+  _scavenge_alot ("ScavengeAlot"),
+  _allocation_profiler ("Allocation Profiler"),
+  _jvmti_force_gc ("JvmtiEnv ForceGarbageCollection"),
+  _gc_locker ("GCLocker Initiated GC"),
+  _heap_inspection ("Heap Inspection Initiated GC"),
+  _heap_dump ("Heap Dump Initiated GC"),
+
+  _no_gc ("No GC"),
+  _no_cause_specified ("Unknown GCCause"),
+  _allocation_failure ("Allocation Failure"),
+
+  _tenured_generation_full ("Tenured Generation Full"),
+  _metadata_GC_threshold ("Metadata GC Threshold"),
+
+  _cms_generation_full ("CMS Generation Full"),
+  _cms_initial_mark ("CMS Initial Mark"),
+  _cms_final_remark ("CMS Final Remark"),
+  _cms_concurrent_mark ("CMS Concurrent Mark"),
+
+  _old_generation_expanded_on_last_scavenge ("Old Generation Expanded On Last Scavenge"),
+  _old_generation_too_full_to_scavenge ("Old Generation Too Full To Scavenge"),
+  _adaptive_size_policy ("Ergonomics"),
+
+  _g1_inc_collection_pause ("G1 Evacuation Pause"),
+  _g1_humongous_allocation ("G1 Humongous Allocation"),
+
+  _last_ditch_collection ("Last ditch collection"),
+  _last_gc_cause ("ILLEGAL VALUE - last gc cause - ILLEGAL VALUE");
+
+  private final String value;
+
+  GCCause(String val) {
+    this.value = val;
+  }
+  public String value() {
+    return value;
+  }
+}
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/agent/src/share/classes/sun/jvm/hotspot/gc_interface/GCName.java	Thu Jun 04 15:15:08 2015 -0700
@@ -0,0 +1,50 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+package sun.jvm.hotspot.gc_interface;
+
+//These definitions should be kept in sync with the definitions in the HotSpot code.
+
+public enum GCName {
+  ParallelOld ("ParallelOld"),
+  SerialOld ("SerialOld"),
+  PSMarkSweep ("PSMarkSweep"),
+  ParallelScavenge ("ParallelScavenge"),
+  DefNew ("DefNew"),
+  ParNew ("ParNew"),
+  G1New ("G1New"),
+  ConcurrentMarkSweep ("ConcurrentMarkSweep"),
+  G1Old ("G1Old"),
+  GCNameEndSentinel ("GCNameEndSentinel");
+
+  private final String value;
+
+  GCName(String val) {
+    this.value = val;
+  }
+  public String value() {
+    return value;
+  }
+}
+
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/agent/src/share/classes/sun/jvm/hotspot/gc_interface/GCWhen.java	Thu Jun 04 15:15:08 2015 -0700
@@ -0,0 +1,45 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+package sun.jvm.hotspot.gc_interface;
+
+//These definitions should be kept in sync with the definitions in the HotSpot code.
+
+public enum GCWhen {
+  BeforeGC ("Before GC"),
+  AfterGC ("After GC"),
+  GCWhenEndSentinel ("GCWhenEndSentinel");
+
+  private final String value;
+
+  GCWhen(String val) {
+    this.value = val;
+  }
+  public String value() {
+    return value;
+  }
+}
+
+
+
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/agent/src/share/classes/sun/jvm/hotspot/gc_interface/ReferenceType.java	Thu Jun 04 15:15:08 2015 -0700
@@ -0,0 +1,45 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+package sun.jvm.hotspot.gc_interface;
+
+//These definitions should be kept in sync with the definitions in the HotSpot code.
+
+public enum ReferenceType {
+  REF_NONE ("None reference"),       // Regular class
+  REF_OTHER ("Other reference"),     // Subclass of java/lang/ref/Reference, but not subclass of one of the classes below
+  REF_SOFT ("Soft reference"),       // Subclass of java/lang/ref/SoftReference
+  REF_WEAK ("Weak reference"),       // Subclass of java/lang/ref/WeakReference
+  REF_FINAL ("Final reference"),     // Subclass of java/lang/ref/FinalReference
+  REF_PHANTOM ("Phantom reference"); // Subclass of java/lang/ref/PhantomReference
+
+  private final String value;
+
+  ReferenceType(String val) {
+    this.value = val;
+  }
+  public String value() {
+    return value;
+  }
+}
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/agent/src/share/classes/sun/jvm/hotspot/memory/AdaptiveFreeList.java	Thu Jun 04 15:15:08 2015 -0700
@@ -0,0 +1,77 @@
+/*
+ * @(#)AdaptiveFreeList.java
+ *
+ * Copyright (c) 2000, 2014, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+package sun.jvm.hotspot.memory;
+
+import java.util.Observable;
+import java.util.Observer;
+
+import sun.jvm.hotspot.debugger.Address;
+import sun.jvm.hotspot.runtime.VM;
+import sun.jvm.hotspot.runtime.VMObject;
+import sun.jvm.hotspot.types.CIntegerField;
+import sun.jvm.hotspot.types.Type;
+import sun.jvm.hotspot.types.TypeDataBase;
+
+public class AdaptiveFreeList extends VMObject {
+  static {
+    VM.registerVMInitializedObserver(new Observer() {
+      public void update(Observable o, Object data) {
+        initialize(VM.getVM().getTypeDataBase());
+      }
+    });
+  }
+
+  private static synchronized void initialize(TypeDataBase db) {
+    Type type = db.lookupType("AdaptiveFreeList<FreeChunk>");
+    sizeField = type.getCIntegerField("_size");
+    countField = type.getCIntegerField("_count");
+    headerSize = type.getSize();
+  }
+
+  // Fields
+  private static CIntegerField sizeField;
+  private static CIntegerField countField;
+  private static long          headerSize;
+
+  //Constructor
+  public AdaptiveFreeList(Address address) {
+    super(address);
+  }
+
+  // Accessors
+  public long size() {
+    return sizeField.getValue(addr);
+  }
+
+  public long count() {
+    return  countField.getValue(addr);
+  }
+
+  public static long sizeOf() {
+    return headerSize;
+  }
+}
--- a/agent/src/share/classes/sun/jvm/hotspot/memory/CompactibleFreeListSpace.java	Thu May 07 18:13:43 2015 -0700
+++ b/agent/src/share/classes/sun/jvm/hotspot/memory/CompactibleFreeListSpace.java	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2003, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2003, 2014, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -24,25 +24,29 @@
 
 package sun.jvm.hotspot.memory;
 
-import java.io.*;
-import java.util.*;
-import sun.jvm.hotspot.debugger.*;
-import sun.jvm.hotspot.oops.*;
-import sun.jvm.hotspot.runtime.*;
-import sun.jvm.hotspot.types.*;
-import sun.jvm.hotspot.utilities.*;
+import java.io.PrintStream;
+import java.util.ArrayList;
+import java.util.Iterator;
+import java.util.List;
+import java.util.Observable;
+import java.util.Observer;
+
+import sun.jvm.hotspot.debugger.Address;
+import sun.jvm.hotspot.debugger.Debugger;
+import sun.jvm.hotspot.oops.ObjectHeap;
+import sun.jvm.hotspot.oops.Oop;
+import sun.jvm.hotspot.runtime.VM;
+import sun.jvm.hotspot.runtime.VMObjectFactory;
+import sun.jvm.hotspot.types.AddressField;
+import sun.jvm.hotspot.types.Type;
+import sun.jvm.hotspot.types.TypeDataBase;
+import sun.jvm.hotspot.utilities.Assert;
 
 public class CompactibleFreeListSpace extends CompactibleSpace {
    private static AddressField collectorField;
-
-   // for free size, three fields
-   //       FreeBlockDictionary* _dictionary;        // ptr to dictionary for large size blocks
-   //       FreeList _indexedFreeList[IndexSetSize]; // indexed array for small size blocks
-   //       LinearAllocBlock _smallLinearAllocBlock; // small linear alloc in TLAB
    private static AddressField indexedFreeListField;
    private static AddressField dictionaryField;
    private static long         smallLinearAllocBlockFieldOffset;
-   private static long indexedFreeListSizeOf;
 
    private int    heapWordSize;     // 4 for 32bit, 8 for 64 bits
    private int    IndexSetStart;    // for small indexed list
@@ -109,11 +113,11 @@
       // small chunks
       long size = 0;
       Address cur = addr.addOffsetTo( indexedFreeListField.getOffset() );
-      cur = cur.addOffsetTo(IndexSetStart*FreeList.sizeOf());
+      cur = cur.addOffsetTo(IndexSetStart*AdaptiveFreeList.sizeOf());
       for (int i=IndexSetStart; i<IndexSetSize; i += IndexSetStride) {
-         FreeList freeList = (FreeList) VMObjectFactory.newObject(FreeList.class, cur);
+         AdaptiveFreeList freeList = (AdaptiveFreeList) VMObjectFactory.newObject(AdaptiveFreeList.class, cur);
          size += i*freeList.count();
-         cur= cur.addOffsetTo(IndexSetStride*FreeList.sizeOf());
+         cur= cur.addOffsetTo(IndexSetStride*AdaptiveFreeList.sizeOf());
       }
 
       // large block
--- a/agent/src/share/classes/sun/jvm/hotspot/memory/FreeList.java	Thu May 07 18:13:43 2015 -0700
+++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
@@ -1,72 +0,0 @@
-/*
- * @(#)FreeList.java
- *
- * Copyright (c) 2000, 2013, Oracle and/or its affiliates. All rights reserved.
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- *
- */
-
-package sun.jvm.hotspot.memory;
-
-import java.util.*;
-import sun.jvm.hotspot.debugger.*;
-import sun.jvm.hotspot.types.*;
-import sun.jvm.hotspot.runtime.*;
-
-public class FreeList extends VMObject {
-   static {
-      VM.registerVMInitializedObserver(new Observer() {
-         public void update(Observable o, Object data) {
-            initialize(VM.getVM().getTypeDataBase());
-         }
-      });
-   }
-
-   private static synchronized void initialize(TypeDataBase db) {
-      Type type = db.lookupType("FreeList<FreeChunk>");
-      sizeField = type.getCIntegerField("_size");
-      countField = type.getCIntegerField("_count");
-      headerSize = type.getSize();
-   }
-
-   // Fields
-   private static CIntegerField sizeField;
-   private static CIntegerField countField;
-   private static long          headerSize;
-
-   //Constructor
-   public FreeList(Address address) {
-     super(address);
-   }
-
-   // Accessors
-   public long size() {
-      return sizeField.getValue(addr);
-   }
-
-   public long count() {
-      return  countField.getValue(addr);
-   }
-
-   public static long sizeOf() {
-     return headerSize;
-  }
-}
--- a/agent/src/share/classes/sun/jvm/hotspot/memory/Universe.java	Thu May 07 18:13:43 2015 -0700
+++ b/agent/src/share/classes/sun/jvm/hotspot/memory/Universe.java	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000, 2012, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2000, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -56,6 +56,12 @@
   private static AddressField narrowKlassBaseField;
   private static CIntegerField narrowKlassShiftField;
 
+  public enum NARROW_OOP_MODE {
+    UnscaledNarrowOop,
+    ZeroBasedNarrowOop,
+    HeapBasedNarrowOop
+  }
+
   static {
     VM.registerVMInitializedObserver(new Observer() {
         public void update(Observable o, Object data) {
@@ -94,7 +100,17 @@
 
   public Universe() {
   }
-
+  public static String narrowOopModeToString(NARROW_OOP_MODE mode) {
+    switch (mode) {
+    case UnscaledNarrowOop:
+      return "32-bits Oops";
+    case ZeroBasedNarrowOop:
+      return "zero based Compressed Oops";
+    case HeapBasedNarrowOop:
+      return "Compressed Oops with base";
+    }
+    return "";
+  }
   public CollectedHeap heap() {
     try {
       return (CollectedHeap) heapConstructor.instantiateWrapperFor(collectedHeapField.getValue());
--- a/agent/src/share/classes/sun/jvm/hotspot/oops/Klass.java	Thu May 07 18:13:43 2015 -0700
+++ b/agent/src/share/classes/sun/jvm/hotspot/oops/Klass.java	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2000, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -55,6 +55,10 @@
     layoutHelper = new IntField(type.getJIntField("_layout_helper"), 0);
     name         = type.getAddressField("_name");
     accessFlags  = new CIntField(type.getCIntegerField("_access_flags"), 0);
+    try {
+      traceIDField  = type.getField("_trace_id");
+    } catch(Exception e) {
+    }
     subklass     = new MetadataField(type.getAddressField("_subklass"), 0);
     nextSibling  = new MetadataField(type.getAddressField("_next_sibling"), 0);
 
@@ -86,6 +90,7 @@
   private static CIntField accessFlags;
   private static MetadataField  subklass;
   private static MetadataField  nextSibling;
+  private static sun.jvm.hotspot.types.Field traceIDField;
 
   private Address getValue(AddressField field) {
     return addr.getAddressAt(field.getOffset());
@@ -107,6 +112,11 @@
   public Klass    getSubklassKlass()    { return (Klass)    subklass.getValue(this);     }
   public Klass    getNextSiblingKlass() { return (Klass)    nextSibling.getValue(this);  }
 
+  public long traceID() {
+    if (traceIDField == null) return 0;
+    return traceIDField.getJLong(addr);
+  }
+
   // computed access flags - takes care of inner classes etc.
   // This is closer to actual source level than getAccessFlags() etc.
   public long computeModifierFlags() {
--- a/agent/src/share/classes/sun/jvm/hotspot/oops/OopUtilities.java	Thu May 07 18:13:43 2015 -0700
+++ b/agent/src/share/classes/sun/jvm/hotspot/oops/OopUtilities.java	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000, 2012, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2000, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -54,6 +54,8 @@
   private static OopField threadNameField;
   private static OopField threadGroupField;
   private static LongField threadEETopField;
+  //tid field is new since 1.5
+  private static LongField threadTIDField;
   // threadStatus field is new since 1.5
   private static IntField threadStatusField;
   // parkBlocker field is new since 1.6
@@ -220,6 +222,7 @@
       threadNameField  = (OopField) k.findField("name", "[C");
       threadGroupField = (OopField) k.findField("group", "Ljava/lang/ThreadGroup;");
       threadEETopField = (LongField) k.findField("eetop", "J");
+      threadTIDField = (LongField) k.findField("tid", "J");
       threadStatusField = (IntField) k.findField("threadStatus", "I");
       threadParkBlockerField = (OopField) k.findField("parkBlocker",
                                      "Ljava/lang/Object;");
@@ -268,6 +271,15 @@
     return VM.getVM().getThreads().createJavaThreadWrapper(addr);
   }
 
+  public static long threadOopGetTID(Oop threadOop) {
+    initThreadFields();
+    if (threadTIDField != null) {
+      return threadTIDField.getValue(threadOop);
+    } else {
+      return 0;
+    }
+  }
+
   /** returns value of java.lang.Thread.threadStatus field */
   public static int threadOopGetThreadStatus(Oop threadOop) {
     initThreadFields();
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/agent/src/share/classes/sun/jvm/hotspot/opto/CompilerPhaseType.java	Thu Jun 04 15:15:08 2015 -0700
@@ -0,0 +1,67 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+package sun.jvm.hotspot.opto;
+
+//These definitions should be kept in sync with the definitions in the HotSpot code.
+
+public enum CompilerPhaseType {
+  PHASE_BEFORE_STRINGOPTS ("Before StringOpts"),
+  PHASE_AFTER_STRINGOPTS ("After StringOpts"),
+  PHASE_BEFORE_REMOVEUSELESS ("Before RemoveUseless"),
+  PHASE_AFTER_PARSING ("After Parsing"),
+  PHASE_ITER_GVN1 ("Iter GVN 1"),
+  PHASE_PHASEIDEAL_BEFORE_EA ("PhaseIdealLoop before EA"),
+  PHASE_ITER_GVN_AFTER_EA ("Iter GVN after EA"),
+  PHASE_ITER_GVN_AFTER_ELIMINATION ("Iter GVN after eliminating allocations and locks"),
+  PHASE_PHASEIDEALLOOP1 ("PhaseIdealLoop 1"),
+  PHASE_PHASEIDEALLOOP2 ("PhaseIdealLoop 2"),
+  PHASE_PHASEIDEALLOOP3 ("PhaseIdealLoop 3"),
+  PHASE_CPP1 ("PhaseCPP 1"),
+  PHASE_ITER_GVN2 ("Iter GVN 2"),
+  PHASE_PHASEIDEALLOOP_ITERATIONS ("PhaseIdealLoop iterations"),
+  PHASE_OPTIMIZE_FINISHED ("Optimize finished"),
+  PHASE_GLOBAL_CODE_MOTION ("Global code motion"),
+  PHASE_FINAL_CODE ("Final Code"),
+  PHASE_AFTER_EA ("After Escape Analysis"),
+  PHASE_BEFORE_CLOOPS ("Before CountedLoop"),
+  PHASE_AFTER_CLOOPS ("After CountedLoop"),
+  PHASE_BEFORE_BEAUTIFY_LOOPS ("Before beautify loops"),
+  PHASE_AFTER_BEAUTIFY_LOOPS ("After beautify loops"),
+  PHASE_BEFORE_MATCHING ("Before Matching"),
+  PHASE_INCREMENTAL_INLINE ("Incremental Inline"),
+  PHASE_INCREMENTAL_BOXING_INLINE ("Incremental Boxing Inline"),
+  PHASE_END ("End"),
+  PHASE_FAILURE ("Failure"),
+  PHASE_NUM_TYPES ("Number of Phase Types");
+
+  private final String value;
+
+  CompilerPhaseType(String val) {
+    this.value = val;
+  }
+  public String value() {
+    return value;
+  }
+}
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/agent/src/share/classes/sun/jvm/hotspot/runtime/Flags.java	Thu Jun 04 15:15:08 2015 -0700
@@ -0,0 +1,48 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+package sun.jvm.hotspot.runtime;
+
+//These definitions should be kept in sync with the definitions in the HotSpot code.
+
+public enum Flags {
+  // value origin
+  DEFAULT ("Default"),
+  COMMAND_LINE ("Command line"),
+  ENVIRON_VAR ("Environment variable"),
+  CONFIG_FILE ("Config file"),
+  MANAGEMENT ("Management"),
+  ERGONOMIC ("Ergonomic"),
+  ATTACH_ON_DEMAND ("Attach on demand"),
+  INTERNAL ("Internal");
+
+  private final String value;
+
+  Flags(String val) {
+    this.value = val;
+  }
+  public String value() {
+    return value;
+  }
+}
--- a/agent/src/share/classes/sun/jvm/hotspot/runtime/Thread.java	Thu May 07 18:13:43 2015 -0700
+++ b/agent/src/share/classes/sun/jvm/hotspot/runtime/Thread.java	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000, 2011, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2000, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -41,6 +41,8 @@
   private static AddressField currentPendingMonitorField;
   private static AddressField currentWaitingMonitorField;
 
+  private static JLongField allocatedBytesField;
+
   static {
     VM.registerVMInitializedObserver(new Observer() {
         public void update(Observable o, Object data) {
@@ -61,6 +63,7 @@
     activeHandlesField = type.getAddressField("_active_handles");
     currentPendingMonitorField = type.getAddressField("_current_pending_monitor");
     currentWaitingMonitorField = type.getAddressField("_current_waiting_monitor");
+    allocatedBytesField = type.getJLongField("_allocated_bytes");
   }
 
   public Thread(Address addr) {
@@ -104,6 +107,10 @@
     return new JNIHandleBlock(a);
   }
 
+  public long allocatedBytes() {
+    return allocatedBytesField.getValue(addr);
+  }
+
   public boolean   isVMThread()                { return false; }
   public boolean   isJavaThread()              { return false; }
   public boolean   isCompilerThread()          { return false; }
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/agent/src/share/classes/sun/jvm/hotspot/runtime/VMOps.java	Thu Jun 04 15:15:08 2015 -0700
@@ -0,0 +1,86 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+package sun.jvm.hotspot.runtime;
+
+//These definitions should be kept in sync with the definitions in the HotSpot code.
+
+public enum VMOps {
+  Dummy,
+  ThreadStop,
+  ThreadDump,
+  PrintThreads,
+  FindDeadlocks,
+  ForceSafepoint,
+  ForceAsyncSafepoint,
+  Deoptimize,
+  DeoptimizeFrame,
+  DeoptimizeAll,
+  ZombieAll,
+  UnlinkSymbols,
+  Verify,
+  PrintJNI,
+  HeapDumper,
+  DeoptimizeTheWorld,
+  CollectForMetadataAllocation,
+  GC_HeapInspection,
+  GenCollectFull,
+  GenCollectFullConcurrent,
+  GenCollectForAllocation,
+  ParallelGCFailedAllocation,
+  ParallelGCSystemGC,
+  CGC_Operation,
+  CMS_Initial_Mark,
+  CMS_Final_Remark,
+  G1CollectFull,
+  G1CollectForAllocation,
+  G1IncCollectionPause,
+  EnableBiasedLocking,
+  RevokeBias,
+  BulkRevokeBias,
+  PopulateDumpSharedSpace,
+  JNIFunctionTableCopier,
+  RedefineClasses,
+  GetOwnedMonitorInfo,
+  GetObjectMonitorUsage,
+  GetCurrentContendedMonitor,
+  GetStackTrace,
+  GetMultipleStackTraces,
+  GetAllStackTraces,
+  GetThreadListStackTraces,
+  GetFrameCount,
+  GetFrameLocation,
+  ChangeBreakpoints,
+  GetOrSetLocal,
+  GetCurrentLocation,
+  EnterInterpOnlyMode,
+  ChangeSingleStep,
+  HeapWalkOperation,
+  HeapIterateOperation,
+  ReportJavaOutOfMemory,
+  JFRCheckpoint,
+  Exit,
+  LinuxDllLoad,
+  Terminating
+}
--- a/agent/src/share/classes/sun/jvm/hotspot/ui/action/HSDBActionManager.java	Thu May 07 18:13:43 2015 -0700
+++ b/agent/src/share/classes/sun/jvm/hotspot/ui/action/HSDBActionManager.java	Thu Jun 04 15:15:08 2015 -0700
@@ -32,10 +32,12 @@
 public class HSDBActionManager extends ActionManager {
 
     public static ActionManager getInstance() {
-        if (manager == null) {
-            manager = new HSDBActionManager();
+        ActionManager m = ActionManager.getInstance();
+        if (m == null) {
+            m = new HSDBActionManager();
+            ActionManager.setInstance(m);
         }
-        return manager;
+        return m;
     }
 
     protected void addActions() {
--- a/agent/src/share/classes/sun/jvm/hotspot/utilities/HeapHprofBinWriter.java	Thu May 07 18:13:43 2015 -0700
+++ b/agent/src/share/classes/sun/jvm/hotspot/utilities/HeapHprofBinWriter.java	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2004, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2004, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -799,6 +799,18 @@
         writeObjectID(klass.getJavaMirror());
 
         ClassData cd = (ClassData) classDataCache.get(klass);
+        if (cd == null) {
+            // The class is not present in the system dictionary, probably Lambda.
+            // Add it to cache here
+            if (klass instanceof InstanceKlass) {
+                InstanceKlass ik = (InstanceKlass) klass;
+                List fields = getInstanceFields(ik);
+                int instSize = getSizeForFields(fields);
+                cd = new ClassData(instSize, fields);
+                classDataCache.put(ik, cd);
+            }
+        }
+
         if (Assert.ASSERTS_ENABLED) {
             Assert.that(cd != null, "can not get class data for " + klass.getName().asString() + klass.getAddress());
         }
--- a/make/aix/makefiles/adlc.make	Thu May 07 18:13:43 2015 -0700
+++ b/make/aix/makefiles/adlc.make	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 #
-# Copyright (c) 1999, 2013, Oracle and/or its affiliates. All rights reserved.
+# Copyright (c) 1999, 2015, Oracle and/or its affiliates. All rights reserved.
 # DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 #
 # This code is free software; you can redistribute it and/or modify it
@@ -140,13 +140,7 @@
 # Note "+="; it is a hook so flags.make can add more flags, like -g or -DFOO.
 ADLCFLAGS += -q -T
 
-# Normally, debugging is done directly on the ad_<arch>*.cpp files.
-# But -g will put #line directives in those files pointing back to <arch>.ad.
-# Some builds of gcc 3.2 have a bug that gets tickled by the extra #line directives
-# so skip it for 3.2 and ealier.
-ifneq "$(shell expr \( $(CC_VER_MAJOR) \> 3 \) \| \( \( $(CC_VER_MAJOR) = 3 \) \& \( $(CC_VER_MINOR) \>= 3 \) \))" "0"
 ADLCFLAGS += -g
-endif
 
 ifdef LP64
 ADLCFLAGS += -D_LP64
--- a/make/aix/makefiles/ppc64.make	Thu May 07 18:13:43 2015 -0700
+++ b/make/aix/makefiles/ppc64.make	Thu Jun 04 15:15:08 2015 -0700
@@ -1,6 +1,6 @@
 #
-# Copyright (c) 2004, 2013, Oracle and/or its affiliates. All rights reserved.
-# Copyright 2012, 2013 SAP AG. All rights reserved.
+# Copyright (c) 2004, 2015, Oracle and/or its affiliates. All rights reserved.
+# Copyright 2012, 2015 SAP AG. 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
@@ -46,7 +46,9 @@
 #  - 1540-1090 (I) The destructor of "..." might not be called.
 #  - 1500-010: (W) WARNING in ...: Infinite loop.  Program may not stop.
 #    There are several infinite loops in the vm, suppress.
-CFLAGS += -qsuppress=1540-1090 -qsuppress=1500-010
+#  - 1540-1639 (I) The behavior of long type bit fields has changed ...
+#                  ... long type bit fields now default to long, not int.
+CFLAGS += -qsuppress=1540-1090 -qsuppress=1500-010 -qsuppress=1540-1639
 
 # Suppress 
 #  - 540-1088 (W) The exception specification is being ignored.
@@ -69,9 +71,6 @@
 OPT_CFLAGS/sharedRuntimeTrig.o = $(OPT_CFLAGS/NOOPT)
 OPT_CFLAGS/sharedRuntimeTrans.o = $(OPT_CFLAGS/NOOPT)
 
-# xlc 10.01 parameters for ipa compile.
-QIPA_COMPILE=$(if $(CXX_IS_V10),-qipa)
-
 # Xlc 10.1 parameters for aggressive optimization:
 # - qhot=level=1: Most aggressive loop optimizations.
 # - qignerrno: Assume errno is not modified by system calls.
@@ -86,7 +85,7 @@
 OPT_CFLAGS/synchronizer.o = $(OPT_CFLAGS) -qnoinline
 
 # Set all the xlC V10.1 options here.
-OPT_CFLAGS += $(QIPA_COMPILE) $(QV10_OPT) $(QV10_OPT_AGGRESSIVE)
+OPT_CFLAGS += $(QV10_OPT) $(QV10_OPT_AGGRESSIVE)
 
 export OBJECT_MODE=64
 
--- a/make/aix/makefiles/xlc.make	Thu May 07 18:13:43 2015 -0700
+++ b/make/aix/makefiles/xlc.make	Thu Jun 04 15:15:08 2015 -0700
@@ -1,6 +1,6 @@
 #
-# Copyright (c) 1999, 2013, Oracle and/or its affiliates. All rights reserved.
-# Copyright (c) 2012, 2013 SAP. All rights reserved.
+# Copyright (c) 1999, 2015, Oracle and/or its affiliates. All rights reserved.
+# Copyright (c) 2012, 2015 SAP. 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
@@ -34,13 +34,17 @@
 
 AS  = $(CC) -c
 
-# get xlc version
-CXX_VERSION   := $(shell $(CXX) -qversion 2>&1 | sed -n 's/.*Version: \([0-9.]*\)/\1/p')
+# get xlc version which comes as VV.RR.MMMM.LLLL where 'VV' is the version,
+# 'RR' is the release, 'MMMM' is the modification and 'LLLL' is the level.
+# We only use 'VV.RR.LLLL' to avoid integer overflows in bash when comparing
+# the version numbers (some shells only support 32-bit integer compares!).
+CXX_VERSION := $(shell $(CXX) -qversion 2>&1 | \
+                   sed -n 's/.*Version: \([0-9]\{2\}\).\([0-9]\{2\}\).[0-9]\{4\}.\([0-9]\{4\}\)/\1\2\3/p')
 
 # xlc 08.00.0000.0023 and higher supports -qtune=balanced
-CXX_SUPPORTS_BALANCED_TUNING=$(shell if [ $(subst .,,$(CXX_VERSION)) -ge 080000000023 ] ; then echo "true" ; fi)
+CXX_SUPPORTS_BALANCED_TUNING := $(shell if [ $(CXX_VERSION) -ge 08000023 ] ; then echo "true" ; fi)
 # xlc 10.01 is used with aggressive optimizations to boost performance
-CXX_IS_V10=$(shell if [ $(subst .,,$(CXX_VERSION)) -ge 100100000000 ] ; then echo "true" ; fi)
+CXX_IS_V10 := $(shell if [ $(CXX_VERSION) -ge 10010000 ] ; then echo "true" ; fi)
 
 # check for precompiled headers support
 
@@ -130,7 +134,7 @@
 # MAPFLAG = -Xlinker --version-script=FILENAME
 
 # Build shared library
-SHARED_FLAG = -q64 -b64 -bexpall -G -bnoentry -qmkshrobj -brtl -bnolibpath
+SHARED_FLAG = -q64 -b64 -bexpall -G -bnoentry -qmkshrobj -brtl -bnolibpath -bernotok
 
 #------------------------------------------------------------------------
 # Debug flags
--- a/make/bsd/makefiles/gcc.make	Thu May 07 18:13:43 2015 -0700
+++ b/make/bsd/makefiles/gcc.make	Thu Jun 04 15:15:08 2015 -0700
@@ -349,7 +349,7 @@
   # The macro takes the version with no dots, ex: 1070
   CFLAGS += -DMAC_OS_X_VERSION_MAX_ALLOWED=$(subst .,,$(MACOSX_VERSION_MIN)) \
             -mmacosx-version-min=$(MACOSX_VERSION_MIN)
-  LDFLAGS += -mmacosx-version-min=$(MACOSX_VERSION_MIN)
+  LFLAGS += -mmacosx-version-min=$(MACOSX_VERSION_MIN)
 endif
 
 
--- a/make/bsd/makefiles/sa.make	Thu May 07 18:13:43 2015 -0700
+++ b/make/bsd/makefiles/sa.make	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 #
-# Copyright (c) 2003, 2012, Oracle and/or its affiliates. All rights reserved.
+# Copyright (c) 2003, 2015, Oracle and/or its affiliates. All rights reserved.
 # DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 #
 # This code is free software; you can redistribute it and/or modify it
@@ -34,6 +34,8 @@
 
 include $(GAMMADIR)/make/sa.files
 
+-include $(HS_ALT_MAKE)/bsd/makefiles/sa.make
+
 TOPDIR    = $(shell echo `pwd`)
 GENERATED = $(TOPDIR)/../generated
 
--- a/make/bsd/makefiles/saproc.make	Thu May 07 18:13:43 2015 -0700
+++ b/make/bsd/makefiles/saproc.make	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 #
-# Copyright (c) 2005, 2013, Oracle and/or its affiliates. All rights reserved.
+# Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved.
 # DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 #
 # This code is free software; you can redistribute it and/or modify it
@@ -64,9 +64,22 @@
 else
   ifeq ($(OS_VENDOR), Darwin)
     SASRCFILES = $(DARWIN_NON_STUB_SASRCFILES)
-    SALIBS = -g -framework Foundation -F/System/Library/Frameworks/JavaVM.framework/Frameworks -framework JavaNativeFoundation -framework Security -framework CoreFoundation
+    SALIBS = -g \
+             -framework Foundation \
+             -framework JavaNativeFoundation \
+             -framework Security \
+             -framework CoreFoundation
     #objc compiler blows up on -march=i586, perhaps it should not be included in the macosx intel 32-bit C++ compiles?
     SAARCH = $(subst -march=i586,,$(ARCHFLAG))
+
+    # This is needed to locate JavaNativeFoundation.framework
+    # JDK 8 doesn't have SYSROOT_CFLAGS, so we'll cobble it together here
+    SA_SYSROOT_FLAGS=
+    ifneq ($(SDKPATH),)
+      SA_SYSROOT_FLAGS += -isysroot "$(SDKPATH)" -iframework"$(SDKPATH)/System/Library/Frameworks"
+    endif
+    # always needed, even if SDKPATH is empty
+    SA_SYSROOT_FLAGS += -F"$(SDKPATH)/System/Library/Frameworks/JavaVM.framework/Frameworks"
   else
     SASRCFILES = $(SASRCDIR)/StubDebuggerLocal.c
     SALIBS = 
@@ -97,17 +110,14 @@
 
 ifneq ($(OS_VENDOR), Darwin)
 SA_LFLAGS = $(MAPFLAG:FILENAME=$(SAMAPFILE))
+else
+# bring in minimum version argument or we'll fail on OSX 10.10
+SA_LFLAGS = $(LFLAGS)
 endif
 SA_LFLAGS += $(LDFLAGS_HASH_STYLE)
 
-ifeq ($(OS_VENDOR), Darwin)
-  BOOT_JAVA_INCLUDES = -I$(BOOT_JAVA_HOME)/include \
-    -I$(BOOT_JAVA_HOME)/include/$(shell uname -s | tr "[:upper:]" "[:lower:]") \
-    -I/System/Library/Frameworks/JavaVM.framework/Headers
-else
-  BOOT_JAVA_INCLUDES = -I$(BOOT_JAVA_HOME)/include \
-    -I$(BOOT_JAVA_HOME)/include/$(shell uname -s | tr "[:upper:]" "[:lower:]")
-endif
+BOOT_JAVA_INCLUDES = -I$(BOOT_JAVA_HOME)/include \
+  -I$(BOOT_JAVA_HOME)/include/$(shell uname -s | tr "[:upper:]" "[:lower:]")
 
 $(LIBSAPROC): $(SASRCFILES) $(SAMAPFILE)
 	$(QUIETLY) if [ "$(BOOT_JAVA_HOME)" = "" ]; then \
@@ -116,6 +126,7 @@
 	fi
 	@echo Making SA debugger back-end...
 	$(QUIETLY) $(CC) -D$(BUILDARCH) -D_GNU_SOURCE                   \
+	           $(SA_SYSROOT_FLAGS)                                  \
 	           $(SYMFLAG) $(SAARCH) $(SHARED_FLAG) $(PICFLAG)       \
 	           -I$(SASRCDIR)                                        \
 	           -I$(GENERATED)                                       \
--- a/make/defs.make	Thu May 07 18:13:43 2015 -0700
+++ b/make/defs.make	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 #
-# Copyright (c) 2006, 2013, Oracle and/or its affiliates. All rights reserved.
+# Copyright (c) 2006, 2015, Oracle and/or its affiliates. All rights reserved.
 # DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 #
 # This code is free software; you can redistribute it and/or modify it
@@ -273,7 +273,7 @@
 
   # Use uname output for SRCARCH, but deal with platform differences. If ARCH
   # is not explicitly listed below, it is treated as x86.
-  SRCARCH     = $(ARCH/$(filter sparc sparc64 ia64 amd64 x86_64 arm ppc ppc64 zero,$(ARCH)))
+  SRCARCH    ?= $(ARCH/$(filter sparc sparc64 ia64 amd64 x86_64 ppc64 zero,$(ARCH)))
   ARCH/       = x86
   ARCH/sparc  = sparc
   ARCH/sparc64= sparc
@@ -281,12 +281,10 @@
   ARCH/amd64  = x86
   ARCH/x86_64 = x86
   ARCH/ppc64  = ppc
-  ARCH/ppc    = ppc
-  ARCH/arm    = arm
   ARCH/zero   = zero
 
   # BUILDARCH is usually the same as SRCARCH, except for sparcv9
-  BUILDARCH = $(SRCARCH)
+  BUILDARCH ?= $(SRCARCH)
   ifeq ($(BUILDARCH), x86)
     ifdef LP64
       BUILDARCH = amd64
@@ -306,18 +304,16 @@
   endif
 
   # LIBARCH is 1:1 mapping from BUILDARCH
-  LIBARCH         = $(LIBARCH/$(BUILDARCH))
+  LIBARCH        ?= $(LIBARCH/$(BUILDARCH))
   LIBARCH/i486    = i386
   LIBARCH/amd64   = amd64
   LIBARCH/sparc   = sparc
   LIBARCH/sparcv9 = sparcv9
   LIBARCH/ia64    = ia64
   LIBARCH/ppc64   = ppc64
-  LIBARCH/ppc     = ppc
-  LIBARCH/arm     = arm
   LIBARCH/zero    = $(ZERO_LIBARCH)
 
-  LP64_ARCH = sparcv9 amd64 ia64 ppc64 zero
+  LP64_ARCH += sparcv9 amd64 ia64 ppc64 zero
 endif
 
 # Required make macro settings for all platforms
--- a/make/linux/makefiles/arm.make	Thu May 07 18:13:43 2015 -0700
+++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
@@ -1,31 +0,0 @@
-#
-# Copyright (c) 2008, 2013, Oracle and/or its affiliates. All rights reserved.
-# DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
-#
-# This code is free software; you can redistribute it and/or modify it
-# under the terms of the GNU General Public License version 2 only, as
-# published by the Free Software Foundation.
-#
-# This code is distributed in the hope that it will be useful, but WITHOUT
-# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
-# FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
-# version 2 for more details (a copy is included in the LICENSE file that
-# accompanied this code).
-#
-# You should have received a copy of the GNU General Public License version
-# 2 along with this work; if not, write to the Free Software Foundation,
-# Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
-#
-# Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
-# or visit www.oracle.com if you need additional information or have any
-# questions.
-#  
-#
-
-Obj_Files += linux_arm.o
-
-ifneq ($(EXT_LIBS_PATH),)
-  LIBS += $(EXT_LIBS_PATH)/sflt_glibc.a 
-endif
-
-CFLAGS += -DVM_LITTLE_ENDIAN
--- a/make/linux/makefiles/build_vm_def.sh	Thu May 07 18:13:43 2015 -0700
+++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
@@ -1,16 +0,0 @@
-#!/bin/sh
-
-# If we're cross compiling use that path for nm
-if [ "$CROSS_COMPILE_ARCH" != "" ]; then 
-NM=$ALT_COMPILER_PATH/nm
-else
-NM=nm
-fi
-
-$NM --defined-only $* \
-    | awk '{
-              if ($3 ~ /^_ZTV/ || $3 ~ /^gHotSpotVM/) print "\t" $3 ";"
-              if ($3 ~ /^UseSharedSpaces$/) print "\t" $3 ";"
-              if ($3 ~ /^_ZN9Arguments17SharedArchivePathE$/) print "\t" $3 ";"
-          }' \
-    | sort -u
--- a/make/linux/makefiles/buildtree.make	Thu May 07 18:13:43 2015 -0700
+++ b/make/linux/makefiles/buildtree.make	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 #
-# Copyright (c) 2005, 2013, Oracle and/or its affiliates. All rights reserved.
+# Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved.
 # DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 #
 # This code is free software; you can redistribute it and/or modify it
@@ -70,6 +70,8 @@
     PLATFORM_FILE = $(GAMMADIR)/make/$(OS_FAMILY)/platform_$(BUILDARCH).suncc
   else
     PLATFORM_FILE = $(GAMMADIR)/make/$(OS_FAMILY)/platform_$(BUILDARCH)
+    ALT_PLATFORM_FILE = $(HS_ALT_MAKE)/$(OS_FAMILY)/platform_$(BUILDARCH)
+    PLATFORM_FILE := $(if $(wildcard $(ALT_PLATFORM_FILE)),$(ALT_PLATFORM_FILE),$(PLATFORM_FILE))
   endif
 endif
 
@@ -202,7 +204,7 @@
 	$(QUIETLY) ( \
 	$(BUILDTREE_COMMENT); \
 	echo; \
-	echo "Platform_file = $(PLATFORM_FILE)" | sed 's|$(GAMMADIR)|$$(GAMMADIR)|'; \
+	echo "Platform_file = $(PLATFORM_FILE)" | sed -e 's|$(HS_ALT_MAKE)|$$(HS_ALT_MAKE)|' -e 's|$(GAMMADIR)|$$(GAMMADIR)|'; \
 	sed -n '/=/s/^ */Platform_/p' < $(PLATFORM_FILE); \
 	echo; \
 	echo "GAMMADIR = $(GAMMADIR)"; \
--- a/make/linux/makefiles/defs.make	Thu May 07 18:13:43 2015 -0700
+++ b/make/linux/makefiles/defs.make	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 #
-# Copyright (c) 2006, 2013, Oracle and/or its affiliates. All rights reserved.
+# Copyright (c) 2006, 2015, Oracle and/or its affiliates. All rights reserved.
 # DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 #
 # This code is free software; you can redistribute it and/or modify it
@@ -109,22 +109,6 @@
   HS_ARCH          = x86
 endif
 
-# ARM
-ifeq ($(ARCH), arm)
-  ARCH_DATA_MODEL  = 32
-  PLATFORM         = linux-arm
-  VM_PLATFORM      = linux_arm
-  HS_ARCH          = arm
-endif
-
-# PPC
-ifeq ($(ARCH), ppc)
-  ARCH_DATA_MODEL  = 32
-  PLATFORM         = linux-ppc
-  VM_PLATFORM      = linux_ppc
-  HS_ARCH          = ppc
-endif
-
 # PPC64
 ifeq ($(ARCH), ppc64)
   ARCH_DATA_MODEL  = 64
--- a/make/linux/makefiles/gcc.make	Thu May 07 18:13:43 2015 -0700
+++ b/make/linux/makefiles/gcc.make	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 #
-# Copyright (c) 1999, 2013, Oracle and/or its affiliates. All rights reserved.
+# Copyright (c) 1999, 2015, Oracle and/or its affiliates. All rights reserved.
 # DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 #
 # This code is free software; you can redistribute it and/or modify it
@@ -176,11 +176,7 @@
 ARCHFLAG/ia64    =
 ARCHFLAG/sparc   = -m32 -mcpu=v9
 ARCHFLAG/sparcv9 = -m64 -mcpu=v9
-ARCHFLAG/arm     =  -fsigned-char
 ARCHFLAG/zero    = $(ZERO_ARCHFLAG)
-ifndef E500V2
-ARCHFLAG/ppc     =  -mcpu=powerpc
-endif
 ARCHFLAG/ppc64   =  -m64
 
 CFLAGS     += $(ARCHFLAG)
@@ -188,10 +184,6 @@
 LFLAGS     += $(ARCHFLAG)
 ASFLAGS    += $(ARCHFLAG)
 
-ifdef E500V2
-CFLAGS += -DE500V2
-endif
-
 # Use C++ Interpreter
 ifdef CC_INTERP
   CFLAGS += -DCC_INTERP
@@ -345,47 +337,41 @@
   # Note: The Itanium gcc compiler crashes when using -gstabs.
   DEBUG_CFLAGS/ia64  = -g
   DEBUG_CFLAGS/amd64 = -g
-  DEBUG_CFLAGS/arm   = -g
-  DEBUG_CFLAGS/ppc   = -g
   DEBUG_CFLAGS/ppc64 = -g
   DEBUG_CFLAGS += $(DEBUG_CFLAGS/$(BUILDARCH))
   ifeq ($(DEBUG_CFLAGS/$(BUILDARCH)),)
       ifeq ($(USE_CLANG), true)
         # Clang doesn't understand -gstabs
-        DEBUG_CFLAGS += -g
+        DEBUG_CFLAGS/$(BUILDARCH) = -g
       else
-        DEBUG_CFLAGS += -gstabs
+        DEBUG_CFLAGS/$(BUILDARCH) = -gstabs
       endif
   endif
   
   ifeq ($(ENABLE_FULL_DEBUG_SYMBOLS),1)
     FASTDEBUG_CFLAGS/ia64  = -g
     FASTDEBUG_CFLAGS/amd64 = -g
-    FASTDEBUG_CFLAGS/arm   = -g
-    FASTDEBUG_CFLAGS/ppc   = -g
     FASTDEBUG_CFLAGS/ppc64 = -g
-    FASTDEBUG_CFLAGS += $(DEBUG_CFLAGS/$(BUILDARCH))
+    FASTDEBUG_CFLAGS += $(FASTDEBUG_CFLAGS/$(BUILDARCH))
     ifeq ($(FASTDEBUG_CFLAGS/$(BUILDARCH)),)
       ifeq ($(USE_CLANG), true)
         # Clang doesn't understand -gstabs
-        FASTDEBUG_CFLAGS += -g
+        FASTDEBUG_CFLAGS/$(BUILDARCH) = -g
       else
-        FASTDEBUG_CFLAGS += -gstabs
+        FASTDEBUG_CFLAGS/$(BUILDARCH) = -gstabs
       endif
     endif
   
