Merge pull request #1292 from kangaroo/aarch64

Initial aarch64/linux support

59 files changed, 2980 insertions, 207 deletions

diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index 6b967e3107..b811aed911 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -7,8 +7,10 @@ if(CLR_CMAKE_PLATFORM_ARCH_AMD64)
   include_directories("debug/inc/amd64")
 elseif(CLR_CMAKE_PLATFORM_ARCH_ARM)
   include_directories("debug/inc/arm")
+elseif(CLR_CMAKE_PLATFORM_ARCH_ARM64)
+  include_directories("debug/inc/arm64")
 else()
-  message(FATAL_ERROR "Only ARM and AMD64 is supported")
+  message(FATAL_ERROR "Only ARM, ARM64 and AMD64 is supported")
 endif()
   
 include_directories("debug/inc/dump")
diff --git a/src/ToolBox/SOS/Strike/util.h b/src/ToolBox/SOS/Strike/util.h
index 7b9d96d997..904edc074f 100644
--- a/src/ToolBox/SOS/Strike/util.h
+++ b/src/ToolBox/SOS/Strike/util.h
@@ -1621,7 +1621,7 @@ BOOL IsMiniDumpFile();
 void ReportOOM();
 
 BOOL SafeReadMemory (TADDR offset, PVOID lpBuffer, ULONG cb, PULONG lpcbBytesRead);
-#if !defined(_TARGET_WIN64_)
+#if !defined(_TARGET_WIN64_) && !defined(_ARM64_)
 // on 64-bit platforms TADDR and CLRDATA_ADDRESS are identical
 inline BOOL SafeReadMemory (CLRDATA_ADDRESS offset, PVOID lpBuffer, ULONG cb, PULONG lpcbBytesRead)
 { return SafeReadMemory(TO_TADDR(offset), lpBuffer, cb, lpcbBytesRead); }
diff --git a/src/ToolBox/SOS/lldbplugin/CMakeLists.txt b/src/ToolBox/SOS/lldbplugin/CMakeLists.txt
index 798c91a142..ce279e9a1c 100644
--- a/src/ToolBox/SOS/lldbplugin/CMakeLists.txt
+++ b/src/ToolBox/SOS/lldbplugin/CMakeLists.txt
@@ -12,6 +12,9 @@ elseif(CLR_CMAKE_PLATFORM_ARCH_ARM)
     add_definitions(-DDBG_TARGET_32BIT=1)
     add_definitions(-DDBG_TARGET_ARM=1)
     add_definitions(-DDBG_TARGET_WIN32=1)
+elseif(CLR_CMAKE_PLATFORM_ARCH_ARM64)
+    message(WARNING "lldb is not supported on linux/aarch64 yet")
+    return()
 endif()
 
 set(ENABLE_LLDBPLUGIN ${CLR_CMAKE_PLATFORM_UNIX} CACHE BOOL "Enable building the SOS plugin for LLDB.")
diff --git a/src/binder/CMakeLists.txt b/src/binder/CMakeLists.txt
index b6e9da3f46..b624d9fafa 100644
--- a/src/binder/CMakeLists.txt
+++ b/src/binder/CMakeLists.txt
@@ -4,8 +4,10 @@ if(CLR_CMAKE_PLATFORM_ARCH_AMD64)
     include_directories(BEFORE "../vm/amd64")
 elseif(CLR_CMAKE_PLATFORM_ARCH_ARM)
     include_directories(BEFORE "../vm/arm")
+elseif(CLR_CMAKE_PLATFORM_ARCH_ARM64)
+    include_directories(BEFORE "../vm/arm64")
 else()
-    message(FATAL_ERROR "Only ARM and AMD64 is supported")
+    message(FATAL_ERROR "Only ARM, ARM64 and AMD64 is supported")
 endif()
 include_directories(BEFORE "../vm")
 include_directories(BEFORE "inc")
diff --git a/src/classlibnative/CMakeLists.txt b/src/classlibnative/CMakeLists.txt
index 00a6941920..683506bba7 100644
--- a/src/classlibnative/CMakeLists.txt
+++ b/src/classlibnative/CMakeLists.txt
@@ -4,8 +4,10 @@ if(CLR_CMAKE_PLATFORM_ARCH_AMD64)
   include_directories(BEFORE "../vm/amd64")
 elseif(CLR_CMAKE_PLATFORM_ARCH_ARM)
   include_directories(BEFORE "../vm/arm")
+elseif(CLR_CMAKE_PLATFORM_ARCH_ARM64)
+  include_directories(BEFORE "../vm/arm64")
 else()
-  message(FATAL_ERROR "Only ARM and AMD64 is supported")
+  message(FATAL_ERROR "Only ARM, ARM64 and AMD64 is supported")
 endif()
 include_directories("../debug/inc")
 include_directories("../debug/inc/dump")
@@ -20,8 +22,12 @@ elseif(CLR_CMAKE_PLATFORM_ARCH_ARM)
   add_definitions(-DDBG_TARGET_32BIT=1)
   add_definitions(-DDBG_TARGET_ARM=1)
   add_definitions(-DDBG_TARGET_WIN32=1)
+elseif(CLR_CMAKE_PLATFORM_ARCH_ARM64)
+  add_definitions(-DDBG_TARGET_64BIT=1)
+  add_definitions(-DDBG_TARGET_ARM64=1)
+  add_definitions(-DDBG_TARGET_WIN64=1)
 else()
-  message(FATAL_ERROR "Only ARM and AMD64 is supported")
+  message(FATAL_ERROR "Only ARM, ARM64 and AMD64 is supported")
 endif()
 
 add_subdirectory(bcltype)
diff --git a/src/debug/daccess/CMakeLists.txt b/src/debug/daccess/CMakeLists.txt
index 0d93d75966..9a61991859 100644
--- a/src/debug/daccess/CMakeLists.txt
+++ b/src/debug/daccess/CMakeLists.txt
@@ -15,8 +15,14 @@ elseif(CLR_CMAKE_PLATFORM_ARCH_ARM)
   add_definitions(-DDBG_TARGET_ARM=1)
   add_definitions(-DDBG_TARGET_WIN32=1)
   add_definitions(-D_WIN32=1)
+elseif(CLR_CMAKE_PLATFORM_ARCH_ARM64)
+  add_definitions(-D_TARGET_ARM64_=1)
+  add_definitions(-DDBG_TARGET_64BIT=1)
+  add_definitions(-DDBG_TARGET_ARM64=1)
+  add_definitions(-DDBG_TARGET_WIN64=1)
+  add_definitions(-D_WIN64=1)
 else()
-  message(FATAL_ERROR "Only ARM and AMD64 is supported")
+  message(FATAL_ERROR "Only ARM, ARM64 and AMD64 is supported")
 endif()
 
 include_directories(BEFORE ${VM_DIR})
diff --git a/src/debug/daccess/daccess.cpp b/src/debug/daccess/daccess.cpp
index b737ecd912..2c8c8323eb 100644
--- a/src/debug/daccess/daccess.cpp
+++ b/src/debug/daccess/daccess.cpp
@@ -5492,15 +5492,17 @@ ClrDataAccess::Initialize(void)
     // Determine our platform based on the pre-processor macros set when we were built
 
 #ifdef FEATURE_PAL
-     #if defined(DBG_TARGET_X86)
-         CorDebugPlatform hostPlatform = CORDB_PLATFORM_POSIX_X86;
-     #elif defined(DBG_TARGET_AMD64)
-         CorDebugPlatform hostPlatform = CORDB_PLATFORM_POSIX_AMD64;
-     #elif defined(DBG_TARGET_ARM)
-         CorDebugPlatform hostPlatform = CORDB_PLATFORM_POSIX_ARM;
-     #else
-         #error Unknown Processor.
-     #endif
+    #if defined(DBG_TARGET_X86)
+        CorDebugPlatform hostPlatform = CORDB_PLATFORM_POSIX_X86;
+    #elif defined(DBG_TARGET_AMD64)
+        CorDebugPlatform hostPlatform = CORDB_PLATFORM_POSIX_AMD64;
+    #elif defined(DBG_TARGET_ARM)
+        CorDebugPlatform hostPlatform = CORDB_PLATFORM_POSIX_ARM;
+    #elif defined(DBG_TARGET_ARM64)
+        CorDebugPlatform hostPlatform = CORDB_PLATFORM_POSIX_ARM64;
+    #else
+        #error Unknown Processor.
+    #endif
 #else
     #if defined(DBG_TARGET_X86)
         CorDebugPlatform hostPlatform = CORDB_PLATFORM_WINDOWS_X86;
diff --git a/src/debug/di/CMakeLists.txt b/src/debug/di/CMakeLists.txt
index 55b5a44657..1c9125adce 100644
--- a/src/debug/di/CMakeLists.txt
+++ b/src/debug/di/CMakeLists.txt
@@ -8,8 +8,13 @@ elseif(CLR_CMAKE_PLATFORM_ARCH_ARM)
   add_definitions(-DDBG_TARGET_32BIT)
   add_definitions(-DDBG_TARGET_WIN32=1)
   add_definitions(-DDBG_TARGET_ARM=1)
+elseif(CLR_CMAKE_PLATFORM_ARCH_ARM64)
+  add_definitions(-D_TARGET_WIN64_=1)
+  add_definitions(-DDBG_TARGET_64BIT)
+  add_definitions(-DDBG_TARGET_WIN64=1)
+  add_definitions(-DDBG_TARGET_ARM64=1)
 else()
-  message(FATAL_ERROR "Only ARM and AMD64 is supported")
+  message(FATAL_ERROR "Only ARM, ARM64 and AMD64 is supported")
 endif()
 
