Merge

.hgtags

@@ -234,3 +234,4 @@ af9a674e12a16da1a4bd53e4990ddb1121a21ef1 jdk8-b109
 b5d2bf482a3ea1cca08c994512804ffbc73de0a1 jdk8-b110
 b9a0f6c693f347a6f4b9bb994957f4eaa05bdedd jdk8-b111
 ad67c34f79c28a8e755f4a49f313868619d6702c jdk8-b112
+4a4dbcf7cb7d3e1a81beaa3b11cd909f69ebc79a jdk8-b113

.hgtags-top-repo

@@ -234,3 +234,4 @@ b7e64be81c8a7690703df5711f4fc2375da8a9cb jdk8-b103
 4faa09c7fe555de086dd9048d3c5cc92317d6f45 jdk8-b110
 d086227bfc45d124f09b3bd72a07956b4073bf71 jdk8-b111
 547316ea137d83d9c63083a9b83db64198fe0c81 jdk8-b112
+6ba4c7cb623ec612031e05cf8bf279d8f407bd1e jdk8-b113

common/autoconf/generated-configure.sh

@@ -869,6 +869,7 @@ SRC_ROOT
 ZERO_ARCHDEF
 DEFINE_CROSS_COMPILE_ARCH
 LP64
+OPENJDK_TARGET_OS_EXPORT_DIR
 OPENJDK_TARGET_OS_API_DIR
 OPENJDK_TARGET_CPU_JLI_CFLAGS
 OPENJDK_TARGET_CPU_OSARCH
@@ -3864,7 +3865,7 @@ fi
 #CUSTOM_AUTOCONF_INCLUDE
 
 # Do not change or remove the following line, it is needed for consistency checks:
-DATE_WHEN_GENERATED=1382540536
+DATE_WHEN_GENERATED=1382702260
 
 ###############################################################################
 #
@@ -7149,6 +7150,13 @@ $as_echo "$COMPILE_TYPE" >&6; }
   fi
 
 
+  if test "x$OPENJDK_TARGET_OS" = xmacosx; then
+      OPENJDK_TARGET_OS_EXPORT_DIR=macosx
+  else
+      OPENJDK_TARGET_OS_EXPORT_DIR=${OPENJDK_TARGET_OS_API_DIR}
+  fi
+
+
   if test "x$OPENJDK_TARGET_CPU_BITS" = x64; then
     A_LP64="LP64:="
     # -D_LP64=1 is only set on linux and mac. Setting on windows causes diff in
@@ -29638,7 +29646,7 @@ fi
       -I${JDK_OUTPUTDIR}/include \
       -I${JDK_OUTPUTDIR}/include/$OPENJDK_TARGET_OS \
       -I${JDK_TOPDIR}/src/share/javavm/export \
-      -I${JDK_TOPDIR}/src/$OPENJDK_TARGET_OS_API_DIR/javavm/export \
+      -I${JDK_TOPDIR}/src/$OPENJDK_TARGET_OS_EXPORT_DIR/javavm/export \
       -I${JDK_TOPDIR}/src/share/native/common \
       -I${JDK_TOPDIR}/src/$OPENJDK_TARGET_OS_API_DIR/native/common"
 
@@ -34245,10 +34253,10 @@ ac_link='$CXX -o conftest$ac_exeext $CXXFLAGS $CPPFLAGS $LDFLAGS conftest.$ac_ex
 ac_compiler_gnu=$ac_cv_cxx_compiler_gnu
 
     PREV_CXXCFLAGS="$CXXFLAGS"
-    PREV_LDFLAGS="$LDFLAGS"
+    PREV_LIBS="$LIBS"
     PREV_CXX="$CXX"
     CXXFLAGS="$CXXFLAGS $FREETYPE_CFLAGS"
-    LDFLAGS="$LDFLAGS $FREETYPE_LIBS"
+    LIBS="$LIBS $FREETYPE_LIBS"
     CXX="$FIXPATH $CXX"
     cat confdefs.h - <<_ACEOF >conftest.$ac_ext
 /* end confdefs.h.  */
@@ -34315,7 +34323,7 @@ fi
 rm -f core conftest.err conftest.$ac_objext \
     conftest$ac_exeext conftest.$ac_ext
     CXXCFLAGS="$PREV_CXXFLAGS"
-    LDFLAGS="$PREV_LDFLAGS"
+    LIBS="$PREV_LIBS"
     CXX="$PREV_CXX"
     ac_ext=cpp
 ac_cpp='$CXXCPP $CPPFLAGS'

common/autoconf/libraries.m4

@@ -481,10 +481,10 @@ AC_DEFUN_ONCE([LIB_SETUP_FREETYPE],
     AC_MSG_CHECKING([if we can compile and link with freetype])
     AC_LANG_PUSH(C++)
     PREV_CXXCFLAGS="$CXXFLAGS"
-    PREV_LDFLAGS="$LDFLAGS"
+    PREV_LIBS="$LIBS"
     PREV_CXX="$CXX"
     CXXFLAGS="$CXXFLAGS $FREETYPE_CFLAGS" 
-    LDFLAGS="$LDFLAGS $FREETYPE_LIBS"
+    LIBS="$LIBS $FREETYPE_LIBS"
     CXX="$FIXPATH $CXX"
     AC_LINK_IFELSE([AC_LANG_SOURCE([[
           #include<ft2build.h>
@@ -508,7 +508,7 @@ AC_DEFUN_ONCE([LIB_SETUP_FREETYPE],
         ]
     )
     CXXCFLAGS="$PREV_CXXFLAGS"
-    LDFLAGS="$PREV_LDFLAGS"
+    LIBS="$PREV_LIBS"
     CXX="$PREV_CXX"
     AC_LANG_POP(C++)
 

common/autoconf/platform.m4

@@ -327,6 +327,13 @@ AC_DEFUN([PLATFORM_SETUP_LEGACY_VARS],
   fi
   AC_SUBST(OPENJDK_TARGET_OS_API_DIR)
 
+  if test "x$OPENJDK_TARGET_OS" = xmacosx; then
+      OPENJDK_TARGET_OS_EXPORT_DIR=macosx
+  else
+      OPENJDK_TARGET_OS_EXPORT_DIR=${OPENJDK_TARGET_OS_API_DIR}
+  fi
+  AC_SUBST(OPENJDK_TARGET_OS_EXPORT_DIR)
+
   if test "x$OPENJDK_TARGET_CPU_BITS" = x64; then
     A_LP64="LP64:="
     # -D_LP64=1 is only set on linux and mac. Setting on windows causes diff in

common/autoconf/spec.gmk.in

@@ -92,6 +92,7 @@ OPENJDK_TARGET_CPU_LEGACY_LIB:=@OPENJDK_TARGET_CPU_LEGACY_LIB@
 OPENJDK_TARGET_CPU_OSARCH:=@OPENJDK_TARGET_CPU_OSARCH@
 OPENJDK_TARGET_CPU_JLI_CFLAGS:=@OPENJDK_TARGET_CPU_JLI_CFLAGS@
 OPENJDK_TARGET_OS_API_DIR:=@OPENJDK_TARGET_OS_API_DIR@
+OPENJDK_TARGET_OS_EXPORT_DIR:=@OPENJDK_TARGET_OS_EXPORT_DIR@
 
 # We are building on this build system.
 # When not cross-compiling, it is the same as the target.

common/autoconf/toolchain.m4

@@ -942,7 +942,7 @@ AC_DEFUN_ONCE([TOOLCHAIN_SETUP_COMPILER_FLAGS_FOR_JDK],
       -I${JDK_OUTPUTDIR}/include \
       -I${JDK_OUTPUTDIR}/include/$OPENJDK_TARGET_OS \
       -I${JDK_TOPDIR}/src/share/javavm/export \
-      -I${JDK_TOPDIR}/src/$OPENJDK_TARGET_OS_API_DIR/javavm/export \
+      -I${JDK_TOPDIR}/src/$OPENJDK_TARGET_OS_EXPORT_DIR/javavm/export \
       -I${JDK_TOPDIR}/src/share/native/common \
       -I${JDK_TOPDIR}/src/$OPENJDK_TARGET_OS_API_DIR/native/common"
 

common/makefiles/JavaCompilation.gmk

@@ -506,30 +506,30 @@ define SetupJavaCompilation
 
     $$($1_BIN)/javac_state: $$($1_SRCS) $$($1_DEPENDS)
 	$(MKDIR) -p $$(@D)
-	$$(call ListPathsSafely,$1_SRCS,\n, >> $$($1_BIN)/_the.batch.tmp)
+	$$(call ListPathsSafely,$1_SRCS,\n, >> $$($1_BIN)/_the.$1_batch.tmp)
 	$(ECHO) Compiling $1
 	($$($1_JVM) $$($1_SJAVAC) \
 	    $$($1_REMOTE) \
 	    -j $(JOBS) \
 	    --permit-unidentified-artifacts \
 	    --permit-sources-without-package \
-	    --compare-found-sources $$($1_BIN)/_the.batch.tmp \
+	    --compare-found-sources $$($1_BIN)/_the.$1_batch.tmp \
 	    --log=$(LOG_LEVEL) \
 	    $$($1_SJAVAC_ARGS) \
 	    $$($1_FLAGS) \
 	    $$($1_HEADERS_ARG) \
 	    -d $$($1_BIN) && \
-	$(MV) $$($1_BIN)/_the.batch.tmp $$($1_BIN)/_the.batch)
+	$(MV) $$($1_BIN)/_the.$1_batch.tmp $$($1_BIN)/_the.$1_batch)
   else
     # Using plain javac to batch compile everything.
-    $1 := $$($1_ALL_COPY_TARGETS) $$($1_ALL_COPY_CLEAN_TARGETS) $$($1_BIN)/_the.batch
+    $1 := $$($1_ALL_COPY_TARGETS) $$($1_ALL_COPY_CLEAN_TARGETS) $$($1_BIN)/_the.$1_batch
 
     # When building in batch, put headers in a temp dir to filter out those that actually
     # changed before copying them to the real header dir.
     ifneq (,$$($1_HEADERS))
       $1_HEADERS_ARG := -h $$($1_HEADERS).tmp
 
-      $$($1_HEADERS)/_the.headers: $$($1_BIN)/_the.batch
+      $$($1_HEADERS)/_the.$1_headers: $$($1_BIN)/_the.$1_batch
 		$(MKDIR) -p $$(@D)
 		for f in `ls $$($1_HEADERS).tmp`; do \
 		  if [ ! -f "$$($1_HEADERS)/$$$$f" ] || [ "`$(DIFF) $$($1_HEADERS)/$$$$f $$($1_HEADERS).tmp/$$$$f`" != "" ]; then \
@@ -539,19 +539,19 @@ define SetupJavaCompilation
 		$(RM) -r $$($1_HEADERS).tmp
 		$(TOUCH) $$@
 
-      $1 += $$($1_HEADERS)/_the.headers
+      $1 += $$($1_HEADERS)/_the.$1_headers
     endif
 
     # When not using sjavac, pass along all sources to javac using an @file.
-    $$($1_BIN)/_the.batch: $$($1_SRCS) $$($1_DEPENDS)
+    $$($1_BIN)/_the.$1_batch: $$($1_SRCS) $$($1_DEPENDS)
 	$(MKDIR) -p $$(@D)
-	$(RM) $$($1_BIN)/_the.batch $$($1_BIN)/_the.batch.tmp
+	$(RM) $$($1_BIN)/_the.$1_batch $$($1_BIN)/_the.$1_batch.tmp
-	$$(call ListPathsSafely,$1_SRCS,\n, >> $$($1_BIN)/_the.batch.tmp)
+	$$(call ListPathsSafely,$1_SRCS,\n, >> $$($1_BIN)/_the.$1_batch.tmp)
-	$(ECHO) Compiling `$(WC) $$($1_BIN)/_the.batch.tmp | $(TR) -s ' ' | $(CUT) -f 2 -d ' '` files for $1
+	$(ECHO) Compiling `$(WC) $$($1_BIN)/_the.$1_batch.tmp | $(TR) -s ' ' | $(CUT) -f 2 -d ' '` files for $1
 	($$($1_JVM) $$($1_JAVAC) $$($1_FLAGS) \
 	    -implicit:none -sourcepath "$$($1_SRCROOTSC)" \
-	    -d $$($1_BIN) $$($1_HEADERS_ARG) @$$($1_BIN)/_the.batch.tmp && \
+	    -d $$($1_BIN) $$($1_HEADERS_ARG) @$$($1_BIN)/_the.$1_batch.tmp && \
-	$(MV) $$($1_BIN)/_the.batch.tmp $$($1_BIN)/_the.batch)
+	$(MV) $$($1_BIN)/_the.$1_batch.tmp $$($1_BIN)/_the.$1_batch)
 
   endif
 

corba/.hgtags

@@ -234,3 +234,4 @@ a4bb3b4500164748a9c33b2283cfda76d89f25ab jdk8-b108
 3d2b7ce93c5c2e3db748f29c3d29620a8b3b748a jdk8-b110
 85c1c94e723582f9a1dd0251502c42b73d6deea7 jdk8-b111
 43cec76d1d62587a07af07e2d9bec93aba2a506b jdk8-b112
+a259ff3e42d91da68f4d4f09d7eb9dc22bc024fc jdk8-b113

hotspot/.hgtags

@@ -387,3 +387,5 @@ f6962730bbde82f279a0ae3a1c14bc5e58096c6e jdk8-b111
 4a845c7a463844cead9e1e1641d6bcfb8a77f1c7 hs25-b54
 0ed9a90f45e1b392c671005f9ee22ce1acf02984 jdk8-b112
 23b8db5ea31d3079f1326afde4cd5c67b1dac49c hs25-b55
+4589b398ab03aba6a5da8c06ff53603488d1b8f4 jdk8-b113
+82a9cdbf683e374a76f2009352de53e16bed5a91 hs25-b56

hotspot/agent/src/os/linux/ps_core.c

@@ -736,26 +736,29 @@ static bool read_lib_segments(struct ps_prochandle* ph, int lib_fd, ELF_EHDR* li
 
       if (existing_map == NULL){
         if (add_map_info(ph, lib_fd, lib_php->p_offset,
-                          target_vaddr, lib_php->p_filesz) == NULL) {
+                          target_vaddr, lib_php->p_memsz) == NULL) {
           goto err;
         }
       } else {
+        // Coredump stores value of p_memsz elf field
+        // rounded up to page boundary.
+
         if ((existing_map->memsz != page_size) &&
             (existing_map->fd != lib_fd) &&
-            (existing_map->memsz != lib_php->p_filesz)){
+            (ROUNDUP(existing_map->memsz, page_size) != ROUNDUP(lib_php->p_memsz, page_size))) {
 
-          print_debug("address conflict @ 0x%lx (size = %ld, flags = %d\n)",
+          print_debug("address conflict @ 0x%lx (existing map size = %ld, size = %ld, flags = %d)\n",
-                        target_vaddr, lib_php->p_filesz, lib_php->p_flags);
+                        target_vaddr, existing_map->memsz, lib_php->p_memsz, lib_php->p_flags);
           goto err;
         }
 
         /* replace PT_LOAD segment with library segment */
         print_debug("overwrote with new address mapping (memsz %ld -> %ld)\n",
-                     existing_map->memsz, lib_php->p_filesz);
+                     existing_map->memsz, ROUNDUP(lib_php->p_memsz, page_size));
 
         existing_map->fd = lib_fd;
         existing_map->offset = lib_php->p_offset;
-        existing_map->memsz = lib_php->p_filesz;
+        existing_map->memsz = ROUNDUP(lib_php->p_memsz, page_size);
       }
     }
 

hotspot/agent/src/share/classes/sun/jvm/hotspot/oops/ConstMethod.java

@@ -51,6 +51,7 @@ public class ConstMethod extends VMObject {
   private static int HAS_GENERIC_SIGNATURE;
   private static int HAS_METHOD_ANNOTATIONS;
   private static int HAS_PARAMETER_ANNOTATIONS;
+  private static int HAS_METHOD_PARAMETERS;
   private static int HAS_DEFAULT_ANNOTATIONS;
   private static int HAS_TYPE_ANNOTATIONS;
 
@@ -70,6 +71,7 @@ public class ConstMethod extends VMObject {
     HAS_GENERIC_SIGNATURE     = db.lookupIntConstant("ConstMethod::_has_generic_signature").intValue();
     HAS_METHOD_ANNOTATIONS    = db.lookupIntConstant("ConstMethod::_has_method_annotations").intValue();
     HAS_PARAMETER_ANNOTATIONS = db.lookupIntConstant("ConstMethod::_has_parameter_annotations").intValue();
+    HAS_METHOD_PARAMETERS = db.lookupIntConstant("ConstMethod::_has_method_parameters").intValue();
     HAS_DEFAULT_ANNOTATIONS   = db.lookupIntConstant("ConstMethod::_has_default_annotations").intValue();
     HAS_TYPE_ANNOTATIONS      = db.lookupIntConstant("ConstMethod::_has_type_annotations").intValue();
 
@@ -85,6 +87,9 @@ public class ConstMethod extends VMObject {
     // start of byte code
     bytecodeOffset = type.getSize();
 
+    type                       = db.lookupType("MethodParametersElement");
+    methodParametersElementSize = type.getSize();
+
     type                       = db.lookupType("CheckedExceptionElement");
     checkedExceptionElementSize = type.getSize();
 
@@ -113,7 +118,7 @@ public class ConstMethod extends VMObject {
 
   // start of bytecode
   private static long bytecodeOffset;
-
+  private static long methodParametersElementSize;
   private static long checkedExceptionElementSize;
   private static long localVariableTableElementSize;
   private static long exceptionTableElementSize;
@@ -387,6 +392,10 @@ public class ConstMethod extends VMObject {
     return ret;
   }
 
+  private boolean hasMethodParameters() {
+    return (getFlags() & HAS_METHOD_PARAMETERS) != 0;
+  }
+
   private boolean hasGenericSignature() {
     return (getFlags() & HAS_GENERIC_SIGNATURE) != 0;
   }
@@ -442,11 +451,41 @@ public class ConstMethod extends VMObject {
     return offsetOfLastU2Element();
   }
 
-  private long offsetOfCheckedExceptionsLength() {
+  private long offsetOfMethodParametersLength() {
+    if (Assert.ASSERTS_ENABLED) {
+      Assert.that(hasMethodParameters(), "should only be called if table is present");
+    }
     return hasGenericSignature() ? offsetOfLastU2Element() - sizeofShort :
                                    offsetOfLastU2Element();
   }
 
+  private int getMethodParametersLength() {
+      if (hasMethodParameters())
+          return (int) getAddress().getCIntegerAt(offsetOfMethodParametersLength(), 2, true);
+      else
+          return 0;
+  }
+
+  // Offset of start of checked exceptions
+  private long offsetOfMethodParameters() {
+    long offset = offsetOfMethodParametersLength();
+    long length = getMethodParametersLength();
+    if (Assert.ASSERTS_ENABLED) {
+      Assert.that(length > 0, "should only be called if method parameter information is present");
+    }
+    offset -= length * methodParametersElementSize;
+    return offset;
+  }
+
+  private long offsetOfCheckedExceptionsLength() {
+    if (hasMethodParameters())
+      return offsetOfMethodParameters() - sizeofShort;
+    else {
+      return hasGenericSignature() ? offsetOfLastU2Element() - sizeofShort :
+                                     offsetOfLastU2Element();
+    }
+  }
+
   private int getCheckedExceptionsLength() {
     if (hasCheckedExceptions()) {
       return (int) getAddress().getCIntegerAt(offsetOfCheckedExceptionsLength(), 2, true);
@@ -496,6 +535,8 @@ public class ConstMethod extends VMObject {
       return offsetOfExceptionTable() - sizeofShort;
     } else if (hasCheckedExceptions()) {
       return offsetOfCheckedExceptions() - sizeofShort;
+    } else if (hasMethodParameters()) {
+      return offsetOfMethodParameters() - sizeofShort;
     } else {
       return hasGenericSignature() ? offsetOfLastU2Element() - sizeofShort :
                                      offsetOfLastU2Element();
@@ -526,6 +567,8 @@ public class ConstMethod extends VMObject {
     }
     if (hasCheckedExceptions()) {
       return offsetOfCheckedExceptions() - sizeofShort;
+    } else if (hasMethodParameters()) {
+      return offsetOfMethodParameters() - sizeofShort;
     } else {
       return hasGenericSignature() ? offsetOfLastU2Element() - sizeofShort :
                                      offsetOfLastU2Element();

hotspot/agent/src/share/classes/sun/jvm/hotspot/tools/ClassLoaderStats.java

@@ -51,8 +51,7 @@ public class ClassLoaderStats extends Tool {
 
    public static void main(String[] args) {
       ClassLoaderStats cls = new ClassLoaderStats();
-      cls.start(args);
+      cls.execute(args);
-      cls.stop();
    }
 
    private static class ClassData {

hotspot/agent/src/share/classes/sun/jvm/hotspot/tools/FinalizerInfo.java

@@ -54,8 +54,7 @@ public class FinalizerInfo extends Tool {
 
     public static void main(String[] args) {
         FinalizerInfo finfo = new FinalizerInfo();
-        finfo.start(args);
+        finfo.execute(args);
-        finfo.stop();
     }
 
     public void run() {

hotspot/agent/src/share/classes/sun/jvm/hotspot/tools/FlagDumper.java

@@ -54,7 +54,6 @@ public class FlagDumper extends Tool {
 
    public static void main(String[] args) {
       FlagDumper fd = new FlagDumper();
-      fd.start(args);
+      fd.execute(args);
-      fd.stop();
    }
 }

hotspot/agent/src/share/classes/sun/jvm/hotspot/tools/HeapDumper.java

@@ -80,8 +80,7 @@ public class HeapDumper extends Tool {
         }
 
         HeapDumper dumper = new HeapDumper(file);
-        dumper.start(args);
+        dumper.execute(args);
-        dumper.stop();
     }
 
 }

hotspot/agent/src/share/classes/sun/jvm/hotspot/tools/HeapSummary.java

@@ -46,8 +46,7 @@ public class HeapSummary extends Tool {
 
    public static void main(String[] args) {
       HeapSummary hs = new HeapSummary();
-      hs.start(args);
+      hs.execute(args);
-      hs.stop();
    }
 
    public void run() {

hotspot/agent/src/share/classes/sun/jvm/hotspot/tools/JInfo.java

@@ -134,8 +134,7 @@ public class JInfo extends Tool {
         }
 
         JInfo jinfo = new JInfo(mode);
-        jinfo.start(args);
+        jinfo.execute(args);
-        jinfo.stop();
     }
 
     private void printVMFlags() {

hotspot/agent/src/share/classes/sun/jvm/hotspot/tools/JMap.java

@@ -136,7 +136,9 @@ public class JMap extends Tool {
                         mode = MODE_HEAP_GRAPH_GXL;
                     } else {
                         System.err.println("unknown heap format:" + format);
-                        return;
+
+                        // Exit with error status
+                        System.exit(1);
                     }
                 } else {
                     copyArgs = false;
@@ -153,8 +155,7 @@ public class JMap extends Tool {
         }
 
         JMap jmap = new JMap(mode);
-        jmap.start(args);
+        jmap.execute(args);
-        jmap.stop();
     }
 
     public boolean writeHeapHprofBin(String fileName) {

hotspot/agent/src/share/classes/sun/jvm/hotspot/tools/JSnap.java

@@ -64,7 +64,6 @@ public class JSnap extends Tool {
 
     public static void main(String[] args) {
         JSnap js = new JSnap();
-        js.start(args);
+        js.execute(args);
-        js.stop();
     }
 }

hotspot/agent/src/share/classes/sun/jvm/hotspot/tools/JStack.java

@@ -89,8 +89,7 @@ public class JStack extends Tool {
         }
 
         JStack jstack = new JStack(mixedMode, concurrentLocks);
-        jstack.start(args);
+        jstack.execute(args);
-        jstack.stop();
     }
 
     private boolean mixedMode;

hotspot/agent/src/share/classes/sun/jvm/hotspot/tools/ObjectHistogram.java

@@ -61,7 +61,6 @@ public class ObjectHistogram extends Tool {
 
    public static void main(String[] args) {
       ObjectHistogram oh = new ObjectHistogram();
-      oh.start(args);
+      oh.execute(args);
-      oh.stop();
    }
 }

hotspot/agent/src/share/classes/sun/jvm/hotspot/tools/PMap.java

@@ -69,7 +69,6 @@ public class PMap extends Tool {
 
    public static void main(String[] args) throws Exception {
       PMap t = new PMap();
-      t.start(args);
+      t.execute(args);
-      t.stop();
    }
 }

hotspot/agent/src/share/classes/sun/jvm/hotspot/tools/PStack.java

@@ -182,8 +182,7 @@ public class PStack extends Tool {
 
    public static void main(String[] args) throws Exception {
       PStack t = new PStack();
-      t.start(args);
+      t.execute(args);
-      t.stop();
    }
 
    // -- Internals only below this point

hotspot/agent/src/share/classes/sun/jvm/hotspot/tools/StackTrace.java

@@ -137,8 +137,7 @@ public class StackTrace extends Tool {
 
    public static void main(String[] args) {
       StackTrace st = new StackTrace();
-      st.start(args);
+      st.execute(args);
-      st.stop();
    }
 
    private boolean verbose;

hotspot/agent/src/share/classes/sun/jvm/hotspot/tools/SysPropsDumper.java

@@ -58,7 +58,6 @@ public class SysPropsDumper extends Tool {
 
    public static void main(String[] args) {
       SysPropsDumper pd = new SysPropsDumper();
-      pd.start(args);
+      pd.execute(args);
-      pd.stop();
    }
 }

hotspot/agent/src/share/classes/sun/jvm/hotspot/tools/Tool.java

@@ -26,6 +26,7 @@ package sun.jvm.hotspot.tools;
 
 import java.io.PrintStream;
 import java.util.Hashtable;
+
 import sun.jvm.hotspot.*;
 import sun.jvm.hotspot.runtime.*;
 import sun.jvm.hotspot.debugger.*;
@@ -105,26 +106,44 @@ public abstract class Tool implements Runnable {
 
       public static void main(String[] args) {
          <derived class> obj = new <derived class>;
-         obj.start(args);
+         obj.execute(args);
       }
 
    */
 
-   protected void stop() {
+   protected void execute(String[] args) {
+       int returnStatus = 1;
+
+       try {
+           returnStatus = start(args);
+       } finally {
+           stop();
+       }
+
+       // Exit with 0 or 1
+       System.exit(returnStatus);
+   }
+
+   public void stop() {
       if (agent != null) {
          agent.detach();
       }
    }
 
-   protected void start(String[] args) {
+   private int start(String[] args) {
+
       if ((args.length < 1) || (args.length > 2)) {
          usage();
-         return;
+         return 1;
       }
 
       // Attempt to handle -h or -help or some invalid flag
-      if (args[0].startsWith("-")) {
+      if (args[0].startsWith("-h")) {
           usage();
+          return 0;
+      } else if (args[0].startsWith("-")) {
+          usage();
+          return 1;
       }
 
       PrintStream err = System.err;
@@ -154,6 +173,7 @@ public abstract class Tool implements Runnable {
 
         default:
            usage();
+           return 1;
       }
 
       agent = new HotSpotAgent();
@@ -191,15 +211,16 @@ public abstract class Tool implements Runnable {
              break;
         }
         if (e.getMessage() != null) {
-          err.print(e.getMessage());
+          err.println(e.getMessage());
           e.printStackTrace();
         }
         err.println();
-        return;
+        return 1;
       }
 
       err.println("Debugger attached successfully.");
       startInternal();
+      return 0;
    }
 
    // When using an existing JVMDebugger.

hotspot/agent/src/share/classes/sun/jvm/hotspot/tools/jcore/ClassDump.java

@@ -177,7 +177,6 @@ public class ClassDump extends Tool {
     public static void main(String[] args) {
 
         ClassDump cd = new ClassDump();
-        cd.start(args);
+        cd.execute(args);
-        cd.stop();
     }
 }

hotspot/agent/src/share/classes/sun/jvm/hotspot/tools/soql/JSDB.java

@@ -42,8 +42,7 @@ public class JSDB extends Tool {
 
     public static void main(String[] args) {
         JSDB jsdb = new JSDB();
-        jsdb.start(args);
+        jsdb.execute(args);
-        jsdb.stop();
     }
 
     public void run() {

hotspot/agent/src/share/classes/sun/jvm/hotspot/tools/soql/SOQL.java

@@ -40,8 +40,7 @@ import sun.jvm.hotspot.utilities.soql.*;
 public class SOQL extends Tool {
    public static void main(String[] args) {
       SOQL soql = new SOQL();
-      soql.start(args);
+      soql.execute(args);
-      soql.stop();
    }
 
    public SOQL() {

hotspot/make/hotspot_version

@@ -35,7 +35,7 @@ HOTSPOT_VM_COPYRIGHT=Copyright 2013
 
 HS_MAJOR_VER=25
 HS_MINOR_VER=0
-HS_BUILD_NUMBER=55
+HS_BUILD_NUMBER=56
 
 JDK_MAJOR_VER=1
 JDK_MINOR_VER=8

hotspot/make/jprt.properties

@@ -24,12 +24,7 @@
 
 # Properties for jprt
 
-# All build result bundles are full jdks, so the 64bit testing does not
+# All build result bundles are full jdks.
-#    need the 32bit sibling bundle installed.
-#    Note: If the hotspot/make/Makefile changed to only bundle the 64bit files
-#          when bundling 64bit, and stripped out the 64bit files from any 32bit
-#          bundles, then this setting would be need to be "true".
-
 jprt.need.sibling.build=false
 
 # At submit time, the release supplied will be in jprt.submit.release
@@ -52,21 +47,11 @@ jprt.sync.push=false
 #       sparc etc.
 