     OPT_CFLAGS/ia64  = -g
     OPT_CFLAGS/amd64 = -g
-    OPT_CFLAGS/arm   = -g
-    OPT_CFLAGS/ppc   = -g
     OPT_CFLAGS/ppc64 = -g
     OPT_CFLAGS += $(OPT_CFLAGS/$(BUILDARCH))
     ifeq ($(OPT_CFLAGS/$(BUILDARCH)),)
       ifeq ($(USE_CLANG), true)
         # Clang doesn't understand -gstabs
-        OPT_CFLAGS += -g
+        OPT_CFLAGS/$(BUILDARCH) = -g
       else
-        OPT_CFLAGS += -gstabs
+        OPT_CFLAGS/$(BUILDARCH) = -gstabs
       endif
     endif
   endif
@@ -409,3 +395,5 @@
 ifndef USE_SUNCC
   CFLAGS += -fno-omit-frame-pointer
 endif
+
+-include $(HS_ALT_MAKE)/linux/makefiles/gcc.make
--- a/make/linux/makefiles/ppc.make	Thu May 07 18:13:43 2015 -0700
+++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
@@ -1,33 +0,0 @@
-#
-# Copyright (c) 2004, 2012, Oracle and/or its affiliates. All rights reserved.
-# DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
-#
-# This code is free software; you can redistribute it and/or modify it
-# under the terms of the GNU General Public License version 2 only, as
-# published by the Free Software Foundation.
-#
-# This code is distributed in the hope that it will be useful, but WITHOUT
-# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
-# FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
-# version 2 for more details (a copy is included in the LICENSE file that
-# accompanied this code).
-#
-# You should have received a copy of the GNU General Public License version
-# 2 along with this work; if not, write to the Free Software Foundation,
-# Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
-#
-# Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
-# or visit www.oracle.com if you need additional information or have any
-# questions.
-#  
-#
-
-# The copied fdlibm routines in sharedRuntimeTrig.o must not be optimized
-OPT_CFLAGS/sharedRuntimeTrig.o = $(OPT_CFLAGS/NOOPT)
-
-# Must also specify if CPU is big endian
-CFLAGS += -DVM_BIG_ENDIAN
-
-ifdef E500V2
-ASFLAGS += -Wa,-mspe -Wa,--defsym -Wa,E500V2=1 
-endif
--- a/make/linux/makefiles/saproc.make	Thu May 07 18:13:43 2015 -0700
+++ b/make/linux/makefiles/saproc.make	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 #
-# Copyright (c) 2005, 2013, Oracle and/or its affiliates. All rights reserved.
+# Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved.
 # DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 #
 # This code is free software; you can redistribute it and/or modify it
@@ -69,19 +69,21 @@
 endif
 
 ifneq ($(ALT_SASRCDIR),)
-ALT_SAINCDIR=-I$(ALT_SASRCDIR)
+ALT_SAINCDIR=-I$(ALT_SASRCDIR) -DALT_SASRCDIR
 else
 ALT_SAINCDIR=
 endif
 SA_LFLAGS = $(MAPFLAG:FILENAME=$(SAMAPFILE)) $(LDFLAGS_HASH_STYLE)
 
+SAARCH ?= $(BUILDARCH)
+
 $(LIBSAPROC): $(SASRCFILES) $(SAMAPFILE)
 	$(QUIETLY) if [ "$(BOOT_JAVA_HOME)" = "" ]; then \
 	  echo "ALT_BOOTDIR, BOOTDIR or JAVA_HOME needs to be defined to build SA"; \
 	  exit 1; \
 	fi
 	@echo Making SA debugger back-end...
-	$(QUIETLY) $(CC) -D$(BUILDARCH) -D_GNU_SOURCE                   \
+	$(QUIETLY) $(CC) -D$(SAARCH) -D_GNU_SOURCE                      \
 		   -D_FILE_OFFSET_BITS=64                               \
                    $(SYMFLAG) $(ARCHFLAG) $(SHARED_FLAG) $(PICFLAG)     \
 	           -I$(SASRCDIR)                                        \
--- a/make/linux/makefiles/vm.make	Thu May 07 18:13:43 2015 -0700
+++ b/make/linux/makefiles/vm.make	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 #
-# Copyright (c) 1999, 2013, Oracle and/or its affiliates. All rights reserved.
+# Copyright (c) 1999, 2015, Oracle and/or its affiliates. All rights reserved.
 # DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 #
 # This code is free software; you can redistribute it and/or modify it
@@ -45,8 +45,9 @@
 ifeq ($(findstring true, $(JVM_VARIANT_ZERO) $(JVM_VARIANT_ZEROSHARK)), true)
   include $(MAKEFILES_DIR)/zeroshark.make
 else
-  include $(MAKEFILES_DIR)/$(BUILDARCH).make
-  -include $(HS_ALT_MAKE)/$(Platform_os_family)/makefiles/$(BUILDARCH).make
+  BUILDARCH_MAKE = $(MAKEFILES_DIR)/$(BUILDARCH).make
+  ALT_BUILDARCH_MAKE = $(HS_ALT_MAKE)/$(Platform_os_family)/makefiles/$(BUILDARCH).make
+  include $(if $(wildcard $(ALT_BUILDARCH_MAKE)),$(ALT_BUILDARCH_MAKE),$(BUILDARCH_MAKE))
 endif
 
 # set VPATH so make knows where to look for source files
@@ -245,8 +246,14 @@
 	rm -f $@
 	cat $^ > $@
 
+VMDEF_PAT  = ^_ZTV
+VMDEF_PAT := ^gHotSpotVM|$(VMDEF_PAT)
+VMDEF_PAT := ^UseSharedSpaces$$|$(VMDEF_PAT)
+VMDEF_PAT := ^_ZN9Arguments17SharedArchivePathE$$|$(VMDEF_PAT)
+
 vm.def: $(Res_Files) $(Obj_Files)
-	sh $(GAMMADIR)/make/linux/makefiles/build_vm_def.sh *.o > $@
+	$(QUIETLY) $(NM) --defined-only $(Obj_Files) | sort -k3 -u | \
+	awk '$$3 ~ /$(VMDEF_PAT)/ { print "\t" $$3 ";" }' > $@
 
 mapfile_ext:
 	rm -f $@
--- a/make/linux/platform_arm	Thu May 07 18:13:43 2015 -0700
+++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
@@ -1,17 +0,0 @@
-os_family = linux
-
-arch = arm
-
-arch_model = arm
-
-os_arch = linux_arm
-
-os_arch_model = linux_arm
-
-lib_arch = arm
-
-compiler = gcc
-
-gnu_dis_arch = arm
-
-sysdefs = -DLINUX -D_GNU_SOURCE -DARM
--- a/make/linux/platform_ppc	Thu May 07 18:13:43 2015 -0700
+++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
@@ -1,17 +0,0 @@
-os_family = linux
-
-arch = ppc
-
-arch_model = ppc_32
-
-os_arch = linux_ppc
-
-os_arch_model = linux_ppc_32
-
-lib_arch = ppc
-
-compiler = gcc
-
-gnu_dis_arch = ppc
-
-sysdefs = -DLINUX -D_GNU_SOURCE -DPPC32
--- a/make/solaris/makefiles/sa.make	Thu May 07 18:13:43 2015 -0700
+++ b/make/solaris/makefiles/sa.make	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 #
-# Copyright (c) 2003, 2012, Oracle and/or its affiliates. All rights reserved.
+# Copyright (c) 2003, 2015, Oracle and/or its affiliates. All rights reserved.
 # DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 #
 # This code is free software; you can redistribute it and/or modify it
@@ -29,8 +29,12 @@
 # and generate JNI header file for native methods.
 
 include $(GAMMADIR)/make/solaris/makefiles/rules.make
+include $(GAMMADIR)/make/defs.make
 AGENT_DIR = $(GAMMADIR)/agent
 include $(GAMMADIR)/make/sa.files
+
+-include $(HS_ALT_MAKE)/solaris/makefiles/sa.make
+
 GENERATED = ../generated
 
 # tools.jar is needed by the JDI - SA binding
--- a/make/windows/makefiles/sa.make	Thu May 07 18:13:43 2015 -0700
+++ b/make/windows/makefiles/sa.make	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 #
-# Copyright (c) 2003, 2013, Oracle and/or its affiliates. All rights reserved.
+# Copyright (c) 2003, 2015, Oracle and/or its affiliates. All rights reserved.
 # DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 #
 # This code is free software; you can redistribute it and/or modify it
@@ -38,6 +38,22 @@
 
 GENERATED = ../generated
 
+HS_COMMON_SRC_REL = src
+
+!if "$(OPENJDK)" != "true"
+HS_ALT_SRC_REL=src/closed
+HS_ALT_SRC = $(WorkSpace)/$(HS_ALT_SRC_REL)
+!ifndef HS_ALT_MAKE
+HS_ALT_MAKE=$(WorkSpace)/make/closed
+!endif
+!endif
+
+HS_COMMON_SRC = $(WorkSpace)/$(HS_COMMON_SRC_REL)
+
+!ifdef HS_ALT_MAKE
+!include $(HS_ALT_MAKE)/windows/makefiles/sa.make
+!endif
+
 # tools.jar is needed by the JDI - SA binding
 SA_CLASSPATH = $(BOOT_JAVA_HOME)/lib/tools.jar
 
--- a/src/cpu/ppc/vm/frame_ppc.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/cpu/ppc/vm/frame_ppc.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -318,3 +318,10 @@
   // unused... but returns fp() to minimize changes introduced by 7087445
   return fp();
 }
+
+#ifndef PRODUCT
+// This is a generic constructor which is only used by pns() in debug.cpp.
+frame::frame(void* sp, void* fp, void* pc) : _sp((intptr_t*)sp), _unextended_sp((intptr_t*)sp) {
+  find_codeblob_and_set_pc_and_deopt_state((address)pc); // also sets _fp and adjusts _unextended_sp
+}
+#endif
--- a/src/cpu/ppc/vm/interpreter_ppc.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/cpu/ppc/vm/interpreter_ppc.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1,6 +1,6 @@
 /*
- * Copyright (c) 1997, 2014, Oracle and/or its affiliates. All rights reserved.
- * Copyright 2012, 2014 SAP AG. All rights reserved.
+ * Copyright (c) 1997, 2015, Oracle and/or its affiliates. All rights reserved.
+ * Copyright 2012, 2015 SAP AG. 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
@@ -459,7 +459,8 @@
 
   // This is not a leaf but we have a JavaFrameAnchor now and we will
   // check (create) exceptions afterward so this is ok.
-  __ call_VM_leaf(CAST_FROM_FN_PTR(address, InterpreterRuntime::throw_AbstractMethodError));
+  __ call_VM_leaf(CAST_FROM_FN_PTR(address, InterpreterRuntime::throw_AbstractMethodError),
+                  R16_thread);
 
   // Pop the C frame and restore LR.
   __ pop_frame();
--- a/src/cpu/ppc/vm/ppc.ad	Thu May 07 18:13:43 2015 -0700
+++ b/src/cpu/ppc/vm/ppc.ad	Thu Jun 04 15:15:08 2015 -0700
@@ -2264,9 +2264,8 @@
 
 // Do we need to mask the count passed to shift instructions or does
 // the cpu only look at the lower 5/6 bits anyway?
-// Off, as masks are generated in expand rules where required.
-// Constant shift counts are handled in Ideal phase.
-const bool Matcher::need_masked_shift_count = false;
+// PowerPC requires masked shift counts.
+const bool Matcher::need_masked_shift_count = true;
 
 // This affects two different things:
 //  - how Decode nodes are matched
@@ -4418,11 +4417,11 @@
   interface(CONST_INTER);
 %}
 
-// constant 'float +0.0'.
+// Float Immediate: +0.0f.
 operand immF_0() %{
-  predicate((n->getf() == 0) &&
-            (fpclassify(n->getf()) == FP_ZERO) && (signbit(n->getf()) == 0));
+  predicate(jint_cast(n->getf()) == 0);
   match(ConF);
+  
   op_cost(0);
   format %{ %}
   interface(CONST_INTER);
--- a/src/cpu/sparc/vm/frame_sparc.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/cpu/sparc/vm/frame_sparc.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -342,7 +342,7 @@
 // constructors
 
 // Construct an unpatchable, deficient frame
-frame::frame(intptr_t* sp, unpatchable_t, address pc, CodeBlob* cb) {
+void frame::init(intptr_t* sp, address pc, CodeBlob* cb) {
 #ifdef _LP64
   assert( (((intptr_t)sp & (wordSize-1)) == 0), "frame constructor passed an invalid sp");
 #endif
@@ -364,6 +364,10 @@
 #endif // ASSERT
 }
 
+frame::frame(intptr_t* sp, unpatchable_t, address pc, CodeBlob* cb) {
+  init(sp, pc, cb);
+}
+
 frame::frame(intptr_t* sp, intptr_t* younger_sp, bool younger_frame_is_interpreted) :
   _sp(sp),
   _younger_sp(younger_sp),
@@ -418,6 +422,13 @@
   }
 }
 
+#ifndef PRODUCT
+// This is a generic constructor which is only used by pns() in debug.cpp.
+frame::frame(void* sp, void* fp, void* pc) {
+  init((intptr_t*)sp, (address)pc, NULL);
+}
+#endif
+
 bool frame::is_interpreted_frame() const  {
   return Interpreter::contains(pc());
 }
--- a/src/cpu/sparc/vm/frame_sparc.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/cpu/sparc/vm/frame_sparc.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -164,6 +164,8 @@
   enum unpatchable_t { unpatchable };
   frame(intptr_t* sp, unpatchable_t, address pc = NULL, CodeBlob* cb = NULL);
 
+  void init(intptr_t* sp, address pc, CodeBlob* cb);
+
   // Walk from sp outward looking for old_sp, and return old_sp's predecessor
   // (i.e. return the sp from the frame where old_sp is the fp).
   // Register windows are assumed to be flushed for the stack in question.
--- a/src/cpu/sparc/vm/sparc.ad	Thu May 07 18:13:43 2015 -0700
+++ b/src/cpu/sparc/vm/sparc.ad	Thu Jun 04 15:15:08 2015 -0700
@@ -3760,13 +3760,9 @@
   interface(CONST_INTER);
 %}
 
+// Double Immediate: +0.0d
 operand immD0() %{
-#ifdef _LP64
-  // on 64-bit architectures this comparision is faster
   predicate(jlong_cast(n->getd()) == 0);
-#else
-  predicate((n->getd() == 0) && (fpclass(n->getd()) == FP_PZERO));
-#endif
   match(ConD);
 
   op_cost(0);
@@ -3783,9 +3779,9 @@
   interface(CONST_INTER);
 %}
 
-// Float Immediate: 0
-operand immF0() %{
-  predicate((n->getf() == 0) && (fpclass(n->getf()) == FP_PZERO));
+// Float Immediate: +0.0f
+operand immF0() %{ 
+  predicate(jint_cast(n->getf()) == 0);  
   match(ConF);
 
   op_cost(0);
--- a/src/cpu/sparc/vm/vm_version_sparc.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/cpu/sparc/vm/vm_version_sparc.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -37,7 +37,7 @@
 
 int VM_Version::_features = VM_Version::unknown_m;
 const char* VM_Version::_features_str = "";
-unsigned int VM_Version::_L2_cache_line_size = 0;
+unsigned int VM_Version::_L2_data_cache_line_size = 0;
 
 void VM_Version::initialize() {
   _features = determine_features();
@@ -363,7 +363,7 @@
 
 #ifndef PRODUCT
   if (PrintMiscellaneous && Verbose) {
-    tty->print_cr("L2 cache line size: %u", L2_cache_line_size());
+    tty->print_cr("L2 data cache line size: %u", L2_data_cache_line_size());
     tty->print("Allocation");
     if (AllocatePrefetchStyle <= 0) {
       tty->print_cr(": no prefetching");
--- a/src/cpu/sparc/vm/vm_version_sparc.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/cpu/sparc/vm/vm_version_sparc.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -96,8 +96,8 @@
   static int  _features;
   static const char* _features_str;
 
-  static unsigned int _L2_cache_line_size;
-  static unsigned int L2_cache_line_size() { return _L2_cache_line_size; }
+  static unsigned int _L2_data_cache_line_size;
+  static unsigned int L2_data_cache_line_size() { return _L2_data_cache_line_size; }
 
   static void print_features();
   static int  determine_features();
@@ -171,7 +171,7 @@
   static const char* cpu_features()     { return _features_str; }
 
   // default prefetch block size on sparc
-  static intx prefetch_data_size()      { return L2_cache_line_size();  }
+  static intx prefetch_data_size()      { return L2_data_cache_line_size();  }
 
   // Prefetch
   static intx prefetch_copy_interval_in_bytes() {
--- a/src/cpu/x86/vm/frame_x86.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/cpu/x86/vm/frame_x86.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -717,3 +717,10 @@
   assert(! is_compiled_frame(), "unknown compiled frame size");
   return fp();
 }
+
+#ifndef PRODUCT
+// This is a generic constructor which is only used by pns() in debug.cpp.
+frame::frame(void* sp, void* fp, void* pc) {
+  init((intptr_t*)sp, (intptr_t*)fp, (address)pc);
+}
+#endif
--- a/src/cpu/x86/vm/frame_x86.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/cpu/x86/vm/frame_x86.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -187,6 +187,8 @@
 
   frame(intptr_t* sp, intptr_t* fp);
 
+  void init(intptr_t* sp, intptr_t* fp, address pc);
+
   // accessors for the instance variables
   // Note: not necessarily the real 'frame pointer' (see real_fp)
   intptr_t*   fp() const { return _fp; }
--- a/src/cpu/x86/vm/frame_x86.inline.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/cpu/x86/vm/frame_x86.inline.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -40,7 +40,7 @@
   _deopt_state = unknown;
 }
 
-inline frame::frame(intptr_t* sp, intptr_t* fp, address pc) {
+inline void frame::init(intptr_t* sp, intptr_t* fp, address pc) {
   _sp = sp;
   _unextended_sp = sp;
   _fp = fp;
@@ -58,6 +58,10 @@
   }
 }
 
+inline frame::frame(intptr_t* sp, intptr_t* fp, address pc) {
+  init(sp, fp, pc);
+}
+
 inline frame::frame(intptr_t* sp, intptr_t* unextended_sp, intptr_t* fp, address pc) {
   _sp = sp;
   _unextended_sp = unextended_sp;
--- a/src/cpu/x86/vm/macroAssembler_x86.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/cpu/x86/vm/macroAssembler_x86.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -6690,7 +6690,7 @@
     subl(cnt2, stride2);
     jccb(Assembler::notZero, COMPARE_WIDE_VECTORS_LOOP);
     // clean upper bits of YMM registers
-    vzeroupper();
+    vpxor(vec1, vec1);
 
     // compare wide vectors tail
     bind(COMPARE_WIDE_TAIL);
@@ -6705,7 +6705,7 @@
     // Identifies the mismatching (higher or lower)16-bytes in the 32-byte vectors.
     bind(VECTOR_NOT_EQUAL);
     // clean upper bits of YMM registers
-    vzeroupper();
+    vpxor(vec1, vec1);
     lea(str1, Address(str1, result, scale));
     lea(str2, Address(str2, result, scale));
     jmp(COMPARE_16_CHARS);
@@ -6964,7 +6964,8 @@
   bind(DONE);
   if (UseAVX >= 2) {
     // clean upper bits of YMM registers
-    vzeroupper();
+    vpxor(vec1, vec1);
+    vpxor(vec2, vec2);
   }
 }
 
@@ -7098,7 +7099,8 @@
 
         BIND(L_check_fill_8_bytes);
         // clean upper bits of YMM registers
-        vzeroupper();
+        movdl(xtmp, value);
+        pshufd(xtmp, xtmp, 0);
       } else {
         // Fill 32-byte chunks
         pshufd(xtmp, xtmp, 0);
@@ -7261,7 +7263,11 @@
     bind(L_copy_16_chars_exit);
     if (UseAVX >= 2) {
       // clean upper bits of YMM registers
-      vzeroupper();
+      vpxor(tmp2Reg, tmp2Reg);
+      vpxor(tmp3Reg, tmp3Reg);
+      vpxor(tmp4Reg, tmp4Reg);
+      movdl(tmp1Reg, tmp5);
+      pshufd(tmp1Reg, tmp1Reg, 0);
     }
     subptr(len, 8);
     jccb(Assembler::greater, L_copy_8_chars_exit);
--- a/src/cpu/x86/vm/stubGenerator_x86_32.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/cpu/x86/vm/stubGenerator_x86_32.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -837,7 +837,8 @@
 
     if (UseUnalignedLoadStores && (UseAVX >= 2)) {
       // clean upper bits of YMM registers
-      __ vzeroupper();
+      __ vpxor(xmm0, xmm0);
+      __ vpxor(xmm1, xmm1);
     }
     __ addl(qword_count, 8);
     __ jccb(Assembler::zero, L_exit);
--- a/src/cpu/x86/vm/stubGenerator_x86_64.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/cpu/x86/vm/stubGenerator_x86_64.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1328,7 +1328,8 @@
       __ BIND(L_end);
       if (UseAVX >= 2) {
         // clean upper bits of YMM registers
-        __ vzeroupper();
+        __ vpxor(xmm0, xmm0);
+        __ vpxor(xmm1, xmm1);
       }
     } else {
       // Copy 32-bytes per iteration
@@ -1405,7 +1406,8 @@
       __ BIND(L_end);
       if (UseAVX >= 2) {
         // clean upper bits of YMM registers
-        __ vzeroupper();
+        __ vpxor(xmm0, xmm0);
+        __ vpxor(xmm1, xmm1);
       }
     } else {
       // Copy 32-bytes per iteration
--- a/src/cpu/x86/vm/vm_version_x86.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/cpu/x86/vm/vm_version_x86.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -570,10 +570,12 @@
   static uint cores_per_cpu()  {
     uint result = 1;
     if (is_intel()) {
-      if (supports_processor_topology()) {
+      bool supports_topology = supports_processor_topology();
+      if (supports_topology) {
         result = _cpuid_info.tpl_cpuidB1_ebx.bits.logical_cpus /
                  _cpuid_info.tpl_cpuidB0_ebx.bits.logical_cpus;
-      } else {
+      }
+      if (!supports_topology || result == 0) {
         result = (_cpuid_info.dcp_cpuid4_eax.bits.cores_per_cpu + 1);
       }
     } else if (is_amd()) {
--- a/src/cpu/zero/vm/cppInterpreter_zero.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/cpu/zero/vm/cppInterpreter_zero.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -730,7 +730,7 @@
     if (method->is_static())
       object = method->constants()->pool_holder()->java_mirror();
     else
-      object = (oop) locals[0];
+      object = (oop) (void*)locals[0];
     monitor->set_obj(object);
   }
 
--- a/src/cpu/zero/vm/frame_zero.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/cpu/zero/vm/frame_zero.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -441,3 +441,10 @@
   // unused... but returns fp() to minimize changes introduced by 7087445
   return fp();
 }
+
+#ifndef PRODUCT
+// This is a generic constructor which is only used by pns() in debug.cpp.
+frame::frame(void* sp, void* fp, void* pc) {
+  Unimplemented();
+}
+#endif
--- a/src/cpu/zero/vm/frame_zero.inline.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/cpu/zero/vm/frame_zero.inline.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -26,6 +26,8 @@
 #ifndef CPU_ZERO_VM_FRAME_ZERO_INLINE_HPP
 #define CPU_ZERO_VM_FRAME_ZERO_INLINE_HPP
 
+#include "code/codeCache.hpp"
+
 // Constructors
 
 inline frame::frame() {
--- a/src/cpu/zero/vm/stack_zero.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/cpu/zero/vm/stack_zero.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -30,7 +30,9 @@
 
 int ZeroStack::suggest_size(Thread *thread) const {
   assert(needs_setup(), "already set up");
-  return align_size_down(abi_stack_available(thread) / 2, wordSize);
+  int abi_available = abi_stack_available(thread);
+  assert(abi_available >= 0, "available abi stack must be >= 0");
+  return align_size_down(abi_available / 2, wordSize);
 }
 
 void ZeroStack::handle_overflow(TRAPS) {
--- a/src/cpu/zero/vm/stack_zero.inline.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/cpu/zero/vm/stack_zero.inline.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -48,9 +48,11 @@
 // to use under normal circumstances.  Note that the returned
 // value can be negative.
 inline int ZeroStack::abi_stack_available(Thread *thread) const {
-  int stack_used = thread->stack_base() - (address) &stack_used;
+  guarantee(Thread::current() == thread, "should run in the same thread");
+  int stack_used = thread->stack_base() - (address) &stack_used
+    + (StackYellowPages+StackRedPages+StackShadowPages) * os::vm_page_size();
   int stack_free = thread->stack_size() - stack_used;
-  return stack_free - shadow_pages_size();
+  return stack_free;
 }
 
 #endif // CPU_ZERO_VM_STACK_ZERO_INLINE_HPP
--- a/src/os/aix/vm/os_aix.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/os/aix/vm/os_aix.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1999, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1999, 2015, Oracle and/or its affiliates. All rights reserved.
  * Copyright 2012, 2014 SAP AG. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
@@ -3727,6 +3727,11 @@
     tty->print_cr("  found:%s", get_signal_handler_name(thisHandler, buf, O_BUFLEN));
     // No need to check this sig any longer
     sigaddset(&check_signal_done, sig);
+    // Running under non-interactive shell, SHUTDOWN2_SIGNAL will be reassigned SIG_IGN
+    if (sig == SHUTDOWN2_SIGNAL && !isatty(fileno(stdin))) {
+      tty->print_cr("Running in non-interactive shell, %s handler is replaced by shell",
+                    exception_name(sig, buf, O_BUFLEN));
+    }
   } else if (os::Aix::get_our_sigflags(sig) != 0 && (int)act.sa_flags != os::Aix::get_our_sigflags(sig)) {
     tty->print("Warning: %s handler flags ", exception_name(sig, buf, O_BUFLEN));
     tty->print("expected:" PTR32_FORMAT, os::Aix::get_our_sigflags(sig));
@@ -3987,11 +3992,6 @@
   return JNI_OK;
 }
 
-// this is called at the end of vm_initialization
-void os::init_3(void) {
-  return;
-}
-
 // Mark the polling page as unreadable
 void os::make_polling_page_unreadable(void) {
   if (!guard_memory((char*)_polling_page, Aix::page_size())) {
--- a/src/os/bsd/vm/os_bsd.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/os/bsd/vm/os_bsd.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1999, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1999, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -1197,12 +1197,18 @@
   guarantee(retval != 0, "just checking");
   return retval;
 
-#elif __FreeBSD__
+#else
+  #ifdef __FreeBSD__
   retval = syscall(SYS_thr_self);
-#elif __OpenBSD__
+  #else
+    #ifdef __OpenBSD__
   retval = syscall(SYS_getthrid);
-#elif __NetBSD__
+    #else
+      #ifdef __NetBSD__
   retval = (pid_t) syscall(SYS__lwp_self);
+      #endif
+    #endif
+  #endif
 #endif
 
   if (retval == -1) {
@@ -3545,6 +3551,11 @@
     tty->print_cr("  found:%s", get_signal_handler_name(thisHandler, buf, O_BUFLEN));
     // No need to check this sig any longer
     sigaddset(&check_signal_done, sig);
+    // Running under non-interactive shell, SHUTDOWN2_SIGNAL will be reassigned SIG_IGN
+    if (sig == SHUTDOWN2_SIGNAL && !isatty(fileno(stdin))) {
+      tty->print_cr("Running in non-interactive shell, %s handler is replaced by shell",
+                    exception_name(sig, buf, O_BUFLEN));
+    }
   } else if(os::Bsd::get_our_sigflags(sig) != 0 && (int)act.sa_flags != os::Bsd::get_our_sigflags(sig)) {
     tty->print("Warning: %s handler flags ", exception_name(sig, buf, O_BUFLEN));
     tty->print("expected:" PTR32_FORMAT, os::Bsd::get_our_sigflags(sig));
@@ -3745,9 +3756,6 @@
   return JNI_OK;
 }
 
-// this is called at the end of vm_initialization
-void os::init_3(void) { }
-
 // Mark the polling page as unreadable
 void os::make_polling_page_unreadable(void) {
   if( !guard_memory((char*)_polling_page, Bsd::page_size()) )
--- a/src/os/linux/vm/os_linux.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/os/linux/vm/os_linux.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1999, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1999, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -160,35 +160,6 @@
 // Declarations
 static void unpackTime(timespec* absTime, bool isAbsolute, jlong time);
 
-#ifdef JAVASE_EMBEDDED
-class MemNotifyThread: public Thread {
-  friend class VMStructs;
- public:
-  virtual void run();
-
- private:
-  static MemNotifyThread* _memnotify_thread;
-  int _fd;
-
- public:
-
-  // Constructor
-  MemNotifyThread(int fd);
-
-  // Tester
-  bool is_memnotify_thread() const { return true; }
-
-  // Printing
-  char* name() const { return (char*)"Linux MemNotify Thread"; }
-
-  // Returns the single instance of the MemNotifyThread
-  static MemNotifyThread* memnotify_thread() { return _memnotify_thread; }
-
-  // Create and start the single instance of MemNotifyThread
-  static void start();
-};
-#endif // JAVASE_EMBEDDED
-
 // utility functions
 
 static int SR_initialize();
@@ -238,44 +209,27 @@
 
 #ifndef SYS_gettid
 // i386: 224, ia64: 1105, amd64: 186, sparc 143
-#ifdef __ia64__
-#define SYS_gettid 1105
-#elif __i386__
-#define SYS_gettid 224
-#elif __amd64__
-#define SYS_gettid 186
-#elif __sparc__
-#define SYS_gettid 143
-#else
-#error define gettid for the arch
+  #ifdef __ia64__
+    #define SYS_gettid 1105
+  #else
+    #ifdef __i386__
+      #define SYS_gettid 224
+    #else
+      #ifdef __amd64__
+        #define SYS_gettid 186
+      #else
+        #ifdef __sparc__
+          #define SYS_gettid 143
+        #else
+          #error define gettid for the arch
+        #endif
+      #endif
+    #endif
+  #endif
 #endif
-#endif
 
 // Cpu architecture string
-#if   defined(ZERO)
-static char cpu_arch[] = ZERO_LIBARCH;
-#elif defined(IA64)
-static char cpu_arch[] = "ia64";
-#elif defined(IA32)
-static char cpu_arch[] = "i386";
-#elif defined(AMD64)
-static char cpu_arch[] = "amd64";
-#elif defined(ARM)
-static char cpu_arch[] = "arm";
-#elif defined(PPC32)
-static char cpu_arch[] = "ppc";
-#elif defined(PPC64)
-static char cpu_arch[] = "ppc64";
-#elif defined(SPARC)
-#  ifdef _LP64
-static char cpu_arch[] = "sparcv9";
-#  else
-static char cpu_arch[] = "sparc";
-#  endif
-#else
-#error Add appropriate cpu_arch setting
-#endif
-
+static char cpu_arch[] = HOTSPOT_LIB_ARCH;
 
 // pid_t gettid()
 //
@@ -4681,6 +4635,11 @@
     tty->print_cr("  found:%s", get_signal_handler_name(thisHandler, buf, O_BUFLEN));
     // No need to check this sig any longer
     sigaddset(&check_signal_done, sig);
+    // Running under non-interactive shell, SHUTDOWN2_SIGNAL will be reassigned SIG_IGN
+    if (sig == SHUTDOWN2_SIGNAL && !isatty(fileno(stdin))) {
+      tty->print_cr("Running in non-interactive shell, %s handler is replaced by shell",
+                    exception_name(sig, buf, O_BUFLEN));
+    }
   } else if(os::Linux::get_our_sigflags(sig) != 0 && (int)act.sa_flags != os::Linux::get_our_sigflags(sig)) {
     tty->print("Warning: %s handler flags ", exception_name(sig, buf, O_BUFLEN));
     tty->print("expected:" PTR32_FORMAT, os::Linux::get_our_sigflags(sig));
@@ -4936,17 +4895,6 @@
   return JNI_OK;
 }
 
-// this is called at the end of vm_initialization
-void os::init_3(void) {
-#ifdef JAVASE_EMBEDDED
-  // Start the MemNotifyThread
-  if (LowMemoryProtection) {
-    MemNotifyThread::start();
-  }
-  return;
-#endif
-}
-
 // Mark the polling page as unreadable
 void os::make_polling_page_unreadable(void) {
   if( !guard_memory((char*)_polling_page, Linux::page_size()) )
@@ -5992,14 +5940,6 @@
 
 extern char** environ;
 
-#ifndef __NR_fork
-#define __NR_fork IA32_ONLY(2) IA64_ONLY(not defined) AMD64_ONLY(57)
-#endif
-
-#ifndef __NR_execve
-#define __NR_execve IA32_ONLY(11) IA64_ONLY(1033) AMD64_ONLY(59)
-#endif
-
 // Run the specified command in a separate process. Return its exit value,
 // or -1 on failure (e.g. can't fork a new process).
 // Unlike system(), this function can be called from signal handler. It
@@ -6007,13 +5947,7 @@
 int os::fork_and_exec(char* cmd) {
   const char * argv[4] = {"sh", "-c", cmd, NULL};
 
-  // fork() in LinuxThreads/NPTL is not async-safe. It needs to run
-  // pthread_atfork handlers and reset pthread library. All we need is a
-  // separate process to execve. Make a direct syscall to fork process.
-  // On IA64 there's no fork syscall, we have to use fork() and hope for
-  // the best...
-  pid_t pid = NOT_IA64(syscall(__NR_fork);)
-              IA64_ONLY(fork();)
+  pid_t pid = fork();
 
   if (pid < 0) {
     // fork failed
@@ -6022,15 +5956,7 @@
   } else if (pid == 0) {
     // child process
 
-    // execve() in LinuxThreads will call pthread_kill_other_threads_np()
-    // first to kill every thread on the thread list. Because this list is
-    // not reset by fork() (see notes above), execve() will instead kill
-    // every thread in the parent process. We know this is the only thread
-    // in the new process, so make a system call directly.
-    // IA64 should use normal execve() from glibc to match the glibc fork()
-    // above.
-    NOT_IA64(syscall(__NR_execve, "/bin/sh", argv, environ);)
-    IA64_ONLY(execve("/bin/sh", (char* const*)argv, environ);)
+    execve("/bin/sh", (char* const*)argv, environ);
 
     // execve failed
     _exit(-1);
@@ -6123,83 +6049,6 @@
   return strlen(buffer);
 }
 
-#ifdef JAVASE_EMBEDDED
-//
-// A thread to watch the '/dev/mem_notify' device, which will tell us when the OS is running low on memory.
-//
-MemNotifyThread* MemNotifyThread::_memnotify_thread = NULL;
-
-// ctor
-//
-MemNotifyThread::MemNotifyThread(int fd): Thread() {
-  assert(memnotify_thread() == NULL, "we can only allocate one MemNotifyThread");
-  _fd = fd;
-
-  if (os::create_thread(this, os::os_thread)) {
-    _memnotify_thread = this;
-    os::set_priority(this, NearMaxPriority);
-    os::start_thread(this);
-  }
-}
-
-// Where all the work gets done
-//
-void MemNotifyThread::run() {
-  assert(this == memnotify_thread(), "expected the singleton MemNotifyThread");
-
-  // Set up the select arguments
-  fd_set rfds;
-  if (_fd != -1) {
-    FD_ZERO(&rfds);
-    FD_SET(_fd, &rfds);
-  }
-
-  // Now wait for the mem_notify device to wake up
-  while (1) {
-    // Wait for the mem_notify device to signal us..
-    int rc = select(_fd+1, _fd != -1 ? &rfds : NULL, NULL, NULL, NULL);
-    if (rc == -1) {
-      perror("select!\n");
-      break;
-    } else if (rc) {
-      //ssize_t free_before = os::available_memory();
-      //tty->print ("Notified: Free: %dK \n",os::available_memory()/1024);
-
-      // The kernel is telling us there is not much memory left...
-      // try to do something about that
-
-      // If we are not already in a GC, try one.
-      if (!Universe::heap()->is_gc_active()) {
-        Universe::heap()->collect(GCCause::_allocation_failure);
-
-        //ssize_t free_after = os::available_memory();
-        //tty->print ("Post-Notify: Free: %dK\n",free_after/1024);
-        //tty->print ("GC freed: %dK\n", (free_after - free_before)/1024);
-      }
-      // We might want to do something like the following if we find the GC's are not helping...
-      // Universe::heap()->size_policy()->set_gc_time_limit_exceeded(true);
-    }
-  }
-}
-
-//
-// See if the /dev/mem_notify device exists, and if so, start a thread to monitor it.
-//
-void MemNotifyThread::start() {
-  int    fd;
-  fd = open ("/dev/mem_notify", O_RDONLY, 0);
-  if (fd < 0) {
-      return;
-  }
-
-  if (memnotify_thread() == NULL) {
-    new MemNotifyThread(fd);
-  }
-}
-
-#endif // JAVASE_EMBEDDED
-
-
 /////////////// Unit tests ///////////////
 
 #ifndef PRODUCT
--- a/src/os/solaris/vm/jvm_solaris.h	Thu May 07 18:13:43 2015 -0700
+++ b/src/os/solaris/vm/jvm_solaris.h	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1998, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1998, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -41,7 +41,9 @@
  * JNI conversion, which should be sorted out later.
  */
 
+#define __USE_LEGACY_PROTOTYPES__
 #include <dirent.h>             /* For DIR */
+#undef __USE_LEGACY_PROTOTYPES__
 #include <sys/param.h>          /* For MAXPATHLEN */
 #include <sys/socket.h>         /* For socklen_t */
 #include <unistd.h>             /* For F_OK, R_OK, W_OK */
--- a/src/os/solaris/vm/os_solaris.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/os/solaris/vm/os_solaris.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1997, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1997, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -2696,29 +2696,30 @@
   }
 }
 
+size_t os::Solaris::page_size_for_alignment(size_t alignment) {
+  assert(is_size_aligned(alignment, (size_t) vm_page_size()),
+         err_msg(SIZE_FORMAT " is not aligned to " SIZE_FORMAT,
+                 alignment, (size_t) vm_page_size()));
+
+  for (int i = 0; _page_sizes[i] != 0; i++) {
+    if (is_size_aligned(alignment, _page_sizes[i])) {
+      return _page_sizes[i];
+    }
+  }
+
+  return (size_t) vm_page_size();
+}
+
 int os::Solaris::commit_memory_impl(char* addr, size_t bytes,
                                     size_t alignment_hint, bool exec) {
   int err = Solaris::commit_memory_impl(addr, bytes, exec);
-  if (err == 0) {
-    if (UseLargePages && (alignment_hint > (size_t)vm_page_size())) {
-      // If the large page size has been set and the VM
-      // is using large pages, use the large page size
-      // if it is smaller than the alignment hint. This is
-      // a case where the VM wants to use a larger alignment size
-      // for its own reasons but still want to use large pages
-      // (which is what matters to setting the mpss range.
-      size_t page_size = 0;
-      if (large_page_size() < alignment_hint) {
-        assert(UseLargePages, "Expected to be here for large page use only");
-        page_size = large_page_size();
-      } else {
-        // If the alignment hint is less than the large page
-        // size, the VM wants a particular alignment (thus the hint)
-        // for internal reasons.  Try to set the mpss range using
-        // the alignment_hint.
-        page_size = alignment_hint;
-      }
-      // Since this is a hint, ignore any failures.
+  if (err == 0 && UseLargePages && alignment_hint > 0) {
+    assert(is_size_aligned(bytes, alignment_hint),
+           err_msg(SIZE_FORMAT " is not aligned to " SIZE_FORMAT, bytes, alignment_hint));
+
+    // The syscall memcntl requires an exact page size (see man memcntl for details).
+    size_t page_size = page_size_for_alignment(alignment_hint);
+    if (page_size > (size_t) vm_page_size()) {
       (void)Solaris::setup_large_pages(addr, bytes, page_size);
     }
   }
@@ -3251,7 +3252,22 @@
   }
 }
 
+bool os::Solaris::is_valid_page_size(size_t bytes) {
+  for (int i = 0; _page_sizes[i] != 0; i++) {
+    if (_page_sizes[i] == bytes) {
+      return true;
+    }
+  }
+  return false;
+}
+
 bool os::Solaris::setup_large_pages(caddr_t start, size_t bytes, size_t align) {
+  assert(is_valid_page_size(align), err_msg(SIZE_FORMAT " is not a valid page size", align));
+  assert(is_ptr_aligned((void*) start, align),
+         err_msg(PTR_FORMAT " is not aligned to " SIZE_FORMAT, p2i((void*) start), align));
+  assert(is_size_aligned(bytes, align),
+         err_msg(SIZE_FORMAT " is not aligned to " SIZE_FORMAT, bytes, align));
+
   // Signal to OS that we want large pages for addresses
   // from addr, addr + bytes
   struct memcntl_mha mpss_struct;
@@ -4577,6 +4593,11 @@
     tty->print_cr("  found:%s", get_signal_handler_name(thisHandler, buf, O_BUFLEN));
     // No need to check this sig any longer
     sigaddset(&check_signal_done, sig);
+    // Running under non-interactive shell, SHUTDOWN2_SIGNAL will be reassigned SIG_IGN
+    if (sig == SHUTDOWN2_SIGNAL && !isatty(fileno(stdin))) {
+      tty->print_cr("Running in non-interactive shell, %s handler is replaced by shell",
+                    exception_name(sig, buf, O_BUFLEN));
+    }
   } else if(os::Solaris::get_our_sigflags(sig) != 0 && act.sa_flags != os::Solaris::get_our_sigflags(sig)) {
     tty->print("Warning: %s handler flags ", exception_name(sig, buf, O_BUFLEN));
     tty->print("expected:" PTR32_FORMAT, os::Solaris::get_our_sigflags(sig));
@@ -5194,10 +5215,6 @@
   return JNI_OK;
 }
 
-void os::init_3(void) {
-  return;
-}
-
 // Mark the polling page as unreadable
 void os::make_polling_page_unreadable(void) {
   if( mprotect((char *)_polling_page, page_size, PROT_NONE) != 0 )
--- a/src/os/solaris/vm/os_solaris.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/os/solaris/vm/os_solaris.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -110,6 +110,8 @@
   static meminfo_func_t _meminfo;
 
   // Large Page Support
+  static bool is_valid_page_size(size_t bytes);
+  static size_t page_size_for_alignment(size_t alignment);
   static bool setup_large_pages(caddr_t start, size_t bytes, size_t align);
 
   static void init_thread_fpu_state(void);
--- a/src/os/windows/vm/os_windows.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/os/windows/vm/os_windows.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -110,11 +110,13 @@
 static FILETIME process_kernel_time;
 
