diff --git a/compiler/dex/quick/x86/assemble_x86.cc b/compiler/dex/quick/x86/assemble_x86.cc
index 0b7b2bd..96dc6ee 100644
--- a/compiler/dex/quick/x86/assemble_x86.cc
+++ b/compiler/dex/quick/x86/assemble_x86.cc
@@ -246,7 +246,9 @@
   UNARY_ENCODING_MAP(Idivmod, 0x7, 0, SETS_CCODES, DaR, kRegRegReg, IS_UNARY_OP | REG_USE0, DaM, kRegRegMem, IS_BINARY_OP | REG_USE0, DaA, kRegRegArray, IS_QUAD_OP | REG_USE01, 0, REG_DEFA_USEA, REG_DEFAD_USEAD, REG_DEFAD_USEAD, "ah:al,ax,", "dx:ax,dx:ax,", "edx:eax,edx:eax,"),
 #undef UNARY_ENCODING_MAP
 
-  { kX86Bswap32R, kRegOpcode, IS_UNARY_OP | REG_DEF0_USE0, { 0, 0, 0x0F, 0xC8, 0, 0, 0, 0 }, "Bswap32R", "!0r" },
+  { kX86Bswap32R, kRegOpcode, IS_UNARY_OP | REG_DEF0_USE0,       { 0, 0, 0x0F, 0xC8, 0, 0, 0, 0 }, "Bswap32R", "!0r" },
+  { kX86Push32R,  kRegOpcode, IS_UNARY_OP | REG_USE0 | IS_STORE, { 0, 0, 0x50, 0,    0, 0, 0, 0 }, "Push32R",  "!0r" },
+  { kX86Pop32R,   kRegOpcode, IS_UNARY_OP | REG_DEF0 | IS_LOAD,  { 0, 0, 0x58, 0,    0, 0, 0, 0 }, "Pop32R",   "!0r" },
 
 #define EXT_0F_ENCODING_MAP(opname, prefix, opcode, reg_def) \
 { kX86 ## opname ## RR, kRegReg,             IS_BINARY_OP   | reg_def | REG_USE01,  { prefix, 0, 0x0F, opcode, 0, 0, 0, 0 }, #opname "RR", "!0r,!1r" }, \
@@ -308,6 +310,8 @@
   { kX86CmpxchgAR, kArrayReg, IS_STORE | IS_QUIN_OP | REG_USE014 | REG_DEFA_USEA | SETS_CCODES, { 0, 0, 0x0F, 0xB1, 0, 0, 0, 0 }, "Cmpxchg", "[!0r+!1r<<!2d+!3d],!4r" },
   { kX86LockCmpxchgMR, kMemReg,   IS_STORE | IS_TERTIARY_OP | REG_USE02 | REG_DEFA_USEA | SETS_CCODES, { 0xF0, 0, 0x0F, 0xB1, 0, 0, 0, 0 }, "Lock Cmpxchg", "[!0r+!1d],!2r" },
   { kX86LockCmpxchgAR, kArrayReg, IS_STORE | IS_QUIN_OP | REG_USE014 | REG_DEFA_USEA | SETS_CCODES, { 0xF0, 0, 0x0F, 0xB1, 0, 0, 0, 0 }, "Lock Cmpxchg", "[!0r+!1r<<!2d+!3d],!4r" },
+  { kX86LockCmpxchg8bM, kMem,   IS_STORE | IS_BINARY_OP | REG_USE0 | REG_DEFAD_USEAD | REG_USEC | REG_USEB | SETS_CCODES, { 0xF0, 0, 0x0F, 0xC7, 0, 1, 0, 0 }, "Lock Cmpxchg8b", "[!0r+!1d]" },
+  { kX86LockCmpxchg8bA, kArray, IS_STORE | IS_QUAD_OP | REG_USE01 | REG_DEFAD_USEAD | REG_USEC | REG_USEB | SETS_CCODES, { 0xF0, 0, 0x0F, 0xC7, 0, 1, 0, 0 }, "Lock Cmpxchg8b", "[!0r+!1r<<!2d+!3d]" },
 