 # Define the Solaris platforms we want for the various releases
-jprt.my.solaris.sparc.jdk8=solaris_sparc_5.10
-jprt.my.solaris.sparc.jdk7=solaris_sparc_5.10
-jprt.my.solaris.sparc.jdk7u8=${jprt.my.solaris.sparc.jdk7}
-jprt.my.solaris.sparc=${jprt.my.solaris.sparc.${jprt.tools.default.release}}
-
 jprt.my.solaris.sparcv9.jdk8=solaris_sparcv9_5.10
 jprt.my.solaris.sparcv9.jdk7=solaris_sparcv9_5.10
 jprt.my.solaris.sparcv9.jdk7u8=${jprt.my.solaris.sparcv9.jdk7}
 jprt.my.solaris.sparcv9=${jprt.my.solaris.sparcv9.${jprt.tools.default.release}}
 
-jprt.my.solaris.i586.jdk8=solaris_i586_5.10
-jprt.my.solaris.i586.jdk7=solaris_i586_5.10
-jprt.my.solaris.i586.jdk7u8=${jprt.my.solaris.i586.jdk7}
-jprt.my.solaris.i586=${jprt.my.solaris.i586.${jprt.tools.default.release}}
-
 jprt.my.solaris.x64.jdk8=solaris_x64_5.10
 jprt.my.solaris.x64.jdk7=solaris_x64_5.10
 jprt.my.solaris.x64.jdk7u8=${jprt.my.solaris.x64.jdk7}
@@ -133,9 +118,7 @@ jprt.my.windows.x64=${jprt.my.windows.x64.${jprt.tools.default.release}}
 # Standard list of jprt build targets for this source tree
 
 jprt.build.targets.standard= \
-    ${jprt.my.solaris.sparc}-{product|fastdebug}, \
     ${jprt.my.solaris.sparcv9}-{product|fastdebug|optimized}, \
-    ${jprt.my.solaris.i586}-{product|fastdebug}, \
     ${jprt.my.solaris.x64}-{product|fastdebug}, \
     ${jprt.my.linux.i586}-{product|fastdebug}, \
     ${jprt.my.linux.x64}-{product|fastdebug|optimized}, \
@@ -145,7 +128,6 @@ jprt.build.targets.standard= \
     ${jprt.my.linux.armvh}-{product|fastdebug}
 
 jprt.build.targets.open= \
-    ${jprt.my.solaris.i586}-{productOpen}, \
     ${jprt.my.solaris.x64}-{debugOpen}, \
     ${jprt.my.linux.x64}-{productOpen}
 
@@ -168,31 +150,6 @@ jprt.build.targets=${jprt.build.targets.${jprt.tools.default.release}}
 
 # Subset lists of test targets for this source tree
 
-jprt.my.solaris.sparc.test.targets= \
-    ${jprt.my.solaris.sparc}-{product|fastdebug}-{c1|c2}-jvm98, \
-    ${jprt.my.solaris.sparc}-{product|fastdebug}-c2-jvm98_nontiered, \
-    ${jprt.my.solaris.sparc}-{product|fastdebug}-{c1|c2}-scimark, \
-    ${jprt.my.solaris.sparc}-product-{c1|c2}-runThese, \
-    ${jprt.my.solaris.sparc}-fastdebug-c1-runThese_Xshare, \
-    ${jprt.my.solaris.sparc}-{product|fastdebug}-{c1|c2}-GCBasher_SerialGC, \
-    ${jprt.my.solaris.sparc}-{product|fastdebug}-{c1|c2}-GCBasher_ParallelGC, \
-    ${jprt.my.solaris.sparc}-{product|fastdebug}-{c1|c2}-GCBasher_ParNewGC, \
-    ${jprt.my.solaris.sparc}-{product|fastdebug}-{c1|c2}-GCBasher_CMS, \
-    ${jprt.my.solaris.sparc}-{product|fastdebug}-{c1|c2}-GCBasher_G1, \
-    ${jprt.my.solaris.sparc}-{product|fastdebug}-{c1|c2}-GCBasher_ParOldGC, \
-    ${jprt.my.solaris.sparc}-{product|fastdebug}-{c1|c2}-GCOld_SerialGC, \
-    ${jprt.my.solaris.sparc}-{product|fastdebug}-{c1|c2}-GCOld_ParallelGC, \
-    ${jprt.my.solaris.sparc}-{product|fastdebug}-{c1|c2}-GCOld_ParNewGC, \
-    ${jprt.my.solaris.sparc}-{product|fastdebug}-{c1|c2}-GCOld_CMS, \
-    ${jprt.my.solaris.sparc}-{product|fastdebug}-{c1|c2}-GCOld_G1, \
-    ${jprt.my.solaris.sparc}-{product|fastdebug}-{c1|c2}-GCOld_ParOldGC, \
-    ${jprt.my.solaris.sparc}-{product|fastdebug}-c2-jbb_default_nontiered, \
-    ${jprt.my.solaris.sparc}-{product|fastdebug}-{c1|c2}-jbb_SerialGC, \
-    ${jprt.my.solaris.sparc}-{product|fastdebug}-{c1|c2}-jbb_ParallelGC, \
-    ${jprt.my.solaris.sparc}-{product|fastdebug}-{c1|c2}-jbb_CMS, \
-    ${jprt.my.solaris.sparc}-{product|fastdebug}-{c1|c2}-jbb_G1, \
-    ${jprt.my.solaris.sparc}-{product|fastdebug}-{c1|c2}-jbb_ParOldGC
-
 jprt.my.solaris.sparcv9.test.targets= \
     ${jprt.my.solaris.sparcv9}-{product|fastdebug}-c2-jvm98, \
     ${jprt.my.solaris.sparcv9}-{product|fastdebug}-c2-jvm98_nontiered, \
@@ -242,37 +199,6 @@ jprt.my.solaris.x64.test.targets= \
     ${jprt.my.solaris.x64}-{product|fastdebug}-c2-GCOld_G1, \
     ${jprt.my.solaris.x64}-{product|fastdebug}-c2-GCOld_ParOldGC
 
-jprt.my.solaris.i586.test.targets= \
-    ${jprt.my.solaris.i586}-{product|fastdebug}-{c1|c2}-jvm98, \
-    ${jprt.my.solaris.i586}-{product|fastdebug}-c2-jvm98_nontiered, \
-    ${jprt.my.solaris.i586}-{product|fastdebug}-{c1|c2}-scimark, \
-    ${jprt.my.solaris.i586}-product-{c1|c2}-runThese_Xcomp, \
-    ${jprt.my.solaris.i586}-fastdebug-c1-runThese_Xcomp, \
-    ${jprt.my.solaris.i586}-fastdebug-c1-runThese_Xshare, \
-    ${jprt.my.solaris.i586}-product-c1-GCBasher_SerialGC, \
-    ${jprt.my.solaris.i586}-product-c1-GCBasher_ParallelGC, \
-    ${jprt.my.solaris.i586}-product-c1-GCBasher_ParNewGC, \
-    ${jprt.my.solaris.i586}-product-c1-GCBasher_CMS, \
-    ${jprt.my.solaris.i586}-product-c1-GCBasher_G1, \
-    ${jprt.my.solaris.i586}-product-c1-GCBasher_ParOldGC, \
-    ${jprt.my.solaris.i586}-fastdebug-c2-GCBasher_SerialGC, \
-    ${jprt.my.solaris.i586}-fastdebug-c2-GCBasher_ParallelGC, \
-    ${jprt.my.solaris.i586}-fastdebug-c2-GCBasher_ParNewGC, \
-    ${jprt.my.solaris.i586}-fastdebug-c2-GCBasher_CMS, \
-    ${jprt.my.solaris.i586}-fastdebug-c2-GCBasher_G1, \
-    ${jprt.my.solaris.i586}-fastdebug-c2-GCBasher_ParOldGC, \
-    ${jprt.my.solaris.i586}-product-c1-GCOld_SerialGC, \
-    ${jprt.my.solaris.i586}-product-c1-GCOld_ParallelGC, \
-    ${jprt.my.solaris.i586}-product-c1-GCOld_ParNewGC, \
-    ${jprt.my.solaris.i586}-product-c1-GCOld_CMS, \
-    ${jprt.my.solaris.i586}-product-c1-GCOld_G1, \
-    ${jprt.my.solaris.i586}-product-c1-GCOld_ParOldGC, \
-    ${jprt.my.solaris.i586}-fastdebug-c2-jbb_default_nontiered, \
-    ${jprt.my.solaris.i586}-fastdebug-c2-jbb_ParallelGC, \
-    ${jprt.my.solaris.i586}-fastdebug-c2-jbb_CMS, \
-    ${jprt.my.solaris.i586}-fastdebug-c2-jbb_G1, \
-    ${jprt.my.solaris.i586}-fastdebug-c2-jbb_ParOldGC
-
 jprt.my.linux.i586.test.targets = \
     ${jprt.my.linux.i586}-{product|fastdebug}-{c1|c2}-jvm98, \
     ${jprt.my.linux.i586}-{product|fastdebug}-c2-jvm98_nontiered, \
@@ -395,7 +321,6 @@ jprt.my.windows.x64.test.targets = \
 # Some basic "smoke" tests for OpenJDK builds
 jprt.test.targets.open = \
     ${jprt.my.solaris.x64}-{productOpen|fastdebugOpen}-c2-jvm98, \
-    ${jprt.my.solaris.i586}-{productOpen|fastdebugOpen}-c2-jvm98, \
     ${jprt.my.linux.x64}-{productOpen|fastdebugOpen}-c2-jvm98
 
 # Testing for actual embedded builds is different to standard
@@ -407,9 +332,7 @@ jprt.my.linux.i586.test.targets.embedded = \
 
 jprt.test.targets.standard = \
   ${jprt.my.linux.i586.test.targets.embedded}, \
-  ${jprt.my.solaris.sparc.test.targets}, \
   ${jprt.my.solaris.sparcv9.test.targets}, \
-  ${jprt.my.solaris.i586.test.targets}, \
   ${jprt.my.solaris.x64.test.targets}, \
   ${jprt.my.linux.i586.test.targets}, \
   ${jprt.my.linux.x64.test.targets}, \
@@ -420,15 +343,12 @@ jprt.test.targets.standard = \
 
 jprt.test.targets.embedded= 		\
   ${jprt.my.linux.i586.test.targets.embedded}, \
-  ${jprt.my.solaris.sparc.test.targets}, \
   ${jprt.my.solaris.sparcv9.test.targets}, \
-  ${jprt.my.solaris.i586.test.targets}, \
   ${jprt.my.solaris.x64.test.targets}, \
   ${jprt.my.linux.x64.test.targets}, \
   ${jprt.my.windows.i586.test.targets}, \
   ${jprt.my.windows.x64.test.targets}
 
-
 jprt.test.targets.jdk8=${jprt.test.targets.standard}
 jprt.test.targets.jdk7=${jprt.test.targets.standard}
 jprt.test.targets.jdk7u8=${jprt.test.targets.jdk7}
@@ -439,15 +359,11 @@ jprt.test.targets=${jprt.test.targets.${jprt.tools.default.release}}
 #jprt.make.rule.test.targets=*-product-*-packtest
 
 jprt.make.rule.test.targets.standard.client = \
-  ${jprt.my.solaris.sparc}-*-c1-clienttest, \
-  ${jprt.my.solaris.i586}-*-c1-clienttest, \
   ${jprt.my.linux.i586}-*-c1-clienttest, \
   ${jprt.my.windows.i586}-*-c1-clienttest
 
 jprt.make.rule.test.targets.standard.server = \
-  ${jprt.my.solaris.sparc}-*-c2-servertest, \
   ${jprt.my.solaris.sparcv9}-*-c2-servertest, \
-  ${jprt.my.solaris.i586}-*-c2-servertest, \
   ${jprt.my.solaris.x64}-*-c2-servertest, \
   ${jprt.my.linux.i586}-*-c2-servertest, \
   ${jprt.my.linux.x64}-*-c2-servertest, \
@@ -456,9 +372,7 @@ jprt.make.rule.test.targets.standard.server = \
   ${jprt.my.windows.x64}-*-c2-servertest
 
 jprt.make.rule.test.targets.standard.internalvmtests = \
-  ${jprt.my.solaris.sparc}-fastdebug-c2-internalvmtests, \
   ${jprt.my.solaris.sparcv9}-fastdebug-c2-internalvmtests, \
-  ${jprt.my.solaris.i586}-fastdebug-c2-internalvmtests, \
   ${jprt.my.solaris.x64}-fastdebug-c2-internalvmtests, \
   ${jprt.my.linux.i586}-fastdebug-c2-internalvmtests, \
   ${jprt.my.linux.x64}-fastdebug-c2-internalvmtests, \
@@ -467,16 +381,12 @@ jprt.make.rule.test.targets.standard.internalvmtests = \
   ${jprt.my.windows.x64}-fastdebug-c2-internalvmtests
 
 jprt.make.rule.test.targets.standard.wbapi = \
-  ${jprt.my.solaris.sparc}-{product|fastdebug}-c2-wbapitest, \
   ${jprt.my.solaris.sparcv9}-{product|fastdebug}-c2-wbapitest, \
-  ${jprt.my.solaris.i586}-{product|fastdebug}-c2-wbapitest, \
   ${jprt.my.solaris.x64}-{product|fastdebug}-c2-wbapitest, \
   ${jprt.my.linux.i586}-{product|fastdebug}-c2-wbapitest, \
   ${jprt.my.linux.x64}-{product|fastdebug}-c2-wbapitest, \
   ${jprt.my.windows.i586}-{product|fastdebug}-c2-wbapitest, \
   ${jprt.my.windows.x64}-{product|fastdebug}-c2-wbapitest, \
-  ${jprt.my.solaris.sparc}-{product|fastdebug}-c1-wbapitest, \
-  ${jprt.my.solaris.i586}-{product|fastdebug}-c1-wbapitest, \
   ${jprt.my.linux.i586}-{product|fastdebug}-c1-wbapitest, \
   ${jprt.my.windows.i586}-{product|fastdebug}-c1-wbapitest
 

hotspot/make/windows/makefiles/trace.make

@@ -40,8 +40,7 @@ TraceGeneratedNames =     \
     traceEventIds.hpp     \
     traceTypes.hpp
 
-
+!if EXISTS($(TraceAltSrcDir))
-!if "$(OPENJDK)" != "true"
 TraceGeneratedNames = $(TraceGeneratedNames) \
     traceRequestables.hpp \
     traceEventControl.hpp \
@@ -56,7 +55,7 @@ TraceGeneratedFiles = \
 	$(TraceOutDir)/traceEventIds.hpp     \
 	$(TraceOutDir)/traceTypes.hpp
 
-!if "$(OPENJDK)" != "true"
+!if EXISTS($(TraceAltSrcDir))
 TraceGeneratedFiles = $(TraceGeneratedFiles) \
 	$(TraceOutDir)/traceRequestables.hpp \
     $(TraceOutDir)/traceEventControl.hpp \
@@ -68,7 +67,7 @@ XSLT = $(QUIETLY) $(REMOTE) $(RUN_JAVA) -classpath $(JvmtiOutDir) jvmtiGen
 XML_DEPS = $(TraceSrcDir)/trace.xml $(TraceSrcDir)/tracetypes.xml \
     $(TraceSrcDir)/trace.dtd $(TraceSrcDir)/xinclude.mod
 
-!if "$(OPENJDK)" != "true"
+!if EXISTS($(TraceAltSrcDir))
 XML_DEPS = $(XML_DEPS) $(TraceAltSrcDir)/traceevents.xml
 !endif
 
@@ -87,7 +86,7 @@ $(TraceOutDir)/traceTypes.hpp: $(TraceSrcDir)/trace.xml $(TraceSrcDir)/traceType
 	@echo Generating $@
 	@$(XSLT) -IN $(TraceSrcDir)/trace.xml -XSL $(TraceSrcDir)/traceTypes.xsl -OUT $(TraceOutDir)/traceTypes.hpp
 
-!if "$(OPENJDK)" == "true"
+!if !EXISTS($(TraceAltSrcDir))
 
 $(TraceOutDir)/traceEventClasses.hpp: $(TraceSrcDir)/trace.xml $(TraceSrcDir)/traceEventClasses.xsl $(XML_DEPS)
 	@echo Generating OpenJDK $@

hotspot/src/cpu/sparc/vm/c1_FrameMap_sparc.cpp

@@ -53,6 +53,8 @@ LIR_Opr FrameMap::map_to_opr(BasicType type, VMRegPair* reg, bool outgoing) {
       opr = as_long_opr(reg);
     } else if (type == T_OBJECT || type == T_ARRAY) {
       opr = as_oop_opr(reg);
+    } else if (type == T_METADATA) {
+      opr = as_metadata_opr(reg);
     } else {
       opr = as_opr(reg);
     }

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

@@ -2565,7 +2565,7 @@ void LIR_Assembler::type_profile_helper(Register mdo, int mdo_offset_bias,
     Address receiver_addr(mdo, md->byte_offset_of_slot(data, ReceiverTypeData::receiver_offset(i)) -
                           mdo_offset_bias);
     __ ld_ptr(receiver_addr, tmp1);
-    __ verify_oop(tmp1);
+    __ verify_klass_ptr(tmp1);
     __ cmp_and_brx_short(recv, tmp1, Assembler::notEqual, Assembler::pt, next_test);
     Address data_addr(mdo, md->byte_offset_of_slot(data, ReceiverTypeData::receiver_count_offset(i)) -
                       mdo_offset_bias);

hotspot/src/cpu/sparc/vm/c1_Runtime1_sparc.cpp

@@ -404,7 +404,9 @@ OopMapSet* Runtime1::generate_code_for(StubID id, StubAssembler* sasm) {
           if (id == fast_new_instance_init_check_id) {
             // make sure the klass is initialized
             __ ldub(G5_klass, in_bytes(InstanceKlass::init_state_offset()), G3_t1);
-            __ cmp_and_br_short(G3_t1, InstanceKlass::fully_initialized, Assembler::notEqual, Assembler::pn, slow_path);
+            __ cmp(G3_t1, InstanceKlass::fully_initialized);
+            __ br(Assembler::notEqual, false, Assembler::pn, slow_path);
+            __ delayed()->nop();
           }
 #ifdef ASSERT
           // assert object can be fast path allocated
@@ -515,7 +517,9 @@ OopMapSet* Runtime1::generate_code_for(StubID id, StubAssembler* sasm) {
 
           // check that array length is small enough for fast path
           __ set(C1_MacroAssembler::max_array_allocation_length, G3_t1);
-          __ cmp_and_br_short(G4_length, G3_t1, Assembler::greaterUnsigned, Assembler::pn, slow_path);
+          __ cmp(G4_length, G3_t1);
+          __ br(Assembler::greaterUnsigned, false, Assembler::pn, slow_path);
+          __ delayed()->nop();
 
           // if we got here then the TLAB allocation failed, so try
           // refilling the TLAB or allocating directly from eden.

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

@@ -3333,7 +3333,8 @@ void MacroAssembler::tlab_refill(Label& retry, Label& try_eden, Label& slow_case
 
   if (CMSIncrementalMode || !Universe::heap()->supports_inline_contig_alloc()) {
     // No allocation in the shared eden.
-    ba_short(slow_case);
+    ba(slow_case);
+    delayed()->nop();
   }
 
   ld_ptr(G2_thread, in_bytes(JavaThread::tlab_top_offset()), top);
@@ -3358,7 +3359,8 @@ void MacroAssembler::tlab_refill(Label& retry, Label& try_eden, Label& slow_case
     add(t2, 1, t2);
     stw(t2, G2_thread, in_bytes(JavaThread::tlab_slow_allocations_offset()));
   }
-  ba_short(try_eden);
+  ba(try_eden);
+  delayed()->nop();
 
   bind(discard_tlab);
   if (TLABStats) {
@@ -3420,7 +3422,8 @@ void MacroAssembler::tlab_refill(Label& retry, Label& try_eden, Label& slow_case
   sub(top, ThreadLocalAllocBuffer::alignment_reserve_in_bytes(), top);
   st_ptr(top, G2_thread, in_bytes(JavaThread::tlab_end_offset()));
   verify_tlab();
-  ba_short(retry);
+  ba(retry);
+  delayed()->nop();
 }
 
 void MacroAssembler::incr_allocated_bytes(RegisterOrConstant size_in_bytes,

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

@@ -2022,6 +2022,10 @@ const RegMask Matcher::mathExactI_result_proj_mask() {
   return G1_REGI_mask();
 }
 
+const RegMask Matcher::mathExactL_result_proj_mask() {
+  return G1_REGL_mask();
+}
+
 const RegMask Matcher::mathExactI_flags_proj_mask() {
   return INT_FLAGS_mask();
 }

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

@@ -1405,6 +1405,15 @@ void Assembler::imull(Register dst, Register src, int value) {
   }
 }
 
+void Assembler::imull(Register dst, Address src) {
+  InstructionMark im(this);
+  prefix(src, dst);
+  emit_int8(0x0F);
+  emit_int8((unsigned char) 0xAF);
+  emit_operand(dst, src);
+}
+
+
 void Assembler::incl(Address dst) {
   // Don't use it directly. Use MacroAssembler::increment() instead.
   InstructionMark im(this);
@@ -5024,6 +5033,14 @@ void Assembler::imulq(Register dst, Register src, int value) {
   }
 }
 
+void Assembler::imulq(Register dst, Address src) {
+  InstructionMark im(this);
+  prefixq(src, dst);
+  emit_int8(0x0F);
+  emit_int8((unsigned char) 0xAF);
+  emit_operand(dst, src);
+}
+
 void Assembler::incl(Register dst) {
   // Don't use it directly. Use MacroAssembler::incrementl() instead.
   // Use two-byte form (one-byte from is a REX prefix in 64-bit mode)

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

@@ -1162,9 +1162,13 @@ private:
 
   void imull(Register dst, Register src);
   void imull(Register dst, Register src, int value);
+  void imull(Register dst, Address src);
 
   void imulq(Register dst, Register src);
   void imulq(Register dst, Register src, int value);
+#ifdef _LP64
+  void imulq(Register dst, Address src);
+#endif
 
 
   // jcc is the generic conditional branch generator to run-

hotspot/src/cpu/x86/vm/bytecodeInterpreter_x86.cpp

@@ -40,11 +40,8 @@
 #include "runtime/synchronizer.hpp"
 #include "runtime/vframeArray.hpp"
 #include "utilities/debug.hpp"
-#ifdef TARGET_ARCH_MODEL_x86_32
+#ifdef TARGET_ARCH_x86
-# include "interp_masm_x86_32.hpp"
+# include "interp_masm_x86.hpp"
-#endif
-#ifdef TARGET_ARCH_MODEL_x86_64
-# include "interp_masm_x86_64.hpp"
 #endif
 
 #ifdef CC_INTERP

hotspot/src/cpu/x86/vm/c1_FrameMap_x86.cpp

@@ -52,6 +52,8 @@ LIR_Opr FrameMap::map_to_opr(BasicType type, VMRegPair* reg, bool) {
 #endif // _LP64
     } else if (type == T_OBJECT || type == T_ARRAY) {
       opr = as_oop_opr(reg);
+    } else if (type == T_METADATA) {
+      opr = as_metadata_opr(reg);
     } else {
       opr = as_opr(reg);
     }

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

@@ -432,15 +432,16 @@ int LIR_Assembler::emit_unwind_handler() {
   int offset = code_offset();
 
   // Fetch the exception from TLS and clear out exception related thread state
-  __ get_thread(rsi);
+  Register thread = NOT_LP64(rsi) LP64_ONLY(r15_thread);
-  __ movptr(rax, Address(rsi, JavaThread::exception_oop_offset()));
+  NOT_LP64(__ get_thread(rsi));
-  __ movptr(Address(rsi, JavaThread::exception_oop_offset()), (intptr_t)NULL_WORD);
+  __ movptr(rax, Address(thread, JavaThread::exception_oop_offset()));
-  __ movptr(Address(rsi, JavaThread::exception_pc_offset()), (intptr_t)NULL_WORD);
+  __ movptr(Address(thread, JavaThread::exception_oop_offset()), (intptr_t)NULL_WORD);
+  __ movptr(Address(thread, JavaThread::exception_pc_offset()), (intptr_t)NULL_WORD);
 
   __ bind(_unwind_handler_entry);
   __ verify_not_null_oop(rax);
   if (method()->is_synchronized() || compilation()->env()->dtrace_method_probes()) {
-    __ mov(rsi, rax);  // Preserve the exception
+    __ mov(rbx, rax);  // Preserve the exception (rbx is always callee-saved)
   }
 
   // Preform needed unlocking
@@ -448,19 +449,24 @@ int LIR_Assembler::emit_unwind_handler() {
   if (method()->is_synchronized()) {
     monitor_address(0, FrameMap::rax_opr);
     stub = new MonitorExitStub(FrameMap::rax_opr, true, 0);
-    __ unlock_object(rdi, rbx, rax, *stub->entry());
+    __ unlock_object(rdi, rsi, rax, *stub->entry());
     __ bind(*stub->continuation());
   }
 
   if (compilation()->env()->dtrace_method_probes()) {
+#ifdef _LP64
+    __ mov(rdi, r15_thread);
+    __ mov_metadata(rsi, method()->constant_encoding());
+#else
     __ get_thread(rax);
     __ movptr(Address(rsp, 0), rax);
     __ mov_metadata(Address(rsp, sizeof(void*)), method()->constant_encoding());
+#endif
     __ call(RuntimeAddress(CAST_FROM_FN_PTR(address, SharedRuntime::dtrace_method_exit)));
   }
 
   if (method()->is_synchronized() || compilation()->env()->dtrace_method_probes()) {
-    __ mov(rax, rsi);  // Restore the exception
+    __ mov(rax, rbx);  // Restore the exception
   }
 
   // remove the activation and dispatch to the unwind handler
@@ -1206,6 +1212,10 @@ void LIR_Assembler::mem2reg(LIR_Opr src, LIR_Opr dest, BasicType type, LIR_Patch
   LIR_Address* addr = src->as_address_ptr();
   Address from_addr = as_Address(addr);
 
+  if (addr->base()->type() == T_OBJECT) {
+    __ verify_oop(addr->base()->as_pointer_register());
+  }
+
   switch (type) {
     case T_BOOLEAN: // fall through
     case T_BYTE:    // fall through

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

@@ -79,7 +79,7 @@ define_pd_global(bool, UseMembar,            false);
 // GC Ergo Flags
 define_pd_global(uintx, CMSYoungGenPerWorker, 64*M);  // default max size of CMS young gen, per GC worker thread
 
-define_pd_global(uintx, TypeProfileLevel, 11);
+define_pd_global(uintx, TypeProfileLevel, 111);
 
 #define ARCH_FLAGS(develop, product, diagnostic, experimental, notproduct) \
                                                                             \

hotspot/src/cpu/x86/vm/interp_masm_x86.cpp

@@ -0,0 +1,229 @@
+/*
+ * Copyright (c) 1997, 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.
+ *
+ */
+
+#include "precompiled.hpp"
+#include "interp_masm_x86.hpp"
+#include "interpreter/interpreter.hpp"
+#include "oops/methodData.hpp"
+
+#ifndef CC_INTERP
+void InterpreterMacroAssembler::profile_obj_type(Register obj, const Address& mdo_addr) {
+  Label update, next, none;
+
+  verify_oop(obj);
+
+  testptr(obj, obj);
+  jccb(Assembler::notZero, update);
+  orptr(mdo_addr, TypeEntries::null_seen);
+  jmpb(next);
+
+  bind(update);
+  load_klass(obj, obj);
+
+  xorptr(obj, mdo_addr);
+  testptr(obj, TypeEntries::type_klass_mask);
+  jccb(Assembler::zero, next); // klass seen before, nothing to
+                               // do. The unknown bit may have been
+                               // set already but no need to check.
+
+  testptr(obj, TypeEntries::type_unknown);
+  jccb(Assembler::notZero, next); // already unknown. Nothing to do anymore.
+
+  cmpptr(mdo_addr, 0);
+  jccb(Assembler::equal, none);
+  cmpptr(mdo_addr, TypeEntries::null_seen);
+  jccb(Assembler::equal, none);
+  // There is a chance that the checks above (re-reading profiling
+  // data from memory) fail if another thread has just set the
+  // profiling to this obj's klass
+  xorptr(obj, mdo_addr);
+  testptr(obj, TypeEntries::type_klass_mask);
+  jccb(Assembler::zero, next);
+
+  // different than before. Cannot keep accurate profile.
+  orptr(mdo_addr, TypeEntries::type_unknown);
+  jmpb(next);
+
+  bind(none);
+  // first time here. Set profile type.
+  movptr(mdo_addr, obj);
+
+  bind(next);
+}
+
+void InterpreterMacroAssembler::profile_arguments_type(Register mdp, Register callee, Register tmp, bool is_virtual) {
+  if (!ProfileInterpreter) {
+    return;
+  }
+
+  if (MethodData::profile_arguments() || MethodData::profile_return()) {
+    Label profile_continue;
+
+    test_method_data_pointer(mdp, profile_continue);
+
+    int off_to_start = is_virtual ? in_bytes(VirtualCallData::virtual_call_data_size()) : in_bytes(CounterData::counter_data_size());
+
+    cmpb(Address(mdp, in_bytes(DataLayout::tag_offset()) - off_to_start), is_virtual ? DataLayout::virtual_call_type_data_tag : DataLayout::call_type_data_tag);
+    jcc(Assembler::notEqual, profile_continue);
+
+    if (MethodData::profile_arguments()) {
+      Label done;
+      int off_to_args = in_bytes(TypeEntriesAtCall::args_data_offset());
+      addptr(mdp, off_to_args);
+
+      for (int i = 0; i < TypeProfileArgsLimit; i++) {
+        if (i > 0 || MethodData::profile_return()) {
+          // If return value type is profiled we may have no argument to profile
+          movptr(tmp, Address(mdp, in_bytes(TypeEntriesAtCall::cell_count_offset())-off_to_args));
+          subl(tmp, i*TypeStackSlotEntries::per_arg_count());
+          cmpl(tmp, TypeStackSlotEntries::per_arg_count());
+          jcc(Assembler::less, done);
+        }
+        movptr(tmp, Address(callee, Method::const_offset()));
+        load_unsigned_short(tmp, Address(tmp, ConstMethod::size_of_parameters_offset()));
+        // stack offset o (zero based) from the start of the argument
+        // list, for n arguments translates into offset n - o - 1 from
+        // the end of the argument list
+        subptr(tmp, Address(mdp, in_bytes(TypeEntriesAtCall::stack_slot_offset(i))-off_to_args));
+        subl(tmp, 1);
+        Address arg_addr = argument_address(tmp);
+        movptr(tmp, arg_addr);
+
+        Address mdo_arg_addr(mdp, in_bytes(TypeEntriesAtCall::argument_type_offset(i))-off_to_args);
+        profile_obj_type(tmp, mdo_arg_addr);
+
+        int to_add = in_bytes(TypeStackSlotEntries::per_arg_size());
+        addptr(mdp, to_add);
+        off_to_args += to_add;
+      }
+
+      if (MethodData::profile_return()) {
+        movptr(tmp, Address(mdp, in_bytes(TypeEntriesAtCall::cell_count_offset())-off_to_args));
+        subl(tmp, TypeProfileArgsLimit*TypeStackSlotEntries::per_arg_count());
+      }
+
+      bind(done);
+
+      if (MethodData::profile_return()) {
+        // We're right after the type profile for the last
+        // argument. tmp is the number of cell left in the
+        // CallTypeData/VirtualCallTypeData to reach its end. Non null
+        // if there's a return to profile.
+        assert(ReturnTypeEntry::static_cell_count() < TypeStackSlotEntries::per_arg_count(), "can't move past ret type");
+        shll(tmp, exact_log2(DataLayout::cell_size));
+        addptr(mdp, tmp);
+      }
+      movptr(Address(rbp, frame::interpreter_frame_mdx_offset * wordSize), mdp);
+    } else {
+      assert(MethodData::profile_return(), "either profile call args or call ret");
+      update_mdp_by_constant(mdp, in_bytes(ReturnTypeEntry::size()));
+    }
+
+    // mdp points right after the end of the
+    // CallTypeData/VirtualCallTypeData, right after the cells for the
+    // return value type if there's one
+
+    bind(profile_continue);
+  }
+}
+
+void InterpreterMacroAssembler::profile_return_type(Register mdp, Register ret, Register tmp) {
+  assert_different_registers(mdp, ret, tmp, _bcp_register);
+  if (ProfileInterpreter && MethodData::profile_return()) {
+    Label profile_continue, done;
+
+    test_method_data_pointer(mdp, profile_continue);
+
+    if (MethodData::profile_return_jsr292_only()) {
+      // If we don't profile all invoke bytecodes we must make sure
+      // it's a bytecode we indeed profile. We can't go back to the
+      // begining of the ProfileData we intend to update to check its
+      // type because we're right after it and we don't known its
+      // length
+      Label do_profile;
+      cmpb(Address(_bcp_register, 0), Bytecodes::_invokedynamic);
+      jcc(Assembler::equal, do_profile);
+      cmpb(Address(_bcp_register, 0), Bytecodes::_invokehandle);
+      jcc(Assembler::equal, do_profile);
+      get_method(tmp);
+      cmpb(Address(tmp, Method::intrinsic_id_offset_in_bytes()), vmIntrinsics::_compiledLambdaForm);
+      jcc(Assembler::notEqual, profile_continue);
+
+      bind(do_profile);
+    }
+
+    Address mdo_ret_addr(mdp, -in_bytes(ReturnTypeEntry::size()));
+    mov(tmp, ret);
+    profile_obj_type(tmp, mdo_ret_addr);
+
+    bind(profile_continue);
+  }
+}
+
+void InterpreterMacroAssembler::profile_parameters_type(Register mdp, Register tmp1, Register tmp2) {
+  if (ProfileInterpreter && MethodData::profile_parameters()) {
+    Label profile_continue, done;
+
+    test_method_data_pointer(mdp, profile_continue);
+
+    // Load the offset of the area within the MDO used for
+    // parameters. If it's negative we're not profiling any parameters
+    movl(tmp1, Address(mdp, in_bytes(MethodData::parameters_type_data_di_offset()) - in_bytes(MethodData::data_offset())));
+    testl(tmp1, tmp1);
+    jcc(Assembler::negative, profile_continue);
+
+    // Compute a pointer to the area for parameters from the offset
+    // and move the pointer to the slot for the last
+    // parameters. Collect profiling from last parameter down.
+    // mdo start + parameters offset + array length - 1
+    addptr(mdp, tmp1);
+    movptr(tmp1, Address(mdp, in_bytes(ArrayData::array_len_offset())));
+    decrement(tmp1, TypeStackSlotEntries::per_arg_count());
+
+    Label loop;
+    bind(loop);
+
+    int off_base = in_bytes(ParametersTypeData::stack_slot_offset(0));
+    int type_base = in_bytes(ParametersTypeData::type_offset(0));
+    Address::ScaleFactor per_arg_scale = Address::times(DataLayout::cell_size);
+    Address arg_off(mdp, tmp1, per_arg_scale, off_base);
+    Address arg_type(mdp, tmp1, per_arg_scale, type_base);
+
+    // load offset on the stack from the slot for this parameter
+    movptr(tmp2, arg_off);
+    negptr(tmp2);
+    // read the parameter from the local area
+    movptr(tmp2, Address(_locals_register, tmp2, Interpreter::stackElementScale()));
+
+    // profile the parameter
+    profile_obj_type(tmp2, arg_type);
+
+    // go to next parameter
+    decrement(tmp1, TypeStackSlotEntries::per_arg_count());
+    jcc(Assembler::positive, loop);
+
+    bind(profile_continue);
+  }
+}
+#endif