 add_definitions(-DFEATURE_METADATA_CUSTOM_DATA_SOURCE -DFEATURE_METADATA_DEBUGGEE_DATA_SOURCE -DFEATURE_NO_HOST -DFEATURE_METADATA_LOAD_TRUSTED_IMAGES)
@@ -66,4 +71,4 @@ endif(WIN32)
 
 add_precompiled_header(stdafx.h stdafx.cpp CORDBDI_SOURCES)
 
-add_library(cordbdi STATIC ${CORDBDI_SOURCES})
\ No newline at end of file
+add_library(cordbdi STATIC ${CORDBDI_SOURCES})
diff --git a/src/debug/di/arm64/cordbregisterset.cpp b/src/debug/di/arm64/cordbregisterset.cpp
index ccddb6fd52..8cd61cafa8 100644
--- a/src/debug/di/arm64/cordbregisterset.cpp
+++ b/src/debug/di/arm64/cordbregisterset.cpp
@@ -8,7 +8,6 @@
 
 //
 //*****************************************************************************
-#include "stdafx.h"
 #include "primitives.h"
 
 
diff --git a/src/debug/di/arm64/primitives.cpp b/src/debug/di/arm64/primitives.cpp
index 66bf4ed17d..f09f32b7d1 100644
--- a/src/debug/di/arm64/primitives.cpp
+++ b/src/debug/di/arm64/primitives.cpp
@@ -5,6 +5,4 @@
 
 // 
 
-#include "stdafx.h"
-
 #include "../../shared/arm64/primitives.cpp"
diff --git a/src/debug/di/shimremotedatatarget.cpp b/src/debug/di/shimremotedatatarget.cpp
index cd1304be40..7885f6bddf 100644
--- a/src/debug/di/shimremotedatatarget.cpp
+++ b/src/debug/di/shimremotedatatarget.cpp
@@ -215,8 +215,10 @@ ShimRemoteDataTarget::GetPlatform(
          *pPlatform = CORDB_PLATFORM_POSIX_X86;
      #elif defined(DBG_TARGET_AMD64)
          *pPlatform = CORDB_PLATFORM_POSIX_AMD64;
-    #elif defined(DBG_TARGET_ARM)
+     #elif defined(DBG_TARGET_ARM)
          *pPlatform = CORDB_PLATFORM_POSIX_ARM;
+     #elif defined(DBG_TARGET_ARM64)
+         *pPlatform = CORDB_PLATFORM_POSIX_ARM64;
      #else
          #error Unknown Processor.
      #endif
diff --git a/src/debug/ee/CMakeLists.txt b/src/debug/ee/CMakeLists.txt
index e37e2573f3..66eb054b1d 100644
--- a/src/debug/ee/CMakeLists.txt
+++ b/src/debug/ee/CMakeLists.txt
@@ -12,8 +12,13 @@ elseif(CLR_CMAKE_PLATFORM_ARCH_ARM)
   add_definitions(-DDBG_TARGET_32BIT=1)
   add_definitions(-DDBG_TARGET_ARM=1)
   add_definitions(-DDBG_TARGET_WIN32=1)
+elseif(CLR_CMAKE_PLATFORM_ARCH_ARM64)
+  add_definitions(-D_TARGET_ARM64_=1)
+  add_definitions(-DDBG_TARGET_64BIT=1)
+  add_definitions(-DDBG_TARGET_ARM64=1)
+  add_definitions(-DDBG_TARGET_WIN64=1)
 else()
-  message(FATAL_ERROR "Only ARM and AMD64 is supported")
+  message(FATAL_ERROR "Only ARM, ARM64 and AMD64 is supported")
 endif()
 
 
diff --git a/src/debug/ee/wks/CMakeLists.txt b/src/debug/ee/wks/CMakeLists.txt
index e2fd7b2bfd..e8ad205fce 100644
--- a/src/debug/ee/wks/CMakeLists.txt
+++ b/src/debug/ee/wks/CMakeLists.txt
@@ -31,6 +31,8 @@ if(CLR_CMAKE_PLATFORM_ARCH_AMD64)
   add_library(cordbee_wks ${CORDBEE_SOURCES_WKS} ../amd64/dbghelpers.S)
 elseif(CLR_CMAKE_PLATFORM_ARCH_ARM)
   add_library(cordbee_wks ${CORDBEE_SOURCES_WKS} ../arm/dbghelpers.S)
+elseif(CLR_CMAKE_PLATFORM_ARCH_ARM64)
+  add_library(cordbee_wks ${CORDBEE_SOURCES_WKS})
 else()
   message(FATAL_ERROR "Only ARM and AMD64 is supported")
 endif()
diff --git a/src/inc/clrnt.h b/src/inc/clrnt.h
index 97d249ad88..fe6c5a87d2 100644
--- a/src/inc/clrnt.h
+++ b/src/inc/clrnt.h
@@ -1012,6 +1012,10 @@ RtlVirtualUnwind (
 
 #ifdef _TARGET_ARM64_
 
+#define UNW_FLAG_NHANDLER               0x0             /* any handler */
+#define UNW_FLAG_EHANDLER               0x1             /* filter handler */
+#define UNW_FLAG_UHANDLER               0x2             /* unwind handler */
+
 // This function returns the RVA of the end of the function (exclusive, so one byte after the actual end)
 // using the unwind info on ARM64. (see ExternalAPIs\Win9CoreSystem\inc\winnt.h)
 FORCEINLINE
@@ -1044,6 +1048,18 @@ typedef struct _UNWIND_INFO {
     // dummy
 } UNWIND_INFO, *PUNWIND_INFO;
 
+PEXCEPTION_ROUTINE
+RtlVirtualUnwind(
+    IN ULONG HandlerType,
+    IN ULONG64 ImageBase,
+    IN ULONG64 ControlPc,
+    IN PRUNTIME_FUNCTION FunctionEntry,
+    IN OUT PCONTEXT ContextRecord,
+    OUT PVOID *HandlerData,
+    OUT PULONG64 EstablisherFrame,
+    IN OUT PKNONVOLATILE_CONTEXT_POINTERS ContextPointers OPTIONAL
+    );
+
 #endif
 
 #endif  // CLRNT_H_
diff --git a/src/inc/volatile.h b/src/inc/volatile.h
index 5310a65ff1..8ec42962b0 100644
--- a/src/inc/volatile.h
+++ b/src/inc/volatile.h
@@ -72,12 +72,12 @@
 #error The Volatile type is currently only defined for Visual C++ and GNU C++
 #endif
 
-#if defined(__GNUC__) && !defined(_X86_) && !defined(_AMD64_) && !defined(_ARM_)
-#error The Volatile type is currently only defined for GCC when targeting x86, AMD64 or ARM CPUs
+#if defined(__GNUC__) && !defined(_X86_) && !defined(_AMD64_) && !defined(_ARM_) && !defined(_ARM64_)
+#error The Volatile type is currently only defined for GCC when targeting x86, AMD64, ARM or ARM64 CPUs
 #endif
 
