8141132: JEP 254: Compact Strings
authorthartmann
Tue, 03 Nov 2015 09:41:03 +0100
changeset 33628 09241459a8b8
parent 33627 c5b7455f846e
child 33629 35bdf0672732
8141132: JEP 254: Compact Strings Summary: Adopt a more space-efficient internal representation for strings. Reviewed-by: alanb, bdelsart, coleenp, iklam, jiangli, jrose, kevinw, naoto, pliden, roland, smarks, twisti Contributed-by: Brent Christian <brent.christian@oracle.com>, Vivek Deshpande <vivek.r.deshpande@intel.com>, Tobias Hartmann <tobias.hartmann@oracle.com>, Charlie Hunt <charlie.hunt@oracle.com>, Vladimir Kozlov <vladimir.kozlov@oracle.com>, Roger Riggs <roger.riggs@oracle.com>, Xueming Shen <xueming.shen@oracle.com>, Aleksey Shipilev <aleksey.shipilev@oracle.com>, Sandhya Viswanathan <sandhya.viswanathan@intel.com>
hotspot/agent/src/share/classes/sun/jvm/hotspot/oops/OopUtilities.java
hotspot/agent/src/share/classes/sun/jvm/hotspot/tools/HeapSummary.java
hotspot/agent/src/share/classes/sun/jvm/hotspot/utilities/Hashtable.java
hotspot/src/cpu/aarch64/vm/aarch64.ad
hotspot/src/cpu/aarch64/vm/c1_LIRAssembler_aarch64.cpp
hotspot/src/cpu/aarch64/vm/globals_aarch64.hpp
hotspot/src/cpu/ppc/vm/globals_ppc.hpp
hotspot/src/cpu/ppc/vm/ppc.ad
hotspot/src/cpu/sparc/vm/assembler_sparc.hpp
hotspot/src/cpu/sparc/vm/c1_LIRAssembler_sparc.cpp
hotspot/src/cpu/sparc/vm/globals_sparc.hpp
hotspot/src/cpu/sparc/vm/macroAssembler_sparc.cpp
hotspot/src/cpu/sparc/vm/macroAssembler_sparc.hpp
hotspot/src/cpu/sparc/vm/sparc.ad
hotspot/src/cpu/x86/vm/assembler_x86.cpp
hotspot/src/cpu/x86/vm/assembler_x86.hpp
hotspot/src/cpu/x86/vm/c1_LIRAssembler_x86.cpp
hotspot/src/cpu/x86/vm/globals_x86.hpp
hotspot/src/cpu/x86/vm/macroAssembler_x86.cpp
hotspot/src/cpu/x86/vm/macroAssembler_x86.hpp
hotspot/src/cpu/x86/vm/x86_32.ad
hotspot/src/cpu/x86/vm/x86_64.ad
hotspot/src/cpu/zero/vm/globals_zero.hpp
hotspot/src/share/vm/adlc/formssel.cpp
hotspot/src/share/vm/adlc/main.cpp
hotspot/src/share/vm/c1/c1_Compiler.cpp
hotspot/src/share/vm/c1/c1_GraphBuilder.cpp
hotspot/src/share/vm/c1/c1_GraphBuilder.hpp
hotspot/src/share/vm/c1/c1_LIRAssembler.hpp
hotspot/src/share/vm/ci/ciTypeArray.cpp
hotspot/src/share/vm/ci/ciTypeArray.hpp
hotspot/src/share/vm/classfile/javaClasses.cpp
hotspot/src/share/vm/classfile/javaClasses.hpp
hotspot/src/share/vm/classfile/stringTable.cpp
hotspot/src/share/vm/classfile/stringTable.hpp
hotspot/src/share/vm/classfile/symbolTable.cpp
hotspot/src/share/vm/classfile/symbolTable.hpp
hotspot/src/share/vm/classfile/vmSymbols.cpp
hotspot/src/share/vm/classfile/vmSymbols.hpp
hotspot/src/share/vm/gc/g1/g1StringDedupTable.cpp
hotspot/src/share/vm/gc/g1/g1StringDedupTable.hpp
hotspot/src/share/vm/memory/filemap.cpp
hotspot/src/share/vm/memory/filemap.hpp
hotspot/src/share/vm/memory/metaspaceShared.cpp
hotspot/src/share/vm/oops/instanceKlass.cpp
hotspot/src/share/vm/oops/typeArrayOop.hpp
hotspot/src/share/vm/opto/c2compiler.cpp
hotspot/src/share/vm/opto/classes.hpp
hotspot/src/share/vm/opto/escape.cpp
hotspot/src/share/vm/opto/gcm.cpp
hotspot/src/share/vm/opto/graphKit.cpp
hotspot/src/share/vm/opto/graphKit.hpp
hotspot/src/share/vm/opto/idealKit.cpp
hotspot/src/share/vm/opto/idealKit.hpp
hotspot/src/share/vm/opto/intrinsicnode.cpp
hotspot/src/share/vm/opto/intrinsicnode.hpp
hotspot/src/share/vm/opto/lcm.cpp
hotspot/src/share/vm/opto/library_call.cpp
hotspot/src/share/vm/opto/loopTransform.cpp
hotspot/src/share/vm/opto/loopnode.cpp
hotspot/src/share/vm/opto/macro.cpp
hotspot/src/share/vm/opto/matcher.cpp
hotspot/src/share/vm/opto/stringopts.cpp
hotspot/src/share/vm/opto/stringopts.hpp
hotspot/src/share/vm/prims/jni.cpp
hotspot/src/share/vm/prims/jvmtiEnv.cpp
hotspot/src/share/vm/prims/jvmtiTagMap.cpp
hotspot/src/share/vm/runtime/globals.hpp
hotspot/src/share/vm/runtime/thread.cpp
hotspot/src/share/vm/utilities/utf8.cpp
hotspot/src/share/vm/utilities/utf8.hpp
hotspot/test/compiler/intrinsics/string/TestStringIntrinsics.java
hotspot/test/runtime/Annotations/TestAnnotatedStringEncoding.java
hotspot/test/runtime/SharedArchiveFile/CdsDifferentCompactStrings.java
--- a/hotspot/agent/src/share/classes/sun/jvm/hotspot/oops/OopUtilities.java	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/agent/src/share/classes/sun/jvm/hotspot/oops/OopUtilities.java	Tue Nov 03 09:41:03 2015 +0100
@@ -40,8 +40,7 @@
   // FIXME: access should be synchronized and cleared when VM is
   // resumed
   // String fields
-  private static IntField offsetField;
-  private static IntField countField;
+  private static ByteField coderField;
   private static OopField valueField;
   // ThreadGroup fields
   private static OopField threadGroupParentField;
@@ -96,20 +95,30 @@
     if (charArray == null) {
       return null;
     }
-    return charArrayToString(charArray, 0, (int) charArray.getLength());
+    int length = (int)charArray.getLength();
+    StringBuffer buf = new StringBuffer(length);
+    for (int i = 0; i < length; i++) {
+      buf.append(charArray.getCharAt(i));
+    }
+    return buf.toString();
   }
 
-  public static String charArrayToString(TypeArray charArray, int offset, int length) {
-    if (charArray == null) {
+  public static String byteArrayToString(TypeArray byteArray, byte coder) {
+    if (byteArray == null) {
       return null;
     }
-    final int limit = offset + length;
-    if (Assert.ASSERTS_ENABLED) {
-      Assert.that(offset >= 0 && limit <= charArray.getLength(), "out of bounds");
-    }
+    int length = (int)byteArray.getLength() >> coder;
     StringBuffer buf = new StringBuffer(length);
-    for (int i = offset; i < limit; i++) {
-      buf.append(charArray.getCharAt(i));
+    if (coder == 0) {
+      // Latin1 encoded
+      for (int i = 0; i < length; i++) {
+        buf.append((char)(byteArray.getByteAt(i) & 0xff));
+      }
+    } else {
+      // UTF16 encoded
+      for (int i = 0; i < length; i++) {
+        buf.append(byteArray.getCharAt(i));
+      }
     }
     return buf.toString();
   }
@@ -141,21 +150,14 @@
   }
 
   public static String stringOopToString(Oop stringOop) {
-    if (offsetField == null) {
-      InstanceKlass k = (InstanceKlass) stringOop.getKlass();
-      offsetField = (IntField) k.findField("offset", "I");   // optional
-      countField  = (IntField) k.findField("count",  "I");   // optional
-      valueField  = (OopField) k.findField("value",  "[C");
-      if (Assert.ASSERTS_ENABLED) {
-         Assert.that(valueField != null, "Field \'value\' of java.lang.String not found");
-      }
+    InstanceKlass k = (InstanceKlass) stringOop.getKlass();
+    coderField  = (ByteField) k.findField("coder", "B");
+    valueField  = (OopField) k.findField("value",  "[B");
+    if (Assert.ASSERTS_ENABLED) {
+       Assert.that(coderField != null, "Field \'coder\' of java.lang.String not found");
+       Assert.that(valueField != null, "Field \'value\' of java.lang.String not found");
     }
-    if (offsetField != null && countField != null) {
-      return charArrayToString((TypeArray) valueField.getValue(stringOop),
-                               offsetField.getValue(stringOop),
-                               countField.getValue(stringOop));
-    }
-    return  charArrayToString((TypeArray) valueField.getValue(stringOop));
+    return byteArrayToString((TypeArray) valueField.getValue(stringOop), coderField.getValue(stringOop));
   }
 
   public static String stringOopToEscapedString(Oop stringOop) {
--- a/hotspot/agent/src/share/classes/sun/jvm/hotspot/tools/HeapSummary.java	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/agent/src/share/classes/sun/jvm/hotspot/tools/HeapSummary.java	Tue Nov 03 09:41:03 2015 +0100
@@ -268,8 +268,8 @@
             VM vm = VM.getVM();
             SystemDictionary sysDict = vm.getSystemDictionary();
             InstanceKlass strKlass = sysDict.getStringKlass();
-            // String has a field named 'value' of type 'char[]'.
-            stringValueField = (OopField) strKlass.findField("value", "[C");
+            // String has a field named 'value' of type 'byte[]'.
+            stringValueField = (OopField) strKlass.findField("value", "[B");
          }
 
          private long stringSize(Instance instance) {
--- a/hotspot/agent/src/share/classes/sun/jvm/hotspot/utilities/Hashtable.java	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/agent/src/share/classes/sun/jvm/hotspot/utilities/Hashtable.java	Tue Nov 03 09:41:03 2015 +0100
@@ -61,9 +61,8 @@
     long h = 0;
     int s = 0;
     int len = buf.length;
-    // Emulate the unsigned int in java_lang_String::hash_code
     while (len-- > 0) {
-      h = 31*h + (0xFFFFFFFFL & buf[s]);
+      h = 31*h + (0xFFL & buf[s]);
       s++;
     }
     return h & 0xFFFFFFFFL;
--- a/hotspot/src/cpu/aarch64/vm/aarch64.ad	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/cpu/aarch64/vm/aarch64.ad	Tue Nov 03 09:41:03 2015 +0100
@@ -14150,6 +14150,7 @@
 instruct string_compare(iRegP_R1 str1, iRegI_R2 cnt1, iRegP_R3 str2, iRegI_R4 cnt2,
                         iRegI_R0 result, iRegP_R10 tmp1, rFlagsReg cr)
 %{
+  predicate(!CompactStrings);
   match(Set result (StrComp (Binary str1 cnt1) (Binary str2 cnt2)));
   effect(KILL tmp1, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL cr);
 
@@ -14165,6 +14166,7 @@
 instruct string_indexof(iRegP_R1 str1, iRegI_R4 cnt1, iRegP_R3 str2, iRegI_R2 cnt2,
        iRegI_R0 result, iRegI tmp1, iRegI tmp2, iRegI tmp3, iRegI tmp4, rFlagsReg cr)
 %{
+  predicate(!CompactStrings);
   match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 cnt2)));
   effect(USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2,
          TEMP tmp1, TEMP tmp2, TEMP tmp3, TEMP tmp4, KILL cr);
@@ -14184,6 +14186,7 @@
                  immI_le_4 int_cnt2, iRegI_R0 result, iRegI tmp1, iRegI tmp2,
                  iRegI tmp3, iRegI tmp4, rFlagsReg cr)
 %{
+  predicate(!CompactStrings);
   match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 int_cnt2)));
   effect(USE_KILL str1, USE_KILL str2, USE_KILL cnt1,
          TEMP tmp1, TEMP tmp2, TEMP tmp3, TEMP tmp4, KILL cr);
@@ -14203,6 +14206,7 @@
 instruct string_equals(iRegP_R1 str1, iRegP_R3 str2, iRegI_R4 cnt,
                         iRegI_R0 result, iRegP_R10 tmp, rFlagsReg cr)
 %{
+  predicate(!CompactStrings);
   match(Set result (StrEquals (Binary str1 str2) cnt));
   effect(KILL tmp, USE_KILL str1, USE_KILL str2, USE_KILL cnt, KILL cr);
 
@@ -14218,6 +14222,7 @@
 instruct array_equals(iRegP_R1 ary1, iRegP_R2 ary2, iRegI_R0 result,
                       iRegP_R10 tmp, rFlagsReg cr)
 %{
+  predicate(((AryEqNode*)n)->encoding() == StrIntrinsicNode::UU);
   match(Set result (AryEq ary1 ary2));
   effect(KILL tmp, USE_KILL ary1, USE_KILL ary2, KILL cr);
 
--- a/hotspot/src/cpu/aarch64/vm/c1_LIRAssembler_aarch64.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/cpu/aarch64/vm/c1_LIRAssembler_aarch64.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -483,15 +483,6 @@
   return offset;
 }
 
-
-// This is the fast version of java.lang.String.compare; it has not
-// OSR-entry and therefore, we generate a slow version for OSR's
-void LIR_Assembler::emit_string_compare(LIR_Opr arg0, LIR_Opr arg1, LIR_Opr dst, CodeEmitInfo* info)  {
-  __ mov(r2, (address)__FUNCTION__);
-  __ call_Unimplemented();
-}
-
-
 void LIR_Assembler::add_debug_info_for_branch(address adr, CodeEmitInfo* info) {
   _masm->code_section()->relocate(adr, relocInfo::poll_type);
   int pc_offset = code_offset();
--- a/hotspot/src/cpu/aarch64/vm/globals_aarch64.hpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/cpu/aarch64/vm/globals_aarch64.hpp	Tue Nov 03 09:41:03 2015 +0100
@@ -79,6 +79,9 @@
 
 define_pd_global(uintx, TypeProfileLevel, 111);
 
+// No performance work done here yet.
+define_pd_global(bool, CompactStrings, false);
+
 // avoid biased locking while we are bootstrapping the aarch64 build
 define_pd_global(bool, UseBiasedLocking, false);
 
--- a/hotspot/src/cpu/ppc/vm/globals_ppc.hpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/cpu/ppc/vm/globals_ppc.hpp	Tue Nov 03 09:41:03 2015 +0100
@@ -72,6 +72,9 @@
 
 define_pd_global(uintx, TypeProfileLevel, 111);
 
+// No performance work done here yet.
+define_pd_global(bool, CompactStrings, false);
+
 // Platform dependent flag handling: flags only defined on this platform.
 #define ARCH_FLAGS(develop, product, diagnostic, experimental, notproduct, range, constraint)  \
                                                                             \
--- a/hotspot/src/cpu/ppc/vm/ppc.ad	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/cpu/ppc/vm/ppc.ad	Tue Nov 03 09:41:03 2015 +0100
@@ -2054,11 +2054,11 @@
     return (UsePopCountInstruction && VM_Version::has_popcntw());
 
   case Op_StrComp:
-    return SpecialStringCompareTo;
+    return SpecialStringCompareTo && !CompactStrings;
   case Op_StrEquals:
-    return SpecialStringEquals;
+    return SpecialStringEquals && !CompactStrings;
   case Op_StrIndexOf:
-    return SpecialStringIndexOf;
+    return SpecialStringIndexOf && !CompactStrings;
   }
 
   return true;  // Per default match rules are supported.
@@ -11077,7 +11077,7 @@
                                   immP needleImm, immL offsetImm, immI_1 needlecntImm,
                                   iRegIdst tmp1, iRegIdst tmp2,
                                   flagsRegCR0 cr0, flagsRegCR1 cr1) %{
-  predicate(SpecialStringIndexOf);  // type check implicit by parameter type, See Matcher::match_rule_supported
+  predicate(SpecialStringIndexOf && !CompactStrings);  // type check implicit by parameter type, See Matcher::match_rule_supported
   match(Set result (StrIndexOf (Binary haystack haycnt) (Binary (AddP needleImm offsetImm) needlecntImm)));
 
   effect(TEMP_DEF result, TEMP tmp1, TEMP tmp2, KILL cr0, KILL cr1);
@@ -11120,7 +11120,7 @@
   effect(USE_KILL needle, /* TDEF needle, */ TEMP_DEF result,
          TEMP tmp1, TEMP tmp2);
   // Required for EA: check if it is still a type_array.
-  predicate(SpecialStringIndexOf && n->in(3)->in(1)->bottom_type()->is_aryptr()->const_oop() &&
+  predicate(SpecialStringIndexOf && !CompactStrings && n->in(3)->in(1)->bottom_type()->is_aryptr()->const_oop() &&
             n->in(3)->in(1)->bottom_type()->is_aryptr()->const_oop()->is_type_array());
   ins_cost(180);
 
@@ -11167,7 +11167,7 @@
   effect(USE_KILL haycnt, /* better: TDEF haycnt, */ TEMP_DEF result,
          TEMP tmp1, TEMP tmp2, TEMP tmp3, TEMP tmp4, TEMP tmp5, KILL cr0, KILL cr1, KILL cr6);
   // Required for EA: check if it is still a type_array.
-  predicate(SpecialStringIndexOf && n->in(3)->in(1)->bottom_type()->is_aryptr()->const_oop() &&
+  predicate(SpecialStringIndexOf && !CompactStrings && n->in(3)->in(1)->bottom_type()->is_aryptr()->const_oop() &&
             n->in(3)->in(1)->bottom_type()->is_aryptr()->const_oop()->is_type_array());
   ins_cost(250);
 
@@ -11200,7 +11200,7 @@
   effect(USE_KILL haycnt, USE_KILL needlecnt, /*better: TDEF haycnt, TDEF needlecnt,*/
          TEMP_DEF result,
          TEMP tmp1, TEMP tmp2, TEMP tmp3, TEMP tmp4, KILL cr0, KILL cr1, KILL cr6);
-  predicate(SpecialStringIndexOf);  // See Matcher::match_rule_supported.
+  predicate(SpecialStringIndexOf && !CompactStrings);  // See Matcher::match_rule_supported.
   ins_cost(300);
 
   ins_alignment(8); // 'compute_padding()' gets called, up to this number-1 nops will get inserted.
@@ -11224,7 +11224,7 @@
   match(Set result (StrEquals (Binary str1 str2) cntImm));
   effect(TEMP_DEF result, TEMP tmp1, TEMP tmp2,
          KILL cr0, KILL cr6, KILL ctr);
-  predicate(SpecialStringEquals);  // See Matcher::match_rule_supported.
+  predicate(SpecialStringEquals && !CompactStrings);  // See Matcher::match_rule_supported.
   ins_cost(250);
 
   ins_alignment(8); // 'compute_padding()' gets called, up to this number-1 nops will get inserted.
@@ -11247,7 +11247,7 @@
   match(Set result (StrEquals (Binary str1 str2) cnt));
   effect(TEMP_DEF result, TEMP tmp1, TEMP tmp2, TEMP tmp3, TEMP tmp4, TEMP tmp5,
          KILL cr0, KILL cr1, KILL cr6, KILL ctr);
-  predicate(SpecialStringEquals);  // See Matcher::match_rule_supported.
+  predicate(SpecialStringEquals && !CompactStrings);  // See Matcher::match_rule_supported.
   ins_cost(300);
 
   ins_alignment(8); // 'compute_padding()' gets called, up to this number-1 nops will get inserted.
@@ -11267,6 +11267,7 @@
 // Use dst register classes if register gets killed, as it is the case for TEMP operands!
 instruct string_compare(rarg1RegP str1, rarg2RegP str2, rarg3RegI cnt1, rarg4RegI cnt2, iRegIdst result,
                         iRegPdst tmp, flagsRegCR0 cr0, regCTR ctr) %{
+  predicate(!CompactStrings);
   match(Set result (StrComp (Binary str1 cnt1) (Binary str2 cnt2)));
   effect(USE_KILL cnt1, USE_KILL cnt2, USE_KILL str1, USE_KILL str2, TEMP_DEF result, TEMP tmp, KILL cr0, KILL ctr);
   ins_cost(300);
--- a/hotspot/src/cpu/sparc/vm/assembler_sparc.hpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/cpu/sparc/vm/assembler_sparc.hpp	Tue Nov 03 09:41:03 2015 +0100
@@ -124,6 +124,8 @@
     impdep1_op3  = 0x36,
     aes3_op3     = 0x36,
     sha_op3      = 0x36,
+    bmask_op3    = 0x36,
+    bshuffle_op3   = 0x36,
     alignaddr_op3  = 0x36,
     faligndata_op3 = 0x36,
     flog3_op3    = 0x36,
@@ -194,6 +196,7 @@
     fnegd_opf          = 0x06,
 
     alignaddr_opf      = 0x18,
+    bmask_opf          = 0x19,
 
     fadds_opf          = 0x41,
     faddd_opf          = 0x42,
@@ -204,6 +207,7 @@
 
     fmuls_opf          = 0x49,
     fmuld_opf          = 0x4a,
+    bshuffle_opf       = 0x4c,
     fdivs_opf          = 0x4d,
     fdivd_opf          = 0x4e,
 
@@ -1226,6 +1230,9 @@
 
   void edge8n( Register s1, Register s2, Register d ) { vis2_only(); emit_int32( op(arith_op) | rd(d) | op3(edge_op3) | rs1(s1) | opf(edge8n_opf) | rs2(s2)); }
 
+  void bmask( Register s1, Register s2, Register d ) { vis2_only(); emit_int32( op(arith_op) | rd(d) | op3(bmask_op3) | rs1(s1) | opf(bmask_opf) | rs2(s2)); }
+  void bshuffle( FloatRegister s1, FloatRegister s2, FloatRegister d ) { vis2_only(); emit_int32( op(arith_op) | fd(d, FloatRegisterImpl::D) | op3(bshuffle_op3) | fs1(s1, FloatRegisterImpl::D) | opf(bshuffle_opf) | fs2(s2, FloatRegisterImpl::D)); }
+
   // VIS3 instructions
 
   void movstosw( FloatRegister s, Register d ) { vis3_only();  emit_int32( op(arith_op) | rd(d) | op3(mftoi_op3) | opf(mstosw_opf) | fs2(s, FloatRegisterImpl::S)); }
--- a/hotspot/src/cpu/sparc/vm/c1_LIRAssembler_sparc.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/cpu/sparc/vm/c1_LIRAssembler_sparc.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -232,118 +232,6 @@
 }
 
 
-// Optimized Library calls
-// This is the fast version of java.lang.String.compare; it has not
-// OSR-entry and therefore, we generate a slow version for OSR's
-void LIR_Assembler::emit_string_compare(LIR_Opr left, LIR_Opr right, LIR_Opr dst, CodeEmitInfo* info) {
-  Register str0 = left->as_register();
-  Register str1 = right->as_register();
-
-  Label Ldone;
-
-  Register result = dst->as_register();
-  {
-    // Get a pointer to the first character of string0 in tmp0
-    //   and get string0.length() in str0
-    // Get a pointer to the first character of string1 in tmp1
-    //   and get string1.length() in str1
-    // Also, get string0.length()-string1.length() in
-    //   o7 and get the condition code set
-    // Note: some instructions have been hoisted for better instruction scheduling
-
-    Register tmp0 = L0;
-    Register tmp1 = L1;
-    Register tmp2 = L2;
-
-    int  value_offset = java_lang_String:: value_offset_in_bytes(); // char array
-    if (java_lang_String::has_offset_field()) {
-      int offset_offset = java_lang_String::offset_offset_in_bytes(); // first character position
-      int  count_offset = java_lang_String:: count_offset_in_bytes();
-      __ load_heap_oop(str0, value_offset, tmp0);
-      __ ld(str0, offset_offset, tmp2);
-      __ add(tmp0, arrayOopDesc::base_offset_in_bytes(T_CHAR), tmp0);
-      __ ld(str0, count_offset, str0);
-      __ sll(tmp2, exact_log2(sizeof(jchar)), tmp2);
-    } else {
-      __ load_heap_oop(str0, value_offset, tmp1);
-      __ add(tmp1, arrayOopDesc::base_offset_in_bytes(T_CHAR), tmp0);
-      __ ld(tmp1, arrayOopDesc::length_offset_in_bytes(), str0);
-    }
-
-    // str1 may be null
-    add_debug_info_for_null_check_here(info);
-
-    if (java_lang_String::has_offset_field()) {
-      int offset_offset = java_lang_String::offset_offset_in_bytes(); // first character position
-      int  count_offset = java_lang_String:: count_offset_in_bytes();
-      __ load_heap_oop(str1, value_offset, tmp1);
-      __ add(tmp0, tmp2, tmp0);
-
-      __ ld(str1, offset_offset, tmp2);
-      __ add(tmp1, arrayOopDesc::base_offset_in_bytes(T_CHAR), tmp1);
-      __ ld(str1, count_offset, str1);
-      __ sll(tmp2, exact_log2(sizeof(jchar)), tmp2);
-      __ add(tmp1, tmp2, tmp1);
-    } else {
-      __ load_heap_oop(str1, value_offset, tmp2);
-      __ add(tmp2, arrayOopDesc::base_offset_in_bytes(T_CHAR), tmp1);
-      __ ld(tmp2, arrayOopDesc::length_offset_in_bytes(), str1);
-    }
-    __ subcc(str0, str1, O7);
-  }
-
-  {
-    // Compute the minimum of the string lengths, scale it and store it in limit
-    Register count0 = I0;
-    Register count1 = I1;
-    Register limit  = L3;
-
-    Label Lskip;
-    __ sll(count0, exact_log2(sizeof(jchar)), limit);             // string0 is shorter
-    __ br(Assembler::greater, true, Assembler::pt, Lskip);
-    __ delayed()->sll(count1, exact_log2(sizeof(jchar)), limit);  // string1 is shorter
-    __ bind(Lskip);
-
-    // If either string is empty (or both of them) the result is the difference in lengths
-    __ cmp(limit, 0);
-    __ br(Assembler::equal, true, Assembler::pn, Ldone);
-    __ delayed()->mov(O7, result);  // result is difference in lengths
-  }
-
-  {
-    // Neither string is empty
-    Label Lloop;
-
-    Register base0 = L0;
-    Register base1 = L1;
-    Register chr0  = I0;
-    Register chr1  = I1;
-    Register limit = L3;
-
-    // Shift base0 and base1 to the end of the arrays, negate limit
-    __ add(base0, limit, base0);
-    __ add(base1, limit, base1);
-    __ neg(limit);  // limit = -min{string0.length(), string1.length()}
-
-    __ lduh(base0, limit, chr0);
-    __ bind(Lloop);
-    __ lduh(base1, limit, chr1);
-    __ subcc(chr0, chr1, chr0);
-    __ br(Assembler::notZero, false, Assembler::pn, Ldone);
-    assert(chr0 == result, "result must be pre-placed");
-    __ delayed()->inccc(limit, sizeof(jchar));
-    __ br(Assembler::notZero, true, Assembler::pt, Lloop);
-    __ delayed()->lduh(base0, limit, chr0);
-  }
-
-  // If strings are equal up to min length, return the length difference.
-  __ mov(O7, result);
-
-  // Otherwise, return the difference between the first mismatched chars.
-  __ bind(Ldone);
-}
-
-
 // --------------------------------------------------------------------------------------------
 
 void LIR_Assembler::monitorexit(LIR_Opr obj_opr, LIR_Opr lock_opr, Register hdr, int monitor_no) {
--- a/hotspot/src/cpu/sparc/vm/globals_sparc.hpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/cpu/sparc/vm/globals_sparc.hpp	Tue Nov 03 09:41:03 2015 +0100
@@ -86,6 +86,8 @@
 
 define_pd_global(uintx, TypeProfileLevel, 111);
 
+define_pd_global(bool, CompactStrings, true);
+
 #define ARCH_FLAGS(develop, product, diagnostic, experimental, notproduct, range, constraint) \
                                                                             \
   product(intx, UseVIS, 99,                                                 \
--- a/hotspot/src/cpu/sparc/vm/macroAssembler_sparc.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/cpu/sparc/vm/macroAssembler_sparc.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -44,6 +44,9 @@
 #include "gc/g1/g1SATBCardTableModRefBS.hpp"
 #include "gc/g1/heapRegion.hpp"
 #endif // INCLUDE_ALL_GCS
+#ifdef COMPILER2
+#include "opto/intrinsicnode.hpp"
+#endif
 
 #ifdef PRODUCT
 #define BLOCK_COMMENT(str) /* nothing */
@@ -4253,27 +4256,385 @@
   }
 }
 
-// Compare char[] arrays aligned to 4 bytes.
-void MacroAssembler::char_arrays_equals(Register ary1, Register ary2,
-                                        Register limit, Register result,
-                                        Register chr1, Register chr2, Label& Ldone) {
-  Label Lvector, Lloop;
-  assert(chr1 == result, "should be the same");
-
-  // Note: limit contains number of bytes (2*char_elements) != 0.
-  andcc(limit, 0x2, chr1); // trailing character ?
+#ifdef COMPILER2
+
+// Compress char[] to byte[] by compressing 16 bytes at once. Return 0 on failure.
+void MacroAssembler::string_compress_16(Register src, Register dst, Register cnt, Register result,
+                                        Register tmp1, Register tmp2, Register tmp3, Register tmp4,
+                                        FloatRegister ftmp1, FloatRegister ftmp2, FloatRegister ftmp3, Label& Ldone) {
+  Label Lloop, Lslow;
+  assert(UseVIS >= 3, "VIS3 is required");
+  assert_different_registers(src, dst, cnt, tmp1, tmp2, tmp3, tmp4, result);
+  assert_different_registers(ftmp1, ftmp2, ftmp3);
+
+  // Check if cnt >= 8 (= 16 bytes)
+  cmp(cnt, 8);
+  br(Assembler::less, false, Assembler::pn, Lslow);
+  delayed()->mov(cnt, result); // copy count
+
+  // Check for 8-byte alignment of src and dst
+  or3(src, dst, tmp1);
+  andcc(tmp1, 7, G0);
+  br(Assembler::notZero, false, Assembler::pn, Lslow);
+  delayed()->nop();
+
+  // Set mask for bshuffle instruction
+  Register mask = tmp4;
+  set(0x13579bdf, mask);
+  bmask(mask, G0, G0);
+
+  // Set mask to 0xff00 ff00 ff00 ff00 to check for non-latin1 characters
+  Assembler::sethi(0xff00fc00, mask); // mask = 0x0000 0000 ff00 fc00
+  add(mask, 0x300, mask);             // mask = 0x0000 0000 ff00 ff00
+  sllx(mask, 32, tmp1);               // tmp1 = 0xff00 ff00 0000 0000
+  or3(mask, tmp1, mask);              // mask = 0xff00 ff00 ff00 ff00
+
+  // Load first 8 bytes
+  ldx(src, 0, tmp1);
+
+  bind(Lloop);
+  // Load next 8 bytes
+  ldx(src, 8, tmp2);
+
+  // Check for non-latin1 character by testing if the most significant byte of a char is set.
+  // Although we have to move the data between integer and floating point registers, this is
+  // still faster than the corresponding VIS instructions (ford/fand/fcmpd).
+  or3(tmp1, tmp2, tmp3);
+  btst(tmp3, mask);
+  // annul zeroing if branch is not taken to preserve original count
+  brx(Assembler::notZero, true, Assembler::pn, Ldone);
+  delayed()->mov(G0, result); // 0 - failed
+
+  // Move bytes into float register
+  movxtod(tmp1, ftmp1);
+  movxtod(tmp2, ftmp2);
+
+  // Compress by copying one byte per char from ftmp1 and ftmp2 to ftmp3
+  bshuffle(ftmp1, ftmp2, ftmp3);
+  stf(FloatRegisterImpl::D, ftmp3, dst, 0);
+
+  // Increment addresses and decrement count
+  inc(src, 16);
+  inc(dst, 8);
+  dec(cnt, 8);
+
+  cmp(cnt, 8);
+  // annul LDX if branch is not taken to prevent access past end of string
+  br(Assembler::greaterEqual, true, Assembler::pt, Lloop);
+  delayed()->ldx(src, 0, tmp1);
+
+  // Fallback to slow version
+  bind(Lslow);
+}
+
+// Compress char[] to byte[]. Return 0 on failure.
+void MacroAssembler::string_compress(Register src, Register dst, Register cnt, Register result, Register tmp, Label& Ldone) {
+  Label Lloop;
+  assert_different_registers(src, dst, cnt, tmp, result);
+
+  lduh(src, 0, tmp);
+
+  bind(Lloop);
+  inc(src, sizeof(jchar));
+  cmp(tmp, 0xff);
+  // annul zeroing if branch is not taken to preserve original count
+  br(Assembler::greater, true, Assembler::pn, Ldone); // don't check xcc
+  delayed()->mov(G0, result); // 0 - failed
+  deccc(cnt);
+  stb(tmp, dst, 0);
+  inc(dst);
+  // annul LDUH if branch is not taken to prevent access past end of string
+  br(Assembler::notZero, true, Assembler::pt, Lloop);
+  delayed()->lduh(src, 0, tmp); // hoisted
+}
+
+// Inflate byte[] to char[] by inflating 16 bytes at once.
+void MacroAssembler::string_inflate_16(Register src, Register dst, Register cnt, Register tmp,
+                                       FloatRegister ftmp1, FloatRegister ftmp2, FloatRegister ftmp3, FloatRegister ftmp4, Label& Ldone) {
+  Label Lloop, Lslow;
+  assert(UseVIS >= 3, "VIS3 is required");
+  assert_different_registers(src, dst, cnt, tmp);
+  assert_different_registers(ftmp1, ftmp2, ftmp3, ftmp4);
+
+  // Check if cnt >= 8 (= 16 bytes)
+  cmp(cnt, 8);
+  br(Assembler::less, false, Assembler::pn, Lslow);
+  delayed()->nop();
+
+  // Check for 8-byte alignment of src and dst
+  or3(src, dst, tmp);
+  andcc(tmp, 7, G0);
+  br(Assembler::notZero, false, Assembler::pn, Lslow);
+  // Initialize float register to zero
+  FloatRegister zerof = ftmp4;
+  delayed()->fzero(FloatRegisterImpl::D, zerof);
+
+  // Load first 8 bytes
+  ldf(FloatRegisterImpl::D, src, 0, ftmp1);
+
+  bind(Lloop);
+  inc(src, 8);
+  dec(cnt, 8);
+
+  // Inflate the string by interleaving each byte from the source array
+  // with a zero byte and storing the result in the destination array.
+  fpmerge(zerof, ftmp1->successor(), ftmp2);
+  stf(FloatRegisterImpl::D, ftmp2, dst, 8);
+  fpmerge(zerof, ftmp1, ftmp3);
+  stf(FloatRegisterImpl::D, ftmp3, dst, 0);
+
+  inc(dst, 16);
+
+  cmp(cnt, 8);
+  // annul LDX if branch is not taken to prevent access past end of string
+  br(Assembler::greaterEqual, true, Assembler::pt, Lloop);
+  delayed()->ldf(FloatRegisterImpl::D, src, 0, ftmp1);
+
+  // Fallback to slow version
+  bind(Lslow);
+}
+
+// Inflate byte[] to char[].
+void MacroAssembler::string_inflate(Register src, Register dst, Register cnt, Register tmp, Label& Ldone) {
+  Label Loop;
+  assert_different_registers(src, dst, cnt, tmp);
+
+  ldub(src, 0, tmp);
+  bind(Loop);
+  inc(src);
+  deccc(cnt);
+  sth(tmp, dst, 0);
+  inc(dst, sizeof(jchar));
+  // annul LDUB if branch is not taken to prevent access past end of string
+  br(Assembler::notZero, true, Assembler::pt, Loop);
+  delayed()->ldub(src, 0, tmp); // hoisted
+}
+
+void MacroAssembler::string_compare(Register str1, Register str2,
+                                    Register cnt1, Register cnt2,
+                                    Register tmp1, Register tmp2,
+                                    Register result, int ae) {
+  Label Ldone, Lloop;
+  assert_different_registers(str1, str2, cnt1, cnt2, tmp1, result);
+  int stride1, stride2;
+
+  // Note: Making use of the fact that compareTo(a, b) == -compareTo(b, a)
+  // we interchange str1 and str2 in the UL case and negate the result.
+  // Like this, str1 is always latin1 encoded, expect for the UU case.
+
+  if (ae == StrIntrinsicNode::LU || ae == StrIntrinsicNode::UL) {
+    srl(cnt2, 1, cnt2);
+  }
+
+  // See if the lengths are different, and calculate min in cnt1.
+  // Save diff in case we need it for a tie-breaker.
+  Label Lskip;
+  Register diff = tmp1;
+  subcc(cnt1, cnt2, diff);
+  br(Assembler::greater, true, Assembler::pt, Lskip);
+  // cnt2 is shorter, so use its count:
+  delayed()->mov(cnt2, cnt1);
+  bind(Lskip);
+
+  // Rename registers
+  Register limit1 = cnt1;
+  Register limit2 = limit1;
+  Register chr1   = result;
+  Register chr2   = cnt2;
+  if (ae == StrIntrinsicNode::LU || ae == StrIntrinsicNode::UL) {
+    // We need an additional register to keep track of two limits
+    assert_different_registers(str1, str2, cnt1, cnt2, tmp1, tmp2, result);
+    limit2 = tmp2;
+  }
+
+  // Is the minimum length zero?
+  cmp(limit1, (int)0); // use cast to resolve overloading ambiguity
+  br(Assembler::equal, true, Assembler::pn, Ldone);
+  // result is difference in lengths
+  if (ae == StrIntrinsicNode::UU) {
+    delayed()->sra(diff, 1, result);  // Divide by 2 to get number of chars
+  } else {
+    delayed()->mov(diff, result);
+  }
+
+  // Load first characters
+  if (ae == StrIntrinsicNode::LL) {
+    stride1 = stride2 = sizeof(jbyte);
+    ldub(str1, 0, chr1);
+    ldub(str2, 0, chr2);
+  } else if (ae == StrIntrinsicNode::UU) {
+    stride1 = stride2 = sizeof(jchar);
+    lduh(str1, 0, chr1);
+    lduh(str2, 0, chr2);
+  } else {
+    stride1 = sizeof(jbyte);
+    stride2 = sizeof(jchar);
+    ldub(str1, 0, chr1);
+    lduh(str2, 0, chr2);
+  }
+
+  // Compare first characters
+  subcc(chr1, chr2, chr1);
+  br(Assembler::notZero, false, Assembler::pt, Ldone);
+  assert(chr1 == result, "result must be pre-placed");
+  delayed()->nop();
+
+  // Check if the strings start at same location
+  cmp(str1, str2);
+  brx(Assembler::equal, true, Assembler::pn, Ldone);
+  delayed()->mov(G0, result);  // result is zero
+
+  // We have no guarantee that on 64 bit the higher half of limit is 0
+  signx(limit1);
+
+  // Get limit
+  if (ae == StrIntrinsicNode::LU || ae == StrIntrinsicNode::UL) {
+    sll(limit1, 1, limit2);
+    subcc(limit2, stride2, chr2);
+  }
+  subcc(limit1, stride1, chr1);
+  br(Assembler::zero, true, Assembler::pn, Ldone);
+  // result is difference in lengths
+  if (ae == StrIntrinsicNode::UU) {
+    delayed()->sra(diff, 1, result);  // Divide by 2 to get number of chars
+  } else {
+    delayed()->mov(diff, result);
+  }
+
+  // Shift str1 and str2 to the end of the arrays, negate limit
+  add(str1, limit1, str1);
+  add(str2, limit2, str2);
+  neg(chr1, limit1);  // limit1 = -(limit1-stride1)
+  if (ae == StrIntrinsicNode::LU || ae == StrIntrinsicNode::UL) {
+    neg(chr2, limit2);  // limit2 = -(limit2-stride2)
+  }
+
+  // Compare the rest of the characters
+  if (ae == StrIntrinsicNode::UU) {
+    lduh(str1, limit1, chr1);
+  } else {
+    ldub(str1, limit1, chr1);
+  }
+
+  bind(Lloop);
+  if (ae == StrIntrinsicNode::LL) {
+    ldub(str2, limit2, chr2);
+  } else {
+    lduh(str2, limit2, chr2);
+  }
+
+  subcc(chr1, chr2, chr1);
+  br(Assembler::notZero, false, Assembler::pt, Ldone);
+  assert(chr1 == result, "result must be pre-placed");
+  delayed()->inccc(limit1, stride1);
+  if (ae == StrIntrinsicNode::LU || ae == StrIntrinsicNode::UL) {
+    inccc(limit2, stride2);
+  }
+
+  // annul LDUB if branch is not taken to prevent access past end of string
+  br(Assembler::notZero, true, Assembler::pt, Lloop);
+  if (ae == StrIntrinsicNode::UU) {
+    delayed()->lduh(str1, limit2, chr1);
+  } else {
+    delayed()->ldub(str1, limit1, chr1);
+  }
+
+  // If strings are equal up to min length, return the length difference.
+  if (ae == StrIntrinsicNode::UU) {
+    // Divide by 2 to get number of chars
+    sra(diff, 1, result);
+  } else {
+    mov(diff, result);
+  }
+
+  // Otherwise, return the difference between the first mismatched chars.
+  bind(Ldone);
+  if(ae == StrIntrinsicNode::UL) {
+    // Negate result (see note above)
+    neg(result);
+  }
+}
+
+void MacroAssembler::array_equals(bool is_array_equ, Register ary1, Register ary2,
+                                  Register limit, Register tmp, Register result, bool is_byte) {
+  Label Ldone, Lvector, Lloop;
+  assert_different_registers(ary1, ary2, limit, tmp, result);
+
+  int length_offset  = arrayOopDesc::length_offset_in_bytes();
+  int base_offset    = arrayOopDesc::base_offset_in_bytes(is_byte ? T_BYTE : T_CHAR);
+
+  if (is_array_equ) {
+    // return true if the same array
+    cmp(ary1, ary2);
+    brx(Assembler::equal, true, Assembler::pn, Ldone);
+    delayed()->add(G0, 1, result); // equal
+
+    br_null(ary1, true, Assembler::pn, Ldone);
+    delayed()->mov(G0, result);    // not equal
+
+    br_null(ary2, true, Assembler::pn, Ldone);
+    delayed()->mov(G0, result);    // not equal
+
+    // load the lengths of arrays
+    ld(Address(ary1, length_offset), limit);
+    ld(Address(ary2, length_offset), tmp);
+
+    // return false if the two arrays are not equal length
+    cmp(limit, tmp);
+    br(Assembler::notEqual, true, Assembler::pn, Ldone);
+    delayed()->mov(G0, result);    // not equal
+  }
+
+  cmp_zero_and_br(Assembler::zero, limit, Ldone, true, Assembler::pn);
+  delayed()->add(G0, 1, result); // zero-length arrays are equal
+
+  if (is_array_equ) {
+    // load array addresses
+    add(ary1, base_offset, ary1);
+    add(ary2, base_offset, ary2);
+  } else {
+    // We have no guarantee that on 64 bit the higher half of limit is 0
+    signx(limit);
+  }
+
+  if (is_byte) {
+    Label Lskip;
+    // check for trailing byte
+    andcc(limit, 0x1, tmp);
+    br(Assembler::zero, false, Assembler::pt, Lskip);
+    delayed()->nop();
+
+    // compare the trailing byte
+    sub(limit, sizeof(jbyte), limit);
+    ldub(ary1, limit, result);
+    ldub(ary2, limit, tmp);
+    cmp(result, tmp);
+    br(Assembler::notEqual, true, Assembler::pt, Ldone);
+    delayed()->mov(G0, result);    // not equal
+
+    // only one byte?
+    cmp_zero_and_br(zero, limit, Ldone, true, Assembler::pn);
+    delayed()->add(G0, 1, result); // zero-length arrays are equal
+    bind(Lskip);
+  } else if (is_array_equ) {
+    // set byte count
+    sll(limit, exact_log2(sizeof(jchar)), limit);
+  }
+
+  // check for trailing character
+  andcc(limit, 0x2, tmp);
   br(Assembler::zero, false, Assembler::pt, Lvector);
   delayed()->nop();
 
   // compare the trailing char
   sub(limit, sizeof(jchar), limit);
-  lduh(ary1, limit, chr1);
-  lduh(ary2, limit, chr2);
-  cmp(chr1, chr2);
+  lduh(ary1, limit, result);
+  lduh(ary2, limit, tmp);
+  cmp(result, tmp);
   br(Assembler::notEqual, true, Assembler::pt, Ldone);
   delayed()->mov(G0, result);     // not equal
 
-  // only one char ?
+  // only one char?
   cmp_zero_and_br(zero, limit, Ldone, true, Assembler::pn);
   delayed()->add(G0, 1, result); // zero-length arrays are equal
 
@@ -4284,21 +4645,23 @@
   add(ary2, limit, ary2);
   neg(limit, limit);
 
-  lduw(ary1, limit, chr1);
+  lduw(ary1, limit, result);
   bind(Lloop);
-  lduw(ary2, limit, chr2);
-  cmp(chr1, chr2);
+  lduw(ary2, limit, tmp);
+  cmp(result, tmp);
   br(Assembler::notEqual, true, Assembler::pt, Ldone);
   delayed()->mov(G0, result);     // not equal
   inccc(limit, 2*sizeof(jchar));
   // annul LDUW if branch is not taken to prevent access past end of array
   br(Assembler::notZero, true, Assembler::pt, Lloop);
-  delayed()->lduw(ary1, limit, chr1); // hoisted
-
-  // Caller should set it:
-  // add(G0, 1, result); // equals
+  delayed()->lduw(ary1, limit, result); // hoisted
+
+  add(G0, 1, result); // equals
+  bind(Ldone);
 }
 
+#endif
+
 // Use BIS for zeroing (count is in bytes).
 void MacroAssembler::bis_zeroing(Register to, Register count, Register temp, Label& Ldone) {
   assert(UseBlockZeroing && VM_Version::has_block_zeroing(), "only works with BIS zeroing");
--- a/hotspot/src/cpu/sparc/vm/macroAssembler_sparc.hpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/cpu/sparc/vm/macroAssembler_sparc.hpp	Tue Nov 03 09:41:03 2015 +0100
@@ -1433,10 +1433,31 @@
   void inc_counter(address counter_addr, Register Rtmp1, Register Rtmp2);
   void inc_counter(int*    counter_addr, Register Rtmp1, Register Rtmp2);
 
-  // Compare char[] arrays aligned to 4 bytes.
-  void char_arrays_equals(Register ary1, Register ary2,
-                          Register limit, Register result,
-                          Register chr1, Register chr2, Label& Ldone);
+#ifdef COMPILER2
+  // Compress char[] to byte[] by compressing 16 bytes at once. Return 0 on failure.
+  void string_compress_16(Register src, Register dst, Register cnt, Register result,
+                          Register tmp1, Register tmp2, Register tmp3, Register tmp4,
+                          FloatRegister ftmp1, FloatRegister ftmp2, FloatRegister ftmp3, Label& Ldone);
+
+  // Compress char[] to byte[]. Return 0 on failure.
+  void string_compress(Register src, Register dst, Register cnt, Register tmp, Register result, Label& Ldone);
+
+  // Inflate byte[] to char[] by inflating 16 bytes at once.
+  void string_inflate_16(Register src, Register dst, Register cnt, Register tmp,
+                         FloatRegister ftmp1, FloatRegister ftmp2, FloatRegister ftmp3, FloatRegister ftmp4, Label& Ldone);
+
+  // Inflate byte[] to char[].
+  void string_inflate(Register src, Register dst, Register cnt, Register tmp, Label& Ldone);
+
+  void string_compare(Register str1, Register str2,
+                      Register cnt1, Register cnt2,
+                      Register tmp1, Register tmp2,
+                      Register result, int ae);
+
+  void array_equals(bool is_array_equ, Register ary1, Register ary2,
+                    Register limit, Register tmp, Register result, bool is_byte);
+#endif
+
   // Use BIS for zeroing
   void bis_zeroing(Register to, Register count, Register temp, Label& Ldone);
 
--- a/hotspot/src/cpu/sparc/vm/sparc.ad	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/cpu/sparc/vm/sparc.ad	Tue Nov 03 09:41:03 2015 +0100
@@ -2905,232 +2905,6 @@
     __ float_cmp( $primary, -1, Fsrc1, Fsrc2, Rdst);
   %}
 