hotspot/src/cpu/x86/vm/interp_masm_x86.hpp

@@ -0,0 +1,60 @@
+/*
+ * Copyright (c) 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.
+ *
+ */
+
+#ifndef CPU_X86_VM_INTERP_MASM_X86_HPP
+#define CPU_X86_VM_INTERP_MASM_X86_HPP
+
+#include "asm/macroAssembler.hpp"
+#include "asm/macroAssembler.inline.hpp"
+#include "interpreter/invocationCounter.hpp"
+#include "runtime/frame.hpp"
+
+// This file specializes the assember with interpreter-specific macros
+
+
+class InterpreterMacroAssembler: public MacroAssembler {
+
+#ifdef TARGET_ARCH_MODEL_x86_32
+# include "interp_masm_x86_32.hpp"
+#endif
+#ifdef TARGET_ARCH_MODEL_x86_64
+# include "interp_masm_x86_64.hpp"
+#endif
+
+ private:
+
+  Register _locals_register; // register that contains the pointer to the locals
+  Register _bcp_register; // register that contains the bcp
+
+ public:
+#ifndef CC_INTERP
+  void profile_obj_type(Register obj, const Address& mdo_addr);
+  void profile_arguments_type(Register mdp, Register callee, Register tmp, bool is_virtual);
+  void profile_return_type(Register mdp, Register ret, Register tmp);
+  void profile_parameters_type(Register mdp, Register tmp1, Register tmp2);
+#endif /* !CC_INTERP */
+
+};
+
+#endif // CPU_X86_VM_INTERP_MASM_X86_HPP

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

@@ -23,7 +23,7 @@
  */
 
 #include "precompiled.hpp"
-#include "interp_masm_x86_32.hpp"
+#include "interp_masm_x86.hpp"
 #include "interpreter/interpreter.hpp"
 #include "interpreter/interpreterRuntime.hpp"
 #include "oops/arrayOop.hpp"
@@ -1046,159 +1046,6 @@ void InterpreterMacroAssembler::profile_not_taken_branch(Register mdp) {
   }
 }
 
-void InterpreterMacroAssembler::profile_obj_type(Register obj, const Address& mdo_addr) {
-  Label update, next, none;
-
-  verify_oop(obj);
-
-  testptr(obj, obj);
-  jccb(Assembler::notZero, update);
-  orptr(mdo_addr, TypeEntries::null_seen);
-  jmpb(next);
-
-  bind(update);
-  load_klass(obj, obj);
-
-  xorptr(obj, mdo_addr);
-  testptr(obj, TypeEntries::type_klass_mask);
-  jccb(Assembler::zero, next); // klass seen before, nothing to
-                               // do. The unknown bit may have been
-                               // set already but no need to check.
-
-  testptr(obj, TypeEntries::type_unknown);
-  jccb(Assembler::notZero, next); // already unknown. Nothing to do anymore.
-
-  cmpptr(mdo_addr, 0);
-  jccb(Assembler::equal, none);
-  cmpptr(mdo_addr, TypeEntries::null_seen);
-  jccb(Assembler::equal, none);
-  // There is a chance that the checks above (re-reading profiling
-  // data from memory) fail if another thread has just set the
-  // profiling to this obj's klass
-  xorptr(obj, mdo_addr);
-  testptr(obj, TypeEntries::type_klass_mask);
-  jccb(Assembler::zero, next);
-
-  // different than before. Cannot keep accurate profile.
-  orptr(mdo_addr, TypeEntries::type_unknown);
-  jmpb(next);
-
-  bind(none);
-  // first time here. Set profile type.
-  movptr(mdo_addr, obj);
-
-  bind(next);
-}
-
-void InterpreterMacroAssembler::profile_arguments_type(Register mdp, Register callee, Register tmp, bool is_virtual) {
-  if (!ProfileInterpreter) {
-    return;
-  }
-
-  if (MethodData::profile_arguments() || MethodData::profile_return()) {
-    Label profile_continue;
-
-    test_method_data_pointer(mdp, profile_continue);
-
-    int off_to_start = is_virtual ? in_bytes(VirtualCallData::virtual_call_data_size()) : in_bytes(CounterData::counter_data_size());
-
-    cmpb(Address(mdp, in_bytes(DataLayout::tag_offset()) - off_to_start), is_virtual ? DataLayout::virtual_call_type_data_tag : DataLayout::call_type_data_tag);
-    jcc(Assembler::notEqual, profile_continue);
-
-    if (MethodData::profile_arguments()) {
-      Label done;
-      int off_to_args = in_bytes(TypeEntriesAtCall::args_data_offset());
-      addptr(mdp, off_to_args);
-
-      for (int i = 0; i < TypeProfileArgsLimit; i++) {
-        if (i > 0 || MethodData::profile_return()) {
-          // If return value type is profiled we may have no argument to profile
-          movl(tmp, Address(mdp, in_bytes(TypeEntriesAtCall::cell_count_offset())-off_to_args));
-          subl(tmp, i*TypeStackSlotEntries::per_arg_count());
-          cmpl(tmp, TypeStackSlotEntries::per_arg_count());
-          jcc(Assembler::less, done);
-        }
-        movptr(tmp, Address(callee, Method::const_offset()));
-        load_unsigned_short(tmp, Address(tmp, ConstMethod::size_of_parameters_offset()));
-        // stack offset o (zero based) from the start of the argument
-        // list, for n arguments translates into offset n - o - 1 from
-        // the end of the argument list
-        subl(tmp, Address(mdp, in_bytes(TypeEntriesAtCall::stack_slot_offset(i))-off_to_args));
-        subl(tmp, 1);
-        Address arg_addr = argument_address(tmp);
-        movptr(tmp, arg_addr);
-
-        Address mdo_arg_addr(mdp, in_bytes(TypeEntriesAtCall::argument_type_offset(i))-off_to_args);
-        profile_obj_type(tmp, mdo_arg_addr);
-
-        int to_add = in_bytes(TypeStackSlotEntries::per_arg_size());
-        addptr(mdp, to_add);
-        off_to_args += to_add;
-      }
-
-      if (MethodData::profile_return()) {
-        movl(tmp, Address(mdp, in_bytes(TypeEntriesAtCall::cell_count_offset())-off_to_args));
-        subl(tmp, TypeProfileArgsLimit*TypeStackSlotEntries::per_arg_count());
-      }
-
-      bind(done);
-
-      if (MethodData::profile_return()) {
-        // We're right after the type profile for the last
-        // argument. tmp is the number of cell left in the
-        // CallTypeData/VirtualCallTypeData to reach its end. Non null
-        // if there's a return to profile.
-        assert(ReturnTypeEntry::static_cell_count() < TypeStackSlotEntries::per_arg_count(), "can't move past ret type");
-        shll(tmp, exact_log2(DataLayout::cell_size));
-        addptr(mdp, tmp);
-      }
-      movptr(Address(rbp, frame::interpreter_frame_mdx_offset * wordSize), mdp);
-    } else {
-      assert(MethodData::profile_return(), "either profile call args or call ret");
-      update_mdp_by_constant(mdp, in_bytes(ReturnTypeEntry::size()));
-    }
-
-    // mdp points right after the end of the
-    // CallTypeData/VirtualCallTypeData, right after the cells for the
-    // return value type if there's one
-
-    bind(profile_continue);
-  }
-}
-
-void InterpreterMacroAssembler::profile_return_type(Register mdp, Register ret, Register tmp) {
-  assert_different_registers(mdp, ret, tmp, rsi);
-  if (ProfileInterpreter && MethodData::profile_return()) {
-    Label profile_continue, done;
-
-    test_method_data_pointer(mdp, profile_continue);
-
-    if (MethodData::profile_return_jsr292_only()) {
-      // If we don't profile all invoke bytecodes we must make sure
-      // it's a bytecode we indeed profile. We can't go back to the
-      // begining of the ProfileData we intend to update to check its
-      // type because we're right after it and we don't known its
-      // length
-      Label do_profile;
-      cmpb(Address(rsi, 0), Bytecodes::_invokedynamic);
-      jcc(Assembler::equal, do_profile);
-      cmpb(Address(rsi, 0), Bytecodes::_invokehandle);
-      jcc(Assembler::equal, do_profile);
-      get_method(tmp);
-      cmpb(Address(tmp, Method::intrinsic_id_offset_in_bytes()), vmIntrinsics::_compiledLambdaForm);
-      jcc(Assembler::notEqual, profile_continue);
-
-      bind(do_profile);
-    }
-
-    Address mdo_ret_addr(mdp, -in_bytes(ReturnTypeEntry::size()));
-    mov(tmp, ret);
-    profile_obj_type(tmp, mdo_ret_addr);
-
-    bind(profile_continue);
-  }
-}
-
 void InterpreterMacroAssembler::profile_call(Register mdp) {
   if (ProfileInterpreter) {
     Label profile_continue;

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

@@ -22,18 +22,6 @@
  *
  */
 
-#ifndef CPU_X86_VM_INTERP_MASM_X86_32_HPP
-#define CPU_X86_VM_INTERP_MASM_X86_32_HPP
-
-#include "asm/macroAssembler.hpp"
-#include "asm/macroAssembler.inline.hpp"
-#include "interpreter/invocationCounter.hpp"
-#include "runtime/frame.hpp"
-
-// This file specializes the assember with interpreter-specific macros
-
-
-class InterpreterMacroAssembler: public MacroAssembler {
 #ifndef CC_INTERP
  protected:
   // Interpreter specific version of call_VM_base
@@ -59,7 +47,7 @@ class InterpreterMacroAssembler: public MacroAssembler {
 #endif /* CC_INTERP */
 
  public:
-  InterpreterMacroAssembler(CodeBuffer* code) : MacroAssembler(code) {}
+  InterpreterMacroAssembler(CodeBuffer* code) : MacroAssembler(code), _locals_register(rdi), _bcp_register(rsi) {}
 
   void load_earlyret_value(TosState state);
 
@@ -215,9 +203,6 @@ class InterpreterMacroAssembler: public MacroAssembler {
 
   void profile_taken_branch(Register mdp, Register bumped_count);
   void profile_not_taken_branch(Register mdp);
-  void profile_obj_type(Register obj, const Address& mdo_addr);
-  void profile_arguments_type(Register mdp, Register callee, Register tmp, bool is_virtual);
-  void profile_return_type(Register mdp, Register ret, Register tmp);
   void profile_call(Register mdp);
   void profile_final_call(Register mdp);
   void profile_virtual_call(Register receiver, Register mdp, Register scratch2,
@@ -236,7 +221,3 @@ class InterpreterMacroAssembler: public MacroAssembler {
   // support for jvmti
   void notify_method_entry();
   void notify_method_exit(TosState state, NotifyMethodExitMode mode);
-
-};
-
-#endif // CPU_X86_VM_INTERP_MASM_X86_32_HPP

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

@@ -23,7 +23,7 @@
  */
 
 #include "precompiled.hpp"
-#include "interp_masm_x86_64.hpp"
+#include "interp_masm_x86.hpp"
 #include "interpreter/interpreter.hpp"
 #include "interpreter/interpreterRuntime.hpp"
 #include "oops/arrayOop.hpp"
@@ -1067,160 +1067,6 @@ void InterpreterMacroAssembler::profile_not_taken_branch(Register mdp) {
   }
 }
 
-void InterpreterMacroAssembler::profile_obj_type(Register obj, const Address& mdo_addr) {
-  Label update, next, none;
-
-  verify_oop(obj);
-
-  testptr(obj, obj);
-  jccb(Assembler::notZero, update);
-  orptr(mdo_addr, TypeEntries::null_seen);
-  jmpb(next);
-
-  bind(update);
-  load_klass(obj, obj);
-
-  xorptr(obj, mdo_addr);
-  testptr(obj, TypeEntries::type_klass_mask);
-  jccb(Assembler::zero, next); // klass seen before, nothing to
-                               // do. The unknown bit may have been
-                               // set already but no need to check.
-
-  testptr(obj, TypeEntries::type_unknown);
-  jccb(Assembler::notZero, next); // already unknown. Nothing to do anymore.
-
-  // There is a chance that by the time we do these checks (re-reading
-  // profiling data from memory) another thread has set the profling
-  // to this obj's klass and we set the profiling as unknow
-  // erroneously
-  cmpptr(mdo_addr, 0);
-  jccb(Assembler::equal, none);
-  cmpptr(mdo_addr, TypeEntries::null_seen);
-  jccb(Assembler::equal, none);
-  // There is a chance that the checks above (re-reading profiling
-  // data from memory) fail if another thread has just set the
-  // profiling to this obj's klass
-  xorptr(obj, mdo_addr);
-  testptr(obj, TypeEntries::type_klass_mask);
-  jccb(Assembler::zero, next);
-
-  // different than before. Cannot keep accurate profile.
-  orptr(mdo_addr, TypeEntries::type_unknown);
-  jmpb(next);
-
-  bind(none);
-  // first time here. Set profile type.
-  movptr(mdo_addr, obj);
-
-  bind(next);
-}
-
-void InterpreterMacroAssembler::profile_arguments_type(Register mdp, Register callee, Register tmp, bool is_virtual) {
-  if (!ProfileInterpreter) {
-    return;
-  }
-
-  if (MethodData::profile_arguments() || MethodData::profile_return()) {
-    Label profile_continue;
-
-    test_method_data_pointer(mdp, profile_continue);
-
-    int off_to_start = is_virtual ? in_bytes(VirtualCallData::virtual_call_data_size()) : in_bytes(CounterData::counter_data_size());
-
-    cmpb(Address(mdp, in_bytes(DataLayout::tag_offset()) - off_to_start), is_virtual ? DataLayout::virtual_call_type_data_tag : DataLayout::call_type_data_tag);
-    jcc(Assembler::notEqual, profile_continue);
-
-    if (MethodData::profile_arguments()) {
-      Label done;
-      int off_to_args = in_bytes(TypeEntriesAtCall::args_data_offset());
-      addptr(mdp, off_to_args);
-
-      for (int i = 0; i < TypeProfileArgsLimit; i++) {
-        if (i > 0 || MethodData::profile_return()) {
-          // If return value type is profiled we may have no argument to profile
-          movq(tmp, Address(mdp, in_bytes(TypeEntriesAtCall::cell_count_offset())-off_to_args));
-          subl(tmp, i*TypeStackSlotEntries::per_arg_count());
-          cmpl(tmp, TypeStackSlotEntries::per_arg_count());
-          jcc(Assembler::less, done);
-        }
-        movptr(tmp, Address(callee, Method::const_offset()));
-        load_unsigned_short(tmp, Address(tmp, ConstMethod::size_of_parameters_offset()));
-        subq(tmp, Address(mdp, in_bytes(TypeEntriesAtCall::stack_slot_offset(i))-off_to_args));
-        subl(tmp, 1);
-        Address arg_addr = argument_address(tmp);
-        movptr(tmp, arg_addr);
-
-        Address mdo_arg_addr(mdp, in_bytes(TypeEntriesAtCall::argument_type_offset(i))-off_to_args);
-        profile_obj_type(tmp, mdo_arg_addr);
-
-        int to_add = in_bytes(TypeStackSlotEntries::per_arg_size());
-        addptr(mdp, to_add);
-        off_to_args += to_add;
-      }
-
-      if (MethodData::profile_return()) {
-        movq(tmp, Address(mdp, in_bytes(TypeEntriesAtCall::cell_count_offset())-off_to_args));
-        subl(tmp, TypeProfileArgsLimit*TypeStackSlotEntries::per_arg_count());
-      }
-
-      bind(done);
-
-      if (MethodData::profile_return()) {
-        // We're right after the type profile for the last
-        // argument. tmp is the number of cell left in the
-        // CallTypeData/VirtualCallTypeData to reach its end. Non null
-        // if there's a return to profile.
-        assert(ReturnTypeEntry::static_cell_count() < TypeStackSlotEntries::per_arg_count(), "can't move past ret type");
-        shll(tmp, exact_log2(DataLayout::cell_size));
-        addptr(mdp, tmp);
-      }
-      movptr(Address(rbp, frame::interpreter_frame_mdx_offset * wordSize), mdp);
-    } else {
-      assert(MethodData::profile_return(), "either profile call args or call ret");
-      update_mdp_by_constant(mdp, in_bytes(ReturnTypeEntry::size()));
-    }
-
-    // mdp points right after the end of the
-    // CallTypeData/VirtualCallTypeData, right after the cells for the
-    // return value type if there's one
-
-    bind(profile_continue);
-  }
-}
-
-void InterpreterMacroAssembler::profile_return_type(Register mdp, Register ret, Register tmp) {
-  assert_different_registers(mdp, ret, tmp, r13);
-  if (ProfileInterpreter && MethodData::profile_return()) {
-    Label profile_continue, done;
-
-    test_method_data_pointer(mdp, profile_continue);
-
-    if (MethodData::profile_return_jsr292_only()) {
-      // If we don't profile all invoke bytecodes we must make sure
-      // it's a bytecode we indeed profile. We can't go back to the
-      // begining of the ProfileData we intend to update to check its
-      // type because we're right after it and we don't known its
-      // length
-      Label do_profile;
-      cmpb(Address(r13, 0), Bytecodes::_invokedynamic);
-      jcc(Assembler::equal, do_profile);
-      cmpb(Address(r13, 0), Bytecodes::_invokehandle);
-      jcc(Assembler::equal, do_profile);
-      get_method(tmp);
-      cmpb(Address(tmp, Method::intrinsic_id_offset_in_bytes()), vmIntrinsics::_compiledLambdaForm);
-      jcc(Assembler::notEqual, profile_continue);
-
-      bind(do_profile);
-    }
-
-    Address mdo_ret_addr(mdp, -in_bytes(ReturnTypeEntry::size()));
-    mov(tmp, ret);
-    profile_obj_type(tmp, mdo_ret_addr);
-
-    bind(profile_continue);
-  }
-}
-
 void InterpreterMacroAssembler::profile_call(Register mdp) {
   if (ProfileInterpreter) {
     Label profile_continue;

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

@@ -22,18 +22,6 @@
  *
  */
 
-#ifndef CPU_X86_VM_INTERP_MASM_X86_64_HPP
-#define CPU_X86_VM_INTERP_MASM_X86_64_HPP
-
-#include "asm/macroAssembler.hpp"
-#include "asm/macroAssembler.inline.hpp"
-#include "interpreter/invocationCounter.hpp"
-#include "runtime/frame.hpp"
-
-// This file specializes the assember with interpreter-specific macros
-
-
-class InterpreterMacroAssembler: public MacroAssembler {
 #ifndef CC_INTERP
  protected:
   // Interpreter specific version of call_VM_base
@@ -55,7 +43,7 @@ class InterpreterMacroAssembler: public MacroAssembler {
 #endif // CC_INTERP
 
  public:
-  InterpreterMacroAssembler(CodeBuffer* code) : MacroAssembler(code) {}
+  InterpreterMacroAssembler(CodeBuffer* code) : MacroAssembler(code), _locals_register(r14), _bcp_register(r13) {}
 
   void load_earlyret_value(TosState state);
 
@@ -224,9 +212,6 @@ class InterpreterMacroAssembler: public MacroAssembler {
 
   void profile_taken_branch(Register mdp, Register bumped_count);
   void profile_not_taken_branch(Register mdp);
-  void profile_obj_type(Register obj, const Address& mdo_addr);
-  void profile_arguments_type(Register mdp, Register callee, Register tmp, bool is_virtual);
-  void profile_return_type(Register mdp, Register ret, Register tmp);
   void profile_call(Register mdp);
   void profile_final_call(Register mdp);
   void profile_virtual_call(Register receiver, Register mdp,
@@ -253,6 +238,3 @@ class InterpreterMacroAssembler: public MacroAssembler {
   // support for jvmti/dtrace
   void notify_method_entry();
   void notify_method_exit(TosState state, NotifyMethodExitMode mode);
-};
-
-#endif // CPU_X86_VM_INTERP_MASM_X86_64_HPP

hotspot/src/cpu/x86/vm/register_definitions_x86.cpp

@@ -26,11 +26,8 @@
 #include "asm/assembler.hpp"
 #include "asm/register.hpp"
 #include "register_x86.hpp"
-#ifdef TARGET_ARCH_MODEL_x86_32
+#ifdef TARGET_ARCH_x86
-# include "interp_masm_x86_32.hpp"
+# include "interp_masm_x86.hpp"
-#endif
-#ifdef TARGET_ARCH_MODEL_x86_64
-# include "interp_masm_x86_64.hpp"
 #endif
 
 REGISTER_DEFINITION(Register, noreg);

hotspot/src/cpu/x86/vm/templateInterpreter_x86.hpp

@@ -34,9 +34,9 @@
   // Run with +PrintInterpreter to get the VM to print out the size.
   // Max size with JVMTI
 #ifdef AMD64
-  const static int InterpreterCodeSize = 208 * 1024;
+  const static int InterpreterCodeSize = 256 * 1024;
 #else
-  const static int InterpreterCodeSize = 176 * 1024;
+  const static int InterpreterCodeSize = 224 * 1024;
 #endif // AMD64
 
 #endif // CPU_X86_VM_TEMPLATEINTERPRETER_X86_HPP

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

@@ -1490,6 +1490,7 @@ address InterpreterGenerator::generate_normal_entry(bool synchronized) {
         in_bytes(JavaThread::do_not_unlock_if_synchronized_offset()));
   __ movbool(do_not_unlock_if_synchronized, true);
 
+  __ profile_parameters_type(rax, rcx, rdx);
   // increment invocation count & check for overflow
   Label invocation_counter_overflow;
   Label profile_method;

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

@@ -1497,6 +1497,7 @@ address InterpreterGenerator::generate_normal_entry(bool synchronized) {
         in_bytes(JavaThread::do_not_unlock_if_synchronized_offset()));
   __ movbool(do_not_unlock_if_synchronized, true);
 
+  __ profile_parameters_type(rax, rcx, rdx);
   // increment invocation count & check for overflow
   Label invocation_counter_overflow;
   Label profile_method;

hotspot/src/cpu/x86/vm/vtableStubs_x86_32.cpp

@@ -25,7 +25,7 @@
 #include "precompiled.hpp"
 #include "asm/macroAssembler.hpp"
 #include "code/vtableStubs.hpp"
-#include "interp_masm_x86_32.hpp"
+#include "interp_masm_x86.hpp"
 #include "memory/resourceArea.hpp"
 #include "oops/instanceKlass.hpp"
 #include "oops/klassVtable.hpp"

hotspot/src/cpu/x86/vm/vtableStubs_x86_64.cpp

@@ -25,7 +25,7 @@
 #include "precompiled.hpp"
 #include "asm/macroAssembler.hpp"
 #include "code/vtableStubs.hpp"
-#include "interp_masm_x86_64.hpp"
+#include "interp_masm_x86.hpp"
 #include "memory/resourceArea.hpp"
 #include "oops/instanceKlass.hpp"
 #include "oops/klassVtable.hpp"

hotspot/src/cpu/x86/vm/x86_32.ad

@@ -1538,6 +1538,11 @@ const RegMask Matcher::mathExactI_result_proj_mask() {
   return EAX_REG_mask();
 }
 
+const RegMask Matcher::mathExactL_result_proj_mask() {
+  ShouldNotReachHere();
+  return RegMask();
+}
+
 const RegMask Matcher::mathExactI_flags_proj_mask() {
   return INT_FLAGS_mask();
 }
@@ -7519,7 +7524,7 @@ instruct cmovL_regUCF(cmpOpUCF cop, eFlagsRegUCF cr, eRegL dst, eRegL src) %{
 //----------Arithmetic Instructions--------------------------------------------
 //----------Addition Instructions----------------------------------------------
 
-instruct addExactI_rReg(eAXRegI dst, rRegI src, eFlagsReg cr)
+instruct addExactI_eReg(eAXRegI dst, rRegI src, eFlagsReg cr)
 %{
   match(AddExactI dst src);
   effect(DEF cr);
@@ -7531,7 +7536,7 @@ instruct addExactI_rReg(eAXRegI dst, rRegI src, eFlagsReg cr)
   ins_pipe(ialu_reg_reg);
 %}
 
-instruct addExactI_rReg_imm(eAXRegI dst, immI src, eFlagsReg cr)
+instruct addExactI_eReg_imm(eAXRegI dst, immI src, eFlagsReg cr)
 %{
   match(AddExactI dst src);
   effect(DEF cr);
@@ -7543,6 +7548,20 @@ instruct addExactI_rReg_imm(eAXRegI dst, immI src, eFlagsReg cr)
   ins_pipe(ialu_reg_reg);
 %}
 
+instruct addExactI_eReg_mem(eAXRegI dst, memory src, eFlagsReg cr)
+%{
+  match(AddExactI dst (LoadI src));
+  effect(DEF cr);
+
+  ins_cost(125);
+  format %{ "ADD    $dst,$src\t# addExact int" %}
+  ins_encode %{
+    __ addl($dst$$Register, $src$$Address);
+  %}
+  ins_pipe( ialu_reg_mem );
+%}
+
+
 // Integer Addition Instructions
 instruct addI_eReg(rRegI dst, rRegI src, eFlagsReg cr) %{
   match(Set dst (AddI dst src));
@@ -7851,6 +7870,44 @@ instruct xchgP( memory mem, pRegP newval) %{
 %}
 
 //----------Subtraction Instructions-------------------------------------------
+
+instruct subExactI_eReg(eAXRegI dst, rRegI src, eFlagsReg cr)
+%{
+  match(SubExactI dst src);
+  effect(DEF cr);
+
+  format %{ "SUB    $dst, $src\t# subExact int" %}
+  ins_encode %{
+    __ subl($dst$$Register, $src$$Register);
+  %}
+  ins_pipe(ialu_reg_reg);
+%}
+
+instruct subExactI_eReg_imm(eAXRegI dst, immI src, eFlagsReg cr)
+%{
+  match(SubExactI dst src);
+  effect(DEF cr);
+
+  format %{ "SUB    $dst, $src\t# subExact int" %}
+  ins_encode %{
+    __ subl($dst$$Register, $src$$constant);
+  %}
+  ins_pipe(ialu_reg_reg);
+%}
+
+instruct subExactI_eReg_mem(eAXRegI dst, memory src, eFlagsReg cr)
+%{
+  match(SubExactI dst (LoadI src));
+  effect(DEF cr);
+
+  ins_cost(125);
+  format %{ "SUB    $dst,$src\t# subExact int" %}
+  ins_encode %{
+    __ subl($dst$$Register, $src$$Address);
+  %}
+  ins_pipe( ialu_reg_mem );
+%}
+
 // Integer Subtraction Instructions
 instruct subI_eReg(rRegI dst, rRegI src, eFlagsReg cr) %{
   match(Set dst (SubI dst src));
@@ -7919,6 +7976,16 @@ instruct negI_eReg(rRegI dst, immI0 zero, eFlagsReg cr) %{
   ins_pipe( ialu_reg );
 %}
 
+instruct negExactI_eReg(eAXRegI dst, eFlagsReg cr) %{
+  match(NegExactI dst);
+  effect(DEF cr);
+
+  format %{ "NEG    $dst\t# negExact int"%}
+  ins_encode %{
+    __ negl($dst$$Register);
+  %}
+  ins_pipe(ialu_reg);
+%}
 