 #ifdef _M_IA64
-#define __CPU__ ia64
-#elif _M_AMD64
-#define __CPU__ amd64
+  #define __CPU__ ia64
 #else
-#define __CPU__ i486
+  #ifdef _M_AMD64
+    #define __CPU__ amd64
+  #else
+    #define __CPU__ i486
+  #endif
 #endif
 
 // save DLL module handle, used by GetModuleFileName
@@ -2138,20 +2140,22 @@
   // at the beginning of the target bundle.
   exceptionInfo->ContextRecord->StIPSR &= 0xFFFFF9FFFFFFFFFF;
   assert(((DWORD64)handler & 0xF) == 0, "Target address must point to the beginning of a bundle!");
-#elif _M_AMD64
+#else
+  #ifdef _M_AMD64
   // Do not blow up if no thread info available.
   if (thread) {
     thread->set_saved_exception_pc((address)(DWORD_PTR)exceptionInfo->ContextRecord->Rip);
   }
   // Set pc to handler
   exceptionInfo->ContextRecord->Rip = (DWORD64)handler;
-#else
+  #else
   // Do not blow up if no thread info available.
   if (thread) {
     thread->set_saved_exception_pc((address)(DWORD_PTR)exceptionInfo->ContextRecord->Eip);
   }
   // Set pc to handler
   exceptionInfo->ContextRecord->Eip = (DWORD)(DWORD_PTR)handler;
+  #endif
 #endif
 
   // Continue the execution
@@ -2250,7 +2254,8 @@
   // (division by zero is handled explicitly)
 #ifdef _M_IA64
   assert(0, "Fix Handle_IDiv_Exception");
-#elif _M_AMD64
+#else
+  #ifdef  _M_AMD64
   PCONTEXT ctx = exceptionInfo->ContextRecord;
   address pc = (address)ctx->Rip;
   assert(pc[0] == 0xF7, "not an idiv opcode");
@@ -2261,7 +2266,7 @@
   ctx->Rax = (DWORD)min_jint;      // result
   ctx->Rdx = (DWORD)0;             // remainder
   // Continue the execution
-#else
+  #else
   PCONTEXT ctx = exceptionInfo->ContextRecord;
   address pc = (address)ctx->Eip;
   assert(pc[0] == 0xF7, "not an idiv opcode");
@@ -2272,6 +2277,7 @@
   ctx->Eax = (DWORD)min_jint;      // result
   ctx->Edx = (DWORD)0;             // remainder
   // Continue the execution
+  #endif
 #endif
   return EXCEPTION_CONTINUE_EXECUTION;
 }
@@ -2351,10 +2357,12 @@
   // This is needed for IA64 because "relocation" / "implicit null check" / "poll instruction"
   // information is saved in the Unix format.
   address pc_unix_format = (address) ((((uint64_t)pc) & 0xFFFFFFFFFFFFFFF0) | ((((uint64_t)pc) & 0xF) >> 2));
-#elif _M_AMD64
+#else
+  #ifdef _M_AMD64
   address pc = (address) exceptionInfo->ContextRecord->Rip;
-#else
+  #else
   address pc = (address) exceptionInfo->ContextRecord->Eip;
+  #endif
 #endif
   Thread* t = ThreadLocalStorage::get_thread_slow();          // slow & steady
 
@@ -4062,10 +4070,6 @@
   return JNI_OK;
 }
 
-void os::init_3(void) {
-  return;
-}
-
 // Mark the polling page as unreadable
 void os::make_polling_page_unreadable(void) {
   DWORD old_status;
--- a/src/os_cpu/bsd_zero/vm/atomic_bsd_zero.inline.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/os_cpu/bsd_zero/vm/atomic_bsd_zero.inline.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -238,7 +238,13 @@
   // operation.  Note that some platforms only support this with the
   // limitation that the only valid value to store is the immediate
   // constant 1.  There is a test for this in JNI_CreateJavaVM().
-  return __sync_lock_test_and_set (dest, exchange_value);
+  jint result = __sync_lock_test_and_set (dest, exchange_value);
+  // All atomic operations are expected to be full memory barriers
+  // (see atomic.hpp). However, __sync_lock_test_and_set is not
+  // a full memory barrier, but an acquire barrier. Hence, this added
+  // barrier.
+  __sync_synchronize();
+  return result;
 #endif // M68K
 #endif // ARM
 }
@@ -251,7 +257,9 @@
 #ifdef M68K
   return m68k_lock_test_and_set(dest, exchange_value);
 #else
-  return __sync_lock_test_and_set (dest, exchange_value);
+  intptr_t result = __sync_lock_test_and_set (dest, exchange_value);
+  __sync_synchronize();
+  return result;
 #endif // M68K
 #endif // ARM
 }
--- a/src/os_cpu/linux_ppc/vm/os_linux_ppc.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/os_cpu/linux_ppc/vm/os_linux_ppc.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1,6 +1,6 @@
 /*
- * Copyright (c) 1997, 2014, Oracle and/or its affiliates. All rights reserved.
- * Copyright 2012, 2014 SAP AG. All rights reserved.
+ * Copyright (c) 1997, 2015, Oracle and/or its affiliates. All rights reserved.
+ * Copyright 2012, 2015 SAP AG. 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
@@ -290,6 +290,7 @@
         goto report_and_die;
       }
 
+      CodeBlob *cb = NULL;
       // Handle signal from NativeJump::patch_verified_entry().
       if (( TrapBasedNotEntrantChecks && sig == SIGTRAP && nativeInstruction_at(pc)->is_sigtrap_zombie_not_entrant()) ||
           (!TrapBasedNotEntrantChecks && sig == SIGILL  && nativeInstruction_at(pc)->is_sigill_zombie_not_entrant())) {
@@ -305,7 +306,10 @@
                // especially when we try to read from the safepoint polling page. So the check
                //   (address)info->si_addr == os::get_standard_polling_page()
                // doesn't work for us. We use:
-               ((NativeInstruction*)pc)->is_safepoint_poll()) {
+               ((NativeInstruction*)pc)->is_safepoint_poll() &&
+               CodeCache::contains((void*) pc) &&
+               ((cb = CodeCache::find_blob(pc)) != NULL) &&
+               cb->is_nmethod()) {
         if (TraceTraps) {
           tty->print_cr("trap: safepoint_poll at " INTPTR_FORMAT " (SIGSEGV)", p2i(pc));
         }
--- a/src/os_cpu/linux_sparc/vm/vm_version_linux_sparc.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/os_cpu/linux_sparc/vm/vm_version_linux_sparc.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -26,8 +26,8 @@
 #include "runtime/os.hpp"
 #include "vm_version_sparc.hpp"
 
-static bool detect_niagara() {
-  char cpu[128];
+static bool cpuinfo_field_contains(const char* field, const char* value) {
+  char line[1024];
   bool rv = false;
 
   FILE* fp = fopen("/proc/cpuinfo", "r");
@@ -35,9 +35,10 @@
     return rv;
   }
 
-  while (!feof(fp)) {
-    if (fscanf(fp, "cpu\t\t: %100[^\n]", &cpu) == 1) {
-      if (strstr(cpu, "Niagara") != NULL) {
+  while (fgets(line, sizeof(line), fp) != NULL) {
+    assert(strlen(line) < sizeof(line) - 1, "buffer line[1024] is too small.");
+    if (strncmp(line, field, strlen(field)) == 0) {
+      if (strstr(line, value) != NULL) {
         rv = true;
       }
       break;
@@ -45,8 +46,15 @@
   }
 
   fclose(fp);
+  return rv;
+}
 
-  return rv;
+static bool detect_niagara() {
+  return cpuinfo_field_contains("cpu", "Niagara");
+}
+
+static bool detect_blkinit() {
+  return cpuinfo_field_contains("cpucaps", "blkinit");
 }
 
 int VM_Version::platform_features(int features) {
@@ -58,5 +66,9 @@
     features = niagara1_m | T_family_m;
   }
 
+  if (detect_blkinit()) {
+    features |= blk_init_instructions_m;
+  }
+
   return features;
 }
--- a/src/os_cpu/linux_zero/vm/atomic_linux_zero.inline.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/os_cpu/linux_zero/vm/atomic_linux_zero.inline.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1,6 +1,6 @@
 /*
- * Copyright (c) 2003, 2013, Oracle and/or its affiliates. All rights reserved.
- * Copyright 2007, 2008, 2011 Red Hat, Inc.
+ * Copyright (c) 2003, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright 2007, 2008, 2011, 2015, 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
@@ -232,7 +232,13 @@
   // operation.  Note that some platforms only support this with the
   // limitation that the only valid value to store is the immediate
   // constant 1.  There is a test for this in JNI_CreateJavaVM().
-  return __sync_lock_test_and_set (dest, exchange_value);
+  jint result = __sync_lock_test_and_set (dest, exchange_value);
+  // All atomic operations are expected to be full memory barriers
+  // (see atomic.hpp). However, __sync_lock_test_and_set is not
+  // a full memory barrier, but an acquire barrier. Hence, this added
+  // barrier.
+  __sync_synchronize();
+  return result;
 #endif // M68K
 #endif // ARM
 }
@@ -245,7 +251,9 @@
 #ifdef M68K
   return m68k_lock_test_and_set(dest, exchange_value);
 #else
-  return __sync_lock_test_and_set (dest, exchange_value);
+  intptr_t result = __sync_lock_test_and_set (dest, exchange_value);
+  __sync_synchronize();
+  return result;
 #endif // M68K
 #endif // ARM
 }
--- a/src/os_cpu/solaris_sparc/vm/os_solaris_sparc.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/os_cpu/solaris_sparc/vm/os_solaris_sparc.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -264,7 +264,7 @@
                 CAST_FROM_FN_PTR(address, os::current_frame));
   if (os::is_first_C_frame(&myframe)) {
     // stack is not walkable
-    return frame(NULL, NULL, NULL);
+    return frame(NULL, NULL, false);
   } else {
     return os::get_sender_for_C_frame(&myframe);
   }
--- a/src/os_cpu/solaris_sparc/vm/vm_version_solaris_sparc.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/os_cpu/solaris_sparc/vm/vm_version_solaris_sparc.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -34,8 +34,7 @@
 #include <dlfcn.h>
 #include <link.h>
 
-extern "C" static int PICL_get_l1_data_cache_line_size_helper(picl_nodehdl_t nodeh, void *result);
-extern "C" static int PICL_get_l2_cache_line_size_helper(picl_nodehdl_t nodeh, void *result);
+extern "C" static int PICL_visit_cpu_helper(picl_nodehdl_t nodeh, void *result);
 
 // Functions from the library we need (signatures should match those in picl.h)
 extern "C" {
@@ -128,60 +127,87 @@
     bool is_inconsistent()  { return _state == INCONSISTENT; }
     void set_inconsistent() { _state = INCONSISTENT;         }
 
-    static int visit(picl_nodehdl_t nodeh, const char* name, void *arg) {
-      UniqueValueVisitor *state = static_cast<UniqueValueVisitor*>(arg);
-      PICL* picl = state->_picl;
-      assert(!state->is_inconsistent(), "Precondition");
+    bool visit(picl_nodehdl_t nodeh, const char* name) {
+      assert(!is_inconsistent(), "Precondition");
       int curr;
-      if (picl->get_int_property(nodeh, name, &curr) == PICL_SUCCESS) {
-        if (!state->is_assigned()) { // first iteration
-          state->set_value(curr);
-        } else if (curr != state->value()) { // following iterations
-          state->set_inconsistent();
+      if (_picl->get_int_property(nodeh, name, &curr) == PICL_SUCCESS) {
+        if (!is_assigned()) { // first iteration
+          set_value(curr);
+        } else if (curr != value()) { // following iterations
+          set_inconsistent();
+        }
+        return true;
+      }
+      return false;
+    }
+  };
+
+  class CPUVisitor {
+    UniqueValueVisitor _l1_visitor;
+    UniqueValueVisitor _l2_visitor;
+    int _limit; // number of times visit() can be run
+  public:
+    CPUVisitor(PICL *picl, int limit) : _l1_visitor(picl), _l2_visitor(picl), _limit(limit) {}
+    static int visit(picl_nodehdl_t nodeh, void *arg) {
+      CPUVisitor *cpu_visitor = static_cast<CPUVisitor*>(arg);
+      UniqueValueVisitor* l1_visitor = cpu_visitor->l1_visitor();
+      UniqueValueVisitor* l2_visitor = cpu_visitor->l2_visitor();
+      if (!l1_visitor->is_inconsistent()) {
+        l1_visitor->visit(nodeh, "l1-dcache-line-size");
+      }
+      static const char* l2_data_cache_line_property_name = NULL;
+      // On the first visit determine the name of the l2 cache line size property and memoize it.
+      if (l2_data_cache_line_property_name == NULL) {
+        assert(!l2_visitor->is_inconsistent(), "First iteration cannot be inconsistent");
+        l2_data_cache_line_property_name = "l2-cache-line-size";
+        if (!l2_visitor->visit(nodeh, l2_data_cache_line_property_name)) {
+          l2_data_cache_line_property_name = "l2-dcache-line-size";
+          l2_visitor->visit(nodeh, l2_data_cache_line_property_name);
+        }
+      } else {
+        if (!l2_visitor->is_inconsistent()) {
+          l2_visitor->visit(nodeh, l2_data_cache_line_property_name);
         }
       }
-      if (state->is_inconsistent()) {
+
+      if (l1_visitor->is_inconsistent() && l2_visitor->is_inconsistent()) {
+        return PICL_WALK_TERMINATE;
+      }
+      cpu_visitor->_limit--;
+      if (cpu_visitor->_limit <= 0) {
         return PICL_WALK_TERMINATE;
       }
       return PICL_WALK_CONTINUE;
     }
+    UniqueValueVisitor* l1_visitor() { return &_l1_visitor; }
+    UniqueValueVisitor* l2_visitor() { return &_l2_visitor; }
   };
-
   int _L1_data_cache_line_size;
-  int _L2_cache_line_size;
+  int _L2_data_cache_line_size;
 public:
-  static int get_l1_data_cache_line_size(picl_nodehdl_t nodeh, void *state) {
-    return UniqueValueVisitor::visit(nodeh, "l1-dcache-line-size", state);
-  }
-  static int get_l2_cache_line_size(picl_nodehdl_t nodeh, void *state) {
-    return UniqueValueVisitor::visit(nodeh, "l2-cache-line-size", state);
+  static int visit_cpu(picl_nodehdl_t nodeh, void *state) {
+    return CPUVisitor::visit(nodeh, state);
   }
 
-  PICL() : _L1_data_cache_line_size(0), _L2_cache_line_size(0), _dl_handle(NULL) {
+  PICL(bool is_fujitsu) : _L1_data_cache_line_size(0), _L2_data_cache_line_size(0), _dl_handle(NULL) {
     if (!open_library()) {
       return;
     }
     if (_picl_initialize() == PICL_SUCCESS) {
       picl_nodehdl_t rooth;
       if (_picl_get_root(&rooth) == PICL_SUCCESS) {
-        UniqueValueVisitor L1_state(this);
-        // Visit all "cpu" class instances
-        _picl_walk_tree_by_class(rooth, "cpu", &L1_state, PICL_get_l1_data_cache_line_size_helper);
-        if (L1_state.is_initial()) { // Still initial, iteration found no values
-          // Try walk all "core" class instances, it might be a Fujitsu machine
-          _picl_walk_tree_by_class(rooth, "core", &L1_state, PICL_get_l1_data_cache_line_size_helper);
+        const char* cpu_class = "cpu";
+        // If it's a Fujitsu machine, it's a "core"
+        if (is_fujitsu) {
+          cpu_class = "core";
         }
-        if (L1_state.is_assigned()) { // Is there a value?
-          _L1_data_cache_line_size = L1_state.value();
+        CPUVisitor cpu_visitor(this, os::processor_count());
+        _picl_walk_tree_by_class(rooth, cpu_class, &cpu_visitor, PICL_visit_cpu_helper);
+        if (cpu_visitor.l1_visitor()->is_assigned()) { // Is there a value?
+          _L1_data_cache_line_size = cpu_visitor.l1_visitor()->value();
         }
-
-        UniqueValueVisitor L2_state(this);
-        _picl_walk_tree_by_class(rooth, "cpu", &L2_state, PICL_get_l2_cache_line_size_helper);
-        if (L2_state.is_initial()) {
-          _picl_walk_tree_by_class(rooth, "core", &L2_state, PICL_get_l2_cache_line_size_helper);
-        }
-        if (L2_state.is_assigned()) {
-          _L2_cache_line_size = L2_state.value();
+        if (cpu_visitor.l2_visitor()->is_assigned()) {
+          _L2_data_cache_line_size = cpu_visitor.l2_visitor()->value();
         }
       }
       _picl_shutdown();
@@ -190,14 +216,12 @@
   }
 
   unsigned int L1_data_cache_line_size() const { return _L1_data_cache_line_size; }
-  unsigned int L2_cache_line_size() const      { return _L2_cache_line_size;      }
+  unsigned int L2_data_cache_line_size() const { return _L2_data_cache_line_size; }
 };
 
-extern "C" static int PICL_get_l1_data_cache_line_size_helper(picl_nodehdl_t nodeh, void *result) {
-  return PICL::get_l1_data_cache_line_size(nodeh, result);
-}
-extern "C" static int PICL_get_l2_cache_line_size_helper(picl_nodehdl_t nodeh, void *result) {
-  return PICL::get_l2_cache_line_size(nodeh, result);
+
+extern "C" static int PICL_visit_cpu_helper(picl_nodehdl_t nodeh, void *result) {
+  return PICL::visit_cpu(nodeh, result);
 }
 
 template<typename FuncType>
@@ -470,8 +494,8 @@
   }
 
   // Figure out cache line sizes using PICL
-  PICL picl;
-  _L2_cache_line_size      = picl.L2_cache_line_size();
+  PICL picl((features & sparc64_family_m) != 0);
+  _L2_data_cache_line_size = picl.L2_data_cache_line_size();
 
   return features;
 }
--- a/src/share/tools/LogCompilation/src/com/sun/hotspot/tools/compiler/LogParser.java	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/tools/LogCompilation/src/com/sun/hotspot/tools/compiler/LogParser.java	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2009, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2009, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -24,12 +24,12 @@
 
 /**
  * A SAX based parser of LogCompilation output from HotSpot.  It takes a complete
- * @author never
  */
 
 package com.sun.hotspot.tools.compiler;
 
 import java.io.FileReader;
+import java.io.PrintStream;
 import java.io.Reader;
 import java.util.ArrayList;
 import java.util.Collections;
@@ -133,6 +133,44 @@
         }
     };
 
+    class Jvms {
+        Jvms(Method method, int bci) {
+            this.method = method;
+            this.bci = bci;
+        }
+        final public Method method;
+        final public int bci;
+        final public String toString() {
+            return "@" + bci + " " + method;
+        }
+    }
+
+    class LockElimination extends BasicLogEvent {
+
+        ArrayList<Jvms> jvms = new ArrayList<Jvms>(1);
+        final String kind;
+        final String classId;
+        final String tagName;
+        LockElimination(String tagName, double start, String id, String kind, String classId) {
+            super(start, id);
+            this.kind = kind;
+            this.classId = classId;
+            this.tagName = tagName;
+        }
+
+        @Override
+        public void print(PrintStream stream) {
+            stream.printf("%s %s %s %s  %.3f ", getId(), tagName, kind, classId, getStart());
+            stream.print(jvms.toString());
+            stream.print("\n");
+        }
+
+        void addJVMS(Method method, int bci) {
+            jvms.add(new Jvms(method, bci));
+        }
+
+    }
+
     private ArrayList<LogEvent> events = new ArrayList<LogEvent>();
 
     private HashMap<String, String> types = new HashMap<String, String>();
@@ -145,6 +183,7 @@
     private Compilation compile;
     private CallSite site;
     private Stack<Phase> phaseStack = new Stack<Phase>();
+    private LockElimination currentLockElimination;
     private UncommonTrapEvent currentTrap;
     private Stack<CallSite> late_inline_scope;
 
@@ -188,7 +227,12 @@
         }
 
         LogParser log = new LogParser();
-        p.parse(new InputSource(reader), log);
+        try {
+            p.parse(new InputSource(reader), log);
+        } catch (Throwable th) {
+            th.printStackTrace();
+            // Carry on with what we've got...
+        }
 
         // Associate compilations with their NMethods
         for (NMethod nm : log.nmethods.values()) {
@@ -370,6 +414,15 @@
                 // uncommon trap inserted during parsing.
                 // ignore for now
             }
+        } else if (qname.startsWith("eliminate_lock")) {
+            String id = atts.getValue("compile_id");
+            if (id != null) {
+                id = makeId(atts);
+                String kind = atts.getValue("kind");
+                String classId = atts.getValue("class_id");
+                currentLockElimination = new LockElimination(qname, Double.parseDouble(search(atts, "stamp")), id, kind, classId);
+                events.add(currentLockElimination);
+            }
         } else if (qname.equals("late_inline")) {
             late_inline_scope = new Stack<CallSite>();
             site = new CallSite(-999, method(search(atts, "method")));
@@ -378,13 +431,14 @@
             // <jvms bci='4' method='java/io/DataInputStream readChar ()C' bytes='40' count='5815' iicount='20815'/>
             if (currentTrap != null) {
                 currentTrap.addJVMS(atts.getValue("method"), Integer.parseInt(atts.getValue("bci")));
+            } else if (currentLockElimination != null) {
+                  currentLockElimination.addJVMS(method(atts.getValue("method")), Integer.parseInt(atts.getValue("bci")));
             } else if (late_inline_scope != null) {
                 bci = Integer.parseInt(search(atts, "bci"));
                 site = new CallSite(bci, method(search(atts, "method")));
                 late_inline_scope.push(site);
             } else {
                 // Ignore <eliminate_allocation type='667'>,
-                //        <eliminate_lock lock='1'>,
                 //        <replace_string_concat arguments='2' string_alloc='0' multiple='0'>
             }
         } else if (qname.equals("nmethod")) {
@@ -437,6 +491,8 @@
             scopes.pop();
         } else if (qname.equals("uncommon_trap")) {
             currentTrap = null;
+        } else if (qname.startsWith("eliminate_lock")) {
+            currentLockElimination = null;
         } else if (qname.equals("late_inline")) {
             // Populate late inlining info.
 
@@ -445,8 +501,8 @@
             CallSite caller = late_inline_scope.pop();
             Method m = compile.getMethod();
             if (m != caller.getMethod()) {
-                System.out.println(m);
-                System.out.println(caller.getMethod() + " bci: " + bci);
+                System.err.println(m);
+                System.err.println(caller.getMethod() + " bci: " + bci);
                 throw new InternalError("call site and late_inline info don't match");
             }
 
--- a/src/share/tools/LogCompilation/src/com/sun/hotspot/tools/compiler/UncommonTrapEvent.java	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/tools/LogCompilation/src/com/sun/hotspot/tools/compiler/UncommonTrapEvent.java	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2009, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2009, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -50,7 +50,7 @@
     }
 
     public void print(PrintStream stream) {
-        stream.printf("%s uncommon trap %s %s\n", getId(), getReason(), getAction());
+        stream.printf("%s uncommon trap %.3f %s %s\n", getId(), getStart(), getReason(), getAction());
         stream.print(getJvms());
     }
 
--- a/src/share/vm/c1/c1_LIR.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/c1/c1_LIR.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2000, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -142,16 +142,11 @@
 
 
 #ifndef PRODUCT
-void LIR_Address::verify() const {
+void LIR_Address::verify0() const {
 #if defined(SPARC) || defined(PPC)
   assert(scale() == times_1, "Scaled addressing mode not available on SPARC/PPC and should not be used");
   assert(disp() == 0 || index()->is_illegal(), "can't have both");
 #endif
-#ifdef ARM
-  assert(disp() == 0 || index()->is_illegal(), "can't have both");
-  // Note: offsets higher than 4096 must not be rejected here. They can
-  // be handled by the back-end or will be rejected if not.
-#endif
 #ifdef _LP64
   assert(base()->is_cpu_register(), "wrong base operand");
   assert(index()->is_illegal() || index()->is_double_cpu(), "wrong index operand");
--- a/src/share/vm/c1/c1_LIR.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/c1/c1_LIR.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2000, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -25,6 +25,7 @@
 #ifndef SHARE_VM_C1_C1_LIR_HPP
 #define SHARE_VM_C1_C1_LIR_HPP
 
+#include "c1/c1_Defs.hpp"
 #include "c1/c1_ValueType.hpp"
 #include "oops/method.hpp"
 
@@ -561,7 +562,13 @@
   virtual BasicType type() const                 { return _type; }
   virtual void print_value_on(outputStream* out) const PRODUCT_RETURN;
 
-  void verify() const PRODUCT_RETURN;
+  void verify0() const PRODUCT_RETURN;
+#if defined(LIR_ADDRESS_PD_VERIFY) && !defined(PRODUCT)
+  void pd_verify() const;
+  void verify() const { pd_verify(); }
+#else
+  void verify() const { verify0(); }
+#endif
 
   static Scale scale(BasicType type);
 };
@@ -610,19 +617,15 @@
                                                                              LIR_OprDesc::float_type           |
                                                                              LIR_OprDesc::fpu_register         |
                                                                              LIR_OprDesc::single_size); }
-#if defined(ARM)
-  static LIR_Opr double_fpu(int reg1, int reg2)    { return (LIR_Opr)((reg1 << LIR_OprDesc::reg1_shift) | (reg2 << LIR_OprDesc::reg2_shift) | LIR_OprDesc::double_type | LIR_OprDesc::fpu_register | LIR_OprDesc::double_size); }
-  static LIR_Opr single_softfp(int reg)            { return (LIR_Opr)((reg  << LIR_OprDesc::reg1_shift) |                                     LIR_OprDesc::float_type  | LIR_OprDesc::cpu_register | LIR_OprDesc::single_size); }
-  static LIR_Opr double_softfp(int reg1, int reg2) { return (LIR_Opr)((reg1 << LIR_OprDesc::reg1_shift) | (reg2 << LIR_OprDesc::reg2_shift) | LIR_OprDesc::double_type | LIR_OprDesc::cpu_register | LIR_OprDesc::double_size); }
-#endif
-#ifdef SPARC
+#if defined(C1_LIR_MD_HPP)
+# include C1_LIR_MD_HPP
+#elif defined(SPARC)
   static LIR_Opr double_fpu(int reg1, int reg2) { return (LIR_Opr)(intptr_t)((reg1 << LIR_OprDesc::reg1_shift) |
                                                                              (reg2 << LIR_OprDesc::reg2_shift) |
                                                                              LIR_OprDesc::double_type          |
                                                                              LIR_OprDesc::fpu_register         |
                                                                              LIR_OprDesc::double_size); }
-#endif
-#ifdef X86
+#elif defined(X86)
   static LIR_Opr double_fpu(int reg)            { return (LIR_Opr)(intptr_t)((reg  << LIR_OprDesc::reg1_shift) |
                                                                              (reg  << LIR_OprDesc::reg2_shift) |
                                                                              LIR_OprDesc::double_type          |
@@ -640,8 +643,7 @@
                                                                              LIR_OprDesc::fpu_register         |
                                                                              LIR_OprDesc::double_size          |
                                                                              LIR_OprDesc::is_xmm_mask); }
-#endif // X86
-#ifdef PPC
+#elif defined(PPC)
   static LIR_Opr double_fpu(int reg)            { return (LIR_Opr)(intptr_t)((reg  << LIR_OprDesc::reg1_shift) |
                                                                              (reg  << LIR_OprDesc::reg2_shift) |
                                                                              LIR_OprDesc::double_type          |
--- a/src/share/vm/c1/c1_LIRGenerator.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/c1/c1_LIRGenerator.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2005, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -23,6 +23,7 @@
  */
 
 #include "precompiled.hpp"
+#include "c1/c1_Defs.hpp"
 #include "c1/c1_Compilation.hpp"
 #include "c1/c1_FrameMap.hpp"
 #include "c1/c1_Instruction.hpp"
@@ -46,10 +47,7 @@
 #define __ gen()->lir()->
 #endif
 
-// TODO: ARM - Use some recognizable constant which still fits architectural constraints
-#ifdef ARM
-#define PATCHED_ADDR  (204)
-#else
+#ifndef PATCHED_ADDR
 #define PATCHED_ADDR  (max_jint)
 #endif
 
@@ -1599,25 +1597,9 @@
   }
   assert(addr->is_register(), "must be a register at this point");
 
-#ifdef ARM
-  // TODO: ARM - move to platform-dependent code
-  LIR_Opr tmp = FrameMap::R14_opr;
-  if (VM_Version::supports_movw()) {
-    __ move((LIR_Opr)card_table_base, tmp);
-  } else {
-    __ move(new LIR_Address(FrameMap::Rthread_opr, in_bytes(JavaThread::card_table_base_offset()), T_ADDRESS), tmp);
-  }
-
-  CardTableModRefBS* ct = (CardTableModRefBS*)_bs;
-  LIR_Address *card_addr = new LIR_Address(tmp, addr, (LIR_Address::Scale) -CardTableModRefBS::card_shift, 0, T_BYTE);
-  if(((int)ct->byte_map_base & 0xff) == 0) {
-    __ move(tmp, card_addr);
-  } else {
-    LIR_Opr tmp_zero = new_register(T_INT);
-    __ move(LIR_OprFact::intConst(0), tmp_zero);
-    __ move(tmp_zero, card_addr);
-  }
-#else // ARM
+#ifdef CARDTABLEMODREF_POST_BARRIER_HELPER
+  CardTableModRef_post_barrier_helper(addr, card_table_base);
+#else
   LIR_Opr tmp = new_pointer_register();
   if (TwoOperandLIRForm) {
     __ move(addr, tmp);
@@ -1633,7 +1615,7 @@
               new LIR_Address(tmp, load_constant(card_table_base),
                               T_BYTE));
   }
-#endif // ARM
+#endif
 }
 
 
@@ -2121,7 +2103,7 @@
   } else {
 #ifdef X86
     addr = new LIR_Address(base_op, index_op, LIR_Address::Scale(log2_scale), 0, dst_type);
-#elif defined(ARM)
+#elif defined(GENERATE_ADDRESS_IS_PREFERRED)
     addr = generate_address(base_op, index_op, log2_scale, 0, dst_type);
 #else
     if (index_op->is_illegal() || log2_scale == 0) {
@@ -2175,6 +2157,9 @@
   LIR_Opr base_op = base.result();
   LIR_Opr index_op = idx.result();
 
+#ifdef GENERATE_ADDRESS_IS_PREFERRED
+  LIR_Address* addr = generate_address(base_op, index_op, log2_scale, 0, x->basic_type());
+#else
 #ifndef _LP64
   if (base_op->type() == T_LONG) {
     base_op = new_register(T_INT);
@@ -2216,6 +2201,7 @@
   }
 
   LIR_Address* addr = new LIR_Address(base_op, index_op, x->basic_type());
+#endif // !GENERATE_ADDRESS_IS_PREFERRED
   __ move(value.result(), addr);
 }
 
@@ -2569,7 +2555,7 @@
     // need to free up storage used for OSR entry point
     LIR_Opr osrBuffer = block()->next()->operand();
     BasicTypeList signature;
-    signature.append(T_INT);
+    signature.append(NOT_LP64(T_INT) LP64_ONLY(T_LONG)); // pass a pointer to osrBuffer
     CallingConvention* cc = frame_map()->c_calling_convention(&signature);
     __ move(osrBuffer, cc->args()->at(0));
     __ call_runtime_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::OSR_migration_end),
--- a/src/share/vm/c1/c1_LIRGenerator.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/c1/c1_LIRGenerator.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2005, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -277,6 +277,9 @@
 
   void G1SATBCardTableModRef_post_barrier(LIR_OprDesc* addr, LIR_OprDesc* new_val);
   void CardTableModRef_post_barrier(LIR_OprDesc* addr, LIR_OprDesc* new_val);
+#ifdef CARDTABLEMODREF_POST_BARRIER_HELPER
+  void CardTableModRef_post_barrier_helper(LIR_OprDesc* addr, LIR_Const* card_table_base);
+#endif
 
 
   static LIR_Opr result_register_for(ValueType* type, bool callee = false);
@@ -550,6 +553,10 @@
 #ifdef ASSERT
   virtual void do_Assert         (Assert*          x);
 #endif
+
+#ifdef C1_LIRGENERATOR_MD_HPP
+#include C1_LIRGENERATOR_MD_HPP
+#endif
 };
 
 
--- a/src/share/vm/c1/c1_LinearScan.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/c1/c1_LinearScan.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2005, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -2138,7 +2138,7 @@
         assert(interval->assigned_regHi() >= pd_first_fpu_reg && interval->assigned_regHi() <= pd_last_fpu_reg, "no fpu register");
         assert(assigned_reg % 2 == 0 && assigned_reg + 1 == interval->assigned_regHi(), "must be sequential and even");
         LIR_Opr result = LIR_OprFact::double_fpu(interval->assigned_regHi() - pd_first_fpu_reg, assigned_reg - pd_first_fpu_reg);
-#elif defined(ARM)
+#elif defined(ARM32)
         assert(assigned_reg >= pd_first_fpu_reg && assigned_reg <= pd_last_fpu_reg, "no fpu register");
         assert(interval->assigned_regHi() >= pd_first_fpu_reg && interval->assigned_regHi() <= pd_last_fpu_reg, "no fpu register");
         assert(assigned_reg % 2 == 0 && assigned_reg + 1 == interval->assigned_regHi(), "must be sequential and even");
@@ -2727,7 +2727,7 @@
 #ifdef SPARC
       assert(opr->fpu_regnrLo() == opr->fpu_regnrHi() + 1, "assumed in calculation (only fpu_regnrHi is used)");
 #endif
-#ifdef ARM
+#ifdef ARM32
       assert(opr->fpu_regnrHi() == opr->fpu_regnrLo() + 1, "assumed in calculation (only fpu_regnrLo is used)");
 #endif
 #ifdef PPC
--- a/src/share/vm/c1/c1_Runtime1.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/c1/c1_Runtime1.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1999, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1999, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -1085,7 +1085,7 @@
 #ifdef ARM
           if((load_klass_or_mirror_patch_id ||
               stub_id == Runtime1::load_appendix_patching_id) &&
-             !VM_Version::supports_movw()) {
+              nativeMovConstReg_at(copy_buff)->is_pc_relative()) {
             nmethod* nm = CodeCache::find_nmethod(instr_pc);
             address addr = NULL;
             assert(nm != NULL, "invalid nmethod_pc");
--- a/src/share/vm/ci/bcEscapeAnalyzer.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/ci/bcEscapeAnalyzer.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -42,7 +42,7 @@
   #define TRACE_BCEA(level, code)
 #endif
 
-// Maintain a map of which aguments a local variable or
+// Maintain a map of which arguments a local variable or
 // stack slot may contain.  In addition to tracking
 // arguments, it tracks two special values, "allocated"
 // which represents any object allocated in the current
@@ -318,14 +318,16 @@
     bool must_record_dependencies = false;
     for (i = arg_size - 1; i >= 0; i--) {
       ArgumentMap arg = state.raw_pop();
-      if (!is_argument(arg))
+      // Check if callee arg is a caller arg or an allocated object
+      bool allocated = arg.contains_allocated();
+      if (!(is_argument(arg) || allocated))
         continue;
       for (int j = 0; j < _arg_size; j++) {
         if (arg.contains(j)) {
           _arg_modified[j] |= analyzer._arg_modified[i];
         }
       }
-      if (!is_arg_stack(arg)) {
+      if (!(is_arg_stack(arg) || allocated)) {
         // arguments have already been recognized as escaping
       } else if (analyzer.is_arg_stack(i) && !analyzer.is_arg_returned(i)) {
         set_method_escape(arg);
--- a/src/share/vm/ci/ciMethod.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/ci/ciMethod.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -70,7 +70,8 @@
 // Loaded method.
 ciMethod::ciMethod(methodHandle h_m, ciInstanceKlass* holder) :
   ciMetadata(h_m()),
-  _holder(holder)
+  _holder(holder),
+  _has_injected_profile(false)
 {
   assert(h_m() != NULL, "no null method");
 
@@ -167,7 +168,8 @@
   _liveness(               NULL),
   _can_be_statically_bound(false),
   _method_blocks(          NULL),
-  _method_data(            NULL)
+  _method_data(            NULL),
+  _has_injected_profile(   false)
 #if defined(COMPILER2) || defined(SHARK)
   ,
   _flow(                   NULL),
@@ -689,7 +691,8 @@
 // via assert_unique_concrete_method or assert_leaf_type.
 ciMethod* ciMethod::find_monomorphic_target(ciInstanceKlass* caller,
                                             ciInstanceKlass* callee_holder,
-                                            ciInstanceKlass* actual_recv) {
+                                            ciInstanceKlass* actual_recv,
+                                            bool check_access) {
   check_is_loaded();
 
   if (actual_recv->is_interface()) {
@@ -697,7 +700,7 @@
     return NULL;
   }
 
-  ciMethod* root_m = resolve_invoke(caller, actual_recv);
+  ciMethod* root_m = resolve_invoke(caller, actual_recv, check_access);
   if (root_m == NULL) {
     // Something went wrong looking up the actual receiver method.
     return NULL;
@@ -776,7 +779,7 @@
 //
 // Given a known receiver klass, find the target for the call.
 // Return NULL if the call has no target or the target is abstract.
-ciMethod* ciMethod::resolve_invoke(ciKlass* caller, ciKlass* exact_receiver) {
+ciMethod* ciMethod::resolve_invoke(ciKlass* caller, ciKlass* exact_receiver, bool check_access) {
    check_is_loaded();
    VM_ENTRY_MARK;
 
@@ -793,9 +796,9 @@
         ||
        InstanceKlass::cast(h_recv())->is_linked() && !exact_receiver->is_interface()) {
      if (holder()->is_interface()) {
-       m = LinkResolver::resolve_interface_call_or_null(h_recv, h_resolved, h_name, h_signature, caller_klass);
+       m = LinkResolver::resolve_interface_call_or_null(h_recv, h_resolved, h_name, h_signature, caller_klass, check_access);
      } else {
-       m = LinkResolver::resolve_virtual_call_or_null(h_recv, h_resolved, h_name, h_signature, caller_klass);
+       m = LinkResolver::resolve_virtual_call_or_null(h_recv, h_resolved, h_name, h_signature, caller_klass, check_access);
      }
    }
 
--- a/src/share/vm/ci/ciMethod.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/ci/ciMethod.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -78,6 +78,7 @@
   bool _is_c1_compilable;
   bool _is_c2_compilable;
   bool _can_be_statically_bound;
+  bool _has_injected_profile;
 
   // Lazy fields, filled in on demand
   address              _code;
@@ -248,11 +249,12 @@
   // its calling environment.
   ciMethod* find_monomorphic_target(ciInstanceKlass* caller,
                                     ciInstanceKlass* callee_holder,
-                                    ciInstanceKlass* actual_receiver);
+                                    ciInstanceKlass* actual_receiver,
+                                    bool check_access = true);
 
   // Given a known receiver klass, find the target for the call.
   // Return NULL if the call has no target or is abstract.
-  ciMethod* resolve_invoke(ciKlass* caller, ciKlass* exact_receiver);
+  ciMethod* resolve_invoke(ciKlass* caller, ciKlass* exact_receiver, bool check_access = true);
 
   // Find the proper vtable index to invoke this method.
   int resolve_vtable_index(ciKlass* caller, ciKlass* receiver);
@@ -281,6 +283,9 @@
   int instructions_size();
   int scale_count(int count, float prof_factor = 1.);  // make MDO count commensurate with IIC
 
+  bool has_injected_profile() const { return _has_injected_profile;     }
+  void set_injected_profile(bool x) {        _has_injected_profile = x; }
+
   // Stack walking support
   bool is_ignored_by_security_stack_walk() const;
 
--- a/src/share/vm/ci/ciTypeFlow.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/ci/ciTypeFlow.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -36,6 +36,7 @@
 #include "interpreter/bytecodes.hpp"
 #include "memory/allocation.inline.hpp"
 #include "opto/compile.hpp"
+#include "opto/node.hpp"
 #include "runtime/deoptimization.hpp"
 #include "utilities/growableArray.hpp"
 
--- a/src/share/vm/classfile/classLoaderData.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/classfile/classLoaderData.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -320,27 +320,6 @@
   }
 }
 
-#ifdef ASSERT
-class AllAliveClosure : public OopClosure {
-  BoolObjectClosure* _is_alive_closure;
-  bool _found_dead;
- public:
-  AllAliveClosure(BoolObjectClosure* is_alive_closure) : _is_alive_closure(is_alive_closure), _found_dead(false) {}
-  template <typename T> void do_oop_work(T* p) {
-    T heap_oop = oopDesc::load_heap_oop(p);
-    if (!oopDesc::is_null(heap_oop)) {
-      oop obj = oopDesc::decode_heap_oop_not_null(heap_oop);
-      if (!_is_alive_closure->do_object_b(obj)) {
-        _found_dead = true;
-      }
-    }
-  }
-  void do_oop(oop* p)       { do_oop_work<oop>(p); }
-  void do_oop(narrowOop* p) { do_oop_work<narrowOop>(p); }
-  bool found_dead()         { return _found_dead; }
-};
-#endif
-
 oop ClassLoaderData::keep_alive_object() const {
   assert(!keep_alive(), "Don't use with CLDs that are artificially kept alive");
   return is_anonymous() ? _klasses->java_mirror() : class_loader();
@@ -350,15 +329,6 @@
   bool alive = keep_alive() // null class loader and incomplete anonymous klasses.
       || is_alive_closure->do_object_b(keep_alive_object());
 
-#ifdef ASSERT
-  if (alive) {
-    AllAliveClosure all_alive_closure(is_alive_closure);
-    KlassToOopClosure klass_closure(&all_alive_closure);
-    const_cast<ClassLoaderData*>(this)->oops_do(&all_alive_closure, &klass_closure, false);
-    assert(!all_alive_closure.found_dead(), err_msg("Found dead oop in alive cld: " PTR_FORMAT, p2i(this)));
-  }
-#endif
-
   return alive;
 }
 
@@ -912,7 +882,7 @@
 }
 
 Klass* ClassLoaderDataGraphKlassIteratorAtomic::next_klass() {
-  Klass* head = (Klass*)_next_klass;
+  Klass* head = _next_klass;
 
   while (head != NULL) {
     Klass* next = next_klass_in_cldg(head);
--- a/src/share/vm/classfile/classLoaderData.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/classfile/classLoaderData.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -307,7 +307,7 @@
 
 // An iterator that distributes Klasses to parallel worker threads.
 class ClassLoaderDataGraphKlassIteratorAtomic : public StackObj {
-  volatile Klass* _next_klass;
+ Klass* volatile _next_klass;
  public:
   ClassLoaderDataGraphKlassIteratorAtomic();
   Klass* next_klass();
--- a/src/share/vm/classfile/defaultMethods.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/classfile/defaultMethods.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2012, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2012, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -1096,6 +1096,7 @@
     }
     // update idnum for new location
     merged_methods->at(i)->set_method_idnum(i);
+    merged_methods->at(i)->set_orig_method_idnum(i);
   }
 
   // Verify correct order
--- a/src/share/vm/classfile/javaClasses.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/classfile/javaClasses.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1997, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1997, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -124,7 +124,7 @@
       tty->print_cr("  name: %s, sig: %s, flags: %08x", fs.name()->as_C_string(), fs.signature()->as_C_string(), fs.access_flags().as_int());
     }
 #endif //PRODUCT
-    fatal("Invalid layout of preloaded class");
+    vm_exit_during_initialization("Invalid layout of preloaded class: use -XX:+TraceClassLoading to see the origin of the problem class");
   }
   dest_offset = fd.offset();
 }
@@ -1278,7 +1278,8 @@
 }
 
 static inline bool version_matches(Method* method, int version) {
-  return (method->constants()->version() == version && version < MAX_VERSION);
+  assert(version < MAX_VERSION, "version is too big");
+  return method != NULL && (method->constants()->version() == version);
 }
 
 static inline int get_line_number(Method* method, int bci) {
@@ -1308,6 +1309,7 @@
   typeArrayOop    _methods;
   typeArrayOop    _bcis;
   objArrayOop     _mirrors;
+  typeArrayOop    _cprefs; // needed to insulate method name against redefinition
   int             _index;
   No_Safepoint_Verifier _nsv;
 