   EXT_0F_ENCODING_MAP(Movzx8,  0x00, 0xB6, REG_DEF0),
   EXT_0F_ENCODING_MAP(Movzx16, 0x00, 0xB7, REG_DEF0),
diff --git a/compiler/dex/quick/x86/int_x86.cc b/compiler/dex/quick/x86/int_x86.cc
index b65fe54..0133a0a 100644
--- a/compiler/dex/quick/x86/int_x86.cc
+++ b/compiler/dex/quick/x86/int_x86.cc
@@ -293,7 +293,26 @@
   // If is_long, high half is in info->args[7]
 
   if (is_long) {
-    LOG(FATAL) << "CAS64: Not implemented";
+    FlushAllRegs();
+    LockCallTemps();
+    NewLIR1(kX86Push32R, rDI);
+    MarkTemp(rDI);
+    LockTemp(rDI);
+    NewLIR1(kX86Push32R, rSI);
+    MarkTemp(rSI);
+    LockTemp(rSI);
+    LoadValueDirectFixed(rl_src_obj, rDI);
+    LoadValueDirectFixed(rl_src_offset, rSI);
+    LoadValueDirectWideFixed(rl_src_expected, rAX, rDX);
+    LoadValueDirectWideFixed(rl_src_new_value, rBX, rCX);
+    NewLIR4(kX86LockCmpxchg8bA, rDI, rSI, 0, 0);
+    FreeTemp(rSI);
+    UnmarkTemp(rSI);
+    NewLIR1(kX86Pop32R, rSI);
+    FreeTemp(rDI);
+    UnmarkTemp(rDI);
+    NewLIR1(kX86Pop32R, rDI);
+    FreeCallTemps();
   } else {
     // EAX must hold expected for CMPXCHG. Neither rl_new_value, nor r_ptr may be in EAX.
     FlushReg(r0);
diff --git a/compiler/dex/quick/x86/target_x86.cc b/compiler/dex/quick/x86/target_x86.cc
index 878fa76..b7a607f 100644
--- a/compiler/dex/quick/x86/target_x86.cc
+++ b/compiler/dex/quick/x86/target_x86.cc
@@ -165,6 +165,10 @@
   if (flags & REG_USED) {
     SetupRegMask(&lir->u.m.use_mask, rDX);
   }
+
+  if (flags & REG_USEB) {
+    SetupRegMask(&lir->u.m.use_mask, rBX);
+  }
 }
 
 /* For dumping instructions */
diff --git a/compiler/dex/quick/x86/x86_dex_file_method_inliner.cc b/compiler/dex/quick/x86/x86_dex_file_method_inliner.cc
index f0e76c9..88e933f 100644
--- a/compiler/dex/quick/x86/x86_dex_file_method_inliner.cc
+++ b/compiler/dex/quick/x86/x86_dex_file_method_inliner.cc
@@ -66,8 +66,8 @@
 
     INTRINSIC(SunMiscUnsafe, CompareAndSwapInt, ObjectJII_Z, kIntrinsicCas,
               kIntrinsicFlagNone),
-    // INTRINSIC(SunMiscUnsafe, CompareAndSwapLong, ObjectJJJ_Z, kIntrinsicCas,
-    //           kIntrinsicFlagIsLong),
+    INTRINSIC(SunMiscUnsafe, CompareAndSwapLong, ObjectJJJ_Z, kIntrinsicCas,
+              kIntrinsicFlagIsLong),
     INTRINSIC(SunMiscUnsafe, CompareAndSwapObject, ObjectJObjectObject_Z, kIntrinsicCas,
               kIntrinsicFlagIsObject),
 
diff --git a/compiler/dex/quick/x86/x86_lir.h b/compiler/dex/quick/x86/x86_lir.h
index bbcae92..5fe76fe 100644
--- a/compiler/dex/quick/x86/x86_lir.h
+++ b/compiler/dex/quick/x86/x86_lir.h
@@ -314,6 +314,7 @@
   UnaryOpcode(kX86Divmod,  DaR, DaM, DaA),
   UnaryOpcode(kX86Idivmod, DaR, DaM, DaA),
   kX86Bswap32R,
+  kX86Push32R, kX86Pop32R,
 #undef UnaryOpcode
 #define Binary0fOpCode(opcode) \
   opcode ## RR, opcode ## RM, opcode ## RA
@@ -355,6 +356,7 @@
   Binary0fOpCode(kX86Imul32),   // 32bit multiply
   kX86CmpxchgRR, kX86CmpxchgMR, kX86CmpxchgAR,  // compare and exchange
   kX86LockCmpxchgMR, kX86LockCmpxchgAR,  // locked compare and exchange
+  kX86LockCmpxchg8bM, kX86LockCmpxchg8bA,  // locked compare and exchange
   Binary0fOpCode(kX86Movzx8),   // zero-extend 8-bit value
   Binary0fOpCode(kX86Movzx16),  // zero-extend 16-bit value
   Binary0fOpCode(kX86Movsx8),   // sign-extend 8-bit value