 #if defined(__GNUC__)
-#if defined(_ARM_)
+#if defined(_ARM_) || defined(_ARM64_)
 // This is functionally equivalent to the MemoryBarrier() macro used on ARM on Windows.
 #define VOLATILE_MEMORY_BARRIER() asm volatile ("dmb sy" : : : "memory")
 #else
diff --git a/src/jit/CMakeLists.txt b/src/jit/CMakeLists.txt
index 2612331d32..96f11b760a 100644
--- a/src/jit/CMakeLists.txt
+++ b/src/jit/CMakeLists.txt
@@ -79,8 +79,17 @@ elseif(CLR_CMAKE_PLATFORM_ARCH_ARM)
     codegenarm.cpp
     unwindarm.cpp
   )
+elseif(CLR_CMAKE_PLATFORM_ARCH_ARM64)
+  set( ARCH_SOURCES
+    codegenarm64.cpp
+    emitarm64.cpp
+    lowerarm64.cpp
+    targetarm64.cpp
+    unwindarm.cpp
+    unwindarm64.cpp
+  )
 else()
-  message(FATAL_ERROR "Only ARM and AMD64 is supported")
+  message(FATAL_ERROR "Only ARM, ARM64 and AMD64 is supported")
 endif()
 
 set( SOURCES
diff --git a/src/jit/codegenarm64.cpp b/src/jit/codegenarm64.cpp
index 1166f793c7..93ad2a3bdf 100644
--- a/src/jit/codegenarm64.cpp
+++ b/src/jit/codegenarm64.cpp
@@ -1424,7 +1424,7 @@ void                CodeGen::genCodeForBBlist()
                 {
                     VarSetOps::AddElemD(compiler, removedGCVars, varIndex);
                 }
-#endif DEBUG
+#endif // DEBUG
                 VarSetOps::RemoveElemD(compiler, gcInfo.gcVarPtrSetCur, varIndex);
             }
             else if (compiler->lvaIsGCTracked(varDsc))
@@ -1434,7 +1434,7 @@ void                CodeGen::genCodeForBBlist()
                 {
                     VarSetOps::AddElemD(compiler, addedGCVars, varIndex);
                 }
-#endif DEBUG
+#endif // DEBUG
                 VarSetOps::AddElemD(compiler, gcInfo.gcVarPtrSetCur, varIndex);
             }
         }
@@ -3341,6 +3341,8 @@ CodeGen::genLclHeap(GenTreePtr tree)
     var_types   type          = genActualType(size->gtType);
     emitAttr    easz          = emitTypeSize(type);
     BasicBlock* endLabel      = nullptr;    
+    BasicBlock* loop          = nullptr;
+    unsigned stackAdjustment  = 0;
     
 #ifdef DEBUG
     // Verify ESP
@@ -3417,7 +3419,7 @@ CodeGen::genLclHeap(GenTreePtr tree)
         inst_RV_IV(INS_AND, regCnt, ~(STACK_ALIGN - 1), emitActualTypeSize(type));
     }
 
-    unsigned stackAdjustment = 0;
+    stackAdjustment = 0;
 #if FEATURE_EH_FUNCLETS 
     // If we have PSPsym, then need to re-locate it after localloc.
     if (hasPspSym)
@@ -3507,7 +3509,7 @@ CodeGen::genLclHeap(GenTreePtr tree)
         genSetRegToIcon(regCnt, amount, ((int)amount == amount)? TYP_INT : TYP_LONG);
     }
 
-    BasicBlock* loop = genCreateTempLabel();
+    loop = genCreateTempLabel();
     if (compiler->info.compInitMem)
     {
         // At this point 'regCnt' is set to the total number of bytes to locAlloc.
diff --git a/src/jit/emitarm64.cpp b/src/jit/emitarm64.cpp
index 3ae8c4ce35..06e4e73d07 100644
--- a/src/jit/emitarm64.cpp
+++ b/src/jit/emitarm64.cpp
@@ -2864,6 +2864,9 @@ emitter::code_t emitter::emitInsCode(instruction ins, insFormat fmt)
         case EA_8BYTE:
             result = INS_OPTS_1D;
             break;
+        default:
+            // TODO-Cleanup: add unreached() here
+            break;
         }
     }
     else if (datasize == EA_16BYTE)
@@ -2882,6 +2885,9 @@ emitter::code_t emitter::emitInsCode(instruction ins, insFormat fmt)
         case EA_8BYTE:
             result = INS_OPTS_2D;
             break;
+        default:
+            // TODO-Cleanup: add unreached() here
+            break;
         }
     }
     return result;
@@ -3095,6 +3101,9 @@ emitter::code_t emitter::emitInsCode(instruction ins, insFormat fmt)
             case EA_8BYTE:
                 result = (index < 1);
                 break;
+            default:
+                // TODO-Cleanup: add unreached() here
+                break;
             }
         }
         else if (datasize == EA_16BYTE)
@@ -3113,6 +3122,9 @@ emitter::code_t emitter::emitInsCode(instruction ins, insFormat fmt)
             case EA_8BYTE:
                 result = (index < 2);
                 break;
+            default:
+                // TODO-Cleanup: add unreached() here
+                break;
             }
         }
     }
@@ -3163,6 +3175,9 @@ void                emitter::emitIns_I(instruction ins,
             assert(!"Instruction cannot be encoded: IF_SI_0A");
         }
         break;
+    default:
+        // TODO-Cleanup: add unreached() here
+        break;
     }
     assert(fmt != IF_NONE);
 
@@ -3205,6 +3220,10 @@ void                emitter::emitIns_R(instruction ins,
         fmt = IF_BR_1A;
         break;
 
+    default:
+        // TODO-Cleanup: add unreached() here
+        break;
+
     }
     assert(fmt != IF_NONE);
 
@@ -3463,6 +3482,10 @@ void                emitter::emitIns_R_I(instruction ins,
         }
         break;
 
+    default:
+        // TODO-Cleanup: add unreached() here
+        break;
+
     }  // end switch (ins)
 
     assert(canEncode);
@@ -3551,6 +3574,11 @@ void                emitter::emitIns_R_F(instruction ins,
             }
         }
         break;
+
+    default:
+        // TODO-Cleanup: add unreached() here
+        break;
+
     }  // end switch (ins)
 
     assert(canEncode);
@@ -3988,6 +4016,10 @@ void                emitter::emitIns_R_R(instruction ins,
         fmt = IF_DV_2J;
         break;
 
+    default:
+        // TODO-Cleanup: add unreached() here
+        break;
+
     }  // end switch (ins)
 
     assert(fmt != IF_NONE);
@@ -4087,6 +4119,10 @@ void                emitter::emitIns_R_I_I(instruction ins,
         }
         break;
 
+    default:
+        // TODO-Cleanup: add unreached() here
+        break;
+
     }  // end switch (ins)
 
     assert(fmt != IF_NONE);
@@ -4527,6 +4563,10 @@ void                emitter::emitIns_R_R_I(instruction ins,
         isLdSt = true;
         break;
 
+    default:
+        // TODO-Cleanup: add unreached() here
+        break;
+
     }  // end switch (ins)
 
     if (isLdSt)
@@ -4932,6 +4972,10 @@ void                emitter::emitIns_R_R_R(instruction ins,
         emitIns_R_R_R_I(ins, attr, reg1, reg2, reg3, 0);
         return;
 
+    default:
+        // TODO-Cleanup: add unreached() here
+        break;
+
     }  // end switch (ins)
 
     assert(fmt != IF_NONE);
@@ -5099,6 +5143,10 @@ void                emitter::emitIns_R_R_R_I(instruction ins,
         isLdSt = true;
         break;
 
+    default:
+        // TODO-Cleanup: add unreached() here
+        break;
+
     }  // end switch (ins)
 
     if (isLdSt)
@@ -5280,6 +5328,10 @@ void                emitter::emitIns_R_R_R_Ext(instruction ins,
         }
         break;
 
+    default:
+        // TODO-Cleanup: add unreached() here
+        break;
+
     }  // end switch (ins)
 
     assert(scale != -1);
@@ -5415,6 +5467,10 @@ void                emitter::emitIns_R_R_I_I(instruction ins,
         fmt = IF_DV_2F;
         break;
 
+    default:
+        // TODO-Cleanup: add unreached() here
+        break;
+
     }
     assert(fmt != IF_NONE);
 
@@ -5478,6 +5534,11 @@ void                emitter::emitIns_R_R_R_R(instruction ins,
     case INS_invalid:
         fmt = IF_NONE;
         break;
+
+    default:
+        // TODO-Cleanup: add unreached() here
+        break;
+
     }
     assert(fmt != IF_NONE);
 
@@ -5522,6 +5583,10 @@ void                emitter::emitIns_R_COND(instruction ins,
         fmt = IF_DR_1D;
         break;
 