-
-  enc_class enc_String_Compare(o0RegP str1, o1RegP str2, g3RegI cnt1, g4RegI cnt2, notemp_iRegI result) %{
-    Label Ldone, Lloop;
-    MacroAssembler _masm(&cbuf);
-
-    Register   str1_reg = reg_to_register_object($str1$$reg);
-    Register   str2_reg = reg_to_register_object($str2$$reg);
-    Register   cnt1_reg = reg_to_register_object($cnt1$$reg);
-    Register   cnt2_reg = reg_to_register_object($cnt2$$reg);
-    Register result_reg = reg_to_register_object($result$$reg);
-
-    assert(result_reg != str1_reg &&
-           result_reg != str2_reg &&
-           result_reg != cnt1_reg &&
-           result_reg != cnt2_reg ,
-           "need different registers");
-
-    // Compute the minimum of the string lengths(str1_reg) and the
-    // difference of the string lengths (stack)
-
-    // See if the lengths are different, and calculate min in str1_reg.
-    // Stash diff in O7 in case we need it for a tie-breaker.
-    Label Lskip;
-    __ subcc(cnt1_reg, cnt2_reg, O7);
-    __ sll(cnt1_reg, exact_log2(sizeof(jchar)), cnt1_reg); // scale the limit
-    __ br(Assembler::greater, true, Assembler::pt, Lskip);
-    // cnt2 is shorter, so use its count:
-    __ delayed()->sll(cnt2_reg, exact_log2(sizeof(jchar)), cnt1_reg); // scale the limit
-    __ bind(Lskip);
-
-    // reallocate cnt1_reg, cnt2_reg, result_reg
-    // Note:  limit_reg holds the string length pre-scaled by 2
-    Register limit_reg =   cnt1_reg;
-    Register  chr2_reg =   cnt2_reg;
-    Register  chr1_reg = result_reg;
-    // str{12} are the base pointers
-
-    // Is the minimum length zero?
-    __ cmp(limit_reg, (int)(0 * sizeof(jchar))); // use cast to resolve overloading ambiguity
-    __ br(Assembler::equal, true, Assembler::pn, Ldone);
-    __ delayed()->mov(O7, result_reg);  // result is difference in lengths
-
-    // Load first characters
-    __ lduh(str1_reg, 0, chr1_reg);
-    __ lduh(str2_reg, 0, chr2_reg);
-
-    // Compare first characters
-    __ subcc(chr1_reg, chr2_reg, chr1_reg);
-    __ br(Assembler::notZero, false, Assembler::pt,  Ldone);
-    assert(chr1_reg == result_reg, "result must be pre-placed");
-    __ delayed()->nop();
-
-    {
-      // Check after comparing first character to see if strings are equivalent
-      Label LSkip2;
-      // Check if the strings start at same location
-      __ cmp(str1_reg, str2_reg);
-      __ brx(Assembler::notEqual, true, Assembler::pt, LSkip2);
-      __ delayed()->nop();
-
-      // Check if the length difference is zero (in O7)
-      __ cmp(G0, O7);
-      __ br(Assembler::equal, true, Assembler::pn, Ldone);
-      __ delayed()->mov(G0, result_reg);  // result is zero
-
-      // Strings might not be equal
-      __ bind(LSkip2);
-    }
-
-    // We have no guarantee that on 64 bit the higher half of limit_reg is 0
-    __ signx(limit_reg);
-
-    __ subcc(limit_reg, 1 * sizeof(jchar), chr1_reg);
-    __ br(Assembler::equal, true, Assembler::pn, Ldone);
-    __ delayed()->mov(O7, result_reg);  // result is difference in lengths
-
-    // Shift str1_reg and str2_reg to the end of the arrays, negate limit
-    __ add(str1_reg, limit_reg, str1_reg);
-    __ add(str2_reg, limit_reg, str2_reg);
-    __ neg(chr1_reg, limit_reg);  // limit = -(limit-2)
-
-    // Compare the rest of the characters
-    __ lduh(str1_reg, limit_reg, chr1_reg);
-    __ bind(Lloop);
-    // __ lduh(str1_reg, limit_reg, chr1_reg); // hoisted
-    __ lduh(str2_reg, limit_reg, chr2_reg);
-    __ subcc(chr1_reg, chr2_reg, chr1_reg);
-    __ br(Assembler::notZero, false, Assembler::pt, Ldone);
-    assert(chr1_reg == result_reg, "result must be pre-placed");
-    __ delayed()->inccc(limit_reg, sizeof(jchar));
-    // annul LDUH if branch is not taken to prevent access past end of string
-    __ br(Assembler::notZero, true, Assembler::pt, Lloop);
-    __ delayed()->lduh(str1_reg, limit_reg, chr1_reg); // hoisted
-
-    // If strings are equal up to min length, return the length difference.
-    __ mov(O7, result_reg);
-
-    // Otherwise, return the difference between the first mismatched chars.
-    __ bind(Ldone);
-  %}
-
-enc_class enc_String_Equals(o0RegP str1, o1RegP str2, g3RegI cnt, notemp_iRegI result) %{
-    Label Lchar, Lchar_loop, Ldone;
-    MacroAssembler _masm(&cbuf);
-
-    Register   str1_reg = reg_to_register_object($str1$$reg);
-    Register   str2_reg = reg_to_register_object($str2$$reg);
-    Register    cnt_reg = reg_to_register_object($cnt$$reg);
-    Register   tmp1_reg = O7;
-    Register result_reg = reg_to_register_object($result$$reg);
-
-    assert(result_reg != str1_reg &&
-           result_reg != str2_reg &&
-           result_reg !=  cnt_reg &&
-           result_reg != tmp1_reg ,
-           "need different registers");
-
-    __ cmp(str1_reg, str2_reg); //same char[] ?
-    __ brx(Assembler::equal, true, Assembler::pn, Ldone);
-    __ delayed()->add(G0, 1, result_reg);
-
-    __ cmp_zero_and_br(Assembler::zero, cnt_reg, Ldone, true, Assembler::pn);
-    __ delayed()->add(G0, 1, result_reg); // count == 0
-
-    //rename registers
-    Register limit_reg =    cnt_reg;
-    Register  chr1_reg = result_reg;
-    Register  chr2_reg =   tmp1_reg;
-
-    // We have no guarantee that on 64 bit the higher half of limit_reg is 0
-    __ signx(limit_reg);
-
-    //check for alignment and position the pointers to the ends
-    __ or3(str1_reg, str2_reg, chr1_reg);
-    __ andcc(chr1_reg, 0x3, chr1_reg);
-    // notZero means at least one not 4-byte aligned.
-    // We could optimize the case when both arrays are not aligned
-    // but it is not frequent case and it requires additional checks.
-    __ br(Assembler::notZero, false, Assembler::pn, Lchar); // char by char compare
-    __ delayed()->sll(limit_reg, exact_log2(sizeof(jchar)), limit_reg); // set byte count
-
-    // Compare char[] arrays aligned to 4 bytes.
-    __ char_arrays_equals(str1_reg, str2_reg, limit_reg, result_reg,
-                          chr1_reg, chr2_reg, Ldone);
-    __ ba(Ldone);
-    __ delayed()->add(G0, 1, result_reg);
-
-    // char by char compare
-    __ bind(Lchar);
-    __ add(str1_reg, limit_reg, str1_reg);
-    __ add(str2_reg, limit_reg, str2_reg);
-    __ neg(limit_reg); //negate count
-
-    __ lduh(str1_reg, limit_reg, chr1_reg);
-    // Lchar_loop
-    __ bind(Lchar_loop);
-    __ lduh(str2_reg, limit_reg, chr2_reg);
-    __ cmp(chr1_reg, chr2_reg);
-    __ br(Assembler::notEqual, true, Assembler::pt, Ldone);
-    __ delayed()->mov(G0, result_reg); //not equal
-    __ inccc(limit_reg, sizeof(jchar));
-    // annul LDUH if branch is not taken to prevent access past end of string
-    __ br(Assembler::notZero, true, Assembler::pt, Lchar_loop);
-    __ delayed()->lduh(str1_reg, limit_reg, chr1_reg); // hoisted
-
-    __ add(G0, 1, result_reg);  //equal
-
-    __ bind(Ldone);
-  %}
-
-enc_class enc_Array_Equals(o0RegP ary1, o1RegP ary2, g3RegP tmp1, notemp_iRegI result) %{
-    Label Lvector, Ldone, Lloop;
-    MacroAssembler _masm(&cbuf);
-
-    Register   ary1_reg = reg_to_register_object($ary1$$reg);
-    Register   ary2_reg = reg_to_register_object($ary2$$reg);
-    Register   tmp1_reg = reg_to_register_object($tmp1$$reg);
-    Register   tmp2_reg = O7;
-    Register result_reg = reg_to_register_object($result$$reg);
-
-    int length_offset  = arrayOopDesc::length_offset_in_bytes();
-    int base_offset    = arrayOopDesc::base_offset_in_bytes(T_CHAR);
-
-    // return true if the same array
-    __ cmp(ary1_reg, ary2_reg);
-    __ brx(Assembler::equal, true, Assembler::pn, Ldone);
-    __ delayed()->add(G0, 1, result_reg); // equal
-
-    __ br_null(ary1_reg, true, Assembler::pn, Ldone);
-    __ delayed()->mov(G0, result_reg);    // not equal
-
-    __ br_null(ary2_reg, true, Assembler::pn, Ldone);
-    __ delayed()->mov(G0, result_reg);    // not equal
-
-    //load the lengths of arrays
-    __ ld(Address(ary1_reg, length_offset), tmp1_reg);
-    __ ld(Address(ary2_reg, length_offset), tmp2_reg);
-
-    // return false if the two arrays are not equal length
-    __ cmp(tmp1_reg, tmp2_reg);
-    __ br(Assembler::notEqual, true, Assembler::pn, Ldone);
-    __ delayed()->mov(G0, result_reg);     // not equal
-
-    __ cmp_zero_and_br(Assembler::zero, tmp1_reg, Ldone, true, Assembler::pn);
-    __ delayed()->add(G0, 1, result_reg); // zero-length arrays are equal
-
-    // load array addresses
-    __ add(ary1_reg, base_offset, ary1_reg);
-    __ add(ary2_reg, base_offset, ary2_reg);
-
-    // renaming registers
-    Register chr1_reg  =  result_reg; // for characters in ary1
-    Register chr2_reg  =  tmp2_reg;   // for characters in ary2
-    Register limit_reg =  tmp1_reg;   // length
-
-    // set byte count
-    __ sll(limit_reg, exact_log2(sizeof(jchar)), limit_reg);
-
-    // Compare char[] arrays aligned to 4 bytes.
-    __ char_arrays_equals(ary1_reg, ary2_reg, limit_reg, result_reg,
-                          chr1_reg, chr2_reg, Ldone);
-    __ add(G0, 1, result_reg); // equals
-
-    __ bind(Ldone);
-  %}
-
   enc_class enc_rethrow() %{
     cbuf.set_insts_mark();
     Register temp_reg = G3;
@@ -10275,33 +10049,204 @@
   ins_pipe(long_memory_op);
 %}
 
-instruct string_compare(o0RegP str1, o1RegP str2, g3RegI cnt1, g4RegI cnt2, notemp_iRegI result,
-                        o7RegI tmp, flagsReg ccr) %{
+instruct string_compareL(o0RegP str1, o1RegP str2, g3RegI cnt1, g4RegI cnt2, notemp_iRegI result,
+                         o7RegI tmp, flagsReg ccr) %{
+  predicate(((StrCompNode*)n)->encoding() == StrIntrinsicNode::LL);
+  match(Set result (StrComp (Binary str1 cnt1) (Binary str2 cnt2)));
+  effect(USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL ccr, KILL tmp);
+  ins_cost(300);
+  format %{ "String Compare byte[] $str1,$cnt1,$str2,$cnt2 -> $result   // KILL $tmp" %}
+  ins_encode %{
+    __ string_compare($str1$$Register, $str2$$Register,
+                      $cnt1$$Register, $cnt2$$Register, 
+                      $tmp$$Register, $tmp$$Register,
+                      $result$$Register, StrIntrinsicNode::LL);
+  %}                    
+  ins_pipe(long_memory_op);
+%}
+
+instruct string_compareU(o0RegP str1, o1RegP str2, g3RegI cnt1, g4RegI cnt2, notemp_iRegI result,
+                         o7RegI tmp, flagsReg ccr) %{
+  predicate(((StrCompNode*)n)->encoding() == StrIntrinsicNode::UU);
   match(Set result (StrComp (Binary str1 cnt1) (Binary str2 cnt2)));
   effect(USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL ccr, KILL tmp);
   ins_cost(300);
-  format %{ "String Compare $str1,$cnt1,$str2,$cnt2 -> $result   // KILL $tmp" %}
-  ins_encode( enc_String_Compare(str1, str2, cnt1, cnt2, result) );
+  format %{ "String Compare char[] $str1,$cnt1,$str2,$cnt2 -> $result   // KILL $tmp" %}
+  ins_encode %{
+    __ string_compare($str1$$Register, $str2$$Register,
+                      $cnt1$$Register, $cnt2$$Register,
+                      $tmp$$Register, $tmp$$Register,
+                      $result$$Register, StrIntrinsicNode::UU);
+  %}                    
   ins_pipe(long_memory_op);
 %}
 
-instruct string_equals(o0RegP str1, o1RegP str2, g3RegI cnt, notemp_iRegI result,
-                       o7RegI tmp, flagsReg ccr) %{
+instruct string_compareLU(o0RegP str1, o1RegP str2, g3RegI cnt1, g4RegI cnt2, notemp_iRegI result,
+                          o7RegI tmp1, g1RegI tmp2, flagsReg ccr) %{
+  predicate(((StrCompNode*)n)->encoding() == StrIntrinsicNode::LU);
+  match(Set result (StrComp (Binary str1 cnt1) (Binary str2 cnt2)));
+  effect(USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL ccr, KILL tmp1, KILL tmp2);
+  ins_cost(300);
+  format %{ "String Compare byte[] $str1,$cnt1,$str2,$cnt2 -> $result   // KILL $tmp1,$tmp2" %}
+  ins_encode %{
+    __ string_compare($str1$$Register, $str2$$Register,
+                      $cnt1$$Register, $cnt2$$Register,
+                      $tmp1$$Register, $tmp2$$Register,
+                      $result$$Register, StrIntrinsicNode::LU);
+  %}                    
+  ins_pipe(long_memory_op);
+%}
+
+instruct string_compareUL(o0RegP str1, o1RegP str2, g3RegI cnt1, g4RegI cnt2, notemp_iRegI result,
+                          o7RegI tmp1, g1RegI tmp2, flagsReg ccr) %{
+  predicate(((StrCompNode*)n)->encoding() == StrIntrinsicNode::UL);
+  match(Set result (StrComp (Binary str1 cnt1) (Binary str2 cnt2)));
+  effect(USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL ccr, KILL tmp1, KILL tmp2);
+  ins_cost(300);
+  format %{ "String Compare byte[] $str1,$cnt1,$str2,$cnt2 -> $result   // KILL $tmp1,$tmp2" %}
+  ins_encode %{
+    __ string_compare($str2$$Register, $str1$$Register,
+                      $cnt2$$Register, $cnt1$$Register, 
+                      $tmp1$$Register, $tmp2$$Register,
+                      $result$$Register, StrIntrinsicNode::UL);
+  %}                    
+  ins_pipe(long_memory_op);
+%}
+
+instruct string_equalsL(o0RegP str1, o1RegP str2, g3RegI cnt, notemp_iRegI result,
+                        o7RegI tmp, flagsReg ccr) %{
+  predicate(((StrEqualsNode*)n)->encoding() == StrIntrinsicNode::LL);
   match(Set result (StrEquals (Binary str1 str2) cnt));
   effect(USE_KILL str1, USE_KILL str2, USE_KILL cnt, KILL tmp, KILL ccr);
   ins_cost(300);
-  format %{ "String Equals $str1,$str2,$cnt -> $result   // KILL $tmp" %}
-  ins_encode( enc_String_Equals(str1, str2, cnt, result) );
+  format %{ "String Equals byte[] $str1,$str2,$cnt -> $result   // KILL $tmp" %}
+  ins_encode %{
+    __ array_equals(false, $str1$$Register, $str2$$Register,
+                    $cnt$$Register, $tmp$$Register,
+                    $result$$Register, true /* byte */);
+  %}
   ins_pipe(long_memory_op);
 %}
 
-instruct array_equals(o0RegP ary1, o1RegP ary2, g3RegI tmp1, notemp_iRegI result,
-                      o7RegI tmp2, flagsReg ccr) %{
+instruct string_equalsU(o0RegP str1, o1RegP str2, g3RegI cnt, notemp_iRegI result,
+                        o7RegI tmp, flagsReg ccr) %{
+  predicate(((StrEqualsNode*)n)->encoding() == StrIntrinsicNode::UU);
+  match(Set result (StrEquals (Binary str1 str2) cnt));
+  effect(USE_KILL str1, USE_KILL str2, USE_KILL cnt, KILL tmp, KILL ccr);
+  ins_cost(300);
+  format %{ "String Equals char[]  $str1,$str2,$cnt -> $result   // KILL $tmp" %}
+  ins_encode %{
+    __ array_equals(false, $str1$$Register, $str2$$Register,
+                    $cnt$$Register, $tmp$$Register,
+                    $result$$Register, false /* byte */);
+  %}
+  ins_pipe(long_memory_op);
+%}
+
+instruct array_equalsB(o0RegP ary1, o1RegP ary2, g3RegI tmp1, notemp_iRegI result,
+                       o7RegI tmp2, flagsReg ccr) %{
+  predicate(((AryEqNode*)n)->encoding() == StrIntrinsicNode::LL);
   match(Set result (AryEq ary1 ary2));
   effect(USE_KILL ary1, USE_KILL ary2, KILL tmp1, KILL tmp2, KILL ccr);
   ins_cost(300);
   format %{ "Array Equals $ary1,$ary2 -> $result   // KILL $tmp1,$tmp2" %}
-  ins_encode( enc_Array_Equals(ary1, ary2, tmp1, result));
+  ins_encode %{
+    __ array_equals(true, $ary1$$Register, $ary2$$Register,
+                    $tmp1$$Register, $tmp2$$Register,
+                    $result$$Register, true /* byte */);
+  %}
+  ins_pipe(long_memory_op);
+%}
+
+instruct array_equalsC(o0RegP ary1, o1RegP ary2, g3RegI tmp1, notemp_iRegI result,
+                       o7RegI tmp2, flagsReg ccr) %{
+  predicate(((AryEqNode*)n)->encoding() == StrIntrinsicNode::UU);
+  match(Set result (AryEq ary1 ary2));
+  effect(USE_KILL ary1, USE_KILL ary2, KILL tmp1, KILL tmp2, KILL ccr);
+  ins_cost(300);
+  format %{ "Array Equals $ary1,$ary2 -> $result   // KILL $tmp1,$tmp2" %}
+  ins_encode %{
+    __ array_equals(true, $ary1$$Register, $ary2$$Register,
+                    $tmp1$$Register, $tmp2$$Register,
+                    $result$$Register, false /* byte */);
+  %}
+  ins_pipe(long_memory_op);
+%}
+
+// char[] to byte[] compression
+instruct string_compress(o0RegP src, o1RegP dst, g3RegI len, notemp_iRegI result, iRegL tmp, flagsReg ccr) %{
+  predicate(UseVIS < 3);
+  match(Set result (StrCompressedCopy src (Binary dst len)));
+  effect(TEMP result, TEMP tmp, USE_KILL src, USE_KILL dst, USE_KILL len, KILL ccr);
+  ins_cost(300);
+  format %{ "String Compress $src,$dst,$len -> $result    // KILL $tmp" %}
+  ins_encode %{
+    Label Ldone;
+    __ signx($len$$Register);
+    __ cmp_zero_and_br(Assembler::zero, $len$$Register, Ldone, false, Assembler::pn);
+    __ delayed()->mov($len$$Register, $result$$Register); // copy count
+    __ string_compress($src$$Register, $dst$$Register, $len$$Register, $result$$Register, $tmp$$Register, Ldone);
+    __ bind(Ldone);
+  %}
+  ins_pipe(long_memory_op);
+%}
+
+// fast char[] to byte[] compression using VIS instructions
+instruct string_compress_fast(o0RegP src, o1RegP dst, g3RegI len, notemp_iRegI result,
+                              iRegL tmp1, iRegL tmp2, iRegL tmp3, iRegL tmp4,
+                              regD ftmp1, regD ftmp2, regD ftmp3, flagsReg ccr) %{
+  predicate(UseVIS >= 3);
+  match(Set result (StrCompressedCopy src (Binary dst len)));
+  effect(TEMP result, TEMP tmp1, TEMP tmp2, TEMP tmp3, TEMP tmp4, TEMP ftmp1, TEMP ftmp2, TEMP ftmp3, USE_KILL src, USE_KILL dst, USE_KILL len, KILL ccr);
+  ins_cost(300);
+  format %{ "String Compress Fast $src,$dst,$len -> $result    // KILL $tmp1,$tmp2,$tmp3,$tmp4,$ftmp1,$ftmp2,$ftmp3" %}
+  ins_encode %{
+    Label Ldone;
+    __ signx($len$$Register);
+    __ string_compress_16($src$$Register, $dst$$Register, $len$$Register, $result$$Register,
+                          $tmp1$$Register, $tmp2$$Register, $tmp3$$Register, $tmp4$$Register,
+                          $ftmp1$$FloatRegister, $ftmp2$$FloatRegister, $ftmp3$$FloatRegister, Ldone);
+    __ cmp_and_brx_short($len$$Register, 0, Assembler::equal, Assembler::pn, Ldone);
+    __ string_compress($src$$Register, $dst$$Register, $len$$Register, $result$$Register, $tmp1$$Register, Ldone);
+    __ bind(Ldone);
+  %}
+  ins_pipe(long_memory_op);
+%}
+
+// byte[] to char[] inflation
+instruct string_inflate(Universe dummy, o0RegP src, o1RegP dst, g3RegI len,
+                        iRegL tmp, flagsReg ccr) %{
+  match(Set dummy (StrInflatedCopy src (Binary dst len)));
+  effect(TEMP tmp, USE_KILL src, USE_KILL dst, USE_KILL len, KILL ccr);
+  ins_cost(300);
+  format %{ "String Inflate $src,$dst,$len    // KILL $tmp" %}
+  ins_encode %{
+    Label Ldone;
+    __ signx($len$$Register);
+    __ cmp_and_brx_short($len$$Register, 0, Assembler::equal, Assembler::pn, Ldone);
+    __ string_inflate($src$$Register, $dst$$Register, $len$$Register, $tmp$$Register, Ldone);
+    __ bind(Ldone);
+  %}
+  ins_pipe(long_memory_op);
+%}
+
+// fast byte[] to char[] inflation using VIS instructions
+instruct string_inflate_fast(Universe dummy, o0RegP src, o1RegP dst, g3RegI len,
+                             iRegL tmp, regD ftmp1, regD ftmp2, regD ftmp3, regD ftmp4, flagsReg ccr) %{
+  predicate(UseVIS >= 3);
+  match(Set dummy (StrInflatedCopy src (Binary dst len)));
+  effect(TEMP tmp, TEMP ftmp1, TEMP ftmp2, TEMP ftmp3, TEMP ftmp4, USE_KILL src, USE_KILL dst, USE_KILL len, KILL ccr);
+  ins_cost(300);
+  format %{ "String Inflate Fast $src,$dst,$len    // KILL $tmp,$ftmp1,$ftmp2,$ftmp3,$ftmp4" %}
+  ins_encode %{
+    Label Ldone;
+    __ signx($len$$Register);
+    __ string_inflate_16($src$$Register, $dst$$Register, $len$$Register, $tmp$$Register,
+                         $ftmp1$$FloatRegister, $ftmp2$$FloatRegister, $ftmp3$$FloatRegister, $ftmp4$$FloatRegister, Ldone);
+    __ cmp_and_brx_short($len$$Register, 0, Assembler::equal, Assembler::pn, Ldone);
+    __ string_inflate($src$$Register, $dst$$Register, $len$$Register, $tmp$$Register, Ldone);
+    __ bind(Ldone);
+  %}
   ins_pipe(long_memory_op);
 %}
 
--- a/hotspot/src/cpu/x86/vm/assembler_x86.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/cpu/x86/vm/assembler_x86.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -3036,6 +3036,35 @@
   emit_int8(imm8);
 }
 
+void Assembler::pcmpeqw(XMMRegister dst, XMMRegister src) {
+  NOT_LP64(assert(VM_Version::supports_sse2(), ""));
+  emit_simd_arith(0x75, dst, src, VEX_SIMD_66,
+                  false, (VM_Version::supports_avx512dq() == false));
+}
+
+void Assembler::vpcmpeqw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len) {
+  assert(UseAVX > 0, "some form of AVX must be enabled");
+  emit_vex_arith(0x75, dst, nds, src, VEX_SIMD_66, vector_len,
+                 false, (VM_Version::supports_avx512dq() == false));
+}
+
+void Assembler::pmovmskb(Register dst, XMMRegister src) {
+  assert(VM_Version::supports_sse2(), "");
+  int encode = simd_prefix_and_encode(as_XMMRegister(dst->encoding()), xnoreg, src, VEX_SIMD_66, true, VEX_OPCODE_0F,
+                                      false, AVX_128bit, (VM_Version::supports_avx512dq() == false));
+  emit_int8((unsigned char)0xD7);
+  emit_int8((unsigned char)(0xC0 | encode));
+}
+
+void Assembler::vpmovmskb(Register dst, XMMRegister src) {
+  assert(VM_Version::supports_avx2(), "");
+  int vector_len = AVX_256bit;
+  int encode = vex_prefix_and_encode(as_XMMRegister(dst->encoding()), xnoreg, src, VEX_SIMD_66,
+                                     vector_len, VEX_OPCODE_0F, true, false);
+  emit_int8((unsigned char)0xD7);
+  emit_int8((unsigned char)(0xC0 | encode));
+}
+
 void Assembler::pextrd(Register dst, XMMRegister src, int imm8) {
   assert(VM_Version::supports_sse4_1(), "");
   int encode = simd_prefix_and_encode(as_XMMRegister(dst->encoding()), xnoreg, src, VEX_SIMD_66, /* no_mask_reg */ true,
@@ -3108,6 +3137,17 @@
   emit_int8((unsigned char)(0xC0 | encode));
 }
 
+void Assembler::vpmovzxbw(XMMRegister dst, Address src) {
+  assert(VM_Version::supports_avx(), "");
+  InstructionMark im(this);
+  bool vector256 = true;
+  assert(dst != xnoreg, "sanity");
+  int dst_enc = dst->encoding();
+  vex_prefix(src, 0, dst_enc, VEX_SIMD_66, VEX_OPCODE_0F_38, false, vector256);
+  emit_int8(0x30);
+  emit_operand(dst, src);
+}
+
 // generic
 void Assembler::pop(Register dst) {
   int encode = prefix_and_encode(dst->encoding());
@@ -5370,6 +5410,16 @@
   emit_int8((unsigned char)(0xC0 | encode));
 }
 
+// duplicate 2-bytes integer data from src into 16 locations in dest
+void Assembler::vpbroadcastw(XMMRegister dst, XMMRegister src) {
+  assert(VM_Version::supports_avx2(), "");
+  bool vector_len = AVX_256bit;
+  int encode = vex_prefix_and_encode(dst, xnoreg, src, VEX_SIMD_66,
+                                     vector_len, VEX_OPCODE_0F_38, false);
+  emit_int8(0x79);
+  emit_int8((unsigned char)(0xC0 | encode));
+}
+
 // duplicate 1-byte integer data from src into 16||32|64 locations in dest : requires AVX512BW and AVX512VL
 void Assembler::evpbroadcastb(XMMRegister dst, XMMRegister src, int vector_len) {
   _instruction_uses_vl = true;
--- a/hotspot/src/cpu/x86/vm/assembler_x86.hpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/cpu/x86/vm/assembler_x86.hpp	Tue Nov 03 09:41:03 2015 +0100
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1997, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1997, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -1682,6 +1682,12 @@
   void pcmpestri(XMMRegister xmm1, XMMRegister xmm2, int imm8);
   void pcmpestri(XMMRegister xmm1, Address src, int imm8);
 
+  void pcmpeqw(XMMRegister dst, XMMRegister src);
+  void vpcmpeqw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
+
+  void pmovmskb(Register dst, XMMRegister src);
+  void vpmovmskb(Register dst, XMMRegister src);
+
   // SSE 4.1 extract
   void pextrd(Register dst, XMMRegister src, int imm8);
   void pextrq(Register dst, XMMRegister src, int imm8);
@@ -1698,6 +1704,8 @@
   void pmovzxbw(XMMRegister dst, XMMRegister src);
   void pmovzxbw(XMMRegister dst, Address src);
 
+  void vpmovzxbw(XMMRegister dst, Address src);
+
 #ifndef _LP64 // no 32bit push/pop on amd64
   void popl(Address dst);
 #endif
@@ -2116,6 +2124,9 @@
   // duplicate 4-bytes integer data from src into 8 locations in dest
   void vpbroadcastd(XMMRegister dst, XMMRegister src);
 
+  // duplicate 2-bytes integer data from src into 16 locations in dest
+  void vpbroadcastw(XMMRegister dst, XMMRegister src);
+
   // duplicate n-bytes integer data from src into vector_len locations in dest
   void evpbroadcastb(XMMRegister dst, XMMRegister src, int vector_len);
   void evpbroadcastb(XMMRegister dst, Address src, int vector_len);
--- a/hotspot/src/cpu/x86/vm/c1_LIRAssembler_x86.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/cpu/x86/vm/c1_LIRAssembler_x86.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -509,86 +509,6 @@
 }
 
 
-// This is the fast version of java.lang.String.compare; it has not
-// OSR-entry and therefore, we generate a slow version for OSR's
-void LIR_Assembler::emit_string_compare(LIR_Opr arg0, LIR_Opr arg1, LIR_Opr dst, CodeEmitInfo* info) {
-  __ movptr (rbx, rcx); // receiver is in rcx
-  __ movptr (rax, arg1->as_register());
-
-  // Get addresses of first characters from both Strings
-  __ load_heap_oop(rsi, Address(rax, java_lang_String::value_offset_in_bytes()));
-  if (java_lang_String::has_offset_field()) {
-    __ movptr     (rcx, Address(rax, java_lang_String::offset_offset_in_bytes()));
-    __ movl       (rax, Address(rax, java_lang_String::count_offset_in_bytes()));
-    __ lea        (rsi, Address(rsi, rcx, Address::times_2, arrayOopDesc::base_offset_in_bytes(T_CHAR)));
-  } else {
-    __ movl       (rax, Address(rsi, arrayOopDesc::length_offset_in_bytes()));
-    __ lea        (rsi, Address(rsi, arrayOopDesc::base_offset_in_bytes(T_CHAR)));
-  }
-
-  // rbx, may be NULL
-  add_debug_info_for_null_check_here(info);
-  __ load_heap_oop(rdi, Address(rbx, java_lang_String::value_offset_in_bytes()));
-  if (java_lang_String::has_offset_field()) {
-    __ movptr     (rcx, Address(rbx, java_lang_String::offset_offset_in_bytes()));
-    __ movl       (rbx, Address(rbx, java_lang_String::count_offset_in_bytes()));
-    __ lea        (rdi, Address(rdi, rcx, Address::times_2, arrayOopDesc::base_offset_in_bytes(T_CHAR)));
-  } else {
-    __ movl       (rbx, Address(rdi, arrayOopDesc::length_offset_in_bytes()));
-    __ lea        (rdi, Address(rdi, arrayOopDesc::base_offset_in_bytes(T_CHAR)));
-  }
-
-  // compute minimum length (in rax) and difference of lengths (on top of stack)
-  __ mov   (rcx, rbx);
-  __ subptr(rbx, rax); // subtract lengths
-  __ push  (rbx);      // result
-  __ cmov  (Assembler::lessEqual, rax, rcx);
-
-  // is minimum length 0?
-  Label noLoop, haveResult;
-  __ testptr (rax, rax);
-  __ jcc (Assembler::zero, noLoop);
-
-  // compare first characters
-  __ load_unsigned_short(rcx, Address(rdi, 0));
-  __ load_unsigned_short(rbx, Address(rsi, 0));
-  __ subl(rcx, rbx);
-  __ jcc(Assembler::notZero, haveResult);
-  // starting loop
-  __ decrement(rax); // we already tested index: skip one
-  __ jcc(Assembler::zero, noLoop);
-
-  // set rsi.edi to the end of the arrays (arrays have same length)
-  // negate the index
-
-  __ lea(rsi, Address(rsi, rax, Address::times_2, type2aelembytes(T_CHAR)));
-  __ lea(rdi, Address(rdi, rax, Address::times_2, type2aelembytes(T_CHAR)));
-  __ negptr(rax);
-
-  // compare the strings in a loop
-
-  Label loop;
-  __ align(wordSize);
-  __ bind(loop);
-  __ load_unsigned_short(rcx, Address(rdi, rax, Address::times_2, 0));
-  __ load_unsigned_short(rbx, Address(rsi, rax, Address::times_2, 0));
-  __ subl(rcx, rbx);
-  __ jcc(Assembler::notZero, haveResult);
-  __ increment(rax);
-  __ jcc(Assembler::notZero, loop);
-
-  // strings are equal up to min length
-
-  __ bind(noLoop);
-  __ pop(rax);
-  return_op(LIR_OprFact::illegalOpr);
-
-  __ bind(haveResult);
-  // leave instruction is going to discard the TOS value
-  __ mov (rax, rcx); // result of call is in rax,
-}
-
-
 void LIR_Assembler::return_op(LIR_Opr result) {
   assert(result->is_illegal() || !result->is_single_cpu() || result->as_register() == rax, "word returns are in rax,");
   if (!result->is_illegal() && result->is_float_kind() && !result->is_xmm_register()) {
--- a/hotspot/src/cpu/x86/vm/globals_x86.hpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/cpu/x86/vm/globals_x86.hpp	Tue Nov 03 09:41:03 2015 +0100
@@ -91,6 +91,8 @@
 
 define_pd_global(uintx, TypeProfileLevel, 111);
 
+define_pd_global(bool, CompactStrings, true);
+
 define_pd_global(bool, PreserveFramePointer, false);
 
 #define ARCH_FLAGS(develop, product, diagnostic, experimental, notproduct, range, constraint) \
--- a/hotspot/src/cpu/x86/vm/macroAssembler_x86.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/cpu/x86/vm/macroAssembler_x86.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -46,6 +46,9 @@
 #include "gc/g1/heapRegion.hpp"
 #endif // INCLUDE_ALL_GCS
 #include "crc32c.h"
+#ifdef COMPILER2
+#include "opto/intrinsicnode.hpp"
+#endif
 
 #ifdef PRODUCT
 #define BLOCK_COMMENT(str) /* nothing */
@@ -6299,25 +6302,34 @@
   }
 }
 
+#ifdef COMPILER2
+
 // IndexOf for constant substrings with size >= 8 chars
 // which don't need to be loaded through stack.
 void MacroAssembler::string_indexofC8(Register str1, Register str2,
                                       Register cnt1, Register cnt2,
                                       int int_cnt2,  Register result,
-                                      XMMRegister vec, Register tmp) {
+                                      XMMRegister vec, Register tmp,
+                                      int ae) {
   ShortBranchVerifier sbv(this);
   assert(UseSSE42Intrinsics, "SSE4.2 is required");
-
-  // This method uses pcmpestri instruction with bound registers
+  assert(ae != StrIntrinsicNode::LU, "Invalid encoding");
+
+  // This method uses the pcmpestri instruction with bound registers
   //   inputs:
   //     xmm - substring
   //     rax - substring length (elements count)
   //     mem - scanned string
   //     rdx - string length (elements count)
   //     0xd - mode: 1100 (substring search) + 01 (unsigned shorts)
+  //     0xc - mode: 1100 (substring search) + 00 (unsigned bytes)
   //   outputs:
   //     rcx - matched index in string
   assert(cnt1 == rdx && cnt2 == rax && tmp == rcx, "pcmpestri");
+  int mode   = (ae == StrIntrinsicNode::LL) ? 0x0c : 0x0d; // bytes or shorts
+  int stride = (ae == StrIntrinsicNode::LL) ? 16 : 8; //UU, UL -> 8
+  Address::ScaleFactor scale1 = (ae == StrIntrinsicNode::LL) ? Address::times_1 : Address::times_2;
+  Address::ScaleFactor scale2 = (ae == StrIntrinsicNode::UL) ? Address::times_1 : scale1;
 
   Label RELOAD_SUBSTR, SCAN_TO_SUBSTR, SCAN_SUBSTR,
         RET_FOUND, RET_NOT_FOUND, EXIT, FOUND_SUBSTR,
@@ -6326,20 +6338,28 @@
   // Note, inline_string_indexOf() generates checks:
   // if (substr.count > string.count) return -1;
   // if (substr.count == 0) return 0;
-  assert(int_cnt2 >= 8, "this code isused only for cnt2 >= 8 chars");
+  assert(int_cnt2 >= stride, "this code is used only for cnt2 >= 8 chars");
 
   // Load substring.
-  movdqu(vec, Address(str2, 0));
+  if (ae == StrIntrinsicNode::UL) {
+    pmovzxbw(vec, Address(str2, 0));
+  } else {
+    movdqu(vec, Address(str2, 0));
+  }
   movl(cnt2, int_cnt2);
   movptr(result, str1); // string addr
 
-  if (int_cnt2 > 8) {
+  if (int_cnt2 > stride) {
     jmpb(SCAN_TO_SUBSTR);
 
     // Reload substr for rescan, this code
     // is executed only for large substrings (> 8 chars)
     bind(RELOAD_SUBSTR);
-    movdqu(vec, Address(str2, 0));
+    if (ae == StrIntrinsicNode::UL) {
+      pmovzxbw(vec, Address(str2, 0));
+    } else {
+      movdqu(vec, Address(str2, 0));
+    }
     negptr(cnt2); // Jumped here with negative cnt2, convert to positive
 
     bind(RELOAD_STR);
@@ -6358,15 +6378,15 @@
     cmpl(cnt1, cnt2);
     jccb(Assembler::negative, RET_NOT_FOUND);  // Left less then substring
 
-    addptr(result, 2);
+    addptr(result, (1<<scale1));
 
   } // (int_cnt2 > 8)
 
   // Scan string for start of substr in 16-byte vectors
   bind(SCAN_TO_SUBSTR);
-  pcmpestri(vec, Address(result, 0), 0x0d);
+  pcmpestri(vec, Address(result, 0), mode);
   jccb(Assembler::below, FOUND_CANDIDATE);   // CF == 1
-  subl(cnt1, 8);
+  subl(cnt1, stride);
   jccb(Assembler::lessEqual, RET_NOT_FOUND); // Scanned full string
   cmpl(cnt1, cnt2);
   jccb(Assembler::negative, RET_NOT_FOUND);  // Left less then substring
@@ -6376,19 +6396,19 @@
   // Found a potential substr
   bind(FOUND_CANDIDATE);
   // Matched whole vector if first element matched (tmp(rcx) == 0).
-  if (int_cnt2 == 8) {
+  if (int_cnt2 == stride) {
     jccb(Assembler::overflow, RET_FOUND);    // OF == 1
   } else { // int_cnt2 > 8
     jccb(Assembler::overflow, FOUND_SUBSTR);
   }
   // After pcmpestri tmp(rcx) contains matched element index
   // Compute start addr of substr
-  lea(result, Address(result, tmp, Address::times_2));
+  lea(result, Address(result, tmp, scale1));
 
   // Make sure string is still long enough
   subl(cnt1, tmp);
   cmpl(cnt1, cnt2);
-  if (int_cnt2 == 8) {
+  if (int_cnt2 == stride) {
     jccb(Assembler::greaterEqual, SCAN_TO_SUBSTR);
   } else { // int_cnt2 > 8
     jccb(Assembler::greaterEqual, MATCH_SUBSTR_HEAD);
@@ -6399,11 +6419,11 @@
   movl(result, -1);
   jmpb(EXIT);
 
-  if (int_cnt2 > 8) {
+  if (int_cnt2 > stride) {
     // This code is optimized for the case when whole substring
     // is matched if its head is matched.
     bind(MATCH_SUBSTR_HEAD);
-    pcmpestri(vec, Address(result, 0), 0x0d);
+    pcmpestri(vec, Address(result, 0), mode);
     // Reload only string if does not match
     jccb(Assembler::noOverflow, RELOAD_STR); // OF == 0
 
@@ -6412,31 +6432,41 @@
     bind(FOUND_SUBSTR);
     // First 8 chars are already matched.
     negptr(cnt2);
-    addptr(cnt2, 8);
+    addptr(cnt2, stride);
 
     bind(SCAN_SUBSTR);
-    subl(cnt1, 8);
-    cmpl(cnt2, -8); // Do not read beyond substring
+    subl(cnt1, stride);
+    cmpl(cnt2, -stride); // Do not read beyond substring
     jccb(Assembler::lessEqual, CONT_SCAN_SUBSTR);
     // Back-up strings to avoid reading beyond substring:
     // cnt1 = cnt1 - cnt2 + 8
     addl(cnt1, cnt2); // cnt2 is negative
-    addl(cnt1, 8);
-    movl(cnt2, 8); negptr(cnt2);
+    addl(cnt1, stride);
+    movl(cnt2, stride); negptr(cnt2);
     bind(CONT_SCAN_SUBSTR);
     if (int_cnt2 < (int)G) {
-      movdqu(vec, Address(str2, cnt2, Address::times_2, int_cnt2*2));
-      pcmpestri(vec, Address(result, cnt2, Address::times_2, int_cnt2*2), 0x0d);
+      int tail_off1 = int_cnt2<<scale1;
+      int tail_off2 = int_cnt2<<scale2;
+      if (ae == StrIntrinsicNode::UL) {
+        pmovzxbw(vec, Address(str2, cnt2, scale2, tail_off2));
+      } else {
+        movdqu(vec, Address(str2, cnt2, scale2, tail_off2));
+      }
+      pcmpestri(vec, Address(result, cnt2, scale1, tail_off1), mode);
     } else {
       // calculate index in register to avoid integer overflow (int_cnt2*2)
       movl(tmp, int_cnt2);
       addptr(tmp, cnt2);
-      movdqu(vec, Address(str2, tmp, Address::times_2, 0));
-      pcmpestri(vec, Address(result, tmp, Address::times_2, 0), 0x0d);
+      if (ae == StrIntrinsicNode::UL) {
+        pmovzxbw(vec, Address(str2, tmp, scale2, 0));
+      } else {
+        movdqu(vec, Address(str2, tmp, scale2, 0));
+      }
+      pcmpestri(vec, Address(result, tmp, scale1, 0), mode);
     }
     // Need to reload strings pointers if not matched whole vector
     jcc(Assembler::noOverflow, RELOAD_SUBSTR); // OF == 0
-    addptr(cnt2, 8);
+    addptr(cnt2, stride);
     jcc(Assembler::negative, SCAN_SUBSTR);
     // Fall through if found full substring
 
@@ -6446,7 +6476,9 @@
   // Found result if we matched full small substring.
   // Compute substr offset
   subptr(result, str1);
-  shrl(result, 1); // index
+  if (ae == StrIntrinsicNode::UU || ae == StrIntrinsicNode::UL) {
+    shrl(result, 1); // index
+  }
   bind(EXIT);
 
 } // string_indexofC8
@@ -6455,9 +6487,12 @@
 void MacroAssembler::string_indexof(Register str1, Register str2,
                                     Register cnt1, Register cnt2,
                                     int int_cnt2,  Register result,
-                                    XMMRegister vec, Register tmp) {
+                                    XMMRegister vec, Register tmp,
+                                    int ae) {
   ShortBranchVerifier sbv(this);
   assert(UseSSE42Intrinsics, "SSE4.2 is required");
+  assert(ae != StrIntrinsicNode::LU, "Invalid encoding");
+
   //
   // int_cnt2 is length of small (< 8 chars) constant substring
   // or (-1) for non constant substring in which case its length
@@ -6467,18 +6502,22 @@
   // if (substr.count > string.count) return -1;
   // if (substr.count == 0) return 0;
   //
-  assert(int_cnt2 == -1 || (0 < int_cnt2 && int_cnt2 < 8), "should be != 0");
-
-  // This method uses pcmpestri instruction with bound registers
+  int stride = (ae == StrIntrinsicNode::LL) ? 16 : 8; //UU, UL -> 8
+  assert(int_cnt2 == -1 || (0 < int_cnt2 && int_cnt2 < stride), "should be != 0");
+  // This method uses the pcmpestri instruction with bound registers
   //   inputs:
   //     xmm - substring
   //     rax - substring length (elements count)
   //     mem - scanned string
   //     rdx - string length (elements count)
   //     0xd - mode: 1100 (substring search) + 01 (unsigned shorts)
+  //     0xc - mode: 1100 (substring search) + 00 (unsigned bytes)
   //   outputs:
   //     rcx - matched index in string
   assert(cnt1 == rdx && cnt2 == rax && tmp == rcx, "pcmpestri");
+  int mode = (ae == StrIntrinsicNode::LL) ? 0x0c : 0x0d; // bytes or shorts
+  Address::ScaleFactor scale1 = (ae == StrIntrinsicNode::LL) ? Address::times_1 : Address::times_2;
+  Address::ScaleFactor scale2 = (ae == StrIntrinsicNode::UL) ? Address::times_1 : scale1;
 
   Label RELOAD_SUBSTR, SCAN_TO_SUBSTR, SCAN_SUBSTR, ADJUST_STR,
         RET_FOUND, RET_NOT_FOUND, CLEANUP, FOUND_SUBSTR,
@@ -6492,23 +6531,40 @@
     movptr(tmp, rsp); // save old SP
 
     if (int_cnt2 > 0) {     // small (< 8 chars) constant substring
-      if (int_cnt2 == 1) {  // One char
+      if (int_cnt2 == (1>>scale2)) { // One byte
+        assert((ae == StrIntrinsicNode::LL || ae == StrIntrinsicNode::UL), "Only possible for latin1 encoding");
+        load_unsigned_byte(result, Address(str2, 0));
+        movdl(vec, result); // move 32 bits
+      } else if (ae == StrIntrinsicNode::LL && int_cnt2 == 3) {  // Three bytes
+        // Not enough header space in 32-bit VM: 12+3 = 15.
+        movl(result, Address(str2, -1));
+        shrl(result, 8);
+        movdl(vec, result); // move 32 bits
+      } else if (ae != StrIntrinsicNode::UL && int_cnt2 == (2>>scale2)) {  // One char
         load_unsigned_short(result, Address(str2, 0));
         movdl(vec, result); // move 32 bits
-      } else if (int_cnt2 == 2) { // Two chars
+      } else if (ae != StrIntrinsicNode::UL && int_cnt2 == (4>>scale2)) { // Two chars
         movdl(vec, Address(str2, 0)); // move 32 bits
-      } else if (int_cnt2 == 4) { // Four chars
+      } else if (ae != StrIntrinsicNode::UL && int_cnt2 == (8>>scale2)) { // Four chars
         movq(vec, Address(str2, 0));  // move 64 bits
-      } else { // cnt2 = { 3, 5, 6, 7 }
+      } else { // cnt2 = { 3, 5, 6, 7 } || (ae == StrIntrinsicNode::UL && cnt2 ={2, ..., 7})
         // Array header size is 12 bytes in 32-bit VM
         // + 6 bytes for 3 chars == 18 bytes,
         // enough space to load vec and shift.
         assert(HeapWordSize*TypeArrayKlass::header_size() >= 12,"sanity");
-        movdqu(vec, Address(str2, (int_cnt2*2)-16));
-        psrldq(vec, 16-(int_cnt2*2));
+        if (ae == StrIntrinsicNode::UL) {
+          int tail_off = int_cnt2-8;
+          pmovzxbw(vec, Address(str2, tail_off));
+          psrldq(vec, -2*tail_off);
+        }
+        else {
+          int tail_off = int_cnt2*(1<<scale2);
+          movdqu(vec, Address(str2, tail_off-16));
+          psrldq(vec, 16-tail_off);
+        }
       }
     } else { // not constant substring
-      cmpl(cnt2, 8);
+      cmpl(cnt2, stride);
       jccb(Assembler::aboveEqual, BIG_STRINGS); // Both strings are big enough
 
       // We can read beyond string if srt+16 does not cross page boundary
@@ -6521,12 +6577,17 @@
 
       // Move small strings to stack to allow load 16 bytes into vec.
       subptr(rsp, 16);
-      int stk_offset = wordSize-2;
+      int stk_offset = wordSize-(1<<scale2);
       push(cnt2);
 
       bind(COPY_SUBSTR);
-      load_unsigned_short(result, Address(str2, cnt2, Address::times_2, -2));
-      movw(Address(rsp, cnt2, Address::times_2, stk_offset), result);
+      if (ae == StrIntrinsicNode::LL || ae == StrIntrinsicNode::UL) {
+        load_unsigned_byte(result, Address(str2, cnt2, scale2, -1));
+        movb(Address(rsp, cnt2, scale2, stk_offset), result);
+      } else if (ae == StrIntrinsicNode::UU) {
+        load_unsigned_short(result, Address(str2, cnt2, scale2, -2));
+        movw(Address(rsp, cnt2, scale2, stk_offset), result);
+      }
       decrement(cnt2);
       jccb(Assembler::notZero, COPY_SUBSTR);
 
@@ -6535,7 +6596,7 @@
     } // non constant
 
     bind(CHECK_STR);
-    cmpl(cnt1, 8);
+    cmpl(cnt1, stride);
     jccb(Assembler::aboveEqual, BIG_STRINGS);
 
     // Check cross page boundary.
@@ -6545,7 +6606,7 @@
     jccb(Assembler::belowEqual, BIG_STRINGS);
 
     subptr(rsp, 16);
-    int stk_offset = -2;
+    int stk_offset = -(1<<scale1);
     if (int_cnt2 < 0) { // not constant
       push(cnt2);
       stk_offset += wordSize;
@@ -6553,8 +6614,13 @@
     movl(cnt2, cnt1);
 
     bind(COPY_STR);
-    load_unsigned_short(result, Address(str1, cnt2, Address::times_2, -2));
-    movw(Address(rsp, cnt2, Address::times_2, stk_offset), result);
+    if (ae == StrIntrinsicNode::LL) {
+      load_unsigned_byte(result, Address(str1, cnt2, scale1, -1));
+      movb(Address(rsp, cnt2, scale1, stk_offset), result);
+    } else {
+      load_unsigned_short(result, Address(str1, cnt2, scale1, -2));
+      movw(Address(rsp, cnt2, scale1, stk_offset), result);
+    }
     decrement(cnt2);
     jccb(Assembler::notZero, COPY_STR);
 
@@ -6566,7 +6632,11 @@
     bind(BIG_STRINGS);
     // Load substring.
     if (int_cnt2 < 0) { // -1
-      movdqu(vec, Address(str2, 0));
+      if (ae == StrIntrinsicNode::UL) {
+        pmovzxbw(vec, Address(str2, 0));
+      } else {
+        movdqu(vec, Address(str2, 0));
+      }
       push(cnt2);       // substr count
       push(str2);       // substr addr
       push(str1);       // string addr
@@ -6597,37 +6667,43 @@
     bind(RELOAD_SUBSTR);
     movptr(str2, Address(rsp, 2*wordSize));
     movl(cnt2, Address(rsp, 3*wordSize));
-    movdqu(vec, Address(str2, 0));
+    if (ae == StrIntrinsicNode::UL) {
+      pmovzxbw(vec, Address(str2, 0));
+    } else {
+      movdqu(vec, Address(str2, 0));
+    }
     // We came here after the beginning of the substring was
     // matched but the rest of it was not so we need to search
     // again. Start from the next element after the previous match.
     subptr(str1, result); // Restore counter
-    shrl(str1, 1);
+    if (ae == StrIntrinsicNode::UU || ae == StrIntrinsicNode::UL) {
+      shrl(str1, 1);
+    }
     addl(cnt1, str1);
     decrementl(cnt1);   // Shift to next element
     cmpl(cnt1, cnt2);
     jccb(Assembler::negative, RET_NOT_FOUND);  // Left less then substring
 
-    addptr(result, 2);
+    addptr(result, (1<<scale1));
   } // non constant
 
   // Scan string for start of substr in 16-byte vectors
   bind(SCAN_TO_SUBSTR);
   assert(cnt1 == rdx && cnt2 == rax && tmp == rcx, "pcmpestri");
-  pcmpestri(vec, Address(result, 0), 0x0d);
+  pcmpestri(vec, Address(result, 0), mode);
   jccb(Assembler::below, FOUND_CANDIDATE);   // CF == 1
-  subl(cnt1, 8);
+  subl(cnt1, stride);
   jccb(Assembler::lessEqual, RET_NOT_FOUND); // Scanned full string
   cmpl(cnt1, cnt2);
   jccb(Assembler::negative, RET_NOT_FOUND);  // Left less then substring
   addptr(result, 16);
 
   bind(ADJUST_STR);
-  cmpl(cnt1, 8); // Do not read beyond string
+  cmpl(cnt1, stride); // Do not read beyond string
   jccb(Assembler::greaterEqual, SCAN_TO_SUBSTR);
   // Back-up string to avoid reading beyond string.
-  lea(result, Address(result, cnt1, Address::times_2, -16));
-  movl(cnt1, 8);
+  lea(result, Address(result, cnt1, scale1, -16));
+  movl(cnt1, stride);
   jmpb(SCAN_TO_SUBSTR);
 
   // Found a potential substr
@@ -6646,13 +6722,12 @@
 
   bind(FOUND_SUBSTR);
   // Compute start addr of substr
-  lea(result, Address(result, tmp, Address::times_2));
-
+  lea(result, Address(result, tmp, scale1));
   if (int_cnt2 > 0) { // Constant substring
     // Repeat search for small substring (< 8 chars)
     // from new point without reloading substring.
     // Have to check that we don't read beyond string.
-    cmpl(tmp, 8-int_cnt2);
+    cmpl(tmp, stride-int_cnt2);
     jccb(Assembler::greater, ADJUST_STR);
     // Fall through if matched whole substring.
   } else { // non constant
@@ -6660,12 +6735,12 @@
 
     addl(tmp, cnt2);
     // Found result if we matched whole substring.
-    cmpl(tmp, 8);
+    cmpl(tmp, stride);
     jccb(Assembler::lessEqual, RET_FOUND);
 
     // Repeat search for small substring (<= 8 chars)
     // from new point 'str1' without reloading substring.
-    cmpl(cnt2, 8);
+    cmpl(cnt2, stride);
     // Have to check that we don't read beyond string.
     jccb(Assembler::lessEqual, ADJUST_STR);
 
@@ -6678,26 +6753,40 @@
     jccb(Assembler::equal, CHECK_NEXT);
 
     bind(SCAN_SUBSTR);
-    pcmpestri(vec, Address(str1, 0), 0x0d);
+    pcmpestri(vec, Address(str1, 0), mode);
     // Need to reload strings pointers if not matched whole vector
     jcc(Assembler::noOverflow, RELOAD_SUBSTR); // OF == 0
 
     bind(CHECK_NEXT);
-    subl(cnt2, 8);
+    subl(cnt2, stride);
     jccb(Assembler::lessEqual, RET_FOUND_LONG); // Found full substring
     addptr(str1, 16);
-    addptr(str2, 16);
-    subl(cnt1, 8);
-    cmpl(cnt2, 8); // Do not read beyond substring
+    if (ae == StrIntrinsicNode::UL) {
+      addptr(str2, 8);
+    } else {
+      addptr(str2, 16);
+    }
+    subl(cnt1, stride);
+    cmpl(cnt2, stride); // Do not read beyond substring
     jccb(Assembler::greaterEqual, CONT_SCAN_SUBSTR);
     // Back-up strings to avoid reading beyond substring.
-    lea(str2, Address(str2, cnt2, Address::times_2, -16));
-    lea(str1, Address(str1, cnt2, Address::times_2, -16));
+
+    if (ae == StrIntrinsicNode::UL) {
+      lea(str2, Address(str2, cnt2, scale2, -8));
+      lea(str1, Address(str1, cnt2, scale1, -16));
+    } else {
+      lea(str2, Address(str2, cnt2, scale2, -16));
+      lea(str1, Address(str1, cnt2, scale1, -16));
+    }
     subl(cnt1, cnt2);
-    movl(cnt2, 8);
-    addl(cnt1, 8);
+    movl(cnt2, stride);
+    addl(cnt1, stride);
     bind(CONT_SCAN_SUBSTR);
-    movdqu(vec, Address(str2, 0));
+    if (ae == StrIntrinsicNode::UL) {
+      pmovzxbw(vec, Address(str2, 0));
+    } else {
+      movdqu(vec, Address(str2, 0));
+    }
     jmpb(SCAN_SUBSTR);
 
     bind(RET_FOUND_LONG);
@@ -6707,20 +6796,143 @@
   bind(RET_FOUND);
   // Compute substr offset
   subptr(result, str1);
-  shrl(result, 1); // index
-
+  if (ae == StrIntrinsicNode::UU || ae == StrIntrinsicNode::UL) {
+    shrl(result, 1); // index
+  }
   bind(CLEANUP);
   pop(rsp); // restore SP
 