@@ -1315,8 +1317,9 @@
 
   enum {
     trace_methods_offset = java_lang_Throwable::trace_methods_offset,
-    trace_bcis_offset = java_lang_Throwable::trace_bcis_offset,
+    trace_bcis_offset    = java_lang_Throwable::trace_bcis_offset,
     trace_mirrors_offset = java_lang_Throwable::trace_mirrors_offset,
+    trace_cprefs_offset  = java_lang_Throwable::trace_cprefs_offset,
     trace_next_offset    = java_lang_Throwable::trace_next_offset,
     trace_size           = java_lang_Throwable::trace_size,
     trace_chunk_size     = java_lang_Throwable::trace_chunk_size
@@ -1338,9 +1341,14 @@
     assert(mirrors != NULL, "mirror array should be initialized in backtrace");
     return mirrors;
   }
+  static typeArrayOop get_cprefs(objArrayHandle chunk) {
+    typeArrayOop cprefs = typeArrayOop(chunk->obj_at(trace_cprefs_offset));
+    assert(cprefs != NULL, "cprefs array should be initialized in backtrace");
+    return cprefs;
+  }
 
   // constructor for new backtrace
-  BacktraceBuilder(TRAPS): _methods(NULL), _bcis(NULL), _head(NULL), _mirrors(NULL) {
+  BacktraceBuilder(TRAPS): _methods(NULL), _bcis(NULL), _head(NULL), _mirrors(NULL), _cprefs(NULL) {
     expand(CHECK);
     _backtrace = _head;
     _index = 0;
@@ -1350,6 +1358,7 @@
     _methods = get_methods(backtrace);
     _bcis = get_bcis(backtrace);
     _mirrors = get_mirrors(backtrace);
+    _cprefs = get_cprefs(backtrace);
     assert(_methods->length() == _bcis->length() &&
            _methods->length() == _mirrors->length(),
            "method and source information arrays should match");
@@ -1375,17 +1384,22 @@
     objArrayOop mirrors = oopFactory::new_objectArray(trace_chunk_size, CHECK);
     objArrayHandle new_mirrors(THREAD, mirrors);
 
+    typeArrayOop cprefs = oopFactory::new_shortArray(trace_chunk_size, CHECK);
+    typeArrayHandle new_cprefs(THREAD, cprefs);
+
     if (!old_head.is_null()) {
       old_head->obj_at_put(trace_next_offset, new_head());
     }
     new_head->obj_at_put(trace_methods_offset, new_methods());
     new_head->obj_at_put(trace_bcis_offset, new_bcis());
     new_head->obj_at_put(trace_mirrors_offset, new_mirrors());
+    new_head->obj_at_put(trace_cprefs_offset, new_cprefs());
 
     _head    = new_head();
     _methods = new_methods();
     _bcis = new_bcis();
     _mirrors = new_mirrors();
+    _cprefs  = new_cprefs();
     _index = 0;
   }
 
@@ -1405,8 +1419,9 @@
       method = mhandle();
     }
 
-    _methods->short_at_put(_index, method->method_idnum());
+    _methods->short_at_put(_index, method->orig_method_idnum());
     _bcis->int_at_put(_index, merge_bci_and_version(bci, method->constants()->version()));
+    _cprefs->short_at_put(_index, method->name_index());
 
     // We need to save the mirrors in the backtrace to keep the class
     // from being unloaded while we still have this stack trace.
@@ -1419,27 +1434,26 @@
 
 // Print stack trace element to resource allocated buffer
 char* java_lang_Throwable::print_stack_element_to_buffer(Handle mirror,
-                                  int method_id, int version, int bci) {
+                                  int method_id, int version, int bci, int cpref) {
 
   // Get strings and string lengths
   InstanceKlass* holder = InstanceKlass::cast(java_lang_Class::as_Klass(mirror()));
   const char* klass_name  = holder->external_name();
   int buf_len = (int)strlen(klass_name);
 
-  // The method id may point to an obsolete method, can't get more stack information
-  Method* method = holder->method_with_idnum(method_id);
-  if (method == NULL) {
-    char* buf = NEW_RESOURCE_ARRAY(char, buf_len + 64);
-    // This is what the java code prints in this case - added Redefined
-    sprintf(buf, "\tat %s.null (Redefined)", klass_name);
-    return buf;
-  }
-
-  char* method_name = method->name()->as_C_string();
+  Method* method = holder->method_with_orig_idnum(method_id, version);
+
+  // The method can be NULL if the requested class version is gone
+  Symbol* sym = (method != NULL) ? method->name() : holder->constants()->symbol_at(cpref);
+  char* method_name = sym->as_C_string();
   buf_len += (int)strlen(method_name);
 
+  // Use a specific ik version as a holder since the mirror might
+  // refer to a version that is now obsolete and no longer accessible
+  // via the previous versions list.
+  holder = holder->get_klass_version(version);
   char* source_file_name = NULL;
-  if (version_matches(method, version)) {
+  if (holder != NULL) {
     Symbol* source = holder->source_file_name();
     if (source != NULL) {
       source_file_name = source->as_C_string();
@@ -1481,17 +1495,18 @@
 }
 
 void java_lang_Throwable::print_stack_element(outputStream *st, Handle mirror,
-                                              int method_id, int version, int bci) {
+                                              int method_id, int version, int bci, int cpref) {
   ResourceMark rm;
-  char* buf = print_stack_element_to_buffer(mirror, method_id, version, bci);
+  char* buf = print_stack_element_to_buffer(mirror, method_id, version, bci, cpref);
   st->print_cr("%s", buf);
 }
 
 void java_lang_Throwable::print_stack_element(outputStream *st, methodHandle method, int bci) {
   Handle mirror = method->method_holder()->java_mirror();
-  int method_id = method->method_idnum();
+  int method_id = method->orig_method_idnum();
   int version = method->constants()->version();
-  print_stack_element(st, mirror, method_id, version, bci);
+  int cpref = method->name_index();
+  print_stack_element(st, mirror, method_id, version, bci, cpref);
 }
 
 const char* java_lang_Throwable::no_stack_trace_message() {
@@ -1516,6 +1531,7 @@
       typeArrayHandle methods (THREAD, BacktraceBuilder::get_methods(result));
       typeArrayHandle bcis (THREAD, BacktraceBuilder::get_bcis(result));
       objArrayHandle mirrors (THREAD, BacktraceBuilder::get_mirrors(result));
+      typeArrayHandle cprefs (THREAD, BacktraceBuilder::get_cprefs(result));
 
       int length = methods()->length();
       for (int index = 0; index < length; index++) {
@@ -1525,7 +1541,8 @@
         int method = methods->short_at(index);
         int version = version_at(bcis->int_at(index));
         int bci = bci_at(bcis->int_at(index));
-        print_stack_element(st, mirror, method, version, bci);
+        int cpref = cprefs->short_at(index);
+        print_stack_element(st, mirror, method, version, bci, cpref);
       }
       result = objArrayHandle(THREAD, objArrayOop(result->obj_at(trace_next_offset)));
     }
@@ -1809,29 +1826,30 @@
   if (chunk == NULL) {
     THROW_(vmSymbols::java_lang_IndexOutOfBoundsException(), NULL);
   }
-  // Get method id, bci, version and mirror from chunk
+  // Get method id, bci, version, mirror and cpref from chunk
   typeArrayOop methods = BacktraceBuilder::get_methods(chunk);
   typeArrayOop bcis = BacktraceBuilder::get_bcis(chunk);
   objArrayOop mirrors = BacktraceBuilder::get_mirrors(chunk);
+  typeArrayOop cprefs = BacktraceBuilder::get_cprefs(chunk);
 
   assert(methods != NULL && bcis != NULL && mirrors != NULL, "sanity check");
 
   int method = methods->short_at(chunk_index);
   int version = version_at(bcis->int_at(chunk_index));
   int bci = bci_at(bcis->int_at(chunk_index));
+  int cpref = cprefs->short_at(chunk_index);
   Handle mirror(THREAD, mirrors->obj_at(chunk_index));
 
   // Chunk can be partial full
   if (mirror.is_null()) {
     THROW_(vmSymbols::java_lang_IndexOutOfBoundsException(), NULL);
   }
-
-  oop element = java_lang_StackTraceElement::create(mirror, method, version, bci, CHECK_0);
+  oop element = java_lang_StackTraceElement::create(mirror, method, version, bci, cpref, CHECK_0);
   return element;
 }
 
 oop java_lang_StackTraceElement::create(Handle mirror, int method_id,
-                                        int version, int bci, TRAPS) {
+                                        int version, int bci, int cpref, TRAPS) {
   // Allocate java.lang.StackTraceElement instance
   Klass* k = SystemDictionary::StackTraceElement_klass();
   assert(k != NULL, "must be loaded in 1.4+");
@@ -1848,17 +1866,13 @@
   oop classname = StringTable::intern((char*) str, CHECK_0);
   java_lang_StackTraceElement::set_declaringClass(element(), classname);
 
-  Method* method = holder->method_with_idnum(method_id);
-  // Method on stack may be obsolete because it was redefined so cannot be
-  // found by idnum.
-  if (method == NULL) {
-    // leave name and fileName null
-    java_lang_StackTraceElement::set_lineNumber(element(), -1);
-    return element();
-  }
+  Method* method = holder->method_with_orig_idnum(method_id, version);
+
+  // The method can be NULL if the requested class version is gone
+  Symbol* sym = (method != NULL) ? method->name() : holder->constants()->symbol_at(cpref);
 
   // Fill in method name
-  oop methodname = StringTable::intern(method->name(), CHECK_0);
+  oop methodname = StringTable::intern(sym, CHECK_0);
   java_lang_StackTraceElement::set_methodName(element(), methodname);
 
   if (!version_matches(method, version)) {
@@ -1867,6 +1881,11 @@
     java_lang_StackTraceElement::set_lineNumber(element(), -1);
   } else {
     // Fill in source file name and line number.
+    // Use a specific ik version as a holder since the mirror might
+    // refer to a version that is now obsolete and no longer accessible
+    // via the previous versions list.
+    holder = holder->get_klass_version(version);
+    assert(holder != NULL, "sanity check");
     Symbol* source = holder->source_file_name();
     if (ShowHiddenFrames && source == NULL)
       source = vmSymbols::unknown_class_name();
@@ -1881,8 +1900,9 @@
 
 oop java_lang_StackTraceElement::create(methodHandle method, int bci, TRAPS) {
   Handle mirror (THREAD, method->method_holder()->java_mirror());
-  int method_id = method->method_idnum();
-  return create(mirror, method_id, method->constants()->version(), bci, THREAD);
+  int method_id = method->orig_method_idnum();
+  int cpref = method->name_index();
+  return create(mirror, method_id, method->constants()->version(), bci, cpref, THREAD);
 }
 
 void java_lang_reflect_AccessibleObject::compute_offsets() {
@@ -2781,33 +2801,6 @@
   return (flags(mname) & (MN_IS_METHOD | MN_IS_CONSTRUCTOR)) > 0;
 }
 
-#if INCLUDE_JVMTI
-// Can be executed on VM thread only
-void java_lang_invoke_MemberName::adjust_vmtarget(oop mname, Method* old_method,
-                                                  Method* new_method, bool* trace_name_printed) {
-  assert(is_method(mname), "wrong type");
-  assert(Thread::current()->is_VM_thread(), "not VM thread");
-
-  Method* target = (Method*)mname->address_field(_vmtarget_offset);
-  if (target == old_method) {
-    mname->address_field_put(_vmtarget_offset, (address)new_method);
-
-    if (RC_TRACE_IN_RANGE(0x00100000, 0x00400000)) {
-      if (!(*trace_name_printed)) {
-        // RC_TRACE_MESG macro has an embedded ResourceMark
-        RC_TRACE_MESG(("adjust: name=%s",
-                       old_method->method_holder()->external_name()));
-        *trace_name_printed = true;
-      }
-      // RC_TRACE macro has an embedded ResourceMark
-      RC_TRACE(0x00400000, ("MemberName method update: %s(%s)",
-                            new_method->name()->as_C_string(),
-                            new_method->signature()->as_C_string()));
-    }
-  }
-}
-#endif // INCLUDE_JVMTI
-
 void java_lang_invoke_MemberName::set_vmtarget(oop mname, Metadata* ref) {
   assert(is_instance(mname), "wrong type");
   // check the type of the vmtarget
@@ -3581,7 +3574,7 @@
     tty->print_cr("  name: %s, sig: %s, flags: %08x", fs.name()->as_C_string(), fs.signature()->as_C_string(), fs.access_flags().as_int());
   }
 #endif //PRODUCT
-  fatal("Invalid layout of preloaded class");
+  vm_exit_during_initialization("Invalid layout of preloaded class: use -XX:+TraceClassLoading to see the origin of the problem class");
   return -1;
 }
 
--- a/src/share/vm/classfile/javaClasses.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/classfile/javaClasses.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1997, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1997, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -484,8 +484,9 @@
     trace_methods_offset = 0,
     trace_bcis_offset    = 1,
     trace_mirrors_offset = 2,
-    trace_next_offset    = 3,
-    trace_size           = 4,
+    trace_cprefs_offset  = 3,
+    trace_next_offset    = 4,
+    trace_size           = 5,
     trace_chunk_size     = 32
   };
 
@@ -496,7 +497,7 @@
   static int static_unassigned_stacktrace_offset;
 
   // Printing
-  static char* print_stack_element_to_buffer(Handle mirror, int method, int version, int bci);
+  static char* print_stack_element_to_buffer(Handle mirror, int method, int version, int bci, int cpref);
   // StackTrace (programmatic access, new since 1.4)
   static void clear_stacktrace(oop throwable);
   // No stack trace available
@@ -517,7 +518,7 @@
   static oop message(Handle throwable);
   static void set_message(oop throwable, oop value);
   static void print_stack_element(outputStream *st, Handle mirror, int method,
-                                  int version, int bci);
+                                  int version, int bci, int cpref);
   static void print_stack_element(outputStream *st, methodHandle method, int bci);
   static void print_stack_usage(Handle stream);
 
@@ -1095,10 +1096,6 @@
 
   static Metadata*      vmtarget(oop mname);
   static void       set_vmtarget(oop mname, Metadata* target);
-#if INCLUDE_JVMTI
-  static void       adjust_vmtarget(oop mname, Method* old_method, Method* new_method,
-                                    bool* trace_name_printed);
-#endif // INCLUDE_JVMTI
 
   static intptr_t       vmindex(oop mname);
   static void       set_vmindex(oop mname, intptr_t index);
@@ -1326,7 +1323,7 @@
   static void set_lineNumber(oop element, int value);
 
   // Create an instance of StackTraceElement
-  static oop create(Handle mirror, int method, int version, int bci, TRAPS);
+  static oop create(Handle mirror, int method, int version, int bci, int cpref, TRAPS);
   static oop create(methodHandle method, int bci, TRAPS);
 
   // Debugging
--- a/src/share/vm/classfile/verifier.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/classfile/verifier.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1998, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1998, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -655,6 +655,7 @@
 
 
     bool this_uninit = false;  // Set to true when invokespecial <init> initialized 'this'
+    bool verified_exc_handlers = false;
 
     // Merge with the next instruction
     {
@@ -686,6 +687,18 @@
         }
       }
 
+      // Look for possible jump target in exception handlers and see if it
+      // matches current_frame.  Do this check here for astore*, dstore*,
+      // fstore*, istore*, and lstore* opcodes because they can change the type
+      // state by adding a local.  JVM Spec says that the incoming type state
+      // should be used for this check.  So, do the check here before a possible
+      // local is added to the type state.
+      if (Bytecodes::is_store_into_local(opcode) && bci >= ex_min && bci < ex_max) {
+        verify_exception_handler_targets(
+          bci, this_uninit, &current_frame, &stackmap_table, CHECK_VERIFY(this));
+        verified_exc_handlers = true;
+      }
+
       switch (opcode) {
         case Bytecodes::_nop :
           no_control_flow = false; break;
@@ -1662,9 +1675,13 @@
       }  // end switch
     }  // end Merge with the next instruction
 
-    // Look for possible jump target in exception handlers and see if it
-    // matches current_frame
-    if (bci >= ex_min && bci < ex_max) {
+    // Look for possible jump target in exception handlers and see if it matches
+    // current_frame.  Don't do this check if it has already been done (for
+    // ([a,d,f,i,l]store* opcodes).  This check cannot be done earlier because
+    // opcodes, such as invokespecial, may set the this_uninit flag.
+    assert(!(verified_exc_handlers && this_uninit),
+      "Exception handler targets got verified before this_uninit got set");
+    if (!verified_exc_handlers && bci >= ex_min && bci < ex_max) {
       verify_exception_handler_targets(
         bci, this_uninit, &current_frame, &stackmap_table, CHECK_VERIFY(this));
     }
@@ -2232,14 +2249,20 @@
 }
 
 // Look at the method's handlers.  If the bci is in the handler's try block
-// then check if the handler_pc is already on the stack.  If not, push it.
+// then check if the handler_pc is already on the stack.  If not, push it
+// unless the handler has already been scanned.
 void ClassVerifier::push_handlers(ExceptionTable* exhandlers,
+                                  GrowableArray<u4>* handler_list,
                                   GrowableArray<u4>* handler_stack,
                                   u4 bci) {
   int exlength = exhandlers->length();
   for(int x = 0; x < exlength; x++) {
     if (bci >= exhandlers->start_pc(x) && bci < exhandlers->end_pc(x)) {
-      handler_stack->append_if_missing(exhandlers->handler_pc(x));
+      u4 exhandler_pc = exhandlers->handler_pc(x);
+      if (!handler_list->contains(exhandler_pc)) {
+        handler_stack->append_if_missing(exhandler_pc);
+        handler_list->append(exhandler_pc);
+      }
     }
   }
 }
@@ -2257,6 +2280,10 @@
   GrowableArray<u4>* bci_stack = new GrowableArray<u4>(30);
   // Create stack for handlers for try blocks containing this handler.
   GrowableArray<u4>* handler_stack = new GrowableArray<u4>(30);
+  // Create list of handlers that have been pushed onto the handler_stack
+  // so that handlers embedded inside of their own TRY blocks only get
+  // scanned once.
+  GrowableArray<u4>* handler_list = new GrowableArray<u4>(30);
   // Create list of visited branch opcodes (goto* and if*).
   GrowableArray<u4>* visited_branches = new GrowableArray<u4>(30);
   ExceptionTable exhandlers(_method());
@@ -2275,7 +2302,7 @@
 
     // If the bytecode is in a TRY block, push its handlers so they
     // will get parsed.
-    push_handlers(&exhandlers, handler_stack, bci);
+    push_handlers(&exhandlers, handler_list, handler_stack, bci);
 
     switch (opcode) {
       case Bytecodes::_if_icmpeq:
@@ -2488,8 +2515,7 @@
     // of the current class.
     VerificationType objectref_type = new_class_type;
     if (name_in_supers(ref_class_type.name(), current_class())) {
-      Klass* ref_klass = load_class(
-        ref_class_type.name(), CHECK_VERIFY(this));
+      Klass* ref_klass = load_class(ref_class_type.name(), CHECK);
       Method* m = InstanceKlass::cast(ref_klass)->uncached_lookup_method(
         vmSymbols::object_initializer_name(),
         cp->signature_ref_at(bcs->get_index_u2()), Klass::normal);
--- a/src/share/vm/classfile/verifier.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/classfile/verifier.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -305,9 +305,10 @@
     bool* this_uninit, constantPoolHandle cp, StackMapTable* stackmap_table,
     TRAPS);
 
-  // Used by ends_in_athrow() to push all handlers that contain bci onto
-  // the handler_stack, if the handler is not already on the stack.
+  // Used by ends_in_athrow() to push all handlers that contain bci onto the
+  // handler_stack, if the handler has not already been pushed on the stack.
   void push_handlers(ExceptionTable* exhandlers,
+                     GrowableArray<u4>* handler_list,
                      GrowableArray<u4>* handler_stack,
                      u4 bci);
 
--- a/src/share/vm/classfile/vmSymbols.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/classfile/vmSymbols.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -244,7 +244,6 @@
   template(returnType_name,                           "returnType")                               \
   template(signature_name,                            "signature")                                \
   template(slot_name,                                 "slot")                                     \
-  template(selectAlternative_name,                    "selectAlternative")                        \
                                                                                                   \
   /* Support for annotations (JDK 1.5 and above) */                                               \
                                                                                                   \
@@ -296,8 +295,7 @@
   template(setTarget_signature,                       "(Ljava/lang/invoke/MethodHandle;)V")       \
   NOT_LP64(  do_alias(intptr_signature,               int_signature)  )                           \
   LP64_ONLY( do_alias(intptr_signature,               long_signature) )                           \
-  template(selectAlternative_signature, "(ZLjava/lang/invoke/MethodHandle;Ljava/lang/invoke/MethodHandle;)Ljava/lang/invoke/MethodHandle;") \
-                                                                      \
+                                                                                                  \
   /* common method and field names */                                                             \
   template(object_initializer_name,                   "<init>")                                   \
   template(class_initializer_name,                    "<clinit>")                                 \
@@ -868,6 +866,12 @@
    do_name(     fullFence_name,                                  "fullFence")                                           \
    do_alias(    fullFence_signature,                              void_method_signature)                                \
                                                                                                                         \
+  /* Custom branch frequencies profiling support for JSR292 */                                                          \
+  do_class(java_lang_invoke_MethodHandleImpl,               "java/lang/invoke/MethodHandleImpl")                        \
+  do_intrinsic(_profileBoolean, java_lang_invoke_MethodHandleImpl, profileBoolean_name, profileBoolean_signature,    F_S)  \
+   do_name(     profileBoolean_name,                               "profileBoolean")                                     \
+   do_signature(profileBoolean_signature,                           "(Z[I)Z")                                            \
+                                                                                                                        \
   /* unsafe memory references (there are a lot of them...) */                                                           \
   do_signature(getObject_signature,       "(Ljava/lang/Object;J)Ljava/lang/Object;")                                    \
   do_signature(putObject_signature,       "(Ljava/lang/Object;JLjava/lang/Object;)V")                                   \
--- a/src/share/vm/code/dependencies.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/code/dependencies.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -811,7 +811,13 @@
     assert((uint)n <= (uint)_num_participants, "oob");
     Method* fm = _found_methods[n];
     assert(n == _num_participants || fm != NULL, "proper usage");
-    assert(fm == NULL || fm->method_holder() == _participants[n], "sanity");
+    if (fm != NULL && fm->method_holder() != _participants[n]) {
+      // Default methods from interfaces can be added to classes. In
+      // that case the holder of the method is not the class but the
+      // interface where it's defined.
+      assert(fm->is_default_method(), "sanity");
+      return NULL;
+    }
     return fm;
   }
 
--- a/src/share/vm/code/vmreg.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/code/vmreg.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1998, 2012, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1998, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -32,25 +32,17 @@
 #ifdef COMPILER2
 #include "opto/adlcVMDeps.hpp"
 #include "utilities/ostream.hpp"
-#ifdef TARGET_ARCH_MODEL_x86_32
+#if defined ADGLOBALS_MD_HPP
+# include ADGLOBALS_MD_HPP
+#elif defined TARGET_ARCH_MODEL_x86_32
 # include "adfiles/adGlobals_x86_32.hpp"
-#endif
-#ifdef TARGET_ARCH_MODEL_x86_64
+#elif defined TARGET_ARCH_MODEL_x86_64
 # include "adfiles/adGlobals_x86_64.hpp"
-#endif
-#ifdef TARGET_ARCH_MODEL_sparc
+#elif defined TARGET_ARCH_MODEL_sparc
 # include "adfiles/adGlobals_sparc.hpp"
-#endif
-#ifdef TARGET_ARCH_MODEL_zero
+#elif defined TARGET_ARCH_MODEL_zero
 # include "adfiles/adGlobals_zero.hpp"
-#endif
-#ifdef TARGET_ARCH_MODEL_arm
-# include "adfiles/adGlobals_arm.hpp"
-#endif
-#ifdef TARGET_ARCH_MODEL_ppc_32
-# include "adfiles/adGlobals_ppc_32.hpp"
-#endif
-#ifdef TARGET_ARCH_MODEL_ppc_64
+#elif defined TARGET_ARCH_MODEL_ppc_64
 # include "adfiles/adGlobals_ppc_64.hpp"
 #endif
 #endif
--- a/src/share/vm/gc_implementation/concurrentMarkSweep/concurrentMarkSweepGeneration.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/gc_implementation/concurrentMarkSweep/concurrentMarkSweepGeneration.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -598,7 +598,7 @@
   _collector_policy(cp),
   _should_unload_classes(CMSClassUnloadingEnabled),
   _concurrent_cycles_since_last_unload(0),
-  _roots_scanning_options(SharedHeap::SO_None),
+  _roots_scanning_options(GenCollectedHeap::SO_None),
   _inter_sweep_estimate(CMS_SweepWeight, CMS_SweepPadding),
   _intra_sweep_estimate(CMS_SweepWeight, CMS_SweepPadding),
   _gc_tracer_cm(new (ResourceObj::C_HEAP, mtGC) CMSTracer()),
@@ -3068,7 +3068,7 @@
   gch->gen_process_roots(_cmsGen->level(),
                          true,   // younger gens are roots
                          true,   // activate StrongRootsScope
-                         SharedHeap::ScanningOption(roots_scanning_options()),
+                         GenCollectedHeap::ScanningOption(roots_scanning_options()),
                          should_unload_classes(),
                          &notOlder,
                          NULL,
@@ -3136,7 +3136,7 @@
   gch->gen_process_roots(_cmsGen->level(),
                          true,   // younger gens are roots
                          true,   // activate StrongRootsScope
-                         SharedHeap::ScanningOption(roots_scanning_options()),
+                         GenCollectedHeap::ScanningOption(roots_scanning_options()),
                          should_unload_classes(),
                          &notOlder,
                          NULL,
@@ -3327,7 +3327,7 @@
 void CMSCollector::setup_cms_unloading_and_verification_state() {
   const  bool should_verify =   VerifyBeforeGC || VerifyAfterGC || VerifyDuringGC
                              || VerifyBeforeExit;
-  const  int  rso           =   SharedHeap::SO_AllCodeCache;
+  const  int  rso           =   GenCollectedHeap::SO_AllCodeCache;
 
   // We set the proper root for this CMS cycle here.
   if (should_unload_classes()) {   // Should unload classes this cycle
@@ -3753,7 +3753,7 @@
       gch->gen_process_roots(_cmsGen->level(),
                              true,   // younger gens are roots
                              true,   // activate StrongRootsScope
-                             SharedHeap::ScanningOption(roots_scanning_options()),
+                             GenCollectedHeap::ScanningOption(roots_scanning_options()),
                              should_unload_classes(),
                              &notOlder,
                              NULL,
@@ -5254,13 +5254,13 @@
   gch->gen_process_roots(_collector->_cmsGen->level(),
                          false,     // yg was scanned above
                          false,     // this is parallel code
-                         SharedHeap::ScanningOption(_collector->CMSCollector::roots_scanning_options()),
+                         GenCollectedHeap::ScanningOption(_collector->CMSCollector::roots_scanning_options()),
                          _collector->should_unload_classes(),
                          &par_mri_cl,
                          NULL,
                          &cld_closure);
   assert(_collector->should_unload_classes()
-         || (_collector->CMSCollector::roots_scanning_options() & SharedHeap::SO_AllCodeCache),
+         || (_collector->CMSCollector::roots_scanning_options() & GenCollectedHeap::SO_AllCodeCache),
          "if we didn't scan the code cache, we have to be ready to drop nmethods with expired weak oops");
   _timer.stop();
   if (PrintCMSStatistics != 0) {
@@ -5390,14 +5390,14 @@
   gch->gen_process_roots(_collector->_cmsGen->level(),
                          false,     // yg was scanned above
                          false,     // this is parallel code
-                         SharedHeap::ScanningOption(_collector->CMSCollector::roots_scanning_options()),
+                         GenCollectedHeap::ScanningOption(_collector->CMSCollector::roots_scanning_options()),
                          _collector->should_unload_classes(),
                          &par_mrias_cl,
                          NULL,
                          NULL);     // The dirty klasses will be handled below
 
   assert(_collector->should_unload_classes()
-         || (_collector->CMSCollector::roots_scanning_options() & SharedHeap::SO_AllCodeCache),
+         || (_collector->CMSCollector::roots_scanning_options() & GenCollectedHeap::SO_AllCodeCache),
          "if we didn't scan the code cache, we have to be ready to drop nmethods with expired weak oops");
   _timer.stop();
   if (PrintCMSStatistics != 0) {
@@ -5982,14 +5982,14 @@
     gch->gen_process_roots(_cmsGen->level(),
                            true,  // younger gens as roots
                            false, // use the local StrongRootsScope
-                           SharedHeap::ScanningOption(roots_scanning_options()),
+                           GenCollectedHeap::ScanningOption(roots_scanning_options()),
                            should_unload_classes(),
                            &mrias_cl,
                            NULL,
                            NULL); // The dirty klasses will be handled below
 
     assert(should_unload_classes()
-           || (roots_scanning_options() & SharedHeap::SO_AllCodeCache),
+           || (roots_scanning_options() & GenCollectedHeap::SO_AllCodeCache),
            "if we didn't scan the code cache, we have to be ready to drop nmethods with expired weak oops");
   }
 
@@ -6634,7 +6634,6 @@
 }
 
 void CMSCollector::do_CMS_operation(CMS_op_type op, GCCause::Cause gc_cause) {
-  gclog_or_tty->date_stamp(PrintGC && PrintGCDateStamps);
   TraceCPUTime tcpu(PrintGCDetails, true, gclog_or_tty);
   GCTraceTime t(GCCauseString("GC", gc_cause), PrintGC, !PrintGCDetails, NULL, _gc_tracer_cm->gc_id());
   TraceCollectorStats tcs(counters());
--- a/src/share/vm/gc_implementation/concurrentMarkSweep/vmCMSOperations.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/gc_implementation/concurrentMarkSweep/vmCMSOperations.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2005, 2011, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -130,8 +130,8 @@
 class VM_GenCollectFullConcurrent: public VM_GC_Operation {
   bool _disabled_icms;
  public:
-  VM_GenCollectFullConcurrent(unsigned int gc_count_before,
-                              unsigned int full_gc_count_before,
+  VM_GenCollectFullConcurrent(uint gc_count_before,
+                              uint full_gc_count_before,
                               GCCause::Cause gc_cause)
     : VM_GC_Operation(gc_count_before, gc_cause, full_gc_count_before, true /* full */),
       _disabled_icms(false)
--- a/src/share/vm/gc_implementation/concurrentMarkSweep/vmStructs_cms.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/gc_implementation/concurrentMarkSweep/vmStructs_cms.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2007, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2007, 2014, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -41,7 +41,7 @@
   nonstatic_field(LinearAllocBlock,            _word_size,                                    size_t)                                \
   nonstatic_field(AFLBinaryTreeDictionary,     _total_size,                                   size_t)                                \
   nonstatic_field(CompactibleFreeListSpace,    _dictionary,                                   AFLBinaryTreeDictionary*)              \
-  nonstatic_field(CompactibleFreeListSpace,    _indexedFreeList[0],                           FreeList<FreeChunk>)                   \
+  nonstatic_field(CompactibleFreeListSpace,    _indexedFreeList[0],                           AdaptiveFreeList<FreeChunk>)           \
   nonstatic_field(CompactibleFreeListSpace,    _smallLinearAllocBlock,                        LinearAllocBlock)
 
 
--- a/src/share/vm/gc_implementation/g1/concurrentMark.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/gc_implementation/g1/concurrentMark.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -114,7 +114,7 @@
 }
 
 size_t CMBitMap::compute_size(size_t heap_size) {
-  return heap_size / mark_distance();
+  return ReservedSpace::allocation_align_size_up(heap_size / mark_distance());
 }
 
 size_t CMBitMap::mark_distance() {
@@ -2640,24 +2640,41 @@
   _nextMarkBitMap  = (CMBitMap*)  temp;
 }
 
-class CMObjectClosure;
-
-// Closure for iterating over objects, currently only used for
-// processing SATB buffers.
-class CMObjectClosure : public ObjectClosure {
+// Closure for marking entries in SATB buffers.
+class CMSATBBufferClosure : public SATBBufferClosure {
 private:
   CMTask* _task;
+  G1CollectedHeap* _g1h;
+
+  // This is very similar to CMTask::deal_with_reference, but with
+  // more relaxed requirements for the argument, so this must be more
+  // circumspect about treating the argument as an object.
+  void do_entry(void* entry) const {
+    _task->increment_refs_reached();
+    HeapRegion* hr = _g1h->heap_region_containing_raw(entry);
+    if (entry < hr->next_top_at_mark_start()) {
+      // Until we get here, we don't know whether entry refers to a valid
+      // object; it could instead have been a stale reference.
+      oop obj = static_cast<oop>(entry);
+      assert(obj->is_oop(true /* ignore mark word */),
+             err_msg("Invalid oop in SATB buffer: " PTR_FORMAT, p2i(obj)));
+      _task->make_reference_grey(obj, hr);
+    }
+  }
 
 public:
-  void do_object(oop obj) {
-    _task->deal_with_reference(obj);
+  CMSATBBufferClosure(CMTask* task, G1CollectedHeap* g1h)
+    : _task(task), _g1h(g1h) { }
+
+  virtual void do_buffer(void** buffer, size_t size) {
+    for (size_t i = 0; i < size; ++i) {
+      do_entry(buffer[i]);
+    }
   }
-
-  CMObjectClosure(CMTask* task) : _task(task) { }
 };
 
 class G1RemarkThreadsClosure : public ThreadClosure {
-  CMObjectClosure _cm_obj;
+  CMSATBBufferClosure _cm_satb_cl;
   G1CMOopClosure _cm_cl;
   MarkingCodeBlobClosure _code_cl;
   int _thread_parity;
@@ -2665,7 +2682,9 @@
 
  public:
   G1RemarkThreadsClosure(G1CollectedHeap* g1h, CMTask* task, bool is_par) :
-    _cm_obj(task), _cm_cl(g1h, g1h->concurrent_mark(), task), _code_cl(&_cm_cl, !CodeBlobToOopClosure::FixRelocations),
+    _cm_satb_cl(task, g1h),
+    _cm_cl(g1h, g1h->concurrent_mark(), task),
+    _code_cl(&_cm_cl, !CodeBlobToOopClosure::FixRelocations),
     _thread_parity(SharedHeap::heap()->strong_roots_parity()), _is_par(is_par) {}
 
   void do_thread(Thread* thread) {
@@ -2681,11 +2700,11 @@
         // live by the SATB invariant but other oops recorded in nmethods may behave differently.
         jt->nmethods_do(&_code_cl);
 
-        jt->satb_mark_queue().apply_closure_and_empty(&_cm_obj);
+        jt->satb_mark_queue().apply_closure_and_empty(&_cm_satb_cl);
       }
     } else if (thread->is_VM_thread()) {
       if (thread->claim_oops_do(_is_par, _thread_parity)) {
-        JavaThread::satb_mark_queue_set().shared_satb_queue()->apply_closure_and_empty(&_cm_obj);
+        JavaThread::satb_mark_queue_set().shared_satb_queue()->apply_closure_and_empty(&_cm_satb_cl);
       }
     }
   }
@@ -3059,9 +3078,7 @@
 #ifndef PRODUCT
 enum VerifyNoCSetOopsPhase {
   VerifyNoCSetOopsStack,
-  VerifyNoCSetOopsQueues,
-  VerifyNoCSetOopsSATBCompleted,
-  VerifyNoCSetOopsSATBThread
+  VerifyNoCSetOopsQueues
 };
 
 class VerifyNoCSetOopsClosure : public OopClosure, public ObjectClosure  {
@@ -3074,8 +3091,6 @@
     switch (_phase) {
     case VerifyNoCSetOopsStack:         return "Stack";
     case VerifyNoCSetOopsQueues:        return "Queue";
-    case VerifyNoCSetOopsSATBCompleted: return "Completed SATB Buffers";
-    case VerifyNoCSetOopsSATBThread:    return "Thread SATB Buffers";
     default:                            ShouldNotReachHere();
     }
     return NULL;
@@ -3102,7 +3117,7 @@
 
   virtual void do_oop(narrowOop* p) {
     // We should not come across narrow oops while scanning marking
-    // stacks and SATB buffers.
+    // stacks
     ShouldNotReachHere();
   }
 
@@ -3111,10 +3126,7 @@
   }
 };
 
-void ConcurrentMark::verify_no_cset_oops(bool verify_stacks,
-                                         bool verify_enqueued_buffers,
-                                         bool verify_thread_buffers,
-                                         bool verify_fingers) {
+void ConcurrentMark::verify_no_cset_oops() {
   assert(SafepointSynchronize::is_at_safepoint(), "should be at a safepoint");
   if (!G1CollectedHeap::heap()->mark_in_progress()) {
     return;
@@ -3122,65 +3134,47 @@
 
   VerifyNoCSetOopsClosure cl;
 
-  if (verify_stacks) {
-    // Verify entries on the global mark stack
-    cl.set_phase(VerifyNoCSetOopsStack);
-    _markStack.oops_do(&cl);
-
-    // Verify entries on the task queues
-    for (uint i = 0; i < _max_worker_id; i += 1) {
-      cl.set_phase(VerifyNoCSetOopsQueues, i);
-      CMTaskQueue* queue = _task_queues->queue(i);
-      queue->oops_do(&cl);
-    }
+  // Verify entries on the global mark stack
+  cl.set_phase(VerifyNoCSetOopsStack);
+  _markStack.oops_do(&cl);
+
+  // Verify entries on the task queues
+  for (uint i = 0; i < _max_worker_id; i += 1) {
+    cl.set_phase(VerifyNoCSetOopsQueues, i);
+    CMTaskQueue* queue = _task_queues->queue(i);
+    queue->oops_do(&cl);
   }
 
-  SATBMarkQueueSet& satb_qs = JavaThread::satb_mark_queue_set();
-
-  // Verify entries on the enqueued SATB buffers
-  if (verify_enqueued_buffers) {
-    cl.set_phase(VerifyNoCSetOopsSATBCompleted);
-    satb_qs.iterate_completed_buffers_read_only(&cl);
+  // Verify the global finger
+  HeapWord* global_finger = finger();
+  if (global_finger != NULL && global_finger < _heap_end) {
+    // The global finger always points to a heap region boundary. We
+    // use heap_region_containing_raw() to get the containing region
+    // given that the global finger could be pointing to a free region
+    // which subsequently becomes continues humongous. If that
+    // happens, heap_region_containing() will return the bottom of the
+    // corresponding starts humongous region and the check below will
+    // not hold any more.
+    // Since we always iterate over all regions, we might get a NULL HeapRegion
+    // here.
+    HeapRegion* global_hr = _g1h->heap_region_containing_raw(global_finger);
+    guarantee(global_hr == NULL || global_finger == global_hr->bottom(),
+              err_msg("global finger: "PTR_FORMAT" region: "HR_FORMAT,
+                      p2i(global_finger), HR_FORMAT_PARAMS(global_hr)));
   }
 
-  // Verify entries on the per-thread SATB buffers
-  if (verify_thread_buffers) {
-    cl.set_phase(VerifyNoCSetOopsSATBThread);
-    satb_qs.iterate_thread_buffers_read_only(&cl);
-  }
-
-  if (verify_fingers) {
-    // Verify the global finger
-    HeapWord* global_finger = finger();
-    if (global_finger != NULL && global_finger < _heap_end) {
-      // The global finger always points to a heap region boundary. We
-      // use heap_region_containing_raw() to get the containing region
-      // given that the global finger could be pointing to a free region
-      // which subsequently becomes continues humongous. If that
-      // happens, heap_region_containing() will return the bottom of the
-      // corresponding starts humongous region and the check below will
-      // not hold any more.
-      // Since we always iterate over all regions, we might get a NULL HeapRegion
-      // here.
-      HeapRegion* global_hr = _g1h->heap_region_containing_raw(global_finger);
-      guarantee(global_hr == NULL || global_finger == global_hr->bottom(),
-                err_msg("global finger: "PTR_FORMAT" region: "HR_FORMAT,
-                        p2i(global_finger), HR_FORMAT_PARAMS(global_hr)));
-    }
-
-    // Verify the task fingers
-    assert(parallel_marking_threads() <= _max_worker_id, "sanity");
-    for (int i = 0; i < (int) parallel_marking_threads(); i += 1) {
-      CMTask* task = _tasks[i];
-      HeapWord* task_finger = task->finger();
-      if (task_finger != NULL && task_finger < _heap_end) {
-        // See above note on the global finger verification.
-        HeapRegion* task_hr = _g1h->heap_region_containing_raw(task_finger);
-        guarantee(task_hr == NULL || task_finger == task_hr->bottom() ||
-                  !task_hr->in_collection_set(),
-                  err_msg("task finger: "PTR_FORMAT" region: "HR_FORMAT,
-                          p2i(task_finger), HR_FORMAT_PARAMS(task_hr)));
-      }
+  // Verify the task fingers
+  assert(parallel_marking_threads() <= _max_worker_id, "sanity");
+  for (int i = 0; i < (int) parallel_marking_threads(); i += 1) {
+    CMTask* task = _tasks[i];
+    HeapWord* task_finger = task->finger();
+    if (task_finger != NULL && task_finger < _heap_end) {
+      // See above note on the global finger verification.
+      HeapRegion* task_hr = _g1h->heap_region_containing_raw(task_finger);
+      guarantee(task_hr == NULL || task_finger == task_hr->bottom() ||
+                !task_hr->in_collection_set(),
+                err_msg("task finger: "PTR_FORMAT" region: "HR_FORMAT,
+                        p2i(task_finger), HR_FORMAT_PARAMS(task_hr)));
     }
   }
 }
@@ -3510,22 +3504,29 @@
 }
 #endif
 
-void CMTask::scan_object(oop obj) {
+template<bool scan>
+inline void CMTask::process_grey_object(oop obj) {
+  assert(scan || obj->is_typeArray(), "Skipping scan of grey non-typeArray");
   assert(_nextMarkBitMap->isMarked((HeapWord*) obj), "invariant");
 
   if (_cm->verbose_high()) {
-    gclog_or_tty->print_cr("[%u] we're scanning object "PTR_FORMAT,
+    gclog_or_tty->print_cr("[%u] processing grey object " PTR_FORMAT,
                            _worker_id, p2i((void*) obj));
   }
 
   size_t obj_size = obj->size();
   _words_scanned += obj_size;
 
-  obj->oop_iterate(_cm_oop_closure);
+  if (scan) {
+    obj->oop_iterate(_cm_oop_closure);
+  }
   statsOnly( ++_objs_scanned );
   check_limits();
 }
 
+template void CMTask::process_grey_object<true>(oop);
+template void CMTask::process_grey_object<false>(oop);
+
 // Closure for iteration over bitmaps
 class CMBitMapClosure : public BitMapClosure {
 private:
@@ -3994,34 +3995,18 @@
   // very counter productive if it did that. :-)
   _draining_satb_buffers = true;
 
-  CMObjectClosure oc(this);
+  CMSATBBufferClosure satb_cl(this, _g1h);
   SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
-  if (G1CollectedHeap::use_parallel_gc_threads()) {
-    satb_mq_set.set_par_closure(_worker_id, &oc);
-  } else {
-    satb_mq_set.set_closure(&oc);
-  }
 
   // This keeps claiming and applying the closure to completed buffers
   // until we run out of buffers or we need to abort.
-  if (G1CollectedHeap::use_parallel_gc_threads()) {
-    while (!has_aborted() &&
-           satb_mq_set.par_apply_closure_to_completed_buffer(_worker_id)) {
-      if (_cm->verbose_medium()) {
-        gclog_or_tty->print_cr("[%u] processed an SATB buffer", _worker_id);
-      }
-      statsOnly( ++_satb_buffers_processed );
-      regular_clock_call();
+  while (!has_aborted() &&
+         satb_mq_set.apply_closure_to_completed_buffer(&satb_cl)) {
+    if (_cm->verbose_medium()) {
+      gclog_or_tty->print_cr("[%u] processed an SATB buffer", _worker_id);
     }
-  } else {
-    while (!has_aborted() &&
-           satb_mq_set.apply_closure_to_completed_buffer()) {
-      if (_cm->verbose_medium()) {
-        gclog_or_tty->print_cr("[%u] processed an SATB buffer", _worker_id);
-      }
-      statsOnly( ++_satb_buffers_processed );
-      regular_clock_call();
-    }
+    statsOnly( ++_satb_buffers_processed );
+    regular_clock_call();
   }
 
   _draining_satb_buffers = false;
@@ -4030,12 +4015,6 @@
          concurrent() ||
          satb_mq_set.completed_buffers_num() == 0, "invariant");
 
-  if (G1CollectedHeap::use_parallel_gc_threads()) {
-    satb_mq_set.set_par_closure(_worker_id, NULL);
-  } else {
-    satb_mq_set.set_closure(NULL);
-  }
-
   // again, this was a potentially expensive operation, decrease the
   // limits to get the regular clock call early
   decrease_limits();
--- a/src/share/vm/gc_implementation/g1/concurrentMark.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/gc_implementation/g1/concurrentMark.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -793,14 +793,9 @@
   }
 
   // Verify that there are no CSet oops on the stacks (taskqueues /
-  // global mark stack), enqueued SATB buffers, per-thread SATB
-  // buffers, and fingers (global / per-task). The boolean parameters
-  // decide which of the above data structures to verify. If marking
-  // is not in progress, it's a no-op.
-  void verify_no_cset_oops(bool verify_stacks,
-                           bool verify_enqueued_buffers,
-                           bool verify_thread_buffers,
-                           bool verify_fingers) PRODUCT_RETURN;
+  // global mark stack) and fingers (global / per-task).
+  // If marking is not in progress, it's a no-op.
+  void verify_no_cset_oops() PRODUCT_RETURN;
 
   bool isPrevMarked(oop p) const {
     assert(p != NULL && p->is_oop(), "expected an oop");
@@ -1108,6 +1103,12 @@
   void regular_clock_call();
   bool concurrent() { return _concurrent; }
 
+  // Test whether obj might have already been passed over by the
+  // mark bitmap scan, and so needs to be pushed onto the mark stack.
+  bool is_below_finger(oop obj, HeapWord* global_finger) const;
+
+  template<bool scan> void process_grey_object(oop obj);
+
 public:
   // It resets the task; it should be called right at the beginning of
   // a marking phase.
@@ -1155,12 +1156,22 @@
 
   void set_cm_oop_closure(G1CMOopClosure* cm_oop_closure);
 
-  // It grays the object by marking it and, if necessary, pushing it
-  // on the local queue
+  // Increment the number of references this task has visited.
+  void increment_refs_reached() { ++_refs_reached; }
+
+  // Grey the object by marking it.  If not already marked, push it on
+  // the local queue if below the finger.
+  // Precondition: obj is in region.
+  // Precondition: obj is below region's NTAMS.
+  inline void make_reference_grey(oop obj, HeapRegion* region);
+
+  // Grey the object (by calling make_grey_reference) if required,
+  // e.g. obj is below its containing region's NTAMS.
+  // Precondition: obj is a valid heap object.
   inline void deal_with_reference(oop obj);
 
   // It scans an object and visits its children.
-  void scan_object(oop obj);
+  void scan_object(oop obj) { process_grey_object<true>(obj); }
 
   // It pushes an object on the local queue.
   inline void push(oop obj);
--- a/src/share/vm/gc_implementation/g1/concurrentMark.inline.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/gc_implementation/g1/concurrentMark.inline.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -259,14 +259,87 @@
              ++_local_pushes );
 }
 
-// This determines whether the method below will check both the local
-// and global fingers when determining whether to push on the stack a
-// gray object (value 1) or whether it will only check the global one
-// (value 0). The tradeoffs are that the former will be a bit more
-// accurate and possibly push less on the stack, but it might also be
-// a little bit slower.
+inline bool CMTask::is_below_finger(oop obj, HeapWord* global_finger) const {
+  // If obj is above the global finger, then the mark bitmap scan
+  // will find it later, and no push is needed.  Similarly, if we have
+  // a current region and obj is between the local finger and the
+  // end of the current region, then no push is needed.  The tradeoff
+  // of checking both vs only checking the global finger is that the
+  // local check will be more accurate and so result in fewer pushes,
+  // but may also be a little slower.
+  HeapWord* objAddr = (HeapWord*)obj;
+  if (_finger != NULL) {
+    // We have a current region.
 