+    default:
+        // TODO-Cleanup: add unreached() here
+        break;
+
     }  // end switch (ins)
 
     assert(fmt != IF_NONE);
@@ -5566,6 +5631,9 @@ void                emitter::emitIns_R_R_COND(instruction ins,
         cfi.cond  = cond;
         fmt = IF_DR_2D;
         break;
+    default:
+        // TODO-Cleanup: add unreached() here
+        break;
 
     }  // end switch (ins)
 
@@ -5616,6 +5684,10 @@ void                emitter::emitIns_R_R_R_COND(instruction ins,
         fmt = IF_DR_3D;
         break;
 
+    default:
+        // TODO-Cleanup: add unreached() here
+        break;
+
     }  // end switch (ins)
 
     assert(fmt != IF_NONE);
@@ -5664,7 +5736,9 @@ void                emitter::emitIns_R_R_FLAGS_COND (instruction ins,
         cfi.cond  = cond;
         fmt = IF_DR_2I;
         break;
-
+    default:
+        // TODO-Cleanup: add unreached() here
+        break;
     }  // end switch (ins)
 
     assert(fmt != IF_NONE);
@@ -5723,7 +5797,9 @@ void                emitter::emitIns_R_I_FLAGS_COND (instruction ins,
             assert(!"Instruction cannot be encoded: ccmp/ccmn imm5");
         }
         break;
-
+    default:
+        // TODO-Cleanup: add unreached() here
+        break;
     }  // end switch (ins)
 
     assert(fmt != IF_NONE);
@@ -5762,7 +5838,9 @@ void            emitter::emitIns_BARR (instruction  ins,
         fmt = IF_SI_0B;
         imm = (ssize_t) barrier;
         break;
-
+    default:
+        // TODO-Cleanup: add unreached() here
+        break;
     }  // end switch (ins)
 
     assert(fmt != IF_NONE);
@@ -6295,6 +6373,9 @@ void                emitter::emitIns_R_L  (instruction   ins,
     case INS_adrp:
         fmt = IF_DI_1E;
         break;
+    default:
+        // TODO-Cleanup: add unreached() here
+        break;
     }
     assert(fmt == IF_DI_1E);
 
@@ -6406,6 +6487,9 @@ void                emitter::emitIns_J(instruction   ins,
         // TODO-ARM64-CQ: fmt = IF_LARGEJMP;  /* Assume the jump will be long */
         fmt = IF_BI_0B;
         break;
+    default:
+        // TODO-Cleanup: add unreached() here
+        break;
     }
     assert((fmt == IF_BI_0A) ||
            (fmt == IF_BI_0B));
@@ -8564,6 +8648,9 @@ size_t              emitter::emitOutputInstr(insGroup  *ig,
         case EA_8BYTE:
             cmode = 0xE;                   // 1110
             break;
+        default:
+            // TODO-Cleanup: add unreached() here
+            break;
         }
 
         code  = emitInsCode(ins, fmt);
@@ -9479,7 +9566,7 @@ void                emitter::emitDispAddrRI(regNumber reg, insOpts opt, ssize_t
         }
         else
         {
-            printf(operStr[1]);
+            printf("%c", operStr[1]);
         }
         emitDispImm(imm, false);
         printf("]");
diff --git a/src/jit/instr.cpp b/src/jit/instr.cpp
index 8486175f10..d1cc16f18e 100644
--- a/src/jit/instr.cpp
+++ b/src/jit/instr.cpp
@@ -3890,7 +3890,7 @@ void                CodeGen::instGen_MemoryBarrier()
 #elif defined (_TARGET_ARM_)
     getEmitter()->emitIns_I(INS_dmb, EA_4BYTE, 0xf);
 #elif defined (_TARGET_ARM64_)
-    NYI_ARM64("instGen_MemoryBarrier");
+    getEmitter()->emitIns_BARR(INS_dmb, INS_BARRIER_ST);
 #else
 #error "Unknown _TARGET_"
 #endif
diff --git a/src/jit/lsra.cpp b/src/jit/lsra.cpp
index 5909922c07..cb475d1ba7 100644
--- a/src/jit/lsra.cpp
+++ b/src/jit/lsra.cpp
@@ -2421,8 +2421,6 @@ LinearScan::getKillSetForNode(GenTree* tree)
         }
         break;
 #endif // _TARGET_XARCH_
-
-#ifdef _TARGET_AMD64_
     case GT_COPYOBJ:
         killMask = compiler->compHelperCallKillSet(CORINFO_HELP_ASSIGN_BYREF);
         break;
@@ -2435,10 +2433,14 @@ LinearScan::getKillSetForNode(GenTree* tree)
             case GenTreeBlkOp::BlkOpKindHelper:
                 killMask = compiler->compHelperCallKillSet(CORINFO_HELP_MEMCPY);
                 break;
+#ifdef _TARGET_AMD64_
             case GenTreeBlkOp::BlkOpKindRepInstr:
                 // rep movs kills RCX, RDI and RSI
                 killMask = RBM_RCX | RBM_RDI | RBM_RSI;
                 break;
+#else
+            case GenTreeBlkOp::BlkOpKindRepInstr:
+#endif
             case GenTreeBlkOp::BlkOpKindUnroll:
             case GenTreeBlkOp::BlkOpKindInvalid:
                 // for these 'cpBlkNode->gtBlkOpKind' kinds, we leave 'killMask' = RBM_NONE
@@ -2455,6 +2457,7 @@ LinearScan::getKillSetForNode(GenTree* tree)
             case GenTreeBlkOp::BlkOpKindHelper:
                 killMask = compiler->compHelperCallKillSet(CORINFO_HELP_MEMSET);
                 break;
+#ifdef _TARGET_AMD64_
             case GenTreeBlkOp::BlkOpKindRepInstr:
                 // rep stos kills RCX and RDI
                 killMask = RBM_RDI;
@@ -2463,6 +2466,9 @@ LinearScan::getKillSetForNode(GenTree* tree)
                     killMask |= RBM_RCX;
                 }
                 break;
+#else
+            case GenTreeBlkOp::BlkOpKindRepInstr:
+#endif
             case GenTreeBlkOp::BlkOpKindUnroll:
             case GenTreeBlkOp::BlkOpKindInvalid:
                 // for these 'cpBlkNode->gtBlkOpKind' kinds, we leave 'killMask' = RBM_NONE
@@ -2470,11 +2476,6 @@ LinearScan::getKillSetForNode(GenTree* tree)
             }
         }
         break;
-#else // !_TARGET_AMD64_
-    case GT_INITBLK:
-    case GT_COPYBLK:
-    case GT_COPYOBJ:
-#endif // !_TARGET_AMD64_
     case GT_LSH:
     case GT_RSH:
     case GT_RSZ:
diff --git a/src/jit/registerarm64.h b/src/jit/registerarm64.h
index 70e6ead666..62d41fbe5e 100644
--- a/src/jit/registerarm64.h
+++ b/src/jit/registerarm64.h
@@ -12,7 +12,7 @@
 #define REGALIAS(alias, realname)
 #endif
 
-#define RMASK(x) (1LL << (x))
+#define RMASK(x) (1ULL << (x))
 
 /*
 REGDEF(name, rnum,       mask, xname, wname) */
@@ -57,7 +57,7 @@ REGALIAS(R29, FP)
 REGALIAS(R30, LR)
 
 #define VBASE 32
-#define VMASK(x) (1LL << (VBASE+(x)))
+#define VMASK(x) (1ULL << (VBASE+(x)))
 
 /*
 REGDEF(name,  rnum,       mask,  xname,  wname) */
diff --git a/src/pal/inc/pal.h b/src/pal/inc/pal.h
index a942ed4f63..9c310a5644 100644
--- a/src/pal/inc/pal.h
+++ b/src/pal/inc/pal.h
@@ -94,8 +94,10 @@ extern "C" {
 #define _M_IA64 64100
 #elif defined(__x86_64__) && !defined(_M_AMD64)
 #define _M_AMD64 100
-#elif defined(__ARM_ARCH) && !defined(_M_ARM)
+#elif defined(__arm__) && !defined(_M_ARM)
 #define _M_ARM 7
+#elif defined(__aarch64__) && !defined(_M_ARM64)
+#define _M_ARM64 1
 #endif
 
 #if defined(_M_IX86) && !defined(_X86_)
@@ -118,6 +120,8 @@ extern "C" {
 #define _AMD64_
 #elif defined(_M_ARM) && !defined(_ARM_)
 #define _ARM_
+#elif defined(_M_ARM64) && !defined(_ARM64_)
+#define _ARM64_
 #endif
 