 } // string_indexof
 
-// Compare strings.
+void MacroAssembler::string_indexof_char(Register str1, Register cnt1, Register ch, Register result,
+                                         XMMRegister vec1, XMMRegister vec2, XMMRegister vec3, Register tmp) {
+  ShortBranchVerifier sbv(this);
+  assert(UseSSE42Intrinsics, "SSE4.2 is required");
+
+  int stride = 8;
+
+  Label FOUND_CHAR, SCAN_TO_CHAR, SCAN_TO_CHAR_LOOP,
+        SCAN_TO_8_CHAR, SCAN_TO_8_CHAR_LOOP, SCAN_TO_16_CHAR_LOOP,
+        RET_NOT_FOUND, SCAN_TO_8_CHAR_INIT,
+        FOUND_SEQ_CHAR, DONE_LABEL;
+
+  movptr(result, str1);
+  if (UseAVX >= 2) {
+    cmpl(cnt1, stride);
+    jccb(Assembler::less, SCAN_TO_CHAR_LOOP);
+    cmpl(cnt1, 2*stride);
+    jccb(Assembler::less, SCAN_TO_8_CHAR_INIT);
+    movdl(vec1, ch);
+    vpbroadcastw(vec1, vec1);
+    vpxor(vec2, vec2);
+    movl(tmp, cnt1);
+    andl(tmp, 0xFFFFFFF0);  //vector count (in chars)
+    andl(cnt1,0x0000000F);  //tail count (in chars)
+
+    bind(SCAN_TO_16_CHAR_LOOP);
+    vmovdqu(vec3, Address(result, 0));
+    vpcmpeqw(vec3, vec3, vec1, true);
+    vptest(vec2, vec3);
+    jcc(Assembler::carryClear, FOUND_CHAR);
+    addptr(result, 32);
+    subl(tmp, 2*stride);
+    jccb(Assembler::notZero, SCAN_TO_16_CHAR_LOOP);
+    jmp(SCAN_TO_8_CHAR);
+    bind(SCAN_TO_8_CHAR_INIT);
+    movdl(vec1, ch);
+    pshuflw(vec1, vec1, 0x00);
+    pshufd(vec1, vec1, 0);
+    pxor(vec2, vec2);
+  }
+  if (UseAVX >= 2 || UseSSE42Intrinsics) {
+    bind(SCAN_TO_8_CHAR);
+    cmpl(cnt1, stride);
+    if (UseAVX >= 2) {
+      jccb(Assembler::less, SCAN_TO_CHAR);
+    }
+    if (!(UseAVX >= 2)) {
+      jccb(Assembler::less, SCAN_TO_CHAR_LOOP);
+      movdl(vec1, ch);
+      pshuflw(vec1, vec1, 0x00);
+      pshufd(vec1, vec1, 0);
+      pxor(vec2, vec2);
+    }
+    movl(tmp, cnt1);
+    andl(tmp, 0xFFFFFFF8);  //vector count (in chars)
+    andl(cnt1,0x00000007);  //tail count (in chars)
+
+    bind(SCAN_TO_8_CHAR_LOOP);
+    movdqu(vec3, Address(result, 0));
+    pcmpeqw(vec3, vec1);
+    ptest(vec2, vec3);
+    jcc(Assembler::carryClear, FOUND_CHAR);
+    addptr(result, 16);
+    subl(tmp, stride);
+    jccb(Assembler::notZero, SCAN_TO_8_CHAR_LOOP);
+  }
+  bind(SCAN_TO_CHAR);
+  testl(cnt1, cnt1);
+  jcc(Assembler::zero, RET_NOT_FOUND);
+
+  bind(SCAN_TO_CHAR_LOOP);
+  load_unsigned_short(tmp, Address(result, 0));
+  cmpl(ch, tmp);
+  jccb(Assembler::equal, FOUND_SEQ_CHAR);
+  addptr(result, 2);
+  subl(cnt1, 1);
+  jccb(Assembler::zero, RET_NOT_FOUND);
+  jmp(SCAN_TO_CHAR_LOOP);
+
+  bind(RET_NOT_FOUND);
+  movl(result, -1);
+  jmpb(DONE_LABEL);
+
+  if (UseAVX >= 2 || UseSSE42Intrinsics) {
+    bind(FOUND_CHAR);
+    if (UseAVX >= 2) {
+      vpmovmskb(tmp, vec3);
+    } else {
+      pmovmskb(tmp, vec3);
+    }
+    bsfl(ch, tmp);
+    addl(result, ch);
+  }
+
+  bind(FOUND_SEQ_CHAR);
+  subptr(result, str1);
+  shrl(result, 1);
+
+  bind(DONE_LABEL);
+} // string_indexof_char
+
+// helper function for string_compare
+void MacroAssembler::load_next_elements(Register elem1, Register elem2, Register str1, Register str2,
+                                        Address::ScaleFactor scale, Address::ScaleFactor scale1,
+                                        Address::ScaleFactor scale2, Register index, int ae) {
+  if (ae == StrIntrinsicNode::LL) {
+    load_unsigned_byte(elem1, Address(str1, index, scale, 0));
+    load_unsigned_byte(elem2, Address(str2, index, scale, 0));
+  } else if (ae == StrIntrinsicNode::UU) {
+    load_unsigned_short(elem1, Address(str1, index, scale, 0));
+    load_unsigned_short(elem2, Address(str2, index, scale, 0));
+  } else {
+    load_unsigned_byte(elem1, Address(str1, index, scale1, 0));
+    load_unsigned_short(elem2, Address(str2, index, scale2, 0));
+  }
+}
+
+// Compare strings, used for char[] and byte[].
 void MacroAssembler::string_compare(Register str1, Register str2,
                                     Register cnt1, Register cnt2, Register result,
-                                    XMMRegister vec1) {
+                                    XMMRegister vec1, int ae) {
   ShortBranchVerifier sbv(this);
   Label LENGTH_DIFF_LABEL, POP_LABEL, DONE_LABEL, WHILE_HEAD_LABEL;
-
+  int stride, stride2, adr_stride, adr_stride1, adr_stride2;
+  Address::ScaleFactor scale, scale1, scale2;
+
+  if (ae == StrIntrinsicNode::LU || ae == StrIntrinsicNode::UL) {
+    shrl(cnt2, 1);
+  }
   // Compute the minimum of the string lengths and the
   // difference of the string lengths (stack).
   // Do the conditional move stuff
@@ -6732,32 +6944,67 @@
   // Is the minimum length zero?
   testl(cnt2, cnt2);
   jcc(Assembler::zero, LENGTH_DIFF_LABEL);
-
-  // Compare first characters
-  load_unsigned_short(result, Address(str1, 0));
-  load_unsigned_short(cnt1, Address(str2, 0));
+  if (ae == StrIntrinsicNode::LL) {
+    // Load first bytes
+    load_unsigned_byte(result, Address(str1, 0));
+    load_unsigned_byte(cnt1, Address(str2, 0));
+  } else if (ae == StrIntrinsicNode::UU) {
+    // Load first characters
+    load_unsigned_short(result, Address(str1, 0));
+    load_unsigned_short(cnt1, Address(str2, 0));
+  } else {
+    load_unsigned_byte(result, Address(str1, 0));
+    load_unsigned_short(cnt1, Address(str2, 0));
+  }
   subl(result, cnt1);
   jcc(Assembler::notZero,  POP_LABEL);
+
+  if (ae == StrIntrinsicNode::UU) {
+    // Divide length by 2 to get number of chars
+    shrl(cnt2, 1);
+  }
   cmpl(cnt2, 1);
   jcc(Assembler::equal, LENGTH_DIFF_LABEL);
 
-  // Check if the strings start at the same location.
-  cmpptr(str1, str2);
-  jcc(Assembler::equal, LENGTH_DIFF_LABEL);
-
-  Address::ScaleFactor scale = Address::times_2;
-  int stride = 8;
+  // Check if the strings start at the same location and setup scale and stride
+  if (ae == StrIntrinsicNode::LL || ae == StrIntrinsicNode::UU) {
+    cmpptr(str1, str2);
+    jcc(Assembler::equal, LENGTH_DIFF_LABEL);
+    if (ae == StrIntrinsicNode::LL) {
+      scale = Address::times_1;
+      stride = 16;
+    } else {
+      scale = Address::times_2;
+      stride = 8;
+    }
+  } else {
+    scale1 = Address::times_1;
+    scale2 = Address::times_2;
+    stride = 8;
+  }
 
   if (UseAVX >= 2 && UseSSE42Intrinsics) {
     Label COMPARE_WIDE_VECTORS, VECTOR_NOT_EQUAL, COMPARE_WIDE_TAIL, COMPARE_SMALL_STR;
     Label COMPARE_WIDE_VECTORS_LOOP, COMPARE_16_CHARS, COMPARE_INDEX_CHAR;
     Label COMPARE_TAIL_LONG;
     int pcmpmask = 0x19;
+    if (ae == StrIntrinsicNode::LL) {
+      pcmpmask &= ~0x01;
+    }
 
     // Setup to compare 16-chars (32-bytes) vectors,
     // start from first character again because it has aligned address.
-    int stride2 = 16;
-    int adr_stride  = stride  << scale;
+    if (ae == StrIntrinsicNode::LL) {
+      stride2 = 32;
+    } else {
+      stride2 = 16;
+    }
+    if (ae == StrIntrinsicNode::LL || ae == StrIntrinsicNode::UU) {
+      adr_stride = stride << scale;
+    } else {
+      adr_stride1 = 8;  //stride << scale1;
+      adr_stride2 = 16; //stride << scale2;
+    }
 
     assert(result == rax && cnt2 == rdx && cnt1 == rcx, "pcmpestri");
     // rax and rdx are used by pcmpestri as elements counters
@@ -6767,26 +7014,39 @@
 
     // fast path : compare first 2 8-char vectors.
     bind(COMPARE_16_CHARS);
-    movdqu(vec1, Address(str1, 0));
+    if (ae == StrIntrinsicNode::LL || ae == StrIntrinsicNode::UU) {
+      movdqu(vec1, Address(str1, 0));
+    } else {
+      pmovzxbw(vec1, Address(str1, 0));
+    }
     pcmpestri(vec1, Address(str2, 0), pcmpmask);
     jccb(Assembler::below, COMPARE_INDEX_CHAR);
 
-    movdqu(vec1, Address(str1, adr_stride));
-    pcmpestri(vec1, Address(str2, adr_stride), pcmpmask);
+    if (ae == StrIntrinsicNode::LL || ae == StrIntrinsicNode::UU) {
+      movdqu(vec1, Address(str1, adr_stride));
+      pcmpestri(vec1, Address(str2, adr_stride), pcmpmask);
+    } else {
+      pmovzxbw(vec1, Address(str1, adr_stride1));
+      pcmpestri(vec1, Address(str2, adr_stride2), pcmpmask);
+    }
     jccb(Assembler::aboveEqual, COMPARE_WIDE_VECTORS);
     addl(cnt1, stride);
 
     // Compare the characters at index in cnt1
-    bind(COMPARE_INDEX_CHAR); //cnt1 has the offset of the mismatching character
-    load_unsigned_short(result, Address(str1, cnt1, scale));
-    load_unsigned_short(cnt2, Address(str2, cnt1, scale));
+    bind(COMPARE_INDEX_CHAR); // cnt1 has the offset of the mismatching character
+    load_next_elements(result, cnt2, str1, str2, scale, scale1, scale2, cnt1, ae);
     subl(result, cnt2);
     jmp(POP_LABEL);
 
     // Setup the registers to start vector comparison loop
     bind(COMPARE_WIDE_VECTORS);
-    lea(str1, Address(str1, result, scale));
-    lea(str2, Address(str2, result, scale));
+    if (ae == StrIntrinsicNode::LL || ae == StrIntrinsicNode::UU) {
+      lea(str1, Address(str1, result, scale));
+      lea(str2, Address(str2, result, scale));
+    } else {
+      lea(str1, Address(str1, result, scale1));
+      lea(str2, Address(str2, result, scale2));
+    }
     subl(result, stride2);
     subl(cnt2, stride2);
     jccb(Assembler::zero, COMPARE_WIDE_TAIL);
@@ -6794,8 +7054,13 @@
 
     //  In a loop, compare 16-chars (32-bytes) at once using (vpxor+vptest)
     bind(COMPARE_WIDE_VECTORS_LOOP);
-    vmovdqu(vec1, Address(str1, result, scale));
-    vpxor(vec1, Address(str2, result, scale));
+    if (ae == StrIntrinsicNode::LL || ae == StrIntrinsicNode::UU) {
+      vmovdqu(vec1, Address(str1, result, scale));
+      vpxor(vec1, Address(str2, result, scale));
+    } else {
+      vpmovzxbw(vec1, Address(str1, result, scale1));
+      vpxor(vec1, Address(str2, result, scale2));
+    }
     vptest(vec1, vec1);
     jccb(Assembler::notZero, VECTOR_NOT_EQUAL);
     addptr(result, stride2);
@@ -6818,8 +7083,13 @@
     bind(VECTOR_NOT_EQUAL);
     // clean upper bits of YMM registers
     vpxor(vec1, vec1);
-    lea(str1, Address(str1, result, scale));
-    lea(str2, Address(str2, result, scale));
+    if (ae == StrIntrinsicNode::LL || ae == StrIntrinsicNode::UU) {
+      lea(str1, Address(str1, result, scale));
+      lea(str2, Address(str2, result, scale));
+    } else {
+      lea(str1, Address(str1, result, scale1));
+      lea(str2, Address(str2, result, scale2));
+    }
     jmp(COMPARE_16_CHARS);
 
     // Compare tail chars, length between 1 to 15 chars
@@ -6828,13 +7098,22 @@
     cmpl(cnt2, stride);
     jccb(Assembler::less, COMPARE_SMALL_STR);
 
-    movdqu(vec1, Address(str1, 0));
+    if (ae == StrIntrinsicNode::LL || ae == StrIntrinsicNode::UU) {
+      movdqu(vec1, Address(str1, 0));
+    } else {
+      pmovzxbw(vec1, Address(str1, 0));
+    }
     pcmpestri(vec1, Address(str2, 0), pcmpmask);
     jcc(Assembler::below, COMPARE_INDEX_CHAR);
     subptr(cnt2, stride);
     jccb(Assembler::zero, LENGTH_DIFF_LABEL);
-    lea(str1, Address(str1, result, scale));
-    lea(str2, Address(str2, result, scale));
+    if (ae == StrIntrinsicNode::LL || ae == StrIntrinsicNode::UU) {
+      lea(str1, Address(str1, result, scale));
+      lea(str2, Address(str2, result, scale));
+    } else {
+      lea(str1, Address(str1, result, scale1));
+      lea(str2, Address(str2, result, scale2));
+    }
     negptr(cnt2);
     jmpb(WHILE_HEAD_LABEL);
 
@@ -6846,10 +7125,17 @@
     // start from first character again because it has aligned address.
     movl(result, cnt2);
     andl(cnt2, ~(stride - 1));   // cnt2 holds the vector count
+    if (ae == StrIntrinsicNode::LL) {
+      pcmpmask &= ~0x01;
+    }
     jccb(Assembler::zero, COMPARE_TAIL);
-
-    lea(str1, Address(str1, result, scale));
-    lea(str2, Address(str2, result, scale));
+    if (ae == StrIntrinsicNode::LL || ae == StrIntrinsicNode::UU) {
+      lea(str1, Address(str1, result, scale));
+      lea(str2, Address(str2, result, scale));
+    } else {
+      lea(str1, Address(str1, result, scale1));
+      lea(str2, Address(str2, result, scale2));
+    }
     negptr(result);
 
     // pcmpestri
@@ -6865,8 +7151,13 @@
     assert(result == rax && cnt2 == rdx && cnt1 == rcx, "pcmpestri");
 
     bind(COMPARE_WIDE_VECTORS);
-    movdqu(vec1, Address(str1, result, scale));
-    pcmpestri(vec1, Address(str2, result, scale), pcmpmask);
+    if (ae == StrIntrinsicNode::LL || ae == StrIntrinsicNode::UU) {
+      movdqu(vec1, Address(str1, result, scale));
+      pcmpestri(vec1, Address(str2, result, scale), pcmpmask);
+    } else {
+      pmovzxbw(vec1, Address(str1, result, scale1));
+      pcmpestri(vec1, Address(str2, result, scale2), pcmpmask);
+    }
     // After pcmpestri cnt1(rcx) contains mismatched element index
 
     jccb(Assembler::below, VECTOR_NOT_EQUAL);  // CF==1
@@ -6881,15 +7172,19 @@
     movl(cnt2, stride);
     movl(result, stride);
     negptr(result);
-    movdqu(vec1, Address(str1, result, scale));
-    pcmpestri(vec1, Address(str2, result, scale), pcmpmask);
+    if (ae == StrIntrinsicNode::LL || ae == StrIntrinsicNode::UU) {
+      movdqu(vec1, Address(str1, result, scale));
+      pcmpestri(vec1, Address(str2, result, scale), pcmpmask);
+    } else {
+      pmovzxbw(vec1, Address(str1, result, scale1));
+      pcmpestri(vec1, Address(str2, result, scale2), pcmpmask);
+    }
     jccb(Assembler::aboveEqual, LENGTH_DIFF_LABEL);
 
     // Mismatched characters in the vectors
     bind(VECTOR_NOT_EQUAL);
     addptr(cnt1, result);
-    load_unsigned_short(result, Address(str1, cnt1, scale));
-    load_unsigned_short(cnt2, Address(str2, cnt1, scale));
+    load_next_elements(result, cnt2, str1, str2, scale, scale1, scale2, cnt1, ae);
     subl(result, cnt2);
     jmpb(POP_LABEL);
 
@@ -6898,15 +7193,19 @@
     // Fallthru to tail compare
   }
   // Shift str2 and str1 to the end of the arrays, negate min
-  lea(str1, Address(str1, cnt2, scale));
-  lea(str2, Address(str2, cnt2, scale));
+  if (ae == StrIntrinsicNode::LL || ae == StrIntrinsicNode::UU) {
+    lea(str1, Address(str1, cnt2, scale));
+    lea(str2, Address(str2, cnt2, scale));
+  } else {
+    lea(str1, Address(str1, cnt2, scale1));
+    lea(str2, Address(str2, cnt2, scale2));
+  }
   decrementl(cnt2);  // first character was compared already
   negptr(cnt2);
 
   // Compare the rest of the elements
   bind(WHILE_HEAD_LABEL);
-  load_unsigned_short(result, Address(str1, cnt2, scale, 0));
-  load_unsigned_short(cnt1, Address(str2, cnt2, scale, 0));
+  load_next_elements(result, cnt1, str1, str2, scale, scale1, scale2, cnt2, ae);
   subl(result, cnt1);
   jccb(Assembler::notZero, POP_LABEL);
   increment(cnt2);
@@ -6915,6 +7214,10 @@
   // Strings are equal up to min length.  Return the length difference.
   bind(LENGTH_DIFF_LABEL);
   pop(result);
+  if (ae == StrIntrinsicNode::UU) {
+    // Divide diff by 2 to get number of chars
+    sarl(result, 1);
+  }
   jmpb(DONE_LABEL);
 
   // Discard the stored length difference
@@ -6923,23 +7226,164 @@
 
   // That's it
   bind(DONE_LABEL);
-}
-
-// Compare char[] arrays aligned to 4 bytes or substrings.
-void MacroAssembler::char_arrays_equals(bool is_array_equ, Register ary1, Register ary2,
-                                        Register limit, Register result, Register chr,
-                                        XMMRegister vec1, XMMRegister vec2) {
+  if(ae == StrIntrinsicNode::UL) {
+    negl(result);
+  }
+}
+
+// Search for Non-ASCII character (Negative byte value) in a byte array,
+// return true if it has any and false otherwise.
+void MacroAssembler::has_negatives(Register ary1, Register len,
+                                   Register result, Register tmp1,
+                                   XMMRegister vec1, XMMRegister vec2) {
+
+  // rsi: byte array
+  // rcx: len
+  // rax: result
   ShortBranchVerifier sbv(this);
-  Label TRUE_LABEL, FALSE_LABEL, DONE, COMPARE_VECTORS, COMPARE_CHAR;
+  assert_different_registers(ary1, len, result, tmp1);
+  assert_different_registers(vec1, vec2);
+  Label TRUE_LABEL, FALSE_LABEL, DONE, COMPARE_CHAR, COMPARE_VECTORS, COMPARE_BYTE;
+
+  // len == 0
+  testl(len, len);
+  jcc(Assembler::zero, FALSE_LABEL);
+
+  movl(result, len); // copy
+
+  if (UseAVX >= 2) {
+    // With AVX2, use 32-byte vector compare
+    Label COMPARE_WIDE_VECTORS, COMPARE_TAIL;
+
+    // Compare 32-byte vectors
+    andl(result, 0x0000001f);  //   tail count (in bytes)
+    andl(len, 0xffffffe0);   // vector count (in bytes)
+    jccb(Assembler::zero, COMPARE_TAIL);
+
+    lea(ary1, Address(ary1, len, Address::times_1));
+    negptr(len);
+
+    movl(tmp1, 0x80808080);   // create mask to test for Unicode chars in vector
+    movdl(vec2, tmp1);
+    vpbroadcastd(vec2, vec2);
+
+    bind(COMPARE_WIDE_VECTORS);
+    vmovdqu(vec1, Address(ary1, len, Address::times_1));
+    vptest(vec1, vec2);
+    jccb(Assembler::notZero, TRUE_LABEL);
+    addptr(len, 32);
+    jcc(Assembler::notZero, COMPARE_WIDE_VECTORS);
+
+    testl(result, result);
+    jccb(Assembler::zero, FALSE_LABEL);
+
+    vmovdqu(vec1, Address(ary1, result, Address::times_1, -32));
+    vptest(vec1, vec2);
+    jccb(Assembler::notZero, TRUE_LABEL);
+    jmpb(FALSE_LABEL);
+
+    bind(COMPARE_TAIL); // len is zero
+    movl(len, result);
+    // Fallthru to tail compare
+  } else if (UseSSE42Intrinsics) {
+    // With SSE4.2, use double quad vector compare
+    Label COMPARE_WIDE_VECTORS, COMPARE_TAIL;
+
+    // Compare 16-byte vectors
+    andl(result, 0x0000000f);  //   tail count (in bytes)
+    andl(len, 0xfffffff0);   // vector count (in bytes)
+    jccb(Assembler::zero, COMPARE_TAIL);
+
+    lea(ary1, Address(ary1, len, Address::times_1));
+    negptr(len);
+
+    movl(tmp1, 0x80808080);
+    movdl(vec2, tmp1);
+    pshufd(vec2, vec2, 0);
+
+    bind(COMPARE_WIDE_VECTORS);
+    movdqu(vec1, Address(ary1, len, Address::times_1));
+    ptest(vec1, vec2);
+    jccb(Assembler::notZero, TRUE_LABEL);
+    addptr(len, 16);
+    jcc(Assembler::notZero, COMPARE_WIDE_VECTORS);
+
+    testl(result, result);
+    jccb(Assembler::zero, FALSE_LABEL);
+
+    movdqu(vec1, Address(ary1, result, Address::times_1, -16));
+    ptest(vec1, vec2);
+    jccb(Assembler::notZero, TRUE_LABEL);
+    jmpb(FALSE_LABEL);
+
+    bind(COMPARE_TAIL); // len is zero
+    movl(len, result);
+    // Fallthru to tail compare
+  }
+
+  // Compare 4-byte vectors
+  andl(len, 0xfffffffc); // vector count (in bytes)
+  jccb(Assembler::zero, COMPARE_CHAR);
+
+  lea(ary1, Address(ary1, len, Address::times_1));
+  negptr(len);
+
+  bind(COMPARE_VECTORS);
+  movl(tmp1, Address(ary1, len, Address::times_1));
+  andl(tmp1, 0x80808080);
+  jccb(Assembler::notZero, TRUE_LABEL);
+  addptr(len, 4);
+  jcc(Assembler::notZero, COMPARE_VECTORS);
+
+  // Compare trailing char (final 2 bytes), if any
+  bind(COMPARE_CHAR);
+  testl(result, 0x2);   // tail  char
+  jccb(Assembler::zero, COMPARE_BYTE);
+  load_unsigned_short(tmp1, Address(ary1, 0));
+  andl(tmp1, 0x00008080);
+  jccb(Assembler::notZero, TRUE_LABEL);
+  subptr(result, 2);
+  lea(ary1, Address(ary1, 2));
+
+  bind(COMPARE_BYTE);
+  testl(result, 0x1);   // tail  byte
+  jccb(Assembler::zero, FALSE_LABEL);
+  load_unsigned_byte(tmp1, Address(ary1, 0));
+  andl(tmp1, 0x00000080);
+  jccb(Assembler::notEqual, TRUE_LABEL);
+  jmpb(FALSE_LABEL);
+
+  bind(TRUE_LABEL);
+  movl(result, 1);   // return true
+  jmpb(DONE);
+
+  bind(FALSE_LABEL);
+  xorl(result, result); // return false
+
+  // That's it
+  bind(DONE);
+  if (UseAVX >= 2) {
+    // clean upper bits of YMM registers
+    vpxor(vec1, vec1);
+    vpxor(vec2, vec2);
+  }
+}
+
+// Compare char[] or byte[] arrays aligned to 4 bytes or substrings.
+void MacroAssembler::arrays_equals(bool is_array_equ, Register ary1, Register ary2,
+                                   Register limit, Register result, Register chr,
+                                   XMMRegister vec1, XMMRegister vec2, bool is_char) {
+  ShortBranchVerifier sbv(this);
+  Label TRUE_LABEL, FALSE_LABEL, DONE, COMPARE_VECTORS, COMPARE_CHAR, COMPARE_BYTE;
 
   int length_offset  = arrayOopDesc::length_offset_in_bytes();
-  int base_offset    = arrayOopDesc::base_offset_in_bytes(T_CHAR);
-
-  // Check the input args
-  cmpptr(ary1, ary2);
-  jcc(Assembler::equal, TRUE_LABEL);
+  int base_offset    = arrayOopDesc::base_offset_in_bytes(is_char ? T_CHAR : T_BYTE);
 
   if (is_array_equ) {
+    // Check the input args
+    cmpptr(ary1, ary2);
+    jcc(Assembler::equal, TRUE_LABEL);
+
     // Need additional checks for arrays_equals.
     testptr(ary1, ary1);
     jcc(Assembler::zero, FALSE_LABEL);
@@ -6962,7 +7406,10 @@
     lea(ary2, Address(ary2, base_offset));
   }
 
-  shll(limit, 1);      // byte count != 0
+  if (is_array_equ && is_char) {
+    // arrays_equals when used for char[].
+    shll(limit, 1);      // byte count != 0
+  }
   movl(result, limit); // copy
 
   if (UseAVX >= 2) {
@@ -6970,7 +7417,7 @@
     Label COMPARE_WIDE_VECTORS, COMPARE_TAIL;
 
     // Compare 32-byte vectors
-    andl(result, 0x0000001e);  //   tail count (in bytes)
+    andl(result, 0x0000001f);  //   tail count (in bytes)
     andl(limit, 0xffffffe0);   // vector count (in bytes)
     jccb(Assembler::zero, COMPARE_TAIL);
 
@@ -7007,7 +7454,7 @@
     Label COMPARE_WIDE_VECTORS, COMPARE_TAIL;
 
     // Compare 16-byte vectors
-    andl(result, 0x0000000e);  //   tail count (in bytes)
+    andl(result, 0x0000000f);  //   tail count (in bytes)
     andl(limit, 0xfffffff0);   // vector count (in bytes)
     jccb(Assembler::zero, COMPARE_TAIL);
 
@@ -7059,12 +7506,26 @@
   // Compare trailing char (final 2 bytes), if any
   bind(COMPARE_CHAR);
   testl(result, 0x2);   // tail  char
-  jccb(Assembler::zero, TRUE_LABEL);
+  jccb(Assembler::zero, COMPARE_BYTE);
   load_unsigned_short(chr, Address(ary1, 0));
   load_unsigned_short(limit, Address(ary2, 0));
   cmpl(chr, limit);
   jccb(Assembler::notEqual, FALSE_LABEL);
 
+  if (is_array_equ && is_char) {
+    bind(COMPARE_BYTE);
+  } else {
+    lea(ary1, Address(ary1, 2));
+    lea(ary2, Address(ary2, 2));
+
+    bind(COMPARE_BYTE);
+    testl(result, 0x1);   // tail  byte
+    jccb(Assembler::zero, TRUE_LABEL);
+    load_unsigned_byte(chr, Address(ary1, 0));
+    load_unsigned_byte(limit, Address(ary2, 0));
+    cmpl(chr, limit);
+    jccb(Assembler::notEqual, FALSE_LABEL);
+  }
   bind(TRUE_LABEL);
   movl(result, 1);   // return true
   jmpb(DONE);
@@ -7081,6 +7542,8 @@
   }
 }
 
+#endif
+
 void MacroAssembler::generate_fill(BasicType t, bool aligned,
                                    Register to, Register value, Register count,
                                    Register rtmp, XMMRegister xtmp) {
@@ -9085,6 +9548,179 @@
 #undef BLOCK_COMMENT
 
 
+// Compress char[] array to byte[].
+void MacroAssembler::char_array_compress(Register src, Register dst, Register len,
+                                         XMMRegister tmp1Reg, XMMRegister tmp2Reg,
+                                         XMMRegister tmp3Reg, XMMRegister tmp4Reg,
+                                         Register tmp5, Register result) {
+  Label copy_chars_loop, return_length, return_zero, done;
+
+  // rsi: src
+  // rdi: dst
+  // rdx: len
+  // rcx: tmp5
+  // rax: result
+
+  // rsi holds start addr of source char[] to be compressed
+  // rdi holds start addr of destination byte[]
+  // rdx holds length
+
+  assert(len != result, "");
+
+  // save length for return
+  push(len);
+
+  if (UseSSE42Intrinsics) {
+    Label copy_32_loop, copy_16, copy_tail;
+
+    movl(result, len);
+    movl(tmp5, 0xff00ff00);   // create mask to test for Unicode chars in vectors
+
+    // vectored compression
+    andl(len, 0xfffffff0);    // vector count (in chars)
+    andl(result, 0x0000000f);    // tail count (in chars)
+    testl(len, len);
+    jccb(Assembler::zero, copy_16);
+
+    // compress 16 chars per iter
+    movdl(tmp1Reg, tmp5);
+    pshufd(tmp1Reg, tmp1Reg, 0);   // store Unicode mask in tmp1Reg
+    pxor(tmp4Reg, tmp4Reg);
+
+    lea(src, Address(src, len, Address::times_2));
+    lea(dst, Address(dst, len, Address::times_1));
+    negptr(len);
+
+    bind(copy_32_loop);
+    movdqu(tmp2Reg, Address(src, len, Address::times_2));     // load 1st 8 characters
+    por(tmp4Reg, tmp2Reg);
+    movdqu(tmp3Reg, Address(src, len, Address::times_2, 16)); // load next 8 characters
+    por(tmp4Reg, tmp3Reg);
+    ptest(tmp4Reg, tmp1Reg);       // check for Unicode chars in next vector
+    jcc(Assembler::notZero, return_zero);
+    packuswb(tmp2Reg, tmp3Reg);    // only ASCII chars; compress each to 1 byte
+    movdqu(Address(dst, len, Address::times_1), tmp2Reg);
+    addptr(len, 16);
+    jcc(Assembler::notZero, copy_32_loop);
+
+    // compress next vector of 8 chars (if any)
+    bind(copy_16);
+    movl(len, result);
+    andl(len, 0xfffffff8);    // vector count (in chars)
+    andl(result, 0x00000007);    // tail count (in chars)
+    testl(len, len);
+    jccb(Assembler::zero, copy_tail);
+
+    movdl(tmp1Reg, tmp5);
+    pshufd(tmp1Reg, tmp1Reg, 0);   // store Unicode mask in tmp1Reg
+    pxor(tmp3Reg, tmp3Reg);
+
+    movdqu(tmp2Reg, Address(src, 0));
+    ptest(tmp2Reg, tmp1Reg);       // check for Unicode chars in vector
+    jccb(Assembler::notZero, return_zero);
+    packuswb(tmp2Reg, tmp3Reg);    // only LATIN1 chars; compress each to 1 byte
+    movq(Address(dst, 0), tmp2Reg);
+    addptr(src, 16);
+    addptr(dst, 8);
+
+    bind(copy_tail);
+    movl(len, result);
+  }
+  // compress 1 char per iter
+  testl(len, len);
+  jccb(Assembler::zero, return_length);
+  lea(src, Address(src, len, Address::times_2));
+  lea(dst, Address(dst, len, Address::times_1));
+  negptr(len);
+
+  bind(copy_chars_loop);
+  load_unsigned_short(result, Address(src, len, Address::times_2));
+  testl(result, 0xff00);      // check if Unicode char
+  jccb(Assembler::notZero, return_zero);
+  movb(Address(dst, len, Address::times_1), result);  // ASCII char; compress to 1 byte
+  increment(len);
+  jcc(Assembler::notZero, copy_chars_loop);
+
+  // if compression succeeded, return length
+  bind(return_length);
+  pop(result);
+  jmpb(done);
+
+  // if compression failed, return 0
+  bind(return_zero);
+  xorl(result, result);
+  addptr(rsp, wordSize);
+
+  bind(done);
+}
+
+// Inflate byte[] array to char[].
+void MacroAssembler::byte_array_inflate(Register src, Register dst, Register len,
+                                        XMMRegister tmp1, Register tmp2) {
+  Label copy_chars_loop, done;
+
+  // rsi: src
+  // rdi: dst
+  // rdx: len
+  // rcx: tmp2
+
+  // rsi holds start addr of source byte[] to be inflated
+  // rdi holds start addr of destination char[]
+  // rdx holds length
+  assert_different_registers(src, dst, len, tmp2);
+
+  if (UseSSE42Intrinsics) {
+    Label copy_8_loop, copy_bytes, copy_tail;
+
+    movl(tmp2, len);
+    andl(tmp2, 0x00000007);   // tail count (in chars)
+    andl(len, 0xfffffff8);    // vector count (in chars)
+    jccb(Assembler::zero, copy_tail);
+
+    // vectored inflation
+    lea(src, Address(src, len, Address::times_1));
+    lea(dst, Address(dst, len, Address::times_2));
+    negptr(len);
+
+    // inflate 8 chars per iter
+    bind(copy_8_loop);
+    pmovzxbw(tmp1, Address(src, len, Address::times_1));  // unpack to 8 words
+    movdqu(Address(dst, len, Address::times_2), tmp1);
+    addptr(len, 8);
+    jcc(Assembler::notZero, copy_8_loop);
+
+    bind(copy_tail);
+    movl(len, tmp2);
+
+    cmpl(len, 4);
+    jccb(Assembler::less, copy_bytes);
+
+    movdl(tmp1, Address(src, 0));  // load 4 byte chars
+    pmovzxbw(tmp1, tmp1);
+    movq(Address(dst, 0), tmp1);
+    subptr(len, 4);
+    addptr(src, 4);
+    addptr(dst, 8);
+
+    bind(copy_bytes);
+  }
+  testl(len, len);
+  jccb(Assembler::zero, done);
+  lea(src, Address(src, len, Address::times_1));
+  lea(dst, Address(dst, len, Address::times_2));
+  negptr(len);
+
+  // inflate 1 char per iter
+  bind(copy_chars_loop);
+  load_unsigned_byte(tmp2, Address(src, len, Address::times_1));  // load byte char
+  movw(Address(dst, len, Address::times_2), tmp2);  // inflate byte char to word
+  increment(len);
+  jcc(Assembler::notZero, copy_chars_loop);
+
+  bind(done);
+}
+
+
 Assembler::Condition MacroAssembler::negate_condition(Assembler::Condition cond) {
   switch (cond) {
     // Note some conditions are synonyms for others
--- a/hotspot/src/cpu/x86/vm/macroAssembler_x86.hpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/cpu/x86/vm/macroAssembler_x86.hpp	Tue Nov 03 09:41:03 2015 +0100
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1997, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1997, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -29,7 +29,6 @@
 #include "utilities/macros.hpp"
 #include "runtime/rtmLocking.hpp"
 
-
 // MacroAssembler extends Assembler by frequently used macros.
 //
 // Instructions for which a 'better' code sequence exists depending
@@ -1212,32 +1211,50 @@
   // clear memory of size 'cnt' qwords, starting at 'base'.
   void clear_mem(Register base, Register cnt, Register rtmp);
 
+#ifdef COMPILER2
+  void string_indexof_char(Register str1, Register cnt1, Register ch, Register result,
+                           XMMRegister vec1, XMMRegister vec2, XMMRegister vec3, Register tmp);
+
   // IndexOf strings.
   // Small strings are loaded through stack if they cross page boundary.
   void string_indexof(Register str1, Register str2,
                       Register cnt1, Register cnt2,
                       int int_cnt2,  Register result,
-                      XMMRegister vec, Register tmp);
+                      XMMRegister vec, Register tmp,
+                      int ae);
 
   // IndexOf for constant substrings with size >= 8 elements
   // which don't need to be loaded through stack.
   void string_indexofC8(Register str1, Register str2,
                       Register cnt1, Register cnt2,
                       int int_cnt2,  Register result,
-                      XMMRegister vec, Register tmp);
+                      XMMRegister vec, Register tmp,
+                      int ae);
 
     // Smallest code: we don't need to load through stack,
     // check string tail.
 
+  // helper function for string_compare
+  void load_next_elements(Register elem1, Register elem2, Register str1, Register str2,
+                          Address::ScaleFactor scale, Address::ScaleFactor scale1,
+                          Address::ScaleFactor scale2, Register index, int ae);
   // Compare strings.
   void string_compare(Register str1, Register str2,
                       Register cnt1, Register cnt2, Register result,
-                      XMMRegister vec1);
+                      XMMRegister vec1, int ae);
 
-  // Compare char[] arrays.
-  void char_arrays_equals(bool is_array_equ, Register ary1, Register ary2,
-                          Register limit, Register result, Register chr,
-                          XMMRegister vec1, XMMRegister vec2);
+  // Search for Non-ASCII character (Negative byte value) in a byte array,
+  // return true if it has any and false otherwise.
+  void has_negatives(Register ary1, Register len,
+                     Register result, Register tmp1,
+                     XMMRegister vec1, XMMRegister vec2);
+
+  // Compare char[] or byte[] arrays.
+  void arrays_equals(bool is_array_equ, Register ary1, Register ary2,
+                     Register limit, Register result, Register chr,
+                     XMMRegister vec1, XMMRegister vec2, bool is_char);
+
+#endif
 
   // Fill primitive arrays
   void generate_fill(BasicType t, bool aligned,
@@ -1332,6 +1349,15 @@
   void fold_8bit_crc32(Register crc, Register table, Register tmp);
   void fold_8bit_crc32(XMMRegister crc, Register table, XMMRegister xtmp, Register tmp);
 
+  // Compress char[] array to byte[].
+  void char_array_compress(Register src, Register dst, Register len,
+                           XMMRegister tmp1, XMMRegister tmp2, XMMRegister tmp3,
+                           XMMRegister tmp4, Register tmp5, Register result);
+
+  // Inflate byte[] array to char[].
+  void byte_array_inflate(Register src, Register dst, Register len,
+                          XMMRegister tmp1, Register tmp2);
+
 #undef VIRTUAL
 
 };
--- a/hotspot/src/cpu/x86/vm/x86_32.ad	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/cpu/x86/vm/x86_32.ad	Tue Nov 03 09:41:03 2015 +0100
@@ -11435,16 +11435,62 @@
   ins_pipe( pipe_slow );
 %}
 
-instruct string_compare(eDIRegP str1, eCXRegI cnt1, eSIRegP str2, eDXRegI cnt2,
-                        eAXRegI result, regD tmp1, eFlagsReg cr) %{
+instruct string_compareL(eDIRegP str1, eCXRegI cnt1, eSIRegP str2, eDXRegI cnt2,
+                         eAXRegI result, regD tmp1, eFlagsReg cr) %{
+  predicate(((StrCompNode*)n)->encoding() == StrIntrinsicNode::LL);
   match(Set result (StrComp (Binary str1 cnt1) (Binary str2 cnt2)));
   effect(TEMP tmp1, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL cr);
 
-  format %{ "String Compare $str1,$cnt1,$str2,$cnt2 -> $result   // KILL $tmp1" %}
+  format %{ "String Compare byte[] $str1,$cnt1,$str2,$cnt2 -> $result   // KILL $tmp1" %}
+  ins_encode %{
+    __ string_compare($str1$$Register, $str2$$Register,
+                      $cnt1$$Register, $cnt2$$Register, $result$$Register,
+                      $tmp1$$XMMRegister, StrIntrinsicNode::LL);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct string_compareU(eDIRegP str1, eCXRegI cnt1, eSIRegP str2, eDXRegI cnt2,
+                         eAXRegI result, regD tmp1, eFlagsReg cr) %{
+  predicate(((StrCompNode*)n)->encoding() == StrIntrinsicNode::UU);
+  match(Set result (StrComp (Binary str1 cnt1) (Binary str2 cnt2)));
+  effect(TEMP tmp1, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL cr);
+
+  format %{ "String Compare char[] $str1,$cnt1,$str2,$cnt2 -> $result   // KILL $tmp1" %}
   ins_encode %{
     __ string_compare($str1$$Register, $str2$$Register,
                       $cnt1$$Register, $cnt2$$Register, $result$$Register,
-                      $tmp1$$XMMRegister);
+                      $tmp1$$XMMRegister, StrIntrinsicNode::UU);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct string_compareLU(eDIRegP str1, eCXRegI cnt1, eSIRegP str2, eDXRegI cnt2,
+                          eAXRegI result, regD tmp1, eFlagsReg cr) %{
+  predicate(((StrCompNode*)n)->encoding() == StrIntrinsicNode::LU);
+  match(Set result (StrComp (Binary str1 cnt1) (Binary str2 cnt2)));
+  effect(TEMP tmp1, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL cr);
+
+  format %{ "String Compare byte[] $str1,$cnt1,$str2,$cnt2 -> $result   // KILL $tmp1" %}
+  ins_encode %{
+    __ string_compare($str1$$Register, $str2$$Register,
+                      $cnt1$$Register, $cnt2$$Register, $result$$Register,
+                      $tmp1$$XMMRegister, StrIntrinsicNode::LU);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct string_compareUL(eSIRegP str1, eDXRegI cnt1, eDIRegP str2, eCXRegI cnt2,
+                          eAXRegI result, regD tmp1, eFlagsReg cr) %{
+  predicate(((StrCompNode*)n)->encoding() == StrIntrinsicNode::UL);
+  match(Set result (StrComp (Binary str1 cnt1) (Binary str2 cnt2)));
+  effect(TEMP tmp1, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL cr);
+
+  format %{ "String Compare byte[] $str1,$cnt1,$str2,$cnt2 -> $result   // KILL $tmp1" %}
+  ins_encode %{
+    __ string_compare($str2$$Register, $str1$$Register,
+                      $cnt2$$Register, $cnt1$$Register, $result$$Register,
+                      $tmp1$$XMMRegister, StrIntrinsicNode::UL);
   %}
   ins_pipe( pipe_slow );
 %}
@@ -11457,21 +11503,78 @@
 
   format %{ "String Equals $str1,$str2,$cnt -> $result    // KILL $tmp1, $tmp2, $tmp3" %}
   ins_encode %{
-    __ char_arrays_equals(false, $str1$$Register, $str2$$Register,
-                          $cnt$$Register, $result$$Register, $tmp3$$Register,
-                          $tmp1$$XMMRegister, $tmp2$$XMMRegister);
+    __ arrays_equals(false, $str1$$Register, $str2$$Register,
+                     $cnt$$Register, $result$$Register, $tmp3$$Register,
+                     $tmp1$$XMMRegister, $tmp2$$XMMRegister, false /* char */);
+  %} 
+
+  ins_pipe( pipe_slow );
+%}
+
+// fast search of substring with known size.
+instruct string_indexof_conL(eDIRegP str1, eDXRegI cnt1, eSIRegP str2, immI int_cnt2,
+                             eBXRegI result, regD vec, eAXRegI cnt2, eCXRegI tmp, eFlagsReg cr) %{
+  predicate(UseSSE42Intrinsics && (((StrIndexOfNode*)n)->encoding() == StrIntrinsicNode::LL));
+  match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 int_cnt2)));
+  effect(TEMP vec, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, KILL cnt2, KILL tmp, KILL cr);
+
+  format %{ "String IndexOf byte[] $str1,$cnt1,$str2,$int_cnt2 -> $result   // KILL $vec, $cnt1, $cnt2, $tmp" %}
+  ins_encode %{
+    int icnt2 = (int)$int_cnt2$$constant;
+    if (icnt2 >= 16) {
+      // IndexOf for constant substrings with size >= 16 elements
+      // which don't need to be loaded through stack.
+      __ string_indexofC8($str1$$Register, $str2$$Register,
+                          $cnt1$$Register, $cnt2$$Register,
+                          icnt2, $result$$Register,
+                          $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::LL);
+    } else {
+      // Small strings are loaded through stack if they cross page boundary.
+      __ string_indexof($str1$$Register, $str2$$Register,
+                        $cnt1$$Register, $cnt2$$Register,
+                        icnt2, $result$$Register,
+                        $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::LL);
+    }
   %}
   ins_pipe( pipe_slow );
 %}
 
 // fast search of substring with known size.
-instruct string_indexof_con(eDIRegP str1, eDXRegI cnt1, eSIRegP str2, immI int_cnt2,
-                            eBXRegI result, regD vec, eAXRegI cnt2, eCXRegI tmp, eFlagsReg cr) %{
-  predicate(UseSSE42Intrinsics);
+instruct string_indexof_conU(eDIRegP str1, eDXRegI cnt1, eSIRegP str2, immI int_cnt2,
+                             eBXRegI result, regD vec, eAXRegI cnt2, eCXRegI tmp, eFlagsReg cr) %{
+  predicate(UseSSE42Intrinsics && (((StrIndexOfNode*)n)->encoding() == StrIntrinsicNode::UU));
   match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 int_cnt2)));
   effect(TEMP vec, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, KILL cnt2, KILL tmp, KILL cr);
 
-  format %{ "String IndexOf $str1,$cnt1,$str2,$int_cnt2 -> $result   // KILL $vec, $cnt1, $cnt2, $tmp" %}
+  format %{ "String IndexOf char[] $str1,$cnt1,$str2,$int_cnt2 -> $result   // KILL $vec, $cnt1, $cnt2, $tmp" %}
+  ins_encode %{
+    int icnt2 = (int)$int_cnt2$$constant;
+    if (icnt2 >= 8) {
+      // IndexOf for constant substrings with size >= 8 elements
+      // which don't need to be loaded through stack.
+      __ string_indexofC8($str1$$Register, $str2$$Register,
+                          $cnt1$$Register, $cnt2$$Register,
+                          icnt2, $result$$Register,
+                          $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::UU);
+    } else {
+      // Small strings are loaded through stack if they cross page boundary.
+      __ string_indexof($str1$$Register, $str2$$Register,
+                        $cnt1$$Register, $cnt2$$Register,
+                        icnt2, $result$$Register,
+                        $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::UU);
+    }
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+// fast search of substring with known size.
+instruct string_indexof_conUL(eDIRegP str1, eDXRegI cnt1, eSIRegP str2, immI int_cnt2,
+                             eBXRegI result, regD vec, eAXRegI cnt2, eCXRegI tmp, eFlagsReg cr) %{
+  predicate(UseSSE42Intrinsics && (((StrIndexOfNode*)n)->encoding() == StrIntrinsicNode::UL));
+  match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 int_cnt2)));
+  effect(TEMP vec, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, KILL cnt2, KILL tmp, KILL cr);
+
+  format %{ "String IndexOf char[] $str1,$cnt1,$str2,$int_cnt2 -> $result   // KILL $vec, $cnt1, $cnt2, $tmp" %}
   ins_encode %{
     int icnt2 = (int)$int_cnt2$$constant;
     if (icnt2 >= 8) {
@@ -11480,47 +11583,154 @@
       __ string_indexofC8($str1$$Register, $str2$$Register,
                           $cnt1$$Register, $cnt2$$Register,
                           icnt2, $result$$Register,
-                          $vec$$XMMRegister, $tmp$$Register);
+                          $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::UL);
     } else {
       // Small strings are loaded through stack if they cross page boundary.
       __ string_indexof($str1$$Register, $str2$$Register,
                         $cnt1$$Register, $cnt2$$Register,
                         icnt2, $result$$Register,
-                        $vec$$XMMRegister, $tmp$$Register);
+                        $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::UL);
     }
   %}
   ins_pipe( pipe_slow );
 %}
 
-instruct string_indexof(eDIRegP str1, eDXRegI cnt1, eSIRegP str2, eAXRegI cnt2,
-                        eBXRegI result, regD vec, eCXRegI tmp, eFlagsReg cr) %{
-  predicate(UseSSE42Intrinsics);
+instruct string_indexofL(eDIRegP str1, eDXRegI cnt1, eSIRegP str2, eAXRegI cnt2,
+                         eBXRegI result, regD vec, eCXRegI tmp, eFlagsReg cr) %{
+  predicate(UseSSE42Intrinsics && (((StrIndexOfNode*)n)->encoding() == StrIntrinsicNode::LL));
   match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 cnt2)));
   effect(TEMP vec, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL tmp, KILL cr);
 
-  format %{ "String IndexOf $str1,$cnt1,$str2,$cnt2 -> $result   // KILL all" %}
+  format %{ "String IndexOf byte[] $str1,$cnt1,$str2,$cnt2 -> $result   // KILL all" %}
   ins_encode %{
     __ string_indexof($str1$$Register, $str2$$Register,
                       $cnt1$$Register, $cnt2$$Register,
                       (-1), $result$$Register,
-                      $vec$$XMMRegister, $tmp$$Register);
+                      $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::LL);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct string_indexofU(eDIRegP str1, eDXRegI cnt1, eSIRegP str2, eAXRegI cnt2,
+                         eBXRegI result, regD vec, eCXRegI tmp, eFlagsReg cr) %{
+  predicate(UseSSE42Intrinsics && (((StrIndexOfNode*)n)->encoding() == StrIntrinsicNode::UU));
+  match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 cnt2)));
+  effect(TEMP vec, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL tmp, KILL cr);
+
+  format %{ "String IndexOf char[] $str1,$cnt1,$str2,$cnt2 -> $result   // KILL all" %}
+  ins_encode %{
+    __ string_indexof($str1$$Register, $str2$$Register,
+                      $cnt1$$Register, $cnt2$$Register,
+                      (-1), $result$$Register,
+                      $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::UU);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct string_indexofUL(eDIRegP str1, eDXRegI cnt1, eSIRegP str2, eAXRegI cnt2,
+                         eBXRegI result, regD vec, eCXRegI tmp, eFlagsReg cr) %{
+  predicate(UseSSE42Intrinsics && (((StrIndexOfNode*)n)->encoding() == StrIntrinsicNode::UL));
+  match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 cnt2)));
+  effect(TEMP vec, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL tmp, KILL cr);
+
+  format %{ "String IndexOf char[] $str1,$cnt1,$str2,$cnt2 -> $result   // KILL all" %}
+  ins_encode %{
+    __ string_indexof($str1$$Register, $str2$$Register,
+                      $cnt1$$Register, $cnt2$$Register,
+                      (-1), $result$$Register,
+                      $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::UL);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct string_indexofU_char(eDIRegP str1, eDXRegI cnt1, eAXRegI ch,
+                              eBXRegI result, regD vec1, regD vec2, regD vec3, eCXRegI tmp, eFlagsReg cr) %{
+  predicate(UseSSE42Intrinsics);
+  match(Set result (StrIndexOfChar (Binary str1 cnt1) ch));
+  effect(TEMP vec1, TEMP vec2, TEMP vec3, USE_KILL str1, USE_KILL cnt1, USE_KILL ch, TEMP tmp, KILL cr);
+  format %{ "String IndexOf char[] $str1,$cnt1,$ch -> $result   // KILL all" %}
+  ins_encode %{
+    __ string_indexof_char($str1$$Register, $cnt1$$Register, $ch$$Register, $result$$Register,
+                           $vec1$$XMMRegister, $vec2$$XMMRegister, $vec3$$XMMRegister, $tmp$$Register);
   %}
   ins_pipe( pipe_slow );
 %}
 
 // fast array equals
-instruct array_equals(eDIRegP ary1, eSIRegP ary2, eAXRegI result,
-                      regD tmp1, regD tmp2, eCXRegI tmp3, eBXRegI tmp4, eFlagsReg cr)
+instruct array_equalsB(eDIRegP ary1, eSIRegP ary2, eAXRegI result,
+                       regD tmp1, regD tmp2, eCXRegI tmp3, eBXRegI tmp4, eFlagsReg cr)
 %{
+  predicate(((AryEqNode*)n)->encoding() == StrIntrinsicNode::LL);
+  match(Set result (AryEq ary1 ary2));
+  effect(TEMP tmp1, TEMP tmp2, USE_KILL ary1, USE_KILL ary2, KILL tmp3, KILL tmp4, KILL cr);
+  //ins_cost(300);
+
+  format %{ "Array Equals byte[] $ary1,$ary2 -> $result   // KILL $tmp1, $tmp2, $tmp3, $tmp4" %}
+  ins_encode %{
+    __ arrays_equals(true, $ary1$$Register, $ary2$$Register,
+                     $tmp3$$Register, $result$$Register, $tmp4$$Register,
+                     $tmp1$$XMMRegister, $tmp2$$XMMRegister, false /* char */);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct array_equalsC(eDIRegP ary1, eSIRegP ary2, eAXRegI result,
+                       regD tmp1, regD tmp2, eCXRegI tmp3, eBXRegI tmp4, eFlagsReg cr)
+%{
+  predicate(((AryEqNode*)n)->encoding() == StrIntrinsicNode::UU);
   match(Set result (AryEq ary1 ary2));
   effect(TEMP tmp1, TEMP tmp2, USE_KILL ary1, USE_KILL ary2, KILL tmp3, KILL tmp4, KILL cr);
   //ins_cost(300);
 