 //----------Multiplication/Division Instructions-------------------------------
 // Integer Multiplication Instructions
@@ -8131,6 +8198,46 @@ instruct mulL_eReg_con(eADXRegL dst, immL_127 src, rRegI tmp, eFlagsReg cr) %{
   ins_pipe( pipe_slow );
 %}
 
+instruct mulExactI_eReg(eAXRegI dst, rRegI src, eFlagsReg cr)
+%{
+  match(MulExactI dst src);
+  effect(DEF cr);
+
+  ins_cost(300);
+  format %{ "IMUL   $dst, $src\t# mulExact int" %}
+  ins_encode %{
+    __ imull($dst$$Register, $src$$Register);
+  %}
+  ins_pipe(ialu_reg_reg_alu0);
+%}
+
+instruct mulExactI_eReg_imm(eAXRegI dst, rRegI src, immI imm, eFlagsReg cr)
+%{
+  match(MulExactI src imm);
+  effect(DEF cr);
+
+  ins_cost(300);
+  format %{ "IMUL   $dst, $src, $imm\t# mulExact int" %}
+  ins_encode %{
+    __ imull($dst$$Register, $src$$Register, $imm$$constant);
+  %}
+  ins_pipe(ialu_reg_reg_alu0);
+%}
+
+instruct mulExactI_eReg_mem(eAXRegI dst, memory src, eFlagsReg cr)
+%{
+  match(MulExactI dst (LoadI src));
+  effect(DEF cr);
+
+  ins_cost(350);
+  format %{ "IMUL   $dst, $src\t# mulExact int" %}
+  ins_encode %{
+    __ imull($dst$$Register, $src$$Address);
+  %}
+  ins_pipe(ialu_reg_mem_alu0);
+%}
+
+
 // Integer DIV with Register
 instruct divI_eReg(eAXRegI rax, eDXRegI rdx, eCXRegI div, eFlagsReg cr) %{
   match(Set rax (DivI rax div));

hotspot/src/cpu/x86/vm/x86_64.ad

@@ -1653,6 +1653,10 @@ const RegMask Matcher::mathExactI_result_proj_mask() {
   return INT_RAX_REG_mask();
 }
 
+const RegMask Matcher::mathExactL_result_proj_mask() {
+  return LONG_RAX_REG_mask();
+}
+
 const RegMask Matcher::mathExactI_flags_proj_mask() {
   return INT_FLAGS_mask();
 }
@@ -6962,6 +6966,58 @@ instruct addExactI_rReg_imm(rax_RegI dst, immI src, rFlagsReg cr)
   ins_pipe(ialu_reg_reg);
 %}
 
+instruct addExactI_rReg_mem(rax_RegI dst, memory src, rFlagsReg cr)
+%{
+  match(AddExactI dst (LoadI src));
+  effect(DEF cr);
+
+  ins_cost(125); // XXX
+  format %{ "addl    $dst, $src\t# addExact int" %}
+  ins_encode %{
+    __ addl($dst$$Register, $src$$Address);
+  %}
+
+  ins_pipe(ialu_reg_mem);
+%}
+
+instruct addExactL_rReg(rax_RegL dst, rRegL src, rFlagsReg cr)
+%{
+  match(AddExactL dst src);
+  effect(DEF cr);
+
+  format %{ "addq    $dst, $src\t# addExact long" %}
+  ins_encode %{
+    __ addq($dst$$Register, $src$$Register);
+  %}
+  ins_pipe(ialu_reg_reg);
+%}
+
+instruct addExactL_rReg_imm(rax_RegL dst, immL32 src, rFlagsReg cr)
+%{
+  match(AddExactL dst src);
+  effect(DEF cr);
+
+  format %{ "addq    $dst, $src\t# addExact long" %}
+  ins_encode %{
+    __ addq($dst$$Register, $src$$constant);
+  %}
+  ins_pipe(ialu_reg_reg);
+%}
+
+instruct addExactL_rReg_mem(rax_RegL dst, memory src, rFlagsReg cr)
+%{
+  match(AddExactL dst (LoadL src));
+  effect(DEF cr);
+
+  ins_cost(125); // XXX
+  format %{ "addq    $dst, $src\t# addExact long" %}
+  ins_encode %{
+    __ addq($dst$$Register, $src$$Address);
+  %}
+
+  ins_pipe(ialu_reg_mem);
+%}
+
 instruct addI_rReg(rRegI dst, rRegI src, rFlagsReg cr)
 %{
   match(Set dst (AddI dst src));
@@ -7574,6 +7630,80 @@ instruct subI_mem_imm(memory dst, immI src, rFlagsReg cr)
   ins_pipe(ialu_mem_imm);
 %}
 
+instruct subExactI_rReg(rax_RegI dst, rRegI src, rFlagsReg cr)
+%{
+  match(SubExactI dst src);
+  effect(DEF cr);
+
+  format %{ "subl    $dst, $src\t# subExact int" %}
+  ins_encode %{
+    __ subl($dst$$Register, $src$$Register);
+  %}
+  ins_pipe(ialu_reg_reg);
+%}
+
+instruct subExactI_rReg_imm(rax_RegI dst, immI src, rFlagsReg cr)
+%{
+  match(SubExactI dst src);
+  effect(DEF cr);
+
+  format %{ "subl    $dst, $src\t# subExact int" %}
+  ins_encode %{
+    __ subl($dst$$Register, $src$$constant);
+  %}
+  ins_pipe(ialu_reg_reg);
+%}
+
+instruct subExactI_rReg_mem(rax_RegI dst, memory src, rFlagsReg cr)
+%{
+  match(SubExactI dst (LoadI src));
+  effect(DEF cr);
+
+  ins_cost(125);
+  format %{ "subl    $dst, $src\t# subExact int" %}
+  ins_encode %{
+    __ subl($dst$$Register, $src$$Address);
+  %}
+  ins_pipe(ialu_reg_mem);
+%}
+
+instruct subExactL_rReg(rax_RegL dst, rRegL src, rFlagsReg cr)
+%{
+  match(SubExactL dst src);
+  effect(DEF cr);
+
+  format %{ "subq    $dst, $src\t# subExact long" %}
+  ins_encode %{
+    __ subq($dst$$Register, $src$$Register);
+  %}
+  ins_pipe(ialu_reg_reg);
+%}
+
+instruct subExactL_rReg_imm(rax_RegL dst, immL32 src, rFlagsReg cr)
+%{
+  match(SubExactL dst (LoadL src));
+  effect(DEF cr);
+
+  format %{ "subq    $dst, $src\t# subExact long" %}
+  ins_encode %{
+    __ subq($dst$$Register, $src$$constant);
+  %}
+  ins_pipe(ialu_reg_reg);
+%}
+
+instruct subExactL_rReg_mem(rax_RegI dst, memory src, rFlagsReg cr)
+%{
+  match(SubExactI dst src);
+  effect(DEF cr);
+
+  ins_cost(125);
+  format %{ "subq    $dst, $src\t# subExact long" %}
+  ins_encode %{
+    __ subq($dst$$Register, $src$$Address);
+  %}
+  ins_pipe(ialu_reg_mem);
+%}
+
 instruct subL_rReg(rRegL dst, rRegL src, rFlagsReg cr)
 %{
   match(Set dst (SubL dst src));
@@ -7690,6 +7820,30 @@ instruct negL_mem(memory dst, immL0 zero, rFlagsReg cr)
   ins_pipe(ialu_reg);
 %}
 
+instruct negExactI_rReg(rax_RegI dst, rFlagsReg cr)
+%{
+  match(NegExactI dst);
+  effect(KILL cr);
+
+  format %{ "negl    $dst\t# negExact int" %}
+  ins_encode %{
+    __ negl($dst$$Register);
+  %}
+  ins_pipe(ialu_reg);
+%}
+
+instruct negExactL_rReg(rax_RegL dst, rFlagsReg cr)
+%{
+  match(NegExactL dst);
+  effect(KILL cr);
+
+  format %{ "negq    $dst\t# negExact long" %}
+  ins_encode %{
+    __ negq($dst$$Register);
+  %}
+  ins_pipe(ialu_reg);
+%}
+
 
 //----------Multiplication/Division Instructions-------------------------------
 // Integer Multiplication Instructions
@@ -7807,6 +7961,86 @@ instruct mulHiL_rReg(rdx_RegL dst, no_rax_RegL src, rax_RegL rax, rFlagsReg cr)
   ins_pipe(ialu_reg_reg_alu0);
 %}
 
+
+instruct mulExactI_rReg(rax_RegI dst, rRegI src, rFlagsReg cr)
+%{
+  match(MulExactI dst src);
+  effect(DEF cr);
+
+  ins_cost(300);
+  format %{ "imull   $dst, $src\t# mulExact int" %}
+  ins_encode %{
+    __ imull($dst$$Register, $src$$Register);
+  %}
+  ins_pipe(ialu_reg_reg_alu0);
+%}
+
+
+instruct mulExactI_rReg_imm(rax_RegI dst, rRegI src, immI imm, rFlagsReg cr)
+%{
+  match(MulExactI src imm);
+  effect(DEF cr);
+
+  ins_cost(300);
+  format %{ "imull   $dst, $src, $imm\t# mulExact int" %}
+  ins_encode %{
+    __ imull($dst$$Register, $src$$Register, $imm$$constant);
+  %}
+  ins_pipe(ialu_reg_reg_alu0);
+%}
+
+instruct mulExactI_rReg_mem(rax_RegI dst, memory src, rFlagsReg cr)
+%{
+  match(MulExactI dst (LoadI src));
+  effect(DEF cr);
+
+  ins_cost(350);
+  format %{ "imull   $dst, $src\t# mulExact int" %}
+  ins_encode %{
+    __ imull($dst$$Register, $src$$Address);
+  %}
+  ins_pipe(ialu_reg_mem_alu0);
+%}
+
+instruct mulExactL_rReg(rax_RegL dst, rRegL src, rFlagsReg cr)
+%{
+  match(MulExactL dst src);
+  effect(DEF cr);
+
+  ins_cost(300);
+  format %{ "imulq   $dst, $src\t# mulExact long" %}
+  ins_encode %{
+    __ imulq($dst$$Register, $src$$Register);
+  %}
+  ins_pipe(ialu_reg_reg_alu0);
+%}
+
+instruct mulExactL_rReg_imm(rax_RegL dst, rRegL src, immL32 imm, rFlagsReg cr)
+%{
+  match(MulExactL src imm);
+  effect(DEF cr);
+
+  ins_cost(300);
+  format %{ "imulq   $dst, $src, $imm\t# mulExact long" %}
+  ins_encode %{
+    __ imulq($dst$$Register, $src$$Register, $imm$$constant);
+  %}
+  ins_pipe(ialu_reg_reg_alu0);
+%}
+
+instruct mulExactL_rReg_mem(rax_RegL dst, memory src, rFlagsReg cr)
+%{
+  match(MulExactL dst (LoadL src));
+  effect(DEF cr);
+
+  ins_cost(350);
+  format %{ "imulq   $dst, $src\t# mulExact long" %}
+  ins_encode %{
+    __ imulq($dst$$Register, $src$$Address);
+  %}
+  ins_pipe(ialu_reg_mem_alu0);
+%}
+
 instruct divI_rReg(rax_RegI rax, rdx_RegI rdx, no_rax_rdx_RegI div,
                    rFlagsReg cr)
 %{

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

@@ -4746,6 +4746,10 @@ int os::fork_and_exec(char* cmd) {
 // as libawt.so, and renamed libawt_xawt.so
 //
 bool os::is_headless_jre() {
+#ifdef __APPLE__
+    // We no longer build headless-only on Mac OS X
+    return false;
+#else
     struct stat statbuf;
     char buf[MAXPATHLEN];
     char libmawtpath[MAXPATHLEN];
@@ -4777,6 +4781,7 @@ bool os::is_headless_jre() {
     if (::stat(libmawtpath, &statbuf) == 0) return false;
 
     return true;
+#endif
 }
 
 // Get the default path to the core file

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

@@ -1193,6 +1193,13 @@ void ArchDesc::buildMustCloneMap(FILE *fp_hpp, FILE *fp_cpp) {
          || strcmp(idealName,"FastLock") == 0
          || strcmp(idealName,"FastUnlock") == 0
          || strcmp(idealName,"AddExactI") == 0
+         || strcmp(idealName,"AddExactL") == 0
+         || strcmp(idealName,"SubExactI") == 0
+         || strcmp(idealName,"SubExactL") == 0
+         || strcmp(idealName,"MulExactI") == 0
+         || strcmp(idealName,"MulExactL") == 0
+         || strcmp(idealName,"NegExactI") == 0
+         || strcmp(idealName,"NegExactL") == 0
          || strcmp(idealName,"FlagsProj") == 0
          || strcmp(idealName,"Bool") == 0
          || strcmp(idealName,"Binary") == 0 ) {

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

@@ -536,12 +536,6 @@ bool InstructForm::rematerialize(FormDict &globals, RegisterForm *registers ) {
   if( data_type != Form::none )
     rematerialize = true;
 
-  // Ugly: until a better fix is implemented, disable rematerialization for
-  // negD nodes because they are proved to be problematic.
-  if (is_ideal_negD()) {
-    return false;
-  }
-
   // Constants
   if( _components.count() == 1 && _components[0]->is(Component::USE_DEF) )
     rematerialize = true;

hotspot/src/share/vm/c1/c1_Compilation.hpp

@@ -238,7 +238,18 @@ class Compilation: public StackObj {
     return env()->comp_level() == CompLevel_full_profile &&
       C1UpdateMethodData && C1ProfileCheckcasts;
   }
-
+  bool profile_parameters() {
+    return env()->comp_level() == CompLevel_full_profile &&
+      C1UpdateMethodData && MethodData::profile_parameters();
+  }
+  bool profile_arguments() {
+    return env()->comp_level() == CompLevel_full_profile &&
+      C1UpdateMethodData && MethodData::profile_arguments();
+  }
+  bool profile_return() {
+    return env()->comp_level() == CompLevel_full_profile &&
+      C1UpdateMethodData && MethodData::profile_return();
+  }
   // will compilation make optimistic assumptions that might lead to
   // deoptimization and that the runtime will account for?
   bool is_optimistic() const                             {

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

@@ -1470,7 +1470,7 @@ void GraphBuilder::method_return(Value x) {
     set_state(state()->caller_state()->copy_for_parsing());
     if (x != NULL) {
       state()->push(x->type(), x);
-      if (profile_calls() && MethodData::profile_return() && x->type()->is_object_kind()) {
+      if (profile_return() && x->type()->is_object_kind()) {
         ciMethod* caller = state()->scope()->method();
         ciMethodData* md = caller->method_data_or_null();
         ciProfileData* data = md->bci_to_data(invoke_bci);
@@ -1672,15 +1672,23 @@ Dependencies* GraphBuilder::dependency_recorder() const {
 }
 
 // How many arguments do we want to profile?
-Values* GraphBuilder::args_list_for_profiling(int& start, bool may_have_receiver) {
+Values* GraphBuilder::args_list_for_profiling(ciMethod* target, int& start, bool may_have_receiver) {
   int n = 0;
-  assert(start == 0, "should be initialized");
+  bool has_receiver = may_have_receiver && Bytecodes::has_receiver(method()->java_code_at_bci(bci()));
-  if (MethodData::profile_arguments()) {
+  start = has_receiver ? 1 : 0;
+  if (profile_arguments()) {
     ciProfileData* data = method()->method_data()->bci_to_data(bci());
     if (data->is_CallTypeData() || data->is_VirtualCallTypeData()) {
       n = data->is_CallTypeData() ? data->as_CallTypeData()->number_of_arguments() : data->as_VirtualCallTypeData()->number_of_arguments();
-      bool has_receiver = may_have_receiver && Bytecodes::has_receiver(method()->java_code_at_bci(bci()));
+    }
-      start = has_receiver ? 1 : 0;
+  }
+  // If we are inlining then we need to collect arguments to profile parameters for the target
+  if (profile_parameters() && target != NULL) {
+    if (target->method_data() != NULL && target->method_data()->parameters_type_data() != NULL) {
+      // The receiver is profiled on method entry so it's included in
+      // the number of parameters but here we're only interested in
+      // actual arguments.
+      n = MAX2(n, target->method_data()->parameters_type_data()->number_of_parameters() - start);
     }
   }
   if (n > 0) {
@@ -1690,9 +1698,9 @@ Values* GraphBuilder::args_list_for_profiling(int& start, bool may_have_receiver
 }
 
 // Collect arguments that we want to profile in a list
-Values* GraphBuilder::collect_args_for_profiling(Values* args, bool may_have_receiver) {
+Values* GraphBuilder::collect_args_for_profiling(Values* args, ciMethod* target, bool may_have_receiver) {
   int start = 0;
-  Values* obj_args = args_list_for_profiling(start, may_have_receiver);
+  Values* obj_args = args_list_for_profiling(target, start, may_have_receiver);
   if (obj_args == NULL) {
     return NULL;
   }
@@ -2006,7 +2014,7 @@ void GraphBuilder::invoke(Bytecodes::Code code) {
       } else if (exact_target != NULL) {
         target_klass = exact_target->holder();
       }
-      profile_call(target, recv, target_klass, collect_args_for_profiling(args, false), false);
+      profile_call(target, recv, target_klass, collect_args_for_profiling(args, NULL, false), false);
     }
   }
 
@@ -2021,7 +2029,7 @@ void GraphBuilder::invoke(Bytecodes::Code code) {
       push(result_type, result);
     }
   }
-  if (profile_calls() && MethodData::profile_return() && result_type->is_object_kind()) {
+  if (profile_return() && result_type->is_object_kind()) {
     profile_return_type(result, target);
   }
 }
@@ -3561,7 +3569,7 @@ bool GraphBuilder::try_inline_intrinsics(ciMethod* callee) {
           recv = args->at(0);
           null_check(recv);
         }
-        profile_call(callee, recv, NULL, collect_args_for_profiling(args, true), true);
+        profile_call(callee, recv, NULL, collect_args_for_profiling(args, callee, true), true);
       }
     }
   }
@@ -3572,7 +3580,7 @@ bool GraphBuilder::try_inline_intrinsics(ciMethod* callee) {
   Value value = append_split(result);
   if (result_type != voidType) push(result_type, value);
 
-  if (callee != method() && profile_calls() && MethodData::profile_return() && result_type->is_object_kind()) {
+  if (callee != method() && profile_return() && result_type->is_object_kind()) {
     profile_return_type(result, callee);
   }
 
@@ -3760,6 +3768,7 @@ bool GraphBuilder::try_inline_full(ciMethod* callee, bool holder_known, Bytecode
 
   // now perform tests that are based on flag settings
   if (callee->force_inline()) {
+    if (inline_level() > MaxForceInlineLevel) INLINE_BAILOUT("MaxForceInlineLevel");
     print_inlining(callee, "force inline by annotation");
   } else if (callee->should_inline()) {
     print_inlining(callee, "force inline by CompileOracle");
@@ -3820,7 +3829,7 @@ bool GraphBuilder::try_inline_full(ciMethod* callee, bool holder_known, Bytecode
 
     if (profile_calls()) {
       int start = 0;
-      Values* obj_args = args_list_for_profiling(start, has_receiver);
+      Values* obj_args = args_list_for_profiling(callee, start, has_receiver);
       if (obj_args != NULL) {
         int s = obj_args->size();
         // if called through method handle invoke, some arguments may have been popped

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

@@ -386,9 +386,12 @@ class GraphBuilder VALUE_OBJ_CLASS_SPEC {
   bool profile_calls()         { return _compilation->profile_calls();         }
   bool profile_inlined_calls() { return _compilation->profile_inlined_calls(); }
   bool profile_checkcasts()    { return _compilation->profile_checkcasts();    }
+  bool profile_parameters()    { return _compilation->profile_parameters();    }
+  bool profile_arguments()     { return _compilation->profile_arguments();     }
+  bool profile_return()        { return _compilation->profile_return();        }
 
-  Values* args_list_for_profiling(int& start, bool may_have_receiver);
+  Values* args_list_for_profiling(ciMethod* target, int& start, bool may_have_receiver);
-  Values* collect_args_for_profiling(Values* args, bool may_have_receiver);
+  Values* collect_args_for_profiling(Values* args, ciMethod* target, bool may_have_receiver);
 
  public:
   NOT_PRODUCT(void print_stats();)

hotspot/src/share/vm/c1/c1_LIR.cpp

@@ -183,10 +183,10 @@ char LIR_OprDesc::type_char(BasicType t) {
     case T_LONG:
     case T_OBJECT:
     case T_ADDRESS:
-    case T_METADATA:
     case T_VOID:
       return ::type2char(t);
-
+    case T_METADATA:
+      return 'M';
     case T_ILLEGAL:
       return '?';
 

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

@@ -1175,7 +1175,7 @@ void LIRGenerator::do_Return(Return* x) {
   if (compilation()->env()->dtrace_method_probes()) {
     BasicTypeList signature;
     signature.append(LP64_ONLY(T_LONG) NOT_LP64(T_INT));    // thread
-    signature.append(T_OBJECT); // Method*
+    signature.append(T_METADATA); // Method*
     LIR_OprList* args = new LIR_OprList();
     args->append(getThreadPointer());
     LIR_Opr meth = new_register(T_METADATA);
@@ -1265,6 +1265,7 @@ void LIRGenerator::do_getClass(Intrinsic* x) {
 
   LIRItem rcvr(x->argument_at(0), this);
   rcvr.load_item();
+  LIR_Opr temp = new_register(T_METADATA);
   LIR_Opr result = rlock_result(x);
 
   // need to perform the null check on the rcvr
@@ -1272,8 +1273,11 @@ void LIRGenerator::do_getClass(Intrinsic* x) {
   if (x->needs_null_check()) {
     info = state_for(x);
   }
-  __ move(new LIR_Address(rcvr.result(), oopDesc::klass_offset_in_bytes(), T_ADDRESS), result, info);
+
-  __ move_wide(new LIR_Address(result, in_bytes(Klass::java_mirror_offset()), T_OBJECT), result);
+  // FIXME T_ADDRESS should actually be T_METADATA but it can't because the
+  // meaning of these two is mixed up (see JDK-8026837).
+  __ move(new LIR_Address(rcvr.result(), oopDesc::klass_offset_in_bytes(), T_ADDRESS), temp, info);
+  __ move_wide(new LIR_Address(temp, in_bytes(Klass::java_mirror_offset()), T_OBJECT), result);
 }
 
 
@@ -2643,6 +2647,39 @@ ciKlass* LIRGenerator::profile_arg_type(ciMethodData* md, int md_base_offset, in
   return result;
 }
 
+// profile parameters on entry to the root of the compilation
+void LIRGenerator::profile_parameters(Base* x) {
+  if (compilation()->profile_parameters()) {
+    CallingConvention* args = compilation()->frame_map()->incoming_arguments();
+    ciMethodData* md = scope()->method()->method_data_or_null();
+    assert(md != NULL, "Sanity");
+
+    if (md->parameters_type_data() != NULL) {
+      ciParametersTypeData* parameters_type_data = md->parameters_type_data();
+      ciTypeStackSlotEntries* parameters =  parameters_type_data->parameters();
+      LIR_Opr mdp = LIR_OprFact::illegalOpr;
+      for (int java_index = 0, i = 0, j = 0; j < parameters_type_data->number_of_parameters(); i++) {
+        LIR_Opr src = args->at(i);
+        assert(!src->is_illegal(), "check");
+        BasicType t = src->type();
+        if (t == T_OBJECT || t == T_ARRAY) {
+          intptr_t profiled_k = parameters->type(j);
+          Local* local = x->state()->local_at(java_index)->as_Local();
+          ciKlass* exact = profile_arg_type(md, md->byte_offset_of_slot(parameters_type_data, ParametersTypeData::type_offset(0)),
+                                            in_bytes(ParametersTypeData::type_offset(j)) - in_bytes(ParametersTypeData::type_offset(0)),
+                                            profiled_k, local, mdp, false, local->declared_type()->as_klass());
+          // If the profile is known statically set it once for all and do not emit any code
+          if (exact != NULL) {
+            md->set_parameter_type(j, exact);
+          }
+          j++;
+        }
+        java_index += type2size[t];
+      }
+    }
+  }
+}
+
 void LIRGenerator::do_Base(Base* x) {
   __ std_entry(LIR_OprFact::illegalOpr);
   // Emit moves from physical registers / stack slots to virtual registers
@@ -2683,7 +2720,7 @@ void LIRGenerator::do_Base(Base* x) {
   if (compilation()->env()->dtrace_method_probes()) {
     BasicTypeList signature;
     signature.append(LP64_ONLY(T_LONG) NOT_LP64(T_INT));    // thread
-    signature.append(T_OBJECT); // Method*
+    signature.append(T_METADATA); // Method*
     LIR_OprList* args = new LIR_OprList();
     args->append(getThreadPointer());
     LIR_Opr meth = new_register(T_METADATA);
@@ -2718,6 +2755,7 @@ void LIRGenerator::do_Base(Base* x) {
 
   // increment invocation counters if needed
   if (!method()->is_accessor()) { // Accessors do not have MDOs, so no counting.
+    profile_parameters(x);
     CodeEmitInfo* info = new CodeEmitInfo(scope()->start()->state()->copy(ValueStack::StateBefore, SynchronizationEntryBCI), NULL, false);
     increment_invocation_counter(info);
   }
@@ -3077,11 +3115,12 @@ void LIRGenerator::do_Intrinsic(Intrinsic* x) {
 }
 
 void LIRGenerator::profile_arguments(ProfileCall* x) {
-  if (MethodData::profile_arguments()) {
+  if (compilation()->profile_arguments()) {
     int bci = x->bci_of_invoke();
     ciMethodData* md = x->method()->method_data_or_null();
     ciProfileData* data = md->bci_to_data(bci);
-    if (data->is_CallTypeData() || data->is_VirtualCallTypeData()) {
+    if ((data->is_CallTypeData() && data->as_CallTypeData()->has_arguments()) ||
+        (data->is_VirtualCallTypeData() && data->as_VirtualCallTypeData()->has_arguments())) {
       ByteSize extra = data->is_CallTypeData() ? CallTypeData::args_data_offset() : VirtualCallTypeData::args_data_offset();
       int base_offset = md->byte_offset_of_slot(data, extra);
       LIR_Opr mdp = LIR_OprFact::illegalOpr;
@@ -3107,6 +3146,71 @@ void LIRGenerator::profile_arguments(ProfileCall* x) {
           md->set_argument_type(bci, i, exact);
         }
       }
+    } else {
+#ifdef ASSERT
+      Bytecodes::Code code = x->method()->raw_code_at_bci(x->bci_of_invoke());
+      int n = x->nb_profiled_args();
+      assert(MethodData::profile_parameters() && x->inlined() &&
+             ((code == Bytecodes::_invokedynamic && n <= 1) || (code == Bytecodes::_invokehandle && n <= 2)),
+             "only at JSR292 bytecodes");
+#endif
+    }
+  }
+}
+
+// profile parameters on entry to an inlined method
+void LIRGenerator::profile_parameters_at_call(ProfileCall* x) {
+  if (compilation()->profile_parameters() && x->inlined()) {
+    ciMethodData* md = x->callee()->method_data_or_null();
+    if (md != NULL) {
+      ciParametersTypeData* parameters_type_data = md->parameters_type_data();
+      if (parameters_type_data != NULL) {
+        ciTypeStackSlotEntries* parameters =  parameters_type_data->parameters();
+        LIR_Opr mdp = LIR_OprFact::illegalOpr;
+        bool has_receiver = !x->callee()->is_static();
+        ciSignature* sig = x->callee()->signature();
+        ciSignatureStream sig_stream(sig, has_receiver ? x->callee()->holder() : NULL);
+        int i = 0; // to iterate on the Instructions
+        Value arg = x->recv();
+        bool not_null = false;
+        int bci = x->bci_of_invoke();
+        Bytecodes::Code bc = x->method()->java_code_at_bci(bci);
+        // The first parameter is the receiver so that's what we start
+        // with if it exists. On exception if method handle call to
+        // virtual method has receiver in the args list
+        if (arg == NULL || !Bytecodes::has_receiver(bc)) {
+          i = 1;
+          arg = x->profiled_arg_at(0);
+          not_null = !x->arg_needs_null_check(0);
+        }
+        int k = 0; // to iterate on the profile data
+        for (;;) {
+          intptr_t profiled_k = parameters->type(k);
+          ciKlass* exact = profile_arg_type(md, md->byte_offset_of_slot(parameters_type_data, ParametersTypeData::type_offset(0)),
+                                            in_bytes(ParametersTypeData::type_offset(k)) - in_bytes(ParametersTypeData::type_offset(0)),
+                                            profiled_k, arg, mdp, not_null, sig_stream.next_klass());
+          // If the profile is known statically set it once for all and do not emit any code
+          if (exact != NULL) {
+            md->set_parameter_type(k, exact);
+          }
+          k++;
+          if (k >= parameters_type_data->number_of_parameters()) {
+#ifdef ASSERT
+            int extra = 0;
+            if (MethodData::profile_arguments() && TypeProfileParmsLimit != -1 &&
+                x->nb_profiled_args() >= TypeProfileParmsLimit &&
+                x->recv() != NULL && Bytecodes::has_receiver(bc)) {
+              extra += 1;
+            }
+            assert(i == x->nb_profiled_args() - extra || (TypeProfileParmsLimit != -1 && TypeProfileArgsLimit > TypeProfileParmsLimit), "unused parameters?");
+#endif
+            break;
+          }
+          arg = x->profiled_arg_at(i);
+          not_null = !x->arg_needs_null_check(i);
+          i++;
+        }
+      }
     }
   }
 }
@@ -3122,6 +3226,11 @@ void LIRGenerator::do_ProfileCall(ProfileCall* x) {
     profile_arguments(x);
   }
 
+  // profile parameters on inlined method entry including receiver
+  if (x->recv() != NULL || x->nb_profiled_args() > 0) {
+    profile_parameters_at_call(x);
+  }
+
   if (x->recv() != NULL) {
     LIRItem value(x->recv(), this);
     value.load_item();
@@ -3222,7 +3331,7 @@ void LIRGenerator::do_RuntimeCall(RuntimeCall* x) {
   BasicTypeList* signature = new BasicTypeList(x->number_of_arguments());
 
   if (x->pass_thread()) {
-    signature->append(T_ADDRESS);
+    signature->append(LP64_ONLY(T_LONG) NOT_LP64(T_INT));    // thread
     args->append(getThreadPointer());
   }
 

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

@@ -436,6 +436,8 @@ class LIRGenerator: public InstructionVisitor, public BlockClosure {
 #endif
   ciKlass* profile_arg_type(ciMethodData* md, int md_first_offset, int md_offset, intptr_t profiled_k, Value arg, LIR_Opr& mdp, bool not_null, ciKlass* signature_k);
   void profile_arguments(ProfileCall* x);
+  void profile_parameters(Base* x);
+  void profile_parameters_at_call(ProfileCall* x);
 
  public:
   Compilation*  compilation() const              { return _compilation; }