 #endif // !_MSC_VER
@@ -3133,6 +3137,184 @@ typedef struct _IMAGE_ARM_RUNTIME_FUNCTION_ENTRY {
     };
 } IMAGE_ARM_RUNTIME_FUNCTION_ENTRY, * PIMAGE_ARM_RUNTIME_FUNCTION_ENTRY;
 
+#elif defined(_ARM64_)
+
+#define CONTEXT_ARM64   0x00400000L
+
+#define CONTEXT_CONTROL (CONTEXT_ARM64 | 0x1L)
+#define CONTEXT_INTEGER (CONTEXT_ARM64 | 0x2L)
+#define CONTEXT_FLOATING_POINT  (CONTEXT_ARM64 | 0x4L)
+#define CONTEXT_DEBUG_REGISTERS (CONTEXT_ARM64 | 0x8L)
+
+#define CONTEXT_FULL (CONTEXT_CONTROL | CONTEXT_INTEGER | CONTEXT_FLOATING_POINT)
+
+#define CONTEXT_ALL (CONTEXT_CONTROL | CONTEXT_INTEGER | CONTEXT_FLOATING_POINT | CONTEXT_DEBUG_REGISTERS)
+
+#define CONTEXT_EXCEPTION_ACTIVE 0x8000000L
+#define CONTEXT_SERVICE_ACTIVE 0x10000000L
+#define CONTEXT_EXCEPTION_REQUEST 0x40000000L
+#define CONTEXT_EXCEPTION_REPORTING 0x80000000L
+
+//
+// This flag is set by the unwinder if it has unwound to a call
+// site, and cleared whenever it unwinds through a trap frame.
+// It is used by language-specific exception handlers to help
+// differentiate exception scopes during dispatching.
+//
+
+#define CONTEXT_UNWOUND_TO_CALL 0x20000000
+
+//
+// Define initial Cpsr/Fpscr value
+//
+
+#define INITIAL_CPSR 0x10
+#define INITIAL_FPSCR 0
+
+// begin_ntoshvp
+
+//
+// Specify the number of breakpoints and watchpoints that the OS
+// will track. Architecturally, ARM64 supports up to 16. In practice,
+// however, almost no one implements more than 4 of each.
+//
+
+#define ARM64_MAX_BREAKPOINTS     8
+#define ARM64_MAX_WATCHPOINTS     2
+
+//
+// Context Frame
+//
+//  This frame has a several purposes: 1) it is used as an argument to
+//  NtContinue, 2) it is used to constuct a call frame for APC delivery,
+//  and 3) it is used in the user level thread creation routines.
+//
+//
+// The flags field within this record controls the contents of a CONTEXT
+// record.
+//
+// If the context record is used as an input parameter, then for each
+// portion of the context record controlled by a flag whose value is
+// set, it is assumed that that portion of the context record contains
+// valid context. If the context record is being used to modify a threads
+// context, then only that portion of the threads context is modified.
+//
+// If the context record is used as an output parameter to capture the
+// context of a thread, then only those portions of the thread's context
+// corresponding to set flags will be returned.
+//
+// CONTEXT_CONTROL specifies Sp, Lr, Pc, and Cpsr
+//
+// CONTEXT_INTEGER specifies R0-R12
+//
+// CONTEXT_FLOATING_POINT specifies Q0-Q15 / D0-D31 / S0-S31
+//
+// CONTEXT_DEBUG_REGISTERS specifies up to 16 of DBGBVR, DBGBCR, DBGWVR,
+//      DBGWCR.
+//
+
+typedef struct _NEON128 {
+    ULONGLONG Low;
+    LONGLONG High;
+} NEON128, *PNEON128;
+
+typedef struct DECLSPEC_ALIGN(16) _CONTEXT {
+
+    //
+    // Control flags.
+    //
+
+    /* +0x000 */ DWORD ContextFlags;
+
+    //
+    // Integer registers
+    //
+
+    /* +0x004 */ DWORD Cpsr;       // NZVF + DAIF + CurrentEL + SPSel
+    /* +0x008 */ DWORD64 X0;
+                 DWORD64 X1;
+                 DWORD64 X2;
+                 DWORD64 X3;
+                 DWORD64 X4;
+                 DWORD64 X5;
+                 DWORD64 X6;
+                 DWORD64 X7;
+                 DWORD64 X8;
+                 DWORD64 X9;
+                 DWORD64 X10;
+                 DWORD64 X11;
+                 DWORD64 X12;
+                 DWORD64 X13;
+                 DWORD64 X14;
+                 DWORD64 X15;
+                 DWORD64 X16;
+                 DWORD64 X17;
+                 DWORD64 X18;
+                 DWORD64 X19;
+                 DWORD64 X20;
+                 DWORD64 X21;
+                 DWORD64 X22;
+                 DWORD64 X23;
+                 DWORD64 X24;
+                 DWORD64 X25;
+                 DWORD64 X26;
+                 DWORD64 X27;
+                 DWORD64 X28;
+    /* +0x0f0 */ DWORD64 Fp;
+    /* +0x0f8 */ DWORD64 Lr;
+    /* +0x100 */ DWORD64 Sp;
+    /* +0x108 */ DWORD64 Pc;
+
+    //
+    // Floating Point/NEON Registers
+    //
+
+    /* +0x110 */ NEON128 V[32];
+    /* +0x310 */ DWORD Fpcr;
+    /* +0x314 */ DWORD Fpsr;
+
+    //
+    // Debug registers
+    //
+
+    /* +0x318 */ DWORD Bcr[ARM64_MAX_BREAKPOINTS];
+    /* +0x338 */ DWORD64 Bvr[ARM64_MAX_BREAKPOINTS];
+    /* +0x378 */ DWORD Wcr[ARM64_MAX_WATCHPOINTS];
+    /* +0x380 */ DWORD64 Wvr[ARM64_MAX_WATCHPOINTS];
+    /* +0x390 */
+
+} CONTEXT, *PCONTEXT, *LPCONTEXT;
+
+//
+// Nonvolatile context pointer record.
+//
+
+typedef struct _KNONVOLATILE_CONTEXT_POINTERS {
+
+    PDWORD64 X19;
+    PDWORD64 X20;
+    PDWORD64 X21;
+    PDWORD64 X22;
+    PDWORD64 X23;
+    PDWORD64 X24;
+    PDWORD64 X25;
+    PDWORD64 X26;
+    PDWORD64 X27;
+    PDWORD64 X28;
+    PDWORD64 Fp;
+    PDWORD64 Lr;
+
+    PDWORD64 D8;
+    PDWORD64 D9;
+    PDWORD64 D10;
+    PDWORD64 D11;
+    PDWORD64 D12;
+    PDWORD64 D13;
+    PDWORD64 D14;
+    PDWORD64 D15;
+
+} KNONVOLATILE_CONTEXT_POINTERS, *PKNONVOLATILE_CONTEXT_POINTERS;
+
 #else
 #error Unknown architecture for defining CONTEXT.
 #endif
@@ -3259,6 +3441,8 @@ PAL_GetLogicalProcessorCacheSizeFromOS();
 #define PAL_CS_NATIVE_DATA_SIZE 96
 #elif defined(__LINUX__) && defined(_ARM_)
 #define PAL_CS_NATIVE_DATA_SIZE 80
+#elif defined(__LINUX__) && defined(_ARM64_)
+#define PAL_CS_NATIVE_DATA_SIZE 116
 #else 
 #warning 
 #error  PAL_CS_NATIVE_DATA_SIZE is not defined for this architecture
diff --git a/src/pal/inc/rt/ntimage.h b/src/pal/inc/rt/ntimage.h
index 77a19a9b9f..d6c273f794 100644
--- a/src/pal/inc/rt/ntimage.h
+++ b/src/pal/inc/rt/ntimage.h
@@ -940,6 +940,25 @@ typedef IMAGE_RELOCATION UNALIGNED *PIMAGE_RELOCATION;
 #define IMAGE_REL_ARM_SECTION           0x000E  // Section table index
 #define IMAGE_REL_ARM_SECREL            0x000F  // Offset within section
 