-  format %{ "Array Equals $ary1,$ary2 -> $result   // KILL $tmp1, $tmp2, $tmp3, $tmp4" %}
-  ins_encode %{
-    __ char_arrays_equals(true, $ary1$$Register, $ary2$$Register,
-                          $tmp3$$Register, $result$$Register, $tmp4$$Register,
-                          $tmp1$$XMMRegister, $tmp2$$XMMRegister);
+  format %{ "Array Equals char[] $ary1,$ary2 -> $result   // KILL $tmp1, $tmp2, $tmp3, $tmp4" %}
+  ins_encode %{
+    __ arrays_equals(true, $ary1$$Register, $ary2$$Register,
+                     $tmp3$$Register, $result$$Register, $tmp4$$Register,
+                     $tmp1$$XMMRegister, $tmp2$$XMMRegister, true /* char */);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct has_negatives(eSIRegP ary1, eCXRegI len, eAXRegI result,
+                      regD tmp1, regD tmp2, eBXRegI tmp3, eFlagsReg cr)
+%{
+  match(Set result (HasNegatives ary1 len));
+  effect(TEMP tmp1, TEMP tmp2, USE_KILL ary1, USE_KILL len, KILL tmp3, KILL cr);
+
+  format %{ "has negatives byte[] $ary1,$len -> $result   // KILL $tmp1, $tmp2, $tmp3" %}
+  ins_encode %{
+    __ has_negatives($ary1$$Register, $len$$Register,
+                     $result$$Register, $tmp3$$Register,
+                     $tmp1$$XMMRegister, $tmp2$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+// fast char[] to byte[] compression
+instruct string_compress(eSIRegP src, eDIRegP dst, eDXRegI len, regD tmp1, regD tmp2, regD tmp3, regD tmp4,
+                         eCXRegI tmp5, eAXRegI result, eFlagsReg cr) %{
+  match(Set result (StrCompressedCopy src (Binary dst len)));
+  effect(TEMP tmp1, TEMP tmp2, TEMP tmp3, TEMP tmp4, USE_KILL src, USE_KILL dst, USE_KILL len, KILL tmp5, KILL cr);
+
+  format %{ "String Compress $src,$dst -> $result    // KILL RAX, RCX, RDX" %}
+  ins_encode %{
+    __ char_array_compress($src$$Register, $dst$$Register, $len$$Register,
+                           $tmp1$$XMMRegister, $tmp2$$XMMRegister, $tmp3$$XMMRegister,
+                           $tmp4$$XMMRegister, $tmp5$$Register, $result$$Register);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+// fast byte[] to char[] inflation
+instruct string_inflate(Universe dummy, eSIRegP src, eDIRegP dst, eDXRegI len,
+                        regD tmp1, eCXRegI tmp2, eFlagsReg cr) %{
+  match(Set dummy (StrInflatedCopy src (Binary dst len)));
+  effect(TEMP tmp1, TEMP tmp2, USE_KILL src, USE_KILL dst, USE_KILL len, KILL cr);
+
+  format %{ "String Inflate $src,$dst    // KILL $tmp1, $tmp2" %}
+  ins_encode %{
+    __ byte_array_inflate($src$$Register, $dst$$Register, $len$$Register,
+                          $tmp1$$XMMRegister, $tmp2$$Register);
   %}
   ins_pipe( pipe_slow );
 %}
--- a/hotspot/src/cpu/x86/vm/x86_64.ad	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/cpu/x86/vm/x86_64.ad	Tue Nov 03 09:41:03 2015 +0100
@@ -10447,30 +10447,108 @@
   ins_pipe( pipe_slow );
 %}
 
-instruct string_compare(rdi_RegP str1, rcx_RegI cnt1, rsi_RegP str2, rdx_RegI cnt2,
-                        rax_RegI result, regD tmp1, rFlagsReg cr)
-%{
+instruct string_compareL(rdi_RegP str1, rcx_RegI cnt1, rsi_RegP str2, rdx_RegI cnt2,
+                         rax_RegI result, regD tmp1, rFlagsReg cr)
+%{
+  predicate(((StrCompNode*)n)->encoding() == StrIntrinsicNode::LL);
   match(Set result (StrComp (Binary str1 cnt1) (Binary str2 cnt2)));
   effect(TEMP tmp1, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL cr);
 
-  format %{ "String Compare $str1,$cnt1,$str2,$cnt2 -> $result   // KILL $tmp1" %}
+  format %{ "String Compare byte[] $str1,$cnt1,$str2,$cnt2 -> $result   // KILL $tmp1" %}
+  ins_encode %{
+    __ string_compare($str1$$Register, $str2$$Register,
+                      $cnt1$$Register, $cnt2$$Register, $result$$Register,
+                      $tmp1$$XMMRegister, StrIntrinsicNode::LL);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct string_compareU(rdi_RegP str1, rcx_RegI cnt1, rsi_RegP str2, rdx_RegI cnt2,
+                         rax_RegI result, regD tmp1, rFlagsReg cr)
+%{
+  predicate(((StrCompNode*)n)->encoding() == StrIntrinsicNode::UU);
+  match(Set result (StrComp (Binary str1 cnt1) (Binary str2 cnt2)));
+  effect(TEMP tmp1, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL cr);
+
+  format %{ "String Compare char[] $str1,$cnt1,$str2,$cnt2 -> $result   // KILL $tmp1" %}
   ins_encode %{
     __ string_compare($str1$$Register, $str2$$Register,
                       $cnt1$$Register, $cnt2$$Register, $result$$Register,
-                      $tmp1$$XMMRegister);
+                      $tmp1$$XMMRegister, StrIntrinsicNode::UU);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct string_compareLU(rdi_RegP str1, rcx_RegI cnt1, rsi_RegP str2, rdx_RegI cnt2,
+                          rax_RegI result, regD tmp1, rFlagsReg cr)
+%{
+  predicate(((StrCompNode*)n)->encoding() == StrIntrinsicNode::LU);
+  match(Set result (StrComp (Binary str1 cnt1) (Binary str2 cnt2)));
+  effect(TEMP tmp1, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL cr);
+
+  format %{ "String Compare byte[] $str1,$cnt1,$str2,$cnt2 -> $result   // KILL $tmp1" %}
+  ins_encode %{
+    __ string_compare($str1$$Register, $str2$$Register,
+                      $cnt1$$Register, $cnt2$$Register, $result$$Register,
+                      $tmp1$$XMMRegister, StrIntrinsicNode::LU);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct string_compareUL(rsi_RegP str1, rdx_RegI cnt1, rdi_RegP str2, rcx_RegI cnt2,
+                          rax_RegI result, regD tmp1, rFlagsReg cr)
+%{
+  predicate(((StrCompNode*)n)->encoding() == StrIntrinsicNode::UL);
+  match(Set result (StrComp (Binary str1 cnt1) (Binary str2 cnt2)));
+  effect(TEMP tmp1, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL cr);
+
+  format %{ "String Compare byte[] $str1,$cnt1,$str2,$cnt2 -> $result   // KILL $tmp1" %}
+  ins_encode %{
+    __ string_compare($str2$$Register, $str1$$Register,
+                      $cnt2$$Register, $cnt1$$Register, $result$$Register,
+                      $tmp1$$XMMRegister, StrIntrinsicNode::UL);
   %}
   ins_pipe( pipe_slow );
 %}
 
 // fast search of substring with known size.
-instruct string_indexof_con(rdi_RegP str1, rdx_RegI cnt1, rsi_RegP str2, immI int_cnt2,
-                            rbx_RegI result, regD vec, rax_RegI cnt2, rcx_RegI tmp, rFlagsReg cr)
-%{
-  predicate(UseSSE42Intrinsics);
+instruct string_indexof_conL(rdi_RegP str1, rdx_RegI cnt1, rsi_RegP str2, immI int_cnt2,
+                             rbx_RegI result, regD vec, rax_RegI cnt2, rcx_RegI tmp, rFlagsReg cr)
+%{
+  predicate(UseSSE42Intrinsics && (((StrIndexOfNode*)n)->encoding() == StrIntrinsicNode::LL));
   match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 int_cnt2)));
   effect(TEMP vec, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, KILL cnt2, KILL tmp, KILL cr);
 
-  format %{ "String IndexOf $str1,$cnt1,$str2,$int_cnt2 -> $result   // KILL $vec, $cnt1, $cnt2, $tmp" %}
+  format %{ "String IndexOf byte[] $str1,$cnt1,$str2,$int_cnt2 -> $result   // KILL $vec, $cnt1, $cnt2, $tmp" %}
+  ins_encode %{
+    int icnt2 = (int)$int_cnt2$$constant;
+    if (icnt2 >= 16) {
+      // IndexOf for constant substrings with size >= 16 elements
+      // which don't need to be loaded through stack.
+      __ string_indexofC8($str1$$Register, $str2$$Register,
+                          $cnt1$$Register, $cnt2$$Register,
+                          icnt2, $result$$Register,
+                          $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::LL);
+    } else {
+      // Small strings are loaded through stack if they cross page boundary.
+      __ string_indexof($str1$$Register, $str2$$Register,
+                        $cnt1$$Register, $cnt2$$Register,
+                        icnt2, $result$$Register,
+                        $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::LL);
+    }
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+// fast search of substring with known size.
+instruct string_indexof_conU(rdi_RegP str1, rdx_RegI cnt1, rsi_RegP str2, immI int_cnt2,
+                             rbx_RegI result, regD vec, rax_RegI cnt2, rcx_RegI tmp, rFlagsReg cr)
+%{
+  predicate(UseSSE42Intrinsics && (((StrIndexOfNode*)n)->encoding() == StrIntrinsicNode::UU));
+  match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 int_cnt2)));
+  effect(TEMP vec, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, KILL cnt2, KILL tmp, KILL cr);
+
+  format %{ "String IndexOf char[] $str1,$cnt1,$str2,$int_cnt2 -> $result   // KILL $vec, $cnt1, $cnt2, $tmp" %}
   ins_encode %{
     int icnt2 = (int)$int_cnt2$$constant;
     if (icnt2 >= 8) {
@@ -10479,31 +10557,108 @@
       __ string_indexofC8($str1$$Register, $str2$$Register,
                           $cnt1$$Register, $cnt2$$Register,
                           icnt2, $result$$Register,
-                          $vec$$XMMRegister, $tmp$$Register);
+                          $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::UU);
+    } else {
+      // Small strings are loaded through stack if they cross page boundary.
+      __ string_indexof($str1$$Register, $str2$$Register,
+                        $cnt1$$Register, $cnt2$$Register,
+                        icnt2, $result$$Register,
+                        $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::UU);
+    }
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+// fast search of substring with known size.
+instruct string_indexof_conUL(rdi_RegP str1, rdx_RegI cnt1, rsi_RegP str2, immI int_cnt2,
+                             rbx_RegI result, regD vec, rax_RegI cnt2, rcx_RegI tmp, rFlagsReg cr)
+%{
+  predicate(UseSSE42Intrinsics && (((StrIndexOfNode*)n)->encoding() == StrIntrinsicNode::UL));
+  match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 int_cnt2)));
+  effect(TEMP vec, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, KILL cnt2, KILL tmp, KILL cr);
+
+  format %{ "String IndexOf char[] $str1,$cnt1,$str2,$int_cnt2 -> $result   // KILL $vec, $cnt1, $cnt2, $tmp" %}
+  ins_encode %{
+    int icnt2 = (int)$int_cnt2$$constant;
+    if (icnt2 >= 8) {
+      // IndexOf for constant substrings with size >= 8 elements
+      // which don't need to be loaded through stack.
+      __ string_indexofC8($str1$$Register, $str2$$Register,
+                          $cnt1$$Register, $cnt2$$Register,
+                          icnt2, $result$$Register,
+                          $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::UL);
     } else {
       // Small strings are loaded through stack if they cross page boundary.
       __ string_indexof($str1$$Register, $str2$$Register,
                         $cnt1$$Register, $cnt2$$Register,
                         icnt2, $result$$Register,
-                        $vec$$XMMRegister, $tmp$$Register);
+                        $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::UL);
     }
   %}
   ins_pipe( pipe_slow );
 %}
 
-instruct string_indexof(rdi_RegP str1, rdx_RegI cnt1, rsi_RegP str2, rax_RegI cnt2,
-                        rbx_RegI result, regD vec, rcx_RegI tmp, rFlagsReg cr)
-%{
-  predicate(UseSSE42Intrinsics);
+instruct string_indexofL(rdi_RegP str1, rdx_RegI cnt1, rsi_RegP str2, rax_RegI cnt2,
+                         rbx_RegI result, regD vec, rcx_RegI tmp, rFlagsReg cr)
+%{
+  predicate(UseSSE42Intrinsics && (((StrIndexOfNode*)n)->encoding() == StrIntrinsicNode::LL));
   match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 cnt2)));
   effect(TEMP vec, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL tmp, KILL cr);
 
-  format %{ "String IndexOf $str1,$cnt1,$str2,$cnt2 -> $result   // KILL all" %}
+  format %{ "String IndexOf byte[] $str1,$cnt1,$str2,$cnt2 -> $result   // KILL all" %}
   ins_encode %{
     __ string_indexof($str1$$Register, $str2$$Register,
                       $cnt1$$Register, $cnt2$$Register,
                       (-1), $result$$Register,
-                      $vec$$XMMRegister, $tmp$$Register);
+                      $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::LL);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct string_indexofU(rdi_RegP str1, rdx_RegI cnt1, rsi_RegP str2, rax_RegI cnt2,
+                         rbx_RegI result, regD vec, rcx_RegI tmp, rFlagsReg cr)
+%{
+  predicate(UseSSE42Intrinsics && (((StrIndexOfNode*)n)->encoding() == StrIntrinsicNode::UU));
+  match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 cnt2)));
+  effect(TEMP vec, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL tmp, KILL cr);
+
+  format %{ "String IndexOf char[] $str1,$cnt1,$str2,$cnt2 -> $result   // KILL all" %}
+  ins_encode %{
+    __ string_indexof($str1$$Register, $str2$$Register,
+                      $cnt1$$Register, $cnt2$$Register,
+                      (-1), $result$$Register,
+                      $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::UU);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct string_indexofUL(rdi_RegP str1, rdx_RegI cnt1, rsi_RegP str2, rax_RegI cnt2,
+                         rbx_RegI result, regD vec, rcx_RegI tmp, rFlagsReg cr)
+%{
+  predicate(UseSSE42Intrinsics && (((StrIndexOfNode*)n)->encoding() == StrIntrinsicNode::UL));
+  match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 cnt2)));
+  effect(TEMP vec, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL tmp, KILL cr);
+
+  format %{ "String IndexOf char[] $str1,$cnt1,$str2,$cnt2 -> $result   // KILL all" %}
+  ins_encode %{
+    __ string_indexof($str1$$Register, $str2$$Register,
+                      $cnt1$$Register, $cnt2$$Register,
+                      (-1), $result$$Register,
+                      $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::UL);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct string_indexofU_char(rdi_RegP str1, rdx_RegI cnt1, rax_RegI ch,
+                              rbx_RegI result, regD vec1, regD vec2, regD vec3, rcx_RegI tmp, rFlagsReg cr)
+%{
+  predicate(UseSSE42Intrinsics);
+  match(Set result (StrIndexOfChar (Binary str1 cnt1) ch));
+  effect(TEMP vec1, TEMP vec2, TEMP vec3, USE_KILL str1, USE_KILL cnt1, USE_KILL ch, TEMP tmp, KILL cr);
+  format %{ "String IndexOf char[] $str1,$cnt1,$ch -> $result   // KILL all" %}
+  ins_encode %{
+    __ string_indexof_char($str1$$Register, $cnt1$$Register, $ch$$Register, $result$$Register,
+                           $vec1$$XMMRegister, $vec2$$XMMRegister, $vec3$$XMMRegister, $tmp$$Register);
   %}
   ins_pipe( pipe_slow );
 %}
@@ -10517,26 +10672,86 @@
 
   format %{ "String Equals $str1,$str2,$cnt -> $result    // KILL $tmp1, $tmp2, $tmp3" %}
   ins_encode %{
-    __ char_arrays_equals(false, $str1$$Register, $str2$$Register,
-                          $cnt$$Register, $result$$Register, $tmp3$$Register,
-                          $tmp1$$XMMRegister, $tmp2$$XMMRegister);
+    __ arrays_equals(false, $str1$$Register, $str2$$Register,
+                     $cnt$$Register, $result$$Register, $tmp3$$Register,
+                     $tmp1$$XMMRegister, $tmp2$$XMMRegister, false /* char */);
   %}
   ins_pipe( pipe_slow );
 %}
 
 // fast array equals
-instruct array_equals(rdi_RegP ary1, rsi_RegP ary2, rax_RegI result,
+instruct array_equalsB(rdi_RegP ary1, rsi_RegP ary2, rax_RegI result,
+                       regD tmp1, regD tmp2, rcx_RegI tmp3, rbx_RegI tmp4, rFlagsReg cr)
+%{
+  predicate(((AryEqNode*)n)->encoding() == StrIntrinsicNode::LL);
+  match(Set result (AryEq ary1 ary2));
+  effect(TEMP tmp1, TEMP tmp2, USE_KILL ary1, USE_KILL ary2, KILL tmp3, KILL tmp4, KILL cr);
+
+  format %{ "Array Equals byte[] $ary1,$ary2 -> $result   // KILL $tmp1, $tmp2, $tmp3, $tmp4" %}
+  ins_encode %{
+    __ arrays_equals(true, $ary1$$Register, $ary2$$Register,
+                     $tmp3$$Register, $result$$Register, $tmp4$$Register,
+                     $tmp1$$XMMRegister, $tmp2$$XMMRegister, false /* char */);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct array_equalsC(rdi_RegP ary1, rsi_RegP ary2, rax_RegI result,
                       regD tmp1, regD tmp2, rcx_RegI tmp3, rbx_RegI tmp4, rFlagsReg cr)
 %{
+  predicate(((AryEqNode*)n)->encoding() == StrIntrinsicNode::UU);
   match(Set result (AryEq ary1 ary2));
   effect(TEMP tmp1, TEMP tmp2, USE_KILL ary1, USE_KILL ary2, KILL tmp3, KILL tmp4, KILL cr);
-  //ins_cost(300);
-
-  format %{ "Array Equals $ary1,$ary2 -> $result   // KILL $tmp1, $tmp2, $tmp3, $tmp4" %}
-  ins_encode %{
-    __ char_arrays_equals(true, $ary1$$Register, $ary2$$Register,
-                          $tmp3$$Register, $result$$Register, $tmp4$$Register,
-                          $tmp1$$XMMRegister, $tmp2$$XMMRegister);
+
+  format %{ "Array Equals char[] $ary1,$ary2 -> $result   // KILL $tmp1, $tmp2, $tmp3, $tmp4" %}
+  ins_encode %{
+    __ arrays_equals(true, $ary1$$Register, $ary2$$Register,
+                     $tmp3$$Register, $result$$Register, $tmp4$$Register,
+                     $tmp1$$XMMRegister, $tmp2$$XMMRegister, true /* char */);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct has_negatives(rsi_RegP ary1, rcx_RegI len, rax_RegI result,
+                      regD tmp1, regD tmp2, rbx_RegI tmp3, rFlagsReg cr)
+%{
+  match(Set result (HasNegatives ary1 len));
+  effect(TEMP tmp1, TEMP tmp2, USE_KILL ary1, USE_KILL len, KILL tmp3, KILL cr);
+
+  format %{ "has negatives byte[] $ary1,$len -> $result   // KILL $tmp1, $tmp2, $tmp3" %}
+  ins_encode %{
+    __ has_negatives($ary1$$Register, $len$$Register,
+                     $result$$Register, $tmp3$$Register,
+                     $tmp1$$XMMRegister, $tmp2$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+// fast char[] to byte[] compression
+instruct string_compress(rsi_RegP src, rdi_RegP dst, rdx_RegI len, regD tmp1, regD tmp2, regD tmp3, regD tmp4,
+                         rcx_RegI tmp5, rax_RegI result, rFlagsReg cr) %{
+  match(Set result (StrCompressedCopy src (Binary dst len)));
+  effect(TEMP tmp1, TEMP tmp2, TEMP tmp3, TEMP tmp4, USE_KILL src, USE_KILL dst, USE_KILL len, KILL tmp5, KILL cr);
+
+  format %{ "String Compress $src,$dst -> $result    // KILL RAX, RCX, RDX" %}
+  ins_encode %{
+    __ char_array_compress($src$$Register, $dst$$Register, $len$$Register,
+                           $tmp1$$XMMRegister, $tmp2$$XMMRegister, $tmp3$$XMMRegister,
+                           $tmp4$$XMMRegister, $tmp5$$Register, $result$$Register);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+// fast byte[] to char[] inflation
+instruct string_inflate(Universe dummy, rsi_RegP src, rdi_RegP dst, rdx_RegI len,
+                        regD tmp1, rcx_RegI tmp2, rFlagsReg cr) %{
+  match(Set dummy (StrInflatedCopy src (Binary dst len)));
+  effect(TEMP tmp1, TEMP tmp2, USE_KILL src, USE_KILL dst, USE_KILL len, KILL cr);
+
+  format %{ "String Inflate $src,$dst    // KILL $tmp1, $tmp2" %}
+  ins_encode %{
+    __ byte_array_inflate($src$$Register, $dst$$Register, $len$$Register,
+                          $tmp1$$XMMRegister, $tmp2$$Register);
   %}
   ins_pipe( pipe_slow );
 %}
--- a/hotspot/src/cpu/zero/vm/globals_zero.hpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/cpu/zero/vm/globals_zero.hpp	Tue Nov 03 09:41:03 2015 +0100
@@ -69,6 +69,9 @@
 
 define_pd_global(bool, PreserveFramePointer, false);
 
+// No performance work done here yet.
+define_pd_global(bool, CompactStrings, false);
+
 #define ARCH_FLAGS(develop, product, diagnostic, experimental, notproduct, range, constraint)  \
                                                                             \
   product(bool, UseFastEmptyMethods, true,                                  \
--- a/hotspot/src/share/vm/adlc/formssel.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/adlc/formssel.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -607,6 +607,8 @@
       ( strcmp(_matrule->_rChild->_opType,"StrComp"    )==0 ||
         strcmp(_matrule->_rChild->_opType,"StrEquals"  )==0 ||
         strcmp(_matrule->_rChild->_opType,"StrIndexOf" )==0 ||
+        strcmp(_matrule->_rChild->_opType,"StrIndexOfChar" )==0 ||
+        strcmp(_matrule->_rChild->_opType,"HasNegatives" )==0 ||
         strcmp(_matrule->_rChild->_opType,"AryEq"      )==0 ))
     return true;
 
@@ -887,11 +889,16 @@
       ( strcmp(_matrule->_rChild->_opType,"AryEq"     )==0 ||
         strcmp(_matrule->_rChild->_opType,"StrComp"   )==0 ||
         strcmp(_matrule->_rChild->_opType,"StrEquals" )==0 ||
+        strcmp(_matrule->_rChild->_opType,"StrInflatedCopy"   )==0 ||
+        strcmp(_matrule->_rChild->_opType,"StrCompressedCopy" )==0 ||
         strcmp(_matrule->_rChild->_opType,"StrIndexOf")==0 ||
+        strcmp(_matrule->_rChild->_opType,"StrIndexOfChar")==0 ||
+        strcmp(_matrule->_rChild->_opType,"HasNegatives")==0 ||
         strcmp(_matrule->_rChild->_opType,"EncodeISOArray")==0)) {
         // String.(compareTo/equals/indexOf) and Arrays.equals
         // and sun.nio.cs.iso8859_1$Encoder.EncodeISOArray
         // take 1 control and 1 memory edges.
+        // Also String.(compressedCopy/inflatedCopy).
     return 2;
   }
 
--- a/hotspot/src/share/vm/adlc/main.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/adlc/main.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -221,6 +221,7 @@
   AD.addInclude(AD._CPP_file, "oops/method.hpp");
   AD.addInclude(AD._CPP_file, "oops/oop.inline.hpp");
   AD.addInclude(AD._CPP_file, "opto/cfgnode.hpp");
+  AD.addInclude(AD._CPP_file, "opto/intrinsicnode.hpp");
   AD.addInclude(AD._CPP_file, "opto/locknode.hpp");
   AD.addInclude(AD._CPP_file, "opto/opcodes.hpp");
   AD.addInclude(AD._CPP_file, "opto/regalloc.hpp");
@@ -256,6 +257,7 @@
   AD.addInclude(AD._DFA_file, "precompiled.hpp");
   AD.addInclude(AD._DFA_file, "adfiles", get_basename(AD._HPP_file._name));
   AD.addInclude(AD._DFA_file, "opto/cfgnode.hpp");  // Use PROB_MAX in predicate.
+  AD.addInclude(AD._DFA_file, "opto/intrinsicnode.hpp");
   AD.addInclude(AD._DFA_file, "opto/matcher.hpp");
   AD.addInclude(AD._DFA_file, "opto/opcodes.hpp");
   AD.addInclude(AD._DFA_file, "opto/convertnode.hpp");
--- a/hotspot/src/share/vm/c1/c1_Compiler.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/c1/c1_Compiler.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -225,6 +225,8 @@
   case vmIntrinsics::_updateByteBufferCRC32:
   case vmIntrinsics::_compareAndSwapInt:
   case vmIntrinsics::_compareAndSwapObject:
+  case vmIntrinsics::_getCharStringU:
+  case vmIntrinsics::_putCharStringU:
 #ifdef TRACE_HAVE_INTRINSICS
   case vmIntrinsics::_classID:
   case vmIntrinsics::_threadID:
--- a/hotspot/src/share/vm/c1/c1_GraphBuilder.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/c1/c1_GraphBuilder.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -3445,6 +3445,8 @@
   case vmIntrinsics::_getAndSetInt       :
   case vmIntrinsics::_getAndSetLong      :
   case vmIntrinsics::_getAndSetObject    : append_unsafe_get_and_set_obj(callee, false); return;
+  case vmIntrinsics::_getCharStringU     : append_char_access(callee, false); return;
+  case vmIntrinsics::_putCharStringU     : append_char_access(callee, true); return;
   default:
     break;
   }
@@ -4179,6 +4181,30 @@
   compilation()->set_has_unsafe_access(true);
 }
 
+void GraphBuilder::append_char_access(ciMethod* callee, bool is_store) {
+  // This intrinsic accesses byte[] array as char[] array. Computing the offsets
+  // correctly requires matched array shapes.
+  assert (arrayOopDesc::base_offset_in_bytes(T_CHAR) == arrayOopDesc::base_offset_in_bytes(T_BYTE),
+          "sanity: byte[] and char[] bases agree");
+  assert (type2aelembytes(T_CHAR) == type2aelembytes(T_BYTE)*2,
+          "sanity: byte[] and char[] scales agree");
+
+  ValueStack* state_before = copy_state_indexed_access();
+  compilation()->set_has_access_indexed(true);
+  Values* args = state()->pop_arguments(callee->arg_size());
+  Value array = args->at(0);
+  Value index = args->at(1);
+  if (is_store) {
+    Value value = args->at(2);
+    Instruction* store = append(new StoreIndexed(array, index, NULL, T_CHAR, value, state_before));
+    store->set_flag(Instruction::NeedsRangeCheckFlag, false);
+    _memory->store_value(value);
+  } else {
+    Instruction* load = append(new LoadIndexed(array, index, NULL, T_CHAR, state_before));
+    load->set_flag(Instruction::NeedsRangeCheckFlag, false);
+    push(load->type(), load);
+  }
+}
 
 void GraphBuilder::print_inlining(ciMethod* callee, const char* msg, bool success) {
   CompileLog* log = compilation()->log();
--- a/hotspot/src/share/vm/c1/c1_GraphBuilder.hpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/c1/c1_GraphBuilder.hpp	Tue Nov 03 09:41:03 2015 +0100
@@ -372,6 +372,7 @@
   void append_unsafe_put_raw(ciMethod* callee, BasicType t);
   void append_unsafe_CAS(ciMethod* callee);
   void append_unsafe_get_and_set_obj(ciMethod* callee, bool is_add);
+  void append_char_access(ciMethod* callee, bool is_store);
 
   void print_inlining(ciMethod* callee, const char* msg = NULL, bool success = true);
 
--- a/hotspot/src/share/vm/c1/c1_LIRAssembler.hpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/c1/c1_LIRAssembler.hpp	Tue Nov 03 09:41:03 2015 +0100
@@ -161,8 +161,6 @@
   // particular sparc uses this for delay slot filling.
   void peephole(LIR_List* list);
 
-  void emit_string_compare(LIR_Opr left, LIR_Opr right, LIR_Opr dst, CodeEmitInfo* info);
-
   void return_op(LIR_Opr result);
 
   // returns offset of poll instruction
--- a/hotspot/src/share/vm/ci/ciTypeArray.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/ci/ciTypeArray.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2005, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -46,3 +46,13 @@
 #endif //ASSERT
   return c;
 }
+
+// ------------------------------------------------------------------
+// ciTypeArray::byte_at
+//
+// Implementation of the byte_at method.
+jbyte ciTypeArray::byte_at(int index) {
+  VM_ENTRY_MARK;
+  assert(index >= 0 && index < length(), "out of range");
+  return get_typeArrayOop()->byte_at(index);
+}
--- a/hotspot/src/share/vm/ci/ciTypeArray.hpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/ci/ciTypeArray.hpp	Tue Nov 03 09:41:03 2015 +0100
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1999, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1999, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -56,6 +56,9 @@
   // array will never change.
   jchar char_at(int index);
 
+  // Return byte at index.
+  jbyte byte_at(int index);
+
 };
 
 #endif // SHARE_VM_CI_CITYPEARRAY_HPP
--- a/hotspot/src/share/vm/classfile/javaClasses.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/classfile/javaClasses.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -147,9 +147,8 @@
 
 
 int java_lang_String::value_offset  = 0;
-int java_lang_String::offset_offset = 0;
-int java_lang_String::count_offset  = 0;
 int java_lang_String::hash_offset   = 0;
+int java_lang_String::coder_offset  = 0;
 
 bool java_lang_String::initialized  = false;
 
@@ -161,16 +160,39 @@
   assert(!initialized, "offsets should be initialized only once");
 
   Klass* k = SystemDictionary::String_klass();
-  compute_offset(value_offset,           k, vmSymbols::value_name(),  vmSymbols::char_array_signature());
-  compute_optional_offset(offset_offset, k, vmSymbols::offset_name(), vmSymbols::int_signature());
-  compute_optional_offset(count_offset,  k, vmSymbols::count_name(),  vmSymbols::int_signature());
+  compute_offset(value_offset,           k, vmSymbols::value_name(),  vmSymbols::byte_array_signature());
   compute_optional_offset(hash_offset,   k, vmSymbols::hash_name(),   vmSymbols::int_signature());
+  compute_optional_offset(coder_offset,  k, vmSymbols::coder_name(),  vmSymbols::byte_signature());
 
   initialized = true;
 }
 
-Handle java_lang_String::basic_create(int length, TRAPS) {
+class CompactStringsFixup : public FieldClosure {
+private:
+  bool _value;
+
+public:
+  CompactStringsFixup(bool value) : _value(value) {}
+
+  void do_field(fieldDescriptor* fd) {
+    if (fd->name() == vmSymbols::compact_strings_name()) {
+      oop mirror = fd->field_holder()->java_mirror();
+      assert(fd->field_holder() == SystemDictionary::String_klass(), "Should be String");
+      assert(mirror != NULL, "String must have mirror already");
+      mirror->bool_field_put(fd->offset(), _value);
+    }
+  }
+};
+
+void java_lang_String::set_compact_strings(bool value) {
+  CompactStringsFixup fix(value);
+  InstanceKlass::cast(SystemDictionary::String_klass())->do_local_static_fields(&fix);
+}
+
+Handle java_lang_String::basic_create(int length, bool is_latin1, TRAPS) {
   assert(initialized, "Must be initialized");
+  assert(CompactStrings || !is_latin1, "Must be UTF16 without CompactStrings");
+
   // Create the String object first, so there's a chance that the String
   // and the char array it points to end up in the same cache line.
   oop obj;
@@ -179,26 +201,44 @@
   // Create the char array.  The String object must be handlized here
   // because GC can happen as a result of the allocation attempt.
   Handle h_obj(THREAD, obj);
-  typeArrayOop buffer;
-    buffer = oopFactory::new_charArray(length, CHECK_NH);
+  int arr_length = is_latin1 ? length : length << 1; // 2 bytes per UTF16.
+  typeArrayOop buffer = oopFactory::new_byteArray(arr_length, CHECK_NH);;
 
   // Point the String at the char array
   obj = h_obj();
   set_value(obj, buffer);
   // No need to zero the offset, allocation zero'ed the entire String object
-  assert(offset(obj) == 0, "initial String offset should be zero");
-//set_offset(obj, 0);
-  set_count(obj, length);
-
+  set_coder(obj, is_latin1 ? CODER_LATIN1 : CODER_UTF16);
   return h_obj;
 }
 
 Handle java_lang_String::create_from_unicode(jchar* unicode, int length, TRAPS) {
-  Handle h_obj = basic_create(length, CHECK_NH);
+  bool is_latin1 = CompactStrings && UNICODE::is_latin1(unicode, length);
+  Handle h_obj = basic_create(length, is_latin1, CHECK_NH);
   typeArrayOop buffer = value(h_obj());
-  for (int index = 0; index < length; index++) {
-    buffer->char_at_put(index, unicode[index]);
+  assert(TypeArrayKlass::cast(buffer->klass())->element_type() == T_BYTE, "only byte[]");
+  if (is_latin1) {
+    for (int index = 0; index < length; index++) {
+      buffer->byte_at_put(index, (jbyte)unicode[index]);
+    }
+  } else {
+    for (int index = 0; index < length; index++) {
+      buffer->char_at_put(index, unicode[index]);
+    }
   }
+
+#ifdef ASSERT
+  {
+    ResourceMark rm;
+    char* expected = UNICODE::as_utf8(unicode, length);
+    char* actual = as_utf8_string(h_obj());
+    if (strcmp(expected, actual) != 0) {
+      tty->print_cr("Unicode conversion failure: %s --> %s", expected, actual);
+      ShouldNotReachHere();
+    }
+  }
+#endif
+
   return h_obj;
 }
 
@@ -211,11 +251,40 @@
   if (utf8_str == NULL) {
     return Handle();
   }
-  int length = UTF8::unicode_length(utf8_str);
-  Handle h_obj = basic_create(length, CHECK_NH);
+  bool has_multibyte, is_latin1;
+  int length = UTF8::unicode_length(utf8_str, is_latin1, has_multibyte);
+  if (!CompactStrings) {
+    has_multibyte = true;
+    is_latin1 = false;
+  }
+
+  Handle h_obj = basic_create(length, is_latin1, CHECK_NH);
   if (length > 0) {
-    UTF8::convert_to_unicode(utf8_str, value(h_obj())->char_at_addr(0), length);
+    if (!has_multibyte) {
+      strncpy((char*)value(h_obj())->byte_at_addr(0), utf8_str, length);
+    } else if (is_latin1) {
+      UTF8::convert_to_unicode(utf8_str, value(h_obj())->byte_at_addr(0), length);
+    } else {
+      UTF8::convert_to_unicode(utf8_str, value(h_obj())->char_at_addr(0), length);
+    }
   }
+
+#ifdef ASSERT
+  // This check is too strict because the input string is not necessarily valid UTF8.
+  // For example, it may be created with arbitrary content via jni_NewStringUTF.
+  /*
+  {
+    ResourceMark rm;
+    const char* expected = utf8_str;
+    char* actual = as_utf8_string(h_obj());
+    if (strcmp(expected, actual) != 0) {
+      tty->print_cr("String conversion failure: %s --> %s", expected, actual);
+      ShouldNotReachHere();
+    }
+  }
+  */
+#endif
+
   return h_obj;
 }
 
@@ -225,11 +294,39 @@
 }
 
 Handle java_lang_String::create_from_symbol(Symbol* symbol, TRAPS) {
-  int length = UTF8::unicode_length((char*)symbol->bytes(), symbol->utf8_length());
-  Handle h_obj = basic_create(length, CHECK_NH);
+  const char* utf8_str = (char*)symbol->bytes();
+  int utf8_len = symbol->utf8_length();
+
+  bool has_multibyte, is_latin1;
+  int length = UTF8::unicode_length(utf8_str, utf8_len, is_latin1, has_multibyte);
+  if (!CompactStrings) {
+    has_multibyte = true;
+    is_latin1 = false;
+  }
+
+  Handle h_obj = basic_create(length, is_latin1, CHECK_NH);
   if (length > 0) {
-    UTF8::convert_to_unicode((char*)symbol->bytes(), value(h_obj())->char_at_addr(0), length);
+    if (!has_multibyte) {
+      strncpy((char*)value(h_obj())->byte_at_addr(0), utf8_str, length);
+    } else if (is_latin1) {
+      UTF8::convert_to_unicode(utf8_str, value(h_obj())->byte_at_addr(0), length);
+    } else {
+      UTF8::convert_to_unicode(utf8_str, value(h_obj())->char_at_addr(0), length);
+    }
   }
+
+#ifdef ASSERT
+  {
+    ResourceMark rm;
+    const char* expected = symbol->as_utf8();
+    char* actual = as_utf8_string(h_obj());
+    if (strncmp(expected, actual, utf8_len) != 0) {
+      tty->print_cr("Symbol conversion failure: %s --> %s", expected, actual);
+      ShouldNotReachHere();
+    }
+  }
+#endif
+
   return h_obj;
 }
 
@@ -261,7 +358,6 @@
 // Converts a Java String to a native C string that can be used for
 // native OS calls.
 char* java_lang_String::as_platform_dependent_str(Handle java_string, TRAPS) {
-
   typedef char* (*to_platform_string_fn_t)(JNIEnv*, jstring, bool*);
   static to_platform_string_fn_t _to_platform_string_fn = NULL;
 
@@ -292,13 +388,15 @@
   oop          obj    = java_string();
   // Typical usage is to convert all '/' to '.' in string.
   typeArrayOop value  = java_lang_String::value(obj);
-  int          offset = java_lang_String::offset(obj);
   int          length = java_lang_String::length(obj);
+  bool      is_latin1 = java_lang_String::is_latin1(obj);
 
   // First check if any from_char exist
   int index; // Declared outside, used later
   for (index = 0; index < length; index++) {
-    if (value->char_at(index + offset) == from_char) {
+    jchar c = !is_latin1 ? value->char_at(index) :
+                  ((jchar) value->byte_at(index)) & 0xff;
+    if (c == from_char) {
       break;
     }
   }
@@ -307,34 +405,66 @@
     return java_string;
   }
 
-  // Create new UNICODE buffer. Must handlize value because GC
+  // Check if result string will be latin1
+  bool to_is_latin1 = false;
+
+  // Replacement char must be latin1
+  if (CompactStrings && UNICODE::is_latin1(to_char)) {
+    if (is_latin1) {
+      // Source string is latin1 as well
+      to_is_latin1 = true;
+    } else if (!UNICODE::is_latin1(from_char)) {
+      // We are replacing an UTF16 char. Scan string to
+      // check if result can be latin1 encoded.
+      to_is_latin1 = true;
+      for (index = 0; index < length; index++) {
+        jchar c = value->char_at(index);
+        if (c != from_char && !UNICODE::is_latin1(c)) {
+          to_is_latin1 = false;
+          break;
+        }
+      }
+    }
+  }
+
+  // Create new UNICODE (or byte) buffer. Must handlize value because GC
   // may happen during String and char array creation.
   typeArrayHandle h_value(THREAD, value);
-  Handle string = basic_create(length, CHECK_NH);
-
+  Handle string = basic_create(length, to_is_latin1, CHECK_NH);
   typeArrayOop from_buffer = h_value();
-  typeArrayOop to_buffer   = java_lang_String::value(string());
+  typeArrayOop to_buffer = java_lang_String::value(string());
 
   // Copy contents
   for (index = 0; index < length; index++) {
-    jchar c = from_buffer->char_at(index + offset);
+    jchar c = (!is_latin1) ? from_buffer->char_at(index) :
+                    ((jchar) from_buffer->byte_at(index)) & 0xff;
     if (c == from_char) {
       c = to_char;
     }
-    to_buffer->char_at_put(index, c);
+    if (!to_is_latin1) {
+      to_buffer->char_at_put(index, c);
+    } else {
+      to_buffer->byte_at_put(index, (jbyte) c);
+    }
   }
   return string;
 }
 
 jchar* java_lang_String::as_unicode_string(oop java_string, int& length, TRAPS) {
   typeArrayOop value  = java_lang_String::value(java_string);
-  int          offset = java_lang_String::offset(java_string);
                length = java_lang_String::length(java_string);
+  bool      is_latin1 = java_lang_String::is_latin1(java_string);
 
   jchar* result = NEW_RESOURCE_ARRAY_RETURN_NULL(jchar, length);
   if (result != NULL) {
-    for (int index = 0; index < length; index++) {
-      result[index] = value->char_at(index + offset);
+    if (!is_latin1) {
+      for (int index = 0; index < length; index++) {
+        result[index] = value->char_at(index);
+      }
+    } else {
+      for (int index = 0; index < length; index++) {
+        result[index] = ((jchar) value->byte_at(index)) & 0xff;
+      }
     }
   } else {
     THROW_MSG_0(vmSymbols::java_lang_OutOfMemoryError(), "could not allocate Unicode string");
@@ -348,21 +478,35 @@
   if (length == 0) return 0;
 
   typeArrayOop value  = java_lang_String::value(java_string);
-  int          offset = java_lang_String::offset(java_string);
-  return java_lang_String::hash_code(value->char_at_addr(offset), length);
+  bool      is_latin1 = java_lang_String::is_latin1(java_string);
+
+  if (is_latin1) {
+    return java_lang_String::hash_code(value->byte_at_addr(0), length);
+  } else {
+    return java_lang_String::hash_code(value->char_at_addr(0), length);
+  }
 }
 
 char* java_lang_String::as_quoted_ascii(oop java_string) {
   typeArrayOop value  = java_lang_String::value(java_string);
-  int          offset = java_lang_String::offset(java_string);
   int          length = java_lang_String::length(java_string);
-
-  jchar* base = (length == 0) ? NULL : value->char_at_addr(offset);
-  if (base == NULL) return NULL;
-
-  int result_length = UNICODE::quoted_ascii_length(base, length) + 1;
-  char* result = NEW_RESOURCE_ARRAY(char, result_length);
-  UNICODE::as_quoted_ascii(base, length, result, result_length);
+  bool      is_latin1 = java_lang_String::is_latin1(java_string);
+
+  if (length == 0) return NULL;
+
+  char* result;
+  int result_length;
+  if (!is_latin1) {
+    jchar* base = value->char_at_addr(0);
+    result_length = UNICODE::quoted_ascii_length(base, length) + 1;
+    result = NEW_RESOURCE_ARRAY(char, result_length);
+    UNICODE::as_quoted_ascii(base, length, result, result_length);
+  } else {
+    jbyte* base = value->byte_at_addr(0);
+    result_length = UNICODE::quoted_ascii_length(base, length) + 1;
+    result = NEW_RESOURCE_ARRAY(char, result_length);
+    UNICODE::as_quoted_ascii(base, length, result, result_length);
+  }
   assert(result_length >= length + 1, "must not be shorter");
   assert(result_length == (int)strlen(result) + 1, "must match");
   return result;
@@ -370,89 +514,141 @@
 
 unsigned int java_lang_String::hash_string(oop java_string) {
   int          length = java_lang_String::length(java_string);
-  // Zero length string doesn't hash necessarily hash to zero.
+  // Zero length string doesn't necessarily hash to zero.
   if (length == 0) {
-    return StringTable::hash_string(NULL, 0);
+    return StringTable::hash_string((jchar*) NULL, 0);
   }
 
   typeArrayOop value  = java_lang_String::value(java_string);
-  int          offset = java_lang_String::offset(java_string);
-  return StringTable::hash_string(value->char_at_addr(offset), length);
+  bool      is_latin1 = java_lang_String::is_latin1(java_string);
+  if (is_latin1) {
+    return StringTable::hash_string(value->byte_at_addr(0), length);
+  } else {
+    return StringTable::hash_string(value->char_at_addr(0), length);
+  }
 }
 
 Symbol* java_lang_String::as_symbol(Handle java_string, TRAPS) {
   oop          obj    = java_string();
   typeArrayOop value  = java_lang_String::value(obj);
-  int          offset = java_lang_String::offset(obj);
   int          length = java_lang_String::length(obj);
-  jchar* base = (length == 0) ? NULL : value->char_at_addr(offset);
-  Symbol* sym = SymbolTable::lookup_unicode(base, length, THREAD);
-  return sym;
+  bool      is_latin1 = java_lang_String::is_latin1(obj);
+  if (!is_latin1) {
+    jchar* base = (length == 0) ? NULL : value->char_at_addr(0);
+    Symbol* sym = SymbolTable::lookup_unicode(base, length, THREAD);
+    return sym;
+  } else {
+    ResourceMark rm;
+    jbyte* position = (length == 0) ? NULL : value->byte_at_addr(0);
+    const char* base = UNICODE::as_utf8(position, length);
+    Symbol* sym = SymbolTable::lookup(base, length, THREAD);
+    return sym;
+  }
 }
 
 Symbol* java_lang_String::as_symbol_or_null(oop java_string) {
   typeArrayOop value  = java_lang_String::value(java_string);
-  int          offset = java_lang_String::offset(java_string);
   int          length = java_lang_String::length(java_string);
-  jchar* base = (length == 0) ? NULL : value->char_at_addr(offset);
-  return SymbolTable::probe_unicode(base, length);
-}
-
+  bool      is_latin1 = java_lang_String::is_latin1(java_string);
+  if (!is_latin1) {
+    jchar* base = (length == 0) ? NULL : value->char_at_addr(0);
+    return SymbolTable::probe_unicode(base, length);
+  } else {
+    ResourceMark rm;
+    jbyte* position = (length == 0) ? NULL : value->byte_at_addr(0);
+    const char* base = UNICODE::as_utf8(position, length);
+    return SymbolTable::probe(base, length);
+  }
+}
 
 int java_lang_String::utf8_length(oop java_string) {
   typeArrayOop value  = java_lang_String::value(java_string);
-  int          offset = java_lang_String::offset(java_string);
   int          length = java_lang_String::length(java_string);
-  jchar* position = (length == 0) ? NULL : value->char_at_addr(offset);
-  return UNICODE::utf8_length(position, length);
+  bool      is_latin1 = java_lang_String::is_latin1(java_string);
+  if (length == 0) {
+    return 0;
+  }
+  if (!is_latin1) {
+    return UNICODE::utf8_length(value->char_at_addr(0), length);
+  } else {
+    return UNICODE::utf8_length(value->byte_at_addr(0), length);
+  }
 }
 
 char* java_lang_String::as_utf8_string(oop java_string) {
   typeArrayOop value  = java_lang_String::value(java_string);
-  int          offset = java_lang_String::offset(java_string);
   int          length = java_lang_String::length(java_string);
-  jchar* position = (length == 0) ? NULL : value->char_at_addr(offset);
-  return UNICODE::as_utf8(position, length);
+  bool      is_latin1 = java_lang_String::is_latin1(java_string);
+  if (!is_latin1) {
+    jchar* position = (length == 0) ? NULL : value->char_at_addr(0);
+    return UNICODE::as_utf8(position, length);
+  } else {
+    jbyte* position = (length == 0) ? NULL : value->byte_at_addr(0);
+    return UNICODE::as_utf8(position, length);
+  }
 }
 
 char* java_lang_String::as_utf8_string(oop java_string, char* buf, int buflen) {
   typeArrayOop value  = java_lang_String::value(java_string);
-  int          offset = java_lang_String::offset(java_string);
   int          length = java_lang_String::length(java_string);
-  jchar* position = (length == 0) ? NULL : value->char_at_addr(offset);
-  return UNICODE::as_utf8(position, length, buf, buflen);
+  bool      is_latin1 = java_lang_String::is_latin1(java_string);
+  if (!is_latin1) {
+    jchar* position = (length == 0) ? NULL : value->char_at_addr(0);
+    return UNICODE::as_utf8(position, length, buf, buflen);
+  } else {
+    jbyte* position = (length == 0) ? NULL : value->byte_at_addr(0);
+    return UNICODE::as_utf8(position, length, buf, buflen);
+  }
 }
 
 char* java_lang_String::as_utf8_string(oop java_string, int start, int len) {
   typeArrayOop value  = java_lang_String::value(java_string);
-  int          offset = java_lang_String::offset(java_string);
   int          length = java_lang_String::length(java_string);
   assert(start + len <= length, "just checking");
-  jchar* position = value->char_at_addr(offset + start);
-  return UNICODE::as_utf8(position, len);
+  bool      is_latin1 = java_lang_String::is_latin1(java_string);
+  if (!is_latin1) {
+    jchar* position = value->char_at_addr(start);
+    return UNICODE::as_utf8(position, len);
+  } else {
+    jbyte* position = value->byte_at_addr(start);
+    return UNICODE::as_utf8(position, len);
+  }
 }
 
 char* java_lang_String::as_utf8_string(oop java_string, int start, int len, char* buf, int buflen) {
   typeArrayOop value  = java_lang_String::value(java_string);
-  int          offset = java_lang_String::offset(java_string);
   int          length = java_lang_String::length(java_string);
   assert(start + len <= length, "just checking");
-  jchar* position = value->char_at_addr(offset + start);
-  return UNICODE::as_utf8(position, len, buf, buflen);
+  bool      is_latin1 = java_lang_String::is_latin1(java_string);
+  if (!is_latin1) {
+    jchar* position = value->char_at_addr(start);
+    return UNICODE::as_utf8(position, len, buf, buflen);
+  } else {
+    jbyte* position = value->byte_at_addr(start);
+    return UNICODE::as_utf8(position, len, buf, buflen);
+  }
 }
 
 bool java_lang_String::equals(oop java_string, jchar* chars, int len) {
   assert(java_string->klass() == SystemDictionary::String_klass(),
          "must be java_string");
   typeArrayOop value  = java_lang_String::value(java_string);
-  int          offset = java_lang_String::offset(java_string);
   int          length = java_lang_String::length(java_string);
   if (length != len) {
     return false;
   }
-  for (int i = 0; i < len; i++) {
-    if (value->char_at(i + offset) != chars[i]) {
-      return false;
+  bool      is_latin1 = java_lang_String::is_latin1(java_string);
+  if (!is_latin1) {
+    for (int i = 0; i < len; i++) {
+      if (value->char_at(i) != chars[i]) {
+        return false;
+      }
+    }
+  } else {
+    for (int i = 0; i < len; i++) {
+      if ((((jchar) value->byte_at(i)) & 0xff) != chars[i]) {
+        return false;
+      }
     }
   }
   return true;
@@ -464,17 +660,20 @@
   assert(str2->klass() == SystemDictionary::String_klass(),
          "must be java String");
   typeArrayOop value1  = java_lang_String::value(str1);
-  int          offset1 = java_lang_String::offset(str1);
   int          length1 = java_lang_String::length(str1);
+  bool       is_latin1 = java_lang_String::is_latin1(str1);
   typeArrayOop value2  = java_lang_String::value(str2);
-  int          offset2 = java_lang_String::offset(str2);
   int          length2 = java_lang_String::length(str2);
-
-  if (length1 != length2) {
+  bool       is_latin2 = java_lang_String::is_latin1(str2);
+
+  if ((length1 != length2) || (is_latin1 != is_latin2)) {
+    // Strings of different size or with different
+    // coders are never equal.
     return false;
   }
-  for (int i = 0; i < length1; i++) {
-    if (value1->char_at(i + offset1) != value2->char_at(i + offset2)) {
+  int blength1 = value1->length();
+  for (int i = 0; i < value1->length(); i++) {
+    if (value1->byte_at(i) != value2->byte_at(i)) {
       return false;
     }
   }
@@ -492,12 +691,13 @@
     return;
   }
 
-  int offset = java_lang_String::offset(java_string);
   int length = java_lang_String::length(java_string);
+  bool is_latin1 = java_lang_String::is_latin1(java_string);
 
   st->print("\"");
   for (int index = 0; index < length; index++) {
-    st->print("%c", value->char_at(index + offset));
+    st->print("%c", (!is_latin1) ?  value->char_at(index) :
+                           ((jchar) value->byte_at(index)) & 0xff );
   }
   st->print("\"");
 }
@@ -1169,10 +1369,13 @@
 
 // ("name as oop" accessor is not necessary)
 
-typeArrayOop java_lang_ThreadGroup::name(oop java_thread_group) {
+const char* java_lang_ThreadGroup::name(oop java_thread_group) {
   oop name = java_thread_group->obj_field(_name_offset);
   // ThreadGroup.name can be null
-  return name == NULL ? (typeArrayOop)NULL : java_lang_String::value(name);
+  if (name != NULL) {
+    return java_lang_String::as_utf8_string(name);
+  }
+  return NULL;
 }
 
 int java_lang_ThreadGroup::nthreads(oop java_thread_group) {
@@ -3541,14 +3744,13 @@
 
   // java.lang.String
 
-  CHECK_OFFSET("java/lang/String", java_lang_String, value, "[C");
-  if (java_lang_String::has_offset_field()) {
-    CHECK_OFFSET("java/lang/String", java_lang_String, offset, "I");
-    CHECK_OFFSET("java/lang/String", java_lang_String, count, "I");
-  }
+  CHECK_OFFSET("java/lang/String", java_lang_String, value, "[B");
   if (java_lang_String::has_hash_field()) {
     CHECK_OFFSET("java/lang/String", java_lang_String, hash, "I");
   }
+  if (java_lang_String::has_coder_field()) {
+    CHECK_OFFSET("java/lang/String", java_lang_String, coder, "B");
+  }
 
   // java.lang.Class
 
--- a/hotspot/src/share/vm/classfile/javaClasses.hpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/classfile/javaClasses.hpp	Tue Nov 03 09:41:03 2015 +0100
@@ -53,28 +53,28 @@
 class java_lang_String : AllStatic {
  private:
   static int value_offset;
-  static int offset_offset;
-  static int count_offset;
   static int hash_offset;
+  static int coder_offset;
 
   static bool initialized;
 
-  static Handle basic_create(int length, TRAPS);
+  static Handle basic_create(int length, bool byte_arr, TRAPS);
 
-  static void set_offset(oop string, int offset) {
+  static void set_coder(oop string, jbyte coder) {
     assert(initialized, "Must be initialized");
-    if (offset_offset > 0) {
-      string->int_field_put(offset_offset, offset);
-    }
-  }
-  static void set_count( oop string, int count) {
-    assert(initialized, "Must be initialized");
-    if (count_offset > 0) {
-      string->int_field_put(count_offset,  count);
+    if (coder_offset > 0) {
+      string->byte_field_put(coder_offset, coder);
     }
   }
 
  public:
+
+  // Coders
+  enum Coder {
+    CODER_LATIN1 =  0,
+    CODER_UTF16  =  1
+  };
+
   static void compute_offsets();
 
   // Instance creation
@@ -86,37 +86,29 @@
   static Handle create_from_platform_dependent_str(const char* str, TRAPS);
   static Handle char_converter(Handle java_string, jchar from_char, jchar to_char, TRAPS);
 