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

@@ -75,9 +75,9 @@
 
 // Map BasicType to spill size in 32-bit words, matching VMReg's notion of words
 #ifdef _LP64
-static int type2spill_size[T_CONFLICT+1]={ -1, 0, 0, 0, 1, 1, 1, 2, 1, 1, 1, 2, 2, 2, 0, 1, -1};
+static int type2spill_size[T_CONFLICT+1]={ -1, 0, 0, 0, 1, 1, 1, 2, 1, 1, 1, 2, 2, 2, 0, 2,  1, 2, 1, -1};
 #else
-static int type2spill_size[T_CONFLICT+1]={ -1, 0, 0, 0, 1, 1, 1, 2, 1, 1, 1, 2, 1, 1, 0, 1, -1};
+static int type2spill_size[T_CONFLICT+1]={ -1, 0, 0, 0, 1, 1, 1, 2, 1, 1, 1, 2, 1, 1, 0, 1, -1, 1, 1, -1};
 #endif
 
 

hotspot/src/share/vm/c1/c1_globals.hpp

@@ -341,6 +341,8 @@
   diagnostic(bool, C1PatchInvokeDynamic, true,                              \
              "Patch invokedynamic appendix not known at compile time")      \
                                                                             \
+  develop(intx, MaxForceInlineLevel, 100,                                   \
+          "maximum number of nested @ForceInline calls that are inlined")   \
                                                                             \
 
 

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

@@ -483,8 +483,7 @@ ciKlass* ciEnv::get_klass_by_index_impl(constantPoolHandle cpool,
     {
       // We have to lock the cpool to keep the oop from being resolved
       // while we are accessing it.
-      oop cplock = cpool->lock();
+        MonitorLockerEx ml(cpool->lock());
-      ObjectLocker ol(cplock, THREAD, cplock != NULL);
       constantTag tag = cpool->tag_at(index);
       if (tag.is_klass()) {
         // The klass has been inserted into the constant pool

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

@@ -565,6 +565,116 @@ void ciCallProfile::add_receiver(ciKlass* receiver, int receiver_count) {
   if (_limit < MorphismLimit) _limit++;
 }
 
+
+void ciMethod::assert_virtual_call_type_ok(int bci) {
+  assert(java_code_at_bci(bci) == Bytecodes::_invokevirtual ||
+         java_code_at_bci(bci) == Bytecodes::_invokeinterface, err_msg("unexpected bytecode %s", Bytecodes::name(java_code_at_bci(bci))));
+}
+
+void ciMethod::assert_call_type_ok(int bci) {
+  assert(java_code_at_bci(bci) == Bytecodes::_invokestatic ||
+         java_code_at_bci(bci) == Bytecodes::_invokespecial ||
+         java_code_at_bci(bci) == Bytecodes::_invokedynamic, err_msg("unexpected bytecode %s", Bytecodes::name(java_code_at_bci(bci))));
+}
+
+/**
+ * Check whether profiling provides a type for the argument i to the
+ * call at bci bci
+ *
+ * @param bci  bci of the call
+ * @param i    argument number
+ * @return     profiled type
+ *
+ * If the profile reports that the argument may be null, return false
+ * at least for now.
+ */
+ciKlass* ciMethod::argument_profiled_type(int bci, int i) {
+  if (MethodData::profile_parameters() && method_data() != NULL && method_data()->is_mature()) {
+    ciProfileData* data = method_data()->bci_to_data(bci);
+    if (data != NULL) {
+      if (data->is_VirtualCallTypeData()) {
+        assert_virtual_call_type_ok(bci);
+        ciVirtualCallTypeData* call = (ciVirtualCallTypeData*)data->as_VirtualCallTypeData();
+        if (i >= call->number_of_arguments()) {
+          return NULL;
+        }
+        ciKlass* type = call->valid_argument_type(i);
+        if (type != NULL && !call->argument_maybe_null(i)) {
+          return type;
+        }
+      } else if (data->is_CallTypeData()) {
+        assert_call_type_ok(bci);
+        ciCallTypeData* call = (ciCallTypeData*)data->as_CallTypeData();
+        if (i >= call->number_of_arguments()) {
+          return NULL;
+        }
+        ciKlass* type = call->valid_argument_type(i);
+        if (type != NULL && !call->argument_maybe_null(i)) {
+          return type;
+        }
+      }
+    }
+  }
+  return NULL;
+}
+
+/**
+ * Check whether profiling provides a type for the return value from
+ * the call at bci bci
+ *
+ * @param bci  bci of the call
+ * @return     profiled type
+ *
+ * If the profile reports that the argument may be null, return false
+ * at least for now.
+ */
+ciKlass* ciMethod::return_profiled_type(int bci) {
+  if (MethodData::profile_return() && method_data() != NULL && method_data()->is_mature()) {
+    ciProfileData* data = method_data()->bci_to_data(bci);
+    if (data != NULL) {
+      if (data->is_VirtualCallTypeData()) {
+        assert_virtual_call_type_ok(bci);
+        ciVirtualCallTypeData* call = (ciVirtualCallTypeData*)data->as_VirtualCallTypeData();
+        ciKlass* type = call->valid_return_type();
+        if (type != NULL && !call->return_maybe_null()) {
+          return type;
+        }
+      } else if (data->is_CallTypeData()) {
+        assert_call_type_ok(bci);
+        ciCallTypeData* call = (ciCallTypeData*)data->as_CallTypeData();
+        ciKlass* type = call->valid_return_type();
+        if (type != NULL && !call->return_maybe_null()) {
+          return type;
+        }
+      }
+    }
+  }
+  return NULL;
+}
+
+/**
+ * Check whether profiling provides a type for the parameter i
+ *
+ * @param i    parameter number
+ * @return     profiled type
+ *
+ * If the profile reports that the argument may be null, return false
+ * at least for now.
+ */
+ciKlass* ciMethod::parameter_profiled_type(int i) {
+  if (MethodData::profile_parameters() && method_data() != NULL && method_data()->is_mature()) {
+    ciParametersTypeData* parameters = method_data()->parameters_type_data();
+    if (parameters != NULL && i < parameters->number_of_parameters()) {
+      ciKlass* type = parameters->valid_parameter_type(i);
+      if (type != NULL && !parameters->parameter_maybe_null(i)) {
+        return type;
+      }
+    }
+  }
+  return NULL;
+}
+
+
 // ------------------------------------------------------------------
 // ciMethod::find_monomorphic_target
 //

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

@@ -117,6 +117,10 @@ class ciMethod : public ciMetadata {
     *bcp = code;
   }
 
+  // Check bytecode and profile data collected are compatible
+  void assert_virtual_call_type_ok(int bci);
+  void assert_call_type_ok(int bci);
+
  public:
   // Basic method information.
   ciFlags flags() const                          { check_is_loaded(); return _flags; }
@@ -230,6 +234,11 @@ class ciMethod : public ciMetadata {
   ciCallProfile call_profile_at_bci(int bci);
   int           interpreter_call_site_count(int bci);
 
+  // Does type profiling provide a useful type at this point?
+  ciKlass*      argument_profiled_type(int bci, int i);
+  ciKlass*      parameter_profiled_type(int i);
+  ciKlass*      return_profiled_type(int bci);
+
   ciField*      get_field_at_bci( int bci, bool &will_link);
   ciMethod*     get_method_at_bci(int bci, bool &will_link, ciSignature* *declared_signature);
 

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

@@ -53,6 +53,7 @@ ciMethodData::ciMethodData(MethodData* md) : ciMetadata(md) {
   _hint_di = first_di();
   // Initialize the escape information (to "don't know.");
   _eflags = _arg_local = _arg_stack = _arg_returned = 0;
+  _parameters = NULL;
 }
 
 // ------------------------------------------------------------------
@@ -74,6 +75,7 @@ ciMethodData::ciMethodData() : ciMetadata(NULL) {
   _hint_di = first_di();
   // Initialize the escape information (to "don't know.");
   _eflags = _arg_local = _arg_stack = _arg_returned = 0;
+  _parameters = NULL;
 }
 
 void ciMethodData::load_data() {
@@ -108,6 +110,12 @@ void ciMethodData::load_data() {
     ci_data = next_data(ci_data);
     data = mdo->next_data(data);
   }
+  if (mdo->parameters_type_data() != NULL) {
+    _parameters = data_layout_at(mdo->parameters_type_data_di());
+    ciParametersTypeData* parameters = new ciParametersTypeData(_parameters);
+    parameters->translate_from(mdo->parameters_type_data());
+  }
+
   // Note:  Extra data are all BitData, and do not need translation.
   _current_mileage = MethodData::mileage_of(mdo->method());
   _invocation_counter = mdo->invocation_count();
@@ -182,6 +190,8 @@ ciProfileData* ciMethodData::data_at(int data_index) {
     return new ciCallTypeData(data_layout);
   case DataLayout::virtual_call_type_data_tag:
     return new ciVirtualCallTypeData(data_layout);
+  case DataLayout::parameters_type_data_tag:
+    return new ciParametersTypeData(data_layout);
   };
 }
 
@@ -318,6 +328,14 @@ void ciMethodData::set_argument_type(int bci, int i, ciKlass* k) {
   }
 }
 
+void ciMethodData::set_parameter_type(int i, ciKlass* k) {
+  VM_ENTRY_MARK;
+  MethodData* mdo = get_MethodData();
+  if (mdo != NULL) {
+    mdo->parameters_type_data()->set_type(i, k->get_Klass());
+  }
+}
+
 void ciMethodData::set_return_type(int bci, ciKlass* k) {
   VM_ENTRY_MARK;
   MethodData* mdo = get_MethodData();
@@ -605,4 +623,9 @@ void ciVirtualCallTypeData::print_data_on(outputStream* st) const {
     ret()->print_data_on(st);
   }
 }
+
+void ciParametersTypeData::print_data_on(outputStream* st) const {
+  st->print_cr("Parametertypes");
+  parameters()->print_data_on(st);
+}
 #endif

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

@@ -43,6 +43,7 @@ class ciMultiBranchData;
 class ciArgInfoData;
 class ciCallTypeData;
 class ciVirtualCallTypeData;
+class ciParametersTypeData;
 
 typedef ProfileData ciProfileData;
 
@@ -99,6 +100,10 @@ public:
     return valid_ciklass(type(i));
   }
 
+  bool maybe_null(int i) const {
+    return was_null_seen(type(i));
+  }
+
 #ifndef PRODUCT
   void print_data_on(outputStream* st) const;
 #endif
@@ -112,6 +117,10 @@ public:
     return valid_ciklass(type());
   }
 
+  bool maybe_null() const {
+    return was_null_seen(type());
+  }
+
 #ifndef PRODUCT
   void print_data_on(outputStream* st) const;
 #endif
@@ -124,7 +133,7 @@ public:
   ciTypeStackSlotEntries* args() const { return (ciTypeStackSlotEntries*)CallTypeData::args(); }
   ciReturnTypeEntry* ret() const { return (ciReturnTypeEntry*)CallTypeData::ret(); }
 
-  void translate_type_data_from(const ProfileData* data) {
+  void translate_from(const ProfileData* data) {
     if (has_arguments()) {
       args()->translate_type_data_from(data->as_CallTypeData()->args());
     }
@@ -153,6 +162,14 @@ public:
     return ret()->valid_type();
   }
 
+  bool argument_maybe_null(int i) const {
+    return args()->maybe_null(i);
+  }
+
+  bool return_maybe_null() const {
+    return ret()->maybe_null();
+  }
+
 #ifndef PRODUCT
   void print_data_on(outputStream* st) const;
 #endif
@@ -259,6 +276,14 @@ public:
     return ret()->valid_type();
   }
 
+  bool argument_maybe_null(int i) const {
+    return args()->maybe_null(i);
+  }
+
+  bool return_maybe_null() const {
+    return ret()->maybe_null();
+  }
+
 #ifndef PRODUCT
   void print_data_on(outputStream* st) const;
 #endif
@@ -290,6 +315,29 @@ public:
   ciArgInfoData(DataLayout* layout) : ArgInfoData(layout) {};
 };
 
+class ciParametersTypeData : public ParametersTypeData {
+public:
+  ciParametersTypeData(DataLayout* layout) : ParametersTypeData(layout) {}
+
+  virtual void translate_from(const ProfileData* data) {
+    parameters()->translate_type_data_from(data->as_ParametersTypeData()->parameters());
+  }
+
+  ciTypeStackSlotEntries* parameters() const { return (ciTypeStackSlotEntries*)ParametersTypeData::parameters(); }
+
+  ciKlass* valid_parameter_type(int i) const {
+    return parameters()->valid_type(i);
+  }
+
+  bool parameter_maybe_null(int i) const {
+    return parameters()->maybe_null(i);
+  }
+
+#ifndef PRODUCT
+  void print_data_on(outputStream* st) const;
+#endif
+};
+
 // ciMethodData
 //
 // This class represents a MethodData* in the HotSpot virtual
@@ -335,6 +383,10 @@ private:
   // Coherent snapshot of original header.
   MethodData _orig;
 
+  // Dedicated area dedicated to parameters. Null if no parameter
+  // profiling for this method.
+  DataLayout* _parameters;
+
   ciMethodData(MethodData* md);
   ciMethodData();
 
@@ -403,6 +455,7 @@ public:
   // If the compiler finds a profiled type that is known statically
   // for sure, set it in the MethodData
   void set_argument_type(int bci, int i, ciKlass* k);
+  void set_parameter_type(int i, ciKlass* k);
   void set_return_type(int bci, ciKlass* k);
 
   void load_data();
@@ -467,6 +520,10 @@ public:
   bool is_arg_returned(int i) const;
   uint arg_modified(int arg) const;
 
+  ciParametersTypeData* parameters_type_data() const {
+    return _parameters != NULL ? new ciParametersTypeData(_parameters) : NULL;
+  }
+
   // Code generation helper
   ByteSize offset_of_slot(ciProfileData* data, ByteSize slot_offset_in_data);
   int      byte_offset_of_slot(ciProfileData* data, ByteSize slot_offset_in_data) { return in_bytes(offset_of_slot(data, slot_offset_in_data)); }

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

@@ -2360,6 +2360,11 @@ methodHandle SystemDictionary::find_method_handle_invoker(Symbol* name,
   objArrayHandle appendix_box = oopFactory::new_objArray(SystemDictionary::Object_klass(), 1, CHECK_(empty));
   assert(appendix_box->obj_at(0) == NULL, "");
 
+  // This should not happen.  JDK code should take care of that.
+  if (accessing_klass.is_null() || method_type.is_null()) {
+    THROW_MSG_(vmSymbols::java_lang_InternalError(), "bad invokehandle", empty);
+  }
+
   // call java.lang.invoke.MethodHandleNatives::linkMethod(... String, MethodType) -> MemberName
   JavaCallArguments args;
   args.push_oop(accessing_klass()->java_mirror());
@@ -2485,6 +2490,9 @@ Handle SystemDictionary::link_method_handle_constant(KlassHandle caller,
   Handle type;
   if (signature->utf8_length() > 0 && signature->byte_at(0) == '(') {
     type = find_method_handle_type(signature, caller, CHECK_(empty));
+  } else if (caller.is_null()) {
+    // This should not happen.  JDK code should take care of that.
+    THROW_MSG_(vmSymbols::java_lang_InternalError(), "bad MH constant", empty);
   } else {
     ResourceMark rm(THREAD);
     SignatureStream ss(signature, false);
@@ -2548,6 +2556,11 @@ methodHandle SystemDictionary::find_dynamic_call_site_invoker(KlassHandle caller
   Handle method_name = java_lang_String::create_from_symbol(name, CHECK_(empty));
   Handle method_type = find_method_handle_type(type, caller, CHECK_(empty));
 
+  // This should not happen.  JDK code should take care of that.
+  if (caller.is_null() || method_type.is_null()) {
+    THROW_MSG_(vmSymbols::java_lang_InternalError(), "bad invokedynamic", empty);
+  }
+
   objArrayHandle appendix_box = oopFactory::new_objArray(SystemDictionary::Object_klass(), 1, CHECK_(empty));
   assert(appendix_box->obj_at(0) == NULL, "");
 

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

@@ -624,6 +624,7 @@
   do_class(java_lang_StrictMath,          "java/lang/StrictMath")                                                       \
   do_signature(double2_double_signature,  "(DD)D")                                                                      \
   do_signature(int2_int_signature,        "(II)I")                                                                      \
+  do_signature(long2_long_signature,      "(JJ)J")                                                                         \
                                                                                                                         \
   /* here are the math names, all together: */                                                                          \
   do_name(abs_name,"abs")       do_name(sin_name,"sin")         do_name(cos_name,"cos")                                 \
@@ -632,8 +633,11 @@
   do_name(exp_name,"exp")       do_name(min_name,"min")         do_name(max_name,"max")                                 \
                                                                                                                         \
   do_name(addExact_name,"addExact")                                                                                     \
-  do_name(subtractExact_name,"subtractExact")                                                                           \
+  do_name(decrementExact_name,"decrementExact")                                                                         \
+  do_name(incrementExact_name,"incrementExact")                                                                         \
   do_name(multiplyExact_name,"multiplyExact")                                                                           \
+  do_name(negateExact_name,"negateExact")                                                                               \
+  do_name(subtractExact_name,"subtractExact")                                                                           \
                                                                                                                         \
   do_intrinsic(_dabs,                     java_lang_Math,         abs_name,   double_double_signature,           F_S)   \
   do_intrinsic(_dsin,                     java_lang_Math,         sin_name,   double_double_signature,           F_S)   \
@@ -647,7 +651,18 @@
   do_intrinsic(_dexp,                     java_lang_Math,         exp_name,   double_double_signature,           F_S)   \
   do_intrinsic(_min,                      java_lang_Math,         min_name,   int2_int_signature,                F_S)   \
   do_intrinsic(_max,                      java_lang_Math,         max_name,   int2_int_signature,                F_S)   \
-  do_intrinsic(_addExact,                 java_lang_Math,         addExact_name, int2_int_signature,             F_S)   \
+  do_intrinsic(_addExactI,                java_lang_Math,         addExact_name, int2_int_signature,             F_S)   \
+  do_intrinsic(_addExactL,                java_lang_Math,         addExact_name, long2_long_signature,           F_S)   \
+  do_intrinsic(_decrementExactI,          java_lang_Math,         decrementExact_name, int_int_signature,        F_S)   \
+  do_intrinsic(_decrementExactL,          java_lang_Math,         decrementExact_name, long2_long_signature,     F_S)   \
+  do_intrinsic(_incrementExactI,          java_lang_Math,         incrementExact_name, int_int_signature,        F_S)   \
+  do_intrinsic(_incrementExactL,          java_lang_Math,         incrementExact_name, long2_long_signature,     F_S)   \
+  do_intrinsic(_multiplyExactI,           java_lang_Math,         multiplyExact_name, int2_int_signature,        F_S)   \
+  do_intrinsic(_multiplyExactL,           java_lang_Math,         multiplyExact_name, long2_long_signature,      F_S)   \
+  do_intrinsic(_negateExactI,             java_lang_Math,         negateExact_name, int_int_signature,           F_S)   \
+  do_intrinsic(_negateExactL,             java_lang_Math,         negateExact_name, long_long_signature,         F_S)   \
+  do_intrinsic(_subtractExactI,           java_lang_Math,         subtractExact_name, int2_int_signature,        F_S)   \
+  do_intrinsic(_subtractExactL,           java_lang_Math,         subtractExact_name, long2_long_signature,      F_S)   \
                                                                                                                         \
   do_intrinsic(_floatToRawIntBits,        java_lang_Float,        floatToRawIntBits_name,   float_int_signature, F_S)   \
    do_name(     floatToRawIntBits_name,                          "floatToRawIntBits")                                   \

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

@@ -1297,13 +1297,6 @@ nmethod* CompileBroker::compile_method(methodHandle method, int osr_bci,
     method->jmethod_id();
   }
 
-  // If the compiler is shut off due to code cache getting full
-  // fail out now so blocking compiles dont hang the java thread
-  if (!should_compile_new_jobs()) {
-    CompilationPolicy::policy()->delay_compilation(method());
-    return NULL;
-  }
-
   // do the compilation
   if (method->is_native()) {
     if (!PreferInterpreterNativeStubs || method->is_method_handle_intrinsic()) {
@@ -1313,11 +1306,22 @@ nmethod* CompileBroker::compile_method(methodHandle method, int osr_bci,
         MutexLocker locker(MethodCompileQueue_lock, THREAD);
         compile_id = assign_compile_id(method, standard_entry_bci);
       }
+      // To properly handle the appendix argument for out-of-line calls we are using a small trampoline that
+      // pops off the appendix argument and jumps to the target (see gen_special_dispatch in SharedRuntime).
+      //
+      // Since normal compiled-to-compiled calls are not able to handle such a thing we MUST generate an adapter
+      // in this case.  If we can't generate one and use it we can not execute the out-of-line method handle calls.
       (void) AdapterHandlerLibrary::create_native_wrapper(method, compile_id);
     } else {
       return NULL;
     }
   } else {
+    // If the compiler is shut off due to code cache getting full
+    // fail out now so blocking compiles dont hang the java thread
+    if (!should_compile_new_jobs()) {
+      CompilationPolicy::policy()->delay_compilation(method());
+      return NULL;
+    }
     compile_method_base(method, osr_bci, comp_level, hot_method, hot_count, comment, THREAD);
   }
 

hotspot/src/share/vm/gc_implementation/g1/g1AllocRegion.hpp

@@ -55,7 +55,7 @@ private:
   // then _alloc_region is NULL and this object should not be used to
   // satisfy allocation requests (it was done this way to force the
   // correct use of init() and release()).
-  HeapRegion* _alloc_region;
+  HeapRegion* volatile _alloc_region;
 
   // It keeps track of the distinct number of regions that are used
   // for allocation in the active interval of this object, i.e.,
@@ -132,8 +132,9 @@ public:
   static void setup(G1CollectedHeap* g1h, HeapRegion* dummy_region);
 
   HeapRegion* get() const {
+    HeapRegion * hr = _alloc_region;
     // Make sure that the dummy region does not escape this class.
-    return (_alloc_region == _dummy_region) ? NULL : _alloc_region;
+    return (hr == _dummy_region) ? NULL : hr;
   }
 
   uint count() { return _count; }

hotspot/src/share/vm/gc_implementation/g1/g1RemSetSummary.cpp

@@ -187,19 +187,23 @@ public:
   size_t code_root_elems() const { return _code_root_elems; }
 
   void print_rs_mem_info_on(outputStream * out, size_t total) {
-    out->print_cr("    %8dK (%5.1f%%) by %zd %s regions", round_to_K(rs_mem_size()), rs_mem_size_percent_of(total), amount(), _name);
+    out->print_cr("    "SIZE_FORMAT_W(8)"K (%5.1f%%) by "SIZE_FORMAT" %s regions",
+        round_to_K(rs_mem_size()), rs_mem_size_percent_of(total), amount(), _name);
   }
 
   void print_cards_occupied_info_on(outputStream * out, size_t total) {
-    out->print_cr("     %8d (%5.1f%%) entries by %zd %s regions", cards_occupied(), cards_occupied_percent_of(total), amount(), _name);
+    out->print_cr("     "SIZE_FORMAT_W(8)" (%5.1f%%) entries by "SIZE_FORMAT" %s regions",
+        cards_occupied(), cards_occupied_percent_of(total), amount(), _name);
   }
 
   void print_code_root_mem_info_on(outputStream * out, size_t total) {
-    out->print_cr("    %8dK (%5.1f%%) by %zd %s regions", round_to_K(code_root_mem_size()), code_root_mem_size_percent_of(total), amount(), _name);
+    out->print_cr("    "SIZE_FORMAT_W(8)"K (%5.1f%%) by "SIZE_FORMAT" %s regions",
+        round_to_K(code_root_mem_size()), code_root_mem_size_percent_of(total), amount(), _name);
   }
 
   void print_code_root_elems_info_on(outputStream * out, size_t total) {
-    out->print_cr("     %8d (%5.1f%%) elements by %zd %s regions", code_root_elems(), code_root_elems_percent_of(total), amount(), _name);
+    out->print_cr("     "SIZE_FORMAT_W(8)" (%5.1f%%) elements by "SIZE_FORMAT" %s regions",
+        code_root_elems(), code_root_elems_percent_of(total), amount(), _name);
   }
 };
 
@@ -327,14 +331,14 @@ void G1RemSetSummary::print_on(outputStream* out) {
   out->print_cr("\n Recent concurrent refinement statistics");
   out->print_cr("  Processed "SIZE_FORMAT" cards",
                 num_concurrent_refined_cards());
-  out->print_cr("  Of %d completed buffers:", num_processed_buf_total());
+  out->print_cr("  Of "SIZE_FORMAT" completed buffers:", num_processed_buf_total());
-  out->print_cr("     %8d (%5.1f%%) by concurrent RS threads.",
+  out->print_cr("     "SIZE_FORMAT_W(8)" (%5.1f%%) by concurrent RS threads.",
                 num_processed_buf_total(),
                 percent_of(num_processed_buf_rs_threads(), num_processed_buf_total()));
-  out->print_cr("     %8d (%5.1f%%) by mutator threads.",
+  out->print_cr("     "SIZE_FORMAT_W(8)" (%5.1f%%) by mutator threads.",
                 num_processed_buf_mutator(),
                 percent_of(num_processed_buf_mutator(), num_processed_buf_total()));
-  out->print_cr("  Did %d coarsenings.", num_coarsenings());
+  out->print_cr("  Did "SIZE_FORMAT" coarsenings.", num_coarsenings());
   out->print_cr("  Concurrent RS threads times (s)");
   out->print("     ");
   for (uint i = 0; i < _num_vtimes; i++) {

hotspot/src/share/vm/interpreter/abstractInterpreter.hpp

@@ -30,11 +30,8 @@
 #include "runtime/thread.inline.hpp"
 #include "runtime/vmThread.hpp"
 #include "utilities/top.hpp"
-#ifdef TARGET_ARCH_MODEL_x86_32
+#ifdef TARGET_ARCH_x86
-# include "interp_masm_x86_32.hpp"
+# include "interp_masm_x86.hpp"
-#endif
-#ifdef TARGET_ARCH_MODEL_x86_64
-# include "interp_masm_x86_64.hpp"
 #endif
 #ifdef TARGET_ARCH_MODEL_sparc
 # include "interp_masm_sparc.hpp"

hotspot/src/share/vm/interpreter/linkResolver.cpp

@@ -1,5 +1,4 @@
 /*
- * Copyright (c) 1997, 2013, Oracle and/or its affiliates. All rights reserved.
  * Copyright (c) 1997, 2013, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
@@ -265,7 +264,7 @@ void LinkResolver::lookup_method_in_klasses(methodHandle& result, KlassHandle kl
 void LinkResolver::lookup_instance_method_in_klasses(methodHandle& result, KlassHandle klass, Symbol* name, Symbol* signature, TRAPS) {
   Method* result_oop = klass->uncached_lookup_method(name, signature);
   result = methodHandle(THREAD, result_oop);
-  while (!result.is_null() && result->is_static()) {
+  while (!result.is_null() && result->is_static() && result->method_holder()->super() != NULL) {
     klass = KlassHandle(THREAD, result->method_holder()->super());
     result = methodHandle(THREAD, klass->uncached_lookup_method(name, signature));
   }
@@ -419,18 +418,28 @@ void LinkResolver::check_method_accessability(KlassHandle ref_klass,
 
   AccessFlags flags = sel_method->access_flags();
 
-  // Special case:  arrays always override "clone". JVMS 2.15.
+  // Special case #1:  arrays always override "clone". JVMS 2.15.
   // If the resolved klass is an array class, and the declaring class
   // is java.lang.Object and the method is "clone", set the flags
   // to public.
+  // Special case #2:  If the resolved klass is an interface, and
+  // the declaring class is java.lang.Object, and the method is
+  // "clone" or "finalize", set the flags to public. If the
+  // resolved interface does not contain "clone" or "finalize"
+  // methods, the method/interface method resolution looks to
+  // the interface's super class, java.lang.Object.  With JDK 8
+  // interface accessability check requirement, special casing
+  // this scenario is necessary to avoid an IAE.
   //
-  // We'll check for the method name first, as that's most likely
+  // We'll check for each method name first and then java.lang.Object
-  // to be false (so we'll short-circuit out of these tests).
+  // to best short-circuit out of these tests.
-  if (sel_method->name() == vmSymbols::clone_name() &&
+  if (((sel_method->name() == vmSymbols::clone_name() &&
-      sel_klass() == SystemDictionary::Object_klass() &&
+        (resolved_klass->oop_is_array() || resolved_klass->is_interface())) ||
-      resolved_klass->oop_is_array()) {
+       (sel_method->name() == vmSymbols::finalize_method_name() &&
+        resolved_klass->is_interface())) &&
+      sel_klass() == SystemDictionary::Object_klass()) {
     // We need to change "protected" to "public".
-    assert(flags.is_protected(), "clone not protected?");
+    assert(flags.is_protected(), "clone or finalize not protected?");
     jint new_flags = flags.as_int();
     new_flags = new_flags & (~JVM_ACC_PROTECTED);
     new_flags = new_flags | JVM_ACC_PUBLIC;

hotspot/src/share/vm/interpreter/templateTable.hpp

@@ -28,11 +28,8 @@
 #include "interpreter/bytecodes.hpp"
 #include "memory/allocation.hpp"
 #include "runtime/frame.hpp"
-#ifdef TARGET_ARCH_MODEL_x86_32
+#ifdef TARGET_ARCH_x86
-# include "interp_masm_x86_32.hpp"
+# include "interp_masm_x86.hpp"
-#endif
-#ifdef TARGET_ARCH_MODEL_x86_64
-# include "interp_masm_x86_64.hpp"
 #endif
 #ifdef TARGET_ARCH_MODEL_sparc
 # include "interp_masm_sparc.hpp"

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

@@ -75,8 +75,7 @@ enum ChunkSizes {    // in words.
   ClassSmallChunk = 256,
   SmallChunk = 512,
   ClassMediumChunk = 4 * K,
-  MediumChunk = 8 * K,
+  MediumChunk = 8 * K
-  HumongousChunkGranularity = 8
 };
 
 static ChunkIndex next_chunk_index(ChunkIndex i) {
@@ -92,6 +91,7 @@ typedef class FreeList<Metachunk> ChunkList;
 
 // Manages the global free lists of chunks.
 class ChunkManager : public CHeapObj<mtInternal> {
+  friend class TestVirtualSpaceNodeTest;
 
   // Free list of chunks of different sizes.
   //   SpecializedChunk
@@ -257,6 +257,8 @@ class VirtualSpaceNode : public CHeapObj<mtClass> {
   // VirtualSpace
   Metachunk* first_chunk() { return (Metachunk*) bottom(); }
 
+  // Committed but unused space in the virtual space
+  size_t free_words_in_vs() const;
  public:
 