+//
+// ARM64 relocation types
+//
+#define IMAGE_REL_ARM64_ABSOLUTE        0x0000
+#define IMAGE_REL_ARM64_ADDR32          0x0001
+#define IMAGE_REL_ARM64_ADDR32NB        0x0002
+#define IMAGE_REL_ARM64_BRANCH26        0x0003
+#define IMAGE_REL_ARM64_PAGEBASE_REL21  0x0004
+#define IMAGE_REL_ARM64_REL21           0x0005
+#define IMAGE_REL_ARM64_PAGEOFFSET_12A  0x0006
+#define IMAGE_REL_ARM64_PAGEOFFSET_12L  0x0007
+#define IMAGE_REL_ARM64_SECREL          0x0008
+#define IMAGE_REL_ARM64_SECREL_LOW12A   0x0009
+#define IMAGE_REL_ARM64_SECREL_HIGH12A  0x000A
+#define IMAGE_REL_ARM64_SECREL_LOW12L   0x000B
+#define IMAGE_REL_ARM64_TOKEN           0x000C
+#define IMAGE_REL_ARM64_SECTION         0x000D
+#define IMAGE_REL_ARM64_ADDR64          0x000E
+
 #define IMAGE_REL_AM_ABSOLUTE           0x0000
 #define IMAGE_REL_AM_ADDR32             0x0001
 #define IMAGE_REL_AM_ADDR32NB           0x0002
diff --git a/src/pal/inc/rt/palrt.h b/src/pal/inc/rt/palrt.h
index 994aa6c0fa..cee51734fa 100644
--- a/src/pal/inc/rt/palrt.h
+++ b/src/pal/inc/rt/palrt.h
@@ -1387,6 +1387,7 @@ typedef VOID (__stdcall *WAITORTIMERCALLBACK)(PVOID, BOOLEAN);
 #define PROCESSOR_ARCHITECTURE_AMD64            9
 #define PROCESSOR_ARCHITECTURE_IA32_ON_WIN64    10
 #define PROCESSOR_ARCHITECTURE_NEUTRAL          11
+#define PROCESSOR_ARCHITECTURE_ARM64            12
 
 #define PROCESSOR_ARCHITECTURE_UNKNOWN 0xFFFF
 
@@ -1636,6 +1637,24 @@ typedef struct _DISPATCHER_CONTEXT {
     DWORD Reserved;
 } DISPATCHER_CONTEXT, *PDISPATCHER_CONTEXT;
 
+#elif defined(_ARM64_)
+
+typedef struct _DISPATCHER_CONTEXT {
+    ULONG64 ControlPc;
+    ULONG64 ImageBase;
+    PRUNTIME_FUNCTION FunctionEntry;
+    ULONG64 EstablisherFrame;
+    ULONG64 TargetPc;
+    PCONTEXT ContextRecord;
+    PEXCEPTION_ROUTINE LanguageHandler;
+    PVOID HandlerData;
+    PUNWIND_HISTORY_TABLE HistoryTable;
+    ULONG64 ScopeIndex;
+    BOOLEAN ControlPcIsUnwound;
+    PBYTE  NonVolatileRegisters;
+    ULONG64 Reserved;
+} DISPATCHER_CONTEXT, *PDISPATCHER_CONTEXT;
+
 #else
 