-  static bool has_offset_field()  {
-    assert(initialized, "Must be initialized");
-    return (offset_offset > 0);
-  }
-
-  static bool has_count_field()  {
-    assert(initialized, "Must be initialized");
-    return (count_offset > 0);
-  }
-
   static bool has_hash_field()  {
     assert(initialized, "Must be initialized");
     return (hash_offset > 0);
   }
+  static bool has_coder_field()  {
+    assert(initialized, "Must be initialized");
+    return (coder_offset > 0);
+  }
+
+  static void set_compact_strings(bool value);
 
   static int value_offset_in_bytes()  {
     assert(initialized && (value_offset > 0), "Must be initialized");
     return value_offset;
   }
-  static int count_offset_in_bytes()  {
-    assert(initialized && (count_offset > 0), "Must be initialized");
-    return count_offset;
-  }
-  static int offset_offset_in_bytes() {
-    assert(initialized && (offset_offset > 0), "Must be initialized");
-    return offset_offset;
-  }
   static int hash_offset_in_bytes()   {
     assert(initialized && (hash_offset > 0), "Must be initialized");
     return hash_offset;
   }
+  static int coder_offset_in_bytes()   {
+    assert(initialized && (coder_offset > 0), "Must be initialized");
+    return coder_offset;
+  }
 
   static void set_value_raw(oop string, typeArrayOop buffer) {
     assert(initialized, "Must be initialized");
@@ -142,28 +134,30 @@
     assert(is_instance(java_string), "must be java_string");
     return java_string->int_field(hash_offset);
   }
-  static int offset(oop java_string) {
+  static bool is_latin1(oop java_string) {
     assert(initialized, "Must be initialized");
     assert(is_instance(java_string), "must be java_string");
-    if (offset_offset > 0) {
-      return java_string->int_field(offset_offset);
+    if (coder_offset > 0) {
+      jbyte coder = java_string->byte_field(coder_offset);
+      assert(CompactStrings || coder == CODER_UTF16, "Must be UTF16 without CompactStrings");
+      return coder == CODER_LATIN1;
     } else {
-      return 0;
+      return false;
     }
   }
   static int length(oop java_string) {
     assert(initialized, "Must be initialized");
     assert(is_instance(java_string), "must be java_string");
-    if (count_offset > 0) {
-      return java_string->int_field(count_offset);
-    } else {
-      typeArrayOop value_array = ((typeArrayOop)java_string->obj_field(value_offset));
-      if (value_array == NULL) {
-        return 0;
-      } else {
-        return value_array->length();
-      }
+    typeArrayOop value_array = ((typeArrayOop)java_string->obj_field(value_offset));
+    if (value_array == NULL) {
+     return 0;
     }
+    int arr_length = value_array->length();
+    if (!is_latin1(java_string)) {
+      assert((arr_length & 1) == 0, "should be even for UTF16 string");
+      arr_length >>= 1; // convert number of bytes to number of elements
+    }
+    return arr_length;
   }
   static int utf8_length(oop java_string);
 
@@ -187,7 +181,7 @@
   // hash P(31) from Kernighan & Ritchie
   //
   // For this reason, THIS ALGORITHM MUST MATCH String.hashCode().
-  template <typename T> static unsigned int hash_code(T* s, int len) {
+  static unsigned int hash_code(const jchar* s, int len) {
     unsigned int h = 0;
     while (len-- > 0) {
       h = 31*h + (unsigned int) *s;
@@ -195,7 +189,18 @@
     }
     return h;
   }
+
+  static unsigned int hash_code(const jbyte* s, int len) {
+    unsigned int h = 0;
+    while (len-- > 0) {
+      h = 31*h + (((unsigned int) *s) & 0xFF);
+      s++;
+    }
+    return h;
+  }
+
   static unsigned int hash_code(oop java_string);
+  static unsigned int latin1_hash_code(typeArrayOop value, int len);
 
   // This is the string hash code used by the StringTable, which may be
   // the same as String.hashCode or an alternate hash code.
@@ -451,7 +456,7 @@
   // parent ThreadGroup
   static oop  parent(oop java_thread_group);
   // name
-  static typeArrayOop name(oop java_thread_group);
+  static const char* name(oop java_thread_group);
   // ("name as oop" accessor is not necessary)
   // Number of threads in group
   static int nthreads(oop java_thread_group);
--- a/hotspot/src/share/vm/classfile/stringTable.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/classfile/stringTable.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -94,11 +94,16 @@
 CompactHashtable<oop, char> StringTable::_shared_table;
 
 // Pick hashing algorithm
-unsigned int StringTable::hash_string(const jchar* s, int len) {
+template<typename T>
+unsigned int StringTable::hash_string(const T* s, int len) {
   return use_alternate_hashcode() ? AltHashing::murmur3_32(seed(), s, len) :
                                     java_lang_String::hash_code(s, len);
 }
 
+// Explicit instantiation for all supported types.
+template unsigned int StringTable::hash_string<jchar>(const jchar* s, int len);
+template unsigned int StringTable::hash_string<jbyte>(const jbyte* s, int len);
+
 oop StringTable::lookup_shared(jchar* name, int len) {
   // java_lang_String::hash_code() was used to compute hash values in the shared table. Don't
   // use the hash value from StringTable::hash_string() as it might use alternate hashcode.
@@ -409,17 +414,25 @@
       for ( ; p != NULL; p = p->next()) {
         oop s = p->literal();
         typeArrayOop value  = java_lang_String::value(s);
-        int          offset = java_lang_String::offset(s);
         int          length = java_lang_String::length(s);
+        bool      is_latin1 = java_lang_String::is_latin1(s);
 
         if (length <= 0) {
           st->print("%d: ", length);
         } else {
           ResourceMark rm(THREAD);
-          jchar* chars = (jchar*)value->char_at_addr(offset);
-          int utf8_length = UNICODE::utf8_length(chars, length);
-          char* utf8_string = NEW_RESOURCE_ARRAY(char, utf8_length + 1);
-          UNICODE::convert_to_utf8(chars, length, utf8_string);
+          int utf8_length;
+          char* utf8_string;
+
+          if (!is_latin1) {
+            jchar* chars = value->char_at_addr(0);
+            utf8_length = UNICODE::utf8_length(chars, length);
+            utf8_string = UNICODE::as_utf8(chars, length);
+          } else {
+            jbyte* bytes = value->byte_at_addr(0);
+            utf8_length = UNICODE::utf8_length(bytes, length);
+            utf8_string = UNICODE::as_utf8(bytes, length);
+          }
 
           st->print("%d: ", utf8_length);
           HashtableTextDump::put_utf8(st, utf8_string, utf8_length);
--- a/hotspot/src/share/vm/classfile/stringTable.hpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/classfile/stringTable.hpp	Tue Nov 03 09:41:03 2015 +0100
@@ -111,7 +111,7 @@
   // Hashing algorithm, used as the hash value used by the
   //     StringTable for bucket selection and comparison (stored in the
   //     HashtableEntry structures).  This is used in the String.intern() method.
-  static unsigned int hash_string(const jchar* s, int len);
+  template<typename T> static unsigned int hash_string(const T* s, int len);
 
   // Internal test.
   static void test_alt_hash() PRODUCT_RETURN;
--- a/hotspot/src/share/vm/classfile/symbolTable.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/classfile/symbolTable.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -244,7 +244,7 @@
 unsigned int SymbolTable::hash_symbol(const char* s, int len) {
   return use_alternate_hashcode() ?
            AltHashing::murmur3_32(seed(), (const jbyte*)s, len) :
-           java_lang_String::hash_code(s, len);
+           java_lang_String::hash_code((const jbyte*)s, len);
 }
 
 
--- a/hotspot/src/share/vm/classfile/symbolTable.hpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/classfile/symbolTable.hpp	Tue Nov 03 09:41:03 2015 +0100
@@ -165,7 +165,7 @@
   // Look up the address of the literal in the SymbolTable for this Symbol*
   static Symbol** lookup_symbol_addr(Symbol* sym);
 
-  // jchar (utf16) version of lookups
+  // jchar (UTF16) version of lookups
   static Symbol* lookup_unicode(const jchar* name, int len, TRAPS);
   static Symbol* lookup_only_unicode(const jchar* name, int len, unsigned int& hash);
 
--- a/hotspot/src/share/vm/classfile/vmSymbols.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/classfile/vmSymbols.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -425,10 +425,26 @@
   // the following switch statement.
   if (!InlineNatives) {
     switch (id) {
-    case vmIntrinsics::_indexOf:
-    case vmIntrinsics::_compareTo:
-    case vmIntrinsics::_equals:
+    case vmIntrinsics::_indexOfL:
+    case vmIntrinsics::_indexOfU:
+    case vmIntrinsics::_indexOfUL:
+    case vmIntrinsics::_indexOfIL:
+    case vmIntrinsics::_indexOfIU:
+    case vmIntrinsics::_indexOfIUL:
+    case vmIntrinsics::_indexOfU_char:
+    case vmIntrinsics::_compareToL:
+    case vmIntrinsics::_compareToU:
+    case vmIntrinsics::_compareToLU:
+    case vmIntrinsics::_compareToUL:
+    case vmIntrinsics::_equalsL:
+    case vmIntrinsics::_equalsU:
     case vmIntrinsics::_equalsC:
+    case vmIntrinsics::_getCharStringU:
+    case vmIntrinsics::_putCharStringU:
+    case vmIntrinsics::_compressStringC:
+    case vmIntrinsics::_compressStringB:
+    case vmIntrinsics::_inflateStringC:
+    case vmIntrinsics::_inflateStringB:
     case vmIntrinsics::_getAndAddInt:
     case vmIntrinsics::_getAndAddLong:
     case vmIntrinsics::_getAndSetInt:
@@ -437,6 +453,7 @@
     case vmIntrinsics::_loadFence:
     case vmIntrinsics::_storeFence:
     case vmIntrinsics::_fullFence:
+    case vmIntrinsics::_hasNegatives:
     case vmIntrinsics::_Reference_get:
       break;
     default:
@@ -619,19 +636,31 @@
     // intrinsic mechanism.
     if (!InlineObjectCopy || !InlineArrayCopy) return true;
     break;
-  case vmIntrinsics::_compareTo:
+  case vmIntrinsics::_compareToL:
+  case vmIntrinsics::_compareToU:
+  case vmIntrinsics::_compareToLU:
+  case vmIntrinsics::_compareToUL:
     if (!SpecialStringCompareTo) return true;
     break;
-  case vmIntrinsics::_indexOf:
+  case vmIntrinsics::_indexOfL:
+  case vmIntrinsics::_indexOfU:
+  case vmIntrinsics::_indexOfUL:
+  case vmIntrinsics::_indexOfIL:
+  case vmIntrinsics::_indexOfIU:
+  case vmIntrinsics::_indexOfIUL:
+  case vmIntrinsics::_indexOfU_char:
     if (!SpecialStringIndexOf) return true;
     break;
-  case vmIntrinsics::_equals:
+  case vmIntrinsics::_equalsL:
+  case vmIntrinsics::_equalsU:
     if (!SpecialStringEquals) return true;
     break;
+  case vmIntrinsics::_equalsB:
   case vmIntrinsics::_equalsC:
     if (!SpecialArraysEquals) return true;
     break;
   case vmIntrinsics::_encodeISOArray:
+  case vmIntrinsics::_encodeByteISOArray:
     if (!SpecialEncodeISOArray) return true;
     break;
   case vmIntrinsics::_getCallerClass:
@@ -666,6 +695,12 @@
   case vmIntrinsics::_subtractExactL:
     if (!UseMathExactIntrinsics || !InlineMathNatives) return true;
     break;
+  case vmIntrinsics::_getCharStringU:
+  case vmIntrinsics::_putCharStringU:
+    // Until JDK-8138651 is fixed, we have to rely on a special flag to
+    // disable these intrinsics for experiments.
+    if (!StringCharIntrinsics) return true;
+    break;
 #endif // COMPILER2
   default:
     return false;
--- a/hotspot/src/share/vm/classfile/vmSymbols.hpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/classfile/vmSymbols.hpp	Tue Nov 03 09:41:03 2015 +0100
@@ -53,6 +53,8 @@
   template(java_lang_Object,                          "java/lang/Object")                         \
   template(java_lang_Class,                           "java/lang/Class")                          \
   template(java_lang_String,                          "java/lang/String")                         \
+  template(java_lang_StringLatin1,                    "java/lang/StringLatin1")                   \
+  template(java_lang_StringUTF16,                     "java/lang/StringUTF16")                    \
   template(java_lang_Thread,                          "java/lang/Thread")                         \
   template(java_lang_ThreadGroup,                     "java/lang/ThreadGroup")                    \
   template(java_lang_Cloneable,                       "java/lang/Cloneable")                      \
@@ -381,9 +383,9 @@
   template(park_event_name,                           "nativeParkEventPointer")                   \
   template(cache_field_name,                          "cache")                                    \
   template(value_name,                                "value")                                    \
-  template(offset_name,                               "offset")                                   \
-  template(count_name,                                "count")                                    \
   template(hash_name,                                 "hash")                                     \
+  template(coder_name,                                "coder")                                    \
+  template(compact_strings_name,                      "COMPACT_STRINGS")                          \
   template(numberOfLeadingZeros_name,                 "numberOfLeadingZeros")                     \
   template(numberOfTrailingZeros_name,                "numberOfTrailingZeros")                    \
   template(bitCount_name,                             "bitCount")                                 \
@@ -837,22 +839,66 @@
                                                                                                                         \
   do_intrinsic(_equalsC,                  java_util_Arrays,       equals_name,    equalsC_signature,             F_S)   \
    do_signature(equalsC_signature,                               "([C[C)Z")                                             \
+  do_intrinsic(_equalsB,                  java_util_Arrays,       equals_name,    equalsB_signature,             F_S)   \
+   do_signature(equalsB_signature,                               "([B[B)Z")                                             \
                                                                                                                         \
-  do_intrinsic(_compareTo,                java_lang_String,       compareTo_name, string_int_signature,          F_R)   \
+  do_intrinsic(_compressStringC,          java_lang_StringUTF16,  compress_name, encodeISOArray_signature,       F_S)   \
+   do_name(     compress_name,                                   "compress")                                            \
+  do_intrinsic(_compressStringB,          java_lang_StringUTF16,  compress_name, indexOfI_signature,             F_S)   \
+  do_intrinsic(_inflateStringC,           java_lang_StringLatin1, inflate_name, inflateC_signature,              F_S)   \
+   do_name(     inflate_name,                                    "inflate")                                             \
+   do_signature(inflateC_signature,                              "([BI[CII)V")                                          \
+  do_intrinsic(_inflateStringB,           java_lang_StringLatin1, inflate_name, inflateB_signature,              F_S)   \
+   do_signature(inflateB_signature,                              "([BI[BII)V")                                          \
+  do_intrinsic(_toBytesStringU,           java_lang_StringUTF16, toBytes_name, toBytesU_signature,               F_S)   \
+   do_name(     toBytes_name,                                    "toBytes")                                             \
+   do_signature(toBytesU_signature,                              "([CII)[B")                                            \
+  do_intrinsic(_getCharsStringU,          java_lang_StringUTF16, getCharsU_name, getCharsU_signature,            F_S)   \
+   do_name(     getCharsU_name,                                  "getChars")                                            \
+   do_signature(getCharsU_signature,                             "([BII[CI)V")                                          \
+  do_intrinsic(_getCharStringU,           java_lang_StringUTF16, getChar_name, getCharStringU_signature,         F_S)   \
+   do_signature(getCharStringU_signature,                        "([BI)C")                                              \
+  do_intrinsic(_putCharStringU,           java_lang_StringUTF16, putChar_name, putCharStringU_signature,         F_S)   \
+   do_signature(putCharStringU_signature,                        "([BII)V")                                             \
+  do_intrinsic(_compareToL,               java_lang_StringLatin1,compareTo_name, compareTo_indexOf_signature,    F_S)   \
+  do_intrinsic(_compareToU,               java_lang_StringUTF16, compareTo_name, compareTo_indexOf_signature,    F_S)   \
+  do_intrinsic(_compareToLU,              java_lang_StringLatin1,compareToLU_name, compareTo_indexOf_signature,  F_S)   \
+  do_intrinsic(_compareToUL,              java_lang_StringUTF16, compareToUL_name, compareTo_indexOf_signature,  F_S)   \
+   do_signature(compareTo_indexOf_signature,                     "([B[B)I")                                             \
    do_name(     compareTo_name,                                  "compareTo")                                           \
-  do_intrinsic(_indexOf,                  java_lang_String,       indexOf_name, string_int_signature,            F_R)   \
+   do_name(     compareToLU_name,                                "compareToUTF16")                                      \
+   do_name(     compareToUL_name,                                "compareToLatin1")                                     \
+  do_intrinsic(_indexOfL,                 java_lang_StringLatin1,indexOf_name, compareTo_indexOf_signature,      F_S)   \
+  do_intrinsic(_indexOfU,                 java_lang_StringUTF16, indexOf_name, compareTo_indexOf_signature,      F_S)   \
+  do_intrinsic(_indexOfUL,                java_lang_StringUTF16, indexOfUL_name, compareTo_indexOf_signature,    F_S)   \
+  do_intrinsic(_indexOfIL,                java_lang_StringLatin1,indexOf_name, indexOfI_signature,               F_S)   \
+  do_intrinsic(_indexOfIU,                java_lang_StringUTF16, indexOf_name, indexOfI_signature,               F_S)   \
+  do_intrinsic(_indexOfIUL,               java_lang_StringUTF16, indexOfUL_name, indexOfI_signature,             F_S)   \
+  do_intrinsic(_indexOfU_char,            java_lang_StringUTF16, indexOfChar_name, indexOfChar_signature,        F_S)   \
    do_name(     indexOf_name,                                    "indexOf")                                             \
-  do_intrinsic(_equals,                   java_lang_String,       equals_name, object_boolean_signature,         F_R)   \
+   do_name(     indexOfChar_name,                                "indexOfChar")                                         \
+   do_name(     indexOfUL_name,                                  "indexOfLatin1")                                       \
+   do_signature(indexOfI_signature,                              "([BI[BII)I")                                          \
+   do_signature(indexOfChar_signature,                           "([BIII)I")                                            \
+  do_intrinsic(_equalsL,                  java_lang_StringLatin1,equals_name, equalsB_signature,                 F_S)   \
+  do_intrinsic(_equalsU,                  java_lang_StringUTF16, equals_name, equalsB_signature,                 F_S)   \
                                                                                                                         \
   do_class(java_nio_Buffer,               "java/nio/Buffer")                                                            \
   do_intrinsic(_checkIndex,               java_nio_Buffer,        checkIndex_name, int_int_signature,            F_R)   \
    do_name(     checkIndex_name,                                 "checkIndex")                                          \
                                                                                                                         \
+  do_class(java_lang_StringCoding,        "java/lang/StringCoding")                                                     \
+  do_intrinsic(_hasNegatives,             java_lang_StringCoding, hasNegatives_name, hasNegatives_signature,     F_S)   \
+   do_name(     hasNegatives_name,                               "hasNegatives")                                        \
+   do_signature(hasNegatives_signature,                          "([BII)Z")                                             \
+                                                                                                                        \
   do_class(sun_nio_cs_iso8859_1_Encoder,  "sun/nio/cs/ISO_8859_1$Encoder")                                              \
   do_intrinsic(_encodeISOArray,     sun_nio_cs_iso8859_1_Encoder, encodeISOArray_name, encodeISOArray_signature, F_S)   \
    do_name(     encodeISOArray_name,                             "implEncodeISOArray")                                  \
    do_signature(encodeISOArray_signature,                        "([CI[BII)I")                                          \
                                                                                                                         \
+  do_intrinsic(_encodeByteISOArray,     java_lang_StringCoding, encodeISOArray_name, indexOfI_signature,         F_S)   \
+                                                                                                                        \
   do_class(java_math_BigInteger,                      "java/math/BigInteger")                                           \
   do_intrinsic(_multiplyToLen,      java_math_BigInteger, multiplyToLen_name, multiplyToLen_signature, F_S)             \
    do_name(     multiplyToLen_name,                             "implMultiplyToLen")                                    \
--- a/hotspot/src/share/vm/gc/g1/g1StringDedupTable.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/gc/g1/g1StringDedupTable.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -198,10 +198,11 @@
   _table = new G1StringDedupTable(_min_size);
 }
 
-void G1StringDedupTable::add(typeArrayOop value, unsigned int hash, G1StringDedupEntry** list) {
+void G1StringDedupTable::add(typeArrayOop value, bool latin1, unsigned int hash, G1StringDedupEntry** list) {
   G1StringDedupEntry* entry = _entry_cache->alloc();
   entry->set_obj(value);
   entry->set_hash(hash);
+  entry->set_latin1(latin1);
   entry->set_next(*list);
   *list = entry;
   _entries++;
@@ -226,15 +227,15 @@
 bool G1StringDedupTable::equals(typeArrayOop value1, typeArrayOop value2) {
   return (value1 == value2 ||
           (value1->length() == value2->length() &&
-           (!memcmp(value1->base(T_CHAR),
-                    value2->base(T_CHAR),
-                    value1->length() * sizeof(jchar)))));
+           (!memcmp(value1->base(T_BYTE),
+                    value2->base(T_BYTE),
+                    value1->length() * sizeof(jbyte)))));
 }
 
-typeArrayOop G1StringDedupTable::lookup(typeArrayOop value, unsigned int hash,
+typeArrayOop G1StringDedupTable::lookup(typeArrayOop value, bool latin1, unsigned int hash,
                                         G1StringDedupEntry** list, uintx &count) {
   for (G1StringDedupEntry* entry = *list; entry != NULL; entry = entry->next()) {
-    if (entry->hash() == hash) {
+    if (entry->hash() == hash && entry->latin1() == latin1) {
       typeArrayOop existing_value = entry->obj();
       if (equals(value, existing_value)) {
         // Match found
@@ -248,13 +249,13 @@
   return NULL;
 }
 
-typeArrayOop G1StringDedupTable::lookup_or_add_inner(typeArrayOop value, unsigned int hash) {
+typeArrayOop G1StringDedupTable::lookup_or_add_inner(typeArrayOop value, bool latin1, unsigned int hash) {
   size_t index = hash_to_index(hash);
   G1StringDedupEntry** list = bucket(index);
   uintx count = 0;
 
   // Lookup in list
-  typeArrayOop existing_value = lookup(value, hash, list, count);
+  typeArrayOop existing_value = lookup(value, latin1, hash, list, count);
 
   // Check if rehash is needed
   if (count > _rehash_threshold) {
@@ -263,7 +264,7 @@
 
   if (existing_value == NULL) {
     // Not found, add new entry
-    add(value, hash, list);
+    add(value, latin1, hash, list);
 
     // Update statistics
     _entries_added++;
@@ -272,15 +273,24 @@
   return existing_value;
 }
 
-unsigned int G1StringDedupTable::hash_code(typeArrayOop value) {
+unsigned int G1StringDedupTable::hash_code(typeArrayOop value, bool latin1) {
   unsigned int hash;
   int length = value->length();
-  const jchar* data = (jchar*)value->base(T_CHAR);
-
-  if (use_java_hash()) {
-    hash = java_lang_String::hash_code(data, length);
+  if (latin1) {
+    const jbyte* data = (jbyte*)value->base(T_BYTE);
+    if (use_java_hash()) {
+      hash = java_lang_String::hash_code(data, length);
+    } else {
+      hash = AltHashing::murmur3_32(_table->_hash_seed, data, length);
+    }
   } else {
-    hash = AltHashing::murmur3_32(_table->_hash_seed, data, length);
+    length /= sizeof(jchar) / sizeof(jbyte); // Convert number of bytes to number of chars
+    const jchar* data = (jchar*)value->base(T_CHAR);
+    if (use_java_hash()) {
+      hash = java_lang_String::hash_code(data, length);
+    } else {
+      hash = AltHashing::murmur3_32(_table->_hash_seed, data, length);
+    }
   }
 
   return hash;
@@ -299,6 +309,7 @@
     return;
   }
 
+  bool latin1 = java_lang_String::is_latin1(java_string);
   unsigned int hash = 0;
 
   if (use_java_hash()) {
@@ -308,7 +319,7 @@
 
   if (hash == 0) {
     // Compute hash
-    hash = hash_code(value);
+    hash = hash_code(value, latin1);
     stat.inc_hashed();
 
     if (use_java_hash() && hash != 0) {
@@ -317,7 +328,7 @@
     }
   }
 
-  typeArrayOop existing_value = lookup_or_add(value, hash);
+  typeArrayOop existing_value = lookup_or_add(value, latin1, hash);
   if (existing_value == value) {
     // Same value, already known
     stat.inc_known();
@@ -459,7 +470,8 @@
             // destination partitions. finish_rehash() will do a single
             // threaded transfer of all entries.
             typeArrayOop value = (typeArrayOop)*p;
-            unsigned int hash = hash_code(value);
+            bool latin1 = (*entry)->latin1();
+            unsigned int hash = hash_code(value, latin1);
             (*entry)->set_hash(hash);
           }
 
@@ -523,7 +535,8 @@
       guarantee(G1CollectedHeap::heap()->is_in_reserved(value), "Object must be on the heap");
       guarantee(!value->is_forwarded(), "Object must not be forwarded");
       guarantee(value->is_typeArray(), "Object must be a typeArrayOop");
-      unsigned int hash = hash_code(value);
+      bool latin1 = (*entry)->latin1();
+      unsigned int hash = hash_code(value, latin1);
       guarantee((*entry)->hash() == hash, "Table entry has inorrect hash");
       guarantee(_table->hash_to_index(hash) == bucket, "Table entry has incorrect index");
       entry = (*entry)->next_addr();
@@ -536,10 +549,12 @@
     G1StringDedupEntry** entry1 = _table->bucket(bucket);
     while (*entry1 != NULL) {
       typeArrayOop value1 = (*entry1)->obj();
+      bool latin1_1 = (*entry1)->latin1();
       G1StringDedupEntry** entry2 = (*entry1)->next_addr();
       while (*entry2 != NULL) {
         typeArrayOop value2 = (*entry2)->obj();
-        guarantee(!equals(value1, value2), "Table entries must not have identical arrays");
+        bool latin1_2 = (*entry2)->latin1();
+        guarantee(latin1_1 != latin1_2 || !equals(value1, value2), "Table entries must not have identical arrays");
         entry2 = (*entry2)->next_addr();
       }
       entry1 = (*entry1)->next_addr();
--- a/hotspot/src/share/vm/gc/g1/g1StringDedupTable.hpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/gc/g1/g1StringDedupTable.hpp	Tue Nov 03 09:41:03 2015 +0100
@@ -40,12 +40,14 @@
 private:
   G1StringDedupEntry* _next;
   unsigned int      _hash;
+  bool              _latin1;
   typeArrayOop      _obj;
 
 public:
   G1StringDedupEntry() :
     _next(NULL),
     _hash(0),
+    _latin1(false),
     _obj(NULL) {
   }
 
@@ -69,6 +71,14 @@
     _hash = hash;
   }
 
+  bool latin1() {
+    return _latin1;
+  }
+
+  void set_latin1(bool latin1) {
+    _latin1 = latin1;
+  }
+
   typeArrayOop obj() {
     return _obj;
   }
@@ -152,7 +162,7 @@
   }
 
   // Adds a new table entry to the given hash bucket.
-  void add(typeArrayOop value, unsigned int hash, G1StringDedupEntry** list);
+  void add(typeArrayOop value, bool latin1, unsigned int hash, G1StringDedupEntry** list);
 
   // Removes the given table entry from the table.
   void remove(G1StringDedupEntry** pentry, uint worker_id);
@@ -162,20 +172,20 @@
 
   // Returns an existing character array in the given hash bucket, or NULL
   // if no matching character array exists.
-  typeArrayOop lookup(typeArrayOop value, unsigned int hash,
+  typeArrayOop lookup(typeArrayOop value, bool latin1, unsigned int hash,
                       G1StringDedupEntry** list, uintx &count);
 
   // Returns an existing character array in the table, or inserts a new
   // table entry if no matching character array exists.
-  typeArrayOop lookup_or_add_inner(typeArrayOop value, unsigned int hash);
+  typeArrayOop lookup_or_add_inner(typeArrayOop value, bool latin1, unsigned int hash);
 
   // Thread safe lookup or add of table entry
-  static typeArrayOop lookup_or_add(typeArrayOop value, unsigned int hash) {
+  static typeArrayOop lookup_or_add(typeArrayOop value, bool latin1, unsigned int hash) {
     // Protect the table from concurrent access. Also note that this lock
     // acts as a fence for _table, which could have been replaced by a new
     // instance if the table was resized or rehashed.
     MutexLockerEx ml(StringDedupTable_lock, Mutex::_no_safepoint_check_flag);
-    return _table->lookup_or_add_inner(value, hash);
+    return _table->lookup_or_add_inner(value, latin1, hash);
   }
 
   // Returns true if the hashtable is currently using a Java compatible
@@ -188,7 +198,7 @@
 
   // Computes the hash code for the given character array, using the
   // currently active hash function and hash seed.
-  static unsigned int hash_code(typeArrayOop value);
+  static unsigned int hash_code(typeArrayOop value, bool latin1);
 
   static uintx unlink_or_oops_do(G1StringDedupUnlinkOrOopsDoClosure* cl,
                                  size_t partition_begin,
--- a/hotspot/src/share/vm/memory/filemap.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/memory/filemap.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -168,6 +168,7 @@
   _version = _current_version;
   _alignment = alignment;
   _obj_alignment = ObjectAlignmentInBytes;
+  _compact_strings = CompactStrings;
   _narrow_oop_mode = Universe::narrow_oop_mode();
   _narrow_oop_shift = Universe::narrow_oop_shift();
   _max_heap_size = MaxHeapSize;
@@ -900,6 +901,13 @@
                   _obj_alignment, ObjectAlignmentInBytes);
     return false;
   }
+  if (_compact_strings != CompactStrings) {
+    FileMapInfo::fail_continue("The shared archive file's CompactStrings setting (%s)"
+                  " does not equal the current CompactStrings setting (%s).",
+                  _compact_strings ? "enabled" : "disabled",
+                  CompactStrings   ? "enabled" : "disabled");
+    return false;
+  }
 
   return true;
 }
--- a/hotspot/src/share/vm/memory/filemap.hpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/memory/filemap.hpp	Tue Nov 03 09:41:03 2015 +0100
@@ -95,6 +95,7 @@
     size_t _alignment;                // how shared archive should be aligned
     int    _obj_alignment;            // value of ObjectAlignmentInBytes
     int    _narrow_oop_shift;         // compressed oop encoding shift
+    bool   _compact_strings;          // value of CompactStrings
     uintx  _max_heap_size;            // java max heap size during dumping
     Universe::NARROW_OOP_MODE _narrow_oop_mode; // compressed oop encoding mode
     int     _narrow_klass_shift;      // save narrow klass base and shift
--- a/hotspot/src/share/vm/memory/metaspaceShared.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/memory/metaspaceShared.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -540,7 +540,7 @@
 
   NOT_PRODUCT(SystemDictionary::verify();)
 
-  // Copy the the symbol table, string table, and the system dictionary to the shared
+  // Copy the symbol table, string table, and the system dictionary to the shared
   // space in usable form.  Copy the hashtable
   // buckets first [read-write], then copy the linked lists of entries
   // [read-only].
--- a/hotspot/src/share/vm/oops/instanceKlass.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/oops/instanceKlass.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -2921,12 +2921,10 @@
 
   if (this == SystemDictionary::String_klass()) {
     typeArrayOop value  = java_lang_String::value(obj);
-    juint        offset = java_lang_String::offset(obj);
     juint        length = java_lang_String::length(obj);
     if (value != NULL &&
         value->is_typeArray() &&
-        offset          <= (juint) value->length() &&
-        offset + length <= (juint) value->length()) {
+        length <= (juint) value->length()) {
       st->print(BULLET"string: ");
       java_lang_String::print(obj, st);
       st->cr();
--- a/hotspot/src/share/vm/oops/typeArrayOop.hpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/oops/typeArrayOop.hpp	Tue Nov 03 09:41:03 2015 +0100
@@ -48,47 +48,47 @@
 
  public:
   jbyte* byte_at_addr(int which) const {
-    assert(is_within_bounds(which), "index out of bounds");
+    assert(is_within_bounds(which), "index %d out of bounds %d", which, length());
     return &byte_base()[which];
   }
 
   jboolean* bool_at_addr(int which) const {
-    assert(is_within_bounds(which), "index out of bounds");
+    assert(is_within_bounds(which), "index %d out of bounds %d", which, length());
     return &bool_base()[which];
   }
 
   jchar* char_at_addr(int which) const {
-    assert(is_within_bounds(which), "index out of bounds");
+    assert(is_within_bounds(which), "index %d out of bounds %d", which, length());
     return &char_base()[which];
   }
 
   jint* int_at_addr(int which) const {
-    assert(is_within_bounds(which), "index out of bounds");
+    assert(is_within_bounds(which), "index %d out of bounds %d", which, length());
     return &int_base()[which];
   }
 
   jshort* short_at_addr(int which) const {
-    assert(is_within_bounds(which), "index out of bounds");
+    assert(is_within_bounds(which), "index %d out of bounds %d", which, length());
     return &short_base()[which];
   }
 
   jushort* ushort_at_addr(int which) const {  // for field descriptor arrays
-    assert(is_within_bounds(which), "index out of bounds");
+    assert(is_within_bounds(which), "index %d out of bounds %d", which, length());
     return (jushort*) &short_base()[which];
   }
 
   jlong* long_at_addr(int which) const {
-    assert(is_within_bounds(which), "index out of bounds");
+    assert(is_within_bounds(which), "index %d out of bounds %d", which, length());
     return &long_base()[which];
   }
 
   jfloat* float_at_addr(int which) const {
-    assert(is_within_bounds(which), "index out of bounds");
+    assert(is_within_bounds(which), "index %d out of bounds %d", which, length());
     return &float_base()[which];
   }
 
   jdouble* double_at_addr(int which) const {
-    assert(is_within_bounds(which), "index out of bounds");
+    assert(is_within_bounds(which), "index %d out of bounds %d", which, length());
     return &double_base()[which];
   }
 
--- a/hotspot/src/share/vm/opto/c2compiler.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/opto/c2compiler.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -181,12 +181,25 @@
   }
 
   switch (id) {
-  case vmIntrinsics::_compareTo:
+  case vmIntrinsics::_compressStringC:
+  case vmIntrinsics::_compressStringB:
+    if (!Matcher::has_match_rule(Op_StrCompressedCopy)) return false;
+    break;
+  case vmIntrinsics::_inflateStringC:
+  case vmIntrinsics::_inflateStringB:
+    if (!Matcher::has_match_rule(Op_StrInflatedCopy)) return false;
+    break;
+  case vmIntrinsics::_compareToL:
+  case vmIntrinsics::_compareToU:
+  case vmIntrinsics::_compareToLU:
+  case vmIntrinsics::_compareToUL:
     if (!Matcher::match_rule_supported(Op_StrComp)) return false;
     break;
-  case vmIntrinsics::_equals:
+  case vmIntrinsics::_equalsL:
+  case vmIntrinsics::_equalsU:
     if (!Matcher::match_rule_supported(Op_StrEquals)) return false;
     break;
+  case vmIntrinsics::_equalsB:
   case vmIntrinsics::_equalsC:
     if (!Matcher::match_rule_supported(Op_AryEq)) return false;
     break;
@@ -194,8 +207,12 @@
     if (StubRoutines::unsafe_arraycopy() == NULL) return false;
     break;
   case vmIntrinsics::_encodeISOArray:
+  case vmIntrinsics::_encodeByteISOArray:
     if (!Matcher::match_rule_supported(Op_EncodeISOArray)) return false;
     break;
+  case vmIntrinsics::_hasNegatives:
+    if (!Matcher::match_rule_supported(Op_HasNegatives))  return false;
+    break;
   case vmIntrinsics::_bitCount_i:
     if (!Matcher::match_rule_supported(Op_PopCountI)) return false;
     break;
@@ -302,7 +319,17 @@
   case vmIntrinsics::_min:
   case vmIntrinsics::_max:
   case vmIntrinsics::_arraycopy:
-  case vmIntrinsics::_indexOf:
+  case vmIntrinsics::_indexOfL:
+  case vmIntrinsics::_indexOfU:
+  case vmIntrinsics::_indexOfUL:
+  case vmIntrinsics::_indexOfIL:
+  case vmIntrinsics::_indexOfIU:
+  case vmIntrinsics::_indexOfIUL:
+  case vmIntrinsics::_indexOfU_char:
+  case vmIntrinsics::_toBytesStringU:
+  case vmIntrinsics::_getCharsStringU:
+  case vmIntrinsics::_getCharStringU:
+  case vmIntrinsics::_putCharStringU:
   case vmIntrinsics::_getObject:
   case vmIntrinsics::_getBoolean:
   case vmIntrinsics::_getByte:
--- a/hotspot/src/share/vm/opto/classes.hpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/opto/classes.hpp	Tue Nov 03 09:41:03 2015 +0100
@@ -136,6 +136,7 @@
 macro(FastUnlock)
 macro(Goto)
 macro(Halt)
+macro(HasNegatives)
 macro(If)
 macro(IfFalse)
 macro(IfTrue)
@@ -246,8 +247,11 @@
 macro(StoreN)
 macro(StoreNKlass)
 macro(StrComp)
+macro(StrCompressedCopy)
 macro(StrEquals)
 macro(StrIndexOf)
+macro(StrIndexOfChar)
+macro(StrInflatedCopy)
 macro(SubD)
 macro(SubF)
 macro(SubI)
--- a/hotspot/src/share/vm/opto/escape.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/opto/escape.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -560,9 +560,13 @@
       break;
     }
     case Op_AryEq:
+    case Op_HasNegatives:
     case Op_StrComp:
     case Op_StrEquals:
     case Op_StrIndexOf:
+    case Op_StrIndexOfChar:
+    case Op_StrInflatedCopy:
+    case Op_StrCompressedCopy:
     case Op_EncodeISOArray: {
       add_local_var(n, PointsToNode::ArgEscape);
       delayed_worklist->push(n); // Process it later.
@@ -743,11 +747,15 @@
       ELSE_FAIL("Op_StoreP");
     }
     case Op_AryEq:
+    case Op_HasNegatives:
     case Op_StrComp:
     case Op_StrEquals:
     case Op_StrIndexOf:
+    case Op_StrIndexOfChar:
+    case Op_StrInflatedCopy:
+    case Op_StrCompressedCopy:
     case Op_EncodeISOArray: {
-      // char[] arrays passed to string intrinsic do not escape but
+      // char[]/byte[] arrays passed to string intrinsic do not escape but
       // they are not scalar replaceable. Adjust escape state for them.
       // Start from in(2) edge since in(1) is memory edge.
       for (uint i = 2; i < n->req(); i++) {
@@ -2722,17 +2730,34 @@
       if (mem->is_LoadStore()) {
         adr = mem->in(MemNode::Address);
       } else {
-        assert(mem->Opcode() == Op_EncodeISOArray, "sanity");
+        assert(mem->Opcode() == Op_EncodeISOArray ||
+               mem->Opcode() == Op_StrCompressedCopy, "sanity");
         adr = mem->in(3); // Memory edge corresponds to destination array
       }
       const Type *at = igvn->type(adr);
       if (at != Type::TOP) {
-        assert (at->isa_ptr() != NULL, "pointer type required.");
+        assert(at->isa_ptr() != NULL, "pointer type required.");
         int idx = C->get_alias_index(at->is_ptr());
-        assert(idx != alias_idx, "Object is not scalar replaceable if a LoadStore node access its field");
-        break;
+        if (idx == alias_idx) {
+          // Assert in debug mode
+          assert(false, "Object is not scalar replaceable if a LoadStore node accesses its field");
+          break; // In product mode return SCMemProj node
+        }
       }
       result = mem->in(MemNode::Memory);
+    } else if (result->Opcode() == Op_StrInflatedCopy) {
+      Node* adr = result->in(3); // Memory edge corresponds to destination array
+      const Type *at = igvn->type(adr);
+      if (at != Type::TOP) {
+        assert(at->isa_ptr() != NULL, "pointer type required.");
+        int idx = C->get_alias_index(at->is_ptr());
+        if (idx == alias_idx) {
+          // Assert in debug mode
+          assert(false, "Object is not scalar replaceable if a StrInflatedCopy node accesses its field");
+          break; // In product mode return SCMemProj node
+        }
+      }
+      result = result->in(MemNode::Memory);
     }
   }
   if (result->is_Phi()) {
@@ -3096,10 +3121,15 @@
         }
       } else {
         uint op = use->Opcode();
-        if (!(op == Op_CmpP || op == Op_Conv2B ||
+        if ((use->in(MemNode::Memory) == n) &&
+            (op == Op_StrCompressedCopy || op == Op_StrInflatedCopy)) {
+          // They overwrite memory edge corresponding to destination array,
+          memnode_worklist.append_if_missing(use);
+        } else if (!(op == Op_CmpP || op == Op_Conv2B ||
               op == Op_CastP2X || op == Op_StoreCM ||
-              op == Op_FastLock || op == Op_AryEq || op == Op_StrComp ||
-              op == Op_StrEquals || op == Op_StrIndexOf)) {
+              op == Op_FastLock || op == Op_AryEq || op == Op_StrComp || op == Op_HasNegatives ||
+              op == Op_StrCompressedCopy || op == Op_StrInflatedCopy ||
+              op == Op_StrEquals || op == Op_StrIndexOf || op == Op_StrIndexOfChar)) {
           n->dump();
           use->dump();
           assert(false, "EA: missing allocation reference path");
@@ -3161,7 +3191,8 @@
       n = n->as_MemBar()->proj_out(TypeFunc::Memory);
       if (n == NULL)
         continue;
-    } else if (n->Opcode() == Op_EncodeISOArray) {
+    } else if (n->Opcode() == Op_StrCompressedCopy ||
+               n->Opcode() == Op_EncodeISOArray) {
       // get the memory projection
       n = n->find_out_with(Op_SCMemProj);
       assert(n->Opcode() == Op_SCMemProj, "memory projection required");
@@ -3216,11 +3247,16 @@
         }
       } else {
         uint op = use->Opcode();
-        if (!(op == Op_StoreCM ||
+        if ((use->in(MemNode::Memory) == n) &&
+            (op == Op_StrCompressedCopy || op == Op_StrInflatedCopy)) {
+          // They overwrite memory edge corresponding to destination array,
+          memnode_worklist.append_if_missing(use);
+        } else if (!(op == Op_StoreCM ||
               (op == Op_CallLeaf && use->as_CallLeaf()->_name != NULL &&
                strcmp(use->as_CallLeaf()->_name, "g1_wb_pre") == 0) ||
-              op == Op_AryEq || op == Op_StrComp ||
-              op == Op_StrEquals || op == Op_StrIndexOf)) {
+              op == Op_AryEq || op == Op_StrComp || op == Op_HasNegatives ||
+              op == Op_StrCompressedCopy || op == Op_StrInflatedCopy ||
+              op == Op_StrEquals || op == Op_StrIndexOf || op == Op_StrIndexOfChar)) {
           n->dump();
           use->dump();
           assert(false, "EA: missing memory path");
--- a/hotspot/src/share/vm/opto/gcm.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/opto/gcm.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1997, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1997, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -504,8 +504,12 @@
          "String equals is a 'load' that does not conflict with any stores");
   assert(load_alias_idx || (load->is_Mach() && load->as_Mach()->ideal_Opcode() == Op_StrIndexOf),
          "String indexOf is a 'load' that does not conflict with any stores");
+  assert(load_alias_idx || (load->is_Mach() && load->as_Mach()->ideal_Opcode() == Op_StrIndexOfChar),
+         "String indexOfChar is a 'load' that does not conflict with any stores");
   assert(load_alias_idx || (load->is_Mach() && load->as_Mach()->ideal_Opcode() == Op_AryEq),
-         "Arrays equals is a 'load' that do not conflict with any stores");
+         "Arrays equals is a 'load' that does not conflict with any stores");
+  assert(load_alias_idx || (load->is_Mach() && load->as_Mach()->ideal_Opcode() == Op_HasNegatives),
+         "HasNegatives is a 'load' that does not conflict with any stores");
 
   if (!C->alias_type(load_alias_idx)->is_rewritable()) {
     // It is impossible to spoil this load by putting stores before it,
--- a/hotspot/src/share/vm/opto/graphKit.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/opto/graphKit.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -4266,35 +4266,11 @@
 #undef __
 
 
-
-Node* GraphKit::load_String_offset(Node* ctrl, Node* str) {
-  if (java_lang_String::has_offset_field()) {
-    int offset_offset = java_lang_String::offset_offset_in_bytes();
-    const TypeInstPtr* string_type = TypeInstPtr::make(TypePtr::NotNull, C->env()->String_klass(),
-                                                       false, NULL, 0);
-    const TypePtr* offset_field_type = string_type->add_offset(offset_offset);
-    int offset_field_idx = C->get_alias_index(offset_field_type);
-    return make_load(ctrl,
-                     basic_plus_adr(str, str, offset_offset),
-                     TypeInt::INT, T_INT, offset_field_idx, MemNode::unordered);
-  } else {
-    return intcon(0);
-  }
-}
-
 Node* GraphKit::load_String_length(Node* ctrl, Node* str) {
-  if (java_lang_String::has_count_field()) {
-    int count_offset = java_lang_String::count_offset_in_bytes();
-    const TypeInstPtr* string_type = TypeInstPtr::make(TypePtr::NotNull, C->env()->String_klass(),
-                                                       false, NULL, 0);
-    const TypePtr* count_field_type = string_type->add_offset(count_offset);
-    int count_field_idx = C->get_alias_index(count_field_type);
-    return make_load(ctrl,
-                     basic_plus_adr(str, str, count_offset),
-                     TypeInt::INT, T_INT, count_field_idx, MemNode::unordered);
-  } else {
-    return load_array_length(load_String_value(ctrl, str));
-  }
+  Node* len = load_array_length(load_String_value(ctrl, str));
+  Node* coder = load_String_coder(ctrl, str);
+  // Divide length by 2 if coder is UTF16
+  return _gvn.transform(new RShiftINode(len, coder));
 }
 
 Node* GraphKit::load_String_value(Node* ctrl, Node* str) {
@@ -4302,9 +4278,9 @@
   const TypeInstPtr* string_type = TypeInstPtr::make(TypePtr::NotNull, C->env()->String_klass(),
                                                      false, NULL, 0);
   const TypePtr* value_field_type = string_type->add_offset(value_offset);
-  const TypeAryPtr*  value_type = TypeAryPtr::make(TypePtr::NotNull,
-                                                   TypeAry::make(TypeInt::CHAR,TypeInt::POS),
-                                                   ciTypeArrayKlass::make(T_CHAR), true, 0);
+  const TypeAryPtr* value_type = TypeAryPtr::make(TypePtr::NotNull,
+                                                  TypeAry::make(TypeInt::BYTE, TypeInt::POS),
+                                                  ciTypeArrayKlass::make(T_BYTE), true, 0);
   int value_field_idx = C->get_alias_index(value_field_type);
   Node* load = make_load(ctrl, basic_plus_adr(str, str, value_offset),
                          value_type, T_OBJECT, value_field_idx, MemNode::unordered);
@@ -4315,14 +4291,21 @@
   return load;
 }
 
-void GraphKit::store_String_offset(Node* ctrl, Node* str, Node* value) {
-  int offset_offset = java_lang_String::offset_offset_in_bytes();
-  const TypeInstPtr* string_type = TypeInstPtr::make(TypePtr::NotNull, C->env()->String_klass(),
-                                                     false, NULL, 0);
-  const TypePtr* offset_field_type = string_type->add_offset(offset_offset);
-  int offset_field_idx = C->get_alias_index(offset_field_type);
-  store_to_memory(ctrl, basic_plus_adr(str, offset_offset),
-                  value, T_INT, offset_field_idx, MemNode::unordered);
+Node* GraphKit::load_String_coder(Node* ctrl, Node* str) {
+  if (java_lang_String::has_coder_field()) {
+    if (!CompactStrings) {
+      return intcon(java_lang_String::CODER_UTF16);
+    }
+    int coder_offset = java_lang_String::coder_offset_in_bytes();
+    const TypeInstPtr* string_type = TypeInstPtr::make(TypePtr::NotNull, C->env()->String_klass(),
+                                                       false, NULL, 0);
+    const TypePtr* coder_field_type = string_type->add_offset(coder_offset);
+    int coder_field_idx = C->get_alias_index(coder_field_type);
+    return make_load(ctrl, basic_plus_adr(str, str, coder_offset),
+                     TypeInt::BYTE, T_BYTE, coder_field_idx, MemNode::unordered);
+  } else {
+    return intcon(0); // false
+  }
 }
 
 void GraphKit::store_String_value(Node* ctrl, Node* str, Node* value) {
@@ -4330,19 +4313,76 @@
   const TypeInstPtr* string_type = TypeInstPtr::make(TypePtr::NotNull, C->env()->String_klass(),
                                                      false, NULL, 0);
   const TypePtr* value_field_type = string_type->add_offset(value_offset);
-
   store_oop_to_object(ctrl, str,  basic_plus_adr(str, value_offset), value_field_type,
-      value, TypeAryPtr::CHARS, T_OBJECT, MemNode::unordered);
+      value, TypeAryPtr::BYTES, T_OBJECT, MemNode::unordered);
 }
 
-void GraphKit::store_String_length(Node* ctrl, Node* str, Node* value) {
-  int count_offset = java_lang_String::count_offset_in_bytes();
+void GraphKit::store_String_coder(Node* ctrl, Node* str, Node* value) {
+  int coder_offset = java_lang_String::coder_offset_in_bytes();
   const TypeInstPtr* string_type = TypeInstPtr::make(TypePtr::NotNull, C->env()->String_klass(),
                                                      false, NULL, 0);
-  const TypePtr* count_field_type = string_type->add_offset(count_offset);
-  int count_field_idx = C->get_alias_index(count_field_type);
-  store_to_memory(ctrl, basic_plus_adr(str, count_offset),
-                  value, T_INT, count_field_idx, MemNode::unordered);
+  const TypePtr* coder_field_type = string_type->add_offset(coder_offset);
+  int coder_field_idx = C->get_alias_index(coder_field_type);
+  store_to_memory(ctrl, basic_plus_adr(str, coder_offset),
+                  value, T_BYTE, coder_field_idx, MemNode::unordered);
+}
+
+Node* GraphKit::compress_string(Node* src, Node* dst, Node* count) {
+  assert(Matcher::match_rule_supported(Op_StrCompressedCopy), "Intrinsic not supported");
+  uint idx = C->get_alias_index(TypeAryPtr::BYTES);
+  StrCompressedCopyNode* str = new StrCompressedCopyNode(control(), memory(idx), src, dst, count);
+  Node* res_mem = _gvn.transform(new SCMemProjNode(str));
+  set_memory(res_mem, idx);
+  return str;
+}
+
+void GraphKit::inflate_string(Node* src, Node* dst, Node* count) {
+  assert(Matcher::match_rule_supported(Op_StrInflatedCopy), "Intrinsic not supported");
+  uint idx = C->get_alias_index(TypeAryPtr::BYTES);
+  StrInflatedCopyNode* str = new StrInflatedCopyNode(control(), memory(idx), src, dst, count);
+  set_memory(_gvn.transform(str), idx);
+}
+
+void GraphKit::inflate_string_slow(Node* src, Node* dst, Node* start, Node* count) {
+  /**
+   * int i_char = start;
+   * for (int i_byte = 0; i_byte < count; i_byte++) {
+   *   dst[i_char++] = (char)(src[i_byte] & 0xff);
+   * }
+   */
+  add_predicate();
+  RegionNode* head = new RegionNode(3);
+  head->init_req(1, control());
+  gvn().set_type(head, Type::CONTROL);
+  record_for_igvn(head);
+
+  Node* i_byte = new PhiNode(head, TypeInt::INT);
+  i_byte->init_req(1, intcon(0));
+  gvn().set_type(i_byte, TypeInt::INT);
+  record_for_igvn(i_byte);
+
+  Node* i_char = new PhiNode(head, TypeInt::INT);
+  i_char->init_req(1, start);
+  gvn().set_type(i_char, TypeInt::INT);
+  record_for_igvn(i_char);
+
+  Node* mem = PhiNode::make(head, memory(TypeAryPtr::BYTES), Type::MEMORY, TypeAryPtr::BYTES);
+  gvn().set_type(mem, Type::MEMORY);
+  record_for_igvn(mem);
+  set_control(head);
+  set_memory(mem, TypeAryPtr::BYTES);
+  Node* ch = load_array_element(control(), src, i_byte, TypeAryPtr::BYTES);
+  Node* st = store_to_memory(control(), array_element_address(dst, i_char, T_BYTE),
+                             AndI(ch, intcon(0xff)), T_CHAR, TypeAryPtr::BYTES, MemNode::unordered);
+
+  IfNode* iff = create_and_map_if(head, Bool(CmpI(i_byte, count), BoolTest::lt), PROB_FAIR, COUNT_UNKNOWN);
+  head->init_req(2, IfTrue(iff));
+  mem->init_req(2, st);
+  i_byte->init_req(2, AddI(i_byte, intcon(1)));
+  i_char->init_req(2, AddI(i_char, intcon(2)));
+
+  set_control(IfFalse(iff));
+  set_memory(st, TypeAryPtr::BYTES);
 }
 
 Node* GraphKit::cast_array_to_stable(Node* ary, const TypeAryPtr* ary_type) {
--- a/hotspot/src/share/vm/opto/graphKit.hpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/opto/graphKit.hpp	Tue Nov 03 09:41:03 2015 +0100
@@ -864,12 +864,14 @@
                   bool deoptimize_on_exception = false);
 
   // java.lang.String helpers
-  Node* load_String_offset(Node* ctrl, Node* str);
   Node* load_String_length(Node* ctrl, Node* str);
   Node* load_String_value(Node* ctrl, Node* str);
-  void store_String_offset(Node* ctrl, Node* str, Node* value);
-  void store_String_length(Node* ctrl, Node* str, Node* value);
+  Node* load_String_coder(Node* ctrl, Node* str);
   void store_String_value(Node* ctrl, Node* str, Node* value);
+  void store_String_coder(Node* ctrl, Node* str, Node* value);
+  Node* compress_string(Node* src, Node* dst, Node* count);
+  void inflate_string(Node* src, Node* dst, Node* count);
+  void inflate_string_slow(Node* src, Node* dst, Node* start, Node* count);
 