   VirtualSpaceNode(size_t byte_size);
@@ -301,7 +303,6 @@ class VirtualSpaceNode : public CHeapObj<mtClass> {
   // used and capacity in this single entry in the list
   size_t used_words_in_vs() const;
   size_t capacity_words_in_vs() const;
-  size_t free_words_in_vs() const;
 
   bool initialize();
 
@@ -319,6 +320,13 @@ class VirtualSpaceNode : public CHeapObj<mtClass> {
   // in the node from any freelist.
   void purge(ChunkManager* chunk_manager);
 
+  // If an allocation doesn't fit in the current node a new node is created.
+  // Allocate chunks out of the remaining committed space in this node
+  // to avoid wasting that memory.
+  // This always adds up because all the chunk sizes are multiples of
+  // the smallest chunk size.
+  void retire(ChunkManager* chunk_manager);
+
 #ifdef ASSERT
   // Debug support
   void mangle();
@@ -461,6 +469,10 @@ class VirtualSpaceList : public CHeapObj<mtClass> {
   // and is typically followed by the allocation of a chunk.
   bool create_new_virtual_space(size_t vs_word_size);
 
+  // Chunk up the unused committed space in the current
+  // virtual space and add the chunks to the free list.
+  void retire_current_virtual_space();
+
  public:
   VirtualSpaceList(size_t word_size);
   VirtualSpaceList(ReservedSpace rs);
@@ -624,11 +636,13 @@ class SpaceManager : public CHeapObj<mtClass> {
   bool is_class() { return _mdtype == Metaspace::ClassType; }
 
   // Accessors
-  size_t specialized_chunk_size() { return SpecializedChunk; }
+  size_t specialized_chunk_size() { return (size_t) is_class() ? ClassSpecializedChunk : SpecializedChunk; }
   size_t small_chunk_size()       { return (size_t) is_class() ? ClassSmallChunk : SmallChunk; }
   size_t medium_chunk_size()      { return (size_t) is_class() ? ClassMediumChunk : MediumChunk; }
   size_t medium_chunk_bunch()     { return medium_chunk_size() * MediumChunkMultiple; }
 
+  size_t smallest_chunk_size()  { return specialized_chunk_size(); }
+
   size_t allocated_blocks_words() const { return _allocated_blocks_words; }
   size_t allocated_blocks_bytes() const { return _allocated_blocks_words * BytesPerWord; }
   size_t allocated_chunks_words() const { return _allocated_chunks_words; }
@@ -1056,6 +1070,35 @@ void VirtualSpaceList::purge(ChunkManager* chunk_manager) {
 #endif
 }
 
+void VirtualSpaceList::retire_current_virtual_space() {
+  assert_lock_strong(SpaceManager::expand_lock());
+
+  VirtualSpaceNode* vsn = current_virtual_space();
+
+  ChunkManager* cm = is_class() ? Metaspace::chunk_manager_class() :
+                                  Metaspace::chunk_manager_metadata();
+
+  vsn->retire(cm);
+}
+
+void VirtualSpaceNode::retire(ChunkManager* chunk_manager) {
+  for (int i = (int)MediumIndex; i >= (int)ZeroIndex; --i) {
+    ChunkIndex index = (ChunkIndex)i;
+    size_t chunk_size = chunk_manager->free_chunks(index)->size();
+
+    while (free_words_in_vs() >= chunk_size) {
+      DEBUG_ONLY(verify_container_count();)
+      Metachunk* chunk = get_chunk_vs(chunk_size);
+      assert(chunk != NULL, "allocation should have been successful");
+
+      chunk_manager->return_chunks(index, chunk);
+      chunk_manager->inc_free_chunks_total(chunk_size);
+      DEBUG_ONLY(verify_container_count();)
+    }
+  }
+  assert(free_words_in_vs() == 0, "should be empty now");
+}
+
 VirtualSpaceList::VirtualSpaceList(size_t word_size) :
                                    _is_class(false),
                                    _virtual_space_list(NULL),
@@ -1181,6 +1224,7 @@ bool VirtualSpaceList::expand_by(size_t min_words, size_t preferred_words) {
   if (vs_expanded) {
     return true;
   }
+  retire_current_virtual_space();
 
   // Get another virtual space.
   size_t grow_vs_words = MAX2((size_t)VirtualSpaceSize, preferred_words);
@@ -1902,12 +1946,12 @@ size_t SpaceManager::calc_chunk_size(size_t word_size) {
     chunk_word_size = medium_chunk_size();
   }
 
-  // Might still need a humongous chunk.  Enforce an
+  // Might still need a humongous chunk.  Enforce
-  // eight word granularity to facilitate reuse (some
+  // humongous allocations sizes to be aligned up to
-  // wastage but better chance of reuse).
+  // the smallest chunk size.
   size_t if_humongous_sized_chunk =
     align_size_up(word_size + Metachunk::overhead(),
-                  HumongousChunkGranularity);
+                  smallest_chunk_size());
   chunk_word_size =
     MAX2((size_t) chunk_word_size, if_humongous_sized_chunk);
 
@@ -2151,10 +2195,10 @@ SpaceManager::~SpaceManager() {
     }
     assert(humongous_chunks->word_size() == (size_t)
            align_size_up(humongous_chunks->word_size(),
-                             HumongousChunkGranularity),
+                             smallest_chunk_size()),
            err_msg("Humongous chunk size is wrong: word size " SIZE_FORMAT
                    " granularity %d",
-                   humongous_chunks->word_size(), HumongousChunkGranularity));
+                   humongous_chunks->word_size(), smallest_chunk_size()));
     Metachunk* next_humongous_chunks = humongous_chunks->next();
     humongous_chunks->container()->dec_container_count();
     chunk_manager()->humongous_dictionary()->return_chunk(humongous_chunks);
@@ -3301,9 +3345,7 @@ MetaWord* Metaspace::allocate(ClassLoaderData* loader_data, size_t word_size,
   }
 
   if (result == NULL) {
-    report_metadata_oome(loader_data, word_size, mdtype, THREAD);
+    report_metadata_oome(loader_data, word_size, mdtype, CHECK_NULL);
-    // Will not reach here.
-    return NULL;
   }
 
   // Zero initialize.
@@ -3494,4 +3536,94 @@ void TestMetaspaceAux_test() {
   TestMetaspaceAuxTest::test();
 }
 
+class TestVirtualSpaceNodeTest {
+  static void chunk_up(size_t words_left, size_t& num_medium_chunks,
+                                          size_t& num_small_chunks,
+                                          size_t& num_specialized_chunks) {
+    num_medium_chunks = words_left / MediumChunk;
+    words_left = words_left % MediumChunk;
+
+    num_small_chunks = words_left / SmallChunk;
+    words_left = words_left % SmallChunk;
+    // how many specialized chunks can we get?
+    num_specialized_chunks = words_left / SpecializedChunk;
+    assert(words_left % SpecializedChunk == 0, "should be nothing left");
+  }
+
+ public:
+  static void test() {
+    MutexLockerEx ml(SpaceManager::expand_lock(), Mutex::_no_safepoint_check_flag);
+    const size_t vsn_test_size_words = MediumChunk  * 4;
+    const size_t vsn_test_size_bytes = vsn_test_size_words * BytesPerWord;
+
+    // The chunk sizes must be multiples of eachother, or this will fail
+    STATIC_ASSERT(MediumChunk % SmallChunk == 0);
+    STATIC_ASSERT(SmallChunk % SpecializedChunk == 0);
+
+    { // No committed memory in VSN
+      ChunkManager cm(SpecializedChunk, SmallChunk, MediumChunk);
+      VirtualSpaceNode vsn(vsn_test_size_bytes);
+      vsn.initialize();
+      vsn.retire(&cm);
+      assert(cm.sum_free_chunks_count() == 0, "did not commit any memory in the VSN");
+    }
+
+    { // All of VSN is committed, half is used by chunks
+      ChunkManager cm(SpecializedChunk, SmallChunk, MediumChunk);
+      VirtualSpaceNode vsn(vsn_test_size_bytes);
+      vsn.initialize();
+      vsn.expand_by(vsn_test_size_words, vsn_test_size_words);
+      vsn.get_chunk_vs(MediumChunk);
+      vsn.get_chunk_vs(MediumChunk);
+      vsn.retire(&cm);
+      assert(cm.sum_free_chunks_count() == 2, "should have been memory left for 2 medium chunks");
+      assert(cm.sum_free_chunks() == 2*MediumChunk, "sizes should add up");
+    }
+
+    { // 4 pages of VSN is committed, some is used by chunks
+      ChunkManager cm(SpecializedChunk, SmallChunk, MediumChunk);
+      VirtualSpaceNode vsn(vsn_test_size_bytes);
+      const size_t page_chunks = 4 * (size_t)os::vm_page_size() / BytesPerWord;
+      assert(page_chunks < MediumChunk, "Test expects medium chunks to be at least 4*page_size");
+      vsn.initialize();
+      vsn.expand_by(page_chunks, page_chunks);
+      vsn.get_chunk_vs(SmallChunk);
+      vsn.get_chunk_vs(SpecializedChunk);
+      vsn.retire(&cm);
+
+      // committed - used = words left to retire
+      const size_t words_left = page_chunks - SmallChunk - SpecializedChunk;
+
+      size_t num_medium_chunks, num_small_chunks, num_spec_chunks;
+      chunk_up(words_left, num_medium_chunks, num_small_chunks, num_spec_chunks);
+
+      assert(num_medium_chunks == 0, "should not get any medium chunks");
+      assert(cm.sum_free_chunks_count() == (num_small_chunks + num_spec_chunks), "should be space for 3 chunks");
+      assert(cm.sum_free_chunks() == words_left, "sizes should add up");
+    }
+
+    { // Half of VSN is committed, a humongous chunk is used
+      ChunkManager cm(SpecializedChunk, SmallChunk, MediumChunk);
+      VirtualSpaceNode vsn(vsn_test_size_bytes);
+      vsn.initialize();
+      vsn.expand_by(MediumChunk * 2, MediumChunk * 2);
+      vsn.get_chunk_vs(MediumChunk + SpecializedChunk); // Humongous chunks will be aligned up to MediumChunk + SpecializedChunk
+      vsn.retire(&cm);
+
+      const size_t words_left = MediumChunk * 2 - (MediumChunk + SpecializedChunk);
+      size_t num_medium_chunks, num_small_chunks, num_spec_chunks;
+      chunk_up(words_left, num_medium_chunks, num_small_chunks, num_spec_chunks);
+
+      assert(num_medium_chunks == 0, "should not get any medium chunks");
+      assert(cm.sum_free_chunks_count() == (num_small_chunks + num_spec_chunks), "should be space for 3 chunks");
+      assert(cm.sum_free_chunks() == words_left, "sizes should add up");
+    }
+
+  }
+};
+
+void TestVirtualSpaceNode_test() {
+  TestVirtualSpaceNodeTest::test();
+}
+
 #endif

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

@@ -40,7 +40,6 @@
 #include "runtime/init.hpp"
 #include "runtime/javaCalls.hpp"
 #include "runtime/signature.hpp"
-#include "runtime/synchronizer.hpp"
 #include "runtime/vframe.hpp"
 
 ConstantPool* ConstantPool::allocate(ClassLoaderData* loader_data, int length, TRAPS) {
@@ -70,6 +69,7 @@ ConstantPool::ConstantPool(Array<u1>* tags) {
 
   // only set to non-zero if constant pool is merged by RedefineClasses
   set_version(0);
+  set_lock(new Monitor(Monitor::nonleaf + 2, "A constant pool lock"));
 
   // initialize tag array
   int length = tags->length();
@@ -95,6 +95,9 @@ void ConstantPool::deallocate_contents(ClassLoaderData* loader_data) {
 void ConstantPool::release_C_heap_structures() {
   // walk constant pool and decrement symbol reference counts
   unreference_symbols();
+
+  delete _lock;
+  set_lock(NULL);
 }
 
 objArrayOop ConstantPool::resolved_references() const {
@@ -151,6 +154,9 @@ void ConstantPool::restore_unshareable_info(TRAPS) {
       ClassLoaderData* loader_data = pool_holder()->class_loader_data();
       set_resolved_references(loader_data->add_handle(refs_handle));
     }
+
+    // Also need to recreate the mutex.  Make sure this matches the constructor
+    set_lock(new Monitor(Monitor::nonleaf + 2, "A constant pool lock"));
   }
 }
 
@@ -161,23 +167,7 @@ void ConstantPool::remove_unshareable_info() {
   set_resolved_reference_length(
     resolved_references() != NULL ? resolved_references()->length() : 0);
   set_resolved_references(NULL);
-}
+  set_lock(NULL);
-
-oop ConstantPool::lock() {
-  if (_pool_holder) {
-    // We re-use the _pool_holder's init_lock to reduce footprint.
-    // Notes on deadlocks:
-    // [1] This lock is a Java oop, so it can be recursively locked by
-    //     the same thread without self-deadlocks.
-    // [2] Deadlock will happen if there is circular dependency between
-    //     the <clinit> of two Java classes. However, in this case,
-    //     the deadlock would have happened long before we reach
-    //     ConstantPool::lock(), so reusing init_lock does not
-    //     increase the possibility of deadlock.
-    return _pool_holder->init_lock();
-  } else {
-    return NULL;
-  }
 }
 
 int ConstantPool::cp_to_object_index(int cp_index) {
@@ -211,9 +201,7 @@ Klass* ConstantPool::klass_at_impl(constantPoolHandle this_oop, int which, TRAPS
 
   Symbol* name = NULL;
   Handle       loader;
-  {
+  {  MonitorLockerEx ml(this_oop->lock());
-    oop cplock = this_oop->lock();
-    ObjectLocker ol(cplock , THREAD, cplock != NULL);
 
     if (this_oop->tag_at(which).is_unresolved_klass()) {
       if (this_oop->tag_at(which).is_unresolved_klass_in_error()) {
@@ -260,8 +248,7 @@ Klass* ConstantPool::klass_at_impl(constantPoolHandle this_oop, int which, TRAPS
 
       bool throw_orig_error = false;
       {
-        oop cplock = this_oop->lock();
+        MonitorLockerEx ml(this_oop->lock());
-        ObjectLocker ol(cplock, THREAD, cplock != NULL);
 
         // some other thread has beaten us and has resolved the class.
         if (this_oop->tag_at(which).is_klass()) {
@@ -329,8 +316,7 @@ Klass* ConstantPool::klass_at_impl(constantPoolHandle this_oop, int which, TRAPS
       }
       return k();
     } else {
-      oop cplock = this_oop->lock();
+      MonitorLockerEx ml(this_oop->lock());
-      ObjectLocker ol(cplock, THREAD, cplock != NULL);
       // Only updated constant pool - if it is resolved.
       do_resolve = this_oop->tag_at(which).is_unresolved_klass();
       if (do_resolve) {
@@ -600,8 +586,7 @@ void ConstantPool::save_and_throw_exception(constantPoolHandle this_oop, int whi
                                      int tag, TRAPS) {
   ResourceMark rm;
   Symbol* error = PENDING_EXCEPTION->klass()->name();
-  oop cplock = this_oop->lock();
+  MonitorLockerEx ml(this_oop->lock());  // lock cpool to change tag.
-  ObjectLocker ol(cplock, THREAD, cplock != NULL);  // lock cpool to change tag.
 
   int error_tag = (tag == JVM_CONSTANT_MethodHandle) ?
            JVM_CONSTANT_MethodHandleInError : JVM_CONSTANT_MethodTypeInError;
@@ -762,8 +747,7 @@ oop ConstantPool::resolve_constant_at_impl(constantPoolHandle this_oop, int inde
   if (cache_index >= 0) {
     // Cache the oop here also.
     Handle result_handle(THREAD, result_oop);
-    oop cplock = this_oop->lock();
+    MonitorLockerEx ml(this_oop->lock());  // don't know if we really need this
-    ObjectLocker ol(cplock, THREAD, cplock != NULL);  // don't know if we really need this
     oop result = this_oop->resolved_references()->obj_at(cache_index);
     // Benign race condition:  resolved_references may already be filled in while we were trying to lock.
     // The important thing here is that all threads pick up the same result.

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

@@ -111,6 +111,7 @@ class ConstantPool : public Metadata {
     int                _version;
   } _saved;
 
+  Monitor*             _lock;
 
   void set_tags(Array<u1>* tags)               { _tags = tags; }
   void tag_at_put(int which, jbyte t)          { tags()->at_put(which, t); }
@@ -843,17 +844,8 @@ class ConstantPool : public Metadata {
 
   void set_resolved_reference_length(int length) { _saved._resolved_reference_length = length; }
   int  resolved_reference_length() const  { return _saved._resolved_reference_length; }
-
+  void set_lock(Monitor* lock)            { _lock = lock; }
-  // lock() may return null -- constant pool updates may happen before this lock is
+  Monitor* lock()                         { return _lock; }
-  // initialized, because the _pool_holder has not been fully initialized and
-  // has not been registered into the system dictionary. In this case, no other
-  // thread can be modifying this constantpool, so no synchronization is
-  // necessary.
-  //
-  // Use cplock() like this:
-  //    oop cplock = cp->lock();
-  //    ObjectLocker ol(cplock , THREAD, cplock != NULL);
-  oop lock();
 
   // Decrease ref counts of symbols that are in the constant pool
   // when the holder class is unloaded

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

@@ -284,8 +284,7 @@ void ConstantPoolCacheEntry::set_method_handle_common(constantPoolHandle cpool,
   // the lock, so that when the losing writer returns, he can use the linked
   // cache entry.
 
-  oop cplock = cpool->lock();
+  MonitorLockerEx ml(cpool->lock());
-  ObjectLocker ol(cplock, Thread::current(), cplock != NULL);
   if (!is_f1_null()) {
     return;
   }

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

@@ -498,13 +498,27 @@ objArrayOop InstanceKlass::signers() const {
 
 oop InstanceKlass::init_lock() const {
   // return the init lock from the mirror
-  return java_lang_Class::init_lock(java_mirror());
+  oop lock = java_lang_Class::init_lock(java_mirror());
+  assert((oop)lock != NULL || !is_not_initialized(), // initialized or in_error state
+         "only fully initialized state can have a null lock");
+  return lock;
+}
+
+// Set the initialization lock to null so the object can be GC'ed.  Any racing
+// threads to get this lock will see a null lock and will not lock.
+// That's okay because they all check for initialized state after getting
+// the lock and return.
+void InstanceKlass::fence_and_clear_init_lock() {
+  // make sure previous stores are all done, notably the init_state.
+  OrderAccess::storestore();
+  java_lang_Class::set_init_lock(java_mirror(), NULL);
+  assert(!is_not_initialized(), "class must be initialized now");
 }
 
 void InstanceKlass::eager_initialize_impl(instanceKlassHandle this_oop) {
   EXCEPTION_MARK;
   oop init_lock = this_oop->init_lock();
-  ObjectLocker ol(init_lock, THREAD);
+  ObjectLocker ol(init_lock, THREAD, init_lock != NULL);
 
   // abort if someone beat us to the initialization
   if (!this_oop->is_not_initialized()) return;  // note: not equivalent to is_initialized()
@@ -523,6 +537,7 @@ void InstanceKlass::eager_initialize_impl(instanceKlassHandle this_oop) {
   } else {
     // linking successfull, mark class as initialized
     this_oop->set_init_state (fully_initialized);
+    this_oop->fence_and_clear_init_lock();
     // trace
     if (TraceClassInitialization) {
       ResourceMark rm(THREAD);
@@ -649,7 +664,7 @@ bool InstanceKlass::link_class_impl(
   // verification & rewriting
   {
     oop init_lock = this_oop->init_lock();
-    ObjectLocker ol(init_lock, THREAD);
+    ObjectLocker ol(init_lock, THREAD, init_lock != NULL);
     // rewritten will have been set if loader constraint error found
     // on an earlier link attempt
     // don't verify or rewrite if already rewritten
@@ -772,7 +787,7 @@ void InstanceKlass::initialize_impl(instanceKlassHandle this_oop, TRAPS) {
   // Step 1
   {
     oop init_lock = this_oop->init_lock();
-    ObjectLocker ol(init_lock, THREAD);
+    ObjectLocker ol(init_lock, THREAD, init_lock != NULL);
 
     Thread *self = THREAD; // it's passed the current thread
 
@@ -920,8 +935,9 @@ void InstanceKlass::set_initialization_state_and_notify(ClassState state, TRAPS)
 
 void InstanceKlass::set_initialization_state_and_notify_impl(instanceKlassHandle this_oop, ClassState state, TRAPS) {
   oop init_lock = this_oop->init_lock();
-  ObjectLocker ol(init_lock, THREAD);
+  ObjectLocker ol(init_lock, THREAD, init_lock != NULL);
   this_oop->set_init_state(state);
+  this_oop->fence_and_clear_init_lock();
   ol.notify_all(CHECK);
 }
 

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

@@ -1023,6 +1023,7 @@ public:
   // It has to be an object not a Mutex because it's held through java calls.
   oop init_lock() const;
 private:
+  void fence_and_clear_init_lock();
 
   // Static methods that are used to implement member methods where an exposed this pointer
   // is needed due to possible GCs

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

@@ -41,7 +41,7 @@
 
 // Some types of data layouts need a length field.
 bool DataLayout::needs_array_len(u1 tag) {
-  return (tag == multi_branch_data_tag) || (tag == arg_info_data_tag);
+  return (tag == multi_branch_data_tag) || (tag == arg_info_data_tag) || (tag == parameters_type_data_tag);
 }
 
 // Perform generic initialization of the data.  More specific
@@ -156,10 +156,13 @@ void JumpData::print_data_on(outputStream* st) const {
 }
 #endif // !PRODUCT
 
-int TypeStackSlotEntries::compute_cell_count(Symbol* signature, int max) {
+int TypeStackSlotEntries::compute_cell_count(Symbol* signature, bool include_receiver, int max) {
+  // Parameter profiling include the receiver
+  int args_count = include_receiver ? 1 : 0;
   ResourceMark rm;
   SignatureStream ss(signature);
-  int args_count = MIN2(ss.reference_parameter_count(), max);
+  args_count += ss.reference_parameter_count();
+  args_count = MIN2(args_count, max);
   return args_count * per_arg_cell_count;
 }
 
@@ -169,7 +172,7 @@ int TypeEntriesAtCall::compute_cell_count(BytecodeStream* stream) {
   Bytecode_invoke inv(stream->method(), stream->bci());
   int args_cell = 0;
   if (arguments_profiling_enabled()) {
-    args_cell = TypeStackSlotEntries::compute_cell_count(inv.signature(), TypeProfileArgsLimit);
+    args_cell = TypeStackSlotEntries::compute_cell_count(inv.signature(), false, TypeProfileArgsLimit);
   }
   int ret_cell = 0;
   if (return_profiling_enabled() && (inv.result_type() == T_OBJECT || inv.result_type() == T_ARRAY)) {
@@ -212,12 +215,19 @@ public:
   int off_at(int i) const { return _offsets.at(i); }
 };
 
-void TypeStackSlotEntries::post_initialize(Symbol* signature, bool has_receiver) {
+void TypeStackSlotEntries::post_initialize(Symbol* signature, bool has_receiver, bool include_receiver) {
   ResourceMark rm;
-  ArgumentOffsetComputer aos(signature, _number_of_entries);
+  int start = 0;
+  // Parameter profiling include the receiver
+  if (include_receiver && has_receiver) {
+    set_stack_slot(0, 0);
+    set_type(0, type_none());
+    start += 1;
+  }
+  ArgumentOffsetComputer aos(signature, _number_of_entries-start);
   aos.total();
-  for (int i = 0; i < _number_of_entries; i++) {
+  for (int i = start; i < _number_of_entries; i++) {
-    set_stack_slot(i, aos.off_at(i) + (has_receiver ? 1 : 0));
+    set_stack_slot(i, aos.off_at(i-start) + (has_receiver ? 1 : 0));
     set_type(i, type_none());
   }
 }
@@ -234,7 +244,7 @@ void CallTypeData::post_initialize(BytecodeStream* stream, MethodData* mdo) {
     assert(count > 0, "room for args type but none found?");
     check_number_of_arguments(count);
 #endif
-    _args.post_initialize(inv.signature(), inv.has_receiver());
+    _args.post_initialize(inv.signature(), inv.has_receiver(), false);
   }
 
   if (has_return()) {
@@ -255,7 +265,7 @@ void VirtualCallTypeData::post_initialize(BytecodeStream* stream, MethodData* md
     assert(count > 0, "room for args type but none found?");
     check_number_of_arguments(count);
 #endif
-    _args.post_initialize(inv.signature(), inv.has_receiver());
+    _args.post_initialize(inv.signature(), inv.has_receiver(), false);
   }
 
   if (has_return()) {
@@ -579,6 +589,34 @@ void ArgInfoData::print_data_on(outputStream* st) const {
 }
 
 #endif
+
+int ParametersTypeData::compute_cell_count(Method* m) {
+  if (!MethodData::profile_parameters_for_method(m)) {
+    return 0;
+  }
+  int max = TypeProfileParmsLimit == -1 ? INT_MAX : TypeProfileParmsLimit;
+  int obj_args = TypeStackSlotEntries::compute_cell_count(m->signature(), !m->is_static(), max);
+  if (obj_args > 0) {
+    return obj_args + 1; // 1 cell for array len
+  }
+  return 0;
+}
+
+void ParametersTypeData::post_initialize(BytecodeStream* stream, MethodData* mdo) {
+  _parameters.post_initialize(mdo->method()->signature(), !mdo->method()->is_static(), true);
+}
+
+bool ParametersTypeData::profiling_enabled() {
+  return MethodData::profile_parameters();
+}
+
+#ifndef PRODUCT
+void ParametersTypeData::print_data_on(outputStream* st) const {
+  st->print("parameter types");
+  _parameters.print_data_on(st);
+}
+#endif
+
 // ==================================================================
 // MethodData*
 //
@@ -741,6 +779,12 @@ int MethodData::compute_allocation_size_in_bytes(methodHandle method) {
   int arg_size = method->size_of_parameters();
   object_size += DataLayout::compute_size_in_bytes(arg_size+1);
 
+  // Reserve room for an area of the MDO dedicated to profiling of
+  // parameters
+  int args_cell = ParametersTypeData::compute_cell_count(method());
+  if (args_cell > 0) {
+    object_size += DataLayout::compute_size_in_bytes(args_cell);
+  }
   return object_size;
 }
 
@@ -915,6 +959,8 @@ ProfileData* DataLayout::data_in() {
     return new CallTypeData(this);
   case DataLayout::virtual_call_type_data_tag:
     return new VirtualCallTypeData(this);
+  case DataLayout::parameters_type_data_tag:
+    return new ParametersTypeData(this);
   };
 }
 
@@ -936,6 +982,9 @@ void MethodData::post_initialize(BytecodeStream* stream) {
     stream->next();
     data->post_initialize(stream, this);
   }
+  if (_parameters_type_data_di != -1) {
+    parameters_type_data()->post_initialize(NULL, this);
+  }
 }
 
 // Initialize the MethodData* corresponding to a given method.
@@ -975,7 +1024,23 @@ MethodData::MethodData(methodHandle method, int size, TRAPS) {
   int arg_size = method->size_of_parameters();
   dp->initialize(DataLayout::arg_info_data_tag, 0, arg_size+1);
 
-  object_size += extra_size + DataLayout::compute_size_in_bytes(arg_size+1);
+  int arg_data_size = DataLayout::compute_size_in_bytes(arg_size+1);
+  object_size += extra_size + arg_data_size;
+
+  int args_cell = ParametersTypeData::compute_cell_count(method());
+  // If we are profiling parameters, we reserver an area near the end
+  // of the MDO after the slots for bytecodes (because there's no bci
+  // for method entry so they don't fit with the framework for the
+  // profiling of bytecodes). We store the offset within the MDO of
+  // this area (or -1 if no parameter is profiled)
+  if (args_cell > 0) {
+    object_size += DataLayout::compute_size_in_bytes(args_cell);
+    _parameters_type_data_di = data_size + extra_size + arg_data_size;
+    DataLayout *dp = data_layout_at(data_size + extra_size + arg_data_size);
+    dp->initialize(DataLayout::parameters_type_data_tag, 0, args_cell);
+  } else {
+    _parameters_type_data_di = -1;
+  }
 
   // Set an initial hint. Don't use set_hint_di() because
   // first_di() may be out of bounds if data_size is 0.
@@ -1134,6 +1199,9 @@ void MethodData::print_value_on(outputStream* st) const {
 void MethodData::print_data_on(outputStream* st) const {
   ResourceMark rm;
   ProfileData* data = first_data();
+  if (_parameters_type_data_di != -1) {
+    parameters_type_data()->print_data_on(st);
+  }
   for ( ; is_valid(data); data = next_data(data)) {
     st->print("%d", dp_to_di(data->dp()));
     st->fill_to(6);
@@ -1222,7 +1290,7 @@ bool MethodData::profile_arguments_for_invoke(methodHandle m, int bci) {
 }
 
 int MethodData::profile_return_flag() {
-  return TypeProfileLevel / 10;
+  return (TypeProfileLevel % 100) / 10;
 }
 
 bool MethodData::profile_return() {
@@ -1249,3 +1317,32 @@ bool MethodData::profile_return_for_invoke(methodHandle m, int bci) {
   assert(profile_return_jsr292_only(), "inconsistent");
   return profile_jsr292(m, bci);
 }
+
+int MethodData::profile_parameters_flag() {
+  return TypeProfileLevel / 100;
+}
+
+bool MethodData::profile_parameters() {
+  return profile_parameters_flag() > no_type_profile && profile_parameters_flag() <= type_profile_all;
+}
+
+bool MethodData::profile_parameters_jsr292_only() {
+  return profile_parameters_flag() == type_profile_jsr292;
+}
+
+bool MethodData::profile_all_parameters() {
+  return profile_parameters_flag() == type_profile_all;
+}
+
+bool MethodData::profile_parameters_for_method(methodHandle m) {
+  if (!profile_parameters()) {
+    return false;
+  }
+
+  if (profile_all_parameters()) {
+    return true;
+  }
+
+  assert(profile_parameters_jsr292_only(), "inconsistent");
+  return m->is_compiled_lambda_form();
+}

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

@@ -119,7 +119,8 @@ public:
     multi_branch_data_tag,
     arg_info_data_tag,
     call_type_data_tag,
-    virtual_call_type_data_tag
+    virtual_call_type_data_tag,
+    parameters_type_data_tag
   };
 
   enum {
@@ -264,6 +265,7 @@ class     BranchData;
 class   ArrayData;
 class     MultiBranchData;
 class     ArgInfoData;
+class     ParametersTypeData;
 
 // ProfileData
 //
@@ -397,6 +399,7 @@ public:
   virtual bool is_ArgInfoData()     const { return false; }
   virtual bool is_CallTypeData()    const { return false; }
   virtual bool is_VirtualCallTypeData()const { return false; }
+  virtual bool is_ParametersTypeData() const { return false; }
 