 typedef struct _DISPATCHER_CONTEXT {
diff --git a/src/pal/inc/unixasmmacros.inc b/src/pal/inc/unixasmmacros.inc
index f996f9505e..4ccb17dd0a 100644
--- a/src/pal/inc/unixasmmacros.inc
+++ b/src/pal/inc/unixasmmacros.inc
@@ -60,4 +60,6 @@
 #include "unixasmmacrosamd64.inc"
 #elif defined(_ARM_)
 #include "unixasmmacrosarm.inc"
+#elif defined(_ARM64_)
+#include "unixasmmacrosarm64.inc"
 #endif
diff --git a/src/pal/inc/unixasmmacrosarm.inc b/src/pal/inc/unixasmmacrosarm.inc
index 40daa7a71c..6418bdcb5d 100644
--- a/src/pal/inc/unixasmmacrosarm.inc
+++ b/src/pal/inc/unixasmmacrosarm.inc
@@ -222,4 +222,4 @@ C_FUNC(\Name\()_End):
         EMIT_BREAKPOINT
 0:
 #endif
-.endm
\ No newline at end of file
+.endm
diff --git a/src/pal/inc/unixasmmacrosarm64.inc b/src/pal/inc/unixasmmacrosarm64.inc
new file mode 100644
index 0000000000..dbed7b6123
--- /dev/null
+++ b/src/pal/inc/unixasmmacrosarm64.inc
@@ -0,0 +1,304 @@
+//
+// Copyright (c) Microsoft. All rights reserved.
+// Licensed under the MIT license. See LICENSE file in the project root for full license information.
+//
+
+.macro PATCH_LABEL Name
+        .global C_FUNC(\Name)
+C_FUNC(\Name):
+.endm
+
+.macro LEAF_ENTRY Name, Section
+        .global C_FUNC(\Name)
+        .type \Name, %function
+C_FUNC(\Name):
+        .cfi_startproc
+.endm
+
+.macro LEAF_END_MARKED Name, Section
+        .global C_FUNC(\Name\()_End)
+C_FUNC(\Name\()_End):
+        .size \Name, .-\Name
+        .cfi_endproc
+.endm
+
+.macro PREPARE_EXTERNAL_VAR Name, HelperReg
+        ldr \HelperReg, [pc, #C_FUNC(\Name)@GOTPCREL]
+.endm
+
+.macro PROLOG_STACK_ALLOC Size
+        sub sp, sp, \Size
+        .cfi_adjust_cfa_offset \Size
+.endm
+
+.macro EPILOG_STACK_FREE Size
+        add sp, sp, \Size
+.endm
+
+.macro EPILOG_STACK_RESTORE
+        mov sp, fp
+.endm
+
+.macro PROLOG_SAVE_REG reg, ofs
+        str \reg, [sp, \ofs]
+.endm
+
+.macro PROLOG_SAVE_REG_PAIR reg1, reg2, ofs
+        stp \reg1, \reg2, [sp, \ofs]
+        .ifc \reg1, fp
+        mov fp, sp
+        .endif
+.endm
+
+.macro PROLOG_SAVE_REG_PAIR_INDEXED reg1, reg2, ofs
+        stp \reg1, \reg2, [sp, \ofs]!
+        .ifc \reg1, fp
+        mov fp, sp
+        .endif
+.endm
+
+.macro EPILOG_RESTORE_REG reg, ofs
+        ldr \reg, [sp, \ofs]
+.endm
+
+.macro EPILOG_RESTORE_REG_PAIR reg1, reg2, ofs
+        ldp \reg1, \reg2, [sp, \ofs]
+.endm
+
+.macro EPILOG_RESTORE_REG_PAIR_INDEXED reg1, reg2, ofs
+        ldp \reg1, \reg2, [sp], \ofs
+.endm
+
+.macro EPILOG_RETURN
+        ret
+.endm
+
+.macro EMIT_BREAKPOINT
+        brk #0
+.endm
+
+//-----------------------------------------------------------------------------
+// Define the prolog for a TransitionFrame-based method. This macro should be called first in the method and
+// comprises the entire prolog (i.e. don't modify SP after calling this).The locals must be 8 byte aligned 
+//
+// Stack layout:
+//
+// (stack parameters)
+// ...
+// fp
+// lr 
+// CalleeSavedRegisters::x28
+// CalleeSavedRegisters::x27
+// CalleeSavedRegisters::x26
+// CalleeSavedRegisters::x25
+// CalleeSavedRegisters::x24
+// CalleeSavedRegisters::x23
+// CalleeSavedRegisters::x22
+// CalleeSavedRegisters::x21
+// CalleeSavedRegisters::x20
+// CalleeSavedRegisters::x19
+// CalleeSavedRegisters::x30 (Lr)
+// CalleeSavedRegisters::x29 (Fp)
+// ArgumentRegisters::x7
+// ArgumentRegisters::x6
+// ArgumentRegisters::x5
+// ArgumentRegisters::x4
+// ArgumentRegisters::x3
+// ArgumentRegisters::x2
+// ArgumentRegisters::x1
+// ArgumentRegisters::x0
+// FloatRegisters::d7
+// FloatRegisters::d6
+// FloatRegisters::d5
+// FloatRegisters::d4
+// FloatRegisters::d3
+// FloatRegisters::d2
+// FloatRegisters::d1
+// FloatRegisters::d0
+.macro PROLOG_WITH_TRANSITION_BLOCK extraLocals, SaveFPArgs
+
+        __PWTB_FloatArgumentRegisters = \extraLocals
+
+        .if ((__PWTB_FloatArgumentRegisters % 16) != 0)
+                __PWTB_FloatArgumentRegisters = __PWTB_FloatArgumentRegisters + 8
+        .endif
+
+        __PWTB_TransitionBlock = __PWTB_FloatArgumentRegisters
+
+        .if \SaveFPArgs > 0
+            __PWTB_TransitionBlock = __PWTB_TransitionBlock + SIZEOF__FloatArgumentRegisters
+        .endif
+
+        __PWTB_StackAlloc = __PWTB_TransitionBlock
+        __PWTB_ArgumentRegisters = __PWTB_StackAlloc + 96 
+
+        PROLOG_SAVE_REG_PAIR_INDEXED   fp, lr, #-160
+        // Spill callee saved registers 
+        PROLOG_SAVE_REG_PAIR   x19, x20, #16
+        PROLOG_SAVE_REG_PAIR   x21, x22, #32
+        PROLOG_SAVE_REG_PAIR   x23, x24, #48
+        PROLOG_SAVE_REG_PAIR   x25, x26, #64
+        PROLOG_SAVE_REG_PAIR   x27, x28, #80
+       
+        // Allocate space for the rest of the frame
+        PROLOG_STACK_ALLOC  __PWTB_StackAlloc
+        
+        // Spill argument registers.
+        SAVE_ARGUMENT_REGISTERS        sp, __PWTB_ArgumentRegisters
+
+        .if \SaveFPArgs > 0
+                SAVE_FLOAT_ARGUMENT_REGISTERS sp, \extraLocals
+        .endif
+
+.endm
+
+//-----------------------------------------------------------------------------
+// The Following sets of SAVE_*_REGISTERS expect the memory to be reserved and 
+// base address to be passed in $reg
+//
+
+// Reserve 64 bytes of memory before calling  SAVE_ARGUMENT_REGISTERS
+.macro SAVE_ARGUMENT_REGISTERS reg, ofs 
+
+        stp                    x0, x1, [\reg, #(\ofs)]
+        stp                    x2, x3, [\reg, #(\ofs + 16)]
+        stp                    x4, x5, [\reg, #(\ofs + 32)]
+        stp                    x6, x7, [\reg, #(\ofs + 48)]
+
+.endm
+
+// Reserve 64 bytes of memory before calling  SAVE_FLOAT_ARGUMENT_REGISTERS
+.macro SAVE_FLOAT_ARGUMENT_REGISTERS reg, ofs 
+
+        stp                    d0, d1, [\reg, #(\ofs)]
+        stp                    d2, d3, [\reg, #(\ofs + 16)]
+        stp                    d4, d5, [\reg, #(\ofs + 32)]
+        stp                    d6, d7, [\reg, #(\ofs + 48)]
+
+.endm
+
+.macro RESTORE_ARGUMENT_REGISTERS reg, ofs 
+
+        ldp                    x0, x1, [\reg, #(\ofs)]
+        ldp                    x2, x3, [\reg, #(\ofs + 16)]
+        ldp                    x4, x5, [\reg, #(\ofs + 32)]
+        ldp                    x6, x7, [\reg, #(\ofs + 48)]
+
+.endm
+
+.macro RESTORE_FLOAT_ARGUMENT_REGISTERS reg, ofs 
+
+        ldp                    d0, d1, [\reg, #(\ofs)]
+        ldp                    d2, d3, [\reg, #(\ofs + 16)]
+        ldp                    d4, d5, [\reg, #(\ofs + 32)]
+        ldp                    d6, d7, [\reg, #(\ofs + 48)]
+
+.endm
+
+.macro EPILOG_BRANCH_REG reg
+
+        br \reg
+
+.endm
+
+//-----------------------------------------------------------------------------
+// Provides a matching epilog to PROLOG_WITH_TRANSITION_BLOCK and ends by preparing for tail-calling.
+// Since this is a tail call argument registers are restored.
+//
+.macro EPILOG_WITH_TRANSITION_BLOCK_TAILCALL extraLocals, SaveFPArgs
+
+        __PWTB_FloatArgumentRegisters = \extraLocals
+
+        .if ((__PWTB_FloatArgumentRegisters % 16) != 0)
+                __PWTB_FloatArgumentRegisters = __PWTB_FloatArgumentRegisters + 8
+        .endif
+
+        __PWTB_TransitionBlock = __PWTB_FloatArgumentRegisters
+
+        .if \SaveFPArgs > 0
+            __PWTB_TransitionBlock = __PWTB_TransitionBlock + SIZEOF__FloatArgumentRegisters
+        .endif
+
+        __PWTB_StackAlloc = __PWTB_TransitionBlock
+        __PWTB_ArgumentRegisters = __PWTB_StackAlloc + 96 
+
+        .if \SaveFPArgs > 0
+                RESTORE_FLOAT_ARGUMENT_REGISTERS  sp, __PWTB_FloatArgumentRegisters
+        .endif
+
+        RESTORE_ARGUMENT_REGISTERS        sp, __PWTB_ArgumentRegisters
+        EPILOG_STACK_FREE                 __PWTB_StackAlloc
+        
+        EPILOG_RESTORE_REG_PAIR   x19, x20, #16
+        EPILOG_RESTORE_REG_PAIR   x21, x22, #32
+        EPILOG_RESTORE_REG_PAIR   x23, x24, #48
+        EPILOG_RESTORE_REG_PAIR   x25, x26, #64
+        EPILOG_RESTORE_REG_PAIR   x27, x28, #80
+        EPILOG_RESTORE_REG_PAIR_INDEXED fp, lr, #160
+
+.endm
+
+// ------------------------------------------------------------------
+// Macro to generate Redirection Stubs
+//
+// $reason : reason for redirection
+//                     Eg. GCThreadControl
+// NOTE: If you edit this macro, make sure you update GetCONTEXTFromRedirectedStubStackFrame.
+// This function is used by both the personality routine and the debugger to retrieve the original CONTEXT.
+.macro GenerateRedirectedHandledJITCaseStub reason
+
+#if NOTYET
+        GBLS __RedirectionStubFuncName
+        GBLS __RedirectionStubEndFuncName
+        GBLS __RedirectionFuncName
+__RedirectionStubFuncName SETS "RedirectedHandledJITCaseFor":CC:"$reason":CC:"_Stub"
+__RedirectionStubEndFuncName SETS "RedirectedHandledJITCaseFor":CC:"$reason":CC:"_StubEnd"
+__RedirectionFuncName SETS "|?RedirectedHandledJITCaseFor":CC:"$reason":CC:"@Thread@@CAXXZ|"
+
+        IMPORT $__RedirectionFuncName
+
+        NESTED_ENTRY $__RedirectionStubFuncName
+        PROLOG_SAVE_REG_PAIR    fp, lr, #-16
+        sub sp, sp, #16                          // stack slot for CONTEXT * and padding 
+
+        //REDIRECTSTUB_SP_OFFSET_CONTEXT is defined in asmconstants.h and is used in GetCONTEXTFromRedirectedStubStackFrame
+        //If CONTEXT is not saved at 0 offset from SP it must be changed as well.
+        ASSERT REDIRECTSTUB_SP_OFFSET_CONTEXT == 0
+
+        // Stack alignment. This check is necessary as this function can be
+        // entered before complete execution of the prolog of another function.
+        and x8, fp, #15
+        sub sp, sp, x8
+
+
+        //
+        // Save a copy of the redirect CONTEXT*.
+        // This is needed for the debugger to unwind the stack.
+        //
+        bl GetCurrentSavedRedirectContext
+        str x0, [sp]
+
+        //
+        // Fetch the interrupted pc and save it as our return address.
+        //
+        ldr x1, [x0, #CONTEXT_Pc]
+        str x1, [fp, #8]
+
+        //
+        // Call target, which will do whatever we needed to do in the context
+        // of the target thread, and will RtlRestoreContext when it is done.
+        //
+        bl $__RedirectionFuncName
+
+        EMIT_BREAKPOINT // Unreachable
+
+// Put a label here to tell the debugger where the end of this function is.
+$__RedirectionStubEndFuncName
+        EXPORT $__RedirectionStubEndFuncName
+
+        NESTED_END
+#else
+        EMIT_BREAKPOINT
+#endif
+
+.endm
diff --git a/src/pal/src/CMakeLists.txt b/src/pal/src/CMakeLists.txt
index 07aa78cce4..0709764a3e 100644
--- a/src/pal/src/CMakeLists.txt
+++ b/src/pal/src/CMakeLists.txt
@@ -30,6 +30,9 @@ if(CMAKE_SYSTEM_PROCESSOR STREQUAL x86_64 OR CMAKE_SYSTEM_PROCESSOR STREQUAL amd
 elseif(CMAKE_SYSTEM_PROCESSOR STREQUAL armv7l)
   set(PAL_CMAKE_PLATFORM_ARCH_ARM 1)
   add_definitions(-D_ARM_)
+elseif(CMAKE_SYSTEM_PROCESSOR STREQUAL aarch64)
+  set(PAL_CMAKE_PLATFORM_ARCH_ARM64 1)
+  add_definitions(-D_ARM64_)
 else()
   message(FATAL_ERROR "Only ARM and AMD64 is supported")
 endif()
@@ -60,6 +63,9 @@ if(PAL_CMAKE_PLATFORM_ARCH_AMD64)
 elseif(PAL_CMAKE_PLATFORM_ARCH_ARM)
   add_definitions(-DBIT32=1)
   add_definitions(-D_WIN32=1)
+elseif(PAL_CMAKE_PLATFORM_ARCH_ARM64)
+  add_definitions(-DBIT64=1)
+  add_definitions(-D_WIN64=1)
 endif()
 