   // Handy for making control flow
   IfNode* create_and_map_if(Node* ctrl, Node* tst, float prob, float cnt) {
--- a/hotspot/src/share/vm/opto/idealKit.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/opto/idealKit.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2005, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -530,7 +530,6 @@
          "call node must be constructed correctly");
 }
 
-
 void IdealKit::make_leaf_call_no_fp(const TypeFunc *slow_call_type,
                               address slow_call,
                               const char *leaf_name,
--- a/hotspot/src/share/vm/opto/idealKit.hpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/opto/idealKit.hpp	Tue Nov 03 09:41:03 2015 +0100
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2005, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -183,6 +183,7 @@
   Node* AddI(Node* l, Node* r) { return transform(new AddINode(l, r)); }
   Node* SubI(Node* l, Node* r) { return transform(new SubINode(l, r)); }
   Node* AndI(Node* l, Node* r) { return transform(new AndINode(l, r)); }
+  Node* OrI(Node* l, Node* r)  { return transform(new OrINode(l, r));  }
   Node* MaxI(Node* l, Node* r) { return transform(new MaxINode(l, r)); }
   Node* LShiftI(Node* l, Node* r) { return transform(new LShiftINode(l, r)); }
   Node* CmpI(Node* l, Node* r) { return transform(new CmpINode(l, r)); }
@@ -256,7 +257,6 @@
                             Node* parm1,
                             Node* parm2,
                             Node* parm3);
-
 };
 
 #endif // SHARE_VM_OPTO_IDEALKIT_HPP
--- a/hotspot/src/share/vm/opto/intrinsicnode.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/opto/intrinsicnode.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -36,7 +36,7 @@
 //------------------------------Ideal------------------------------------------
 // Return a node which is more "ideal" than the current node.  Strip out
 // control copies
-Node *StrIntrinsicNode::Ideal(PhaseGVN *phase, bool can_reshape) {
+Node* StrIntrinsicNode::Ideal(PhaseGVN* phase, bool can_reshape) {
   if (remove_dead_region(phase, can_reshape)) return this;
   // Don't bother trying to transform a dead node
   if (in(0) && in(0)->is_top())  return NULL;
@@ -55,11 +55,29 @@
 }
 
 //------------------------------Value------------------------------------------
-const Type *StrIntrinsicNode::Value( PhaseTransform *phase ) const {
+const Type* StrIntrinsicNode::Value(PhaseTransform* phase) const {
   if (in(0) && phase->type(in(0)) == Type::TOP) return Type::TOP;
   return bottom_type();
 }
 
+uint StrIntrinsicNode::size_of() const { return sizeof(*this); }
+
+//=============================================================================
+//------------------------------Ideal------------------------------------------
+// Return a node which is more "ideal" than the current node.  Strip out
+// control copies
+Node* StrCompressedCopyNode::Ideal(PhaseGVN* phase, bool can_reshape) {
+  return remove_dead_region(phase, can_reshape) ? this : NULL;
+}
+
+//=============================================================================
+//------------------------------Ideal------------------------------------------
+// Return a node which is more "ideal" than the current node.  Strip out
+// control copies
+Node* StrInflatedCopyNode::Ideal(PhaseGVN* phase, bool can_reshape) {
+  return remove_dead_region(phase, can_reshape) ? this : NULL;
+}
+
 //=============================================================================
 //------------------------------match_edge-------------------------------------
 // Do not match memory edge
@@ -70,12 +88,12 @@
 //------------------------------Ideal------------------------------------------
 // Return a node which is more "ideal" than the current node.  Strip out
 // control copies
-Node *EncodeISOArrayNode::Ideal(PhaseGVN *phase, bool can_reshape) {
+Node* EncodeISOArrayNode::Ideal(PhaseGVN* phase, bool can_reshape) {
   return remove_dead_region(phase, can_reshape) ? this : NULL;
 }
 
 //------------------------------Value------------------------------------------
-const Type *EncodeISOArrayNode::Value(PhaseTransform *phase) const {
+const Type* EncodeISOArrayNode::Value(PhaseTransform* phase) const {
   if (in(0) && phase->type(in(0)) == Type::TOP) return Type::TOP;
   return bottom_type();
 }
--- a/hotspot/src/share/vm/opto/intrinsicnode.hpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/opto/intrinsicnode.hpp	Tue Nov 03 09:41:03 2015 +0100
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -35,75 +35,131 @@
 // hit (cache is checked with exposed code in gen_subtype_check()).  Return
 // not zero for a miss or zero for a hit.
 class PartialSubtypeCheckNode : public Node {
-  public:
+ public:
   PartialSubtypeCheckNode(Node* c, Node* sub, Node* super) : Node(c,sub,super) {}
   virtual int Opcode() const;
-  virtual const Type *bottom_type() const { return TypeRawPtr::BOTTOM; }
+  virtual const Type* bottom_type() const { return TypeRawPtr::BOTTOM; }
   virtual uint ideal_reg() const { return Op_RegP; }
 };
 
 //------------------------------StrIntrinsic-------------------------------
-// Base class for Ideal nodes used in String instrinsic code.
+// Base class for Ideal nodes used in String intrinsic code.
 class StrIntrinsicNode: public Node {
-  public:
+ public:
+  // Possible encodings of the two parameters passed to the string intrinsic.
+  // 'L' stands for Latin1 and 'U' stands for UTF16. For example, 'LU' means that
+  // the first string is Latin1 encoded and the second string is UTF16 encoded.
+  typedef enum ArgEncoding { LL, LU, UL, UU, none } ArgEnc;
+
+ protected:
+  // Encoding of strings. Used to select the right version of the intrinsic.
+  const ArgEncoding _encoding;
+  virtual uint size_of() const;
+
+ public:
   StrIntrinsicNode(Node* control, Node* char_array_mem,
-                   Node* s1, Node* c1, Node* s2, Node* c2):
-  Node(control, char_array_mem, s1, c1, s2, c2) {
+                   Node* s1, Node* c1, Node* s2, Node* c2, ArgEncoding encoding):
+  Node(control, char_array_mem, s1, c1, s2, c2), _encoding(encoding) {
   }
 
   StrIntrinsicNode(Node* control, Node* char_array_mem,
-                   Node* s1, Node* s2, Node* c):
-  Node(control, char_array_mem, s1, s2, c) {
+                   Node* s1, Node* s2, Node* c, ArgEncoding encoding):
+  Node(control, char_array_mem, s1, s2, c), _encoding(encoding) {
   }
 
   StrIntrinsicNode(Node* control, Node* char_array_mem,
-                   Node* s1, Node* s2):
-  Node(control, char_array_mem, s1, s2) {
+                   Node* s1, Node* s2, ArgEncoding encoding):
+  Node(control, char_array_mem, s1, s2), _encoding(encoding) {
   }
 
   virtual bool depends_only_on_test() const { return false; }
-  virtual const TypePtr* adr_type() const { return TypeAryPtr::CHARS; }
+  virtual const TypePtr* adr_type() const { return TypeAryPtr::BYTES; }
   virtual uint match_edge(uint idx) const;
   virtual uint ideal_reg() const { return Op_RegI; }
-  virtual Node *Ideal(PhaseGVN *phase, bool can_reshape);
-  virtual const Type *Value(PhaseTransform *phase) const;
+  virtual Node* Ideal(PhaseGVN* phase, bool can_reshape);
+  virtual const Type* Value(PhaseTransform* phase) const;
+  ArgEncoding encoding() const { return _encoding; }
 };
 
 //------------------------------StrComp-------------------------------------
 class StrCompNode: public StrIntrinsicNode {
-  public:
+ public:
   StrCompNode(Node* control, Node* char_array_mem,
-              Node* s1, Node* c1, Node* s2, Node* c2):
-  StrIntrinsicNode(control, char_array_mem, s1, c1, s2, c2) {};
+              Node* s1, Node* c1, Node* s2, Node* c2, ArgEncoding encoding):
+  StrIntrinsicNode(control, char_array_mem, s1, c1, s2, c2, encoding) {};
   virtual int Opcode() const;
   virtual const Type* bottom_type() const { return TypeInt::INT; }
 };
 
 //------------------------------StrEquals-------------------------------------
 class StrEqualsNode: public StrIntrinsicNode {
-  public:
+ public:
   StrEqualsNode(Node* control, Node* char_array_mem,
-                Node* s1, Node* s2, Node* c):
-  StrIntrinsicNode(control, char_array_mem, s1, s2, c) {};
+                Node* s1, Node* s2, Node* c, ArgEncoding encoding):
+  StrIntrinsicNode(control, char_array_mem, s1, s2, c, encoding) {};
   virtual int Opcode() const;
   virtual const Type* bottom_type() const { return TypeInt::BOOL; }
 };
 
 //------------------------------StrIndexOf-------------------------------------
 class StrIndexOfNode: public StrIntrinsicNode {
-  public:
+ public:
   StrIndexOfNode(Node* control, Node* char_array_mem,
-                 Node* s1, Node* c1, Node* s2, Node* c2):
-  StrIntrinsicNode(control, char_array_mem, s1, c1, s2, c2) {};
+                 Node* s1, Node* c1, Node* s2, Node* c2, ArgEncoding encoding):
+  StrIntrinsicNode(control, char_array_mem, s1, c1, s2, c2, encoding) {};
+  virtual int Opcode() const;
+  virtual const Type* bottom_type() const { return TypeInt::INT; }
+};
+
+//------------------------------StrIndexOfChar-------------------------------------
+class StrIndexOfCharNode: public StrIntrinsicNode {
+ public:
+  StrIndexOfCharNode(Node* control, Node* char_array_mem,
+                     Node* s1, Node* c1, Node* c, ArgEncoding encoding):
+  StrIntrinsicNode(control, char_array_mem, s1, c1, c, encoding) {};
   virtual int Opcode() const;
   virtual const Type* bottom_type() const { return TypeInt::INT; }
 };
 
+//--------------------------StrCompressedCopy-------------------------------
+class StrCompressedCopyNode: public StrIntrinsicNode {
+ public:
+  StrCompressedCopyNode(Node* control, Node* arymem,
+                        Node* s1, Node* s2, Node* c):
+  StrIntrinsicNode(control, arymem, s1, s2, c, none) {};
+  virtual int Opcode() const;
+  virtual const Type* bottom_type() const { return TypeInt::INT; }
+  virtual const TypePtr* adr_type() const { return TypePtr::BOTTOM; }
+  virtual Node* Ideal(PhaseGVN* phase, bool can_reshape);
+};
+
+//--------------------------StrInflatedCopy---------------------------------
+class StrInflatedCopyNode: public StrIntrinsicNode {
+ public:
+  StrInflatedCopyNode(Node* control, Node* arymem,
+                      Node* s1, Node* s2, Node* c):
+  StrIntrinsicNode(control, arymem, s1, s2, c, none) {};
+  virtual int Opcode() const;
+  virtual const Type* bottom_type() const { return Type::MEMORY; }
+  virtual const TypePtr* adr_type() const { return TypePtr::BOTTOM; }
+  virtual Node* Ideal(PhaseGVN* phase, bool can_reshape);
+};
+
 //------------------------------AryEq---------------------------------------
 class AryEqNode: public StrIntrinsicNode {
-  public:
-  AryEqNode(Node* control, Node* char_array_mem, Node* s1, Node* s2):
-  StrIntrinsicNode(control, char_array_mem, s1, s2) {};
+ public:
+  AryEqNode(Node* control, Node* char_array_mem,
+            Node* s1, Node* s2, ArgEncoding encoding):
+  StrIntrinsicNode(control, char_array_mem, s1, s2, encoding) {};
+  virtual int Opcode() const;
+  virtual const Type* bottom_type() const { return TypeInt::BOOL; }
+};
+
+//------------------------------HasNegatives---------------------------------
+class HasNegativesNode: public StrIntrinsicNode {
+ public:
+  HasNegativesNode(Node* control, Node* char_array_mem, Node* s1, Node* c1):
+  StrIntrinsicNode(control, char_array_mem, s1, c1, none) {};
   virtual int Opcode() const;
   virtual const Type* bottom_type() const { return TypeInt::BOOL; }
 };
@@ -112,16 +168,16 @@
 //------------------------------EncodeISOArray--------------------------------
 // encode char[] to byte[] in ISO_8859_1
 class EncodeISOArrayNode: public Node {
-  public:
-  EncodeISOArrayNode(Node *control, Node* arymem, Node* s1, Node* s2, Node* c): Node(control, arymem, s1, s2, c) {};
+ public:
+  EncodeISOArrayNode(Node* control, Node* arymem, Node* s1, Node* s2, Node* c): Node(control, arymem, s1, s2, c) {};
   virtual int Opcode() const;
   virtual bool depends_only_on_test() const { return false; }
   virtual const Type* bottom_type() const { return TypeInt::INT; }
   virtual const TypePtr* adr_type() const { return TypePtr::BOTTOM; }
   virtual uint match_edge(uint idx) const;
   virtual uint ideal_reg() const { return Op_RegI; }
-  virtual Node *Ideal(PhaseGVN *phase, bool can_reshape);
-  virtual const Type *Value(PhaseTransform *phase) const;
+  virtual Node* Ideal(PhaseGVN* phase, bool can_reshape);
+  virtual const Type* Value(PhaseTransform* phase) const;
 };
 
 #endif // SHARE_VM_OPTO_INTRINSICNODE_HPP
--- a/hotspot/src/share/vm/opto/lcm.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/opto/lcm.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -196,8 +196,12 @@
     case Op_StrComp:
     case Op_StrEquals:
     case Op_StrIndexOf:
+    case Op_StrIndexOfChar:
     case Op_AryEq:
+    case Op_StrInflatedCopy:
+    case Op_StrCompressedCopy:
     case Op_EncodeISOArray:
+    case Op_HasNegatives:
       // Not a legit memory op for implicit null check regardless of
       // embedded loads
       continue;
--- a/hotspot/src/share/vm/opto/library_call.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/opto/library_call.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -200,12 +200,16 @@
   }
   Node * load_field_from_object(Node * fromObj, const char * fieldName, const char * fieldTypeString, bool is_exact, bool is_static, ciInstanceKlass * fromKls);
 
-  Node* make_string_method_node(int opcode, Node* str1_start, Node* cnt1, Node* str2_start, Node* cnt2);
-  Node* make_string_method_node(int opcode, Node* str1, Node* str2);
-  bool inline_string_compareTo();
-  bool inline_string_indexOf();
-  Node* string_indexOf(Node* string_object, ciTypeArray* target_array, jint offset, jint cache_i, jint md2_i);
-  bool inline_string_equals();
+  Node* make_string_method_node(int opcode, Node* str1_start, Node* cnt1, Node* str2_start, Node* cnt2, StrIntrinsicNode::ArgEnc ae);
+  bool inline_string_compareTo(StrIntrinsicNode::ArgEnc ae);
+  bool inline_string_indexOf(StrIntrinsicNode::ArgEnc ae);
+  bool inline_string_indexOfI(StrIntrinsicNode::ArgEnc ae);
+  bool inline_string_indexOfChar();
+  bool inline_string_equals(StrIntrinsicNode::ArgEnc ae);
+  bool inline_string_toBytesU();
+  bool inline_string_getCharsU();
+  bool inline_string_copy(bool compress);
+  bool inline_string_char_access(bool is_store);
   Node* round_double_node(Node* n);
   bool runtime_math(const TypeFunc* call_type, address funcAddr, const char* funcName);
   bool inline_math_native(vmIntrinsics::ID id);
@@ -251,7 +255,7 @@
   bool inline_native_newArray();
   bool inline_native_getLength();
   bool inline_array_copyOf(bool is_copyOfRange);
-  bool inline_array_equals();
+  bool inline_array_equals(StrIntrinsicNode::ArgEnc ae);
   void copy_to_clone(Node* obj, Node* alloc_obj, Node* obj_size, bool is_array, bool card_mark);
   bool inline_native_clone(bool is_virtual);
   bool inline_native_Reflection_getCallerClass();
@@ -298,6 +302,7 @@
   bool inline_updateBytesAdler32();
   bool inline_updateByteBufferAdler32();
   bool inline_multiplyToLen();
+  bool inline_hasNegatives();
   bool inline_squareToLen();
   bool inline_mulAdd();
   bool inline_montgomeryMultiply();
@@ -458,6 +463,7 @@
 bool LibraryCallKit::try_to_inline(int predicate) {
   // Handle symbolic names for otherwise undistinguished boolean switches:
   const bool is_store       = true;
+  const bool is_compress    = true;
   const bool is_native_ptr  = true;
   const bool is_static      = true;
   const bool is_volatile    = true;
@@ -512,9 +518,31 @@
 
   case vmIntrinsics::_arraycopy:                return inline_arraycopy();
 
-  case vmIntrinsics::_compareTo:                return inline_string_compareTo();
-  case vmIntrinsics::_indexOf:                  return inline_string_indexOf();
-  case vmIntrinsics::_equals:                   return inline_string_equals();
+  case vmIntrinsics::_compareToL:               return inline_string_compareTo(StrIntrinsicNode::LL);
+  case vmIntrinsics::_compareToU:               return inline_string_compareTo(StrIntrinsicNode::UU);
+  case vmIntrinsics::_compareToLU:              return inline_string_compareTo(StrIntrinsicNode::LU);
+  case vmIntrinsics::_compareToUL:              return inline_string_compareTo(StrIntrinsicNode::UL);
+
+  case vmIntrinsics::_indexOfL:                 return inline_string_indexOf(StrIntrinsicNode::LL);
+  case vmIntrinsics::_indexOfU:                 return inline_string_indexOf(StrIntrinsicNode::UU);
+  case vmIntrinsics::_indexOfUL:                return inline_string_indexOf(StrIntrinsicNode::UL);
+  case vmIntrinsics::_indexOfIL:                return inline_string_indexOfI(StrIntrinsicNode::LL);
+  case vmIntrinsics::_indexOfIU:                return inline_string_indexOfI(StrIntrinsicNode::UU);
+  case vmIntrinsics::_indexOfIUL:               return inline_string_indexOfI(StrIntrinsicNode::UL);
+  case vmIntrinsics::_indexOfU_char:            return inline_string_indexOfChar();
+
+  case vmIntrinsics::_equalsL:                  return inline_string_equals(StrIntrinsicNode::LL);
+  case vmIntrinsics::_equalsU:                  return inline_string_equals(StrIntrinsicNode::UU);
+
+  case vmIntrinsics::_toBytesStringU:           return inline_string_toBytesU();
+  case vmIntrinsics::_getCharsStringU:          return inline_string_getCharsU();
+  case vmIntrinsics::_getCharStringU:           return inline_string_char_access(!is_store);
+  case vmIntrinsics::_putCharStringU:           return inline_string_char_access( is_store);
+
+  case vmIntrinsics::_compressStringC:
+  case vmIntrinsics::_compressStringB:          return inline_string_copy( is_compress);
+  case vmIntrinsics::_inflateStringC:
+  case vmIntrinsics::_inflateStringB:           return inline_string_copy(!is_compress);
 
   case vmIntrinsics::_getObject:                return inline_unsafe_access(!is_native_ptr, !is_store, T_OBJECT,  !is_volatile);
   case vmIntrinsics::_getBoolean:               return inline_unsafe_access(!is_native_ptr, !is_store, T_BOOLEAN, !is_volatile);
@@ -617,7 +645,8 @@
   case vmIntrinsics::_getLength:                return inline_native_getLength();
   case vmIntrinsics::_copyOf:                   return inline_array_copyOf(false);
   case vmIntrinsics::_copyOfRange:              return inline_array_copyOf(true);
-  case vmIntrinsics::_equalsC:                  return inline_array_equals();
+  case vmIntrinsics::_equalsB:                  return inline_array_equals(StrIntrinsicNode::LL);
+  case vmIntrinsics::_equalsC:                  return inline_array_equals(StrIntrinsicNode::UU);
   case vmIntrinsics::_clone:                    return inline_native_clone(intrinsic()->is_virtual());
 
   case vmIntrinsics::_isAssignableFrom:         return inline_native_subtype_check();
@@ -687,6 +716,7 @@
     return inline_ghash_processBlocks();
 
   case vmIntrinsics::_encodeISOArray:
+  case vmIntrinsics::_encodeByteISOArray:
     return inline_encodeISOArray();
 
   case vmIntrinsics::_updateCRC32:
@@ -711,6 +741,9 @@
   case vmIntrinsics::_isCompileConstant:
     return inline_isCompileConstant();
 
+  case vmIntrinsics::_hasNegatives:
+    return inline_hasNegatives();
+
   default:
     // If you get here, it may be that someone has added a new intrinsic
     // to the list in vmSymbols.hpp without implementing it here.
@@ -876,75 +909,24 @@
 
 
 //------------------------------make_string_method_node------------------------
-// Helper method for String intrinsic functions. This version is called
-// with str1 and str2 pointing to String object nodes.
-//
-Node* LibraryCallKit::make_string_method_node(int opcode, Node* str1, Node* str2) {
-  Node* no_ctrl = NULL;
-
-  // Get start addr of string
-  Node* str1_value   = load_String_value(no_ctrl, str1);
-  Node* str1_offset  = load_String_offset(no_ctrl, str1);
-  Node* str1_start   = array_element_address(str1_value, str1_offset, T_CHAR);
-
-  // Get length of string 1
-  Node* str1_len  = load_String_length(no_ctrl, str1);
-
-  Node* str2_value   = load_String_value(no_ctrl, str2);
-  Node* str2_offset  = load_String_offset(no_ctrl, str2);
-  Node* str2_start   = array_element_address(str2_value, str2_offset, T_CHAR);
-
-  Node* str2_len = NULL;
-  Node* result = NULL;
-
-  switch (opcode) {
-  case Op_StrIndexOf:
-    // Get length of string 2
-    str2_len = load_String_length(no_ctrl, str2);
-
-    result = new StrIndexOfNode(control(), memory(TypeAryPtr::CHARS),
-                                str1_start, str1_len, str2_start, str2_len);
-    break;
-  case Op_StrComp:
-    // Get length of string 2
-    str2_len = load_String_length(no_ctrl, str2);
-
-    result = new StrCompNode(control(), memory(TypeAryPtr::CHARS),
-                             str1_start, str1_len, str2_start, str2_len);
-    break;
-  case Op_StrEquals:
-    result = new StrEqualsNode(control(), memory(TypeAryPtr::CHARS),
-                               str1_start, str2_start, str1_len);
-    break;
-  default:
-    ShouldNotReachHere();
-    return NULL;
-  }
-
-  // All these intrinsics have checks.
-  C->set_has_split_ifs(true); // Has chance for split-if optimization
-
-  return _gvn.transform(result);
-}
-
-// Helper method for String intrinsic functions. This version is called
-// with str1 and str2 pointing to char[] nodes, with cnt1 and cnt2 pointing
-// to Int nodes containing the lenghts of str1 and str2.
-//
-Node* LibraryCallKit::make_string_method_node(int opcode, Node* str1_start, Node* cnt1, Node* str2_start, Node* cnt2) {
+// Helper method for String intrinsic functions. This version is called with
+// str1 and str2 pointing to byte[] nodes containing Latin1 or UTF16 encoded
+// characters (depending on 'is_byte'). cnt1 and cnt2 are pointing to Int nodes
+// containing the lengths of str1 and str2.
+Node* LibraryCallKit::make_string_method_node(int opcode, Node* str1_start, Node* cnt1, Node* str2_start, Node* cnt2, StrIntrinsicNode::ArgEnc ae) {
   Node* result = NULL;
   switch (opcode) {
   case Op_StrIndexOf:
-    result = new StrIndexOfNode(control(), memory(TypeAryPtr::CHARS),
-                                str1_start, cnt1, str2_start, cnt2);
+    result = new StrIndexOfNode(control(), memory(TypeAryPtr::BYTES),
+                                str1_start, cnt1, str2_start, cnt2, ae);
     break;
   case Op_StrComp:
-    result = new StrCompNode(control(), memory(TypeAryPtr::CHARS),
-                             str1_start, cnt1, str2_start, cnt2);
+    result = new StrCompNode(control(), memory(TypeAryPtr::BYTES),
+                             str1_start, cnt1, str2_start, cnt2, ae);
     break;
   case Op_StrEquals:
-    result = new StrEqualsNode(control(), memory(TypeAryPtr::CHARS),
-                               str1_start, str2_start, cnt1);
+    result = new StrEqualsNode(control(), memory(TypeAryPtr::BYTES),
+                               str1_start, str2_start, cnt1, ae);
     break;
   default:
     ShouldNotReachHere();
@@ -958,98 +940,54 @@
 }
 
 //------------------------------inline_string_compareTo------------------------
-// public int java.lang.String.compareTo(String anotherString);
-bool LibraryCallKit::inline_string_compareTo() {
-  Node* receiver = null_check(argument(0));
-  Node* arg      = null_check(argument(1));
-  if (stopped()) {
-    return true;
-  }
-  set_result(make_string_method_node(Op_StrComp, receiver, arg));
+bool LibraryCallKit::inline_string_compareTo(StrIntrinsicNode::ArgEnc ae) {
+  Node* arg1 = argument(0);
+  Node* arg2 = argument(1);
+
+  // Get start addr and length of first argument
+  Node* arg1_start  = array_element_address(arg1, intcon(0), T_BYTE);
+  Node* arg1_cnt    = load_array_length(arg1);
+
+  // Get start addr and length of second argument
+  Node* arg2_start  = array_element_address(arg2, intcon(0), T_BYTE);
+  Node* arg2_cnt    = load_array_length(arg2);
+
+  Node* result = make_string_method_node(Op_StrComp, arg1_start, arg1_cnt, arg2_start, arg2_cnt, ae);
+  set_result(result);
   return true;
 }
 
 //------------------------------inline_string_equals------------------------
-bool LibraryCallKit::inline_string_equals() {
-  Node* receiver = null_check_receiver();
-  // NOTE: Do not null check argument for String.equals() because spec
-  // allows to specify NULL as argument.
-  Node* argument = this->argument(1);
-  if (stopped()) {
-    return true;
-  }
+bool LibraryCallKit::inline_string_equals(StrIntrinsicNode::ArgEnc ae) {
+  Node* arg1 = argument(0);
+  Node* arg2 = argument(1);
 
   // paths (plus control) merge
-  RegionNode* region = new RegionNode(5);
+  RegionNode* region = new RegionNode(3);
   Node* phi = new PhiNode(region, TypeInt::BOOL);
 
-  // does source == target string?
-  Node* cmp = _gvn.transform(new CmpPNode(receiver, argument));
-  Node* bol = _gvn.transform(new BoolNode(cmp, BoolTest::eq));
-
-  Node* if_eq = generate_slow_guard(bol, NULL);
-  if (if_eq != NULL) {
-    // receiver == argument
-    phi->init_req(2, intcon(1));
-    region->init_req(2, if_eq);
-  }
-
-  // get String klass for instanceOf
-  ciInstanceKlass* klass = env()->String_klass();
-
   if (!stopped()) {
-    Node* inst = gen_instanceof(argument, makecon(TypeKlassPtr::make(klass)));
-    Node* cmp  = _gvn.transform(new CmpINode(inst, intcon(1)));
-    Node* bol  = _gvn.transform(new BoolNode(cmp, BoolTest::ne));
-
-    Node* inst_false = generate_guard(bol, NULL, PROB_MIN);
-    //instanceOf == true, fallthrough
-
-    if (inst_false != NULL) {
-      phi->init_req(3, intcon(0));
-      region->init_req(3, inst_false);
-    }
-  }
-
-  if (!stopped()) {
-    const TypeOopPtr* string_type = TypeOopPtr::make_from_klass(klass);
-
-    // Properly cast the argument to String
-    argument = _gvn.transform(new CheckCastPPNode(control(), argument, string_type));
-    // This path is taken only when argument's type is String:NotNull.
-    argument = cast_not_null(argument, false);
-
-    Node* no_ctrl = NULL;
-
-    // Get start addr of receiver
-    Node* receiver_val    = load_String_value(no_ctrl, receiver);
-    Node* receiver_offset = load_String_offset(no_ctrl, receiver);
-    Node* receiver_start = array_element_address(receiver_val, receiver_offset, T_CHAR);
-
-    // Get length of receiver
-    Node* receiver_cnt  = load_String_length(no_ctrl, receiver);
-
-    // Get start addr of argument
-    Node* argument_val    = load_String_value(no_ctrl, argument);
-    Node* argument_offset = load_String_offset(no_ctrl, argument);
-    Node* argument_start = array_element_address(argument_val, argument_offset, T_CHAR);
-
-    // Get length of argument
-    Node* argument_cnt  = load_String_length(no_ctrl, argument);
-
-    // Check for receiver count != argument count
-    Node* cmp = _gvn.transform(new CmpINode(receiver_cnt, argument_cnt));
+    // Get start addr and length of first argument
+    Node* arg1_start  = array_element_address(arg1, intcon(0), T_BYTE);
+    Node* arg1_cnt    = load_array_length(arg1);
+
+    // Get start addr and length of second argument
+    Node* arg2_start  = array_element_address(arg2, intcon(0), T_BYTE);
+    Node* arg2_cnt    = load_array_length(arg2);
+
+    // Check for arg1_cnt != arg2_cnt
+    Node* cmp = _gvn.transform(new CmpINode(arg1_cnt, arg2_cnt));
     Node* bol = _gvn.transform(new BoolNode(cmp, BoolTest::ne));
     Node* if_ne = generate_slow_guard(bol, NULL);
     if (if_ne != NULL) {
-      phi->init_req(4, intcon(0));
-      region->init_req(4, if_ne);
+      phi->init_req(2, intcon(0));
+      region->init_req(2, if_ne);
     }
 
     // Check for count == 0 is done by assembler code for StrEquals.
 
     if (!stopped()) {
-      Node* equals = make_string_method_node(Op_StrEquals, receiver_start, receiver_cnt, argument_start, argument_cnt);
+      Node* equals = make_string_method_node(Op_StrEquals, arg1_start, arg1_cnt, arg2_start, arg2_cnt, ae);
       phi->init_req(1, equals);
       region->init_req(1, control());
     }
@@ -1064,289 +1002,462 @@
 }
 
 //------------------------------inline_array_equals----------------------------
-bool LibraryCallKit::inline_array_equals() {
+bool LibraryCallKit::inline_array_equals(StrIntrinsicNode::ArgEnc ae) {
+  assert(ae == StrIntrinsicNode::UU || ae == StrIntrinsicNode::LL, "unsupported array types");
   Node* arg1 = argument(0);
   Node* arg2 = argument(1);
-  set_result(_gvn.transform(new AryEqNode(control(), memory(TypeAryPtr::CHARS), arg1, arg2)));
+
+  const TypeAryPtr* mtype = (ae == StrIntrinsicNode::UU) ? TypeAryPtr::CHARS : TypeAryPtr::BYTES;
+  set_result(_gvn.transform(new AryEqNode(control(), memory(mtype), arg1, arg2, ae)));
   return true;
 }
 
-// Java version of String.indexOf(constant string)
-// class StringDecl {
-//   StringDecl(char[] ca) {
-//     offset = 0;
-//     count = ca.length;
-//     value = ca;
-//   }
-//   int offset;
-//   int count;
-//   char[] value;
-// }
-//
-// static int string_indexOf_J(StringDecl string_object, char[] target_object,
-//                             int targetOffset, int cache_i, int md2) {
-//   int cache = cache_i;
-//   int sourceOffset = string_object.offset;
-//   int sourceCount = string_object.count;
-//   int targetCount = target_object.length;
-//
-//   int targetCountLess1 = targetCount - 1;
-//   int sourceEnd = sourceOffset + sourceCount - targetCountLess1;
-//
-//   char[] source = string_object.value;
-//   char[] target = target_object;
-//   int lastChar = target[targetCountLess1];
-//
-//  outer_loop:
-//   for (int i = sourceOffset; i < sourceEnd; ) {
-//     int src = source[i + targetCountLess1];
-//     if (src == lastChar) {
-//       // With random strings and a 4-character alphabet,
-//       // reverse matching at this point sets up 0.8% fewer
-//       // frames, but (paradoxically) makes 0.3% more probes.
-//       // Since those probes are nearer the lastChar probe,
-//       // there is may be a net D$ win with reverse matching.
-//       // But, reversing loop inhibits unroll of inner loop
-//       // for unknown reason.  So, does running outer loop from
-//       // (sourceOffset - targetCountLess1) to (sourceOffset + sourceCount)
-//       for (int j = 0; j < targetCountLess1; j++) {
-//         if (target[targetOffset + j] != source[i+j]) {
-//           if ((cache & (1 << source[i+j])) == 0) {
-//             if (md2 < j+1) {
-//               i += j+1;
-//               continue outer_loop;
-//             }
-//           }
-//           i += md2;
-//           continue outer_loop;
-//         }
-//       }
-//       return i - sourceOffset;
-//     }
-//     if ((cache & (1 << src)) == 0) {
-//       i += targetCountLess1;
-//     } // using "i += targetCount;" and an "else i++;" causes a jump to jump.
-//     i++;
-//   }
-//   return -1;
-// }
-
-//------------------------------string_indexOf------------------------
-Node* LibraryCallKit::string_indexOf(Node* string_object, ciTypeArray* target_array, jint targetOffset_i,
-                                     jint cache_i, jint md2_i) {
-
-  Node* no_ctrl  = NULL;
-  float likely   = PROB_LIKELY(0.9);
-  float unlikely = PROB_UNLIKELY(0.9);
-
-  const int nargs = 0; // no arguments to push back for uncommon trap in predicate
-
-  Node* source        = load_String_value(no_ctrl, string_object);
-  Node* sourceOffset  = load_String_offset(no_ctrl, string_object);
-  Node* sourceCount   = load_String_length(no_ctrl, string_object);
-
-  Node* target = _gvn.transform( makecon(TypeOopPtr::make_from_constant(target_array, true)));
-  jint target_length = target_array->length();
-  const TypeAry* target_array_type = TypeAry::make(TypeInt::CHAR, TypeInt::make(0, target_length, Type::WidenMin));
-  const TypeAryPtr* target_type = TypeAryPtr::make(TypePtr::BotPTR, target_array_type, target_array->klass(), true, Type::OffsetBot);
-
-  // String.value field is known to be @Stable.
-  if (UseImplicitStableValues) {
-    target = cast_array_to_stable(target, target_type);
-  }
-
-  IdealKit kit(this, false, true);
-#define __ kit.
-  Node* zero             = __ ConI(0);
-  Node* one              = __ ConI(1);
-  Node* cache            = __ ConI(cache_i);
-  Node* md2              = __ ConI(md2_i);
-  Node* lastChar         = __ ConI(target_array->char_at(target_length - 1));
-  Node* targetCountLess1 = __ ConI(target_length - 1);
-  Node* targetOffset     = __ ConI(targetOffset_i);
-  Node* sourceEnd        = __ SubI(__ AddI(sourceOffset, sourceCount), targetCountLess1);
-
-  IdealVariable rtn(kit), i(kit), j(kit); __ declarations_done();
-  Node* outer_loop = __ make_label(2 /* goto */);
-  Node* return_    = __ make_label(1);
-
-  __ set(rtn,__ ConI(-1));
-  __ loop(this, nargs, i, sourceOffset, BoolTest::lt, sourceEnd); {
-       Node* i2  = __ AddI(__ value(i), targetCountLess1);
-       // pin to prohibit loading of "next iteration" value which may SEGV (rare)
-       Node* src = load_array_element(__ ctrl(), source, i2, TypeAryPtr::CHARS);
-       __ if_then(src, BoolTest::eq, lastChar, unlikely); {
-         __ loop(this, nargs, j, zero, BoolTest::lt, targetCountLess1); {
-              Node* tpj = __ AddI(targetOffset, __ value(j));
-              Node* targ = load_array_element(no_ctrl, target, tpj, target_type);
-              Node* ipj  = __ AddI(__ value(i), __ value(j));
-              Node* src2 = load_array_element(no_ctrl, source, ipj, TypeAryPtr::CHARS);
-              __ if_then(targ, BoolTest::ne, src2); {
-                __ if_then(__ AndI(cache, __ LShiftI(one, src2)), BoolTest::eq, zero); {
-                  __ if_then(md2, BoolTest::lt, __ AddI(__ value(j), one)); {
-                    __ increment(i, __ AddI(__ value(j), one));
-                    __ goto_(outer_loop);
-                  } __ end_if(); __ dead(j);
-                }__ end_if(); __ dead(j);
-                __ increment(i, md2);
-                __ goto_(outer_loop);
-              }__ end_if();
-              __ increment(j, one);
-         }__ end_loop(); __ dead(j);
-         __ set(rtn, __ SubI(__ value(i), sourceOffset)); __ dead(i);
-         __ goto_(return_);
-       }__ end_if();
-       __ if_then(__ AndI(cache, __ LShiftI(one, src)), BoolTest::eq, zero, likely); {
-         __ increment(i, targetCountLess1);
-       }__ end_if();
-       __ increment(i, one);
-       __ bind(outer_loop);
-  }__ end_loop(); __ dead(i);
-  __ bind(return_);
-
-  // Final sync IdealKit and GraphKit.
-  final_sync(kit);
-  Node* result = __ value(rtn);
-#undef __
-  C->set_has_loops(true);
-  return result;
+//------------------------------inline_hasNegatives------------------------------
+bool LibraryCallKit::inline_hasNegatives() {
+  if (too_many_traps(Deoptimization::Reason_intrinsic))  return false;
+
+  assert(callee()->signature()->size() == 3, "hasNegatives has 3 parameters");
+  // no receiver since it is static method
+  Node* ba         = argument(0);
+  Node* offset     = argument(1);
+  Node* len        = argument(2);
+
+  RegionNode* bailout = new RegionNode(1);
+  record_for_igvn(bailout);
+
+  // offset must not be negative.
+  generate_negative_guard(offset, bailout);
+
+  // offset + length must not exceed length of ba.
+  generate_limit_guard(offset, len, load_array_length(ba), bailout);
+
+  if (bailout->req() > 1) {
+    PreserveJVMState pjvms(this);
+    set_control(_gvn.transform(bailout));
+    uncommon_trap(Deoptimization::Reason_intrinsic,
+                  Deoptimization::Action_maybe_recompile);
+  }
+  if (!stopped()) {
+    Node* ba_start = array_element_address(ba, offset, T_BYTE);
+    Node* result = new HasNegativesNode(control(), memory(TypeAryPtr::BYTES), ba_start, len);
+    set_result(_gvn.transform(result));
+  }
+  return true;
 }
 
 //------------------------------inline_string_indexOf------------------------
-bool LibraryCallKit::inline_string_indexOf() {
-  Node* receiver = argument(0);
-  Node* arg      = argument(1);
-
-  Node* result;
-  if (Matcher::has_match_rule(Op_StrIndexOf) &&
-      UseSSE42Intrinsics) {
-    // Generate SSE4.2 version of indexOf
-    // We currently only have match rules that use SSE4.2
-
-    receiver = null_check(receiver);
-    arg      = null_check(arg);
-    if (stopped()) {
-      return true;
-    }
-
-    // Make the merge point
-    RegionNode* result_rgn = new RegionNode(4);
-    Node*       result_phi = new PhiNode(result_rgn, TypeInt::INT);
-    Node* no_ctrl  = NULL;
-
-    // Get start addr of source string
-    Node* source = load_String_value(no_ctrl, receiver);
-    Node* source_offset = load_String_offset(no_ctrl, receiver);
-    Node* source_start = array_element_address(source, source_offset, T_CHAR);
-
-    // Get length of source string
-    Node* source_cnt  = load_String_length(no_ctrl, receiver);
-
-    // Get start addr of substring
-    Node* substr = load_String_value(no_ctrl, arg);
-    Node* substr_offset = load_String_offset(no_ctrl, arg);
-    Node* substr_start = array_element_address(substr, substr_offset, T_CHAR);
-
-    // Get length of source string
-    Node* substr_cnt  = load_String_length(no_ctrl, arg);
-
-    // Check for substr count > string count
-    Node* cmp = _gvn.transform(new CmpINode(substr_cnt, source_cnt));
-    Node* bol = _gvn.transform(new BoolNode(cmp, BoolTest::gt));
-    Node* if_gt = generate_slow_guard(bol, NULL);
-    if (if_gt != NULL) {
-      result_phi->init_req(2, intcon(-1));
-      result_rgn->init_req(2, if_gt);
-    }
-
-    if (!stopped()) {
-      // Check for substr count == 0
-      cmp = _gvn.transform(new CmpINode(substr_cnt, intcon(0)));
-      bol = _gvn.transform(new BoolNode(cmp, BoolTest::eq));
-      Node* if_zero = generate_slow_guard(bol, NULL);
-      if (if_zero != NULL) {
-        result_phi->init_req(3, intcon(0));
-        result_rgn->init_req(3, if_zero);
-      }
-    }
-
-    if (!stopped()) {
-      result = make_string_method_node(Op_StrIndexOf, source_start, source_cnt, substr_start, substr_cnt);
-      result_phi->init_req(1, result);
-      result_rgn->init_req(1, control());
+bool LibraryCallKit::inline_string_indexOf(StrIntrinsicNode::ArgEnc ae) {
+  if (!Matcher::has_match_rule(Op_StrIndexOf) || !UseSSE42Intrinsics) {
+    return false;
+  }
+  Node* src = argument(0);
+  Node* tgt = argument(1);
+
+  // Make the merge point
+  RegionNode* result_rgn = new RegionNode(4);
+  Node*       result_phi = new PhiNode(result_rgn, TypeInt::INT);
+
+  // Get start addr and length of source string
+  Node* src_start = array_element_address(src, intcon(0), T_BYTE);
+  Node* src_count = load_array_length(src);
+
+  // Get start addr and length of substring
+  Node* tgt_start = array_element_address(tgt, intcon(0), T_BYTE);
+  Node* tgt_count = load_array_length(tgt);
+
+  if (ae == StrIntrinsicNode::UU || ae == StrIntrinsicNode::UL) {
+    // Divide src size by 2 if String is UTF16 encoded
+    src_count = _gvn.transform(new RShiftINode(src_count, intcon(1)));
+  }
+  if (ae == StrIntrinsicNode::UU) {
+    // Divide substring size by 2 if String is UTF16 encoded
+    tgt_count = _gvn.transform(new RShiftINode(tgt_count, intcon(1)));
+  }
+
+  // Check for substr count > string count
+  Node* cmp = _gvn.transform(new CmpINode(tgt_count, src_count));
+  Node* bol = _gvn.transform(new BoolNode(cmp, BoolTest::gt));
+  Node* if_gt = generate_slow_guard(bol, NULL);
+  if (if_gt != NULL) {
+    result_phi->init_req(2, intcon(-1));
+    result_rgn->init_req(2, if_gt);
+  }
+
+  if (!stopped()) {
+    // Check for substr count == 0
+    cmp = _gvn.transform(new CmpINode(tgt_count, intcon(0)));
+    bol = _gvn.transform(new BoolNode(cmp, BoolTest::eq));
+    Node* if_zero = generate_slow_guard(bol, NULL);
+    if (if_zero != NULL) {
+      result_phi->init_req(3, intcon(0));
+      result_rgn->init_req(3, if_zero);
     }
-    set_control(_gvn.transform(result_rgn));
-    record_for_igvn(result_rgn);
-    result = _gvn.transform(result_phi);
-
-  } else { // Use LibraryCallKit::string_indexOf
-    // don't intrinsify if argument isn't a constant string.
-    if (!arg->is_Con()) {
-     return false;
-    }
-    const TypeOopPtr* str_type = _gvn.type(arg)->isa_oopptr();
-    if (str_type == NULL) {
-      return false;
-    }
-    ciInstanceKlass* klass = env()->String_klass();
-    ciObject* str_const = str_type->const_oop();
-    if (str_const == NULL || str_const->klass() != klass) {
-      return false;
-    }
-    ciInstance* str = str_const->as_instance();
-    assert(str != NULL, "must be instance");
-
-    ciObject* v = str->field_value_by_offset(java_lang_String::value_offset_in_bytes()).as_object();
-    ciTypeArray* pat = v->as_type_array(); // pattern (argument) character array
-
-    int o;
-    int c;
-    if (java_lang_String::has_offset_field()) {
-      o = str->field_value_by_offset(java_lang_String::offset_offset_in_bytes()).as_int();
-      c = str->field_value_by_offset(java_lang_String::count_offset_in_bytes()).as_int();
-    } else {
-      o = 0;
-      c = pat->length();
+  }
+
+  if (!stopped()) {
+    Node* result = make_string_method_node(Op_StrIndexOf, src_start, src_count, tgt_start, tgt_count, ae);
+    result_phi->init_req(1, result);
+    result_rgn->init_req(1, control());
+  }
+  set_control(_gvn.transform(result_rgn));
+  record_for_igvn(result_rgn);
+  set_result(_gvn.transform(result_phi));
+
+  return true;
+}
+
+//-----------------------------inline_string_indexOf-----------------------
+bool LibraryCallKit::inline_string_indexOfI(StrIntrinsicNode::ArgEnc ae) {
+  if (!Matcher::has_match_rule(Op_StrIndexOf) || !UseSSE42Intrinsics) {
+    return false;
+  }
+  assert(callee()->signature()->size() == 5, "String.indexOf() has 5 arguments");
+  Node* src         = argument(0); // byte[]
+  Node* src_count   = argument(1);
+  Node* tgt         = argument(2); // byte[]
+  Node* tgt_count   = argument(3);
+  Node* from_index  = argument(4);
+
+  // Java code which calls this method has range checks for from_index value.
+  src_count = _gvn.transform(new SubINode(src_count, from_index));
+
+  // Multiply byte array index by 2 if String is UTF16 encoded
+  Node* src_offset = (ae == StrIntrinsicNode::LL) ? from_index : _gvn.transform(new LShiftINode(from_index, intcon(1)));
+  Node* src_start = array_element_address(src, src_offset, T_BYTE);
+  Node* tgt_start = array_element_address(tgt, intcon(0), T_BYTE);
+
+  Node* result = make_string_method_node(Op_StrIndexOf, src_start, src_count, tgt_start, tgt_count, ae);
+
+  // The result is index relative to from_index if substring was found, -1 otherwise.
+  // Generate code which will fold into cmove.
+  RegionNode* region = new RegionNode(3);
+  Node* phi = new PhiNode(region, TypeInt::INT);
+
+  Node* cmp = _gvn.transform(new CmpINode(result, intcon(0)));
+  Node* bol = _gvn.transform(new BoolNode(cmp, BoolTest::lt));
+
+  Node* if_lt = generate_slow_guard(bol, NULL);
+  if (if_lt != NULL) {
+    // result == -1
+    phi->init_req(2, result);
+    region->init_req(2, if_lt);
+  }
+  if (!stopped()) {
+    result = _gvn.transform(new AddINode(result, from_index));
+    phi->init_req(1, result);
+    region->init_req(1, control());
+  }
+
+  set_control(_gvn.transform(region));
+  record_for_igvn(region);
+  set_result(_gvn.transform(phi));
+
+  return true;
+}
+
+//-----------------------------inline_string_indexOfChar-----------------------
+bool LibraryCallKit::inline_string_indexOfChar() {
+  if (!Matcher::has_match_rule(Op_StrIndexOfChar) || !(UseSSE > 4)) {
+    return false;
+  }
+  assert(callee()->signature()->size() == 4, "String.indexOfChar() has 4 arguments");
+  Node* src         = argument(0); // byte[]
+  Node* tgt         = argument(1); // tgt is int ch
+  Node* from_index  = argument(2);
+  Node* max         = argument(3);
+
+  Node* src_offset = _gvn.transform(new LShiftINode(from_index, intcon(1)));
+  Node* src_start = array_element_address(src, src_offset, T_BYTE);
+
+  Node* src_count = _gvn.transform(new SubINode(max, from_index));
+
+  RegionNode* region = new RegionNode(3);
+  Node* phi = new PhiNode(region, TypeInt::INT);
+
+  Node* result = new StrIndexOfCharNode(control(), memory(TypeAryPtr::BYTES), src_start, src_count, tgt, StrIntrinsicNode::none);
+  C->set_has_split_ifs(true); // Has chance for split-if optimization
+  _gvn.transform(result);
+
+  Node* cmp = _gvn.transform(new CmpINode(result, intcon(0)));
+  Node* bol = _gvn.transform(new BoolNode(cmp, BoolTest::lt));
+
+  Node* if_lt = generate_slow_guard(bol, NULL);
+  if (if_lt != NULL) {
+    // result == -1
+    phi->init_req(2, result);
+    region->init_req(2, if_lt);
+  }
+  if (!stopped()) {
+    result = _gvn.transform(new AddINode(result, from_index));
+    phi->init_req(1, result);
+    region->init_req(1, control());
+  }
+  set_control(_gvn.transform(region));
+  record_for_igvn(region);
+  set_result(_gvn.transform(phi));
+
+  return true;
+}
+//---------------------------inline_string_copy---------------------
+// compressIt == true --> generate a compressed copy operation (compress char[]/byte[] to byte[])
+//   int StringUTF16.compress(char[] src, int srcOff, byte[] dst, int dstOff, int len)
+//   int StringUTF16.compress(byte[] src, int srcOff, byte[] dst, int dstOff, int len)
+// compressIt == false --> generate an inflated copy operation (inflate byte[] to char[]/byte[])
+//   void StringLatin1.inflate(byte[] src, int srcOff, char[] dst, int dstOff, int len)
+//   void StringLatin1.inflate(byte[] src, int srcOff, byte[] dst, int dstOff, int len)
+bool LibraryCallKit::inline_string_copy(bool compress) {
+  int nargs = 5;  // 2 oops, 3 ints
+  assert(callee()->signature()->size() == nargs, "string copy has 5 arguments");
+
+  Node* src         = argument(0);
+  Node* src_offset  = argument(1);
+  Node* dst         = argument(2);
+  Node* dst_offset  = argument(3);
+  Node* length      = argument(4);
+
+  // Check for allocation before we add nodes that would confuse
+  // tightly_coupled_allocation()
+  AllocateArrayNode* alloc = tightly_coupled_allocation(dst, NULL);
+
+  // Figure out the size and type of the elements we will be copying.
+  const Type* src_type = src->Value(&_gvn);
+  const Type* dst_type = dst->Value(&_gvn);
+  BasicType src_elem = src_type->isa_aryptr()->klass()->as_array_klass()->element_type()->basic_type();
+  BasicType dst_elem = dst_type->isa_aryptr()->klass()->as_array_klass()->element_type()->basic_type();
+  assert((compress && dst_elem == T_BYTE && (src_elem == T_BYTE || src_elem == T_CHAR)) ||
+         (!compress && src_elem == T_BYTE && (dst_elem == T_BYTE || dst_elem == T_CHAR)),
+         "Unsupported array types for inline_string_copy");
+
+  // Convert char[] offsets to byte[] offsets
+  if (compress && src_elem == T_BYTE) {
+    src_offset = _gvn.transform(new LShiftINode(src_offset, intcon(1)));
+  } else if (!compress && dst_elem == T_BYTE) {
+    dst_offset = _gvn.transform(new LShiftINode(dst_offset, intcon(1)));
+  }
+
+  Node* src_start = array_element_address(src, src_offset, src_elem);
+  Node* dst_start = array_element_address(dst, dst_offset, dst_elem);
+  // 'src_start' points to src array + scaled offset
+  // 'dst_start' points to dst array + scaled offset
+  Node* count;
+  if (compress) {
+    count = compress_string(src_start, dst_start, length);
+  } else {
+    inflate_string(src_start, dst_start, length);
+  }
+
+  if (alloc != NULL) {
+    if (alloc->maybe_set_complete(&_gvn)) {
+      // "You break it, you buy it."
+      InitializeNode* init = alloc->initialization();
+      assert(init->is_complete(), "we just did this");
+      init->set_complete_with_arraycopy();
+      assert(dst->is_CheckCastPP(), "sanity");
+      assert(dst->in(0)->in(0) == init, "dest pinned");
     }
-
-    // constant strings have no offset and count == length which
-    // simplifies the resulting code somewhat so lets optimize for that.
-    if (o != 0 || c != pat->length()) {
-     return false;
+    // Do not let stores that initialize this object be reordered with
+    // a subsequent store that would make this object accessible by
+    // other threads.
+    // Record what AllocateNode this StoreStore protects so that
+    // escape analysis can go from the MemBarStoreStoreNode to the
+    // AllocateNode and eliminate the MemBarStoreStoreNode if possible
+    // based on the escape status of the AllocateNode.
+    insert_mem_bar(Op_MemBarStoreStore, alloc->proj_out(AllocateNode::RawAddress));
+  }
+  if (compress) {
+    set_result(_gvn.transform(count));
+  }
+  return true;
+}
+
+#ifdef _LP64
+#define XTOP ,top() /*additional argument*/
+#else  //_LP64
+#define XTOP        /*no additional argument*/
+#endif //_LP64
+
+//------------------------inline_string_toBytesU--------------------------
+// public static byte[] StringUTF16.toBytes(char[] value, int off, int len)
+bool LibraryCallKit::inline_string_toBytesU() {
+  // Get the arguments.
+  Node* value     = argument(0);
+  Node* offset    = argument(1);
+  Node* length    = argument(2);
+
+  Node* newcopy = NULL;
+
+  // Set the original stack and the reexecute bit for the interpreter to reexecute
+  // the bytecode that invokes StringUTF16.toBytes() if deoptimization happens.
+  { PreserveReexecuteState preexecs(this);
+    jvms()->set_should_reexecute(true);
+
+    // Check if a null path was taken unconditionally.
+    value = null_check(value);
+
+    RegionNode* bailout = new RegionNode(1);
+    record_for_igvn(bailout);
+
+    // Make sure that resulting byte[] length does not overflow Integer.MAX_VALUE
+    generate_negative_guard(length, bailout);
+    generate_limit_guard(length, intcon(0), intcon(max_jint/2), bailout);
+
+    if (bailout->req() > 1) {
+      PreserveJVMState pjvms(this);
+      set_control(_gvn.transform(bailout));
+      uncommon_trap(Deoptimization::Reason_intrinsic,
+                    Deoptimization::Action_maybe_recompile);
     }
-
-    receiver = null_check(receiver, T_OBJECT);
-    // NOTE: No null check on the argument is needed since it's a constant String oop.
-    if (stopped()) {
-      return true;
-    }
-
-    // The null string as a pattern always returns 0 (match at beginning of string)
-    if (c == 0) {
-      set_result(intcon(0));
-      return true;
+    if (stopped()) return true;
+
+    // Range checks are done by caller.
+
+    Node* size = _gvn.transform(new LShiftINode(length, intcon(1)));
+    Node* klass_node = makecon(TypeKlassPtr::make(ciTypeArrayKlass::make(T_BYTE)));
+    newcopy = new_array(klass_node, size, 0);  // no arguments to push
+    AllocateArrayNode* alloc = tightly_coupled_allocation(newcopy, NULL);
+
+    // Calculate starting addresses.
+    Node* src_start = array_element_address(value, offset, T_CHAR);
+    Node* dst_start = basic_plus_adr(newcopy, arrayOopDesc::base_offset_in_bytes(T_BYTE));
+
+    // Check if src array address is aligned to HeapWordSize (dst is always aligned)
+    const TypeInt* toffset = gvn().type(offset)->is_int();
+    bool aligned = toffset->is_con() && ((toffset->get_con() * type2aelembytes(T_CHAR)) % HeapWordSize == 0);
+
+    // Figure out which arraycopy runtime method to call (disjoint, uninitialized).
+    const char* copyfunc_name = "arraycopy";
+    address     copyfunc_addr = StubRoutines::select_arraycopy_function(T_CHAR, aligned, true, copyfunc_name, true);
+    Node* call = make_runtime_call(RC_LEAF|RC_NO_FP,
+                      OptoRuntime::fast_arraycopy_Type(),
+                      copyfunc_addr, copyfunc_name, TypeRawPtr::BOTTOM,
+                      src_start, dst_start, ConvI2X(length) XTOP);
+    // Do not let reads from the cloned object float above the arraycopy.
+    if (alloc != NULL) {
+      if (alloc->maybe_set_complete(&_gvn)) {
+        // "You break it, you buy it."
+        InitializeNode* init = alloc->initialization();
+        assert(init->is_complete(), "we just did this");
+        init->set_complete_with_arraycopy();
+        assert(newcopy->is_CheckCastPP(), "sanity");
+        assert(newcopy->in(0)->in(0) == init, "dest pinned");
+      }
+      // Do not let stores that initialize this object be reordered with
+      // a subsequent store that would make this object accessible by
+      // other threads.
+      // Record what AllocateNode this StoreStore protects so that
+      // escape analysis can go from the MemBarStoreStoreNode to the
+      // AllocateNode and eliminate the MemBarStoreStoreNode if possible
+      // based on the escape status of the AllocateNode.
+      insert_mem_bar(Op_MemBarStoreStore, alloc->proj_out(AllocateNode::RawAddress));
+    } else {
+      insert_mem_bar(Op_MemBarCPUOrder);
     }
-
-    // Generate default indexOf
-    jchar lastChar = pat->char_at(o + (c - 1));
-    int cache = 0;
-    int i;
-    for (i = 0; i < c - 1; i++) {
-      assert(i < pat->length(), "out of range");
-      cache |= (1 << (pat->char_at(o + i) & (sizeof(cache) * BitsPerByte - 1)));
+  } // original reexecute is set back here
+
+  C->set_has_split_ifs(true); // Has chance for split-if optimization
+  if (!stopped()) {
+    set_result(newcopy);
+  }
+  return true;
+}
+
+//------------------------inline_string_getCharsU--------------------------
+// public void StringUTF16.getChars(byte[] value, int srcBegin, int srcEnd, char dst[], int dstBegin)
+bool LibraryCallKit::inline_string_getCharsU() {
+  if (too_many_traps(Deoptimization::Reason_intrinsic))  return false;
+
+  // Get the arguments.
+  Node* value     = argument(0);
+  Node* src_begin = argument(1);
+  Node* src_end   = argument(2); // exclusive offset (i < src_end)
+  Node* dst       = argument(3);
+  Node* dst_begin = argument(4);
+
+  // Check for allocation before we add nodes that would confuse
+  // tightly_coupled_allocation()
+  AllocateArrayNode* alloc = tightly_coupled_allocation(dst, NULL);
+
+  // Check if a null path was taken unconditionally.
+  value = null_check(value);
+  dst = null_check(dst);
+  if (stopped()) {
+    return true;
+  }
+
+  // Range checks are done by caller.
+
+  // Get length and convert char[] offset to byte[] offset
+  Node* length = _gvn.transform(new SubINode(src_end, src_begin));
+  src_begin = _gvn.transform(new LShiftINode(src_begin, intcon(1)));
+
+  if (!stopped()) {
+    // Calculate starting addresses.
+    Node* src_start = array_element_address(value, src_begin, T_BYTE);
+    Node* dst_start = array_element_address(dst, dst_begin, T_CHAR);
+
+    // Check if array addresses are aligned to HeapWordSize
+    const TypeInt* tsrc = gvn().type(src_begin)->is_int();
+    const TypeInt* tdst = gvn().type(dst_begin)->is_int();
+    bool aligned = tsrc->is_con() && ((tsrc->get_con() * type2aelembytes(T_BYTE)) % HeapWordSize == 0) &&
+                   tdst->is_con() && ((tdst->get_con() * type2aelembytes(T_CHAR)) % HeapWordSize == 0);
+
+    // Figure out which arraycopy runtime method to call (disjoint, uninitialized).
+    const char* copyfunc_name = "arraycopy";
+    address     copyfunc_addr = StubRoutines::select_arraycopy_function(T_CHAR, aligned, true, copyfunc_name, true);
+    Node* call = make_runtime_call(RC_LEAF|RC_NO_FP,
+                      OptoRuntime::fast_arraycopy_Type(),
+                      copyfunc_addr, copyfunc_name, TypeRawPtr::BOTTOM,
+                      src_start, dst_start, ConvI2X(length) XTOP);
+    // Do not let reads from the cloned object float above the arraycopy.
+    if (alloc != NULL) {
+      if (alloc->maybe_set_complete(&_gvn)) {
+        // "You break it, you buy it."
+        InitializeNode* init = alloc->initialization();
+        assert(init->is_complete(), "we just did this");
+        init->set_complete_with_arraycopy();
+        assert(dst->is_CheckCastPP(), "sanity");
+        assert(dst->in(0)->in(0) == init, "dest pinned");
+      }
+      // Do not let stores that initialize this object be reordered with
+      // a subsequent store that would make this object accessible by
+      // other threads.
+      // Record what AllocateNode this StoreStore protects so that
+      // escape analysis can go from the MemBarStoreStoreNode to the
+      // AllocateNode and eliminate the MemBarStoreStoreNode if possible
+      // based on the escape status of the AllocateNode.
+      insert_mem_bar(Op_MemBarStoreStore, alloc->proj_out(AllocateNode::RawAddress));
+    } else {
+      insert_mem_bar(Op_MemBarCPUOrder);
     }
-
-    int md2 = c;
-    for (i = 0; i < c - 1; i++) {
-      assert(i < pat->length(), "out of range");
-      if (pat->char_at(o + i) == lastChar) {
-        md2 = (c - 1) - i;
-      }
-    }
-
-    result = string_indexOf(receiver, pat, o, cache, md2);
-  }
-  set_result(result);
+  }
+
+  C->set_has_split_ifs(true); // Has chance for split-if optimization
+  return true;
+}
+
+//----------------------inline_string_char_access----------------------------
+// Store/Load char to/from byte[] array.
+// static void StringUTF16.putChar(byte[] val, int index, int c)
+// static char StringUTF16.getChar(byte[] val, int index)
+bool LibraryCallKit::inline_string_char_access(bool is_store) {
+  Node* value  = argument(0);
+  Node* index  = argument(1);
+  Node* ch = is_store ? argument(2) : NULL;
+
+  // This intrinsic accesses byte[] array as char[] array. Computing the offsets
+  // correctly requires matched array shapes.
+  assert (arrayOopDesc::base_offset_in_bytes(T_CHAR) == arrayOopDesc::base_offset_in_bytes(T_BYTE),
+          "sanity: byte[] and char[] bases agree");
+  assert (type2aelembytes(T_CHAR) == type2aelembytes(T_BYTE)*2,
+          "sanity: byte[] and char[] scales agree");
+
+  Node* adr = array_element_address(value, index, T_CHAR);
+  if (is_store) {
+    (void) store_to_memory(control(), adr, ch, T_CHAR, TypeAryPtr::BYTES, MemNode::unordered);
+  } else {
+    ch = make_load(control(), adr, TypeInt::CHAR, T_CHAR, MemNode::unordered);
+    set_result(ch);
+  }
   return true;
 }
 