-#define _CHECK_BOTH_FINGERS_      1
+    // Finger and region values are all NULL or all non-NULL.  We
+    // use _finger to check since we immediately use its value.
+    assert(_curr_region != NULL, "invariant");
+    assert(_region_limit != NULL, "invariant");
+    assert(_region_limit <= global_finger, "invariant");
+
+    // True if obj is less than the local finger, or is between
+    // the region limit and the global finger.
+    if (objAddr < _finger) {
+      return true;
+    } else if (objAddr < _region_limit) {
+      return false;
+    } // Else check global finger.
+  }
+  // Check global finger.
+  return objAddr < global_finger;
+}
+
+inline void CMTask::make_reference_grey(oop obj, HeapRegion* hr) {
+  if (_cm->par_mark_and_count(obj, hr, _marked_bytes_array, _card_bm)) {
+
+    if (_cm->verbose_high()) {
+      gclog_or_tty->print_cr("[%u] marked object " PTR_FORMAT,
+                             _worker_id, p2i(obj));
+    }
+
+    // No OrderAccess:store_load() is needed. It is implicit in the
+    // CAS done in CMBitMap::parMark() call in the routine above.
+    HeapWord* global_finger = _cm->finger();
+
+    // We only need to push a newly grey object on the mark
+    // stack if it is in a section of memory the mark bitmap
+    // scan has already examined.  Mark bitmap scanning
+    // maintains progress "fingers" for determining that.
+    //
+    // Notice that the global finger might be moving forward
+    // concurrently. This is not a problem. In the worst case, we
+    // mark the object while it is above the global finger and, by
+    // the time we read the global finger, it has moved forward
+    // past this object. In this case, the object will probably
+    // be visited when a task is scanning the region and will also
+    // be pushed on the stack. So, some duplicate work, but no
+    // correctness problems.
+    if (is_below_finger(obj, global_finger)) {
+      if (obj->is_typeArray()) {
+        // Immediately process arrays of primitive types, rather
+        // than pushing on the mark stack.  This keeps us from
+        // adding humongous objects to the mark stack that might
+        // be reclaimed before the entry is processed - see
+        // selection of candidates for eager reclaim of humongous
+        // objects.  The cost of the additional type test is
+        // mitigated by avoiding a trip through the mark stack,
+        // by only doing a bookkeeping update and avoiding the
+        // actual scan of the object - a typeArray contains no
+        // references, and the metadata is built-in.
+        process_grey_object<false>(obj);
+      } else {
+        if (_cm->verbose_high()) {
+          gclog_or_tty->print_cr("[%u] below a finger (local: " PTR_FORMAT
+                                 ", global: " PTR_FORMAT ") pushing "
+                                 PTR_FORMAT " on mark stack",
+                                 _worker_id, p2i(_finger),
+                                 p2i(global_finger), p2i(obj));
+        }
+        push(obj);
+      }
+    }
+  }
+}
 
 inline void CMTask::deal_with_reference(oop obj) {
   if (_cm->verbose_high()) {
@@ -274,7 +347,7 @@
                            _worker_id, p2i((void*) obj));
   }
 
-  ++_refs_reached;
+  increment_refs_reached();
 
   HeapWord* objAddr = (HeapWord*) obj;
   assert(obj->is_oop_or_null(true /* ignore mark word */), "Error");
@@ -286,62 +359,7 @@
       // anything with it).
       HeapRegion* hr = _g1h->heap_region_containing_raw(obj);
       if (!hr->obj_allocated_since_next_marking(obj)) {
-        if (_cm->verbose_high()) {
-          gclog_or_tty->print_cr("[%u] "PTR_FORMAT" is not considered marked",
-                                 _worker_id, p2i((void*) obj));
-        }
-
-        // we need to mark it first
-        if (_cm->par_mark_and_count(obj, hr, _marked_bytes_array, _card_bm)) {
-          // No OrderAccess:store_load() is needed. It is implicit in the
-          // CAS done in CMBitMap::parMark() call in the routine above.
-          HeapWord* global_finger = _cm->finger();
-
-#if _CHECK_BOTH_FINGERS_
-          // we will check both the local and global fingers
-
-          if (_finger != NULL && objAddr < _finger) {
-            if (_cm->verbose_high()) {
-              gclog_or_tty->print_cr("[%u] below the local finger ("PTR_FORMAT"), "
-                                     "pushing it", _worker_id, p2i(_finger));
-            }
-            push(obj);
-          } else if (_curr_region != NULL && objAddr < _region_limit) {
-            // do nothing
-          } else if (objAddr < global_finger) {
-            // Notice that the global finger might be moving forward
-            // concurrently. This is not a problem. In the worst case, we
-            // mark the object while it is above the global finger and, by
-            // the time we read the global finger, it has moved forward
-            // passed this object. In this case, the object will probably
-            // be visited when a task is scanning the region and will also
-            // be pushed on the stack. So, some duplicate work, but no
-            // correctness problems.
-
-            if (_cm->verbose_high()) {
-              gclog_or_tty->print_cr("[%u] below the global finger "
-                                     "("PTR_FORMAT"), pushing it",
-                                     _worker_id, p2i(global_finger));
-            }
-            push(obj);
-          } else {
-            // do nothing
-          }
-#else // _CHECK_BOTH_FINGERS_
-          // we will only check the global finger
-
-          if (objAddr < global_finger) {
-            // see long comment above
-
-            if (_cm->verbose_high()) {
-              gclog_or_tty->print_cr("[%u] below the global finger "
-                                     "("PTR_FORMAT"), pushing it",
-                                     _worker_id, p2i(global_finger));
-            }
-            push(obj);
-          }
-#endif // _CHECK_BOTH_FINGERS_
-        }
+        make_reference_grey(obj, hr);
       }
     }
   }
--- a/src/share/vm/gc_implementation/g1/g1AllocRegion.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/gc_implementation/g1/g1AllocRegion.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -254,25 +254,23 @@
 HeapRegion* SurvivorGCAllocRegion::allocate_new_region(size_t word_size,
                                                        bool force) {
   assert(!force, "not supported for GC alloc regions");
-  return _g1h->new_gc_alloc_region(word_size, count(), GCAllocForSurvived);
+  return _g1h->new_gc_alloc_region(word_size, count(), InCSetState::Young);
 }
 
 void SurvivorGCAllocRegion::retire_region(HeapRegion* alloc_region,
                                           size_t allocated_bytes) {
-  _g1h->retire_gc_alloc_region(alloc_region, allocated_bytes,
-                               GCAllocForSurvived);
+  _g1h->retire_gc_alloc_region(alloc_region, allocated_bytes, InCSetState::Young);
 }
 
 HeapRegion* OldGCAllocRegion::allocate_new_region(size_t word_size,
                                                   bool force) {
   assert(!force, "not supported for GC alloc regions");
-  return _g1h->new_gc_alloc_region(word_size, count(), GCAllocForTenured);
+  return _g1h->new_gc_alloc_region(word_size, count(), InCSetState::Old);
 }
 
 void OldGCAllocRegion::retire_region(HeapRegion* alloc_region,
                                      size_t allocated_bytes) {
-  _g1h->retire_gc_alloc_region(alloc_region, allocated_bytes,
-                               GCAllocForTenured);
+  _g1h->retire_gc_alloc_region(alloc_region, allocated_bytes, InCSetState::Old);
 }
 
 HeapRegion* OldGCAllocRegion::release() {
--- a/src/share/vm/gc_implementation/g1/g1Allocator.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/gc_implementation/g1/g1Allocator.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -59,7 +59,7 @@
       !(retained_region->top() == retained_region->end()) &&
       !retained_region->is_empty() &&
       !retained_region->isHumongous()) {
-    retained_region->record_top_and_timestamp();
+    retained_region->record_timestamp();
     // The retained region was added to the old region set when it was
     // retired. We have to remove it now, since we don't allow regions
     // we allocate to in the region sets. We'll re-add it later, when
@@ -94,6 +94,9 @@
   // want either way so no reason to check explicitly for either
   // condition.
   _retained_old_gc_alloc_region = old_gc_alloc_region(context)->release();
+  if (_retained_old_gc_alloc_region != NULL) {
+    _retained_old_gc_alloc_region->record_retained_region();
+  }
 
   if (ResizePLAB) {
     _g1h->_survivor_plab_stats.adjust_desired_plab_sz(no_of_gc_workers);
@@ -110,15 +113,16 @@
 G1ParGCAllocBuffer::G1ParGCAllocBuffer(size_t gclab_word_size) :
   ParGCAllocBuffer(gclab_word_size), _retired(true) { }
 
-HeapWord* G1ParGCAllocator::allocate_slow(GCAllocPurpose purpose, size_t word_sz, AllocationContext_t context) {
-  HeapWord* obj = NULL;
-  size_t gclab_word_size = _g1h->desired_plab_sz(purpose);
+HeapWord* G1ParGCAllocator::allocate_direct_or_new_plab(InCSetState dest,
+                                                        size_t word_sz,
+                                                        AllocationContext_t context) {
+  size_t gclab_word_size = _g1h->desired_plab_sz(dest);
   if (word_sz * 100 < gclab_word_size * ParallelGCBufferWastePct) {
-    G1ParGCAllocBuffer* alloc_buf = alloc_buffer(purpose, context);
+    G1ParGCAllocBuffer* alloc_buf = alloc_buffer(dest, context);
     add_to_alloc_buffer_waste(alloc_buf->words_remaining());
     alloc_buf->retire(false /* end_of_gc */, false /* retain */);
 
-    HeapWord* buf = _g1h->par_allocate_during_gc(purpose, gclab_word_size, context);
+    HeapWord* buf = _g1h->par_allocate_during_gc(dest, gclab_word_size, context);
     if (buf == NULL) {
       return NULL; // Let caller handle allocation failure.
     }
@@ -126,30 +130,33 @@
     alloc_buf->set_word_size(gclab_word_size);
     alloc_buf->set_buf(buf);
 
-    obj = alloc_buf->allocate(word_sz);
+    HeapWord* const obj = alloc_buf->allocate(word_sz);
     assert(obj != NULL, "buffer was definitely big enough...");
+    return obj;
   } else {
-    obj = _g1h->par_allocate_during_gc(purpose, word_sz, context);
+    return _g1h->par_allocate_during_gc(dest, word_sz, context);
   }
-  return obj;
 }
 
 G1DefaultParGCAllocator::G1DefaultParGCAllocator(G1CollectedHeap* g1h) :
-            G1ParGCAllocator(g1h),
-            _surviving_alloc_buffer(g1h->desired_plab_sz(GCAllocForSurvived)),
-            _tenured_alloc_buffer(g1h->desired_plab_sz(GCAllocForTenured)) {
-
-  _alloc_buffers[GCAllocForSurvived] = &_surviving_alloc_buffer;
-  _alloc_buffers[GCAllocForTenured]  = &_tenured_alloc_buffer;
-
+  G1ParGCAllocator(g1h),
+  _surviving_alloc_buffer(g1h->desired_plab_sz(InCSetState::Young)),
+  _tenured_alloc_buffer(g1h->desired_plab_sz(InCSetState::Old)) {
+  for (uint state = 0; state < InCSetState::Num; state++) {
+    _alloc_buffers[state] = NULL;
+  }
+  _alloc_buffers[InCSetState::Young] = &_surviving_alloc_buffer;
+  _alloc_buffers[InCSetState::Old]  = &_tenured_alloc_buffer;
 }
 
 void G1DefaultParGCAllocator::retire_alloc_buffers() {
-  for (int ap = 0; ap < GCAllocPurposeCount; ++ap) {
-    size_t waste = _alloc_buffers[ap]->words_remaining();
-    add_to_alloc_buffer_waste(waste);
-    _alloc_buffers[ap]->flush_stats_and_retire(_g1h->stats_for_purpose((GCAllocPurpose)ap),
-                                               true /* end_of_gc */,
-                                               false /* retain */);
+  for (uint state = 0; state < InCSetState::Num; state++) {
+    G1ParGCAllocBuffer* const buf = _alloc_buffers[state];
+    if (buf != NULL) {
+      add_to_alloc_buffer_waste(buf->words_remaining());
+      buf->flush_stats_and_retire(_g1h->alloc_buffer_stats(state),
+                                  true /* end_of_gc */,
+                                  false /* retain */);
+    }
   }
 }
--- a/src/share/vm/gc_implementation/g1/g1Allocator.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/gc_implementation/g1/g1Allocator.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -27,14 +27,9 @@
 
 #include "gc_implementation/g1/g1AllocationContext.hpp"
 #include "gc_implementation/g1/g1AllocRegion.hpp"
+#include "gc_implementation/g1/g1InCSetState.hpp"
 #include "gc_implementation/shared/parGCAllocBuffer.hpp"
 
-enum GCAllocPurpose {
-  GCAllocForTenured,
-  GCAllocForSurvived,
-  GCAllocPurposeCount
-};
-
 // Base class for G1 allocators.
 class G1Allocator : public CHeapObj<mtGC> {
   friend class VMStructs;
@@ -178,20 +173,40 @@
 protected:
   G1CollectedHeap* _g1h;
 
+  // The survivor alignment in effect in bytes.
+  // == 0 : don't align survivors
+  // != 0 : align survivors to that alignment
+  // These values were chosen to favor the non-alignment case since some
+  // architectures have a special compare against zero instructions.
+  const uint _survivor_alignment_bytes;
+
   size_t _alloc_buffer_waste;
   size_t _undo_waste;
 
   void add_to_alloc_buffer_waste(size_t waste) { _alloc_buffer_waste += waste; }
   void add_to_undo_waste(size_t waste)         { _undo_waste += waste; }
 
-  HeapWord* allocate_slow(GCAllocPurpose purpose, size_t word_sz, AllocationContext_t context);
+  virtual void retire_alloc_buffers() = 0;
+  virtual G1ParGCAllocBuffer* alloc_buffer(InCSetState dest, AllocationContext_t context) = 0;
 
-  virtual void retire_alloc_buffers() = 0;
-  virtual G1ParGCAllocBuffer* alloc_buffer(GCAllocPurpose purpose, AllocationContext_t context) = 0;
+  // Calculate the survivor space object alignment in bytes. Returns that or 0 if
+  // there are no restrictions on survivor alignment.
+  static uint calc_survivor_alignment_bytes() {
+    assert(SurvivorAlignmentInBytes >= ObjectAlignmentInBytes, "sanity");
+    if (SurvivorAlignmentInBytes == ObjectAlignmentInBytes) {
+      // No need to align objects in the survivors differently, return 0
+      // which means "survivor alignment is not used".
+      return 0;
+    } else {
+      assert(SurvivorAlignmentInBytes > 0, "sanity");
+      return SurvivorAlignmentInBytes;
+    }
+  }
 
 public:
   G1ParGCAllocator(G1CollectedHeap* g1h) :
-    _g1h(g1h), _alloc_buffer_waste(0), _undo_waste(0) {
+    _g1h(g1h), _survivor_alignment_bytes(calc_survivor_alignment_bytes()),
+    _alloc_buffer_waste(0), _undo_waste(0) {
   }
 
   static G1ParGCAllocator* create_allocator(G1CollectedHeap* g1h);
@@ -199,24 +214,40 @@
   size_t alloc_buffer_waste() { return _alloc_buffer_waste; }
   size_t undo_waste() {return _undo_waste; }
 
-  HeapWord* allocate(GCAllocPurpose purpose, size_t word_sz, AllocationContext_t context) {
-    HeapWord* obj = NULL;
-    if (purpose == GCAllocForSurvived) {
-      obj = alloc_buffer(purpose, context)->allocate_aligned(word_sz, SurvivorAlignmentInBytes);
+  // Allocate word_sz words in dest, either directly into the regions or by
+  // allocating a new PLAB. Returns the address of the allocated memory, NULL if
+  // not successful.
+  HeapWord* allocate_direct_or_new_plab(InCSetState dest,
+                                        size_t word_sz,
+                                        AllocationContext_t context);
+
+  // Allocate word_sz words in the PLAB of dest.  Returns the address of the
+  // allocated memory, NULL if not successful.
+  HeapWord* plab_allocate(InCSetState dest,
+                          size_t word_sz,
+                          AllocationContext_t context) {
+    G1ParGCAllocBuffer* buffer = alloc_buffer(dest, context);
+    if (_survivor_alignment_bytes == 0) {
+      return buffer->allocate(word_sz);
     } else {
-      obj = alloc_buffer(purpose, context)->allocate(word_sz);
+      return buffer->allocate_aligned(word_sz, _survivor_alignment_bytes);
     }
+  }
+
+  HeapWord* allocate(InCSetState dest, size_t word_sz,
+                     AllocationContext_t context) {
+    HeapWord* const obj = plab_allocate(dest, word_sz, context);
     if (obj != NULL) {
       return obj;
     }
-    return allocate_slow(purpose, word_sz, context);
+    return allocate_direct_or_new_plab(dest, word_sz, context);
   }
 
-  void undo_allocation(GCAllocPurpose purpose, HeapWord* obj, size_t word_sz, AllocationContext_t context) {
-    if (alloc_buffer(purpose, context)->contains(obj)) {
-      assert(alloc_buffer(purpose, context)->contains(obj + word_sz - 1),
+  void undo_allocation(InCSetState dest, HeapWord* obj, size_t word_sz, AllocationContext_t context) {
+    if (alloc_buffer(dest, context)->contains(obj)) {
+      assert(alloc_buffer(dest, context)->contains(obj + word_sz - 1),
              "should contain whole object");
-      alloc_buffer(purpose, context)->undo_allocation(obj, word_sz);
+      alloc_buffer(dest, context)->undo_allocation(obj, word_sz);
     } else {
       CollectedHeap::fill_with_object(obj, word_sz);
       add_to_undo_waste(word_sz);
@@ -227,13 +258,17 @@
 class G1DefaultParGCAllocator : public G1ParGCAllocator {
   G1ParGCAllocBuffer  _surviving_alloc_buffer;
   G1ParGCAllocBuffer  _tenured_alloc_buffer;
-  G1ParGCAllocBuffer* _alloc_buffers[GCAllocPurposeCount];
+  G1ParGCAllocBuffer* _alloc_buffers[InCSetState::Num];
 
 public:
   G1DefaultParGCAllocator(G1CollectedHeap* g1h);
 
-  virtual G1ParGCAllocBuffer* alloc_buffer(GCAllocPurpose purpose, AllocationContext_t context) {
-    return _alloc_buffers[purpose];
+  virtual G1ParGCAllocBuffer* alloc_buffer(InCSetState dest, AllocationContext_t context) {
+    assert(dest.is_valid(),
+           err_msg("Allocation buffer index out-of-bounds: " CSETSTATE_FORMAT, dest.value()));
+    assert(_alloc_buffers[dest.value()] != NULL,
+           err_msg("Allocation buffer is NULL: " CSETSTATE_FORMAT, dest.value()));
+    return _alloc_buffers[dest.value()];
   }
 
   virtual void retire_alloc_buffers() ;
--- a/src/share/vm/gc_implementation/g1/g1BiasedArray.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/gc_implementation/g1/g1BiasedArray.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -75,7 +75,7 @@
     assert((uintptr_t)end % mapping_granularity_in_bytes == 0,
       err_msg("end mapping area address must be a multiple of mapping granularity %zd, is " PTR_FORMAT,
         mapping_granularity_in_bytes, p2i(end)));
-    size_t num_target_elems = (end - bottom) / (mapping_granularity_in_bytes / HeapWordSize);
+    size_t num_target_elems = pointer_delta(end, bottom, mapping_granularity_in_bytes);
     idx_t bias = (uintptr_t)bottom / mapping_granularity_in_bytes;
     address base = create_new_base_array(num_target_elems, target_elem_size_in_bytes);
     initialize_base(base, num_target_elems, bias, target_elem_size_in_bytes, log2_intptr(mapping_granularity_in_bytes));
--- a/src/share/vm/gc_implementation/g1/g1CollectedHeap.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/gc_implementation/g1/g1CollectedHeap.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2001, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2001, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -46,6 +46,7 @@
 #include "gc_implementation/g1/g1ParScanThreadState.inline.hpp"
 #include "gc_implementation/g1/g1RegionToSpaceMapper.hpp"
 #include "gc_implementation/g1/g1RemSet.inline.hpp"
+#include "gc_implementation/g1/g1RootProcessor.hpp"
 #include "gc_implementation/g1/g1StringDedup.hpp"
 #include "gc_implementation/g1/g1YCTypes.hpp"
 #include "gc_implementation/g1/heapRegion.inline.hpp"
@@ -85,18 +86,6 @@
 // apply to TLAB allocation, which is not part of this interface: it
 // is done by clients of this interface.)
 
-// Notes on implementation of parallelism in different tasks.
-//
-// G1ParVerifyTask uses heap_region_par_iterate_chunked() for parallelism.
-// The number of GC workers is passed to heap_region_par_iterate_chunked().
-// It does use run_task() which sets _n_workers in the task.
-// G1ParTask executes g1_process_roots() ->
-// SharedHeap::process_roots() which calls eventually to
-// CardTableModRefBS::par_non_clean_card_iterate_work() which uses
-// SequentialSubTasksDone.  SharedHeap::process_roots() also
-// directly uses SubTasksDone (_process_strong_tasks field in SharedHeap).
-//
-
 // Local to this file.
 
 class RefineCardTableEntryClosure: public CardTableEntryClosure {
@@ -364,7 +353,7 @@
   HeapRegion* lists[] = {_head,   _survivor_head};
   const char* names[] = {"YOUNG", "SURVIVOR"};
 
-  for (unsigned int list = 0; list < ARRAY_SIZE(lists); ++list) {
+  for (uint list = 0; list < ARRAY_SIZE(lists); ++list) {
     gclog_or_tty->print_cr("%s LIST CONTENTS", names[list]);
     HeapRegion *curr = lists[list];
     if (curr == NULL)
@@ -838,8 +827,8 @@
   assert_heap_not_locked_and_not_at_safepoint();
   assert(!isHumongous(word_size), "we do not allow humongous TLABs");
 
-  unsigned int dummy_gc_count_before;
-  int dummy_gclocker_retry_count = 0;
+  uint dummy_gc_count_before;
+  uint dummy_gclocker_retry_count = 0;
   return attempt_allocation(word_size, &dummy_gc_count_before, &dummy_gclocker_retry_count);
 }
 
@@ -849,8 +838,8 @@
   assert_heap_not_locked_and_not_at_safepoint();
 
   // Loop until the allocation is satisfied, or unsatisfied after GC.
-  for (int try_count = 1, gclocker_retry_count = 0; /* we'll return */; try_count += 1) {
-    unsigned int gc_count_before;
+  for (uint try_count = 1, gclocker_retry_count = 0; /* we'll return */; try_count += 1) {
+    uint gc_count_before;
 
     HeapWord* result = NULL;
     if (!isHumongous(word_size)) {
@@ -902,8 +891,8 @@
 
 HeapWord* G1CollectedHeap::attempt_allocation_slow(size_t word_size,
                                                    AllocationContext_t context,
-                                                   unsigned int *gc_count_before_ret,
-                                                   int* gclocker_retry_count_ret) {
+                                                   uint* gc_count_before_ret,
+                                                   uint* gclocker_retry_count_ret) {
   // Make sure you read the note in attempt_allocation_humongous().
 
   assert_heap_not_locked_and_not_at_safepoint();
@@ -920,7 +909,7 @@
   HeapWord* result = NULL;
   for (int try_count = 1; /* we'll return */; try_count += 1) {
     bool should_try_gc;
-    unsigned int gc_count_before;
+    uint gc_count_before;
 
     {
       MutexLockerEx x(Heap_lock);
@@ -964,7 +953,7 @@
     if (should_try_gc) {
       bool succeeded;
       result = do_collection_pause(word_size, gc_count_before, &succeeded,
-          GCCause::_g1_inc_collection_pause);
+                                   GCCause::_g1_inc_collection_pause);
       if (result != NULL) {
         assert(succeeded, "only way to get back a non-NULL result");
         return result;
@@ -1018,8 +1007,8 @@
 }
 
 HeapWord* G1CollectedHeap::attempt_allocation_humongous(size_t word_size,
-                                                        unsigned int * gc_count_before_ret,
-                                                        int* gclocker_retry_count_ret) {
+                                                        uint* gc_count_before_ret,
+                                                        uint* gclocker_retry_count_ret) {
   // The structure of this method has a lot of similarities to
   // attempt_allocation_slow(). The reason these two were not merged
   // into a single one is that such a method would require several "if
@@ -1052,7 +1041,7 @@
   HeapWord* result = NULL;
   for (int try_count = 1; /* we'll return */; try_count += 1) {
     bool should_try_gc;
-    unsigned int gc_count_before;
+    uint gc_count_before;
 
     {
       MutexLockerEx x(Heap_lock);
@@ -1090,7 +1079,7 @@
 
       bool succeeded;
       result = do_collection_pause(word_size, gc_count_before, &succeeded,
-          GCCause::_g1_humongous_allocation);
+                                   GCCause::_g1_humongous_allocation);
       if (result != NULL) {
         assert(succeeded, "only way to get back a non-NULL result");
         return result;
@@ -1297,7 +1286,6 @@
 
     // Timing
     assert(gc_cause() != GCCause::_java_lang_system_gc || explicit_gc, "invariant");
-    gclog_or_tty->date_stamp(G1Log::fine() && PrintGCDateStamps);
     TraceCPUTime tcpu(G1Log::finer(), true, gclog_or_tty);
 
     {
@@ -1855,7 +1843,6 @@
   _is_alive_closure_stw(this),
   _ref_processor_cm(NULL),
   _ref_processor_stw(NULL),
-  _process_strong_tasks(new SubTasksDone(G1H_PS_NumElements)),
   _bot_shared(NULL),
   _evac_failure_scan_stack(NULL),
   _mark_in_progress(false),
@@ -1866,7 +1853,7 @@
   _secondary_free_list("Secondary Free List", new SecondaryFreeRegionListMtSafeChecker()),
   _old_set("Old Set", false /* humongous */, new OldRegionSetMtSafeChecker()),
   _humongous_set("Master Humongous Set", true /* humongous */, new HumongousRegionSetMtSafeChecker()),
-  _humongous_is_live(),
+  _humongous_reclaim_candidates(),
   _has_humongous_reclaim_candidates(false),
   _free_regions_coming(false),
   _young_list(new YoungList(this)),
@@ -1878,6 +1865,7 @@
   _old_marking_cycles_started(0),
   _old_marking_cycles_completed(0),
   _concurrent_cycle_started(false),
+  _heap_summary_sent(false),
   _in_cset_fast_test(),
   _dirty_cards_region_list(NULL),
   _worker_cset_start_region(NULL),
@@ -1888,9 +1876,6 @@
   _gc_tracer_cm(new (ResourceObj::C_HEAP, mtGC) G1OldTracer()) {
 
   _g1h = this;
-  if (_process_strong_tasks == NULL || !_process_strong_tasks->valid()) {
-    vm_exit_during_initialization("Failed necessary allocation.");
-  }
 
   _allocator = G1Allocator::create_allocator(_g1h);
   _humongous_object_threshold_in_words = HeapRegion::GrainWords / 2;
@@ -1902,7 +1887,7 @@
   assert(n_rem_sets > 0, "Invariant.");
 
   _worker_cset_start_region = NEW_C_HEAP_ARRAY(HeapRegion*, n_queues, mtGC);
-  _worker_cset_start_region_time_stamp = NEW_C_HEAP_ARRAY(unsigned int, n_queues, mtGC);
+  _worker_cset_start_region_time_stamp = NEW_C_HEAP_ARRAY(uint, n_queues, mtGC);
   _evacuation_failed_info_array = NEW_C_HEAP_ARRAY(EvacuationFailedInfo, n_queues, mtGC);
 
   for (int i = 0; i < n_queues; i++) {
@@ -1919,6 +1904,26 @@
   guarantee(_task_queues != NULL, "task_queues allocation failure.");
 }
 
+G1RegionToSpaceMapper* G1CollectedHeap::create_aux_memory_mapper(const char* description,
+                                                                 size_t size,
+                                                                 size_t translation_factor) {
+  size_t preferred_page_size = os::page_size_for_region_unaligned(size, 1);
+  // Allocate a new reserved space, preferring to use large pages.
+  ReservedSpace rs(size, preferred_page_size);
+  G1RegionToSpaceMapper* result  =
+    G1RegionToSpaceMapper::create_mapper(rs,
+                                         size,
+                                         rs.alignment(),
+                                         HeapRegion::GrainBytes,
+                                         translation_factor,
+                                         mtGC);
+  if (TracePageSizes) {
+    gclog_or_tty->print_cr("G1 '%s': pg_sz=" SIZE_FORMAT " base=" PTR_FORMAT " size=" SIZE_FORMAT " alignment=" SIZE_FORMAT " reqsize=" SIZE_FORMAT,
+                           description, preferred_page_size, p2i(rs.base()), rs.size(), rs.alignment(), size);
+  }
+  return result;
+}
+
 jint G1CollectedHeap::initialize() {
   CollectedHeap::pre_initialize();
   os::enable_vtime();
@@ -1992,57 +1997,35 @@
   ReservedSpace g1_rs = heap_rs.first_part(max_byte_size);
   G1RegionToSpaceMapper* heap_storage =
     G1RegionToSpaceMapper::create_mapper(g1_rs,
+                                         g1_rs.size(),
                                          UseLargePages ? os::large_page_size() : os::vm_page_size(),
                                          HeapRegion::GrainBytes,
                                          1,
                                          mtJavaHeap);
   heap_storage->set_mapping_changed_listener(&_listener);
 
-  // Reserve space for the block offset table. We do not support automatic uncommit
-  // for the card table at this time. BOT only.
-  ReservedSpace bot_rs(G1BlockOffsetSharedArray::compute_size(g1_rs.size() / HeapWordSize));
+  // Create storage for the BOT, card table, card counts table (hot card cache) and the bitmaps.
   G1RegionToSpaceMapper* bot_storage =
-    G1RegionToSpaceMapper::create_mapper(bot_rs,
-                                         os::vm_page_size(),
-                                         HeapRegion::GrainBytes,
-                                         G1BlockOffsetSharedArray::N_bytes,
-                                         mtGC);
+    create_aux_memory_mapper("Block offset table",
+                             G1BlockOffsetSharedArray::compute_size(g1_rs.size() / HeapWordSize),
+                             G1BlockOffsetSharedArray::N_bytes);
 
   ReservedSpace cardtable_rs(G1SATBCardTableLoggingModRefBS::compute_size(g1_rs.size() / HeapWordSize));
   G1RegionToSpaceMapper* cardtable_storage =
-    G1RegionToSpaceMapper::create_mapper(cardtable_rs,
-                                         os::vm_page_size(),
-                                         HeapRegion::GrainBytes,
-                                         G1BlockOffsetSharedArray::N_bytes,
-                                         mtGC);
-
-  // Reserve space for the card counts table.
-  ReservedSpace card_counts_rs(G1BlockOffsetSharedArray::compute_size(g1_rs.size() / HeapWordSize));
+    create_aux_memory_mapper("Card table",
+                             G1SATBCardTableLoggingModRefBS::compute_size(g1_rs.size() / HeapWordSize),
+                             G1BlockOffsetSharedArray::N_bytes);
+
   G1RegionToSpaceMapper* card_counts_storage =
-    G1RegionToSpaceMapper::create_mapper(card_counts_rs,
-                                         os::vm_page_size(),
-                                         HeapRegion::GrainBytes,
-                                         G1BlockOffsetSharedArray::N_bytes,
-                                         mtGC);
-
-  // Reserve space for prev and next bitmap.
+    create_aux_memory_mapper("Card counts table",
+                             G1BlockOffsetSharedArray::compute_size(g1_rs.size() / HeapWordSize),
+                             G1BlockOffsetSharedArray::N_bytes);
+
   size_t bitmap_size = CMBitMap::compute_size(g1_rs.size());
-
-  ReservedSpace prev_bitmap_rs(ReservedSpace::allocation_align_size_up(bitmap_size));
   G1RegionToSpaceMapper* prev_bitmap_storage =
-    G1RegionToSpaceMapper::create_mapper(prev_bitmap_rs,
-                                         os::vm_page_size(),
-                                         HeapRegion::GrainBytes,
-                                         CMBitMap::mark_distance(),
-                                         mtGC);
-
-  ReservedSpace next_bitmap_rs(ReservedSpace::allocation_align_size_up(bitmap_size));
+    create_aux_memory_mapper("Prev Bitmap", bitmap_size, CMBitMap::mark_distance());
   G1RegionToSpaceMapper* next_bitmap_storage =
-    G1RegionToSpaceMapper::create_mapper(next_bitmap_rs,
-                                         os::vm_page_size(),
-                                         HeapRegion::GrainBytes,
-                                         CMBitMap::mark_distance(),
-                                         mtGC);
+    create_aux_memory_mapper("Next Bitmap", bitmap_size, CMBitMap::mark_distance());
 
   _hrm.initialize(heap_storage, prev_bitmap_storage, next_bitmap_storage, bot_storage, cardtable_storage, card_counts_storage);
   g1_barrier_set()->initialize(cardtable_storage);
@@ -2065,8 +2048,14 @@
 
   _g1h = this;
 
-  _in_cset_fast_test.initialize(_hrm.reserved().start(), _hrm.reserved().end(), HeapRegion::GrainBytes);
-  _humongous_is_live.initialize(_hrm.reserved().start(), _hrm.reserved().end(), HeapRegion::GrainBytes);
+  {
+    HeapWord* start = _hrm.reserved().start();
+    HeapWord* end = _hrm.reserved().end();
+    size_t granularity = HeapRegion::GrainBytes;
+
+    _in_cset_fast_test.initialize(start, end, granularity);
+    _humongous_reclaim_candidates.initialize(start, end, granularity);
+  }
 
   // Create the ConcurrentMark data structure and thread.
   // (Must do this late, so that "max_regions" is defined.)
@@ -2158,11 +2147,6 @@
   }
 }
 
-void G1CollectedHeap::clear_humongous_is_live_table() {
-  guarantee(G1ReclaimDeadHumongousObjectsAtYoungGC, "Should only be called if true");
-  _humongous_is_live.clear();
-}
-
 size_t G1CollectedHeap::conservative_max_heap_alignment() {
   return HeapRegion::max_region_size();
 }
@@ -2299,11 +2283,11 @@
   hot_card_cache->drain(worker_i, g1_rem_set(), into_cset_dcq);
 
   DirtyCardQueueSet& dcqs = JavaThread::dirty_card_queue_set();
-  int n_completed_buffers = 0;
+  size_t n_completed_buffers = 0;
   while (dcqs.apply_closure_to_completed_buffer(cl, worker_i, 0, true)) {
     n_completed_buffers++;
   }
-  g1_policy()->phase_times()->record_update_rs_processed_buffers(worker_i, n_completed_buffers);
+  g1_policy()->phase_times()->record_thread_work_item(G1GCPhaseTimes::UpdateRS, worker_i, n_completed_buffers);
   dcqs.clear_n_completed_buffers();
   assert(!dcqs.completed_buffers_exist_dirty(), "Completed buffers exist!");
 }
@@ -2448,13 +2432,24 @@
     _gc_timer_cm->register_gc_end();
     _gc_tracer_cm->report_gc_end(_gc_timer_cm->gc_end(), _gc_timer_cm->time_partitions());
 
+    // Clear state variables to prepare for the next concurrent cycle.
     _concurrent_cycle_started = false;
+    _heap_summary_sent = false;
   }
 }
 
 void G1CollectedHeap::trace_heap_after_concurrent_cycle() {
   if (_concurrent_cycle_started) {
-    trace_heap_after_gc(_gc_tracer_cm);
+    // This function can be called when:
+    //  the cleanup pause is run
+    //  the concurrent cycle is aborted before the cleanup pause.
+    //  the concurrent cycle is aborted after the cleanup pause,
+    //   but before the concurrent cycle end has been registered.
+    // Make sure that we only send the heap information once.
+    if (!_heap_summary_sent) {
+      trace_heap_after_gc(_gc_tracer_cm);
+      _heap_summary_sent = true;
+    }
   }
 }
 
@@ -2477,9 +2472,9 @@
 void G1CollectedHeap::collect(GCCause::Cause cause) {
   assert_heap_not_locked();
 
-  unsigned int gc_count_before;
-  unsigned int old_marking_count_before;
-  unsigned int full_gc_count_before;
+  uint gc_count_before;
+  uint old_marking_count_before;
+  uint full_gc_count_before;
   bool retry_gc;
 
   do {
@@ -3292,11 +3287,12 @@
     G1VerifyCodeRootOopClosure codeRootsCl(this, &rootsCl, vo);
     G1VerifyCodeRootBlobClosure blobsCl(&codeRootsCl);
 
-    process_all_roots(true,            // activate StrongRootsScope
-                      SO_AllCodeCache, // roots scanning options
-                      &rootsCl,
-                      &cldCl,
-                      &blobsCl);
+    {
+      G1RootProcessor root_processor(this);
+      root_processor.process_all_roots(&rootsCl,
+                                       &cldCl,
+                                       &blobsCl);
+    }
 
     bool failures = rootsCl.failures() || codeRootsCl.failures();
 
@@ -3616,7 +3612,7 @@
 }
 
 HeapWord* G1CollectedHeap::do_collection_pause(size_t word_size,
-                                               unsigned int gc_count_before,
+                                               uint gc_count_before,
                                                bool* succeeded,
                                                GCCause::Cause gc_cause) {
   assert_heap_not_locked_and_not_at_safepoint();
@@ -3671,18 +3667,73 @@
   return g1_rem_set()->cardsScanned();
 }
 