 
   BitData* as_BitData() const {
@@ -447,6 +450,10 @@ public:
     assert(is_VirtualCallTypeData(), "wrong type");
     return is_VirtualCallTypeData() ? (VirtualCallTypeData*)this : NULL;
   }
+  ParametersTypeData* as_ParametersTypeData() const {
+    assert(is_ParametersTypeData(), "wrong type");
+    return is_ParametersTypeData() ? (ParametersTypeData*)this : NULL;
+  }
 
 
   // Subclass specific initialization
@@ -767,9 +774,9 @@ public:
   TypeStackSlotEntries(int base_off, int nb_entries)
     : TypeEntries(base_off), _number_of_entries(nb_entries) {}
 
-  static int compute_cell_count(Symbol* signature, int max);
+  static int compute_cell_count(Symbol* signature, bool include_receiver, int max);
 
-  void post_initialize(Symbol* signature, bool has_receiver);
+  void post_initialize(Symbol* signature, bool has_receiver, bool include_receiver);
 
   // offset of cell for stack slot for entry i within this block of cells for a TypeStackSlotEntries
   static int stack_slot_local_offset(int i) {
@@ -946,17 +953,6 @@ private:
     assert(number_of_arguments() == total, "should be set in DataLayout::initialize");
   }
 
-protected:
-  // An entry for a return value takes less space than an entry for an
-  // argument so if the number of cells exceeds the number of cells
-  // needed for an argument, this object contains type information for
-  // at least one argument.
-  bool has_arguments() const {
-    bool res = cell_count_no_header() >= TypeStackSlotEntries::per_arg_count();
-    assert (!res || TypeEntriesAtCall::arguments_profiling_enabled(), "no profiling of arguments");
-    return res;
-  }
-
 public:
   CallTypeData(DataLayout* layout) :
     CounterData(layout),
@@ -1017,6 +1013,16 @@ public:
     _ret.set_type(TypeEntries::with_status(k, current));
   }
 
+  // An entry for a return value takes less space than an entry for an
+  // argument so if the number of cells exceeds the number of cells
+  // needed for an argument, this object contains type information for
+  // at least one argument.
+  bool has_arguments() const {
+    bool res = cell_count_no_header() >= TypeStackSlotEntries::per_arg_count();
+    assert (!res || TypeEntriesAtCall::arguments_profiling_enabled(), "no profiling of arguments");
+    return res;
+  }
+
   // An entry for a return value takes less space than an entry for an
   // argument, so if the remainder of the number of cells divided by
   // the number of cells for an argument is not null, a return value
@@ -1213,17 +1219,6 @@ private:
     assert(number_of_arguments() == total, "should be set in DataLayout::initialize");
   }
 
-protected:
-  // An entry for a return value takes less space than an entry for an
-  // argument so if the number of cells exceeds the number of cells
-  // needed for an argument, this object contains type information for
-  // at least one argument.
-  bool has_arguments() const {
-    bool res = cell_count_no_header() >= TypeStackSlotEntries::per_arg_count();
-    assert (!res || TypeEntriesAtCall::arguments_profiling_enabled(), "no profiling of arguments");
-    return res;
-  }
-
 public:
   VirtualCallTypeData(DataLayout* layout) :
     VirtualCallData(layout),
@@ -1294,6 +1289,16 @@ public:
     return res;
   }
 
+  // An entry for a return value takes less space than an entry for an
+  // argument so if the number of cells exceeds the number of cells
+  // needed for an argument, this object contains type information for
+  // at least one argument.
+  bool has_arguments() const {
+    bool res = cell_count_no_header() >= TypeStackSlotEntries::per_arg_count();
+    assert (!res || TypeEntriesAtCall::arguments_profiling_enabled(), "no profiling of arguments");
+    return res;
+  }
+
   // Code generation support
   static ByteSize args_data_offset() {
     return cell_offset(VirtualCallData::static_cell_count()) + TypeEntriesAtCall::args_data_offset();
@@ -1662,6 +1667,75 @@ public:
 #endif
 };
 
+// ParametersTypeData
+//
+// A ParametersTypeData is used to access profiling information about
+// types of parameters to a method
+class ParametersTypeData : public ArrayData {
+
+private:
+  TypeStackSlotEntries _parameters;
+
+  static int stack_slot_local_offset(int i) {
+    assert_profiling_enabled();
+    return array_start_off_set + TypeStackSlotEntries::stack_slot_local_offset(i);
+  }
+
+  static int type_local_offset(int i) {
+    assert_profiling_enabled();
+    return array_start_off_set + TypeStackSlotEntries::type_local_offset(i);
+  }
+
+  static bool profiling_enabled();
+  static void assert_profiling_enabled() {
+    assert(profiling_enabled(), "method parameters profiling should be on");
+  }
+
+public:
+  ParametersTypeData(DataLayout* layout) : ArrayData(layout), _parameters(1, number_of_parameters()) {
+    assert(layout->tag() == DataLayout::parameters_type_data_tag, "wrong type");
+    // Some compilers (VC++) don't want this passed in member initialization list
+    _parameters.set_profile_data(this);
+  }
+
+  static int compute_cell_count(Method* m);
+
+  virtual bool is_ParametersTypeData() const { return true; }
+
+  virtual void post_initialize(BytecodeStream* stream, MethodData* mdo);
+
+  int number_of_parameters() const {
+    return array_len() / TypeStackSlotEntries::per_arg_count();
+  }
+
+  const TypeStackSlotEntries* parameters() const { return &_parameters; }
+
+  uint stack_slot(int i) const {
+    return _parameters.stack_slot(i);
+  }
+
+  void set_type(int i, Klass* k) {
+    intptr_t current = _parameters.type(i);
+    _parameters.set_type(i, TypeEntries::with_status((intptr_t)k, current));
+  }
+
+  virtual void clean_weak_klass_links(BoolObjectClosure* is_alive_closure) {
+    _parameters.clean_weak_klass_links(is_alive_closure);
+  }
+
+#ifndef PRODUCT
+  virtual void print_data_on(outputStream* st) const;
+#endif
+
+  static ByteSize stack_slot_offset(int i) {
+    return cell_offset(stack_slot_local_offset(i));
+  }
+
+  static ByteSize type_offset(int i) {
+    return cell_offset(type_local_offset(i));
+  }
+};
+
 // MethodData*
 //
 // A MethodData* holds information which has been collected about
@@ -1773,6 +1847,10 @@ private:
   // Size of _data array in bytes.  (Excludes header and extra_data fields.)
   int _data_size;
 
+  // data index for the area dedicated to parameters. -1 if no
+  // parameter profiling.
+  int _parameters_type_data_di;
+
   // Beginning of the data entries
   intptr_t _data[1];
 
@@ -1842,6 +1920,9 @@ private:
   static int profile_return_flag();
   static bool profile_all_return();
   static bool profile_return_for_invoke(methodHandle m, int bci);
+  static int profile_parameters_flag();
+  static bool profile_parameters_jsr292_only();
+  static bool profile_all_parameters();
 
 public:
   static int header_size() {
@@ -2048,6 +2129,16 @@ public:
     }
   }
 
+  // Return pointer to area dedicated to parameters in MDO
+  ParametersTypeData* parameters_type_data() const {
+    return _parameters_type_data_di != -1 ? data_layout_at(_parameters_type_data_di)->data_in()->as_ParametersTypeData() : NULL;
+  }
+
+  int parameters_type_data_di() const {
+    assert(_parameters_type_data_di != -1, "no args type data");
+    return _parameters_type_data_di;
+  }
+
   // Support for code generation
   static ByteSize data_offset() {
     return byte_offset_of(MethodData, _data[0]);
@@ -2060,6 +2151,10 @@ public:
     return byte_offset_of(MethodData, _backedge_counter);
   }
 
+  static ByteSize parameters_type_data_di_offset() {
+    return byte_offset_of(MethodData, _parameters_type_data_di);
+  }
+
   // Deallocation support - no pointer fields to deallocate
   void deallocate_contents(ClassLoaderData* loader_data) {}
 
@@ -2083,8 +2178,10 @@ public:
   void verify_on(outputStream* st);
   void verify_data_on(outputStream* st);
 
+  static bool profile_parameters_for_method(methodHandle m);
   static bool profile_arguments();
   static bool profile_return();
+  static bool profile_parameters();
   static bool profile_return_jsr292_only();
 };
 

hotspot/src/share/vm/opto/c2_globals.hpp

@@ -638,7 +638,13 @@
           "Find best control for expensive operations")                     \
                                                                             \
   product(bool, UseMathExactIntrinsics, true,                               \
-          "Enables intrinsification of various java.lang.Math funcitons")
+          "Enables intrinsification of various java.lang.Math functions")   \
+                                                                            \
+  experimental(bool, ReplaceInParentMaps, false,                            \
+          "Propagate type improvements in callers of inlinee if possible")  \
+                                                                            \
+  experimental(bool, UseTypeSpeculation, false,                             \
+          "Speculatively propagate types from profiles")
 
 C2_FLAGS(DECLARE_DEVELOPER_FLAG, DECLARE_PD_DEVELOPER_FLAG, DECLARE_PRODUCT_FLAG, DECLARE_PD_PRODUCT_FLAG, DECLARE_DIAGNOSTIC_FLAG, DECLARE_EXPERIMENTAL_FLAG, DECLARE_NOTPRODUCT_FLAG)
 

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

@@ -63,12 +63,12 @@ public:
   }
 
   virtual bool      is_parse() const           { return true; }
-  virtual JVMState* generate(JVMState* jvms);
+  virtual JVMState* generate(JVMState* jvms, Parse* parent_parser);
   int is_osr() { return _is_osr; }
 
 };
 
-JVMState* ParseGenerator::generate(JVMState* jvms) {
+JVMState* ParseGenerator::generate(JVMState* jvms, Parse* parent_parser) {
   Compile* C = Compile::current();
 
   if (is_osr()) {
@@ -80,7 +80,7 @@ JVMState* ParseGenerator::generate(JVMState* jvms) {
     return NULL;  // bailing out of the compile; do not try to parse
   }
 
-  Parse parser(jvms, method(), _expected_uses);
+  Parse parser(jvms, method(), _expected_uses, parent_parser);
   // Grab signature for matching/allocation
 #ifdef ASSERT
   if (parser.tf() != (parser.depth() == 1 ? C->tf() : tf())) {
@@ -119,12 +119,12 @@ class DirectCallGenerator : public CallGenerator {
       _separate_io_proj(separate_io_proj)
   {
   }
-  virtual JVMState* generate(JVMState* jvms);
+  virtual JVMState* generate(JVMState* jvms, Parse* parent_parser);
 
   CallStaticJavaNode* call_node() const { return _call_node; }
 };
 
-JVMState* DirectCallGenerator::generate(JVMState* jvms) {
+JVMState* DirectCallGenerator::generate(JVMState* jvms, Parse* parent_parser) {
   GraphKit kit(jvms);
   bool is_static = method()->is_static();
   address target = is_static ? SharedRuntime::get_resolve_static_call_stub()
@@ -171,10 +171,10 @@ public:
            vtable_index >= 0, "either invalid or usable");
   }
   virtual bool      is_virtual() const          { return true; }
-  virtual JVMState* generate(JVMState* jvms);
+  virtual JVMState* generate(JVMState* jvms, Parse* parent_parser);
 };
 
-JVMState* VirtualCallGenerator::generate(JVMState* jvms) {
+JVMState* VirtualCallGenerator::generate(JVMState* jvms, Parse* parent_parser) {
   GraphKit kit(jvms);
   Node* receiver = kit.argument(0);
 
@@ -276,7 +276,7 @@ class LateInlineCallGenerator : public DirectCallGenerator {
   // Convert the CallStaticJava into an inline
   virtual void do_late_inline();
 
-  virtual JVMState* generate(JVMState* jvms) {
+  virtual JVMState* generate(JVMState* jvms, Parse* parent_parser) {
     Compile *C = Compile::current();
     C->print_inlining_skip(this);
 
@@ -290,7 +290,7 @@ class LateInlineCallGenerator : public DirectCallGenerator {
     // that the late inlining logic can distinguish between fall
     // through and exceptional uses of the memory and io projections
     // as is done for allocations and macro expansion.
-    return DirectCallGenerator::generate(jvms);
+    return DirectCallGenerator::generate(jvms, parent_parser);
   }
 
   virtual void print_inlining_late(const char* msg) {
@@ -389,7 +389,7 @@ void LateInlineCallGenerator::do_late_inline() {
   }
 
   // Now perform the inling using the synthesized JVMState
-  JVMState* new_jvms = _inline_cg->generate(jvms);
+  JVMState* new_jvms = _inline_cg->generate(jvms, NULL);
   if (new_jvms == NULL)  return;  // no change
   if (C->failing())      return;
 
@@ -429,8 +429,8 @@ class LateInlineMHCallGenerator : public LateInlineCallGenerator {
 
   virtual bool is_mh_late_inline() const { return true; }
 
-  virtual JVMState* generate(JVMState* jvms) {
+  virtual JVMState* generate(JVMState* jvms, Parse* parent_parser) {
-    JVMState* new_jvms = LateInlineCallGenerator::generate(jvms);
+    JVMState* new_jvms = LateInlineCallGenerator::generate(jvms, parent_parser);
     if (_input_not_const) {
       // inlining won't be possible so no need to enqueue right now.
       call_node()->set_generator(this);
@@ -477,15 +477,17 @@ class LateInlineStringCallGenerator : public LateInlineCallGenerator {
   LateInlineStringCallGenerator(ciMethod* method, CallGenerator* inline_cg) :
     LateInlineCallGenerator(method, inline_cg) {}
 
-  virtual JVMState* generate(JVMState* jvms) {
+  virtual JVMState* generate(JVMState* jvms, Parse* parent_parser) {
     Compile *C = Compile::current();
     C->print_inlining_skip(this);
 
     C->add_string_late_inline(this);
 
-    JVMState* new_jvms =  DirectCallGenerator::generate(jvms);
+    JVMState* new_jvms =  DirectCallGenerator::generate(jvms, parent_parser);
     return new_jvms;
   }
+
+  virtual bool is_string_late_inline() const { return true; }
 };
 
 CallGenerator* CallGenerator::for_string_late_inline(ciMethod* method, CallGenerator* inline_cg) {
@@ -498,13 +500,13 @@ class LateInlineBoxingCallGenerator : public LateInlineCallGenerator {
   LateInlineBoxingCallGenerator(ciMethod* method, CallGenerator* inline_cg) :
     LateInlineCallGenerator(method, inline_cg) {}
 
-  virtual JVMState* generate(JVMState* jvms) {
+  virtual JVMState* generate(JVMState* jvms, Parse* parent_parser) {
     Compile *C = Compile::current();
     C->print_inlining_skip(this);
 
     C->add_boxing_late_inline(this);
 
-    JVMState* new_jvms =  DirectCallGenerator::generate(jvms);
+    JVMState* new_jvms =  DirectCallGenerator::generate(jvms, parent_parser);
     return new_jvms;
   }
 };
@@ -540,7 +542,7 @@ public:
   virtual bool      is_virtual() const          { return _is_virtual; }
   virtual bool      is_deferred() const         { return true; }
 
-  virtual JVMState* generate(JVMState* jvms);
+  virtual JVMState* generate(JVMState* jvms, Parse* parent_parser);
 };
 
 
@@ -550,12 +552,12 @@ CallGenerator* CallGenerator::for_warm_call(WarmCallInfo* ci,
   return new WarmCallGenerator(ci, if_cold, if_hot);
 }
 
-JVMState* WarmCallGenerator::generate(JVMState* jvms) {
+JVMState* WarmCallGenerator::generate(JVMState* jvms, Parse* parent_parser) {
   Compile* C = Compile::current();
   if (C->log() != NULL) {
     C->log()->elem("warm_call bci='%d'", jvms->bci());
   }
-  jvms = _if_cold->generate(jvms);
+  jvms = _if_cold->generate(jvms, parent_parser);
   if (jvms != NULL) {
     Node* m = jvms->map()->control();
     if (m->is_CatchProj()) m = m->in(0);  else m = C->top();
@@ -616,7 +618,7 @@ public:
   virtual bool      is_inline()    const    { return _if_hit->is_inline(); }
   virtual bool      is_deferred()  const    { return _if_hit->is_deferred(); }
 
-  virtual JVMState* generate(JVMState* jvms);
+  virtual JVMState* generate(JVMState* jvms, Parse* parent_parser);
 };
 
 
@@ -628,7 +630,7 @@ CallGenerator* CallGenerator::for_predicted_call(ciKlass* predicted_receiver,
 }
 
 
-JVMState* PredictedCallGenerator::generate(JVMState* jvms) {
+JVMState* PredictedCallGenerator::generate(JVMState* jvms, Parse* parent_parser) {
   GraphKit kit(jvms);
   PhaseGVN& gvn = kit.gvn();
   // We need an explicit receiver null_check before checking its type.
@@ -656,7 +658,7 @@ JVMState* PredictedCallGenerator::generate(JVMState* jvms) {
   { PreserveJVMState pjvms(&kit);
     kit.set_control(slow_ctl);
     if (!kit.stopped()) {
-      slow_jvms = _if_missed->generate(kit.sync_jvms());
+      slow_jvms = _if_missed->generate(kit.sync_jvms(), parent_parser);
       if (kit.failing())
         return NULL;  // might happen because of NodeCountInliningCutoff
       assert(slow_jvms != NULL, "must be");
@@ -677,12 +679,12 @@ JVMState* PredictedCallGenerator::generate(JVMState* jvms) {
   kit.replace_in_map(receiver, exact_receiver);
 
   // Make the hot call:
-  JVMState* new_jvms = _if_hit->generate(kit.sync_jvms());
+  JVMState* new_jvms = _if_hit->generate(kit.sync_jvms(), parent_parser);
   if (new_jvms == NULL) {
     // Inline failed, so make a direct call.
     assert(_if_hit->is_inline(), "must have been a failed inline");
     CallGenerator* cg = CallGenerator::for_direct_call(_if_hit->method());
-    new_jvms = cg->generate(kit.sync_jvms());
+    new_jvms = cg->generate(kit.sync_jvms(), parent_parser);
   }
   kit.add_exception_states_from(new_jvms);
   kit.set_jvms(new_jvms);
@@ -773,7 +775,7 @@ CallGenerator* CallGenerator::for_method_handle_inline(JVMState* jvms, ciMethod*
         ciMethod* target = oop_ptr->const_oop()->as_method_handle()->get_vmtarget();
         guarantee(!target->is_method_handle_intrinsic(), "should not happen");  // XXX remove
         const int vtable_index = Method::invalid_vtable_index;
-        CallGenerator* cg = C->call_generator(target, vtable_index, false, jvms, true, PROB_ALWAYS, true, true);
+        CallGenerator* cg = C->call_generator(target, vtable_index, false, jvms, true, PROB_ALWAYS, NULL, true, true);
         assert(!cg->is_late_inline() || cg->is_mh_late_inline(), "no late inline here");
         if (cg != NULL && cg->is_inline())
           return cg;
@@ -829,6 +831,7 @@ CallGenerator* CallGenerator::for_method_handle_inline(JVMState* jvms, ciMethod*
         int  vtable_index       = Method::invalid_vtable_index;
         bool call_does_dispatch = false;
 
+        ciKlass* speculative_receiver_type = NULL;
         if (is_virtual_or_interface) {
           ciInstanceKlass* klass = target->holder();
           Node*             receiver_node = kit.argument(0);
@@ -837,9 +840,12 @@ CallGenerator* CallGenerator::for_method_handle_inline(JVMState* jvms, ciMethod*
           target = C->optimize_virtual_call(caller, jvms->bci(), klass, target, receiver_type,
                                             is_virtual,
                                             call_does_dispatch, vtable_index);  // out-parameters
+          // We lack profiling at this call but type speculation may
+          // provide us with a type
+          speculative_receiver_type = receiver_type->speculative_type();
         }
 
-        CallGenerator* cg = C->call_generator(target, vtable_index, call_does_dispatch, jvms, true, PROB_ALWAYS, true, true);
+        CallGenerator* cg = C->call_generator(target, vtable_index, call_does_dispatch, jvms, true, PROB_ALWAYS, speculative_receiver_type, true, true);
         assert(!cg->is_late_inline() || cg->is_mh_late_inline(), "no late inline here");
         if (cg != NULL && cg->is_inline())
           return cg;
@@ -874,7 +880,7 @@ public:
   virtual bool      is_inlined()   const    { return true; }
   virtual bool      is_intrinsic() const    { return true; }
 
-  virtual JVMState* generate(JVMState* jvms);
+  virtual JVMState* generate(JVMState* jvms, Parse* parent_parser);
 };
 
 
@@ -884,7 +890,7 @@ CallGenerator* CallGenerator::for_predicted_intrinsic(CallGenerator* intrinsic,
 }
 
 
-JVMState* PredictedIntrinsicGenerator::generate(JVMState* jvms) {
+JVMState* PredictedIntrinsicGenerator::generate(JVMState* jvms, Parse* parent_parser) {
   GraphKit kit(jvms);
   PhaseGVN& gvn = kit.gvn();
 
@@ -904,7 +910,7 @@ JVMState* PredictedIntrinsicGenerator::generate(JVMState* jvms) {
     PreserveJVMState pjvms(&kit);
     kit.set_control(slow_ctl);
     if (!kit.stopped()) {
-      slow_jvms = _cg->generate(kit.sync_jvms());
+      slow_jvms = _cg->generate(kit.sync_jvms(), parent_parser);
       if (kit.failing())
         return NULL;  // might happen because of NodeCountInliningCutoff
       assert(slow_jvms != NULL, "must be");
@@ -922,12 +928,12 @@ JVMState* PredictedIntrinsicGenerator::generate(JVMState* jvms) {
   }
 
   // Generate intrinsic code:
-  JVMState* new_jvms = _intrinsic->generate(kit.sync_jvms());
+  JVMState* new_jvms = _intrinsic->generate(kit.sync_jvms(), parent_parser);
   if (new_jvms == NULL) {
     // Intrinsic failed, so use slow code or make a direct call.
     if (slow_map == NULL) {
       CallGenerator* cg = CallGenerator::for_direct_call(method());
-      new_jvms = cg->generate(kit.sync_jvms());
+      new_jvms = cg->generate(kit.sync_jvms(), parent_parser);
     } else {
       kit.set_jvms(slow_jvms);
       return kit.transfer_exceptions_into_jvms();
@@ -997,7 +1003,7 @@ public:
   virtual bool      is_virtual() const          { ShouldNotReachHere(); return false; }
   virtual bool      is_trap() const             { return true; }
 
-  virtual JVMState* generate(JVMState* jvms);
+  virtual JVMState* generate(JVMState* jvms, Parse* parent_parser);
 };
 
 
@@ -1009,7 +1015,7 @@ CallGenerator::for_uncommon_trap(ciMethod* m,
 }
 
 
-JVMState* UncommonTrapCallGenerator::generate(JVMState* jvms) {
+JVMState* UncommonTrapCallGenerator::generate(JVMState* jvms, Parse* parent_parser) {
   GraphKit kit(jvms);
   // Take the trap with arguments pushed on the stack.  (Cf. null_check_receiver).
   int nargs = method()->arg_size();

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

@@ -31,6 +31,8 @@
 #include "opto/type.hpp"
 #include "runtime/deoptimization.hpp"
 
+class Parse;
+
 //---------------------------CallGenerator-------------------------------------
 // The subclasses of this class handle generation of ideal nodes for
 // call sites and method entry points.
@@ -72,6 +74,7 @@ class CallGenerator : public ResourceObj {
   virtual bool      is_late_inline() const      { return false; }
   // same but for method handle calls
   virtual bool      is_mh_late_inline() const   { return false; }
+  virtual bool      is_string_late_inline() const{ return false; }
 
   // for method handle calls: have we tried inlinining the call already?
   virtual bool      already_attempted() const   { ShouldNotReachHere(); return false; }
@@ -108,7 +111,7 @@ class CallGenerator : public ResourceObj {
   //
   // If the result is NULL, it means that this CallGenerator was unable
   // to handle the given call, and another CallGenerator should be consulted.
-  virtual JVMState* generate(JVMState* jvms) = 0;
+  virtual JVMState* generate(JVMState* jvms, Parse* parent_parser) = 0;
 
   // How to generate a call site that is inlined:
   static CallGenerator* for_inline(ciMethod* m, float expected_uses = -1);

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

@@ -30,6 +30,7 @@ macro(AbsF)
 macro(AbsI)
 macro(AddD)
 macro(AddExactI)
+macro(AddExactL)
 macro(AddF)
 macro(AddI)
 macro(AddL)
@@ -170,6 +171,8 @@ macro(LoopLimit)
 macro(Mach)
 macro(MachProj)
 macro(MathExact)
+macro(MathExactI)
+macro(MathExactL)
 macro(MaxI)
 macro(MemBarAcquire)
 macro(MemBarAcquireLock)
@@ -189,12 +192,16 @@ macro(MoveF2I)
 macro(MoveL2D)
 macro(MoveD2L)
 macro(MulD)
+macro(MulExactI)
+macro(MulExactL)
 macro(MulF)
 macro(MulHiL)
 macro(MulI)
 macro(MulL)
 macro(Multi)
 macro(NegD)
+macro(NegExactI)
+macro(NegExactL)
 macro(NegF)
 macro(NeverBranch)
 macro(Opaque1)
@@ -244,6 +251,8 @@ macro(StrComp)
 macro(StrEquals)
 macro(StrIndexOf)
 macro(SubD)
+macro(SubExactI)
+macro(SubExactL)
 macro(SubF)
 macro(SubI)
 macro(SubL)

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

@@ -655,7 +655,8 @@ Compile::Compile( ciEnv* ci_env, C2Compiler* compiler, ciMethod* target, int osr
                   _inlining_progress(false),
                   _inlining_incrementally(false),
                   _print_inlining_list(NULL),
-                  _print_inlining_idx(0) {
+                  _print_inlining_idx(0),
+                  _preserve_jvm_state(0) {
   C = this;
 
   CompileWrapper cw(this);
@@ -763,7 +764,7 @@ Compile::Compile( ciEnv* ci_env, C2Compiler* compiler, ciMethod* target, int osr
       return;
     }
     JVMState* jvms = build_start_state(start(), tf());
-    if ((jvms = cg->generate(jvms)) == NULL) {
+    if ((jvms = cg->generate(jvms, NULL)) == NULL) {
       record_method_not_compilable("method parse failed");
       return;
     }
@@ -940,7 +941,8 @@ Compile::Compile( ciEnv* ci_env,
     _inlining_progress(false),
     _inlining_incrementally(false),
     _print_inlining_list(NULL),
-    _print_inlining_idx(0) {
+    _print_inlining_idx(0),
+    _preserve_jvm_state(0) {
   C = this;
 
 #ifndef PRODUCT
@@ -1358,7 +1360,7 @@ const TypePtr *Compile::flatten_alias_type( const TypePtr *tj ) const {
     // During the 2nd round of IterGVN, NotNull castings are removed.
     // Make sure the Bottom and NotNull variants alias the same.
     // Also, make sure exact and non-exact variants alias the same.
-    if( ptr == TypePtr::NotNull || ta->klass_is_exact() ) {
+    if (ptr == TypePtr::NotNull || ta->klass_is_exact() || ta->speculative() != NULL) {
       tj = ta = TypeAryPtr::make(TypePtr::BotPTR,ta->ary(),ta->klass(),false,offset);
     }
   }
@@ -1383,6 +1385,9 @@ const TypePtr *Compile::flatten_alias_type( const TypePtr *tj ) const {
       // Also, make sure exact and non-exact variants alias the same.
       tj = to = TypeInstPtr::make(TypePtr::BotPTR,to->klass(),false,0,offset);
     }
+    if (to->speculative() != NULL) {
+      tj = to = TypeInstPtr::make(to->ptr(),to->klass(),to->klass_is_exact(),to->const_oop(),to->offset(), to->instance_id());
+    }
     // Canonicalize the holder of this field
     if (offset >= 0 && offset < instanceOopDesc::base_offset_in_bytes()) {
       // First handle header references such as a LoadKlassNode, even if the
@@ -2011,6 +2016,12 @@ void Compile::Optimize() {
     if (failing())  return;
   }
 
+  // Remove the speculative part of types and clean up the graph from
+  // the extra CastPP nodes whose only purpose is to carry them. Do
+  // that early so that optimizations are not disrupted by the extra
+  // CastPP nodes.
+  remove_speculative_types(igvn);
+
   // No more new expensive nodes will be added to the list from here
   // so keep only the actual candidates for optimizations.
   cleanup_expensive_nodes(igvn);
@@ -3004,6 +3015,10 @@ void Compile::final_graph_reshaping_impl( Node *n, Final_Reshape_Counts &frc) {
       if (result != NULL) {
         for (DUIterator_Fast jmax, j = result->fast_outs(jmax); j < jmax; j++) {
           Node* out = result->fast_out(j);
+          // Phi nodes shouldn't be moved. They would only match below if they
+          // had the same control as the MathExactNode. The only time that
+          // would happen is if the Phi is also an input to the MathExact
+          if (!out->is_Phi()) {
             if (out->in(0) == NULL) {
               out->set_req(0, non_throwing);
             } else if (out->in(0) == ctrl) {
@@ -3012,6 +3027,7 @@ void Compile::final_graph_reshaping_impl( Node *n, Final_Reshape_Counts &frc) {
           }
         }
       }
+    }
     break;
   default:
     assert( !n->is_Call(), "" );
@@ -3792,6 +3808,45 @@ void Compile::add_expensive_node(Node * n) {
   }
 }
 
+/**
+ * Remove the speculative part of types and clean up the graph
+ */
+void Compile::remove_speculative_types(PhaseIterGVN &igvn) {
+  if (UseTypeSpeculation) {
+    Unique_Node_List worklist;
+    worklist.push(root());
+    int modified = 0;
+    // Go over all type nodes that carry a speculative type, drop the
+    // speculative part of the type and enqueue the node for an igvn
+    // which may optimize it out.
+    for (uint next = 0; next < worklist.size(); ++next) {
+      Node *n  = worklist.at(next);
+      if (n->is_Type() && n->as_Type()->type()->isa_oopptr() != NULL &&
+          n->as_Type()->type()->is_oopptr()->speculative() != NULL) {
+        TypeNode* tn = n->as_Type();
+        const TypeOopPtr* t = tn->type()->is_oopptr();
+        bool in_hash = igvn.hash_delete(n);
+        assert(in_hash, "node should be in igvn hash table");
+        tn->set_type(t->remove_speculative());
+        igvn.hash_insert(n);
+        igvn._worklist.push(n); // give it a chance to go away
+        modified++;
+      }
+      uint max = n->len();
+      for( uint i = 0; i < max; ++i ) {
+        Node *m = n->in(i);
+        if (not_a_node(m))  continue;
+        worklist.push(m);
+      }
+    }
+    // Drop the speculative part of all types in the igvn's type table
+    igvn.remove_speculative_types();
+    if (modified > 0) {
+      igvn.optimize();
+    }
+  }
+}
+
 // Auxiliary method to support randomized stressing/fuzzing.
 //
 // This method can be called the arbitrary number of times, with current count