@@ -4189,12 +4300,6 @@
   return true;
 }
 
-#ifdef _LP64
-#define XTOP ,top() /*additional argument*/
-#else  //_LP64
-#define XTOP        /*no additional argument*/
-#endif //_LP64
-
 //----------------------inline_unsafe_copyMemory-------------------------
 // public native void sun.misc.Unsafe.copyMemory(Object srcBase, long srcOffset, Object destBase, long destOffset, long bytes);
 bool LibraryCallKit::inline_unsafe_copyMemory() {
@@ -5003,10 +5108,11 @@
   // Figure out the size and type of the elements we will be copying.
   BasicType src_elem = src_type->isa_aryptr()->klass()->as_array_klass()->element_type()->basic_type();
   BasicType dst_elem = dst_type->isa_aryptr()->klass()->as_array_klass()->element_type()->basic_type();
-  if (src_elem != T_CHAR || dst_elem != T_BYTE) {
+  if (!((src_elem == T_CHAR) || (src_elem== T_BYTE)) || dst_elem != T_BYTE) {
     return false;
   }
-  Node* src_start = array_element_address(src, src_offset, src_elem);
+
+  Node* src_start = array_element_address(src, src_offset, T_CHAR);
   Node* dst_start = array_element_address(dst, dst_offset, dst_elem);
   // 'src_start' points to src array + scaled offset
   // 'dst_start' points to dst array + scaled offset
@@ -5124,7 +5230,7 @@
 
 //-------------inline_squareToLen------------------------------------
 bool LibraryCallKit::inline_squareToLen() {
-  assert(UseSquareToLenIntrinsic, "not implementated on this platform");
+  assert(UseSquareToLenIntrinsic, "not implemented on this platform");
 
   address stubAddr = StubRoutines::squareToLen();
   if (stubAddr == NULL) {
@@ -5170,7 +5276,7 @@
 
 //-------------inline_mulAdd------------------------------------------
 bool LibraryCallKit::inline_mulAdd() {
-  assert(UseMulAddIntrinsic, "not implementated on this platform");
+  assert(UseMulAddIntrinsic, "not implemented on this platform");
 
   address stubAddr = StubRoutines::mulAdd();
   if (stubAddr == NULL) {
--- a/hotspot/src/share/vm/opto/loopTransform.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/opto/loopTransform.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -622,8 +622,10 @@
       case Op_StrComp:
       case Op_StrEquals:
       case Op_StrIndexOf:
+      case Op_StrIndexOfChar:
       case Op_EncodeISOArray:
-      case Op_AryEq: {
+      case Op_AryEq:
+      case Op_HasNegatives: {
         return false;
       }
 #if INCLUDE_RTM_OPT
@@ -741,8 +743,10 @@
       case Op_StrComp:
       case Op_StrEquals:
       case Op_StrIndexOf:
+      case Op_StrIndexOfChar:
       case Op_EncodeISOArray:
-      case Op_AryEq: {
+      case Op_AryEq:
+      case Op_HasNegatives: {
         // Do not unroll a loop with String intrinsics code.
         // String intrinsics are large and have loops.
         return false;
--- a/hotspot/src/share/vm/opto/loopnode.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/opto/loopnode.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -3494,7 +3494,9 @@
     case Op_StrComp:            // Does a bunch of load-like effects
     case Op_StrEquals:
     case Op_StrIndexOf:
+    case Op_StrIndexOfChar:
     case Op_AryEq:
+    case Op_HasNegatives:
       pinned = false;
     }
     if( pinned ) {
--- a/hotspot/src/share/vm/opto/macro.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/opto/macro.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -379,13 +379,25 @@
       if (mem->is_LoadStore()) {
         adr = mem->in(MemNode::Address);
       } else {
-        assert(mem->Opcode() == Op_EncodeISOArray, "sanity");
+        assert(mem->Opcode() == Op_EncodeISOArray ||
+               mem->Opcode() == Op_StrCompressedCopy, "sanity");
         adr = mem->in(3); // Destination array
       }
       const TypePtr* atype = adr->bottom_type()->is_ptr();
       int adr_idx = phase->C->get_alias_index(atype);
       if (adr_idx == alias_idx) {
-        assert(false, "Object is not scalar replaceable if a LoadStore node access its field");
+        DEBUG_ONLY(mem->dump();)
+        assert(false, "Object is not scalar replaceable if a LoadStore node accesses its field");
+        return NULL;
+      }
+      mem = mem->in(MemNode::Memory);
+   } else if (mem->Opcode() == Op_StrInflatedCopy) {
+      Node* adr = mem->in(3); // Destination array
+      const TypePtr* atype = adr->bottom_type()->is_ptr();
+      int adr_idx = phase->C->get_alias_index(atype);
+      if (adr_idx == alias_idx) {
+        DEBUG_ONLY(mem->dump();)
+        assert(false, "Object is not scalar replaceable if a StrInflatedCopy node accesses its field");
         return NULL;
       }
       mem = mem->in(MemNode::Memory);
@@ -516,8 +528,10 @@
         }
         values.at_put(j, val);
       } else if (val->Opcode() == Op_SCMemProj) {
-        assert(val->in(0)->is_LoadStore() || val->in(0)->Opcode() == Op_EncodeISOArray, "sanity");
-        assert(false, "Object is not scalar replaceable if a LoadStore node access its field");
+        assert(val->in(0)->is_LoadStore() ||
+               val->in(0)->Opcode() == Op_EncodeISOArray ||
+               val->in(0)->Opcode() == Op_StrCompressedCopy, "sanity");
+        assert(false, "Object is not scalar replaceable if a LoadStore node accesses its field");
         return NULL;
       } else if (val->is_ArrayCopy()) {
         Node* res = make_arraycopy_load(val->as_ArrayCopy(), offset, val->in(0), ft, phi_type, alloc);
--- a/hotspot/src/share/vm/opto/matcher.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/opto/matcher.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -936,9 +936,13 @@
     case Op_StrComp:
     case Op_StrEquals:
     case Op_StrIndexOf:
+    case Op_StrIndexOfChar:
     case Op_AryEq:
+    case Op_HasNegatives:
     case Op_MemBarVolatile:
     case Op_MemBarCPUOrder: // %%% these ideals should have narrower adr_type?
+    case Op_StrInflatedCopy:
+    case Op_StrCompressedCopy:
     case Op_EncodeISOArray:
       nidx = Compile::AliasIdxTop;
       nat = NULL;
@@ -2156,7 +2160,11 @@
       case Op_StrComp:
       case Op_StrEquals:
       case Op_StrIndexOf:
+      case Op_StrIndexOfChar:
       case Op_AryEq:
+      case Op_HasNegatives:
+      case Op_StrInflatedCopy:
+      case Op_StrCompressedCopy:
       case Op_EncodeISOArray:
         set_shared(n); // Force result into register (it will be anyways)
         break;
@@ -2336,7 +2344,8 @@
         n->del_req(3);
         break;
       }
-      case Op_StrEquals: {
+      case Op_StrEquals:
+      case Op_StrIndexOfChar: {
         Node *pair1 = new BinaryNode(n->in(2),n->in(3));
         n->set_req(2,pair1);
         n->set_req(3,n->in(4));
@@ -2353,6 +2362,8 @@
         n->del_req(4);
         break;
       }
+      case Op_StrCompressedCopy:
+      case Op_StrInflatedCopy:
       case Op_EncodeISOArray: {
         // Restructure into a binary tree for Matching.
         Node* pair = new BinaryNode(n->in(3), n->in(4));
--- a/hotspot/src/share/vm/opto/stringopts.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/opto/stringopts.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -598,7 +598,7 @@
   }
 
   // Collect the types needed to talk about the various slices of memory
-  char_adr_idx = C->get_alias_index(TypeAryPtr::CHARS);
+  byte_adr_idx = C->get_alias_index(TypeAryPtr::BYTES);
 
   // For each locally allocated StringBuffer see if the usages can be
   // collapsed into a single String construction.
@@ -1128,6 +1128,25 @@
 }
 
 Node* PhaseStringOpts::int_stringSize(GraphKit& kit, Node* arg) {
+  if (arg->is_Con()) {
+    // Constant integer. Compute constant length using Integer.sizeTable
+    int arg_val = arg->get_int();
+    int count = 1;
+    if (arg_val < 0) {
+      arg_val = -arg_val;
+      count++;
+    }
+
+    ciArray* size_table = (ciArray*)size_table_field->constant_value().as_object();
+    for (int i = 0; i < size_table->length(); i++) {
+      if (arg_val <= size_table->element_value(i).as_int()) {
+        count += i;
+        break;
+      }
+    }
+    return __ intcon(count);
+  }
+
   RegionNode *final_merge = new RegionNode(3);
   kit.gvn().set_type(final_merge, Type::CONTROL);
   Node* final_size = new PhiNode(final_merge, TypeInt::INT);
@@ -1212,77 +1231,34 @@
   return final_size;
 }
 
-void PhaseStringOpts::int_getChars(GraphKit& kit, Node* arg, Node* char_array, Node* start, Node* end) {
-  RegionNode *final_merge = new RegionNode(4);
-  kit.gvn().set_type(final_merge, Type::CONTROL);
-  Node *final_mem = PhiNode::make(final_merge, kit.memory(char_adr_idx), Type::MEMORY, TypeAryPtr::CHARS);
-  kit.gvn().set_type(final_mem, Type::MEMORY);
-
-  // need to handle Integer.MIN_VALUE specially because negating doesn't make it positive
-  {
-    // i == MIN_VALUE
-    IfNode* iff = kit.create_and_map_if(kit.control(),
-                                        __ Bool(__ CmpI(arg, __ intcon(0x80000000)), BoolTest::ne),
-                                        PROB_FAIR, COUNT_UNKNOWN);
-
-    Node* old_mem = kit.memory(char_adr_idx);
-
-    kit.set_control(__ IfFalse(iff));
-    if (kit.stopped()) {
-      // Statically not equal to MIN_VALUE so this path is dead
-      final_merge->init_req(3, kit.control());
-    } else {
-      copy_string(kit, __ makecon(TypeInstPtr::make(C->env()->the_min_jint_string())),
-                  char_array, start);
-      final_merge->init_req(3, kit.control());
-      final_mem->init_req(3, kit.memory(char_adr_idx));
-    }
-
-    kit.set_control(__ IfTrue(iff));
-    kit.set_memory(old_mem, char_adr_idx);
-  }
-
-
-  // Simplified version of Integer.getChars
-
-  // int q, r;
-  // int charPos = index;
-  Node* charPos = end;
-
-  // char sign = 0;
-
-  Node* i = arg;
-  Node* sign = __ intcon(0);
-
+// Simplified version of Integer.getChars
+void PhaseStringOpts::getChars(GraphKit& kit, Node* arg, Node* dst_array, BasicType bt, Node* end, Node* final_merge, Node* final_mem, int merge_index) {
   // if (i < 0) {
   //     sign = '-';
   //     i = -i;
   // }
-  {
-    IfNode* iff = kit.create_and_map_if(kit.control(),
-                                        __ Bool(__ CmpI(arg, __ intcon(0)), BoolTest::lt),
-                                        PROB_FAIR, COUNT_UNKNOWN);
+  IfNode* iff = kit.create_and_map_if(kit.control(), __ Bool(__ CmpI(arg, __ intcon(0)), BoolTest::lt),
+                                      PROB_FAIR, COUNT_UNKNOWN);
 
-    RegionNode *merge = new RegionNode(3);
-    kit.gvn().set_type(merge, Type::CONTROL);
-    i = new PhiNode(merge, TypeInt::INT);
-    kit.gvn().set_type(i, TypeInt::INT);
-    sign = new PhiNode(merge, TypeInt::INT);
-    kit.gvn().set_type(sign, TypeInt::INT);
+  RegionNode* merge = new RegionNode(3);
+  kit.gvn().set_type(merge, Type::CONTROL);
+  Node* i = new PhiNode(merge, TypeInt::INT);
+  kit.gvn().set_type(i, TypeInt::INT);
+  Node* sign = new PhiNode(merge, TypeInt::INT);
+  kit.gvn().set_type(sign, TypeInt::INT);
 
-    merge->init_req(1, __ IfTrue(iff));
-    i->init_req(1, __ SubI(__ intcon(0), arg));
-    sign->init_req(1, __ intcon('-'));
-    merge->init_req(2, __ IfFalse(iff));
-    i->init_req(2, arg);
-    sign->init_req(2, __ intcon(0));
+  merge->init_req(1, __ IfTrue(iff));
+  i->init_req(1, __ SubI(__ intcon(0), arg));
+  sign->init_req(1, __ intcon('-'));
+  merge->init_req(2, __ IfFalse(iff));
+  i->init_req(2, arg);
+  sign->init_req(2, __ intcon(0));
 
-    kit.set_control(merge);
+  kit.set_control(merge);
 
-    C->record_for_igvn(merge);
-    C->record_for_igvn(i);
-    C->record_for_igvn(sign);
-  }
+  C->record_for_igvn(merge);
+  C->record_for_igvn(i);
+  C->record_for_igvn(sign);
 
   // for (;;) {
   //     q = i / 10;
@@ -1292,126 +1268,409 @@
   //     if (i == 0) break;
   // }
 
-  {
-    // Add loop predicate first.
-    kit.add_predicate();
+  // Add loop predicate first.
+  kit.add_predicate();
+
+  RegionNode* head = new RegionNode(3);
+  head->init_req(1, kit.control());
 
-    RegionNode *head = new RegionNode(3);
-    head->init_req(1, kit.control());
-    kit.gvn().set_type(head, Type::CONTROL);
-    Node *i_phi = new PhiNode(head, TypeInt::INT);
-    i_phi->init_req(1, i);
-    kit.gvn().set_type(i_phi, TypeInt::INT);
-    charPos = PhiNode::make(head, charPos);
-    kit.gvn().set_type(charPos, TypeInt::INT);
-    Node *mem = PhiNode::make(head, kit.memory(char_adr_idx), Type::MEMORY, TypeAryPtr::CHARS);
-    kit.gvn().set_type(mem, Type::MEMORY);
-    kit.set_control(head);
-    kit.set_memory(mem, char_adr_idx);
+  kit.gvn().set_type(head, Type::CONTROL);
+  Node* i_phi = new PhiNode(head, TypeInt::INT);
+  i_phi->init_req(1, i);
+  kit.gvn().set_type(i_phi, TypeInt::INT);
+  Node* charPos = new PhiNode(head, TypeInt::INT);
+  charPos->init_req(1, end);
+  kit.gvn().set_type(charPos, TypeInt::INT);
+  Node* mem = PhiNode::make(head, kit.memory(byte_adr_idx), Type::MEMORY, TypeAryPtr::BYTES);
+  kit.gvn().set_type(mem, Type::MEMORY);
 
-    Node* q = __ DivI(NULL, i_phi, __ intcon(10));
-    Node* r = __ SubI(i_phi, __ AddI(__ LShiftI(q, __ intcon(3)),
-                                     __ LShiftI(q, __ intcon(1))));
-    Node* m1 = __ SubI(charPos, __ intcon(1));
-    Node* ch = __ AddI(r, __ intcon('0'));
-
-    Node* st = __ store_to_memory(kit.control(), kit.array_element_address(char_array, m1, T_CHAR),
-                                  ch, T_CHAR, char_adr_idx, MemNode::unordered);
-
+  kit.set_control(head);
+  kit.set_memory(mem, byte_adr_idx);
 
-    IfNode* iff = kit.create_and_map_if(head, __ Bool(__ CmpI(q, __ intcon(0)), BoolTest::ne),
-                                        PROB_FAIR, COUNT_UNKNOWN);
-    Node* ne = __ IfTrue(iff);
-    Node* eq = __ IfFalse(iff);
+  Node* q = __ DivI(kit.null(), i_phi, __ intcon(10));
+  Node* r = __ SubI(i_phi, __ AddI(__ LShiftI(q, __ intcon(3)),
+                                   __ LShiftI(q, __ intcon(1))));
+  Node* index = __ SubI(charPos, __ intcon((bt == T_BYTE) ? 1 : 2));
+  Node* ch = __ AddI(r, __ intcon('0'));
+  Node* st = __ store_to_memory(kit.control(), kit.array_element_address(dst_array, index, T_BYTE),
+                                ch, bt, byte_adr_idx, MemNode::unordered);
 
-    head->init_req(2, ne);
-    mem->init_req(2, st);
-    i_phi->init_req(2, q);
-    charPos->init_req(2, m1);
-
-    charPos = m1;
+  iff = kit.create_and_map_if(head, __ Bool(__ CmpI(q, __ intcon(0)), BoolTest::ne),
+                              PROB_FAIR, COUNT_UNKNOWN);
+  Node* ne = __ IfTrue(iff);
+  Node* eq = __ IfFalse(iff);
 
-    kit.set_control(eq);
-    kit.set_memory(st, char_adr_idx);
+  head->init_req(2, ne);
+  mem->init_req(2, st);
+
+  i_phi->init_req(2, q);
+  charPos->init_req(2, index);
+  charPos = index;
 
-    C->record_for_igvn(head);
-    C->record_for_igvn(mem);
-    C->record_for_igvn(i_phi);
-    C->record_for_igvn(charPos);
-  }
+  kit.set_control(eq);
+  kit.set_memory(st, byte_adr_idx);
 
-  {
-    // if (sign != 0) {
-    //     buf [--charPos] = sign;
-    // }
-    IfNode* iff = kit.create_and_map_if(kit.control(),
-                                        __ Bool(__ CmpI(sign, __ intcon(0)), BoolTest::ne),
-                                        PROB_FAIR, COUNT_UNKNOWN);
+  C->record_for_igvn(head);
+  C->record_for_igvn(mem);
+  C->record_for_igvn(i_phi);
+  C->record_for_igvn(charPos);
 
-    final_merge->init_req(2, __ IfFalse(iff));
-    final_mem->init_req(2, kit.memory(char_adr_idx));
+  // if (sign != 0) {
+  //     buf [--charPos] = sign;
+  // }
+  iff = kit.create_and_map_if(kit.control(), __ Bool(__ CmpI(sign, __ intcon(0)), BoolTest::ne),
+                              PROB_FAIR, COUNT_UNKNOWN);
+
+  final_merge->init_req(merge_index + 2, __ IfFalse(iff));
+  final_mem->init_req(merge_index + 2, kit.memory(byte_adr_idx));
 
-    kit.set_control(__ IfTrue(iff));
-    if (kit.stopped()) {
-      final_merge->init_req(1, C->top());
-      final_mem->init_req(1, C->top());
-    } else {
-      Node* m1 = __ SubI(charPos, __ intcon(1));
-      Node* st = __ store_to_memory(kit.control(), kit.array_element_address(char_array, m1, T_CHAR),
-                                    sign, T_CHAR, char_adr_idx, MemNode::unordered);
+  kit.set_control(__ IfTrue(iff));
+  if (kit.stopped()) {
+    final_merge->init_req(merge_index + 1, C->top());
+    final_mem->init_req(merge_index + 1, C->top());
+  } else {
+    Node* index = __ SubI(charPos, __ intcon((bt == T_BYTE) ? 1 : 2));
+    st = __ store_to_memory(kit.control(), kit.array_element_address(dst_array, index, T_BYTE),
+                            sign, bt, byte_adr_idx, MemNode::unordered);
 
-      final_merge->init_req(1, kit.control());
-      final_mem->init_req(1, st);
-    }
-
-    kit.set_control(final_merge);
-    kit.set_memory(final_mem, char_adr_idx);
-
-    C->record_for_igvn(final_merge);
-    C->record_for_igvn(final_mem);
+    final_merge->init_req(merge_index + 1, kit.control());
+    final_mem->init_req(merge_index + 1, st);
   }
 }
 
+// Copy the characters representing arg into dst_array starting at start
+Node* PhaseStringOpts::int_getChars(GraphKit& kit, Node* arg, Node* dst_array, Node* dst_coder, Node* start, Node* size) {
+  bool dcon = dst_coder->is_Con();
+  bool dbyte = dcon ? (dst_coder->get_int() == java_lang_String::CODER_LATIN1) : false;
+  Node* end = __ AddI(start, __ LShiftI(size, dst_coder));
 
-Node* PhaseStringOpts::copy_string(GraphKit& kit, Node* str, Node* char_array, Node* start) {
-  Node* string = str;
-  Node* offset = kit.load_String_offset(kit.control(), string);
-  Node* count  = kit.load_String_length(kit.control(), string);
-  Node* value  = kit.load_String_value (kit.control(), string);
+  // The final_merge node has 4 entries in case the encoding is known:
+  // (0) Control, (1) result w/ sign, (2) result w/o sign, (3) result for Integer.min_value
+  // or 6 entries in case the encoding is not known:
+  // (0) Control, (1) Latin1 w/ sign, (2) Latin1 w/o sign, (3) min_value, (4) UTF16 w/ sign, (5) UTF16 w/o sign
+  RegionNode* final_merge = new RegionNode(dcon ? 4 : 6);
+  kit.gvn().set_type(final_merge, Type::CONTROL);
+
+  Node* final_mem = PhiNode::make(final_merge, kit.memory(byte_adr_idx), Type::MEMORY, TypeAryPtr::BYTES);
+  kit.gvn().set_type(final_mem, Type::MEMORY);
+
+  // need to handle arg == Integer.MIN_VALUE specially because negating doesn't make it positive
+  IfNode* iff = kit.create_and_map_if(kit.control(), __ Bool(__ CmpI(arg, __ intcon(0x80000000)), BoolTest::ne),
+                                      PROB_FAIR, COUNT_UNKNOWN);
+
+  Node* old_mem = kit.memory(byte_adr_idx);
+
+  kit.set_control(__ IfFalse(iff));
+  if (kit.stopped()) {
+    // Statically not equal to MIN_VALUE so this path is dead
+    final_merge->init_req(3, kit.control());
+  } else {
+    copy_string(kit, __ makecon(TypeInstPtr::make(C->env()->the_min_jint_string())),
+                dst_array, dst_coder, start);
+    final_merge->init_req(3, kit.control());
+    final_mem->init_req(3, kit.memory(byte_adr_idx));
+  }
+
+  kit.set_control(__ IfTrue(iff));
+  kit.set_memory(old_mem, byte_adr_idx);
+
+  if (!dcon) {
+    // Check encoding of destination
+    iff = kit.create_and_map_if(kit.control(), __ Bool(__ CmpI(dst_coder, __ intcon(0)), BoolTest::eq),
+                                PROB_FAIR, COUNT_UNKNOWN);
+    old_mem = kit.memory(byte_adr_idx);
+  }
+  if (!dcon || dbyte) {
+    // Destination is Latin1,
+    if (!dcon) {
+      kit.set_control(__ IfTrue(iff));
+    }
+    getChars(kit, arg, dst_array, T_BYTE, end, final_merge, final_mem);
+  }
+  if (!dcon || !dbyte) {
+    // Destination is UTF16
+    int merge_index = 0;
+    if (!dcon) {
+      kit.set_control(__ IfFalse(iff));
+      kit.set_memory(old_mem, byte_adr_idx);
+      merge_index = 3; // Account for Latin1 case
+    }
+    getChars(kit, arg, dst_array, T_CHAR, end, final_merge, final_mem, merge_index);
+  }
+
+  // Final merge point for Latin1 and UTF16 case
+  kit.set_control(final_merge);
+  kit.set_memory(final_mem, byte_adr_idx);
+
+  C->record_for_igvn(final_merge);
+  C->record_for_igvn(final_mem);
+  return end;
+}
+
+// Copy 'count' bytes/chars from src_array to dst_array starting at index start
+void PhaseStringOpts::arraycopy(GraphKit& kit, IdealKit& ideal, Node* src_array, Node* dst_array, BasicType elembt, Node* start, Node* count) {
+  assert(elembt == T_BYTE || elembt == T_CHAR, "Invalid type for arraycopy");
+
+  if (elembt == T_CHAR) {
+    // Get number of chars
+    count = __ RShiftI(count, __ intcon(1));
+  }
+
+  Node* extra = NULL;
+#ifdef _LP64
+  count = __ ConvI2L(count);
+  extra = C->top();
+#endif
+
+  Node* src_ptr = __ array_element_address(src_array, __ intcon(0), T_BYTE);
+  Node* dst_ptr = __ array_element_address(dst_array, start, T_BYTE);
+  // Check if destination address is aligned to HeapWordSize
+  const TypeInt* tdst = __ gvn().type(start)->is_int();
+  bool aligned = tdst->is_con() && ((tdst->get_con() * type2aelembytes(T_BYTE)) % HeapWordSize == 0);
+  // Figure out which arraycopy runtime method to call (disjoint, uninitialized).
+  const char* copyfunc_name = "arraycopy";
+  address     copyfunc_addr = StubRoutines::select_arraycopy_function(elembt, aligned, true, copyfunc_name, true);
+  ideal.make_leaf_call_no_fp(OptoRuntime::fast_arraycopy_Type(), copyfunc_addr, copyfunc_name,
+                             TypeAryPtr::BYTES, src_ptr, dst_ptr, count, extra);
+}
+
+#undef __
+#define __ ideal.
+
+// Copy contents of a Latin1 encoded string from src_array to dst_array
+void PhaseStringOpts::copy_latin1_string(GraphKit& kit, IdealKit& ideal, Node* src_array, IdealVariable& count,
+                                         Node* dst_array, Node* dst_coder, Node* start) {
+  bool dcon = dst_coder->is_Con();
+  bool dbyte = dcon ? (dst_coder->get_int() == java_lang_String::CODER_LATIN1) : false;
 
-  // copy the contents
-  if (offset->is_Con() && count->is_Con() && value->is_Con() && count->get_int() < unroll_string_copy_length) {
+  if (!dcon) {
+    __ if_then(dst_coder, BoolTest::eq, __ ConI(java_lang_String::CODER_LATIN1));
+  }
+  if (!dcon || dbyte) {
+    // Destination is Latin1. Simply emit a byte arraycopy.
+    arraycopy(kit, ideal, src_array, dst_array, T_BYTE, start, __ value(count));
+  }
+  if (!dcon) {
+    __ else_();
+  }
+  if (!dcon || !dbyte) {
+    // Destination is UTF16. Inflate src_array into dst_array.
+    kit.sync_kit(ideal);
+    if (Matcher::match_rule_supported(Op_StrInflatedCopy)) {
+      // Use fast intrinsic
+      Node* src = kit.array_element_address(src_array, kit.intcon(0), T_BYTE);
+      Node* dst = kit.array_element_address(dst_array, start, T_BYTE);
+      kit.inflate_string(src, dst, __ value(count));
+    } else {
+      // No intrinsic available, use slow method
+      kit.inflate_string_slow(src_array, dst_array, start, __ value(count));
+    }
+    ideal.sync_kit(&kit);
+    // Multiply count by two since we now need two bytes per char
+    __ set(count, __ LShiftI(__ value(count), __ ConI(1)));
+  }
+  if (!dcon) {
+    __ end_if();
+  }
+}
+
+// Read two bytes from index and index+1 and convert them to a char
+static jchar readChar(ciTypeArray* array, int index) {
+  int shift_high, shift_low;
+#ifdef VM_LITTLE_ENDIAN
+    shift_high = 0;
+    shift_low = 8;
+#else
+    shift_high = 8;
+    shift_low = 0;
+#endif
+
+  jchar b1 = ((jchar) array->byte_at(index)) & 0xff;
+  jchar b2 = ((jchar) array->byte_at(index+1)) & 0xff;
+  return (b1 << shift_high) | (b2 << shift_low);
+}
+
+// Copy contents of constant src_array to dst_array by emitting individual stores
+void PhaseStringOpts::copy_constant_string(GraphKit& kit, IdealKit& ideal, ciTypeArray* src_array, IdealVariable& count,
+                                           bool src_is_byte, Node* dst_array, Node* dst_coder, Node* start) {
+  bool dcon = dst_coder->is_Con();
+  bool dbyte = dcon ? (dst_coder->get_int() == java_lang_String::CODER_LATIN1) : false;
+  int length = src_array->length();
+
+  if (!dcon) {
+    __ if_then(dst_coder, BoolTest::eq, __ ConI(java_lang_String::CODER_LATIN1));
+  }
+  if (!dcon || dbyte) {
+    // Destination is Latin1. Copy each byte of src_array into dst_array.
+    Node* index = start;
+    for (int i = 0; i < length; i++) {
+      Node* adr = kit.array_element_address(dst_array, index, T_BYTE);
+      Node* val = __ ConI(src_array->byte_at(i));
+      __ store(__ ctrl(), adr, val, T_BYTE, byte_adr_idx, MemNode::unordered);
+      index = __ AddI(index, __ ConI(1));
+    }
+  }
+  if (!dcon) {
+    __ else_();
+  }
+  if (!dcon || !dbyte) {
+    // Destination is UTF16. Copy each char of src_array into dst_array.
+    Node* index = start;
+    for (int i = 0; i < length; i++) {
+      Node* adr = kit.array_element_address(dst_array, index, T_BYTE);
+      jchar val;
+      if (src_is_byte) {
+        val = src_array->byte_at(i);
+      } else {
+        val = readChar(src_array, i++);
+      }
+      __ store(__ ctrl(), adr, __ ConI(val), T_CHAR, byte_adr_idx, MemNode::unordered);
+      index = __ AddI(index, __ ConI(2));
+    }
+    if (src_is_byte) {
+      // Multiply count by two since we now need two bytes per char
+      __ set(count, __ ConI(2 * length));
+    }
+  }
+  if (!dcon) {
+    __ end_if();
+  }
+}
+
+// Compress copy contents of the byte/char String str into dst_array starting at index start.
+Node* PhaseStringOpts::copy_string(GraphKit& kit, Node* str, Node* dst_array, Node* dst_coder, Node* start) {
+  Node* src_array = kit.load_String_value(kit.control(), str);
+
+  IdealKit ideal(&kit, true, true);
+  IdealVariable count(ideal); __ declarations_done();
+
+  if (str->is_Con()) {
+    // Constant source string
+    const TypeOopPtr* t = kit.gvn().type(src_array)->isa_oopptr();
+    ciTypeArray* src_array_type = t->const_oop()->as_type_array();
+
+    // Check encoding of constant string
+    bool src_is_byte = (get_constant_coder(kit, str) == java_lang_String::CODER_LATIN1);
+
     // For small constant strings just emit individual stores.
     // A length of 6 seems like a good space/speed tradeof.
-    int c = count->get_int();
-    int o = offset->get_int();
-    const TypeOopPtr* t = kit.gvn().type(value)->isa_oopptr();
-    ciTypeArray* value_array = t->const_oop()->as_type_array();
-    for (int e = 0; e < c; e++) {
-      __ store_to_memory(kit.control(), kit.array_element_address(char_array, start, T_CHAR),
-                         __ intcon(value_array->char_at(o + e)), T_CHAR, char_adr_idx,
-                         MemNode::unordered);
-      start = __ AddI(start, __ intcon(1));
+    __ set(count, __ ConI(src_array_type->length()));
+    int src_len = src_array_type->length() / (src_is_byte ? 1 : 2);
+    if (src_len < unroll_string_copy_length) {
+      // Small constant string
+      copy_constant_string(kit, ideal, src_array_type, count, src_is_byte, dst_array, dst_coder, start);
+    } else if (src_is_byte) {
+      // Source is Latin1
+      copy_latin1_string(kit, ideal, src_array, count, dst_array, dst_coder, start);
+    } else {
+      // Source is UTF16 (destination too). Simply emit a char arraycopy.
+      arraycopy(kit, ideal, src_array, dst_array, T_CHAR, start, __ value(count));
     }
   } else {
-    Node* src_ptr = kit.array_element_address(value, offset, T_CHAR);
-    Node* dst_ptr = kit.array_element_address(char_array, start, T_CHAR);
-    Node* c = count;
-    Node* extra = NULL;
-#ifdef _LP64
-    c = __ ConvI2L(c);
-    extra = C->top();
-#endif
-    Node* call = kit.make_runtime_call(GraphKit::RC_LEAF|GraphKit::RC_NO_FP,
-                                       OptoRuntime::fast_arraycopy_Type(),
-                                       CAST_FROM_FN_PTR(address, StubRoutines::jshort_disjoint_arraycopy()),
-                                       "jshort_disjoint_arraycopy", TypeAryPtr::CHARS,
-                                       src_ptr, dst_ptr, c, extra);
-    start = __ AddI(start, count);
+    Node* size = kit.load_array_length(src_array);
+    __ set(count, size);
+    // Non-constant source string
+    if (CompactStrings) {
+      // Emit runtime check for coder
+      Node* coder = kit.load_String_coder(__ ctrl(), str);
+      __ if_then(coder, BoolTest::eq, __ ConI(java_lang_String::CODER_LATIN1)); {
+        // Source is Latin1
+        copy_latin1_string(kit, ideal, src_array, count, dst_array, dst_coder, start);
+      } __ else_();
+    }
+    // Source is UTF16 (destination too). Simply emit a char arraycopy.
+    arraycopy(kit, ideal, src_array, dst_array, T_CHAR, start, __ value(count));
+
+    if (CompactStrings) {
+      __ end_if();
+    }
   }
-  return start;
+
+  // Finally sync IdealKit and GraphKit.
+  kit.sync_kit(ideal);
+  return __ AddI(start, __ value(count));
 }
 
+// Compress copy the char into dst_array at index start.
+Node* PhaseStringOpts::copy_char(GraphKit& kit, Node* val, Node* dst_array, Node* dst_coder, Node* start) {
+  bool dcon = (dst_coder != NULL) && dst_coder->is_Con();
+  bool dbyte = dcon ? (dst_coder->get_int() == java_lang_String::CODER_LATIN1) : false;
+
+  IdealKit ideal(&kit, true, true);
+  IdealVariable end(ideal); __ declarations_done();
+  Node* adr = kit.array_element_address(dst_array, start, T_BYTE);
+  if (!dcon){
+    __ if_then(dst_coder, BoolTest::eq, __ ConI(java_lang_String::CODER_LATIN1));
+  }
+  if (!dcon || dbyte) {
+    // Destination is Latin1. Store a byte.
+    __ store(__ ctrl(), adr, val, T_BYTE, byte_adr_idx, MemNode::unordered);
+    __ set(end, __ AddI(start, __ ConI(1)));
+  }
+  if (!dcon) {
+    __ else_();
+  }
+  if (!dcon || !dbyte) {
+    // Destination is UTF16. Store a char.
+    __ store(__ ctrl(), adr, val, T_CHAR, byte_adr_idx, MemNode::unordered);
+    __ set(end, __ AddI(start, __ ConI(2)));
+  }
+  if (!dcon) {
+    __ end_if();
+  }
+  // Finally sync IdealKit and GraphKit.
+  kit.sync_kit(ideal);
+  return __ value(end);
+}
+
+#undef __
+#define __ kit.
+
+// Allocate a byte array of specified length.
+Node* PhaseStringOpts::allocate_byte_array(GraphKit& kit, IdealKit* ideal, Node* length) {
+  if (ideal != NULL) {
+    // Sync IdealKit and graphKit.
+    kit.sync_kit(*ideal);
+  }
+  Node* byte_array = NULL;
+  {
+    PreserveReexecuteState preexecs(&kit);
+    // The original jvms is for an allocation of either a String or
+    // StringBuffer so no stack adjustment is necessary for proper
+    // reexecution.  If we deoptimize in the slow path the bytecode
+    // will be reexecuted and the char[] allocation will be thrown away.
+    kit.jvms()->set_should_reexecute(true);
+    byte_array = kit.new_array(__ makecon(TypeKlassPtr::make(ciTypeArrayKlass::make(T_BYTE))),
+                               length, 1);
+  }
+
+  // Mark the allocation so that zeroing is skipped since the code
+  // below will overwrite the entire array
+  AllocateArrayNode* byte_alloc = AllocateArrayNode::Ideal_array_allocation(byte_array, _gvn);
+  byte_alloc->maybe_set_complete(_gvn);
+
+  if (ideal != NULL) {
+    // Sync IdealKit and graphKit.
+    ideal->sync_kit(&kit);
+  }
+  return byte_array;
+}
+
+jbyte PhaseStringOpts::get_constant_coder(GraphKit& kit, Node* str) {
+  assert(str->is_Con(), "String must be constant");
+  const TypeOopPtr* str_type = kit.gvn().type(str)->isa_oopptr();
+  ciInstance* str_instance = str_type->const_oop()->as_instance();
+  jbyte coder = str_instance->field_value_by_offset(java_lang_String::coder_offset_in_bytes()).as_byte();
+  assert(CompactStrings || (coder == java_lang_String::CODER_UTF16), "Strings must be UTF16 encoded");
+  return coder;
+}
+
+int PhaseStringOpts::get_constant_length(GraphKit& kit, Node* str) {
+  assert(str->is_Con(), "String must be constant");
+  Node* src_array = kit.load_String_value(kit.control(), str);
+  const TypeOopPtr* t = kit.gvn().type(src_array)->isa_oopptr();
+  return t->const_oop()->as_type_array()->length();
+}
 
 void PhaseStringOpts::replace_string_concat(StringConcat* sc) {
   // Log a little info about the transformation
@@ -1445,7 +1704,6 @@
   jvms->set_map(map);
   map->ensure_stack(jvms, jvms->method()->max_stack());
 
-
   // disconnect all the old StringBuilder calls from the graph
   sc->eliminate_unneeded_control();
 
@@ -1473,7 +1731,17 @@
   // are need for the copying phase.
   Node* string_sizes = new Node(args);
 
+  Node* coder = __ intcon(0);
   Node* length = __ intcon(0);
+  // If at least one argument is UTF16 encoded, we can fix the encoding.
+  bool coder_fixed = false;
+
+  if (!CompactStrings) {
+    // Fix encoding of result string to UTF16
+    coder_fixed = true;
+    coder = __ intcon(java_lang_String::CODER_UTF16);
+  }
+
   for (int argi = 0; argi < sc->num_arguments(); argi++) {
     Node* arg = sc->argument(argi);
     switch (sc->mode(argi)) {
@@ -1491,7 +1759,7 @@
         const Type* type = kit.gvn().type(arg);
         assert(type != TypePtr::NULL_PTR, "missing check");
         if (!type->higher_equal(TypeInstPtr::NOTNULL)) {
-          // Null check with uncommont trap since
+          // Null check with uncommon trap since
           // StringBuilder(null) throws exception.
           // Use special uncommon trap instead of
           // calling normal do_null_check().
@@ -1509,11 +1777,13 @@
       case StringConcat::StringMode: {
         const Type* type = kit.gvn().type(arg);
         Node* count = NULL;
+        Node* arg_coder = NULL;
         if (type == TypePtr::NULL_PTR) {
           // replace the argument with the null checked version
           arg = null_string;
           sc->set_argument(argi, arg);
           count = kit.load_String_length(kit.control(), arg);
+          arg_coder = kit.load_String_coder(kit.control(), arg);
         } else if (!type->higher_equal(TypeInstPtr::NOTNULL)) {
           // s = s != null ? s : "null";
           // length = length + (s.count - s.offset);
@@ -1537,11 +1807,32 @@
           arg = phi;
           sc->set_argument(argi, arg);
           count = kit.load_String_length(kit.control(), arg);
+          arg_coder = kit.load_String_coder(kit.control(), arg);
         } else {
           // A corresponding nullcheck will be connected during IGVN MemNode::Ideal_common_DU_postCCP
           // kit.control might be a different test, that can be hoisted above the actual nullcheck
           // in case, that the control input is not null, Ideal_common_DU_postCCP will not look for a nullcheck.
           count = kit.load_String_length(NULL, arg);
+          arg_coder = kit.load_String_coder(NULL, arg);
+        }
+        if (arg->is_Con()) {
+          // Constant string. Get constant coder and length.
+          jbyte const_coder = get_constant_coder(kit, arg);
+          int const_length = get_constant_length(kit, arg);
+          if (const_coder == java_lang_String::CODER_LATIN1) {
+            // Can be latin1 encoded
+            arg_coder = __ intcon(const_coder);
+            count = __ intcon(const_length);
+          } else {
+            // Found UTF16 encoded string. Fix result array encoding to UTF16.
+            coder_fixed = true;
+            coder = __ intcon(const_coder);
+            count = __ intcon(const_length / 2);
+          }
+        }
+
+        if (!coder_fixed) {
+          coder = __ OrI(coder, arg_coder);
         }
         length = __ AddI(length, count);
         string_sizes->init_req(argi, NULL);
@@ -1549,6 +1840,34 @@
       }
       case StringConcat::CharMode: {
         // one character only
+        const TypeInt* t = kit.gvn().type(arg)->is_int();
+        if (!coder_fixed && t->is_con()) {
+          // Constant char
+          if (t->get_con() <= 255) {
+            // Can be latin1 encoded
+            coder = __ OrI(coder, __ intcon(java_lang_String::CODER_LATIN1));
+          } else {
+            // Must be UTF16 encoded. Fix result array encoding to UTF16.
+            coder_fixed = true;
+            coder = __ intcon(java_lang_String::CODER_UTF16);
+          }
+        } else if (!coder_fixed) {
+          // Not constant
+#undef __
+#define __ ideal.
+          IdealKit ideal(&kit, true, true);
+          IdealVariable char_coder(ideal); __ declarations_done();
+          // Check if character can be latin1 encoded
+          __ if_then(arg, BoolTest::le, __ ConI(0xFF));
+            __ set(char_coder, __ ConI(java_lang_String::CODER_LATIN1));
+          __ else_();
+            __ set(char_coder, __ ConI(java_lang_String::CODER_UTF16));
+          __ end_if();
+          kit.sync_kit(ideal);
+          coder = __ OrI(coder, __ value(char_coder));
+#undef __
+#define __ kit.
+        }
         length = __ AddI(length, __ intcon(1));
         break;
       }
@@ -1576,54 +1895,37 @@
 
   Node* result;
   if (!kit.stopped()) {
-    Node* char_array = NULL;
+    assert(CompactStrings || (coder->is_Con() && coder->get_int() == java_lang_String::CODER_UTF16),
+           "Result string must be UTF16 encoded if CompactStrings is disabled");
+
+    Node* dst_array = NULL;
     if (sc->num_arguments() == 1 &&
-          (sc->mode(0) == StringConcat::StringMode ||
-           sc->mode(0) == StringConcat::StringNullCheckMode)) {
+        (sc->mode(0) == StringConcat::StringMode ||
+         sc->mode(0) == StringConcat::StringNullCheckMode)) {
       // Handle the case when there is only a single String argument.
       // In this case, we can just pull the value from the String itself.
-      char_array = kit.load_String_value(kit.control(), sc->argument(0));
+      dst_array = kit.load_String_value(kit.control(), sc->argument(0));
     } else {
-      // length now contains the number of characters needed for the
-      // char[] so create a new AllocateArray for the char[]
-      {
-        PreserveReexecuteState preexecs(&kit);
-        // The original jvms is for an allocation of either a String or
-        // StringBuffer so no stack adjustment is necessary for proper
-        // reexecution.  If we deoptimize in the slow path the bytecode
-        // will be reexecuted and the char[] allocation will be thrown away.
-        kit.jvms()->set_should_reexecute(true);
-        char_array = kit.new_array(__ makecon(TypeKlassPtr::make(ciTypeArrayKlass::make(T_CHAR))),
-                                   length, 1);
-      }
+      // Allocate destination byte array according to coder
+      dst_array = allocate_byte_array(kit, NULL, __ LShiftI(length, coder));
 
-      // Mark the allocation so that zeroing is skipped since the code
-      // below will overwrite the entire array
-      AllocateArrayNode* char_alloc = AllocateArrayNode::Ideal_array_allocation(char_array, _gvn);
-      char_alloc->maybe_set_complete(_gvn);
-
-      // Now copy the string representations into the final char[]
+      // Now copy the string representations into the final byte[]
       Node* start = __ intcon(0);
       for (int argi = 0; argi < sc->num_arguments(); argi++) {
         Node* arg = sc->argument(argi);
         switch (sc->mode(argi)) {
           case StringConcat::IntMode: {
-            Node* end = __ AddI(start, string_sizes->in(argi));
-            // getChars words backwards so pass the ending point as well as the start
-            int_getChars(kit, arg, char_array, start, end);
-            start = end;
+            start = int_getChars(kit, arg, dst_array, coder, start, string_sizes->in(argi));
             break;
           }
           case StringConcat::StringNullCheckMode:
           case StringConcat::StringMode: {
-            start = copy_string(kit, arg, char_array, start);
+            start = copy_string(kit, arg, dst_array, coder, start);
             break;
           }
           case StringConcat::CharMode: {
-            __ store_to_memory(kit.control(), kit.array_element_address(char_array, start, T_CHAR),
-                               arg, T_CHAR, char_adr_idx, MemNode::unordered);
-            start = __ AddI(start, __ intcon(1));
-            break;
+            start = copy_char(kit, arg, dst_array, coder, start);
+          break;
           }
           default:
             ShouldNotReachHere();
@@ -1642,12 +1944,9 @@
       result = kit.new_instance(__ makecon(TypeKlassPtr::make(C->env()->String_klass())));
     }
 