-bool G1CollectedHeap::humongous_region_is_always_live(uint index) {
-  HeapRegion* region = region_at(index);
-  assert(region->startsHumongous(), "Must start a humongous object");
-  return oop(region->bottom())->is_objArray() || !region->rem_set()->is_empty();
-}
-
 class RegisterHumongousWithInCSetFastTestClosure : public HeapRegionClosure {
  private:
   size_t _total_humongous;
   size_t _candidate_humongous;
+
+  DirtyCardQueue _dcq;
+
+  // We don't nominate objects with many remembered set entries, on
+  // the assumption that such objects are likely still live.
+  bool is_remset_small(HeapRegion* region) const {
+    HeapRegionRemSet* const rset = region->rem_set();
+    return G1EagerReclaimHumongousObjectsWithStaleRefs
+      ? rset->occupancy_less_or_equal_than(G1RSetSparseRegionEntries)
+      : rset->is_empty();
+  }
+
+  bool is_typeArray_region(HeapRegion* region) const {
+    return oop(region->bottom())->is_typeArray();
+  }
+
+  bool humongous_region_is_candidate(G1CollectedHeap* heap, HeapRegion* region) const {
+    assert(region->startsHumongous(), "Must start a humongous object");
+
+    // Candidate selection must satisfy the following constraints
+    // while concurrent marking is in progress:
+    //
+    // * In order to maintain SATB invariants, an object must not be
+    // reclaimed if it was allocated before the start of marking and
+    // has not had its references scanned.  Such an object must have
+    // its references (including type metadata) scanned to ensure no
+    // live objects are missed by the marking process.  Objects
+    // allocated after the start of concurrent marking don't need to
+    // be scanned.
+    //
+    // * An object must not be reclaimed if it is on the concurrent
+    // mark stack.  Objects allocated after the start of concurrent
+    // marking are never pushed on the mark stack.
+    //
+    // Nominating only objects allocated after the start of concurrent
+    // marking is sufficient to meet both constraints.  This may miss
+    // some objects that satisfy the constraints, but the marking data
+    // structures don't support efficiently performing the needed
+    // additional tests or scrubbing of the mark stack.
+    //
+    // However, we presently only nominate is_typeArray() objects.
+    // A humongous object containing references induces remembered
+    // set entries on other regions.  In order to reclaim such an
+    // object, those remembered sets would need to be cleaned up.
+    //
+    // We also treat is_typeArray() objects specially, allowing them
+    // to be reclaimed even if allocated before the start of
+    // concurrent mark.  For this we rely on mark stack insertion to
+    // exclude is_typeArray() objects, preventing reclaiming an object
+    // that is in the mark stack.  We also rely on the metadata for
+    // such objects to be built-in and so ensured to be kept live.
+    // Frequent allocation and drop of large binary blobs is an
+    // important use case for eager reclaim, and this special handling
+    // may reduce needed headroom.
+
+    return is_typeArray_region(region) && is_remset_small(region);
+  }
+
  public:
-  RegisterHumongousWithInCSetFastTestClosure() : _total_humongous(0), _candidate_humongous(0) {
+  RegisterHumongousWithInCSetFastTestClosure()
+  : _total_humongous(0),
+    _candidate_humongous(0),
+    _dcq(&JavaThread::dirty_card_queue_set()) {
   }
 
   virtual bool doHeapRegion(HeapRegion* r) {
@@ -3691,14 +3742,33 @@
     }
     G1CollectedHeap* g1h = G1CollectedHeap::heap();
 
-    uint region_idx = r->hrm_index();
-    bool is_candidate = !g1h->humongous_region_is_always_live(region_idx);
-    // Is_candidate already filters out humongous regions with some remembered set.
-    // This will not lead to humongous object that we mistakenly keep alive because
-    // during young collection the remembered sets will only be added to.
+    bool is_candidate = humongous_region_is_candidate(g1h, r);
+    uint rindex = r->hrm_index();
+    g1h->set_humongous_reclaim_candidate(rindex, is_candidate);
     if (is_candidate) {
-      g1h->register_humongous_region_with_in_cset_fast_test(region_idx);
       _candidate_humongous++;
+      g1h->register_humongous_region_with_in_cset_fast_test(rindex);
+      // Is_candidate already filters out humongous object with large remembered sets.
+      // If we have a humongous object with a few remembered sets, we simply flush these
+      // remembered set entries into the DCQS. That will result in automatic
+      // re-evaluation of their remembered set entries during the following evacuation
+      // phase.
+      if (!r->rem_set()->is_empty()) {
+        guarantee(r->rem_set()->occupancy_less_or_equal_than(G1RSetSparseRegionEntries),
+                  "Found a not-small remembered set here. This is inconsistent with previous assumptions.");
+        G1SATBCardTableLoggingModRefBS* bs = g1h->g1_barrier_set();
+        HeapRegionRemSetIterator hrrs(r->rem_set());
+        size_t card_index;
+        while (hrrs.has_next(card_index)) {
+          jbyte* card_ptr = (jbyte*)bs->byte_for_index(card_index);
+          if (*card_ptr != CardTableModRefBS::dirty_card_val()) {
+            *card_ptr = CardTableModRefBS::dirty_card_val();
+            _dcq.enqueue(card_ptr);
+          }
+        }
+        r->rem_set()->clear_locked();
+      }
+      assert(r->rem_set()->is_empty(), "At this point any humongous candidate remembered set must be empty.");
     }
     _total_humongous++;
 
@@ -3707,23 +3777,29 @@
 
   size_t total_humongous() const { return _total_humongous; }
   size_t candidate_humongous() const { return _candidate_humongous; }
+
+  void flush_rem_set_entries() { _dcq.flush(); }
 };
 
 void G1CollectedHeap::register_humongous_regions_with_in_cset_fast_test() {
-  if (!G1ReclaimDeadHumongousObjectsAtYoungGC) {
-    g1_policy()->phase_times()->record_fast_reclaim_humongous_stats(0, 0);
+  if (!G1EagerReclaimHumongousObjects) {
+    g1_policy()->phase_times()->record_fast_reclaim_humongous_stats(0.0, 0, 0);
     return;
   }
-
+  double time = os::elapsed_counter();
+
+  // Collect reclaim candidate information and register candidates with cset.
   RegisterHumongousWithInCSetFastTestClosure cl;
   heap_region_iterate(&cl);
-  g1_policy()->phase_times()->record_fast_reclaim_humongous_stats(cl.total_humongous(),
+
+  time = ((double)(os::elapsed_counter() - time) / os::elapsed_frequency()) * 1000.0;
+  g1_policy()->phase_times()->record_fast_reclaim_humongous_stats(time,
+                                                                  cl.total_humongous(),
                                                                   cl.candidate_humongous());
   _has_humongous_reclaim_candidates = cl.candidate_humongous() > 0;
 
-  if (_has_humongous_reclaim_candidates) {
-    clear_humongous_is_live_table();
-  }
+  // Finally flush all remembered set entries to re-check into the global DCQS.
+  cl.flush_rem_set_entries();
 }
 
 void
@@ -3901,10 +3977,10 @@
 
     TraceCPUTime tcpu(G1Log::finer(), true, gclog_or_tty);
 
-    int active_workers = (G1CollectedHeap::use_parallel_gc_threads() ?
+    uint active_workers = (G1CollectedHeap::use_parallel_gc_threads() ?
                                 workers()->active_workers() : 1);
     double pause_start_sec = os::elapsedTime();
-    g1_policy()->phase_times()->note_gc_start(active_workers);
+    g1_policy()->phase_times()->note_gc_start(active_workers, mark_in_progress());
     log_gc_header();
 
     TraceCollectorStats tcs(g1mm()->incremental_collection_counters());
@@ -4014,15 +4090,12 @@
 
         register_humongous_regions_with_in_cset_fast_test();
 
+        assert(check_cset_fast_test(), "Inconsistency in the InCSetState table.");
+
         _cm->note_start_of_gc();
-        // We should not verify the per-thread SATB buffers given that
-        // we have not filtered them yet (we'll do so during the
-        // GC). We also call this after finalize_cset() to
+        // We call this after finalize_cset() to
         // ensure that the CSet has been finalized.
-        _cm->verify_no_cset_oops(true  /* verify_stacks */,
-                                 true  /* verify_enqueued_buffers */,
-                                 false /* verify_thread_buffers */,
-                                 true  /* verify_fingers */);
+        _cm->verify_no_cset_oops();
 
         if (_hr_printer.is_active()) {
           HeapRegion* hr = g1_policy()->collection_set();
@@ -4045,16 +4118,6 @@
         // Actually do the work...
         evacuate_collection_set(evacuation_info);
 
-        // We do this to mainly verify the per-thread SATB buffers
-        // (which have been filtered by now) since we didn't verify
-        // them earlier. No point in re-checking the stacks / enqueued
-        // buffers given that the CSet has not changed since last time
-        // we checked.
-        _cm->verify_no_cset_oops(false /* verify_stacks */,
-                                 false /* verify_enqueued_buffers */,
-                                 true  /* verify_thread_buffers */,
-                                 true  /* verify_fingers */);
-
         free_collection_set(g1_policy()->collection_set(), evacuation_info);
 
         eagerly_reclaim_humongous_regions();
@@ -4137,10 +4200,7 @@
 
         // We redo the verification but now wrt to the new CSet which
         // has just got initialized after the previous CSet was freed.
-        _cm->verify_no_cset_oops(true  /* verify_stacks */,
-                                 true  /* verify_enqueued_buffers */,
-                                 true  /* verify_thread_buffers */,
-                                 true  /* verify_fingers */);
+        _cm->verify_no_cset_oops();
         _cm->note_end_of_gc();
 
         // This timing is only used by the ergonomics to handle our pause target.
@@ -4243,29 +4303,6 @@
   return true;
 }
 
-size_t G1CollectedHeap::desired_plab_sz(GCAllocPurpose purpose)
-{
-  size_t gclab_word_size;
-  switch (purpose) {
-    case GCAllocForSurvived:
-      gclab_word_size = _survivor_plab_stats.desired_plab_sz();
-      break;
-    case GCAllocForTenured:
-      gclab_word_size = _old_plab_stats.desired_plab_sz();
-      break;
-    default:
-      assert(false, "unknown GCAllocPurpose");
-      gclab_word_size = _old_plab_stats.desired_plab_sz();
-      break;
-  }
-
-  // Prevent humongous PLAB sizes for two reasons:
-  // * PLABs are allocated using a similar paths as oops, but should
-  //   never be in a humongous region
-  // * Allowing humongous PLABs needlessly churns the region free lists
-  return MIN2(_humongous_object_threshold_in_words, gclab_word_size);
-}
-
 void G1CollectedHeap::init_for_evac_failure(OopsInHeapRegionClosure* cl) {
   _drain_in_progress = false;
   set_evac_failure_closure(cl);
@@ -4405,35 +4442,6 @@
   }
 }
 
-HeapWord* G1CollectedHeap::par_allocate_during_gc(GCAllocPurpose purpose,
-                                                  size_t word_size,
-                                                  AllocationContext_t context) {
-  if (purpose == GCAllocForSurvived) {
-    HeapWord* result = survivor_attempt_allocation(word_size, context);
-    if (result != NULL) {
-      return result;
-    } else {
-      // Let's try to allocate in the old gen in case we can fit the
-      // object there.
-      return old_attempt_allocation(word_size, context);
-    }
-  } else {
-    assert(purpose ==  GCAllocForTenured, "sanity");
-    HeapWord* result = old_attempt_allocation(word_size, context);
-    if (result != NULL) {
-      return result;
-    } else {
-      // Let's try to allocate in the survivors in case we can fit the
-      // object there.
-      return survivor_attempt_allocation(word_size, context);
-    }
-  }
-
-  ShouldNotReachHere();
-  // Trying to keep some compilers happy.
-  return NULL;
-}
-
 void G1ParCopyHelper::mark_object(oop obj) {
   assert(!_g1->heap_region_containing(obj)->in_collection_set(), "should not mark objects in the CSet");
 
@@ -4476,14 +4484,14 @@
 
   assert(_worker_id == _par_scan_state->queue_num(), "sanity");
 
-  G1CollectedHeap::in_cset_state_t state = _g1->in_cset_state(obj);
-
-  if (state == G1CollectedHeap::InCSet) {
+  const InCSetState state = _g1->in_cset_state(obj);
+  if (state.is_in_cset()) {
     oop forwardee;
-    if (obj->is_forwarded()) {
-      forwardee = obj->forwardee();
+    markOop m = obj->mark();
+    if (m->is_marked()) {
+      forwardee = (oop) m->decode_pointer();
     } else {
-      forwardee = _par_scan_state->copy_to_survivor_space(obj);
+      forwardee = _par_scan_state->copy_to_survivor_space(state, obj, m);
     }
     assert(forwardee != NULL, "forwardee should not be NULL");
     oopDesc::encode_store_heap_oop(p, forwardee);
@@ -4497,7 +4505,7 @@
       do_klass_barrier(p, forwardee);
     }
   } else {
-    if (state == G1CollectedHeap::IsHumongous) {
+    if (state.is_humongous()) {
       _g1->set_humongous_is_live(obj);
     }
     // The object is not in collection set. If we're a root scanning
@@ -4582,60 +4590,11 @@
   }
 };
 
-class G1CodeBlobClosure : public CodeBlobClosure {
-  class HeapRegionGatheringOopClosure : public OopClosure {
-    G1CollectedHeap* _g1h;
-    OopClosure* _work;
-    nmethod* _nm;
-
-    template <typename T>
-    void do_oop_work(T* p) {
-      _work->do_oop(p);
-      T oop_or_narrowoop = oopDesc::load_heap_oop(p);
-      if (!oopDesc::is_null(oop_or_narrowoop)) {
-        oop o = oopDesc::decode_heap_oop_not_null(oop_or_narrowoop);
-        HeapRegion* hr = _g1h->heap_region_containing_raw(o);
-        assert(!_g1h->obj_in_cs(o) || hr->rem_set()->strong_code_roots_list_contains(_nm), "if o still in CS then evacuation failed and nm must already be in the remset");
-        hr->add_strong_code_root(_nm);
-      }
-    }
-
-  public:
-    HeapRegionGatheringOopClosure(OopClosure* oc) : _g1h(G1CollectedHeap::heap()), _work(oc), _nm(NULL) {}
-
-    void do_oop(oop* o) {
-      do_oop_work(o);
-    }
-
-    void do_oop(narrowOop* o) {
-      do_oop_work(o);
-    }
-
-    void set_nm(nmethod* nm) {
-      _nm = nm;
-    }
-  };
-
-  HeapRegionGatheringOopClosure _oc;
-public:
-  G1CodeBlobClosure(OopClosure* oc) : _oc(oc) {}
-
-  void do_code_blob(CodeBlob* cb) {
-    nmethod* nm = cb->as_nmethod_or_null();
-    if (nm != NULL) {
-      if (!nm->test_set_oops_do_mark()) {
-        _oc.set_nm(nm);
-        nm->oops_do(&_oc);
-        nm->fix_oop_relocations();
-      }
-    }
-  }
-};
-
 class G1ParTask : public AbstractGangTask {
 protected:
   G1CollectedHeap*       _g1h;
   RefToScanQueueSet      *_queues;
+  G1RootProcessor*       _root_processor;
   ParallelTaskTerminator _terminator;
   uint _n_workers;
 
@@ -4643,10 +4602,11 @@
   Mutex* stats_lock() { return &_stats_lock; }
 
 public:
-  G1ParTask(G1CollectedHeap* g1h, RefToScanQueueSet *task_queues)
+  G1ParTask(G1CollectedHeap* g1h, RefToScanQueueSet *task_queues, G1RootProcessor* root_processor)
     : AbstractGangTask("G1 collection"),
       _g1h(g1h),
       _queues(task_queues),
+      _root_processor(root_processor),
       _terminator(0, _queues),
       _stats_lock(Mutex::leaf, "parallel G1 stats lock", true)
   {}
@@ -4660,13 +4620,7 @@
   ParallelTaskTerminator* terminator() { return &_terminator; }
 
   virtual void set_for_termination(int active_workers) {
-    // This task calls set_n_termination() in par_non_clean_card_iterate_work()
-    // in the young space (_par_seq_tasks) in the G1 heap
-    // for SequentialSubTasksDone.
-    // This task also uses SubTasksDone in SharedHeap and G1CollectedHeap
-    // both of which need setting by set_n_termination().
-    _g1h->SharedHeap::set_n_termination(active_workers);
-    _g1h->set_n_termination(active_workers);
+    _root_processor->set_num_workers(active_workers);
     terminator()->reset_for_reuse(active_workers);
     _n_workers = active_workers;
   }
@@ -4703,8 +4657,7 @@
   void work(uint worker_id) {
     if (worker_id >= _n_workers) return;  // no work needed this round
 
-    double start_time_ms = os::elapsedTime() * 1000.0;
-    _g1h->g1_policy()->phase_times()->record_gc_worker_start_time(worker_id, start_time_ms);
+    _g1h->g1_policy()->phase_times()->record_time_secs(G1GCPhaseTimes::GCWorkerStart, worker_id, os::elapsedTime());
 
     {
       ResourceMark rm;
@@ -4736,24 +4689,21 @@
                                                                                     false, // Process all klasses.
                                                                                     true); // Need to claim CLDs.
 
-      G1CodeBlobClosure scan_only_code_cl(&scan_only_root_cl);
-      G1CodeBlobClosure scan_mark_code_cl(&scan_mark_root_cl);
-      // IM Weak code roots are handled later.
-
       OopClosure* strong_root_cl;
       OopClosure* weak_root_cl;
       CLDClosure* strong_cld_cl;
       CLDClosure* weak_cld_cl;
-      CodeBlobClosure* strong_code_cl;
+
+      bool trace_metadata = false;
 
       if (_g1h->g1_policy()->during_initial_mark_pause()) {
         // We also need to mark copied objects.
         strong_root_cl = &scan_mark_root_cl;
         strong_cld_cl  = &scan_mark_cld_cl;
-        strong_code_cl = &scan_mark_code_cl;
         if (ClassUnloadingWithConcurrentMark) {
           weak_root_cl = &scan_mark_weak_root_cl;
           weak_cld_cl  = &scan_mark_weak_cld_cl;
+          trace_metadata = true;
         } else {
           weak_root_cl = &scan_mark_root_cl;
           weak_cld_cl  = &scan_mark_cld_cl;
@@ -4763,31 +4713,32 @@
         weak_root_cl   = &scan_only_root_cl;
         strong_cld_cl  = &scan_only_cld_cl;
         weak_cld_cl    = &scan_only_cld_cl;
-        strong_code_cl = &scan_only_code_cl;
       }
 
-
-      G1ParPushHeapRSClosure  push_heap_rs_cl(_g1h, &pss);
-
       pss.start_strong_roots();
-      _g1h->g1_process_roots(strong_root_cl,
-                             weak_root_cl,
-                             &push_heap_rs_cl,
-                             strong_cld_cl,
-                             weak_cld_cl,
-                             strong_code_cl,
-                             worker_id);
-
+
+      _root_processor->evacuate_roots(strong_root_cl,
+                                      weak_root_cl,
+                                      strong_cld_cl,
+                                      weak_cld_cl,
+                                      trace_metadata,
+                                      worker_id);
+
+      G1ParPushHeapRSClosure push_heap_rs_cl(_g1h, &pss);
+      _root_processor->scan_remembered_sets(&push_heap_rs_cl,
+                                            weak_root_cl,
+                                            worker_id);
       pss.end_strong_roots();
 
       {
         double start = os::elapsedTime();
         G1ParEvacuateFollowersClosure evac(_g1h, &pss, _queues, &_terminator);
         evac.do_void();
-        double elapsed_ms = (os::elapsedTime()-start)*1000.0;
-        double term_ms = pss.term_time()*1000.0;
-        _g1h->g1_policy()->phase_times()->add_obj_copy_time(worker_id, elapsed_ms-term_ms);
-        _g1h->g1_policy()->phase_times()->record_termination(worker_id, term_ms, pss.term_attempts());
+        double elapsed_sec = os::elapsedTime() - start;
+        double term_sec = pss.term_time();
+        _g1h->g1_policy()->phase_times()->add_time_secs(G1GCPhaseTimes::ObjCopy, worker_id, elapsed_sec - term_sec);
+        _g1h->g1_policy()->phase_times()->record_time_secs(G1GCPhaseTimes::Termination, worker_id, term_sec);
+        _g1h->g1_policy()->phase_times()->record_thread_work_item(G1GCPhaseTimes::Termination, worker_id, pss.term_attempts());
       }
       _g1h->g1_policy()->record_thread_age_table(pss.age_table());
       _g1h->update_surviving_young_words(pss.surviving_young_words()+1);
@@ -4803,100 +4754,10 @@
       // destructors are executed here and are included as part of the
       // "GC Worker Time".
     }
-
-    double end_time_ms = os::elapsedTime() * 1000.0;
-    _g1h->g1_policy()->phase_times()->record_gc_worker_end_time(worker_id, end_time_ms);
+    _g1h->g1_policy()->phase_times()->record_time_secs(G1GCPhaseTimes::GCWorkerEnd, worker_id, os::elapsedTime());
   }
 };
 
-// *** Common G1 Evacuation Stuff
-
-// This method is run in a GC worker.
-
-void
-G1CollectedHeap::
-g1_process_roots(OopClosure* scan_non_heap_roots,
-                 OopClosure* scan_non_heap_weak_roots,
-                 OopsInHeapRegionClosure* scan_rs,
-                 CLDClosure* scan_strong_clds,
-                 CLDClosure* scan_weak_clds,
-                 CodeBlobClosure* scan_strong_code,
-                 uint worker_i) {
-
-  // First scan the shared roots.
-  double ext_roots_start = os::elapsedTime();
-  double closure_app_time_sec = 0.0;
-
-  bool during_im = _g1h->g1_policy()->during_initial_mark_pause();
-  bool trace_metadata = during_im && ClassUnloadingWithConcurrentMark;
-
-  BufferingOopClosure buf_scan_non_heap_roots(scan_non_heap_roots);
-  BufferingOopClosure buf_scan_non_heap_weak_roots(scan_non_heap_weak_roots);
-
-  process_roots(false, // no scoping; this is parallel code
-                SharedHeap::SO_None,
-                &buf_scan_non_heap_roots,
-                &buf_scan_non_heap_weak_roots,
-                scan_strong_clds,
-                // Unloading Initial Marks handle the weak CLDs separately.
-                (trace_metadata ? NULL : scan_weak_clds),
-                scan_strong_code);
-
-  // Now the CM ref_processor roots.
-  if (!_process_strong_tasks->is_task_claimed(G1H_PS_refProcessor_oops_do)) {
-    // We need to treat the discovered reference lists of the
-    // concurrent mark ref processor as roots and keep entries
-    // (which are added by the marking threads) on them live
-    // until they can be processed at the end of marking.
-    ref_processor_cm()->weak_oops_do(&buf_scan_non_heap_roots);
-  }
-
-  if (trace_metadata) {
-    // Barrier to make sure all workers passed
-    // the strong CLD and strong nmethods phases.
-    active_strong_roots_scope()->wait_until_all_workers_done_with_threads(n_par_threads());
-
-    // Now take the complement of the strong CLDs.
-    ClassLoaderDataGraph::roots_cld_do(NULL, scan_weak_clds);
-  }
-
-  // Finish up any enqueued closure apps (attributed as object copy time).
-  buf_scan_non_heap_roots.done();
-  buf_scan_non_heap_weak_roots.done();
-
-  double obj_copy_time_sec = buf_scan_non_heap_roots.closure_app_seconds()
-      + buf_scan_non_heap_weak_roots.closure_app_seconds();
-
-  g1_policy()->phase_times()->record_obj_copy_time(worker_i, obj_copy_time_sec * 1000.0);
-
-  double ext_root_time_ms =
-    ((os::elapsedTime() - ext_roots_start) - obj_copy_time_sec) * 1000.0;
-
-  g1_policy()->phase_times()->record_ext_root_scan_time(worker_i, ext_root_time_ms);
-
-  // During conc marking we have to filter the per-thread SATB buffers
-  // to make sure we remove any oops into the CSet (which will show up
-  // as implicitly live).
-  double satb_filtering_ms = 0.0;
-  if (!_process_strong_tasks->is_task_claimed(G1H_PS_filter_satb_buffers)) {
-    if (mark_in_progress()) {
-      double satb_filter_start = os::elapsedTime();
-
-      JavaThread::satb_mark_queue_set().filter_thread_buffers();
-
-      satb_filtering_ms = (os::elapsedTime() - satb_filter_start) * 1000.0;
-    }
-  }
-  g1_policy()->phase_times()->record_satb_filtering_time(worker_i, satb_filtering_ms);
-
-  // Now scan the complement of the collection set.
-  G1CodeBlobClosure scavenge_cs_nmethods(scan_non_heap_weak_roots);
-
-  g1_rem_set()->oops_into_collection_set_do(scan_rs, &scavenge_cs_nmethods, worker_i);
-
-  _process_strong_tasks->all_tasks_completed();
-}
-
 class G1StringSymbolTableUnlinkTask : public AbstractGangTask {
 private:
   BoolObjectClosure* _is_alive;
@@ -5310,7 +5171,8 @@
   G1RedirtyLoggedCardsTask(DirtyCardQueueSet* queue) : AbstractGangTask("Redirty Cards"), _queue(queue) { }
 
   virtual void work(uint worker_id) {
-    double start_time = os::elapsedTime();
+    G1GCPhaseTimes* phase_times = G1CollectedHeap::heap()->g1_policy()->phase_times();
+    G1GCParPhaseTimesTracker x(phase_times, G1GCPhaseTimes::RedirtyCards, worker_id);
 
     RedirtyLoggedCardTableEntryClosure cl;
     if (G1CollectedHeap::heap()->use_parallel_gc_threads()) {
@@ -5319,9 +5181,7 @@
       _queue->apply_closure_to_all_completed_buffers(&cl);
     }
 
-    G1GCPhaseTimes* timer = G1CollectedHeap::heap()->g1_policy()->phase_times();
-    timer->record_redirty_logged_cards_time_ms(worker_id, (os::elapsedTime() - start_time) * 1000.0);
-    timer->record_redirty_logged_cards_processed_cards(worker_id, cl.num_processed());
+    phase_times->record_thread_work_item(G1GCPhaseTimes::RedirtyCards, worker_id, cl.num_processed());
   }
 };
 
@@ -5382,17 +5242,17 @@
     oop obj = *p;
     assert(obj != NULL, "the caller should have filtered out NULL values");
 
-    G1CollectedHeap::in_cset_state_t cset_state = _g1->in_cset_state(obj);
-    if (cset_state == G1CollectedHeap::InNeither) {
+    const InCSetState cset_state = _g1->in_cset_state(obj);
+    if (!cset_state.is_in_cset_or_humongous()) {
       return;
     }
-    if (cset_state == G1CollectedHeap::InCSet) {
+    if (cset_state.is_in_cset()) {
       assert( obj->is_forwarded(), "invariant" );
       *p = obj->forwardee();
     } else {
       assert(!obj->is_forwarded(), "invariant" );
-      assert(cset_state == G1CollectedHeap::IsHumongous,
-             err_msg("Only allowed InCSet state is IsHumongous, but is %d", cset_state));
+      assert(cset_state.is_humongous(),
+             err_msg("Only allowed InCSet state is IsHumongous, but is %d", cset_state.value()));
       _g1->set_humongous_is_live(obj);
     }
   }
@@ -5885,7 +5745,6 @@
     n_workers = 1;
   }
 
-  G1ParTask g1_par_task(this, _task_queues);
 
   init_for_evac_failure(NULL);
 
@@ -5896,7 +5755,8 @@
   double end_par_time_sec;
 
   {
-    StrongRootsScope srs(this);
+    G1RootProcessor root_processor(this);
+    G1ParTask g1_par_task(this, _task_queues, &root_processor);
     // InitialMark needs claim bits to keep track of the marked-through CLDs.
     if (g1_policy()->during_initial_mark_pause()) {
       ClassLoaderDataGraph::clear_claimed_marks();
@@ -5917,18 +5777,20 @@
     end_par_time_sec = os::elapsedTime();
 
     // Closing the inner scope will execute the destructor
-    // for the StrongRootsScope object. We record the current
+    // for the G1RootProcessor object. We record the current
     // elapsed time before closing the scope so that time
-    // taken for the SRS destructor is NOT included in the
+    // taken for the destructor is NOT included in the
     // reported parallel time.
   }
 
+  G1GCPhaseTimes* phase_times = g1_policy()->phase_times();
+
   double par_time_ms = (end_par_time_sec - start_par_time_sec) * 1000.0;
-  g1_policy()->phase_times()->record_par_time(par_time_ms);
+  phase_times->record_par_time(par_time_ms);
 
   double code_root_fixup_time_ms =
         (os::elapsedTime() - end_par_time_sec) * 1000.0;
-  g1_policy()->phase_times()->record_code_root_fixup_time(code_root_fixup_time_ms);
+  phase_times->record_code_root_fixup_time(code_root_fixup_time_ms);
 
   set_par_threads(0);
 
@@ -5939,14 +5801,15 @@
   // not copied during the pause.
   process_discovered_references(n_workers);
 
-  // Weak root processing.
-  {
+  if (G1StringDedup::is_enabled()) {
+    double fixup_start = os::elapsedTime();
+
     G1STWIsAliveClosure is_alive(this);
     G1KeepAliveClosure keep_alive(this);
-    JNIHandles::weak_oops_do(&is_alive, &keep_alive);
-    if (G1StringDedup::is_enabled()) {
-      G1StringDedup::unlink_or_oops_do(&is_alive, &keep_alive);
-    }
+    G1StringDedup::unlink_or_oops_do(&is_alive, &keep_alive, true, phase_times);
+
+    double fixup_time_ms = (os::elapsedTime() - fixup_start) * 1000.0;
+    phase_times->record_string_dedup_fixup_time(fixup_time_ms);
   }
 
   _allocator->release_gc_alloc_regions(n_workers, evacuation_info);
@@ -6214,6 +6077,70 @@
   heap_region_iterate(&cl);
   guarantee(!cl.failures(), "bitmap verification");
 }
+
+class G1CheckCSetFastTableClosure : public HeapRegionClosure {
+ private:
+  bool _failures;
+ public:
+  G1CheckCSetFastTableClosure() : HeapRegionClosure(), _failures(false) { }
+
+  virtual bool doHeapRegion(HeapRegion* hr) {
+    uint i = hr->hrm_index();
+    InCSetState cset_state = (InCSetState) G1CollectedHeap::heap()->_in_cset_fast_test.get_by_index(i);
+    if (hr->isHumongous()) {
+      if (hr->in_collection_set()) {
+        gclog_or_tty->print_cr("\n## humongous region %u in CSet", i);
+        _failures = true;
+        return true;
+      }
+      if (cset_state.is_in_cset()) {
+        gclog_or_tty->print_cr("\n## inconsistent cset state %d for humongous region %u", cset_state.value(), i);
+        _failures = true;
+        return true;
+      }
+      if (hr->continuesHumongous() && cset_state.is_humongous()) {
+        gclog_or_tty->print_cr("\n## inconsistent cset state %d for continues humongous region %u", cset_state.value(), i);
+        _failures = true;
+        return true;
+      }
+    } else {
+      if (cset_state.is_humongous()) {
+        gclog_or_tty->print_cr("\n## inconsistent cset state %d for non-humongous region %u", cset_state.value(), i);
+        _failures = true;
+        return true;
+      }
+      if (hr->in_collection_set() != cset_state.is_in_cset()) {
+        gclog_or_tty->print_cr("\n## in CSet %d / cset state %d inconsistency for region %u",
+                               hr->in_collection_set(), cset_state.value(), i);
+       _failures = true;
+       return true;
+      }
+      if (cset_state.is_in_cset()) {
+        if (hr->is_young() != (cset_state.is_young())) {
+          gclog_or_tty->print_cr("\n## is_young %d / cset state %d inconsistency for region %u",
+                                 hr->is_young(), cset_state.value(), i);
+          _failures = true;
+          return true;
+        }
+        if (hr->is_old() != (cset_state.is_old())) {
+          gclog_or_tty->print_cr("\n## is_old %d / cset state %d inconsistency for region %u",
+                                 hr->is_old(), cset_state.value(), i);
+          _failures = true;
+          return true;
+        }
+      }
+    }
+    return false;
+  }
+
+  bool failures() const { return _failures; }
+};
+
+bool G1CollectedHeap::check_cset_fast_test() {
+  G1CheckCSetFastTableClosure cl;
+  _hrm.iterate(&cl);
+  return !cl.failures();
+}
 #endif // PRODUCT
 
 void G1CollectedHeap::cleanUpCardTable() {
@@ -6409,47 +6336,47 @@
     // are completely up-to-date wrt to references to the humongous object.
     //
     // Other implementation considerations:
-    // - never consider object arrays: while they are a valid target, they have not
-    // been observed to be used as temporary objects.
-    // - they would also pose considerable effort for cleaning up the the remembered
-    // sets.
-    // While this cleanup is not strictly necessary to be done (or done instantly),
-    // given that their occurrence is very low, this saves us this additional
-    // complexity.
+    // - never consider object arrays at this time because they would pose
+    // considerable effort for cleaning up the the remembered sets. This is
+    // required because stale remembered sets might reference locations that
+    // are currently allocated into.
     uint region_idx = r->hrm_index();
-    if (g1h->humongous_is_live(region_idx) ||
-        g1h->humongous_region_is_always_live(region_idx)) {
-
-      if (G1TraceReclaimDeadHumongousObjectsAtYoungGC) {
-        gclog_or_tty->print_cr("Live humongous %d region %d with remset "SIZE_FORMAT" code roots "SIZE_FORMAT" is marked %d live-other %d obj array %d",
-                               r->isHumongous(),
+    if (!g1h->is_humongous_reclaim_candidate(region_idx) ||
+        !r->rem_set()->is_empty()) {
+
+      if (G1TraceEagerReclaimHumongousObjects) {
+        gclog_or_tty->print_cr("Live humongous region %u size "SIZE_FORMAT" start "PTR_FORMAT" length "UINT32_FORMAT" with remset "SIZE_FORMAT" code roots "SIZE_FORMAT" is marked %d reclaim candidate %d type array %d",
                                region_idx,
+                               obj->size()*HeapWordSize,
+                               r->bottom(),
+                               r->region_num(),
                                r->rem_set()->occupied(),
                                r->rem_set()->strong_code_roots_list_length(),
                                next_bitmap->isMarked(r->bottom()),
-                               g1h->humongous_is_live(region_idx),
-                               obj->is_objArray()
+                               g1h->is_humongous_reclaim_candidate(region_idx),
+                               obj->is_typeArray()
                               );
       }
 
       return false;
     }
 
-    guarantee(!obj->is_objArray(),
-              err_msg("Eagerly reclaiming object arrays is not supported, but the object "PTR_FORMAT" is.",
+    guarantee(obj->is_typeArray(),
+              err_msg("Only eagerly reclaiming type arrays is supported, but the object "
+                      PTR_FORMAT " is not.",
                       r->bottom()));
 
-    if (G1TraceReclaimDeadHumongousObjectsAtYoungGC) {
-      gclog_or_tty->print_cr("Reclaim humongous region %d start "PTR_FORMAT" region %d length "UINT32_FORMAT" with remset "SIZE_FORMAT" code roots "SIZE_FORMAT" is marked %d live-other %d obj array %d",
-                             r->isHumongous(),
+    if (G1TraceEagerReclaimHumongousObjects) {
+      gclog_or_tty->print_cr("Dead humongous region %u size "SIZE_FORMAT" start "PTR_FORMAT" length "UINT32_FORMAT" with remset "SIZE_FORMAT" code roots "SIZE_FORMAT" is marked %d reclaim candidate %d type array %d",
+                             region_idx,
+                             obj->size()*HeapWordSize,
                              r->bottom(),
-                             region_idx,
                              r->region_num(),
                              r->rem_set()->occupied(),
                              r->rem_set()->strong_code_roots_list_length(),
                              next_bitmap->isMarked(r->bottom()),
-                             g1h->humongous_is_live(region_idx),
-                             obj->is_objArray()
+                             g1h->is_humongous_reclaim_candidate(region_idx),
+                             obj->is_typeArray()
                             );
     }
     // Need to clear mark bit of the humongous object if already set.
@@ -6480,7 +6407,8 @@
 void G1CollectedHeap::eagerly_reclaim_humongous_regions() {
   assert_at_safepoint(true);
 
-  if (!G1ReclaimDeadHumongousObjectsAtYoungGC || !_has_humongous_reclaim_candidates) {
+  if (!G1EagerReclaimHumongousObjects ||
+      (!_has_humongous_reclaim_candidates && !G1TraceEagerReclaimHumongousObjects)) {
     g1_policy()->phase_times()->record_fast_reclaim_humongous_time_ms(0.0, 0);
     return;
   }
@@ -6793,20 +6721,20 @@
 
 HeapRegion* G1CollectedHeap::new_gc_alloc_region(size_t word_size,
                                                  uint count,
-                                                 GCAllocPurpose ap) {
+                                                 InCSetState dest) {
   assert(FreeList_lock->owned_by_self(), "pre-condition");
 
-  if (count < g1_policy()->max_regions(ap)) {
-    bool survivor = (ap == GCAllocForSurvived);
+  if (count < g1_policy()->max_regions(dest)) {
+    const bool is_survivor = (dest.is_young());
     HeapRegion* new_alloc_region = new_region(word_size,
-                                              !survivor,
+                                              !is_survivor,
                                               true /* do_expand */);
     if (new_alloc_region != NULL) {
       // We really only need to do this for old regions given that we
       // should never scan survivors. But it doesn't hurt to do it
       // for survivors too.
-      new_alloc_region->record_top_and_timestamp();
-      if (survivor) {
+      new_alloc_region->record_timestamp();
+      if (is_survivor) {
         new_alloc_region->set_survivor();
         _hr_printer.alloc(new_alloc_region, G1HRPrinter::Survivor);
         check_bitmaps("Survivor Region Allocation", new_alloc_region);
@@ -6818,8 +6746,6 @@
       bool during_im = g1_policy()->during_initial_mark_pause();
       new_alloc_region->note_start_of_copying(during_im);
       return new_alloc_region;
-    } else {
-      g1_policy()->note_alloc_region_limit_reached(ap);
     }
   }
   return NULL;
@@ -6827,11 +6753,11 @@
 
 void G1CollectedHeap::retire_gc_alloc_region(HeapRegion* alloc_region,
                                              size_t allocated_bytes,
-                                             GCAllocPurpose ap) {
+                                             InCSetState dest) {
   bool during_im = g1_policy()->during_initial_mark_pause();
   alloc_region->note_end_of_copying(during_im);
   g1_policy()->record_bytes_copied_during_gc(allocated_bytes);
-  if (ap == GCAllocForSurvived) {
+  if (dest.is_young()) {
     young_list()->add_survivor_region(alloc_region);
   } else {
     _old_set.add(alloc_region);
--- a/src/share/vm/gc_implementation/g1/g1CollectedHeap.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/gc_implementation/g1/g1CollectedHeap.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
   /*
- * Copyright (c) 2001, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2001, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -32,6 +32,7 @@
 #include "gc_implementation/g1/g1AllocRegion.hpp"
 #include "gc_implementation/g1/g1BiasedArray.hpp"
 #include "gc_implementation/g1/g1HRPrinter.hpp"
+#include "gc_implementation/g1/g1InCSetState.hpp"
 #include "gc_implementation/g1/g1MonitoringSupport.hpp"
 #include "gc_implementation/g1/g1SATBCardTableModRefBS.hpp"
 #include "gc_implementation/g1/g1YCTypes.hpp"
@@ -212,6 +213,9 @@
   // Other related classes.
   friend class G1MarkSweep;
 
+  // Testing classes.
+  friend class G1CheckCSetFastTableClosure;
+
 private:
   // The one and only G1CollectedHeap, so static functions can find it.
   static G1CollectedHeap* _g1h;
@@ -229,7 +233,6 @@
   // It keeps track of the humongous regions.
   HeapRegionSet _humongous_set;
 
-  void clear_humongous_is_live_table();
   void eagerly_reclaim_humongous_regions();
 
   // The number of regions we could create by expansion.
@@ -299,22 +302,26 @@
   // Helper for monitoring and management support.
   G1MonitoringSupport* _g1mm;
 
-  // Records whether the region at the given index is kept live by roots or
-  // references from the young generation.
-  class HumongousIsLiveBiasedMappedArray : public G1BiasedMappedArray<bool> {
+  // Records whether the region at the given index is (still) a
+  // candidate for eager reclaim.  Only valid for humongous start
+  // regions; other regions have unspecified values.  Humongous start
+  // regions are initialized at start of collection pause, with
+  // candidates removed from the set as they are found reachable from
+  // roots or the young generation.
+  class HumongousReclaimCandidates : public G1BiasedMappedArray<bool> {
    protected:
     bool default_value() const { return false; }
    public:
     void clear() { G1BiasedMappedArray<bool>::clear(); }
-    void set_live(uint region) {
-      set_by_index(region, true);
+    void set_candidate(uint region, bool value) {
+      set_by_index(region, value);
     }
-    bool is_live(uint region) {
+    bool is_candidate(uint region) {
       return get_by_index(region);
     }
   };
 
-  HumongousIsLiveBiasedMappedArray _humongous_is_live;
+  HumongousReclaimCandidates _humongous_reclaim_candidates;
   // Stores whether during humongous object registration we found candidate regions.
   // If not, we can skip a few steps.
   bool _has_humongous_reclaim_candidates;
@@ -339,13 +346,14 @@
 
   // Keeps track of how many "old marking cycles" (i.e., Full GCs or
   // concurrent cycles) we have started.
-  volatile unsigned int _old_marking_cycles_started;
+  volatile uint _old_marking_cycles_started;
 
   // Keeps track of how many "old marking cycles" (i.e., Full GCs or
   // concurrent cycles) we have completed.
-  volatile unsigned int _old_marking_cycles_completed;
+  volatile uint _old_marking_cycles_completed;
 
   bool _concurrent_cycle_started;
+  bool _heap_summary_sent;
 
   // This is a non-product method that is helpful for testing. It is
   // called at the end of a GC and artificially expands the heap by
@@ -362,6 +370,12 @@
   // heap after a compaction.
   void print_hrm_post_compaction();
 
+  // Create a memory mapper for auxiliary data structures of the given size and
+  // translation factor.
+  static G1RegionToSpaceMapper* create_aux_memory_mapper(const char* description,
+                                                         size_t size,
+                                                         size_t translation_factor);
+
   double verify(bool guard, const char* msg);
   void verify_before_gc();
   void verify_after_gc();
@@ -510,22 +524,22 @@
   // the mutator alloc region without taking the Heap_lock. This
   // should only be used for non-humongous allocations.
   inline HeapWord* attempt_allocation(size_t word_size,
-                                      unsigned int* gc_count_before_ret,
-                                      int* gclocker_retry_count_ret);
+                                      uint* gc_count_before_ret,
+                                      uint* gclocker_retry_count_ret);
 
   // Second-level mutator allocation attempt: take the Heap_lock and
   // retry the allocation attempt, potentially scheduling a GC
   // pause. This should only be used for non-humongous allocations.
   HeapWord* attempt_allocation_slow(size_t word_size,
                                     AllocationContext_t context,
-                                    unsigned int* gc_count_before_ret,
-                                    int* gclocker_retry_count_ret);
+                                    uint* gc_count_before_ret,
+                                    uint* gclocker_retry_count_ret);
 
   // Takes the Heap_lock and attempts a humongous allocation. It can
   // potentially schedule a GC pause.
   HeapWord* attempt_allocation_humongous(size_t word_size,
-                                         unsigned int* gc_count_before_ret,
-                                         int* gclocker_retry_count_ret);
+                                         uint* gc_count_before_ret,
+                                         uint* gclocker_retry_count_ret);
 
   // Allocation attempt that should be called during safepoints (e.g.,
   // at the end of a successful GC). expect_null_mutator_alloc_region
@@ -545,15 +559,9 @@
   // allocation region, either by picking one or expanding the
   // heap, and then allocate a block of the given size. The block
   // may not be a humongous - it must fit into a single heap region.
-  HeapWord* par_allocate_during_gc(GCAllocPurpose purpose,
-                                   size_t word_size,
-                                   AllocationContext_t context);
-
-  HeapWord* allocate_during_gc_slow(GCAllocPurpose purpose,
-                                    HeapRegion*    alloc_region,
-                                    bool           par,
-                                    size_t         word_size);
-
+  inline HeapWord* par_allocate_during_gc(InCSetState dest,
+                                          size_t word_size,
+                                          AllocationContext_t context);
   // Ensure that no further allocations can happen in "r", bearing in mind
   // that parallel threads might be attempting allocations.
   void par_allocate_remaining_space(HeapRegion* r);
@@ -575,9 +583,9 @@
 