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

@@ -424,6 +424,11 @@ class Compile : public Phase {
   static int cmp_expensive_nodes(Node** n1, Node** n2);
   // Expensive nodes list already sorted?
   bool expensive_nodes_sorted() const;
+  // Remove the speculative part of types and clean up the graph
+  void remove_speculative_types(PhaseIterGVN &igvn);
+
+  // Are we within a PreserveJVMState block?
+  int _preserve_jvm_state;
 
  public:
 
@@ -820,7 +825,9 @@ class Compile : public Phase {
 
   // Decide how to build a call.
   // The profile factor is a discount to apply to this site's interp. profile.
-  CallGenerator*    call_generator(ciMethod* call_method, int vtable_index, bool call_does_dispatch, JVMState* jvms, bool allow_inline, float profile_factor, bool allow_intrinsics = true, bool delayed_forbidden = false);
+  CallGenerator*    call_generator(ciMethod* call_method, int vtable_index, bool call_does_dispatch,
+                                   JVMState* jvms, bool allow_inline, float profile_factor, ciKlass* speculative_receiver_type = NULL,
+                                   bool allow_intrinsics = true, bool delayed_forbidden = false);
   bool should_delay_inlining(ciMethod* call_method, JVMState* jvms) {
     return should_delay_string_inlining(call_method, jvms) ||
            should_delay_boxing_inlining(call_method, jvms);
@@ -1156,6 +1163,21 @@ class Compile : public Phase {
 
   // Auxiliary method for randomized fuzzing/stressing
   static bool randomized_select(int count);
+
+  // enter a PreserveJVMState block
+  void inc_preserve_jvm_state() {
+    _preserve_jvm_state++;
+  }
+
+  // exit a PreserveJVMState block
+  void dec_preserve_jvm_state() {
+    _preserve_jvm_state--;
+    assert(_preserve_jvm_state >= 0, "_preserve_jvm_state shouldn't be negative");
+  }
+
+  bool has_preserve_jvm_state() const {
+    return _preserve_jvm_state > 0;
+  }
 };
 
 #endif // SHARE_VM_OPTO_COMPILE_HPP

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

@@ -63,7 +63,8 @@ void trace_type_profile(Compile* C, ciMethod *method, int depth, int bci, ciMeth
 
 CallGenerator* Compile::call_generator(ciMethod* callee, int vtable_index, bool call_does_dispatch,
                                        JVMState* jvms, bool allow_inline,
-                                       float prof_factor, bool allow_intrinsics, bool delayed_forbidden) {
+                                       float prof_factor, ciKlass* speculative_receiver_type,
+                                       bool allow_intrinsics, bool delayed_forbidden) {
   ciMethod*       caller   = jvms->method();
   int             bci      = jvms->bci();
   Bytecodes::Code bytecode = caller->java_code_at_bci(bci);
@@ -117,7 +118,7 @@ CallGenerator* Compile::call_generator(ciMethod* callee, int vtable_index, bool
       if (cg->is_predicted()) {
         // Code without intrinsic but, hopefully, inlined.
         CallGenerator* inline_cg = this->call_generator(callee,
-              vtable_index, call_does_dispatch, jvms, allow_inline, prof_factor, false);
+              vtable_index, call_does_dispatch, jvms, allow_inline, prof_factor, speculative_receiver_type, false);
         if (inline_cg != NULL) {
           cg = CallGenerator::for_predicted_intrinsic(cg, inline_cg);
         }
@@ -212,8 +213,24 @@ CallGenerator* Compile::call_generator(ciMethod* callee, int vtable_index, bool
       // The major receiver's count >= TypeProfileMajorReceiverPercent of site_count.
       bool have_major_receiver = (100.*profile.receiver_prob(0) >= (float)TypeProfileMajorReceiverPercent);
       ciMethod* receiver_method = NULL;
-      if (have_major_receiver || profile.morphism() == 1 ||
+
-          (profile.morphism() == 2 && UseBimorphicInlining)) {
+      int morphism = profile.morphism();
+      if (speculative_receiver_type != NULL) {
+        // We have a speculative type, we should be able to resolve
+        // the call. We do that before looking at the profiling at
+        // this invoke because it may lead to bimorphic inlining which
+        // a speculative type should help us avoid.
+        receiver_method = callee->resolve_invoke(jvms->method()->holder(),
+                                                 speculative_receiver_type);
+        if (receiver_method == NULL) {
+          speculative_receiver_type = NULL;
+        } else {
+          morphism = 1;
+        }
+      }
+      if (receiver_method == NULL &&
+          (have_major_receiver || morphism == 1 ||
+           (morphism == 2 && UseBimorphicInlining))) {
         // receiver_method = profile.method();
         // Profiles do not suggest methods now.  Look it up in the major receiver.
         receiver_method = callee->resolve_invoke(jvms->method()->holder(),
@@ -227,7 +244,7 @@ CallGenerator* Compile::call_generator(ciMethod* callee, int vtable_index, bool
           // Look up second receiver.
           CallGenerator* next_hit_cg = NULL;
           ciMethod* next_receiver_method = NULL;
-          if (profile.morphism() == 2 && UseBimorphicInlining) {
+          if (morphism == 2 && UseBimorphicInlining) {
             next_receiver_method = callee->resolve_invoke(jvms->method()->holder(),
                                                                profile.receiver(1));
             if (next_receiver_method != NULL) {
@@ -242,11 +259,10 @@ CallGenerator* Compile::call_generator(ciMethod* callee, int vtable_index, bool
             }
           }
           CallGenerator* miss_cg;
-          Deoptimization::DeoptReason reason = (profile.morphism() == 2) ?
+          Deoptimization::DeoptReason reason = morphism == 2 ?
                                     Deoptimization::Reason_bimorphic :
                                     Deoptimization::Reason_class_check;
-          if (( profile.morphism() == 1 ||
+          if ((morphism == 1 || (morphism == 2 && next_hit_cg != NULL)) &&
-               (profile.morphism() == 2 && next_hit_cg != NULL) ) &&
               !too_many_traps(jvms->method(), jvms->bci(), reason)
              ) {
             // Generate uncommon trap for class check failure path
@@ -260,6 +276,7 @@ CallGenerator* Compile::call_generator(ciMethod* callee, int vtable_index, bool
           }
           if (miss_cg != NULL) {
             if (next_hit_cg != NULL) {
+              assert(speculative_receiver_type == NULL, "shouldn't end up here if we used speculation");
               trace_type_profile(C, jvms->method(), jvms->depth() - 1, jvms->bci(), next_receiver_method, profile.receiver(1), site_count, profile.receiver_count(1));
               // We don't need to record dependency on a receiver here and below.
               // Whenever we inline, the dependency is added by Parse::Parse().
@@ -267,7 +284,9 @@ CallGenerator* Compile::call_generator(ciMethod* callee, int vtable_index, bool
             }
             if (miss_cg != NULL) {
               trace_type_profile(C, jvms->method(), jvms->depth() - 1, jvms->bci(), receiver_method, profile.receiver(0), site_count, receiver_count);
-              CallGenerator* cg = CallGenerator::for_predicted_call(profile.receiver(0), miss_cg, hit_cg, profile.receiver_prob(0));
+              ciKlass* k = speculative_receiver_type != NULL ? speculative_receiver_type : profile.receiver(0);
+              float hit_prob = speculative_receiver_type != NULL ? 1.0 : profile.receiver_prob(0);
+              CallGenerator* cg = CallGenerator::for_predicted_call(k, miss_cg, hit_cg, hit_prob);
               if (cg != NULL)  return cg;
             }
           }
@@ -446,6 +465,8 @@ void Parse::do_call() {
   int       vtable_index       = Method::invalid_vtable_index;
   bool      call_does_dispatch = false;
 
+  // Speculative type of the receiver if any
+  ciKlass* speculative_receiver_type = NULL;
   if (is_virtual_or_interface) {
     Node* receiver_node             = stack(sp() - nargs);
     const TypeOopPtr* receiver_type = _gvn.type(receiver_node)->isa_oopptr();
@@ -453,6 +474,7 @@ void Parse::do_call() {
     callee = C->optimize_virtual_call(method(), bci(), klass, orig_callee, receiver_type,
                                       is_virtual,
                                       call_does_dispatch, vtable_index);  // out-parameters
+    speculative_receiver_type = receiver_type != NULL ? receiver_type->speculative_type() : NULL;
   }
 
   // Note:  It's OK to try to inline a virtual call.
@@ -468,7 +490,7 @@ void Parse::do_call() {
   // Decide call tactic.
   // This call checks with CHA, the interpreter profile, intrinsics table, etc.
   // It decides whether inlining is desirable or not.
-  CallGenerator* cg = C->call_generator(callee, vtable_index, call_does_dispatch, jvms, try_inline, prof_factor());
+  CallGenerator* cg = C->call_generator(callee, vtable_index, call_does_dispatch, jvms, try_inline, prof_factor(), speculative_receiver_type);
 
   // NOTE:  Don't use orig_callee and callee after this point!  Use cg->method() instead.
   orig_callee = callee = NULL;
@@ -477,6 +499,10 @@ void Parse::do_call() {
   // Round double arguments before call
   round_double_arguments(cg->method());
 
+  // Feed profiling data for arguments to the type system so it can
+  // propagate it as speculative types
+  record_profiled_arguments_for_speculation(cg->method(), bc());
+
 #ifndef PRODUCT
   // bump global counters for calls
   count_compiled_calls(/*at_method_entry*/ false, cg->is_inline());
@@ -491,11 +517,18 @@ void Parse::do_call() {
   // save across call, for a subsequent cast_not_null.
   Node* receiver = has_receiver ? argument(0) : NULL;
 
+  // The extra CheckCastPP for speculative types mess with PhaseStringOpts
+  if (receiver != NULL && !call_does_dispatch && !cg->is_string_late_inline()) {
+    // Feed profiling data for a single receiver to the type system so
+    // it can propagate it as a speculative type
+    receiver = record_profiled_receiver_for_speculation(receiver);
+  }
+
   // Bump method data counters (We profile *before* the call is made
   // because exceptions don't return to the call site.)
   profile_call(receiver);
 
-  JVMState* new_jvms = cg->generate(jvms);
+  JVMState* new_jvms = cg->generate(jvms, this);
   if (new_jvms == NULL) {
     // When inlining attempt fails (e.g., too many arguments),
     // it may contaminate the current compile state, making it
@@ -508,8 +541,8 @@ void Parse::do_call() {
     // the call site, perhaps because it did not match a pattern the
     // intrinsic was expecting to optimize. Should always be possible to
     // get a normal java call that may inline in that case
-    cg = C->call_generator(cg->method(), vtable_index, call_does_dispatch, jvms, try_inline, prof_factor(), /* allow_intrinsics= */ false);
+    cg = C->call_generator(cg->method(), vtable_index, call_does_dispatch, jvms, try_inline, prof_factor(), speculative_receiver_type, /* allow_intrinsics= */ false);
-    if ((new_jvms = cg->generate(jvms)) == NULL) {
+    if ((new_jvms = cg->generate(jvms, this)) == NULL) {
       guarantee(failing(), "call failed to generate:  calls should work");
       return;
     }
@@ -607,6 +640,16 @@ void Parse::do_call() {
       null_assert(peek());
       set_bci(iter().cur_bci()); // put it back
     }
+    BasicType ct = ctype->basic_type();
+    if (ct == T_OBJECT || ct == T_ARRAY) {
+      ciKlass* better_type = method()->return_profiled_type(bci());
+      if (UseTypeSpeculation && better_type != NULL) {
+        // If profiling reports a single type for the return value,
+        // feed it to the type system so it can propagate it as a
+        // speculative type
+        record_profile_for_speculation(stack(sp()-1), better_type);
+      }
+    }
   }
 
   // Restart record of parsing work after possible inlining of call

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

@@ -639,6 +639,7 @@ PreserveJVMState::PreserveJVMState(GraphKit* kit, bool clone_map) {
   _map    = kit->map();   // preserve the map
   _sp     = kit->sp();
   kit->set_map(clone_map ? kit->clone_map() : NULL);
+  Compile::current()->inc_preserve_jvm_state();
 #ifdef ASSERT
   _bci    = kit->bci();
   Parse* parser = kit->is_Parse();
@@ -656,6 +657,7 @@ PreserveJVMState::~PreserveJVMState() {
 #endif
   kit->set_map(_map);
   kit->set_sp(_sp);
+  Compile::current()->dec_preserve_jvm_state();
 }
 
 
@@ -1373,17 +1375,70 @@ Node* GraphKit::cast_not_null(Node* obj, bool do_replace_in_map) {
 
 //--------------------------replace_in_map-------------------------------------
 void GraphKit::replace_in_map(Node* old, Node* neww) {
-  this->map()->replace_edge(old, neww);
+  if (old == neww) {
+    return;
+  }
+
+  map()->replace_edge(old, neww);
 
   // Note: This operation potentially replaces any edge
   // on the map.  This includes locals, stack, and monitors
   // of the current (innermost) JVM state.
 
-  // We can consider replacing in caller maps.
+  if (!ReplaceInParentMaps) {
-  // The idea would be that an inlined function's null checks
+    return;
-  // can be shared with the entire inlining tree.
+  }
-  // The expense of doing this is that the PreserveJVMState class
+
-  // would have to preserve caller states too, with a deep copy.
+  // PreserveJVMState doesn't do a deep copy so we can't modify
+  // parents
+  if (Compile::current()->has_preserve_jvm_state()) {
+    return;
+  }
+
+  Parse* parser = is_Parse();
+  bool progress = true;
+  Node* ctrl = map()->in(0);
+  // Follow the chain of parsers and see whether the update can be
+  // done in the map of callers. We can do the replace for a caller if
+  // the current control post dominates the control of a caller.
+  while (parser != NULL && parser->caller() != NULL && progress) {
+    progress = false;
+    Node* parent_map = parser->caller()->map();
+    assert(parser->exits().map()->jvms()->depth() == parser->caller()->depth(), "map mismatch");
+
+    Node* parent_ctrl = parent_map->in(0);
+
+    while (parent_ctrl->is_Region()) {
+      Node* n = parent_ctrl->as_Region()->is_copy();
+      if (n == NULL) {
+        break;
+      }
+      parent_ctrl = n;
+    }
+
+    for (;;) {
+      if (ctrl == parent_ctrl) {
+        // update the map of the exits which is the one that will be
+        // used when compilation resume after inlining
+        parser->exits().map()->replace_edge(old, neww);
+        progress = true;
+        break;
+      }
+      if (ctrl->is_Proj() && ctrl->as_Proj()->is_uncommon_trap_if_pattern(Deoptimization::Reason_none)) {
+        ctrl = ctrl->in(0)->in(0);
+      } else if (ctrl->is_Region()) {
+        Node* n = ctrl->as_Region()->is_copy();
+        if (n == NULL) {
+          break;
+        }
+        ctrl = n;
+      } else {
+        break;
+      }
+    }
+
+    parser = parser->parent_parser();
+  }
 }
 
 
@@ -2043,6 +2098,104 @@ void GraphKit::round_double_arguments(ciMethod* dest_method) {
   }
 }
 
+/**
+ * Record profiling data exact_kls for Node n with the type system so
+ * that it can propagate it (speculation)
+ *
+ * @param n          node that the type applies to
+ * @param exact_kls  type from profiling
+ *
+ * @return           node with improved type
+ */
+Node* GraphKit::record_profile_for_speculation(Node* n, ciKlass* exact_kls) {
+  const TypeOopPtr* current_type = _gvn.type(n)->isa_oopptr();
+  assert(UseTypeSpeculation, "type speculation must be on");
+  if (exact_kls != NULL &&
+      // nothing to improve if type is already exact
+      (current_type == NULL ||
+       (!current_type->klass_is_exact() &&
+        (current_type->speculative() == NULL ||
+         !current_type->speculative()->klass_is_exact())))) {
+    const TypeKlassPtr* tklass = TypeKlassPtr::make(exact_kls);
+    const TypeOopPtr* xtype = tklass->as_instance_type();
+    assert(xtype->klass_is_exact(), "Should be exact");
+
+    // Build a type with a speculative type (what we think we know
+    // about the type but will need a guard when we use it)
+    const TypeOopPtr* spec_type = TypeOopPtr::make(TypePtr::BotPTR, Type::OffsetBot, TypeOopPtr::InstanceBot, xtype);
+    // We're changing the type, we need a new cast node to carry the
+    // new type. The new type depends on the control: what profiling
+    // tells us is only valid from here as far as we can tell.
+    Node* cast = new(C) CastPPNode(n, spec_type);
+    cast->init_req(0, control());
+    cast = _gvn.transform(cast);
+    replace_in_map(n, cast);
+    n = cast;
+  }
+  return n;
+}
+
+/**
+ * Record profiling data from receiver profiling at an invoke with the
+ * type system so that it can propagate it (speculation)
+ *
+ * @param n  receiver node
+ *
+ * @return           node with improved type
+ */
+Node* GraphKit::record_profiled_receiver_for_speculation(Node* n) {
+  if (!UseTypeSpeculation) {
+    return n;
+  }
+  ciKlass* exact_kls = profile_has_unique_klass();
+  return record_profile_for_speculation(n, exact_kls);
+}
+
+/**
+ * Record profiling data from argument profiling at an invoke with the
+ * type system so that it can propagate it (speculation)
+ *
+ * @param dest_method  target method for the call
+ * @param bc           what invoke bytecode is this?
+ */
+void GraphKit::record_profiled_arguments_for_speculation(ciMethod* dest_method, Bytecodes::Code bc) {
+  if (!UseTypeSpeculation) {
+    return;
+  }
+  const TypeFunc* tf    = TypeFunc::make(dest_method);
+  int             nargs = tf->_domain->_cnt - TypeFunc::Parms;
+  int skip = Bytecodes::has_receiver(bc) ? 1 : 0;
+  for (int j = skip, i = 0; j < nargs && i < TypeProfileArgsLimit; j++) {
+    const Type *targ = tf->_domain->field_at(j + TypeFunc::Parms);
+    if (targ->basic_type() == T_OBJECT || targ->basic_type() == T_ARRAY) {
+      ciKlass* better_type = method()->argument_profiled_type(bci(), i);
+      if (better_type != NULL) {
+        record_profile_for_speculation(argument(j), better_type);
+      }
+      i++;
+    }
+  }
+}
+
+/**
+ * Record profiling data from parameter profiling at an invoke with
+ * the type system so that it can propagate it (speculation)
+ */
+void GraphKit::record_profiled_parameters_for_speculation() {
+  if (!UseTypeSpeculation) {
+    return;
+  }
+  for (int i = 0, j = 0; i < method()->arg_size() ; i++) {
+    if (_gvn.type(local(i))->isa_oopptr()) {
+      ciKlass* better_type = method()->parameter_profiled_type(j);
+      if (better_type != NULL) {
+        record_profile_for_speculation(local(i), better_type);
+      }
+      j++;
+    }
+  }
+}
+
 void GraphKit::round_double_result(ciMethod* dest_method) {
   // A non-strict method may return a double value which has an extended
   // exponent, but this must not be visible in a caller which is 'strict'
@@ -2580,10 +2733,10 @@ bool GraphKit::seems_never_null(Node* obj, ciProfileData* data) {
 // If the profile has seen exactly one type, narrow to exactly that type.
 // Subsequent type checks will always fold up.
 Node* GraphKit::maybe_cast_profiled_receiver(Node* not_null_obj,
-                                             ciProfileData* data,
+                                             ciKlass* require_klass,
-                                             ciKlass* require_klass) {
+                                            ciKlass* spec_klass,
+                                             bool safe_for_replace) {
   if (!UseTypeProfile || !TypeProfileCasts) return NULL;
-  if (data == NULL)  return NULL;
 
   // Make sure we haven't already deoptimized from this tactic.
   if (too_many_traps(Deoptimization::Reason_class_check))
@@ -2591,15 +2744,15 @@ Node* GraphKit::maybe_cast_profiled_receiver(Node* not_null_obj,
 
   // (No, this isn't a call, but it's enough like a virtual call
   // to use the same ciMethod accessor to get the profile info...)
-  ciCallProfile profile = method()->call_profile_at_bci(bci());
+  // If we have a speculative type use it instead of profiling (which
-  if (profile.count() >= 0 &&         // no cast failures here
+  // may not help us)
-      profile.has_receiver(0) &&
+  ciKlass* exact_kls = spec_klass == NULL ? profile_has_unique_klass() : spec_klass;
-      profile.morphism() == 1) {
+  if (exact_kls != NULL) {// no cast failures here
-    ciKlass* exact_kls = profile.receiver(0);
     if (require_klass == NULL ||
         static_subtype_check(require_klass, exact_kls) == SSC_always_true) {
-      // If we narrow the type to match what the type profile sees,
+      // If we narrow the type to match what the type profile sees or
-      // we can then remove the rest of the cast.
+      // the speculative type, we can then remove the rest of the
+      // cast.
       // This is a win, even if the exact_kls is very specific,
       // because downstream operations, such as method calls,
       // will often benefit from the sharper type.
@@ -2611,7 +2764,9 @@ Node* GraphKit::maybe_cast_profiled_receiver(Node* not_null_obj,
         uncommon_trap(Deoptimization::Reason_class_check,
                       Deoptimization::Action_maybe_recompile);
       }
+      if (safe_for_replace) {
         replace_in_map(not_null_obj, exact_obj);
+      }
       return exact_obj;
     }
     // assert(ssc == SSC_always_true)... except maybe the profile lied to us.
@@ -2620,11 +2775,59 @@ Node* GraphKit::maybe_cast_profiled_receiver(Node* not_null_obj,
   return NULL;
 }
 
+/**
+ * Cast obj to type and emit guard unless we had too many traps here
+ * already
+ *
+ * @param obj       node being casted
+ * @param type      type to cast the node to
+ * @param not_null  true if we know node cannot be null
+ */
+Node* GraphKit::maybe_cast_profiled_obj(Node* obj,
+                                        ciKlass* type,
+                                        bool not_null) {
+  // type == NULL if profiling tells us this object is always null
+  if (type != NULL) {
+    if (!too_many_traps(Deoptimization::Reason_null_check) &&
+        !too_many_traps(Deoptimization::Reason_class_check)) {
+      Node* not_null_obj = NULL;
+      // not_null is true if we know the object is not null and
+      // there's no need for a null check
+      if (!not_null) {
+        Node* null_ctl = top();
+        not_null_obj = null_check_oop(obj, &null_ctl, true, true);
+        assert(null_ctl->is_top(), "no null control here");
+      } else {
+        not_null_obj = obj;
+      }
+
+      Node* exact_obj = not_null_obj;
+      ciKlass* exact_kls = type;
+      Node* slow_ctl  = type_check_receiver(exact_obj, exact_kls, 1.0,
+                                            &exact_obj);
+      {
+        PreserveJVMState pjvms(this);
+        set_control(slow_ctl);
+        uncommon_trap(Deoptimization::Reason_class_check,
+                      Deoptimization::Action_maybe_recompile);
+      }
+      replace_in_map(not_null_obj, exact_obj);
+      obj = exact_obj;
+    }
+  } else {
+    if (!too_many_traps(Deoptimization::Reason_null_assert)) {
+      Node* exact_obj = null_assert(obj);
+      replace_in_map(obj, exact_obj);
+      obj = exact_obj;
+    }
+  }
+  return obj;
+}
 
 //-------------------------------gen_instanceof--------------------------------
 // Generate an instance-of idiom.  Used by both the instance-of bytecode
 // and the reflective instance-of call.
-Node* GraphKit::gen_instanceof(Node* obj, Node* superklass) {
+Node* GraphKit::gen_instanceof(Node* obj, Node* superklass, bool safe_for_replace) {
   kill_dead_locals();           // Benefit all the uncommon traps
   assert( !stopped(), "dead parse path should be checked in callers" );
   assert(!TypePtr::NULL_PTR->higher_equal(_gvn.type(superklass)->is_klassptr()),
@@ -2637,10 +2840,8 @@ Node* GraphKit::gen_instanceof(Node* obj, Node* superklass) {
   C->set_has_split_ifs(true); // Has chance for split-if optimization
 
   ciProfileData* data = NULL;
-  bool safe_for_replace = false;
   if (java_bc() == Bytecodes::_instanceof) {  // Only for the bytecode
     data = method()->method_data()->bci_to_data(bci());
-    safe_for_replace = true;
   }
   bool never_see_null = (ProfileDynamicTypes  // aggressive use of profile
                          && seems_never_null(obj, data));
@@ -2664,15 +2865,38 @@ Node* GraphKit::gen_instanceof(Node* obj, Node* superklass) {
     phi   ->del_req(_null_path);
   }
 
-  if (ProfileDynamicTypes && data != NULL) {
+  // Do we know the type check always succeed?
-    Node* cast_obj = maybe_cast_profiled_receiver(not_null_obj, data, NULL);
+  bool known_statically = false;
+  if (_gvn.type(superklass)->singleton()) {
+    ciKlass* superk = _gvn.type(superklass)->is_klassptr()->klass();
+    ciKlass* subk = _gvn.type(obj)->is_oopptr()->klass();
+    if (subk != NULL && subk->is_loaded()) {
+      int static_res = static_subtype_check(superk, subk);
+      known_statically = (static_res == SSC_always_true || static_res == SSC_always_false);
+    }
+  }
+
+  if (known_statically && UseTypeSpeculation) {
+    // If we know the type check always succeed then we don't use the
+    // profiling data at this bytecode. Don't lose it, feed it to the
+    // type system as a speculative type.
+    not_null_obj = record_profiled_receiver_for_speculation(not_null_obj);
+  } else {
+    const TypeOopPtr* obj_type = _gvn.type(obj)->is_oopptr();
+    // We may not have profiling here or it may not help us. If we
+    // have a speculative type use it to perform an exact cast.
+    ciKlass* spec_obj_type = obj_type->speculative_type();
+    if (spec_obj_type != NULL || (ProfileDynamicTypes && data != NULL)) {
+      Node* cast_obj = maybe_cast_profiled_receiver(not_null_obj, NULL, spec_obj_type, safe_for_replace);
       if (stopped()) {            // Profile disagrees with this path.
         set_control(null_ctl);    // Null is the only remaining possibility.
         return intcon(0);
       }
-    if (cast_obj != NULL)
+      if (cast_obj != NULL) {
         not_null_obj = cast_obj;
       }
+    }
+  }
 
   // Load the object's klass
   Node* obj_klass = load_object_klass(not_null_obj);
@@ -2718,7 +2942,10 @@ Node* GraphKit::gen_checkcast(Node *obj, Node* superklass,
     if (objtp != NULL && objtp->klass() != NULL) {
       switch (static_subtype_check(tk->klass(), objtp->klass())) {
       case SSC_always_true:
-        return obj;
+        // If we know the type check always succeed then we don't use
+        // the profiling data at this bytecode. Don't lose it, feed it
+        // to the type system as a speculative type.
+        return record_profiled_receiver_for_speculation(obj);
       case SSC_always_false:
         // It needs a null check because a null will *pass* the cast check.
         // A non-null value will always produce an exception.
@@ -2767,12 +2994,17 @@ Node* GraphKit::gen_checkcast(Node *obj, Node* superklass,
   }
 
   Node* cast_obj = NULL;
-  if (data != NULL &&
+  const TypeOopPtr* obj_type = _gvn.type(obj)->is_oopptr();
+  // We may not have profiling here or it may not help us. If we have
+  // a speculative type use it to perform an exact cast.
+  ciKlass* spec_obj_type = obj_type->speculative_type();
+  if (spec_obj_type != NULL ||
+      (data != NULL &&
        // Counter has never been decremented (due to cast failure).
        // ...This is a reasonable thing to expect.  It is true of
        // all casts inserted by javac to implement generic types.
-      data->as_CounterData()->count() >= 0) {
+       data->as_CounterData()->count() >= 0)) {
-    cast_obj = maybe_cast_profiled_receiver(not_null_obj, data, tk->klass());
+    cast_obj = maybe_cast_profiled_receiver(not_null_obj, tk->klass(), spec_obj_type, safe_for_replace);
     if (cast_obj != NULL) {
       if (failure_control != NULL) // failure is now impossible
         (*failure_control) = top();

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

@@ -386,10 +386,33 @@ class GraphKit : public Phase {
   // Check the null_seen bit.
   bool seems_never_null(Node* obj, ciProfileData* data);
 
+  // Check for unique class for receiver at call
+  ciKlass* profile_has_unique_klass() {
+    ciCallProfile profile = method()->call_profile_at_bci(bci());
+    if (profile.count() >= 0 &&         // no cast failures here
+        profile.has_receiver(0) &&
+        profile.morphism() == 1) {
+      return profile.receiver(0);
+    }
+    return NULL;
+  }
+
+  // record type from profiling with the type system
+  Node* record_profile_for_speculation(Node* n, ciKlass* exact_kls);
+  Node* record_profiled_receiver_for_speculation(Node* n);
+  void record_profiled_arguments_for_speculation(ciMethod* dest_method, Bytecodes::Code bc);
+  void record_profiled_parameters_for_speculation();
+
   // Use the type profile to narrow an object type.
   Node* maybe_cast_profiled_receiver(Node* not_null_obj,
-                                     ciProfileData* data,
+                                     ciKlass* require_klass,
-                                     ciKlass* require_klass);
+                                    ciKlass* spec,
+                                     bool safe_for_replace);
+
+  // Cast obj to type and emit guard unless we had too many traps here already
+  Node* maybe_cast_profiled_obj(Node* obj,
+                                ciKlass* type,
+                                bool not_null = false);
 
   // Cast obj to not-null on this path
   Node* cast_not_null(Node* obj, bool do_replace_in_map = true);
@@ -775,7 +798,7 @@ class GraphKit : public Phase {
 
   // Generate an instance-of idiom.  Used by both the instance-of bytecode
   // and the reflective instance-of call.
-  Node* gen_instanceof( Node *subobj, Node* superkls );
+  Node* gen_instanceof(Node *subobj, Node* superkls, bool safe_for_replace = false);
 
   // Generate a check-cast idiom.  Used by both the check-cast bytecode
   // and the array-store bytecode

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

@@ -1019,7 +1019,7 @@ void IfNode::dominated_by( Node *prev_dom, PhaseIterGVN *igvn ) {
   // be skipped. For example, range check predicate has two checks
   // for lower and upper bounds.
   ProjNode* unc_proj = proj_out(1 - prev_dom->as_Proj()->_con)->as_Proj();
-  if (PhaseIdealLoop::is_uncommon_trap_proj(unc_proj, Deoptimization::Reason_predicate))
+  if (unc_proj->is_uncommon_trap_proj(Deoptimization::Reason_predicate))
     prev_dom = idom;
 
   // Now walk the current IfNode's projections.