-    // Intialize the string
-    if (java_lang_String::has_offset_field()) {
-      kit.store_String_offset(kit.control(), result, __ intcon(0));
-      kit.store_String_length(kit.control(), result, length);
-    }
-    kit.store_String_value(kit.control(), result, char_array);
+    // Initialize the string
+    kit.store_String_value(kit.control(), result, dst_array);
+    kit.store_String_coder(kit.control(), result, coder);
   } else {
     result = C->top();
   }
--- a/hotspot/src/share/vm/opto/stringopts.hpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/opto/stringopts.hpp	Tue Nov 03 09:41:03 2015 +0100
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2009, 2012, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2009, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -29,6 +29,7 @@
 #include "opto/phaseX.hpp"
 
 class StringConcat;
+class IdealVariable;
 
 class PhaseStringOpts : public Phase {
   friend class StringConcat;
@@ -40,7 +41,7 @@
   Unique_Node_List dead_worklist;
 
   // Memory slices needed for code gen
-  int char_adr_idx;
+  int byte_adr_idx;
 
   // Integer.sizeTable - used for int to String conversion
   ciField* size_table_field;
@@ -64,11 +65,37 @@
   // Compute the number of characters required to represent the int value
   Node* int_stringSize(GraphKit& kit, Node* value);
 
-  // Copy the characters representing value into char_array starting at start
-  void int_getChars(GraphKit& kit, Node* value, Node* char_array, Node* start, Node* end);
+  // Simplified version of Integer.getChars
+  void getChars(GraphKit& kit, Node* arg, Node* dst_array, BasicType bt, Node* end, Node* final_merge, Node* final_mem, int merge_index = 0);
+
+  // Copy the characters representing arg into dst_array starting at start
+  Node* int_getChars(GraphKit& kit, Node* arg, Node* dst_array, Node* dst_coder, Node* start, Node* size);
+
+  // Copy contents of the String str into dst_array starting at index start.
+  Node* copy_string(GraphKit& kit, Node* str, Node* dst_array, Node* dst_coder, Node* start);
+
+  // Copy 'count' bytes/chars from src_array to dst_array starting at index start
+  void arraycopy(GraphKit& kit, IdealKit& ideal, Node* src_array, Node* dst_array, BasicType elembt, Node* start, Node* count);
+
+  // Copy contents of constant src_array to dst_array by emitting individual stores
+  void copy_constant_string(GraphKit& kit, IdealKit& ideal, ciTypeArray* src_array, IdealVariable& count,
+                            bool src_is_byte, Node* dst_array, Node* dst_coder, Node* start);
 
-  // Copy of the contents of the String str into char_array starting at index start.
-  Node* copy_string(GraphKit& kit, Node* str, Node* char_array, Node* start);
+  // Copy contents of a Latin1 encoded string from src_array to dst_array
+  void copy_latin1_string(GraphKit& kit, IdealKit& ideal, Node* src_array, IdealVariable& count,
+                          Node* dst_array, Node* dst_coder, Node* start);
+
+  // Copy the char into dst_array at index start.
+  Node* copy_char(GraphKit& kit, Node* val, Node* dst_array, Node* dst_coder, Node* start);
+
+  // Allocate a byte array of specified length.
+  Node* allocate_byte_array(GraphKit& kit, IdealKit* ideal, Node* length);
+
+  // Returns the coder of a constant string
+  jbyte get_constant_coder(GraphKit& kit, Node* str);
+
+  // Returns the length of a constant string
+  int get_constant_length(GraphKit& kit, Node* str);
 
   // Clean up any leftover nodes
   void record_dead_node(Node* node);
--- a/hotspot/src/share/vm/prims/jni.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/prims/jni.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -2474,12 +2474,18 @@
   typeArrayOop s_value = java_lang_String::value(s);
   if (s_value != NULL) {
     int s_len = java_lang_String::length(s);
-    int s_offset = java_lang_String::offset(s);
+    bool is_latin1 = java_lang_String::is_latin1(s);
     buf = NEW_C_HEAP_ARRAY_RETURN_NULL(jchar, s_len + 1, mtInternal);  // add one for zero termination
     /* JNI Specification states return NULL on OOM */
     if (buf != NULL) {
       if (s_len > 0) {
-        memcpy(buf, s_value->char_at_addr(s_offset), sizeof(jchar)*s_len);
+        if (!is_latin1) {
+          memcpy(buf, s_value->char_at_addr(0), sizeof(jchar)*s_len);
+        } else {
+          for (int i = 0; i < s_len; i++) {
+            buf[i] = ((jchar) s_value->byte_at(i)) & 0xff;
+          }
+        }
       }
       buf[s_len] = 0;
       //%note jni_5
@@ -3118,9 +3124,15 @@
     THROW(vmSymbols::java_lang_StringIndexOutOfBoundsException());
   } else {
     if (len > 0) {
-      int s_offset = java_lang_String::offset(s);
       typeArrayOop s_value = java_lang_String::value(s);
-      memcpy(buf, s_value->char_at_addr(s_offset+start), sizeof(jchar)*len);
+      bool is_latin1 = java_lang_String::is_latin1(s);
+      if (!is_latin1) {
+        memcpy(buf, s_value->char_at_addr(start), sizeof(jchar)*len);
+      } else {
+        for (int i = 0; i < len; i++) {
+          buf[i] = ((jchar) s_value->byte_at(i + start)) & 0xff;
+        }
+      }
     }
   }
 JNI_END
@@ -3186,18 +3198,23 @@
   JNIWrapper("GetStringCritical");
   HOTSPOT_JNI_GETSTRINGCRITICAL_ENTRY(env, string, (uintptr_t *) isCopy);
   GC_locker::lock_critical(thread);
-  if (isCopy != NULL) {
-    *isCopy = JNI_FALSE;
-  }
   oop s = JNIHandles::resolve_non_null(string);
-  int s_len = java_lang_String::length(s);
   typeArrayOop s_value = java_lang_String::value(s);
-  int s_offset = java_lang_String::offset(s);
+  bool is_latin1 = java_lang_String::is_latin1(s);
+  if (isCopy != NULL) {
+    *isCopy = is_latin1 ? JNI_TRUE : JNI_FALSE;
+  }
   const jchar* ret;
-  if (s_len > 0) {
-    ret = s_value->char_at_addr(s_offset);
+  if (!is_latin1) {
+    ret = s_value->char_at_addr(0);
   } else {
-    ret = (jchar*) s_value->base(T_CHAR);
+    // Inflate latin1 encoded string to UTF16
+    int s_len = java_lang_String::length(s);
+    jchar* buf = NEW_C_HEAP_ARRAY(jchar, s_len, mtInternal);
+    for (int i = 0; i < s_len; i++) {
+      buf[i] = ((jchar) s_value->byte_at(i)) & 0xff;
+    }
+    ret = &buf[0];
   }
  HOTSPOT_JNI_GETSTRINGCRITICAL_RETURN((uint16_t *) ret);
   return ret;
@@ -3207,7 +3224,14 @@
 JNI_ENTRY(void, jni_ReleaseStringCritical(JNIEnv *env, jstring str, const jchar *chars))
   JNIWrapper("ReleaseStringCritical");
   HOTSPOT_JNI_RELEASESTRINGCRITICAL_ENTRY(env, str, (uint16_t *) chars);
-  // The str and chars arguments are ignored
+  // The str and chars arguments are ignored for UTF16 strings
+  oop s = JNIHandles::resolve_non_null(str);
+  bool is_latin1 = java_lang_String::is_latin1(s);
+  if (is_latin1) {
+    // For latin1 string, free jchar array allocated by earlier call to GetStringCritical.
+    // This assumes that ReleaseStringCritical bookends GetStringCritical.
+    FREE_C_HEAP_ARRAY(jchar, chars);
+  }
   GC_locker::unlock_critical(thread);
 HOTSPOT_JNI_RELEASESTRINGCRITICAL_RETURN();
 JNI_END
--- a/hotspot/src/share/vm/prims/jvmtiEnv.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/prims/jvmtiEnv.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -966,7 +966,7 @@
     if (name() != NULL) {
       n = java_lang_String::as_utf8_string(name());
     } else {
-      n = UNICODE::as_utf8(NULL, 0);
+      n = UNICODE::as_utf8((jchar*) NULL, 0);
     }
 
     info_ptr->name = (char *) jvmtiMalloc(strlen(n)+1);
@@ -1187,15 +1187,14 @@
   Handle group_obj (current_thread, JNIHandles::resolve_external_guard(group));
   NULL_CHECK(group_obj(), JVMTI_ERROR_INVALID_THREAD_GROUP);
 
-  typeArrayHandle name;
+  const char* name;
   Handle parent_group;
   bool is_daemon;
   ThreadPriority max_priority;
 
   { MutexLocker mu(Threads_lock);
 
-    name         = typeArrayHandle(current_thread,
-                                   java_lang_ThreadGroup::name(group_obj()));
+    name         = java_lang_ThreadGroup::name(group_obj());
     parent_group = Handle(current_thread, java_lang_ThreadGroup::parent(group_obj()));
     is_daemon    = java_lang_ThreadGroup::is_daemon(group_obj());
     max_priority = java_lang_ThreadGroup::maxPriority(group_obj());
@@ -1205,11 +1204,10 @@
   info_ptr->max_priority = max_priority;
   info_ptr->parent       = jni_reference(parent_group);
 
-  if (name() != NULL) {
-    const char* n = UNICODE::as_utf8((jchar*) name->base(T_CHAR), name->length());
-    info_ptr->name = (char *)jvmtiMalloc(strlen(n)+1);
+  if (name != NULL) {
+    info_ptr->name = (char*)jvmtiMalloc(strlen(name)+1);
     NULL_CHECK(info_ptr->name, JVMTI_ERROR_OUT_OF_MEMORY);
-    strcpy(info_ptr->name, n);
+    strcpy(info_ptr->name, name);
   } else {
     info_ptr->name = NULL;
   }
--- a/hotspot/src/share/vm/prims/jvmtiTagMap.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/prims/jvmtiTagMap.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -1057,21 +1057,36 @@
   // get the string value and length
   // (string value may be offset from the base)
   int s_len = java_lang_String::length(str);
-  int s_offset = java_lang_String::offset(str);
+  bool is_latin1 = java_lang_String::is_latin1(str);
   jchar* value;
   if (s_len > 0) {
-    value = s_value->char_at_addr(s_offset);
+    if (!is_latin1) {
+      value = s_value->char_at_addr(0);
+    } else {
+      // Inflate latin1 encoded string to UTF16
+      jchar* buf = NEW_C_HEAP_ARRAY(jchar, s_len, mtInternal);
+      for (int i = 0; i < s_len; i++) {
+        buf[i] = ((jchar) s_value->byte_at(i)) & 0xff;
+      }
+      value = &buf[0];
+    }
   } else {
+    // Don't use char_at_addr(0) if length is 0
     value = (jchar*) s_value->base(T_CHAR);
   }
 
   // invoke the callback
-  return (*cb)(wrapper->klass_tag(),
-               wrapper->obj_size(),
-               wrapper->obj_tag_p(),
-               value,
-               (jint)s_len,
-               user_data);
+  jint res = (*cb)(wrapper->klass_tag(),
+                   wrapper->obj_size(),
+                   wrapper->obj_tag_p(),
+                   value,
+                   (jint)s_len,
+                   user_data);
+
+  if (is_latin1 && s_len > 0) {
+    FREE_C_HEAP_ARRAY(jchar, value);
+  }
+  return res;
 }
 
 // helper function to invoke string primitive value callback
--- a/hotspot/src/share/vm/runtime/globals.hpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/runtime/globals.hpp	Tue Nov 03 09:41:03 2015 +0100
@@ -2889,6 +2889,9 @@
   product(bool, AggressiveOpts, false,                                      \
           "Enable aggressive optimizations - see arguments.cpp")            \
                                                                             \
+  product_pd(bool, CompactStrings,                                          \
+          "Enable Strings to use single byte chars in backing store")       \
+                                                                            \
   product_pd(uintx, TypeProfileLevel,                                       \
           "=XYZ, with Z: Type profiling of arguments at call; "             \
                      "Y: Type profiling of return value at call; "          \
@@ -4259,6 +4262,9 @@
              "Use the FP register for holding the frame pointer "           \
              "and not as a general purpose register.")                      \
                                                                             \
+  diagnostic(bool, StringCharIntrinsics, true,                              \
+             "Inline String*.getChar/putChar intrinsics.")                  \
+                                                                            \
   diagnostic(bool, CheckIntrinsics, true,                                   \
              "When a class C is loaded, check that "                        \
              "(1) all intrinsics defined by the VM for class C are present "\
--- a/hotspot/src/share/vm/runtime/thread.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/runtime/thread.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -2933,12 +2933,8 @@
   if (thread_obj != NULL) {
     oop thread_group = java_lang_Thread::threadGroup(thread_obj);
     if (thread_group != NULL) {
-      typeArrayOop name = java_lang_ThreadGroup::name(thread_group);
       // ThreadGroup.name can be null
-      if (name != NULL) {
-        const char* str = UNICODE::as_utf8((jchar*) name->base(T_CHAR), name->length());
-        return str;
-      }
+      return java_lang_ThreadGroup::name(thread_group);
     }
   }
   return NULL;
@@ -2952,12 +2948,8 @@
     if (thread_group != NULL) {
       oop parent = java_lang_ThreadGroup::parent(thread_group);
       if (parent != NULL) {
-        typeArrayOop name = java_lang_ThreadGroup::name(parent);
         // ThreadGroup.name can be null
-        if (name != NULL) {
-          const char* str = UNICODE::as_utf8((jchar*) name->base(T_CHAR), name->length());
-          return str;
-        }
+        return java_lang_ThreadGroup::name(parent);
       }
     }
   }
@@ -3304,6 +3296,9 @@
 
   initialize_class(vmSymbols::java_lang_String(), CHECK);
 
+  // Inject CompactStrings value after the static initializers for String ran.
+  java_lang_String::set_compact_strings(CompactStrings);
+
   // Initialize java_lang.System (needed before creating the thread)
   initialize_class(vmSymbols::java_lang_System(), CHECK);
   // The VM creates & returns objects of this class. Make sure it's initialized.
--- a/hotspot/src/share/vm/utilities/utf8.cpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/utilities/utf8.cpp	Tue Nov 03 09:41:03 2015 +0100
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1997, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1997, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -27,7 +27,7 @@
 
 // Assume the utf8 string is in legal form and has been
 // checked in the class file parser/format checker.
-char* UTF8::next(const char* str, jchar* value) {
+template<typename T> char* UTF8::next(const char* str, T* value) {
   unsigned const char *ptr = (const unsigned char *)str;
   unsigned char ch, ch2, ch3;
   int length = -1;              /* bad length */
@@ -68,11 +68,11 @@
   } /* end of switch */
 
   if (length <= 0) {
-    *value = ptr[0];    /* default bad result; */
+    *value = (T)ptr[0];    /* default bad result; */
     return (char*)(ptr + 1); // make progress somehow
   }
 
-  *value = result;
+  *value = (T)result;
 
   // The assert is correct but the .class file is wrong
   // assert(UNICODE::utf8_size(result) == length, "checking reverse computation");
@@ -96,12 +96,22 @@
 // Count bytes of the form 10xxxxxx and deduct this count
 // from the total byte count.  The utf8 string must be in
 // legal form which has been verified in the format checker.
-int UTF8::unicode_length(const char* str, int len) {
+int UTF8::unicode_length(const char* str, int len, bool& is_latin1, bool& has_multibyte) {
   int num_chars = len;
+  has_multibyte = false;
+  is_latin1 = true;
+  unsigned char prev = 0;
   for (int i = 0; i < len; i++) {
-    if ((str[i] & 0xC0) == 0x80) {
+    unsigned char c = str[i];
+    if ((c & 0xC0) == 0x80) {
+      // Multibyte, check if valid latin1 character.
+      has_multibyte = true;
+      if (prev > 0xC3) {
+        is_latin1 = false;
+      }
       --num_chars;
     }
+    prev = c;
   }
   return num_chars;
 }
@@ -110,17 +120,28 @@
 // 10xxxxxx which only appear in multibyte characters.
 // The utf8 string must be in legal form and has been
 // verified in the format checker.
-int UTF8::unicode_length(const char* str) {
+int UTF8::unicode_length(const char* str, bool& is_latin1, bool& has_multibyte) {
   int num_chars = 0;
+  has_multibyte = false;
+  is_latin1 = true;
+  unsigned char prev = 0;
   for (const char* p = str; *p; p++) {
-    if (((*p) & 0xC0) != 0x80) {
+    unsigned char c = (*p);
+    if ((c & 0xC0) == 0x80) {
+      // Multibyte, check if valid latin1 character.
+      has_multibyte = true;
+      if (prev > 0xC3) {
+        is_latin1 = false;
+      }
+    } else {
       num_chars++;
     }
+    prev = c;
   }
   return num_chars;
 }
 
-// Writes a jchar a utf8 and returns the end
+// Writes a jchar as utf8 and returns the end
 static u_char* utf8_write(u_char* base, jchar ch) {
   if ((ch != 0) && (ch <=0x7f)) {
     base[0] = (u_char) ch;
@@ -145,7 +166,7 @@
   return base + 3;
 }
 
-void UTF8::convert_to_unicode(const char* utf8_str, jchar* unicode_str, int unicode_length) {
+template<typename T> void UTF8::convert_to_unicode(const char* utf8_str, T* unicode_str, int unicode_length) {
   unsigned char ch;
   const char *ptr = utf8_str;
   int index = 0;
@@ -153,7 +174,7 @@
   /* ASCII case loop optimization */
   for (; index < unicode_length; index++) {
     if((ch = ptr[0]) > 0x7F) { break; }
-    unicode_str[index] = ch;
+    unicode_str[index] = (T)ch;
     ptr = (const char *)(ptr + 1);
   }
 
@@ -162,6 +183,12 @@
   }
 }
 
+// Explicit instantiation for all supported string types.
+template char* UTF8::next<jchar>(const char* str, jchar* value);
+template char* UTF8::next<jbyte>(const char* str, jbyte* value);
+template void UTF8::convert_to_unicode<jchar>(const char* utf8_str, jchar* unicode_str, int unicode_length);
+template void UTF8::convert_to_unicode<jbyte>(const char* utf8_str, jbyte* unicode_str, int unicode_length);
+
 // returns the quoted ascii length of a 0-terminated utf8 string
 int UTF8::quoted_ascii_length(const char* utf8_str, int utf8_length) {
   const char *ptr = utf8_str;
@@ -306,9 +333,20 @@
                  + ((str[4] & 0x0f) << 6)  + (str[5] & 0x3f);
 }
 
-
 //-------------------------------------------------------------------------------------
 
+bool UNICODE::is_latin1(jchar c) {
+  return (c <= 0x00FF);
+}
+
+bool UNICODE::is_latin1(jchar* base, int length) {
+  for (int index = 0; index < length; index++) {
+    if (base[index] > 0x00FF) {
+      return false;
+    }
+  }
+  return true;
+}
 
 int UNICODE::utf8_size(jchar c) {
   if ((0x0001 <= c) && (c <= 0x007F)) return 1;
@@ -316,6 +354,11 @@
   return 3;
 }
 
+int UNICODE::utf8_size(jbyte c) {
+  if (c >= 0x0001) return 1;
+  return 2;
+}
+
 int UNICODE::utf8_length(jchar* base, int length) {
   int result = 0;
   for (int index = 0; index < length; index++) {
@@ -327,6 +370,15 @@
   return result;
 }
 
+int UNICODE::utf8_length(jbyte* base, int length) {
+  int result = 0;
+  for (int index = 0; index < length; index++) {
+    jbyte c = base[index];
+    result += utf8_size(c);
+  }
+  return result;
+}
+
 char* UNICODE::as_utf8(jchar* base, int length) {
   int utf8_len = utf8_length(base, length);
   u_char* buf = NEW_RESOURCE_ARRAY(u_char, utf8_len + 1);
@@ -335,6 +387,26 @@
   return result;
 }
 
+char* UNICODE::as_utf8(jbyte* base, int length) {
+  int utf8_len = utf8_length(base, length);
+  u_char* result = NEW_RESOURCE_ARRAY(u_char, utf8_len + 1);
+  u_char* p = result;
+  if (utf8_len == length) {
+    for (int index = 0; index < length; index++) {
+      *p++ = base[index];
+    }
+  } else {
+    // Unicode string contains U+0000 which should
+    // be encoded as 0xC080 in "modified" UTF8.
+    for (int index = 0; index < length; index++) {
+      p = utf8_write(p, ((jchar) base[index]) & 0xff);
+    }
+  }
+  *p = '\0';
+  assert(p == &result[utf8_len], "length prediction must be correct");
+  return (char*) result;
+}
+
 char* UNICODE::as_utf8(jchar* base, int length, char* buf, int buflen) {
   u_char* p = (u_char*)buf;
   for (int index = 0; index < length; index++) {
@@ -347,6 +419,26 @@
   return buf;
 }
 
+char* UNICODE::as_utf8(jbyte* base, int length, char* buf, int buflen) {
+  u_char* p = (u_char*)buf;
+  u_char* end = (u_char*)buf + buflen;
+  for (int index = 0; index < length; index++) {
+    jbyte c = base[index];
+    int sz = utf8_size(c);
+    buflen -= sz;
+    if (buflen <= 0) break; // string is truncated
+    if (sz == 1) {
+      *p++ = c;
+    } else {
+      // Unicode string contains U+0000 which should
+      // be encoded as 0xC080 in "modified" UTF8.
+      p = utf8_write(p, ((jchar) c) & 0xff);
+    }
+  }
+  *p = '\0';
+  return buf;
+}
+
 void UNICODE::convert_to_utf8(const jchar* base, int length, char* utf8_buffer) {
   for(int index = 0; index < length; index++) {
     utf8_buffer = (char*)utf8_write((u_char*)utf8_buffer, base[index]);
@@ -355,10 +447,11 @@
 }
 
 // returns the quoted ascii length of a unicode string
-int UNICODE::quoted_ascii_length(jchar* base, int length) {
+template<typename T>
+int UNICODE::quoted_ascii_length(T* base, int length) {
   int result = 0;
   for (int i = 0; i < length; i++) {
-    jchar c = base[i];
+    T c = base[i];
     if (c >= 32 && c < 127) {
       result++;
     } else {
@@ -368,12 +461,13 @@
   return result;
 }
 
-// converts a utf8 string to quoted ascii
-void UNICODE::as_quoted_ascii(const jchar* base, int length, char* buf, int buflen) {
+// converts a unicode string to quoted ascii
+template<typename T>
+void UNICODE::as_quoted_ascii(const T* base, int length, char* buf, int buflen) {
   char* p = buf;
   char* end = buf + buflen;
   for (int index = 0; index < length; index++) {
-    jchar c = base[index];
+    T c = base[index];
     if (c >= 32 && c < 127) {
       if (p + 1 >= end) break;      // string is truncated
       *p++ = (char)c;
@@ -386,6 +480,13 @@
   *p = '\0';
 }
 
+// Explicit instantiation for all supported types.
+template int UNICODE::quoted_ascii_length<jbyte>(jbyte* base, int length);
+template int UNICODE::quoted_ascii_length<jchar>(jchar* base, int length);
+template void UNICODE::as_quoted_ascii<jbyte>(const jbyte* base, int length, char* buf, int buflen);
+template void UNICODE::as_quoted_ascii<jchar>(const jchar* base, int length, char* buf, int buflen);
+
+
 #ifndef PRODUCT
 void TestAsUtf8() {
   char res[60];
--- a/hotspot/src/share/vm/utilities/utf8.hpp	Mon Nov 02 12:34:27 2015 +0000
+++ b/hotspot/src/share/vm/utilities/utf8.hpp	Tue Nov 03 09:41:03 2015 +0100
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1997, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1997, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -33,13 +33,21 @@
 class UTF8 : AllStatic {
  public:
   // returns the unicode length of a 0-terminated utf8 string
-  static int unicode_length(const char* utf8_str);
+  static int unicode_length(const char* utf8_str) {
+    bool is_latin1, has_multibyte;
+    return unicode_length(utf8_str, is_latin1, has_multibyte);
+  }
+  static int unicode_length(const char* utf8_str, bool& is_latin1, bool& has_multibyte);
 
   // returns the unicode length of a non-0-terminated utf8 string
-  static int unicode_length(const char* utf8_str, int len);
+  static int unicode_length(const char* utf8_str, int len) {
+    bool is_latin1, has_multibyte;
+    return unicode_length(utf8_str, len, is_latin1, has_multibyte);
+  }
+  static int unicode_length(const char* utf8_str, int len, bool& is_latin1, bool& has_multibyte);
 
   // converts a utf8 string to a unicode string
-  static void convert_to_unicode(const char* utf8_str, jchar* unicode_buffer, int unicode_length);
+  template<typename T> static void convert_to_unicode(const char* utf8_str, T* unicode_str, int unicode_length);
 
   // returns the quoted ascii length of a utf8 string
   static int quoted_ascii_length(const char* utf8_str, int utf8_length);
@@ -53,7 +61,7 @@
 
   // decodes the current utf8 character, stores the result in value,
   // and returns the end of the current utf8 chararacter.
-  static char* next(const char* str, jchar* value);
+  template<typename T> static char* next(const char* str, T* value);
 
   // decodes the current utf8 character, gets the supplementary character instead of
   // the surrogate pair when seeing a supplementary character in string,
@@ -76,11 +84,19 @@
 
 class UNICODE : AllStatic {
  public:
+  // checks if the given unicode character can be encoded as latin1
+  static bool is_latin1(jchar c);
+
+  // checks if the given string can be encoded as latin1
+  static bool is_latin1(jchar* base, int length);
+
   // returns the utf8 size of a unicode character
   static int utf8_size(jchar c);
+  static int utf8_size(jbyte c);
 
   // returns the utf8 length of a unicode string
   static int utf8_length(jchar* base, int length);
+  static int utf8_length(jbyte* base, int length);
 
   // converts a unicode string to utf8 string
   static void convert_to_utf8(const jchar* base, int length, char* utf8_buffer);
@@ -88,13 +104,15 @@
   // converts a unicode string to a utf8 string; result is allocated
   // in resource area unless a buffer is provided.
   static char* as_utf8(jchar* base, int length);
+  static char* as_utf8(jbyte* base, int length);
   static char* as_utf8(jchar* base, int length, char* buf, int buflen);
+  static char* as_utf8(jbyte* base, int length, char* buf, int buflen);
 
   // returns the quoted ascii length of a unicode string
-  static int quoted_ascii_length(jchar* base, int length);
+  template<typename T> static int quoted_ascii_length(T* base, int length);
 
-  // converts a utf8 string to quoted ascii
-  static void as_quoted_ascii(const jchar* base, int length, char* buf, int buflen);
+  // converts a unicode string to quoted ascii
+  template<typename T> static void as_quoted_ascii(const T* base, int length, char* buf, int buflen);
 };
 
 #endif // SHARE_VM_UTILITIES_UTF8_HPP
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/test/compiler/intrinsics/string/TestStringIntrinsics.java	Tue Nov 03 09:41:03 2015 +0100
@@ -0,0 +1,435 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+
+import java.lang.annotation.*;
+import java.lang.reflect.*;
+import java.util.Arrays;
+
+/*
+ * @test
+ * @bug 8054307
+ * @summary Tests correctness of string related intrinsics and C2 optimizations.
+ * @run main/timeout=240 TestStringIntrinsics
+ */
+public class TestStringIntrinsics {
+
+    public enum Operation {
+        ARR_EQUALS_B, ARR_EQUALS_C, EQUALS, COMPARE_TO, INDEX_OF, INDEX_OF_CON_U, INDEX_OF_CON_L,
+        INDEX_OF_CON_UL, CONCAT, CONCAT_C, CONCAT_I, CONCAT_M, INDEX_OF_CHAR
+    }
+
+    @Retention(RetentionPolicy.RUNTIME)
+    @Target(ElementType.METHOD)
+    @interface Test {
+        Operation op();
+        String constString() default "";
+        String[] inStrings() default {};
+        char[] inChars() default {};
+        int[] inInts() default {};
+        String[] outStrings() default {};
+    }
+
+    public static void main(String[] args) throws Exception {
+        new TestStringIntrinsics().run();
+    }
+
+    public void run() throws Exception {
+        // Build latin1 and UTF16 strings
+        StringBuilder latin1Builder = new StringBuilder();
+        for (int i = 0; i <= 255; ++i) {
+            latin1Builder.append((char) i);
+        }
+        String latin1 = latin1Builder.toString();
+        StringBuilder utf16Builder = new StringBuilder();
+        for (int i = 0; i <= 10000; ++i) {
+            utf16Builder.append((char) i);
+        }
+        String utf16 = utf16Builder.toString();
+
+        // Invoke test methods
+        for (Method m : TestStringIntrinsics.class.getMethods()) {
+            if (m.isAnnotationPresent(Test.class)) {
+                System.out.print("Checking " + m.getName() + "... ");
+                Operation op = m.getAnnotation(Test.class).op();
+                Test antn = m.getAnnotation(Test.class);
+                if (isStringConcatTest(op)) {
+                    checkStringConcat(op, m, antn);
+                } else {
+                    checkIntrinsics(op, m, latin1, utf16, antn);
+                }
+                System.out.println("Done.");
+            }
+        }
+    }
+
+    private boolean isStringConcatTest(Operation op) {
+        return op == Operation.CONCAT ||
+               op == Operation.CONCAT_C ||
+               op == Operation.CONCAT_I ||
+               op == Operation.CONCAT_M;
+    }
+
+    /**
+     * Checks correctness of the String.equals, String.compareTo and String.indexOf intrinsics.
+     * -XX:SpecialStringEquals
+     * -XX:SpecialStringCompareTo
+     * -XX:SpecialStringIndexOf
+     */
+    private void checkIntrinsics(Operation op, Method m, String latin1, String utf16, Test antn) throws Exception {
+        for (int i = 0; i < 50_000; ++i) {
+            // Copy and permute latin1 and UTF16 string
+            char[] arrL = latin1.toCharArray();
+            int indexL = i % arrL.length;
+            int mod = (arrL.length - arrL[indexL]);
+            int incL = i % ((mod != 0) ? mod : 1);
+            arrL[indexL] = (char) ((int) arrL[indexL] + incL);
+            String latin1Copy = String.valueOf(arrL);
+
+            char[] arrU = utf16.toCharArray();
+            int indexU = i % arrU.length;
+            mod = (arrU.length - arrU[indexU]);
+            int incU = i % ((mod != 0) ? mod : 1);
+            arrU[indexU] = (char) ((int) arrU[indexU] + incU);
+            String utf16Copy = String.valueOf(arrU);
+
+            switch (op) {
+            case ARR_EQUALS_B:
+                invokeAndCheck(m, (incL == 0), latin1.getBytes("ISO-8859-1"), latin1Copy.getBytes("ISO-8859-1"));
+                invokeAndCheck(m, true, new byte[] {1, 2, 3}, new byte[] {1, 2, 3});
+                invokeAndCheck(m, true, new byte[] {1}, new byte[] {1});
+                invokeAndCheck(m, true, new byte[] {}, new byte[] {});
+                break;
+            case ARR_EQUALS_C:
+                invokeAndCheck(m, (incU == 0), utf16.toCharArray(), arrU);
+                break;
+            case EQUALS:
+                invokeAndCheck(m, (incL == 0), latin1, latin1Copy);
+                invokeAndCheck(m, false, latin1, "");
+                invokeAndCheck(m, false, "", latin1);
+
+                invokeAndCheck(m, (incU == 0), utf16, utf16Copy);
+                invokeAndCheck(m, false, utf16, "");
+                invokeAndCheck(m, false, "", utf16);
+
+                invokeAndCheck(m, false, latin1, utf16);
+                break;
+            case COMPARE_TO:
+                invokeAndCheck(m, -incL, latin1, latin1Copy);
+                invokeAndCheck(m, latin1.length(), latin1, "");
+
+                invokeAndCheck(m, -incU, utf16, utf16Copy);
+                invokeAndCheck(m, utf16.length(), utf16, "");
+
+                // Cross coder
+                char cL = latin1.charAt(indexL);
+                char cU = utf16.charAt(indexU);
+                invokeAndCheck(m, cL - cU, latin1, latin1.replace(cL, cU));
+                invokeAndCheck(m, cU - cL, utf16, utf16.replace(cU, cL));
+
+                // Different lengths
+                invokeAndCheck(m, 1, "ABCD", "ABC");
+                invokeAndCheck(m, -1, "\uff21\uff22\uff23", "\uff21\uff22\uff23\uff24");
+                invokeAndCheck(m, 1, "ABC\uff24", "ABC");
+                invokeAndCheck(m, 3, "ABC\uff24\uff25\uff26", "ABC");
+                invokeAndCheck(m, -1, "ABC","ABC\uff24");
+                invokeAndCheck(m, -3, "ABC","ABC\uff24\uff25\uff26");
+                break;
+            case INDEX_OF:
+                invokeAndCheck(m, indexL, latin1, latin1.substring(indexL), (indexL > 42) ? 42 : 0);
+                invokeAndCheck(m, 0, latin1, "", 0);
+
+                invokeAndCheck(m, indexU, utf16, utf16.substring(indexU), (indexU > 42) ? 42 : 0);
+                invokeAndCheck(m, 0, utf16, "", 0);
+
+                // Cross coder
+                invokeAndCheck(m, -1, latin1.substring(0, indexL), utf16.substring(indexU), (indexL > 42) ? 42 : 0);
+                // Skip latin1 chars in utf16 string
+                int start = 256;
+                int end = indexU > start ? indexU : start;
+                invokeAndCheck(m, end-start, utf16.substring(start, end) + latin1.substring(indexL), latin1.substring(indexL), 0);
+                break;
+            case INDEX_OF_CON_L:
+                invokeAndCheck(m, antn.constString(), latin1);
+                break;
+            case INDEX_OF_CON_U:
+                invokeAndCheck(m, antn.constString(), utf16);
+                break;
+            case INDEX_OF_CON_UL:
+                invokeAndCheck(m, antn.constString(), utf16);
+                break;
+            case INDEX_OF_CHAR:
+                invokeAndCheck(m, 7, "abcdefg\uD800\uDC00", 65536, 0);
+                invokeAndCheck(m, -1, "abcdefg\uD800\uDC01", 65536, 0);
+                invokeAndCheck(m, -1, "abcdefg\uD800", 65536, 0);
+                invokeAndCheck(m, 3, "abc\u0107", 263, 0);
+                invokeAndCheck(m, -1, "abc\u0108", 263, 0);
+                invokeAndCheck(m, 7, "abcdefg\u0107", 263, 0);
+                invokeAndCheck(m, 7, "abcdefg\u0107", 263, -1);
+                invokeAndCheck(m, 0, "\u0107", 263, 0);
+                break;
+            default:
+                throw new RuntimeException("Unexpected operation.");
+            }
+        }
+    }
+
+    /**
+     * Checks correctness of the C2 string concatenation optimization.
+     * -XX:OptimizeStringConcat
+     */
+    private void checkStringConcat(Operation op, Method m, Test antn) throws Exception {
+        for (int i = 0; i < 50_000; ++i) {
+            String[] result = antn.outStrings();
+            switch(op) {
+            case CONCAT:
+                String[] strs = antn.inStrings();
+                for (int j = 0; j < strs.length; ++j) {
+                    invokeAndCheck(m, result[j], strs[j]);
+                }
+                break;
+            case CONCAT_C:
+                char[] ch = antn.inChars();
+                for (int j = 0; j < ch.length; ++j) {
+                    invokeAndCheck(m, result[j], ch[j]);
+                }
+                break;
+            case CONCAT_I:
+                int[] k = antn.inInts();
+                for (int j = 0; j < k.length; ++j) {
+                    invokeAndCheck(m, result[j], k[j]);
+                }
+                break;
+            case CONCAT_M:
+                strs = antn.inStrings();
+                ch = antn.inChars();
+                k = antn.inInts();
+                for (int j = 0; j < strs.length; ++j) {
+                    invokeAndCheck(m, result[j], strs[j], ch[j], k[j]);
+                }
+                break;
+            default:
+                throw new RuntimeException("Unexpected operation.");
+            }
+        }
+    }
+
+    /**
+     * Invokes method 'm' by passing arguments 'args' and checks if the
+     * returned value equals 'expectedResult'.
+     */
+    private void invokeAndCheck(Method m, Object expectedResult, Object... args) throws Exception {
+        Object result = m.invoke(null, args);
+        if (!result.equals(expectedResult)) {
+//            System.out.println("Expected:");
+//            System.out.println(expectedResult);
+//            System.out.println("Returned:");
+//            System.out.println(result);
+            throw new RuntimeException("Result of '" + m.getName() + "' not equal to expected value.");
+        }
+    }
+
+    /*
+     * Constants
+     */
+    static final char charU = '\uff21';
+    static final char charL = 'A';
+    static final String emptyString = "";
+    static final String stringL = "abcdefghijklmnop";
+    static final String stringSmallL = "abc";
+    static final String stringU = "\u0f21\u0f22\u0f23\u0f24\u0f25\u0f26\u0f27\u0f28";
+    static final String stringSmallU = "\u0f21\u0f22\u0f23";
+    static final int constInt = 123;
+    static final int constIntNeg = -123;
+
+    /*
+     * Arrays.equals
+     */
+    @Test(op = Operation.ARR_EQUALS_B)
+    public static boolean arrayEqualsB(byte[] a, byte[] b) {
+      return Arrays.equals(a, b);
+    }
+
+    @Test(op = Operation.ARR_EQUALS_C)
+    public static boolean arrayEqualsC(char[] a, char[] b) {
+      return Arrays.equals(a, b);
+    }
+
+    /*
+     * String.equals
+     */
+    @Test(op = Operation.EQUALS)
+    public static boolean equals(String a, String b) {
+        return a.equals(b);
+    }
+
+    /*
+     * String.compareTo
+     */
+    @Test(op = Operation.COMPARE_TO)
+    public static int compareTo(String a, String b) {
+        return a.compareTo(b);
+    }
+
+    /*
+     * String.indexOf
+     */
+    @Test(op = Operation.INDEX_OF)
+    public static int indexOf(String a, String b, int from) {
+        return a.indexOf(b, from);
+    }
+
+    @Test(op = Operation.INDEX_OF_CON_U, constString = stringSmallU)
+    public static String indexOfConstU(String a) {
+        int result = a.indexOf(stringSmallU);
+        return a.substring(result, result + stringSmallU.length());
+    }
+
+    @Test(op = Operation.INDEX_OF_CON_U, constString = stringU)
+    public static String indexOfConstLargeU(String a) {
+        int result = a.indexOf(stringU);
+        return a.substring(result, result + stringU.length());
+    }
+
+    @Test(op = Operation.INDEX_OF_CON_U, constString = emptyString)
+    public static String indexOfConstEmptyU(String a) {
+        int result = a.indexOf(emptyString);
+        return a.substring(result, result + emptyString.length());
+    }
+
+    @Test(op = Operation.INDEX_OF_CON_L, constString = stringSmallL)
+    public static String indexOfConstL(String a) {
+        int result = a.indexOf(stringSmallL);
+        return a.substring(result, result + stringSmallL.length());
+    }
+
+    @Test(op = Operation.INDEX_OF_CON_L, constString = stringL)
+    public static String indexOfConstLargeL(String a) {
+        int result = a.indexOf(stringL);
+        return a.substring(result, result + stringL.length());
+    }
+
+    @Test(op = Operation.INDEX_OF_CON_L, constString = emptyString)
+    public static String indexOfConstEmptyL(String a) {
+        int result = a.indexOf(emptyString);
+        return a.substring(result, result + emptyString.length());
+    }
+
+    @Test(op = Operation.INDEX_OF_CON_UL, constString = stringSmallL)
+    public static String indexOfConstUL(String a) {
+        int result = a.indexOf(stringSmallL);
+        return a.substring(result, result + stringSmallL.length());
+    }
+
+    @Test(op = Operation.INDEX_OF_CON_UL, constString = stringL)
+    public static String indexOfConstLargeUL(String a) {
+        int result = a.indexOf(stringL);
+        return a.substring(result, result + stringL.length());
+    }
+
+    @Test(op = Operation.INDEX_OF_CHAR)
+    public static int indexOfChar(String a, int ch, int from) {
+        return a.indexOf(ch, from);
+    }
+
+    /*
+     * String concatenation optimization
+     */
+    @Test(op = Operation.CONCAT, inStrings = {"ABC", "\uff21\uff22\uff23"}, outStrings = {"ABC", "\uff21\uff22\uff23"})
+    public static String concatString(String a) {
+        return new StringBuilder().append(a).toString();
+    }
+
+    @Test(op = Operation.CONCAT, inStrings = {""}, outStrings = {""})
+    public static String concatStringEmpty(String a) {
+        return new StringBuilder().toString();
+    }
+
+    @Test(op = Operation.CONCAT, inStrings = {""}, outStrings = {"null"})
+    public static String concatStringNull(String a) {
+        return new StringBuilder().append((String)null).toString();
+    }
+
+    @Test(op = Operation.CONCAT, inStrings = {"ABC", "\uff21\uff22\uff23"}, outStrings = {"abcdefghijklmnopABCabc", "abcdefghijklmnop\uff21\uff22\uff23abc"})
+    public static String concatStringConstL(String a) {
+        return new StringBuilder().append(stringL).append(a).append(stringSmallL).toString();
+    }
+
+    @Test(op = Operation.CONCAT, inStrings = {"ABC", "\uff21\uff22\uff23"}, outStrings = {"\u0f21\u0f22\u0f23ABC\u0f21\u0f22\u0f23\u0f24\u0f25\u0f26\u0f27\u0f28", "\u0f21\u0f22\u0f23\uff21\uff22\uff23\u0f21\u0f22\u0f23\u0f24\u0f25\u0f26\u0f27\u0f28"})
+    public static String concatStringConstU(String a) {
+        return new StringBuilder().append(stringSmallU).append(a).append(stringU).toString();
+    }
+
+    @Test(op = Operation.CONCAT_C, inChars = {'A', '\uff21'}, outStrings = {"A", "\uff21"})
+    public static String concatChar(char a) {
+        return new StringBuilder().append(a).toString();
+    }
+
+    @Test(op = Operation.CONCAT_C, inChars = {'A', '\uff21'}, outStrings = {"abcdefghijklmnopAabcA\uff21", "abcdefghijklmnop\uff21abcA\uff21"})
+    public static String concatCharConstL(char a) {
+        return new StringBuilder().append(stringL).append(a).append(stringSmallL).append(charL).append(charU).toString();
+    }
+
+    @Test(op = Operation.CONCAT_C, inChars = {'A', '\uff21'}, outStrings = {"\u0f21\u0f22\u0f23A\u0f21\u0f22\u0f23\u0f24\u0f25\u0f26\u0f27\u0f28\uff21A", "\u0f21\u0f22\u0f23\uff21\u0f21\u0f22\u0f23\u0f24\u0f25\u0f26\u0f27\u0f28\uff21A"})
+    public static String concatCharConstU(char a) {
+        return new StringBuilder().append(stringSmallU).append(a).append(stringU).append(charU).append(charL).toString();
+    }
+
+    @Test(op = Operation.CONCAT_I, inInts = {Integer.MIN_VALUE, -42, 42, Integer.MAX_VALUE}, outStrings = {"-2147483648", "-42", "42", "2147483647"})
+    public static String concatInt(int a) {
+        return new StringBuilder().append(a).toString();
+    }
+
+    @Test(op = Operation.CONCAT_I, inInts = {Integer.MIN_VALUE, -42, 42, Integer.MAX_VALUE}, outStrings = {"abcdefghijklmnop-2147483648abc123-123", "abcdefghijklmnop-42abc123-123", "abcdefghijklmnop42abc123-123", "abcdefghijklmnop2147483647abc123-123"})
+    public static String concatIntConstL(int b) {
+        return new StringBuilder().append(stringL).append(b).append(stringSmallL).append(constInt).append(constIntNeg).toString();
+    }
+
+    @Test(op = Operation.CONCAT_I, inInts = {Integer.MIN_VALUE, -42, 42, Integer.MAX_VALUE}, outStrings = {"\u0f21\u0f22\u0f23-2147483648\u0f21\u0f22\u0f23\u0f24\u0f25\u0f26\u0f27\u0f28123-123", "\u0f21\u0f22\u0f23-42\u0f21\u0f22\u0f23\u0f24\u0f25\u0f26\u0f27\u0f28123-123", "\u0f21\u0f22\u0f2342\u0f21\u0f22\u0f23\u0f24\u0f25\u0f26\u0f27\u0f28123-123", "\u0f21\u0f22\u0f232147483647\u0f21\u0f22\u0f23\u0f24\u0f25\u0f26\u0f27\u0f28123-123"})
+    public static String concatIntConstU(int b) {
+        return new StringBuilder().append(stringSmallU).append(b).append(stringU).append(constInt).append(constIntNeg).toString();
+    }
+
+    @Test(op = Operation.CONCAT, inStrings = {""}, outStrings = {"nullabcabcdefghijklmnopA123-123"})
+    public static String concatConstL(String a) {
+        return new StringBuilder().append((String)null).append(stringSmallL).append(stringL).append(charL).append(constInt).append(constIntNeg).toString();
+    }
+
+    @Test(op = Operation.CONCAT, inStrings = {""}, outStrings = {"nullabcabcdefghijklmnop\u0f21\u0f22\u0f23\u0f21\u0f22\u0f23\u0f24\u0f25\u0f26\u0f27\u0f28A\uff21123-123"})
+    public static String concatConstU(String a) {
+        return new StringBuilder().append((String)null).append(stringSmallL).append(stringL).append(stringSmallU).append(stringU).append(charL).append(charU).append(constInt).append(constIntNeg).toString();
+    }
+
+    @Test(op = Operation.CONCAT_M,
+          inStrings = {"ABCDEFG", "ABCDEFG", "\u0f21\u0f22\u0f23\u0f24\u0f25\u0f26\u0f27\u0f28", "\u0f21\u0f22\u0f23\u0f24\u0f25\u0f26\u0f27\u0f28"},
+          inChars = {'A', '\uff21', 'A', '\uff21'},
+          inInts = {Integer.MIN_VALUE, Integer.MAX_VALUE, Integer.MIN_VALUE, Integer.MAX_VALUE},
+          outStrings = {"ABCDEFGA-2147483648nullabcdefghijklmnop123-123A\u0f21\u0f22\u0f23\u0f24\u0f25\u0f26\u0f27\u0f28\uff21ABCDEFGA-2147483648null",
+                        "ABCDEFG\uff212147483647nullabcdefghijklmnop123-123A\u0f21\u0f22\u0f23\u0f24\u0f25\u0f26\u0f27\u0f28\uff21ABCDEFG\uff212147483647null",
+                        "\u0f21\u0f22\u0f23\u0f24\u0f25\u0f26\u0f27\u0f28A-2147483648nullabcdefghijklmnop123-123A\u0f21\u0f22\u0f23\u0f24\u0f25\u0f26\u0f27\u0f28\uff21\u0f21\u0f22\u0f23\u0f24\u0f25\u0f26\u0f27\u0f28A-2147483648null",
+            "\u0f21\u0f22\u0f23\u0f24\u0f25\u0f26\u0f27\u0f28\uff212147483647nullabcdefghijklmnop123-123A\u0f21\u0f22\u0f23\u0f24\u0f25\u0f26\u0f27\u0f28\uff21\u0f21\u0f22\u0f23\u0f24\u0f25\u0f26\u0f27\u0f28\uff212147483647null"})
+    public static String concatMixed(String a, char b, int c) {
+        return new StringBuilder().append(a).append(b).append(c).append((String)null)
+                .append(stringL).append(constInt).append(constIntNeg).append(charL).append(stringU).append(charU)
+                .append(a).append(b).append(c).append((String)null).toString();
+    }
+}
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/test/runtime/Annotations/TestAnnotatedStringEncoding.java	Tue Nov 03 09:41:03 2015 +0100
@@ -0,0 +1,150 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+
+import java.lang.annotation.*;
+import java.lang.reflect.*;
+
+/*
+ * @test
+ * @bug 8054307
+ * @summary Tests the correct encoding of latin1/UTF16 Strings used in annotations.
+ */
+public class TestAnnotatedStringEncoding {
+    @Retention(RetentionPolicy.RUNTIME)
+    @Target(ElementType.METHOD)
+    @interface Test {
+        String str();
+        int index();
+    }
+
+    public static void main(String[] args) throws Exception {
+        new TestAnnotatedStringEncoding().run();
+    }
+
+    public void run() {
+        // Iterate over annotated methods and retrieve the string
+        for (Method m : this.getClass().getMethods()) {
+            if (m.isAnnotationPresent(Test.class)) {
+                // Check if string equals expected value
+                Test test = m.getAnnotation(Test.class);
+                String str = test.str();
+                int index = test.index();
+                if (!str.equals(strValue[index])) {
+                    throw new RuntimeException(m.getName() + " failed: \"" + str + "\" (0x" + Integer.toHexString(str.charAt(0)) +
+                            ") does not equal \"" + strValue[index] + "\" (0x" + Integer.toHexString(strValue[index].charAt(0)) + ") .");
+                }
+            }
+        }
+        System.out.println("Test passed.");
+    }
+
+    public static String[] strValue = {
+        "\u0000", "\u0020", "\u0021", "\u0080",
+        "\u00FF", "\u0100", "\u017F", "\u01FF",
+        "\u07FF", "\u0800", "\uC280", "\uC2BF",
+        "\uC380", "\uC3BF", "\uC5BF", "\uFFFF",
+        "\u10000", "\u1FFFFF", "\u200000",
+        "\u3FFFFFF", "\u4000000", "\u7FFFFFFF",
+        "ab\uff23\uff24ef\uff27", "\uff21\uff22cd\uff25g", "\u00FF\u00FF\u00FF", "\u00A1\u00A1\u00A1\u00A1", ""};
+
+    @Test(str = "\u0000", index = 0)
+    public static void check0() { }
+
+    @Test(str = "\u0020", index = 1)
+    public static void check1() { }
+
+    @Test(str = "\u0021", index = 2)
+    public static void check2() { }
+
+    @Test(str = "\u0080", index = 3)
+    public static void check3() { }
+
+    @Test(str = "\u00FF", index = 4)
+    public static void check4() { }
+
+    @Test(str = "\u0100", index = 5)
+    public static void check5() { }
+
+    @Test(str = "\u017F", index = 6)
+    public static void check6() { }
+
+    @Test(str = "\u01FF", index = 7)
+    public static void check7() { }
+
+    @Test(str = "\u07FF", index = 8)
+    public static void check8() { }
+
+    @Test(str = "\u0800", index = 9)
+    public static void check9() { }
+
+    @Test(str = "\uC280", index = 10)
+    public static void check10() { }
+
+    @Test(str = "\uC2BF", index = 11)
+    public static void check11() { }
+
+    @Test(str = "\uC380", index = 12)
+    public static void check12() { }
+
+    @Test(str = "\uC3BF", index = 13)
+    public static void check13() { }
+
+    @Test(str = "\uC5BF", index = 14)
+    public static void check14() { }
+
+    @Test(str = "\uFFFF", index = 15)
+    public static void check15() { }
+
+    @Test(str = "\u10000", index = 16)
+    public static void check16() { }
+
+    @Test(str = "\u1FFFFF", index = 17)
+    public static void check17() { }
+
+    @Test(str = "\u200000", index = 18)
+    public static void check18() { }
+
+    @Test(str = "\u3FFFFFF", index = 19)
+    public static void check19() { }
+
+    @Test(str = "\u4000000", index = 20)
+    public static void check20() { }
+
+    @Test(str = "\u7FFFFFFF", index = 21)
+    public static void check21() { }
+
+    @Test(str = "ab\uff23\uff24ef\uff27", index = 22)
+    public static void check22() { }
+
+    @Test(str = "\uff21\uff22cd\uff25g", index = 23)
+    public static void check23() { }
+
+    @Test(str = "\u00FF\u00FF\u00FF", index = 24)
+    public static void check24() { }
+
+    @Test(str = "\u00A1\u00A1\u00A1\u00A1", index = 25)
+    public static void check25() { }
+
+    @Test(str = "", index = 26)
+    public static void check26() { }
+}
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/test/runtime/SharedArchiveFile/CdsDifferentCompactStrings.java	Tue Nov 03 09:41:03 2015 +0100
@@ -0,0 +1,75 @@
+/*
+ * Copyright (c) 2013, 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * 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.
+ */
+
+/*
+ * @test CdsDifferentCompactStrings
+ * @summary CDS (class data sharing) requires the same -XX:[+-]CompactStrings
+ *          setting between archive creation time and load time.
+ * @library /testlibrary
+ * @modules java.base/sun.misc
+ *          java.management
+ */
+
+import jdk.test.lib.*;
+
+public class CdsDifferentCompactStrings {
+    public static void main(String[] args) throws Exception {
+        createAndLoadSharedArchive("+", "-");
+        createAndLoadSharedArchive("-", "+");
+    }
+
+    private static void createAndLoadSharedArchive(String create, String load)
+        throws Exception
+    {
+        String createCompactStringsArgument = "-XX:" + create + "CompactStrings";
+        String loadCompactStringsArgument   = "-XX:" + load   + "CompactStrings";
+
+        String filename = "./CdsDifferentCompactStrings" + create + ".jsa";
+
+        ProcessBuilder pb = ProcessTools.createJavaProcessBuilder(
+            "-XX:+UnlockDiagnosticVMOptions",
+            "-XX:SharedArchiveFile=" + filename,
+            "-Xshare:dump",
+            createCompactStringsArgument);
+
+        OutputAnalyzer output = new OutputAnalyzer(pb.start());
+        output.shouldContain("Loading classes to share");
+        output.shouldHaveExitValue(0);
+
+        pb = ProcessTools.createJavaProcessBuilder(
+            "-XX:+UnlockDiagnosticVMOptions",
+            "-XX:SharedArchiveFile=" + filename,
+            "-Xshare:on",
+            loadCompactStringsArgument,
+            "-version");
+
+        output = new OutputAnalyzer(pb.start());
+        try {
+            output.shouldContain("The shared archive file's CompactStrings " +
+                "setting .* does not equal the current CompactStrings setting");
+        } catch (RuntimeException e) {
+            output.shouldContain("Unable to use shared archive");
+        }
+        output.shouldHaveExitValue(1);
+    }
+}