   // For GC alloc regions.
   HeapRegion* new_gc_alloc_region(size_t word_size, uint count,
-                                  GCAllocPurpose ap);
+                                  InCSetState dest);
   void retire_gc_alloc_region(HeapRegion* alloc_region,
-                              size_t allocated_bytes, GCAllocPurpose ap);
+                              size_t allocated_bytes, InCSetState dest);
 
   // - if explicit_gc is true, the GC is for a System.gc() or a heap
   //   inspection request and should collect the entire heap
@@ -638,26 +646,11 @@
   // (Rounds up to a HeapRegion boundary.)
   bool expand(size_t expand_bytes);
 
-  // Returns the PLAB statistics given a purpose.
-  PLABStats* stats_for_purpose(GCAllocPurpose purpose) {
-    PLABStats* stats = NULL;
+  // Returns the PLAB statistics for a given destination.
+  inline PLABStats* alloc_buffer_stats(InCSetState dest);
 
-    switch (purpose) {
-    case GCAllocForSurvived:
-      stats = &_survivor_plab_stats;
-      break;
-    case GCAllocForTenured:
-      stats = &_old_plab_stats;
-      break;
-    default:
-      assert(false, "unrecognized GCAllocPurpose");
-    }
-
-    return stats;
-  }
-
-  // Determines PLAB size for a particular allocation purpose.
-  size_t desired_plab_sz(GCAllocPurpose purpose);
+  // Determines PLAB size for a given destination.
+  inline size_t desired_plab_sz(InCSetState dest);
 
   inline AllocationContextStats& allocation_context_stats();
 
@@ -665,15 +658,15 @@
   virtual void gc_prologue(bool full);
   virtual void gc_epilogue(bool full);
 
+  // Modify the reclaim candidate set and test for presence.
+  // These are only valid for starts_humongous regions.
+  inline void set_humongous_reclaim_candidate(uint region, bool value);
+  inline bool is_humongous_reclaim_candidate(uint region);
+
+  // Remove from the reclaim candidate set.  Also remove from the
+  // collection set so that later encounters avoid the slow path.
   inline void set_humongous_is_live(oop obj);
 
-  bool humongous_is_live(uint region) {
-    return _humongous_is_live.is_live(region);
-  }
-
-  // Returns whether the given region (which must be a humongous (start) region)
-  // is to be considered conservatively live regardless of any other conditions.
-  bool humongous_region_is_always_live(uint index);
   // Register the given region to be part of the collection set.
   inline void register_humongous_region_with_in_cset_fast_test(uint index);
   // Register regions with humongous objects (actually on the start region) in
@@ -681,8 +674,11 @@
   void register_humongous_regions_with_in_cset_fast_test();
   // We register a region with the fast "in collection set" test. We
   // simply set to true the array slot corresponding to this region.
-  void register_region_with_in_cset_fast_test(HeapRegion* r) {
-    _in_cset_fast_test.set_in_cset(r->hrm_index());
+  void register_young_region_with_in_cset_fast_test(HeapRegion* r) {
+    _in_cset_fast_test.set_in_young(r->hrm_index());
+  }
+  void register_old_region_with_in_cset_fast_test(HeapRegion* r) {
+    _in_cset_fast_test.set_in_old(r->hrm_index());
   }
 
   // This is a fast test on whether a reference points into the
@@ -714,7 +710,7 @@
   // +ExplicitGCInvokesConcurrent).
   void increment_old_marking_cycles_completed(bool concurrent);
 
-  unsigned int old_marking_cycles_completed() {
+  uint old_marking_cycles_completed() {
     return _old_marking_cycles_completed;
   }
 
@@ -773,7 +769,7 @@
   // methods that call do_collection_pause() release the Heap_lock
   // before the call, so it's easy to read gc_count_before just before.
   HeapWord* do_collection_pause(size_t         word_size,
-                                unsigned int   gc_count_before,
+                                uint           gc_count_before,
                                 bool*          succeeded,
                                 GCCause::Cause gc_cause);
 
@@ -812,22 +808,6 @@
   // statistics or updating free lists.
   void abandon_collection_set(HeapRegion* cs_head);
 
-  // Applies "scan_non_heap_roots" to roots outside the heap,
-  // "scan_rs" to roots inside the heap (having done "set_region" to
-  // indicate the region in which the root resides),
-  // and does "scan_metadata" If "scan_rs" is
-  // NULL, then this step is skipped.  The "worker_i"
-  // param is for use with parallel roots processing, and should be
-  // the "i" of the calling parallel worker thread's work(i) function.
-  // In the sequential case this param will be ignored.
-  void g1_process_roots(OopClosure* scan_non_heap_roots,
-                        OopClosure* scan_non_heap_weak_roots,
-                        OopsInHeapRegionClosure* scan_rs,
-                        CLDClosure* scan_strong_clds,
-                        CLDClosure* scan_weak_clds,
-                        CodeBlobClosure* scan_strong_code,
-                        uint worker_i);
-
   // The concurrent marker (and the thread it runs in.)
   ConcurrentMark* _cm;
   ConcurrentMarkThread* _cmThread;
@@ -1012,23 +992,12 @@
   // The heap region entry for a given worker is valid iff
   // the associated time stamp value matches the current value
   // of G1CollectedHeap::_gc_time_stamp.
-  unsigned int* _worker_cset_start_region_time_stamp;
-
-  enum G1H_process_roots_tasks {
-    G1H_PS_filter_satb_buffers,
-    G1H_PS_refProcessor_oops_do,
-    // Leave this one last.
-    G1H_PS_NumElements
-  };
-
-  SubTasksDone* _process_strong_tasks;
+  uint* _worker_cset_start_region_time_stamp;
 
   volatile bool _free_regions_coming;
 
 public:
 
-  SubTasksDone* process_strong_tasks() { return _process_strong_tasks; }
-
   void set_refine_cte_cl_concurrency(bool concurrent);
 
   RefToScanQueue *task_queue(int i) const;
@@ -1061,21 +1030,11 @@
   // Initialize weak reference processing.
   virtual void ref_processing_init();
 
-  void set_par_threads(uint t) {
-    SharedHeap::set_par_threads(t);
-    // Done in SharedHeap but oddly there are
-    // two _process_strong_tasks's in a G1CollectedHeap
-    // so do it here too.
-    _process_strong_tasks->set_n_threads(t);
-  }
-
+  // Explicitly import set_par_threads into this scope
+  using SharedHeap::set_par_threads;
   // Set _n_par_threads according to a policy TBD.
   void set_par_threads();
 
-  void set_n_termination(int t) {
-    _process_strong_tasks->set_n_threads(t);
-  }
-
   virtual CollectedHeap::Name kind() const {
     return CollectedHeap::G1CollectedHeap;
   }
@@ -1150,6 +1109,10 @@
   // The number of regions that are completely free.
   uint num_free_regions() const { return _hrm.num_free_regions(); }
 
+  MemoryUsage get_auxiliary_data_memory_usage() const {
+    return _hrm.get_auxiliary_data_memory_usage();
+  }
+
   // The number of regions that are not completely free.
   uint num_used_regions() const { return num_regions() - num_free_regions(); }
 
@@ -1182,6 +1145,9 @@
   // appropriate error messages and crash.
   void check_bitmaps(const char* caller) PRODUCT_RETURN;
 
+  // Do sanity check on the contents of the in-cset fast test table.
+  bool check_cset_fast_test() PRODUCT_RETURN_( return true; );
+
   // verify_region_sets() performs verification over the region
   // lists. It will be compiled in the product code to be used when
   // necessary (i.e., during heap verification).
@@ -1277,53 +1243,15 @@
 
   inline bool is_in_cset_or_humongous(const oop obj);
 
-  enum in_cset_state_t {
-   InNeither,           // neither in collection set nor humongous
-   InCSet,              // region is in collection set only
-   IsHumongous          // region is a humongous start region
-  };
  private:
-  // Instances of this class are used for quick tests on whether a reference points
-  // into the collection set or is a humongous object (points into a humongous
-  // object).
-  // Each of the array's elements denotes whether the corresponding region is in
-  // the collection set or a humongous region.
-  // We use this to quickly reclaim humongous objects: by making a humongous region
-  // succeed this test, we sort-of add it to the collection set. During the reference
-  // iteration closures, when we see a humongous region, we simply mark it as
-  // referenced, i.e. live.
-  class G1FastCSetBiasedMappedArray : public G1BiasedMappedArray<char> {
-   protected:
-    char default_value() const { return G1CollectedHeap::InNeither; }
-   public:
-    void set_humongous(uintptr_t index) {
-      assert(get_by_index(index) != InCSet, "Should not overwrite InCSet values");
-      set_by_index(index, G1CollectedHeap::IsHumongous);
-    }
-
-    void clear_humongous(uintptr_t index) {
-      set_by_index(index, G1CollectedHeap::InNeither);
-    }
-
-    void set_in_cset(uintptr_t index) {
-      assert(get_by_index(index) != G1CollectedHeap::IsHumongous, "Should not overwrite IsHumongous value");
-      set_by_index(index, G1CollectedHeap::InCSet);
-    }
-
-    bool is_in_cset_or_humongous(HeapWord* addr) const { return get_by_address(addr) != G1CollectedHeap::InNeither; }
-    bool is_in_cset(HeapWord* addr) const { return get_by_address(addr) == G1CollectedHeap::InCSet; }
-    G1CollectedHeap::in_cset_state_t at(HeapWord* addr) const { return (G1CollectedHeap::in_cset_state_t)get_by_address(addr); }
-    void clear() { G1BiasedMappedArray<char>::clear(); }
-  };
-
   // This array is used for a quick test on whether a reference points into
   // the collection set or not. Each of the array's elements denotes whether the
   // corresponding region is in the collection set or not.
-  G1FastCSetBiasedMappedArray _in_cset_fast_test;
+  G1InCSetStateFastTestBiasedMappedArray _in_cset_fast_test;
 
  public:
 
-  inline in_cset_state_t in_cset_state(const oop obj);
+  inline InCSetState in_cset_state(const oop obj);
 
   // Return "TRUE" iff the given object address is in the reserved
   // region of g1.
--- a/src/share/vm/gc_implementation/g1/g1CollectedHeap.inline.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/gc_implementation/g1/g1CollectedHeap.inline.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2001, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2001, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -35,6 +35,41 @@
 #include "runtime/orderAccess.inline.hpp"
 #include "utilities/taskqueue.hpp"
 
+PLABStats* G1CollectedHeap::alloc_buffer_stats(InCSetState dest) {
+  switch (dest.value()) {
+    case InCSetState::Young:
+      return &_survivor_plab_stats;
+    case InCSetState::Old:
+      return &_old_plab_stats;
+    default:
+      ShouldNotReachHere();
+      return NULL; // Keep some compilers happy
+  }
+}
+
+size_t G1CollectedHeap::desired_plab_sz(InCSetState dest) {
+  size_t gclab_word_size = alloc_buffer_stats(dest)->desired_plab_sz();
+  // Prevent humongous PLAB sizes for two reasons:
+  // * PLABs are allocated using a similar paths as oops, but should
+  //   never be in a humongous region
+  // * Allowing humongous PLABs needlessly churns the region free lists
+  return MIN2(_humongous_object_threshold_in_words, gclab_word_size);
+}
+
+HeapWord* G1CollectedHeap::par_allocate_during_gc(InCSetState dest,
+                                                  size_t word_size,
+                                                  AllocationContext_t context) {
+  switch (dest.value()) {
+    case InCSetState::Young:
+      return survivor_attempt_allocation(word_size, context);
+    case InCSetState::Old:
+      return old_attempt_allocation(word_size, context);
+    default:
+      ShouldNotReachHere();
+      return NULL; // Keep some compilers happy
+  }
+}
+
 // Inline functions for G1CollectedHeap
 
 inline AllocationContextStats& G1CollectedHeap::allocation_context_stats() {
@@ -96,8 +131,8 @@
 }
 
 inline HeapWord* G1CollectedHeap::attempt_allocation(size_t word_size,
-                                                     unsigned int* gc_count_before_ret,
-                                                     int* gclocker_retry_count_ret) {
+                                                     uint* gc_count_before_ret,
+                                                     uint* gclocker_retry_count_ret) {
   assert_heap_not_locked_and_not_at_safepoint();
   assert(!isHumongous(word_size), "attempt_allocation() should not "
          "be called for humongous allocation requests");
@@ -203,7 +238,7 @@
   return _in_cset_fast_test.is_in_cset_or_humongous((HeapWord*)obj);
 }
 
-G1CollectedHeap::in_cset_state_t G1CollectedHeap::in_cset_state(const oop obj) {
+InCSetState G1CollectedHeap::in_cset_state(const oop obj) {
   return _in_cset_fast_test.at((HeapWord*)obj);
 }
 
@@ -313,20 +348,30 @@
   return is_obj_ill(obj, heap_region_containing(obj));
 }
 
+inline void G1CollectedHeap::set_humongous_reclaim_candidate(uint region, bool value) {
+  assert(_hrm.at(region)->startsHumongous(), "Must start a humongous object");
+  _humongous_reclaim_candidates.set_candidate(region, value);
+}
+
+inline bool G1CollectedHeap::is_humongous_reclaim_candidate(uint region) {
+  assert(_hrm.at(region)->startsHumongous(), "Must start a humongous object");
+  return _humongous_reclaim_candidates.is_candidate(region);
+}
+
 inline void G1CollectedHeap::set_humongous_is_live(oop obj) {
   uint region = addr_to_region((HeapWord*)obj);
-  // We not only set the "live" flag in the humongous_is_live table, but also
+  // Clear the flag in the humongous_reclaim_candidates table.  Also
   // reset the entry in the _in_cset_fast_test table so that subsequent references
   // to the same humongous object do not go into the slow path again.
   // This is racy, as multiple threads may at the same time enter here, but this
   // is benign.
-  // During collection we only ever set the "live" flag, and only ever clear the
+  // During collection we only ever clear the "candidate" flag, and only ever clear the
   // entry in the in_cset_fast_table.
   // We only ever evaluate the contents of these tables (in the VM thread) after
   // having synchronized the worker threads with the VM thread, or in the same
   // thread (i.e. within the VM thread).
-  if (!_humongous_is_live.is_live(region)) {
-    _humongous_is_live.set_live(region);
+  if (is_humongous_reclaim_candidate(region)) {
+    set_humongous_reclaim_candidate(region, false);
     _in_cset_fast_test.clear_humongous(region);
   }
 }
--- a/src/share/vm/gc_implementation/g1/g1CollectorPolicy.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/gc_implementation/g1/g1CollectorPolicy.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1084,7 +1084,7 @@
   if (update_stats) {
     double cost_per_card_ms = 0.0;
     if (_pending_cards > 0) {
-      cost_per_card_ms = phase_times()->average_last_update_rs_time() / (double) _pending_cards;
+      cost_per_card_ms = phase_times()->average_time_ms(G1GCPhaseTimes::UpdateRS) / (double) _pending_cards;
       _cost_per_card_ms_seq->add(cost_per_card_ms);
     }
 
@@ -1092,7 +1092,7 @@
 
     double cost_per_entry_ms = 0.0;
     if (cards_scanned > 10) {
-      cost_per_entry_ms = phase_times()->average_last_scan_rs_time() / (double) cards_scanned;
+      cost_per_entry_ms = phase_times()->average_time_ms(G1GCPhaseTimes::ScanRS) / (double) cards_scanned;
       if (_last_gc_was_young) {
         _cost_per_entry_ms_seq->add(cost_per_entry_ms);
       } else {
@@ -1134,7 +1134,7 @@
     double cost_per_byte_ms = 0.0;
 
     if (copied_bytes > 0) {
-      cost_per_byte_ms = phase_times()->average_last_obj_copy_time() / (double) copied_bytes;
+      cost_per_byte_ms = phase_times()->average_time_ms(G1GCPhaseTimes::ObjCopy) / (double) copied_bytes;
       if (_in_marking_window) {
         _cost_per_byte_ms_during_cm_seq->add(cost_per_byte_ms);
       } else {
@@ -1143,8 +1143,8 @@
     }
 
     double all_other_time_ms = pause_time_ms -
-      (phase_times()->average_last_update_rs_time() + phase_times()->average_last_scan_rs_time()
-      + phase_times()->average_last_obj_copy_time() + phase_times()->average_last_termination_time());
+      (phase_times()->average_time_ms(G1GCPhaseTimes::UpdateRS) + phase_times()->average_time_ms(G1GCPhaseTimes::ScanRS) +
+          phase_times()->average_time_ms(G1GCPhaseTimes::ObjCopy) + phase_times()->average_time_ms(G1GCPhaseTimes::Termination));
 
     double young_other_time_ms = 0.0;
     if (young_cset_region_length() > 0) {
@@ -1185,8 +1185,8 @@
 
   // Note that _mmu_tracker->max_gc_time() returns the time in seconds.
   double update_rs_time_goal_ms = _mmu_tracker->max_gc_time() * MILLIUNITS * G1RSetUpdatingPauseTimePercent / 100.0;
-  adjust_concurrent_refinement(phase_times()->average_last_update_rs_time(),
-                               phase_times()->sum_last_update_rs_processed_buffers(), update_rs_time_goal_ms);
+  adjust_concurrent_refinement(phase_times()->average_time_ms(G1GCPhaseTimes::UpdateRS),
+                               phase_times()->sum_thread_work_items(G1GCPhaseTimes::UpdateRS), update_rs_time_goal_ms);
 
   _collectionSetChooser->verify();
 }
@@ -1437,18 +1437,6 @@
   return young_list_length < young_list_max_length;
 }
 
-uint G1CollectorPolicy::max_regions(int purpose) {
-  switch (purpose) {
-    case GCAllocForSurvived:
-      return _max_survivor_regions;
-    case GCAllocForTenured:
-      return REGIONS_UNLIMITED;
-    default:
-      ShouldNotReachHere();
-      return REGIONS_UNLIMITED;
-  };
-}
-
 void G1CollectorPolicy::update_max_gc_locker_expansion() {
   uint expansion_region_num = 0;
   if (GCLockerEdenExpansionPercent > 0) {
@@ -1683,7 +1671,7 @@
   hr->set_next_in_collection_set(_collection_set);
   _collection_set = hr;
   _collection_set_bytes_used_before += hr->used();
-  _g1->register_region_with_in_cset_fast_test(hr);
+  _g1->register_old_region_with_in_cset_fast_test(hr);
   size_t rs_length = hr->rem_set()->occupied();
   _recorded_rs_lengths += rs_length;
   _old_cset_region_length += 1;
@@ -1816,7 +1804,7 @@
   hr->set_in_collection_set(true);
   assert( hr->next_in_collection_set() == NULL, "invariant");
 
-  _g1->register_region_with_in_cset_fast_test(hr);
+  _g1->register_young_region_with_in_cset_fast_test(hr);
 }
 
 // Add the region at the RHS of the incremental cset
@@ -2189,19 +2177,19 @@
     _other.add(pause_time_ms - phase_times->accounted_time_ms());
     _root_region_scan_wait.add(phase_times->root_region_scan_wait_time_ms());
     _parallel.add(phase_times->cur_collection_par_time_ms());
-    _ext_root_scan.add(phase_times->average_last_ext_root_scan_time());
-    _satb_filtering.add(phase_times->average_last_satb_filtering_times_ms());
-    _update_rs.add(phase_times->average_last_update_rs_time());
-    _scan_rs.add(phase_times->average_last_scan_rs_time());
-    _obj_copy.add(phase_times->average_last_obj_copy_time());
-    _termination.add(phase_times->average_last_termination_time());
+    _ext_root_scan.add(phase_times->average_time_ms(G1GCPhaseTimes::ExtRootScan));
+    _satb_filtering.add(phase_times->average_time_ms(G1GCPhaseTimes::SATBFiltering));
+    _update_rs.add(phase_times->average_time_ms(G1GCPhaseTimes::UpdateRS));
+    _scan_rs.add(phase_times->average_time_ms(G1GCPhaseTimes::ScanRS));
+    _obj_copy.add(phase_times->average_time_ms(G1GCPhaseTimes::ObjCopy));
+    _termination.add(phase_times->average_time_ms(G1GCPhaseTimes::Termination));
 
-    double parallel_known_time = phase_times->average_last_ext_root_scan_time() +
-      phase_times->average_last_satb_filtering_times_ms() +
-      phase_times->average_last_update_rs_time() +
-      phase_times->average_last_scan_rs_time() +
-      phase_times->average_last_obj_copy_time() +
-      + phase_times->average_last_termination_time();
+    double parallel_known_time = phase_times->average_time_ms(G1GCPhaseTimes::ExtRootScan) +
+      phase_times->average_time_ms(G1GCPhaseTimes::SATBFiltering) +
+      phase_times->average_time_ms(G1GCPhaseTimes::UpdateRS) +
+      phase_times->average_time_ms(G1GCPhaseTimes::ScanRS) +
+      phase_times->average_time_ms(G1GCPhaseTimes::ObjCopy) +
+      phase_times->average_time_ms(G1GCPhaseTimes::Termination);
 
     double parallel_other_time = phase_times->cur_collection_par_time_ms() - parallel_known_time;
     _parallel_other.add(parallel_other_time);
--- a/src/share/vm/gc_implementation/g1/g1CollectorPolicy.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/gc_implementation/g1/g1CollectorPolicy.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -877,28 +877,20 @@
 public:
   uint tenuring_threshold() const { return _tenuring_threshold; }
 
-  inline GCAllocPurpose
-    evacuation_destination(HeapRegion* src_region, uint age, size_t word_sz) {
-      if (age < _tenuring_threshold && src_region->is_young()) {
-        return GCAllocForSurvived;
-      } else {
-        return GCAllocForTenured;
-      }
-  }
-
-  inline bool track_object_age(GCAllocPurpose purpose) {
-    return purpose == GCAllocForSurvived;
-  }
-
   static const uint REGIONS_UNLIMITED = (uint) -1;
 
-  uint max_regions(int purpose);
-
-  // The limit on regions for a particular purpose is reached.
-  void note_alloc_region_limit_reached(int purpose) {
-    if (purpose == GCAllocForSurvived) {
-      _tenuring_threshold = 0;
+  uint max_regions(InCSetState dest) {
+    switch (dest.value()) {
+      case InCSetState::Young:
+        return _max_survivor_regions;
+      case InCSetState::Old:
+        return REGIONS_UNLIMITED;
+      default:
+        assert(false, err_msg("Unknown dest state: " CSETSTATE_FORMAT, dest.value()));
+        break;
     }
+    // keep some compilers happy
+    return 0;
   }
 
   void note_start_adding_survivor_regions() {
--- a/src/share/vm/gc_implementation/g1/g1GCPhaseTimes.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/gc_implementation/g1/g1GCPhaseTimes.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -22,12 +22,13 @@
  *
  */
 
-
 #include "precompiled.hpp"
 #include "gc_implementation/g1/g1CollectedHeap.inline.hpp"
 #include "gc_implementation/g1/g1GCPhaseTimes.hpp"
 #include "gc_implementation/g1/g1Log.hpp"
 #include "gc_implementation/g1/g1StringDedup.hpp"
+#include "memory/allocation.hpp"
+#include "runtime/os.hpp"
 
 // Helper class for avoiding interleaved logging
 class LineBuffer: public StackObj {
@@ -70,184 +71,258 @@
     va_end(ap);
   }
 
+  void print_cr() {
+    gclog_or_tty->print_cr("%s", _buffer);
+    _cur = _indent_level * INDENT_CHARS;
+  }
+
   void append_and_print_cr(const char* format, ...)  ATTRIBUTE_PRINTF(2, 3) {
     va_list ap;
     va_start(ap, format);
     vappend(format, ap);
     va_end(ap);
-    gclog_or_tty->print_cr("%s", _buffer);
-    _cur = _indent_level * INDENT_CHARS;
+    print_cr();
   }
 };
 
-PRAGMA_DIAG_PUSH
-PRAGMA_FORMAT_NONLITERAL_IGNORED
 template <class T>
-void WorkerDataArray<T>::print(int level, const char* title) {
-  if (_length == 1) {
-    // No need for min, max, average and sum for only one worker
-    LineBuffer buf(level);
-    buf.append("[%s:  ", title);
-    buf.append(_print_format, _data[0]);
-    buf.append_and_print_cr("]");
-    return;
+class WorkerDataArray  : public CHeapObj<mtGC> {
+  friend class G1GCParPhasePrinter;
+  T*          _data;
+  uint        _length;
+  const char* _title;
+  bool        _print_sum;
+  int         _log_level;
+  uint        _indent_level;
+  bool        _enabled;
+
+  WorkerDataArray<size_t>* _thread_work_items;
+
+  NOT_PRODUCT(T uninitialized();)
+
+  // We are caching the sum and average to only have to calculate them once.
+  // This is not done in an MT-safe way. It is intended to allow single
+  // threaded code to call sum() and average() multiple times in any order
+  // without having to worry about the cost.
+  bool   _has_new_data;
+  T      _sum;
+  T      _min;
+  T      _max;
+  double _average;
+
+ public:
+  WorkerDataArray(uint length, const char* title, bool print_sum, int log_level, uint indent_level) :
+    _title(title), _length(0), _print_sum(print_sum), _log_level(log_level), _indent_level(indent_level),
+    _has_new_data(true), _thread_work_items(NULL), _enabled(true) {
+    assert(length > 0, "Must have some workers to store data for");
+    _length = length;
+    _data = NEW_C_HEAP_ARRAY(T, _length, mtGC);
   }
 
-  T min = _data[0];
-  T max = _data[0];
-  T sum = 0;
-
-  LineBuffer buf(level);
-  buf.append("[%s:", title);
-  for (uint i = 0; i < _length; ++i) {
-    T val = _data[i];
-    min = MIN2(val, min);
-    max = MAX2(val, max);
-    sum += val;
-    if (G1Log::finest()) {
-      buf.append("  ");
-      buf.append(_print_format, val);
-    }
+  ~WorkerDataArray() {
+    FREE_C_HEAP_ARRAY(T, _data, mtGC);
   }
 
-  if (G1Log::finest()) {
-    buf.append_and_print_cr("%s", "");
+  void link_thread_work_items(WorkerDataArray<size_t>* thread_work_items) {
+    _thread_work_items = thread_work_items;
   }
 
-  double avg = (double)sum / (double)_length;
-  buf.append(" Min: ");
-  buf.append(_print_format, min);
-  buf.append(", Avg: ");
-  buf.append("%.1lf", avg); // Always print average as a double
-  buf.append(", Max: ");
-  buf.append(_print_format, max);
-  buf.append(", Diff: ");
-  buf.append(_print_format, max - min);
-  if (_print_sum) {
-    // for things like the start and end times the sum is not
-    // that relevant
-    buf.append(", Sum: ");
-    buf.append(_print_format, sum);
+  WorkerDataArray<size_t>* thread_work_items() { return _thread_work_items; }
+
+  void set(uint worker_i, T value) {
+    assert(worker_i < _length, err_msg("Worker %d is greater than max: %d", worker_i, _length));
+    assert(_data[worker_i] == WorkerDataArray<T>::uninitialized(), err_msg("Overwriting data for worker %d in %s", worker_i, _title));
+    _data[worker_i] = value;
+    _has_new_data = true;
   }
-  buf.append_and_print_cr("]");
-}
-PRAGMA_DIAG_POP
+
+  void set_thread_work_item(uint worker_i, size_t value) {
+    assert(_thread_work_items != NULL, "No sub count");
+    _thread_work_items->set(worker_i, value);
+  }
+
+  T get(uint worker_i) {
+    assert(worker_i < _length, err_msg("Worker %d is greater than max: %d", worker_i, _length));
+    assert(_data[worker_i] != WorkerDataArray<T>::uninitialized(), err_msg("No data added for worker %d", worker_i));
+    return _data[worker_i];
+  }
+
+  void add(uint worker_i, T value) {
+    assert(worker_i < _length, err_msg("Worker %d is greater than max: %d", worker_i, _length));
+    assert(_data[worker_i] != WorkerDataArray<T>::uninitialized(), err_msg("No data to add to for worker %d", worker_i));
+    _data[worker_i] += value;
+    _has_new_data = true;
+  }
+
+  double average(){
+    calculate_totals();
+    return _average;
+  }
+
+  T sum() {
+    calculate_totals();
+    return _sum;
+  }
+
+  T minimum() {
+    calculate_totals();
+    return _min;
+  }
+
+  T maximum() {
+    calculate_totals();
+    return _max;
+  }
+
+  void reset() PRODUCT_RETURN;
+  void verify() PRODUCT_RETURN;
+
+  void set_enabled(bool enabled) { _enabled = enabled; }
+
+  int log_level() { return _log_level;  }
+
+ private:
+
+  void calculate_totals(){
+    if (!_has_new_data) {
+      return;
+    }
+
+    _sum = (T)0;
+    _min = _data[0];
+    _max = _min;
+    for (uint i = 0; i < _length; ++i) {
+      T val = _data[i];
+      _sum += val;
+      _min = MIN2(_min, val);
+      _max = MAX2(_max, val);
+    }
+    _average = (double)_sum / (double)_length;
+    _has_new_data = false;
+  }
+};
+
 
 #ifndef PRODUCT
 
-template <> const int WorkerDataArray<int>::_uninitialized = -1;
-template <> const double WorkerDataArray<double>::_uninitialized = -1.0;
-template <> const size_t WorkerDataArray<size_t>::_uninitialized = (size_t)-1;
+template <>
+size_t WorkerDataArray<size_t>::uninitialized() {
+  return (size_t)-1;
+}
+
+template <>
+double WorkerDataArray<double>::uninitialized() {
+  return -1.0;
+}
 
 template <class T>
 void WorkerDataArray<T>::reset() {
   for (uint i = 0; i < _length; i++) {
-    _data[i] = (T)_uninitialized;
+    _data[i] = WorkerDataArray<T>::uninitialized();
+  }
+  if (_thread_work_items != NULL) {
+    _thread_work_items->reset();
   }
 }
 
 template <class T>
 void WorkerDataArray<T>::verify() {
+  if (!_enabled) {
+    return;
+  }
+
   for (uint i = 0; i < _length; i++) {
-    assert(_data[i] != _uninitialized,
-        err_msg("Invalid data for worker " UINT32_FORMAT ", data: %lf, uninitialized: %lf",
-            i, (double)_data[i], (double)_uninitialized));
+    assert(_data[i] != WorkerDataArray<T>::uninitialized(),
+        err_msg("Invalid data for worker %u in '%s'", i, _title));
+  }
+  if (_thread_work_items != NULL) {
+    _thread_work_items->verify();
   }
 }
 
 #endif
 
 G1GCPhaseTimes::G1GCPhaseTimes(uint max_gc_threads) :
-  _max_gc_threads(max_gc_threads),
-  _last_gc_worker_start_times_ms(_max_gc_threads, "%.1lf", false),
-  _last_ext_root_scan_times_ms(_max_gc_threads, "%.1lf"),
-  _last_satb_filtering_times_ms(_max_gc_threads, "%.1lf"),
-  _last_update_rs_times_ms(_max_gc_threads, "%.1lf"),
-  _last_update_rs_processed_buffers(_max_gc_threads, "%d"),
-  _last_scan_rs_times_ms(_max_gc_threads, "%.1lf"),
-  _last_strong_code_root_scan_times_ms(_max_gc_threads, "%.1lf"),
-  _last_obj_copy_times_ms(_max_gc_threads, "%.1lf"),
-  _last_termination_times_ms(_max_gc_threads, "%.1lf"),
-  _last_termination_attempts(_max_gc_threads, SIZE_FORMAT),
-  _last_gc_worker_end_times_ms(_max_gc_threads, "%.1lf", false),
-  _last_gc_worker_times_ms(_max_gc_threads, "%.1lf"),
-  _last_gc_worker_other_times_ms(_max_gc_threads, "%.1lf"),
-  _last_redirty_logged_cards_time_ms(_max_gc_threads, "%.1lf"),
-  _last_redirty_logged_cards_processed_cards(_max_gc_threads, SIZE_FORMAT),
-  _cur_string_dedup_queue_fixup_worker_times_ms(_max_gc_threads, "%.1lf"),
-  _cur_string_dedup_table_fixup_worker_times_ms(_max_gc_threads, "%.1lf")
+  _max_gc_threads(max_gc_threads)
 {
   assert(max_gc_threads > 0, "Must have some GC threads");
+
+  _gc_par_phases[GCWorkerStart] = new WorkerDataArray<double>(max_gc_threads, "GC Worker Start (ms)", false, G1Log::LevelFiner, 2);
+  _gc_par_phases[ExtRootScan] = new WorkerDataArray<double>(max_gc_threads, "Ext Root Scanning (ms)", true, G1Log::LevelFiner, 2);
+
+  // Root scanning phases
+  _gc_par_phases[ThreadRoots] = new WorkerDataArray<double>(max_gc_threads, "Thread Roots (ms)", true, G1Log::LevelFinest, 3);
+  _gc_par_phases[StringTableRoots] = new WorkerDataArray<double>(max_gc_threads, "StringTable Roots (ms)", true, G1Log::LevelFinest, 3);
+  _gc_par_phases[UniverseRoots] = new WorkerDataArray<double>(max_gc_threads, "Universe Roots (ms)", true, G1Log::LevelFinest, 3);
+  _gc_par_phases[JNIRoots] = new WorkerDataArray<double>(max_gc_threads, "JNI Handles Roots (ms)", true, G1Log::LevelFinest, 3);
+  _gc_par_phases[ObjectSynchronizerRoots] = new WorkerDataArray<double>(max_gc_threads, "ObjectSynchronizer Roots (ms)", true, G1Log::LevelFinest, 3);
+  _gc_par_phases[FlatProfilerRoots] = new WorkerDataArray<double>(max_gc_threads, "FlatProfiler Roots (ms)", true, G1Log::LevelFinest, 3);
+  _gc_par_phases[ManagementRoots] = new WorkerDataArray<double>(max_gc_threads, "Management Roots (ms)", true, G1Log::LevelFinest, 3);
+  _gc_par_phases[SystemDictionaryRoots] = new WorkerDataArray<double>(max_gc_threads, "SystemDictionary Roots (ms)", true, G1Log::LevelFinest, 3);
+  _gc_par_phases[CLDGRoots] = new WorkerDataArray<double>(max_gc_threads, "CLDG Roots (ms)", true, G1Log::LevelFinest, 3);
+  _gc_par_phases[JVMTIRoots] = new WorkerDataArray<double>(max_gc_threads, "JVMTI Roots (ms)", true, G1Log::LevelFinest, 3);
+  _gc_par_phases[CodeCacheRoots] = new WorkerDataArray<double>(max_gc_threads, "CodeCache Roots (ms)", true, G1Log::LevelFinest, 3);
+  _gc_par_phases[CMRefRoots] = new WorkerDataArray<double>(max_gc_threads, "CM RefProcessor Roots (ms)", true, G1Log::LevelFinest, 3);
+  _gc_par_phases[WaitForStrongCLD] = new WorkerDataArray<double>(max_gc_threads, "Wait For Strong CLD (ms)", true, G1Log::LevelFinest, 3);
+  _gc_par_phases[WeakCLDRoots] = new WorkerDataArray<double>(max_gc_threads, "Weak CLD Roots (ms)", true, G1Log::LevelFinest, 3);
+  _gc_par_phases[SATBFiltering] = new WorkerDataArray<double>(max_gc_threads, "SATB Filtering (ms)", true, G1Log::LevelFinest, 3);
+
+  _gc_par_phases[UpdateRS] = new WorkerDataArray<double>(max_gc_threads, "Update RS (ms)", true, G1Log::LevelFiner, 2);
+  _gc_par_phases[ScanRS] = new WorkerDataArray<double>(max_gc_threads, "Scan RS (ms)", true, G1Log::LevelFiner, 2);
+  _gc_par_phases[CodeRoots] = new WorkerDataArray<double>(max_gc_threads, "Code Root Scanning (ms)", true, G1Log::LevelFiner, 2);
+  _gc_par_phases[ObjCopy] = new WorkerDataArray<double>(max_gc_threads, "Object Copy (ms)", true, G1Log::LevelFiner, 2);
+  _gc_par_phases[Termination] = new WorkerDataArray<double>(max_gc_threads, "Termination (ms)", true, G1Log::LevelFiner, 2);
+  _gc_par_phases[GCWorkerTotal] = new WorkerDataArray<double>(max_gc_threads, "GC Worker Total (ms)", true, G1Log::LevelFiner, 2);
+  _gc_par_phases[GCWorkerEnd] = new WorkerDataArray<double>(max_gc_threads, "GC Worker End (ms)", false, G1Log::LevelFiner, 2);
+  _gc_par_phases[Other] = new WorkerDataArray<double>(max_gc_threads, "GC Worker Other (ms)", true, G1Log::LevelFiner, 2);
+
+  _update_rs_processed_buffers = new WorkerDataArray<size_t>(max_gc_threads, "Processed Buffers", true, G1Log::LevelFiner, 3);
+  _gc_par_phases[UpdateRS]->link_thread_work_items(_update_rs_processed_buffers);
+
+  _termination_attempts = new WorkerDataArray<size_t>(max_gc_threads, "Termination Attempts", true, G1Log::LevelFinest, 3);
+  _gc_par_phases[Termination]->link_thread_work_items(_termination_attempts);
+
+  _gc_par_phases[StringDedupQueueFixup] = new WorkerDataArray<double>(max_gc_threads, "Queue Fixup (ms)", true, G1Log::LevelFiner, 2);
+  _gc_par_phases[StringDedupTableFixup] = new WorkerDataArray<double>(max_gc_threads, "Table Fixup (ms)", true, G1Log::LevelFiner, 2);
+
+  _gc_par_phases[RedirtyCards] = new WorkerDataArray<double>(max_gc_threads, "Parallel Redirty", true, G1Log::LevelFinest, 3);
+  _redirtied_cards = new WorkerDataArray<size_t>(max_gc_threads, "Redirtied Cards", true, G1Log::LevelFinest, 3);
+  _gc_par_phases[RedirtyCards]->link_thread_work_items(_redirtied_cards);
 }
 
-void G1GCPhaseTimes::note_gc_start(uint active_gc_threads) {
+void G1GCPhaseTimes::note_gc_start(uint active_gc_threads, bool mark_in_progress) {
   assert(active_gc_threads > 0, "The number of threads must be > 0");
-  assert(active_gc_threads <= _max_gc_threads, "The number of active threads must be <= the max nubmer of threads");
+  assert(active_gc_threads <= _max_gc_threads, "The number of active threads must be <= the max number of threads");
   _active_gc_threads = active_gc_threads;
 
-  _last_gc_worker_start_times_ms.reset();
-  _last_ext_root_scan_times_ms.reset();
-  _last_satb_filtering_times_ms.reset();
-  _last_update_rs_times_ms.reset();
-  _last_update_rs_processed_buffers.reset();
-  _last_scan_rs_times_ms.reset();
-  _last_strong_code_root_scan_times_ms.reset();
-  _last_obj_copy_times_ms.reset();
-  _last_termination_times_ms.reset();
-  _last_termination_attempts.reset();
-  _last_gc_worker_end_times_ms.reset();
-  _last_gc_worker_times_ms.reset();
-  _last_gc_worker_other_times_ms.reset();
+  for (int i = 0; i < GCParPhasesSentinel; i++) {
+    _gc_par_phases[i]->reset();
+  }
 
-  _last_redirty_logged_cards_time_ms.reset();
-  _last_redirty_logged_cards_processed_cards.reset();
-
+  _gc_par_phases[StringDedupQueueFixup]->set_enabled(G1StringDedup::is_enabled());
+  _gc_par_phases[StringDedupTableFixup]->set_enabled(G1StringDedup::is_enabled());
 }
 
 void G1GCPhaseTimes::note_gc_end() {
-  _last_gc_worker_start_times_ms.verify();
-  _last_ext_root_scan_times_ms.verify();
-  _last_satb_filtering_times_ms.verify();
-  _last_update_rs_times_ms.verify();
-  _last_update_rs_processed_buffers.verify();
-  _last_scan_rs_times_ms.verify();
-  _last_strong_code_root_scan_times_ms.verify();
-  _last_obj_copy_times_ms.verify();
-  _last_termination_times_ms.verify();
-  _last_termination_attempts.verify();
-  _last_gc_worker_end_times_ms.verify();
+  for (uint i = 0; i < _active_gc_threads; i++) {
+    double worker_time = _gc_par_phases[GCWorkerEnd]->get(i) - _gc_par_phases[GCWorkerStart]->get(i);
+    record_time_secs(GCWorkerTotal, i , worker_time);
 
-  for (uint i = 0; i < _active_gc_threads; i++) {
-    double worker_time = _last_gc_worker_end_times_ms.get(i) - _last_gc_worker_start_times_ms.get(i);
-    _last_gc_worker_times_ms.set(i, worker_time);
+    double worker_known_time =
+        _gc_par_phases[ExtRootScan]->get(i) +
+        _gc_par_phases[SATBFiltering]->get(i) +
+        _gc_par_phases[UpdateRS]->get(i) +
+        _gc_par_phases[ScanRS]->get(i) +
+        _gc_par_phases[CodeRoots]->get(i) +
+        _gc_par_phases[ObjCopy]->get(i) +
+        _gc_par_phases[Termination]->get(i);
 
-    double worker_known_time = _last_ext_root_scan_times_ms.get(i) +
-                               _last_satb_filtering_times_ms.get(i) +
-                               _last_update_rs_times_ms.get(i) +
-                               _last_scan_rs_times_ms.get(i) +
-                               _last_strong_code_root_scan_times_ms.get(i) +
-                               _last_obj_copy_times_ms.get(i) +
-                               _last_termination_times_ms.get(i);
-
-    double worker_other_time = worker_time - worker_known_time;
-    _last_gc_worker_other_times_ms.set(i, worker_other_time);
+    record_time_secs(Other, i, worker_time - worker_known_time);
   }
 
-  _last_gc_worker_times_ms.verify();
-  _last_gc_worker_other_times_ms.verify();
-
-  _last_redirty_logged_cards_time_ms.verify();
-  _last_redirty_logged_cards_processed_cards.verify();
-}
-
-void G1GCPhaseTimes::note_string_dedup_fixup_start() {
-  _cur_string_dedup_queue_fixup_worker_times_ms.reset();
-  _cur_string_dedup_table_fixup_worker_times_ms.reset();
-}
-
-void G1GCPhaseTimes::note_string_dedup_fixup_end() {
-  _cur_string_dedup_queue_fixup_worker_times_ms.verify();
-  _cur_string_dedup_table_fixup_worker_times_ms.verify();
+  for (int i = 0; i < GCParPhasesSentinel; i++) {
+    _gc_par_phases[i]->verify();
+  }
 }
 
 void G1GCPhaseTimes::print_stats(int level, const char* str, double value) {
@@ -259,7 +334,7 @@
 }
 
 void G1GCPhaseTimes::print_stats(int level, const char* str, double value, uint workers) {
-  LineBuffer(level).append_and_print_cr("[%s: %.1lf ms, GC Workers: " UINT32_FORMAT "]", str, value, workers);
+  LineBuffer(level).append_and_print_cr("[%s: %.1lf ms, GC Workers: %u]", str, value, workers);
 }
 
 double G1GCPhaseTimes::accounted_time_ms() {
@@ -287,46 +362,172 @@
     return misc_time_ms;
 }
 
+// record the time a phase took in seconds
+void G1GCPhaseTimes::record_time_secs(GCParPhases phase, uint worker_i, double secs) {
+  _gc_par_phases[phase]->set(worker_i, secs);
+}
+
+// add a number of seconds to a phase
+void G1GCPhaseTimes::add_time_secs(GCParPhases phase, uint worker_i, double secs) {
+  _gc_par_phases[phase]->add(worker_i, secs);
+}
+
+void G1GCPhaseTimes::record_thread_work_item(GCParPhases phase, uint worker_i, size_t count) {
+  _gc_par_phases[phase]->set_thread_work_item(worker_i, count);
+}
+
+// return the average time for a phase in milliseconds
+double G1GCPhaseTimes::average_time_ms(GCParPhases phase) {
+  return _gc_par_phases[phase]->average() * 1000.0;
+}
+
+double G1GCPhaseTimes::get_time_ms(GCParPhases phase, uint worker_i) {
+  return _gc_par_phases[phase]->get(worker_i) * 1000.0;
+}
+
+double G1GCPhaseTimes::sum_time_ms(GCParPhases phase) {
+  return _gc_par_phases[phase]->sum() * 1000.0;
+}
+
+double G1GCPhaseTimes::min_time_ms(GCParPhases phase) {
+  return _gc_par_phases[phase]->minimum() * 1000.0;
+}
+
+double G1GCPhaseTimes::max_time_ms(GCParPhases phase) {
+  return _gc_par_phases[phase]->maximum() * 1000.0;
+}
+
+size_t G1GCPhaseTimes::get_thread_work_item(GCParPhases phase, uint worker_i) {
+  assert(_gc_par_phases[phase]->thread_work_items() != NULL, "No sub count");
+  return _gc_par_phases[phase]->thread_work_items()->get(worker_i);
+}
+
+size_t G1GCPhaseTimes::sum_thread_work_items(GCParPhases phase) {
+  assert(_gc_par_phases[phase]->thread_work_items() != NULL, "No sub count");
+  return _gc_par_phases[phase]->thread_work_items()->sum();
+}
+
+double G1GCPhaseTimes::average_thread_work_items(GCParPhases phase) {
+  assert(_gc_par_phases[phase]->thread_work_items() != NULL, "No sub count");
+  return _gc_par_phases[phase]->thread_work_items()->average();
+}
+
+size_t G1GCPhaseTimes::min_thread_work_items(GCParPhases phase) {
+  assert(_gc_par_phases[phase]->thread_work_items() != NULL, "No sub count");
+  return _gc_par_phases[phase]->thread_work_items()->minimum();
+}
+
+size_t G1GCPhaseTimes::max_thread_work_items(GCParPhases phase) {
+  assert(_gc_par_phases[phase]->thread_work_items() != NULL, "No sub count");
+  return _gc_par_phases[phase]->thread_work_items()->maximum();
+}
+
+class G1GCParPhasePrinter : public StackObj {
+  G1GCPhaseTimes* _phase_times;
+ public:
+  G1GCParPhasePrinter(G1GCPhaseTimes* phase_times) : _phase_times(phase_times) {}
+
+  void print(G1GCPhaseTimes::GCParPhases phase_id) {
+    WorkerDataArray<double>* phase = _phase_times->_gc_par_phases[phase_id];
+
+    if (phase->_log_level > G1Log::level() || !phase->_enabled) {
+      return;
+    }
+
+    if (phase->_length == 1) {
+      print_single_length(phase_id, phase);
+    } else {
+      print_multi_length(phase_id, phase);
+    }
+  }
+
+ private:
+
+  void print_single_length(G1GCPhaseTimes::GCParPhases phase_id, WorkerDataArray<double>* phase) {
+    // No need for min, max, average and sum for only one worker
+    LineBuffer buf(phase->_indent_level);
+    buf.append_and_print_cr("[%s:  %.1lf]", phase->_title, _phase_times->get_time_ms(phase_id, 0));
+
+    if (phase->_thread_work_items != NULL) {
+      LineBuffer buf2(phase->_thread_work_items->_indent_level);
+      buf2.append_and_print_cr("[%s:  "SIZE_FORMAT"]", phase->_thread_work_items->_title, _phase_times->sum_thread_work_items(phase_id));
+    }
+  }
+
+  void print_time_values(LineBuffer& buf, G1GCPhaseTimes::GCParPhases phase_id, WorkerDataArray<double>* phase) {
+    for (uint i = 0; i < phase->_length; ++i) {
+      buf.append("  %.1lf", _phase_times->get_time_ms(phase_id, i));
+    }
+    buf.print_cr();
+  }
+
+  void print_count_values(LineBuffer& buf, G1GCPhaseTimes::GCParPhases phase_id, WorkerDataArray<size_t>* thread_work_items) {
+    for (uint i = 0; i < thread_work_items->_length; ++i) {
+      buf.append("  " SIZE_FORMAT, _phase_times->get_thread_work_item(phase_id, i));
+    }
+    buf.print_cr();
+  }
+
+  void print_thread_work_items(G1GCPhaseTimes::GCParPhases phase_id, WorkerDataArray<size_t>* thread_work_items) {
+    LineBuffer buf(thread_work_items->_indent_level);
+    buf.append("[%s:", thread_work_items->_title);
+
+    if (G1Log::finest()) {
+      print_count_values(buf, phase_id, thread_work_items);
+    }
+
+    assert(thread_work_items->_print_sum, err_msg("%s does not have print sum true even though it is a count", thread_work_items->_title));
+
+    buf.append_and_print_cr(" Min: " SIZE_FORMAT ", Avg: %.1lf, Max: " SIZE_FORMAT ", Diff: " SIZE_FORMAT ", Sum: " SIZE_FORMAT "]",
+        _phase_times->min_thread_work_items(phase_id), _phase_times->average_thread_work_items(phase_id), _phase_times->max_thread_work_items(phase_id),
+        _phase_times->max_thread_work_items(phase_id) - _phase_times->min_thread_work_items(phase_id), _phase_times->sum_thread_work_items(phase_id));
+  }
+
+  void print_multi_length(G1GCPhaseTimes::GCParPhases phase_id, WorkerDataArray<double>* phase) {
+    LineBuffer buf(phase->_indent_level);
+    buf.append("[%s:", phase->_title);
+
+    if (G1Log::finest()) {
+      print_time_values(buf, phase_id, phase);
+    }
+
+    buf.append(" Min: %.1lf, Avg: %.1lf, Max: %.1lf, Diff: %.1lf",
+        _phase_times->min_time_ms(phase_id), _phase_times->average_time_ms(phase_id), _phase_times->max_time_ms(phase_id),
+        _phase_times->max_time_ms(phase_id) - _phase_times->min_time_ms(phase_id));
+
+    if (phase->_print_sum) {
+      // for things like the start and end times the sum is not
+      // that relevant
+      buf.append(", Sum: %.1lf", _phase_times->sum_time_ms(phase_id));
+    }
+
+    buf.append_and_print_cr("]");
+
+    if (phase->_thread_work_items != NULL) {
+      print_thread_work_items(phase_id, phase->_thread_work_items);
+    }
+  }
+};
+
 void G1GCPhaseTimes::print(double pause_time_sec) {
+  G1GCParPhasePrinter par_phase_printer(this);
+
   if (_root_region_scan_wait_time_ms > 0.0) {
     print_stats(1, "Root Region Scan Waiting", _root_region_scan_wait_time_ms);
   }
-  if (G1CollectedHeap::use_parallel_gc_threads()) {
-    print_stats(1, "Parallel Time", _cur_collection_par_time_ms, _active_gc_threads);
-    _last_gc_worker_start_times_ms.print(2, "GC Worker Start (ms)");
-    _last_ext_root_scan_times_ms.print(2, "Ext Root Scanning (ms)");
-    if (_last_satb_filtering_times_ms.sum() > 0.0) {
-      _last_satb_filtering_times_ms.print(2, "SATB Filtering (ms)");
-    }
-    _last_update_rs_times_ms.print(2, "Update RS (ms)");
-      _last_update_rs_processed_buffers.print(3, "Processed Buffers");
-    _last_scan_rs_times_ms.print(2, "Scan RS (ms)");
-    _last_strong_code_root_scan_times_ms.print(2, "Code Root Scanning (ms)");
-    _last_obj_copy_times_ms.print(2, "Object Copy (ms)");
-    _last_termination_times_ms.print(2, "Termination (ms)");
-    if (G1Log::finest()) {
-      _last_termination_attempts.print(3, "Termination Attempts");
-    }
-    _last_gc_worker_other_times_ms.print(2, "GC Worker Other (ms)");
-    _last_gc_worker_times_ms.print(2, "GC Worker Total (ms)");
-    _last_gc_worker_end_times_ms.print(2, "GC Worker End (ms)");
-  } else {
-    _last_ext_root_scan_times_ms.print(1, "Ext Root Scanning (ms)");
-    if (_last_satb_filtering_times_ms.sum() > 0.0) {
-      _last_satb_filtering_times_ms.print(1, "SATB Filtering (ms)");
-    }
-    _last_update_rs_times_ms.print(1, "Update RS (ms)");
-      _last_update_rs_processed_buffers.print(2, "Processed Buffers");
-    _last_scan_rs_times_ms.print(1, "Scan RS (ms)");
-    _last_strong_code_root_scan_times_ms.print(1, "Code Root Scanning (ms)");
-    _last_obj_copy_times_ms.print(1, "Object Copy (ms)");
+
+  print_stats(1, "Parallel Time", _cur_collection_par_time_ms, _active_gc_threads);
+  for (int i = 0; i <= GCMainParPhasesLast; i++) {
+    par_phase_printer.print((GCParPhases) i);
   }
+
   print_stats(1, "Code Root Fixup", _cur_collection_code_root_fixup_time_ms);
   print_stats(1, "Code Root Purge", _cur_strong_code_root_purge_time_ms);
   if (G1StringDedup::is_enabled()) {
     print_stats(1, "String Dedup Fixup", _cur_string_dedup_fixup_time_ms, _active_gc_threads);
-    _cur_string_dedup_queue_fixup_worker_times_ms.print(2, "Queue Fixup (ms)");
-    _cur_string_dedup_table_fixup_worker_times_ms.print(2, "Table Fixup (ms)");
+    for (int i = StringDedupPhasesFirst; i <= StringDedupPhasesLast; i++) {
+      par_phase_printer.print((GCParPhases) i);
+    }
   }
   print_stats(1, "Clear CT", _cur_clear_ct_time_ms);
   double misc_time_ms = pause_time_sec * MILLIUNITS - accounted_time_ms();
@@ -350,15 +551,16 @@
   print_stats(2, "Ref Proc", _cur_ref_proc_time_ms);
   print_stats(2, "Ref Enq", _cur_ref_enq_time_ms);
   print_stats(2, "Redirty Cards", _recorded_redirty_logged_cards_time_ms);
-  if (G1Log::finest()) {
-    _last_redirty_logged_cards_time_ms.print(3, "Parallel Redirty");
-    _last_redirty_logged_cards_processed_cards.print(3, "Redirtied Cards");
-  }
-  if (G1ReclaimDeadHumongousObjectsAtYoungGC) {
-    print_stats(2, "Humongous Reclaim", _cur_fast_reclaim_humongous_time_ms);
+  par_phase_printer.print(RedirtyCards);
+
+  if (G1EagerReclaimHumongousObjects) {
+    print_stats(2, "Humongous Register", _cur_fast_reclaim_humongous_register_time_ms);
     if (G1Log::finest()) {
       print_stats(3, "Humongous Total", _cur_fast_reclaim_humongous_total);
       print_stats(3, "Humongous Candidate", _cur_fast_reclaim_humongous_candidates);
+    }
+    print_stats(2, "Humongous Reclaim", _cur_fast_reclaim_humongous_time_ms);
+    if (G1Log::finest()) {
       print_stats(3, "Humongous Reclaimed", _cur_fast_reclaim_humongous_reclaimed);
     }
   }
@@ -373,3 +575,17 @@
     print_stats(2, "Verify After", _cur_verify_after_time_ms);
   }
 }
+
+G1GCParPhaseTimesTracker::G1GCParPhaseTimesTracker(G1GCPhaseTimes* phase_times, G1GCPhaseTimes::GCParPhases phase, uint worker_id) :
+    _phase_times(phase_times), _phase(phase), _worker_id(worker_id) {
+  if (_phase_times != NULL) {
+    _start_time = os::elapsedTime();
+  }
+}
+
+G1GCParPhaseTimesTracker::~G1GCParPhaseTimesTracker() {
+  if (_phase_times != NULL) {
+    _phase_times->record_time_secs(_phase, _worker_id, os::elapsedTime() - _start_time);
+  }
+}
+
--- a/src/share/vm/gc_implementation/g1/g1GCPhaseTimes.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/gc_implementation/g1/g1GCPhaseTimes.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -26,106 +26,60 @@
 #define SHARE_VM_GC_IMPLEMENTATION_G1_G1GCPHASETIMESLOG_HPP
 
 #include "memory/allocation.hpp"
-#include "gc_interface/gcCause.hpp"
 
-template <class T>
-class WorkerDataArray  : public CHeapObj<mtGC> {
-  T*          _data;
-  uint        _length;
-  const char* _print_format;
-  bool        _print_sum;
+class LineBuffer;
 
-  NOT_PRODUCT(static const T _uninitialized;)
-
-  // We are caching the sum and average to only have to calculate them once.
-  // This is not done in an MT-safe way. It is intended to allow single
-  // threaded code to call sum() and average() multiple times in any order
-  // without having to worry about the cost.
-  bool   _has_new_data;
-  T      _sum;
-  double _average;
-
- public:
-  WorkerDataArray(uint length, const char* print_format, bool print_sum = true) :
-  _length(length), _print_format(print_format), _print_sum(print_sum), _has_new_data(true) {
-    assert(length > 0, "Must have some workers to store data for");
-    _data = NEW_C_HEAP_ARRAY(T, _length, mtGC);
-  }
-
-  ~WorkerDataArray() {
-    FREE_C_HEAP_ARRAY(T, _data, mtGC);
-  }
-
-  void set(uint worker_i, T value) {
-    assert(worker_i < _length, err_msg("Worker %d is greater than max: %d", worker_i, _length));
-    assert(_data[worker_i] == (T)-1, err_msg("Overwriting data for worker %d", worker_i));
-    _data[worker_i] = value;
-    _has_new_data = true;
-  }
-
-  T get(uint worker_i) {
-    assert(worker_i < _length, err_msg("Worker %d is greater than max: %d", worker_i, _length));
-    assert(_data[worker_i] != (T)-1, err_msg("No data to add to for worker %d", worker_i));
-    return _data[worker_i];
-  }
-
-  void add(uint worker_i, T value) {
-    assert(worker_i < _length, err_msg("Worker %d is greater than max: %d", worker_i, _length));
-    assert(_data[worker_i] != (T)-1, err_msg("No data to add to for worker %d", worker_i));
-    _data[worker_i] += value;
-    _has_new_data = true;
-  }
-
-  double average(){
-    if (_has_new_data) {
-      calculate_totals();
-    }
-    return _average;
-  }
-
-  T sum() {
-    if (_has_new_data) {
-      calculate_totals();
-    }
-    return _sum;
-  }
-
-  void print(int level, const char* title);
-
-  void reset() PRODUCT_RETURN;
-  void verify() PRODUCT_RETURN;
-
- private:
-
-  void calculate_totals(){
-    _sum = (T)0;
-    for (uint i = 0; i < _length; ++i) {
-      _sum += _data[i];
-    }
-    _average = (double)_sum / (double)_length;
-    _has_new_data = false;
-  }
-};
+template <class T> class WorkerDataArray;
 
 class G1GCPhaseTimes : public CHeapObj<mtGC> {
+  friend class G1GCParPhasePrinter;
 
- private:
   uint _active_gc_threads;
   uint _max_gc_threads;
 
-  WorkerDataArray<double> _last_gc_worker_start_times_ms;
-  WorkerDataArray<double> _last_ext_root_scan_times_ms;
-  WorkerDataArray<double> _last_satb_filtering_times_ms;
-  WorkerDataArray<double> _last_update_rs_times_ms;
-  WorkerDataArray<int>    _last_update_rs_processed_buffers;
-  WorkerDataArray<double> _last_scan_rs_times_ms;
-  WorkerDataArray<double> _last_strong_code_root_scan_times_ms;
-  WorkerDataArray<double> _last_obj_copy_times_ms;
-  WorkerDataArray<double> _last_termination_times_ms;
-  WorkerDataArray<size_t> _last_termination_attempts;
-  WorkerDataArray<double> _last_gc_worker_end_times_ms;
-  WorkerDataArray<double> _last_gc_worker_times_ms;
-  WorkerDataArray<double> _last_gc_worker_other_times_ms;
+ public:
+  enum GCParPhases {
+    GCWorkerStart,
+    ExtRootScan,
+    ThreadRoots,
+    StringTableRoots,
+    UniverseRoots,
+    JNIRoots,
+    ObjectSynchronizerRoots,
+    FlatProfilerRoots,
+    ManagementRoots,
+    SystemDictionaryRoots,
+    CLDGRoots,
+    JVMTIRoots,
+    CodeCacheRoots,
+    CMRefRoots,
+    WaitForStrongCLD,
+    WeakCLDRoots,
+    SATBFiltering,
+    UpdateRS,
+    ScanRS,
+    CodeRoots,
+    ObjCopy,
+    Termination,
+    Other,
+    GCWorkerTotal,
+    GCWorkerEnd,
+    StringDedupQueueFixup,
+    StringDedupTableFixup,
+    RedirtyCards,
+    GCParPhasesSentinel
+  };
+
+ private:
+  // Markers for grouping the phases in the GCPhases enum above
+  static const int GCMainParPhasesLast = GCWorkerEnd;
+  static const int StringDedupPhasesFirst = StringDedupQueueFixup;
+  static const int StringDedupPhasesLast = StringDedupTableFixup;
+
+  WorkerDataArray<double>* _gc_par_phases[GCParPhasesSentinel];
+  WorkerDataArray<size_t>* _update_rs_processed_buffers;
+  WorkerDataArray<size_t>* _termination_attempts;
+  WorkerDataArray<size_t>* _redirtied_cards;
 
   double _cur_collection_par_time_ms;
   double _cur_collection_code_root_fixup_time_ms;
@@ -135,9 +89,7 @@
   double _cur_evac_fail_restore_remsets;
   double _cur_evac_fail_remove_self_forwards;
 
-  double                  _cur_string_dedup_fixup_time_ms;
-  WorkerDataArray<double> _cur_string_dedup_queue_fixup_worker_times_ms;
-  WorkerDataArray<double> _cur_string_dedup_table_fixup_worker_times_ms;
+  double _cur_string_dedup_fixup_time_ms;
 
   double _cur_clear_ct_time_ms;
   double _cur_ref_proc_time_ms;
@@ -149,14 +101,13 @@
   double _recorded_young_cset_choice_time_ms;
   double _recorded_non_young_cset_choice_time_ms;
 
-  WorkerDataArray<double> _last_redirty_logged_cards_time_ms;
-  WorkerDataArray<size_t> _last_redirty_logged_cards_processed_cards;
   double _recorded_redirty_logged_cards_time_ms;
 
   double _recorded_young_free_cset_time_ms;
   double _recorded_non_young_free_cset_time_ms;
 
   double _cur_fast_reclaim_humongous_time_ms;
+  double _cur_fast_reclaim_humongous_register_time_ms;
   size_t _cur_fast_reclaim_humongous_total;
   size_t _cur_fast_reclaim_humongous_candidates;
   size_t _cur_fast_reclaim_humongous_reclaimed;
@@ -171,54 +122,34 @@
 
  public:
   G1GCPhaseTimes(uint max_gc_threads);
-  void note_gc_start(uint active_gc_threads);
+  void note_gc_start(uint active_gc_threads, bool mark_in_progress);
   void note_gc_end();
   void print(double pause_time_sec);
 
-  void record_gc_worker_start_time(uint worker_i, double ms) {
-    _last_gc_worker_start_times_ms.set(worker_i, ms);
-  }
+  // record the time a phase took in seconds
+  void record_time_secs(GCParPhases phase, uint worker_i, double secs);
 
-  void record_ext_root_scan_time(uint worker_i, double ms) {
-    _last_ext_root_scan_times_ms.set(worker_i, ms);
-  }
+  // add a number of seconds to a phase
+  void add_time_secs(GCParPhases phase, uint worker_i, double secs);
 
-  void record_satb_filtering_time(uint worker_i, double ms) {
-    _last_satb_filtering_times_ms.set(worker_i, ms);
-  }
+  void record_thread_work_item(GCParPhases phase, uint worker_i, size_t count);
 
-  void record_update_rs_time(uint worker_i, double ms) {
-    _last_update_rs_times_ms.set(worker_i, ms);
-  }
+  // return the average time for a phase in milliseconds
+  double average_time_ms(GCParPhases phase);
 
-  void record_update_rs_processed_buffers(uint worker_i, int processed_buffers) {
-    _last_update_rs_processed_buffers.set(worker_i, processed_buffers);
-  }
+  size_t sum_thread_work_items(GCParPhases phase);
 
-  void record_scan_rs_time(uint worker_i, double ms) {
-    _last_scan_rs_times_ms.set(worker_i, ms);
-  }
+ private:
+  double get_time_ms(GCParPhases phase, uint worker_i);
+  double sum_time_ms(GCParPhases phase);
+  double min_time_ms(GCParPhases phase);
+  double max_time_ms(GCParPhases phase);
+  size_t get_thread_work_item(GCParPhases phase, uint worker_i);
+  double average_thread_work_items(GCParPhases phase);
+  size_t min_thread_work_items(GCParPhases phase);
+  size_t max_thread_work_items(GCParPhases phase);
 
-  void record_strong_code_root_scan_time(uint worker_i, double ms) {
-    _last_strong_code_root_scan_times_ms.set(worker_i, ms);
-  }
-
-  void record_obj_copy_time(uint worker_i, double ms) {
-    _last_obj_copy_times_ms.set(worker_i, ms);
-  }
-
-  void add_obj_copy_time(uint worker_i, double ms) {
-    _last_obj_copy_times_ms.add(worker_i, ms);
-  }
-
-  void record_termination(uint worker_i, double ms, size_t attempts) {
-    _last_termination_times_ms.set(worker_i, ms);
-    _last_termination_attempts.set(worker_i, attempts);
-  }
-
-  void record_gc_worker_end_time(uint worker_i, double ms) {
-    _last_gc_worker_end_times_ms.set(worker_i, ms);
-  }
+ public:
 
   void record_clear_ct_time(double ms) {
     _cur_clear_ct_time_ms = ms;
@@ -248,21 +179,10 @@
     _cur_evac_fail_remove_self_forwards = ms;
   }
 
-  void note_string_dedup_fixup_start();
-  void note_string_dedup_fixup_end();
-
   void record_string_dedup_fixup_time(double ms) {
     _cur_string_dedup_fixup_time_ms = ms;
   }
 
-  void record_string_dedup_queue_fixup_worker_time(uint worker_id, double ms) {
-    _cur_string_dedup_queue_fixup_worker_times_ms.set(worker_id, ms);
-  }
-
-  void record_string_dedup_table_fixup_worker_time(uint worker_id, double ms) {
-    _cur_string_dedup_table_fixup_worker_times_ms.set(worker_id, ms);
-  }
-
   void record_ref_proc_time(double ms) {
     _cur_ref_proc_time_ms = ms;
   }
@@ -283,7 +203,8 @@
     _recorded_non_young_free_cset_time_ms = time_ms;
   }
 
-  void record_fast_reclaim_humongous_stats(size_t total, size_t candidates) {
+  void record_fast_reclaim_humongous_stats(double time_ms, size_t total, size_t candidates) {
+    _cur_fast_reclaim_humongous_register_time_ms = time_ms;
     _cur_fast_reclaim_humongous_total = total;
     _cur_fast_reclaim_humongous_candidates = candidates;
   }
@@ -301,14 +222,6 @@
     _recorded_non_young_cset_choice_time_ms = time_ms;
   }
 
-  void record_redirty_logged_cards_time_ms(uint worker_i, double time_ms) {
-    _last_redirty_logged_cards_time_ms.set(worker_i, time_ms);
-  }
-
-  void record_redirty_logged_cards_processed_cards(uint worker_i, size_t processed_buffers) {
-    _last_redirty_logged_cards_processed_cards.set(worker_i, processed_buffers);
-  }
-
   void record_redirty_logged_cards_time_ms(double time_ms) {
     _recorded_redirty_logged_cards_time_ms = time_ms;
   }
@@ -362,38 +275,16 @@
   double fast_reclaim_humongous_time_ms() {
     return _cur_fast_reclaim_humongous_time_ms;
   }
+};
 
-  double average_last_update_rs_time() {
-    return _last_update_rs_times_ms.average();
-  }
-
-  int sum_last_update_rs_processed_buffers() {
-    return _last_update_rs_processed_buffers.sum();
-  }
-
-  double average_last_scan_rs_time(){
-    return _last_scan_rs_times_ms.average();
-  }
-
-  double average_last_strong_code_root_scan_time(){
-    return _last_strong_code_root_scan_times_ms.average();
-  }
-
-  double average_last_obj_copy_time() {
-    return _last_obj_copy_times_ms.average();
-  }
-
-  double average_last_termination_time() {
-    return _last_termination_times_ms.average();
-  }
-
-  double average_last_ext_root_scan_time() {
-    return _last_ext_root_scan_times_ms.average();
-  }
-
-  double average_last_satb_filtering_times_ms() {
-    return _last_satb_filtering_times_ms.average();
-  }
+class G1GCParPhaseTimesTracker : public StackObj {
+  double _start_time;
+  G1GCPhaseTimes::GCParPhases _phase;
+  G1GCPhaseTimes* _phase_times;
+  uint _worker_id;
+public:
+  G1GCParPhaseTimesTracker(G1GCPhaseTimes* phase_times, G1GCPhaseTimes::GCParPhases phase, uint worker_id);
+  ~G1GCParPhaseTimesTracker();
 };
 
 #endif // SHARE_VM_GC_IMPLEMENTATION_G1_G1GCPHASETIMESLOG_HPP
--- a/src/share/vm/gc_implementation/g1/g1HotCardCache.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/gc_implementation/g1/g1HotCardCache.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2013, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -36,16 +36,13 @@
   if (default_use_cache()) {
     _use_cache = true;
 
-    _hot_cache_size = (1 << G1ConcRSLogCacheSize);
+    _hot_cache_size = (size_t)1 << G1ConcRSLogCacheSize;
     _hot_cache = NEW_C_HEAP_ARRAY(jbyte*, _hot_cache_size, mtGC);
 
-    _n_hot = 0;
-    _hot_cache_idx = 0;
+    reset_hot_cache_internal();
 
     // For refining the cards in the hot cache in parallel
-    uint n_workers = (ParallelGCThreads > 0 ?
-                        _g1h->workers()->total_workers() : 1);
-    _hot_cache_par_chunk_size = MAX2(1, _hot_cache_size / (int)n_workers);
+    _hot_cache_par_chunk_size = (int)(ParallelGCThreads > 0 ? ClaimChunkSize : _hot_cache_size);
     _hot_cache_par_claimed_idx = 0;
 
     _card_counts.initialize(card_counts_storage);
@@ -66,26 +63,21 @@
     // return it for immediate refining.
     return card_ptr;
   }
+  // Otherwise, the card is hot.
+  size_t index = Atomic::add_ptr((intptr_t)1, (volatile intptr_t*)&_hot_cache_idx) - 1;
+  size_t masked_index = index & (_hot_cache_size - 1);
+  jbyte* current_ptr = _hot_cache[masked_index];
 
-  // Otherwise, the card is hot.
-  jbyte* res = NULL;
-  MutexLockerEx x(HotCardCache_lock, Mutex::_no_safepoint_check_flag);
-  if (_n_hot == _hot_cache_size) {
-    res = _hot_cache[_hot_cache_idx];
-    _n_hot--;
-  }
-
-  // Now _n_hot < _hot_cache_size, and we can insert at _hot_cache_idx.
-  _hot_cache[_hot_cache_idx] = card_ptr;
-  _hot_cache_idx++;
-
-  if (_hot_cache_idx == _hot_cache_size) {
-    // Wrap around
-    _hot_cache_idx = 0;
-  }
-  _n_hot++;
-
-  return res;
+  // Try to store the new card pointer into the cache. Compare-and-swap to guard
+  // against the unlikely event of a race resulting in another card pointer to
+  // have already been written to the cache. In this case we will return
+  // card_ptr in favor of the other option, which would be starting over. This
+  // should be OK since card_ptr will likely be the older card already when/if
+  // this ever happens.
+  jbyte* previous_ptr = (jbyte*)Atomic::cmpxchg_ptr(card_ptr,
+                                                    &_hot_cache[masked_index],
+                                                    current_ptr);
+  return (previous_ptr == current_ptr) ? previous_ptr : card_ptr;
 }
 
 void G1HotCardCache::drain(uint worker_i,
@@ -98,38 +90,37 @@
 
   assert(_hot_cache != NULL, "Logic");
   assert(!use_cache(), "cache should be disabled");
-  int start_idx;
+  while (_hot_cache_par_claimed_idx < _hot_cache_size) {
+    size_t end_idx = Atomic::add_ptr((intptr_t)_hot_cache_par_chunk_size,
+                                     (volatile intptr_t*)&_hot_cache_par_claimed_idx);
+    size_t start_idx = end_idx - _hot_cache_par_chunk_size;
+    // The current worker has successfully claimed the chunk [start_idx..end_idx)
+    end_idx = MIN2(end_idx, _hot_cache_size);
+    for (size_t i = start_idx; i < end_idx; i++) {
+      jbyte* card_ptr = _hot_cache[i];
+      if (card_ptr != NULL) {
+        if (g1rs->refine_card(card_ptr, worker_i, true)) {
+          // The part of the heap spanned by the card contains references
+          // that point into the current collection set.
+          // We need to record the card pointer in the DirtyCardQueueSet
+          // that we use for such cards.
+          //
+          // The only time we care about recording cards that contain
+          // references that point into the collection set is during
+          // RSet updating while within an evacuation pause.
+          // In this case worker_i should be the id of a GC worker thread
+          assert(SafepointSynchronize::is_at_safepoint(), "Should be at a safepoint");
+          assert(worker_i < ParallelGCThreads,
+                 err_msg("incorrect worker id: %u", worker_i));
 
-  while ((start_idx = _hot_cache_par_claimed_idx) < _n_hot) { // read once
-    int end_idx = start_idx + _hot_cache_par_chunk_size;
-
-    if (start_idx ==
-        Atomic::cmpxchg(end_idx, &_hot_cache_par_claimed_idx, start_idx)) {
-      // The current worker has successfully claimed the chunk [start_idx..end_idx)
-      end_idx = MIN2(end_idx, _n_hot);
-      for (int i = start_idx; i < end_idx; i++) {
-        jbyte* card_ptr = _hot_cache[i];
-        if (card_ptr != NULL) {
-          if (g1rs->refine_card(card_ptr, worker_i, true)) {
-            // The part of the heap spanned by the card contains references
-            // that point into the current collection set.
-            // We need to record the card pointer in the DirtyCardQueueSet
-            // that we use for such cards.
-            //
-            // The only time we care about recording cards that contain
-            // references that point into the collection set is during
-            // RSet updating while within an evacuation pause.
-            // In this case worker_i should be the id of a GC worker thread
-            assert(SafepointSynchronize::is_at_safepoint(), "Should be at a safepoint");
-            assert(worker_i < (ParallelGCThreads == 0 ? 1 : ParallelGCThreads),
-                   err_msg("incorrect worker id: "UINT32_FORMAT, worker_i));
-
-            into_cset_dcq->enqueue(card_ptr);
-          }
+          into_cset_dcq->enqueue(card_ptr);
         }
+      } else {
+        break;
       }
     }
   }
+
   // The existing entries in the hot card cache, which were just refined
   // above, are discarded prior to re-enabling the cache near the end of the GC.
 }
--- a/src/share/vm/gc_implementation/g1/g1HotCardCache.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/gc_implementation/g1/g1HotCardCache.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2013, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -54,21 +54,33 @@
 // code, increasing throughput.
 
 class G1HotCardCache: public CHeapObj<mtGC> {
-  G1CollectedHeap*   _g1h;
+
+  G1CollectedHeap*  _g1h;
+
+  bool              _use_cache;
+
+  G1CardCounts      _card_counts;
 
   // The card cache table
-  jbyte**      _hot_cache;
+  jbyte**           _hot_cache;
 
-  int          _hot_cache_size;
-  int          _n_hot;
-  int          _hot_cache_idx;
+  size_t            _hot_cache_size;
 
-  int          _hot_cache_par_chunk_size;
-  volatile int _hot_cache_par_claimed_idx;
+  int               _hot_cache_par_chunk_size;
 
-  bool         _use_cache;
+  // Avoids false sharing when concurrently updating _hot_cache_idx or
+  // _hot_cache_par_claimed_idx. These are never updated at the same time
+  // thus it's not necessary to separate them as well
+  char _pad_before[DEFAULT_CACHE_LINE_SIZE];
 
-  G1CardCounts _card_counts;
+  volatile size_t _hot_cache_idx;
+
+  volatile size_t _hot_cache_par_claimed_idx;
+
+  char _pad_after[DEFAULT_CACHE_LINE_SIZE];
+
+  // The number of cached cards a thread claims when flushing the cache
+  static const int ClaimChunkSize = 32;
 
   bool default_use_cache() const {
     return (G1ConcRSLogCacheSize > 0);
@@ -110,16 +122,25 @@
   void reset_hot_cache() {
     assert(SafepointSynchronize::is_at_safepoint(), "Should be at a safepoint");
     assert(Thread::current()->is_VM_thread(), "Current thread should be the VMthread");
-    _hot_cache_idx = 0; _n_hot = 0;
+    if (default_use_cache()) {
+        reset_hot_cache_internal();
+    }
   }
 
-  bool hot_cache_is_empty() { return _n_hot == 0; }
-
   // Zeros the values in the card counts table for entire committed heap
   void reset_card_counts();
 
   // Zeros the values in the card counts table for the given region
   void reset_card_counts(HeapRegion* hr);
+
+ private:
+  void reset_hot_cache_internal() {
+    assert(_hot_cache != NULL, "Logic");
+    _hot_cache_idx = 0;
+    for (size_t i = 0; i < _hot_cache_size; i++) {
+      _hot_cache[i] = NULL;
+    }
+  }
 };
 
 #endif // SHARE_VM_GC_IMPLEMENTATION_G1_G1HOTCARDCACHE_HPP
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/src/share/vm/gc_implementation/g1/g1InCSetState.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -0,0 +1,132 @@
+/*
+ * Copyright (c) 2014, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+#ifndef SHARE_VM_GC_IMPLEMENTATION_G1_G1INCSETSTATE_HPP
+#define SHARE_VM_GC_IMPLEMENTATION_G1_G1INCSETSTATE_HPP
+
+#include "gc_implementation/g1/g1BiasedArray.hpp"
+#include "memory/allocation.hpp"
+
+// Per-region state during garbage collection.
+struct InCSetState {
+ public:
+  // We use different types to represent the state value. Particularly SPARC puts
+  // values in structs from "left to right", i.e. MSB to LSB. This results in many
+  // unnecessary shift operations when loading and storing values of this type.
+  // This degrades performance significantly (>10%) on that platform.
+  // Other tested ABIs do not seem to have this problem, and actually tend to
+  // favor smaller types, so we use the smallest usable type there.
+#ifdef SPARC
+  #define CSETSTATE_FORMAT INTPTR_FORMAT
+  typedef intptr_t in_cset_state_t;
+#else
+  #define CSETSTATE_FORMAT "%d"
+  typedef int8_t in_cset_state_t;
+#endif
+ private:
+  in_cset_state_t _value;
+ public:
+  enum {
+    // Selection of the values were driven to micro-optimize the encoding and
+    // frequency of the checks.
+    // The most common check is whether the region is in the collection set or not.
+    // This encoding allows us to use an != 0 check which in some architectures
+    // (x86*) can be encoded slightly more efficently than a normal comparison
+    // against zero.
+    // The same situation occurs when checking whether the region is humongous
+    // or not, which is encoded by values < 0.
+    // The other values are simply encoded in increasing generation order, which
+    // makes getting the next generation fast by a simple increment.
+    Humongous    = -1,    // The region is humongous - note that actually any value < 0 would be possible here.
+    NotInCSet    =  0,    // The region is not in the collection set.
+    Young        =  1,    // The region is in the collection set and a young region.
+    Old          =  2,    // The region is in the collection set and an old region.
+    Num
+  };
+
+  InCSetState(in_cset_state_t value = NotInCSet) : _value(value) {
+    assert(is_valid(), err_msg("Invalid state %d", _value));
+  }
+
+  in_cset_state_t value() const        { return _value; }
+
+  void set_old()                       { _value = Old; }
+
+  bool is_in_cset_or_humongous() const { return _value != NotInCSet; }
+  bool is_in_cset() const              { return _value > NotInCSet; }
+  bool is_humongous() const            { return _value < NotInCSet; }
+  bool is_young() const                { return _value == Young; }
+  bool is_old() const                  { return _value == Old; }
+
+#ifdef ASSERT
+  bool is_default() const              { return !is_in_cset_or_humongous(); }
+  bool is_valid() const                { return (_value >= Humongous) && (_value < Num); }
+  bool is_valid_gen() const            { return (_value >= Young && _value <= Old); }
+#endif
+};
+
+// Instances of this class are used for quick tests on whether a reference points
+// into the collection set and into which generation or is a humongous object
+//
+// Each of the array's elements indicates whether the corresponding region is in
+// the collection set and if so in which generation, or a humongous region.
+//
+// We use this to speed up reference processing during young collection and
+// quickly reclaim humongous objects. For the latter, by making a humongous region
+// succeed this test, we sort-of add it to the collection set. During the reference
+// iteration closures, when we see a humongous region, we then simply mark it as
+// referenced, i.e. live.
+class G1InCSetStateFastTestBiasedMappedArray : public G1BiasedMappedArray<InCSetState> {
+ protected:
+  InCSetState default_value() const { return InCSetState::NotInCSet; }
+ public:
+  void set_humongous(uintptr_t index) {
+    assert(get_by_index(index).is_default(),
+           err_msg("State at index " INTPTR_FORMAT" should be default but is " CSETSTATE_FORMAT, index, get_by_index(index).value()));
+    set_by_index(index, InCSetState::Humongous);
+  }
+
+  void clear_humongous(uintptr_t index) {
+    set_by_index(index, InCSetState::NotInCSet);
+  }
+
+  void set_in_young(uintptr_t index) {
+    assert(get_by_index(index).is_default(),
+           err_msg("State at index " INTPTR_FORMAT" should be default but is " CSETSTATE_FORMAT, index, get_by_index(index).value()));
+    set_by_index(index, InCSetState::Young);
+  }
+
+  void set_in_old(uintptr_t index) {
+    assert(get_by_index(index).is_default(),
+           err_msg("State at index " INTPTR_FORMAT" should be default but is " CSETSTATE_FORMAT, index, get_by_index(index).value()));
+    set_by_index(index, InCSetState::Old);
+  }
+
+  bool is_in_cset_or_humongous(HeapWord* addr) const { return at(addr).is_in_cset_or_humongous(); }
+  bool is_in_cset(HeapWord* addr) const { return at(addr).is_in_cset(); }
+  InCSetState at(HeapWord* addr) const { return get_by_address(addr); }
+  void clear() { G1BiasedMappedArray<InCSetState>::clear(); }
+};
+
+#endif // SHARE_VM_GC_IMPLEMENTATION_G1_G1INCSETSTATE_HPP
--- a/src/share/vm/gc_implementation/g1/g1Log.hpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/gc_implementation/g1/g1Log.hpp	Thu Jun 04 15:15:08 2015 -0700
@@ -28,6 +28,7 @@
 #include "memory/allocation.hpp"
 
 class G1Log : public AllStatic {
+ public:
   typedef enum {
     LevelNone,
     LevelFine,
@@ -35,6 +36,7 @@
     LevelFinest
   } LogLevel;
 
+ private:
   static LogLevel _level;
 
  public:
@@ -50,6 +52,10 @@
     return _level == LevelFinest;
   }
 
+  static LogLevel level() {
+    return _level;
+  }
+
   static void init();
 };
 
--- a/src/share/vm/gc_implementation/g1/g1MarkSweep.cpp	Thu May 07 18:13:43 2015 -0700
+++ b/src/share/vm/gc_implementation/g1/g1MarkSweep.cpp	Thu Jun 04 15:15:08 2015 -0700
@@ -31,6 +31,7 @@
 #include "code/icBuffer.hpp"
 #include "gc_implementation/g1/g1Log.hpp"
 #include "gc_implementation/g1/g1MarkSweep.hpp"
+#include "gc_implementation/g1/g1RootProcessor.hpp"
 #include "gc_implementation/g1/g1StringDedup.hpp"
 #include "gc_implementation/shared/gcHeapSummary.hpp"
 #include "gc_implementation/shared/gcTimer.hpp"
@@ -126,21 +127,22 @@
   GCTraceTime tm("phase 1", G1Log::fine() && Verbose, true, gc_timer(), gc_tracer()->gc_id());
   GenMarkSweep::trace(" 1");
 
-  SharedHeap* sh = SharedHeap::heap();
+  G1CollectedHeap* g1h = G1CollectedHeap::heap();
 
   // Need cleared claim bits for the roots processing
   ClassLoaderDataGraph::clear_claimed_marks();
 
   MarkingCodeBlobClosure follow_code_closure(&GenMarkSweep::follow_root_closure, !CodeBlobToOopClosure::FixRelocations);
-  sh->process_strong_roots(true,   // activate StrongRootsScope
-                           SharedHeap::SO_None,
-                           &GenMarkSweep::follow_root_closure,
-                           &GenMarkSweep::follow_cld_closure,
-                           &follow_code_closure);
+  {
+    G1RootProcessor root_processor(g1h);
+    root_processor.process_strong_roots(&GenMarkSweep::follow_root_closure,
+                                        &GenMarkSweep::follow_cld_closure,
+                                        &follow_code_closure);
+  }
 
   // Process reference objects found during marking
   ReferenceProcessor* rp = GenMarkSweep::ref_processor();
-  assert(rp == G1CollectedHeap::heap()->ref_processor_stw(), "Sanity");
+  assert(rp == g1h->ref_processor_stw(), "Sanity");
 
   rp->setup_policy(clear_all_softrefs);
   const ReferenceProcessorStats& stats =
@@ -226,6 +228,12 @@
   }
 };
 
+class G1AlwaysTrueClosure: public BoolObjectClosure {
+public:
+  bool do_object_b(oop p) { return true; }
+};
+static G1AlwaysTrueClosure always_true;
+
 void G1MarkSweep::mark_sweep_phase3() {
   G1CollectedHeap* g1h = G1CollectedHeap::heap();
 
@@ -233,24 +241,23 @@
   GCTraceTime tm("phase 3", G1Log::fine() && Verbose, true, gc_timer(), gc_tracer()->gc_id());
   GenMarkSweep::trace("3");
 
-  SharedHeap* sh = SharedHeap::heap();
-
   // Need cleared claim bits for the roots processing
   ClassLoaderDataGraph::clear_claimed_marks();
 
   CodeBlobToOopClosure adjust_code_closure(&GenMarkSweep::adjust_pointer_closure, CodeBlobToOopClosure::FixRelocations);
-  sh->process_all_roots(true,  // activate StrongRootsScope
-                        SharedHeap::SO_AllCodeCache,
-                        &GenMarkSweep::adjust_pointer_closure,
-                        &GenMarkSweep::adjust_cld_closure,
-                        &adjust_code_closure);
+  {
+    G1RootProcessor root_processor(g1h);
+    root_processor.process_all_roots(&GenMarkSweep::adjust_pointer_closure,
+                                     &GenMarkSweep::adjust_cld_closure,
+                                     &adjust_code_closure);
+  }
 
   assert(GenMarkSweep::ref_processor() == g1h->ref_processor_stw(), "Sanity");
   g1h->ref_processor_stw()->weak_oops_do(&GenMarkSweep::adjust_pointer_closure);
 
   // Now adjust pointers in remaining weak roots.  (All of which should
   // have been cleared if they pointed to non-surviving objects.)
-  sh->process_weak_roots(&GenMarkSweep::adjust_pointer_closure);
+  JNIHandles::weak_oops_do(&always_true, &GenMarkSweep::adjust_pointer_closure);