 
@@ -76,6 +82,11 @@ elseif(PAL_CMAKE_PLATFORM_ARCH_ARM)
     arch/arm/context2.S
     arch/arm/processor.cpp
   )
+elseif(PAL_CMAKE_PLATFORM_ARCH_ARM64)
+  set(ARCH_SOURCES
+    arch/arm64/context2.S
+    arch/arm64/processor.cpp
+  )
 endif()
 
 set(SOURCES
diff --git a/src/pal/src/arch/arm64/context2.S b/src/pal/src/arch/arm64/context2.S
new file mode 100644
index 0000000000..3d9cfb94bd
--- /dev/null
+++ b/src/pal/src/arch/arm64/context2.S
@@ -0,0 +1,301 @@
+//
+// Copyright (c) Microsoft. All rights reserved.
+// Licensed under the MIT license. See LICENSE file in the project root for full license information.
+//
+//
+// Implementation of _CONTEXT_CaptureContext for the ARM platform.
+// This function is processor dependent.  It is used by exception handling,
+// and is always apply to the current thread.
+//
+
+#include "unixasmmacros.inc"
+
+#define CONTEXT_ARM64   0x00400000L
+
+#define CONTEXT_CONTROL (CONTEXT_ARM64 | 0x1L)
+#define CONTEXT_INTEGER (CONTEXT_ARM64 | 0x2L)
+#define CONTEXT_FLOATING_POINT  (CONTEXT_ARM64 | 0x4L)
+#define CONTEXT_DEBUG_REGISTERS (CONTEXT_ARM64 | 0x8L)
+
+#define CONTEXT_FULL (CONTEXT_CONTROL | CONTEXT_INTEGER | CONTEXT_FLOATING_POINT)
+
+#define CONTEXT_ContextFlags 0
+#define CONTEXT_Cpsr         CONTEXT_ContextFlags+4
+#define CONTEXT_X0           CONTEXT_Cpsr+4
+#define CONTEXT_X1           CONTEXT_X0+8
+#define CONTEXT_X2           CONTEXT_X1+8
+#define CONTEXT_X3           CONTEXT_X2+8
+#define CONTEXT_X4           CONTEXT_X3+8
+#define CONTEXT_X5           CONTEXT_X4+8
+#define CONTEXT_X6           CONTEXT_X5+8
+#define CONTEXT_X7           CONTEXT_X6+8
+#define CONTEXT_X8           CONTEXT_X7+8
+#define CONTEXT_X9           CONTEXT_X8+8
+#define CONTEXT_X10          CONTEXT_X9+8
+#define CONTEXT_X11          CONTEXT_X10+8
+#define CONTEXT_X12          CONTEXT_X11+8
+#define CONTEXT_X13          CONTEXT_X12+8
+#define CONTEXT_X14          CONTEXT_X13+8
+#define CONTEXT_X15          CONTEXT_X14+8
+#define CONTEXT_X16          CONTEXT_X15+8
+#define CONTEXT_X17          CONTEXT_X16+8
+#define CONTEXT_X18          CONTEXT_X17+8
+#define CONTEXT_X19          CONTEXT_X18+8
+#define CONTEXT_X20          CONTEXT_X19+8
+#define CONTEXT_X21          CONTEXT_X20+8
+#define CONTEXT_X22          CONTEXT_X21+8
+#define CONTEXT_X23          CONTEXT_X22+8
+#define CONTEXT_X24          CONTEXT_X23+8
+#define CONTEXT_X25          CONTEXT_X24+8
+#define CONTEXT_X26          CONTEXT_X25+8
+#define CONTEXT_X27          CONTEXT_X26+8
+#define CONTEXT_X28          CONTEXT_X27+8
+#define CONTEXT_Fp           CONTEXT_X28+8
+#define CONTEXT_Lr           CONTEXT_Fp+8
+#define CONTEXT_Sp           CONTEXT_Lr+8
+#define CONTEXT_Pc           CONTEXT_Sp+8
+#define CONTEXT_NEON_OFFSET  CONTEXT_Pc+8
+#define CONTEXT_V0           0
+#define CONTEXT_V1           CONTEXT_V0+16
+#define CONTEXT_V2           CONTEXT_V1+16
+#define CONTEXT_V3           CONTEXT_V2+16
+#define CONTEXT_V4           CONTEXT_V3+16
+#define CONTEXT_V5           CONTEXT_V4+16
+#define CONTEXT_V6           CONTEXT_V5+16
+#define CONTEXT_V7           CONTEXT_V6+16
+#define CONTEXT_V8           CONTEXT_V7+16
+#define CONTEXT_V9           CONTEXT_V8+16
+#define CONTEXT_V10          CONTEXT_V9+16
+#define CONTEXT_V11          CONTEXT_V10+16
+#define CONTEXT_V12          CONTEXT_V11+16
+#define CONTEXT_V13          CONTEXT_V12+16
+#define CONTEXT_V14          CONTEXT_V13+16
+#define CONTEXT_V15          CONTEXT_V14+16
+#define CONTEXT_V16          CONTEXT_V15+16
+#define CONTEXT_V17          CONTEXT_V16+16
+#define CONTEXT_V18          CONTEXT_V17+16
+#define CONTEXT_V19          CONTEXT_V18+16
+#define CONTEXT_V20          CONTEXT_V19+16
+#define CONTEXT_V21          CONTEXT_V20+16
+#define CONTEXT_V22          CONTEXT_V21+16
+#define CONTEXT_V23          CONTEXT_V22+16
+#define CONTEXT_V24          CONTEXT_V23+16
+#define CONTEXT_V25          CONTEXT_V24+16
+#define CONTEXT_V26          CONTEXT_V25+16
+#define CONTEXT_V27          CONTEXT_V26+16
+#define CONTEXT_V28          CONTEXT_V27+16
+#define CONTEXT_V29          CONTEXT_V28+16
+#define CONTEXT_V30          CONTEXT_V29+16
+#define CONTEXT_V31          CONTEXT_V30+16
+#define CONTEXT_FLOAT_CONTROL_OFFSET  CONTEXT_V31
+#define CONTEXT_Fpcr         0
+#define CONTEXT_Fpsr         CONTEXT_Fpcr+4
+
+// Incoming:
+//  x0: Context*
+//
+LEAF_ENTRY CONTEXT_CaptureContext, _TEXT
+    sub sp, sp, #32
+    // save x1, x2 and x3 on stack so we can use them as scratch
+    stp x1, x2, [sp]
+    str x3, [sp, 16]
+    // save the current flags on the stack
+    mrs x1, nzcv
+    str x1, [sp, 24]
+
+    ldr w1, [x0, CONTEXT_ContextFlags]
+    // clangs assembler doesn't seem to support the mov Wx, imm32 yet
+    movz w2, #0x40, lsl #16
+    movk w2, #0x1
+    mov w3, w2
+    and w2, w1, w2
+    cmp w2, w3
+    b.ne LOCAL_LABEL(Done_CONTEXT_CONTROL)
+
+    // save the cpsr
+    ldr x2, [sp, 24]
+    str w2, [x0, CONTEXT_Cpsr]
+    stp fp, lr, [x0, CONTEXT_Fp]
+    add sp, sp, #32
+    mov x2, sp
+    stp x2, lr, [x0, CONTEXT_Sp]
+    sub sp, sp, #32
+
+LOCAL_LABEL(Done_CONTEXT_CONTROL): 
+    // we dont clobber x1 in the CONTEXT_CONTROL case
+    // ldr w1, [x0, CONTEXT_ContextFlags]
+    // clangs assembler doesn't seem to support the mov Wx, imm32 yet
+    movz w2, #0x40, lsl #16
+    movk w2, #0x2
+    mov w3, w2
+    and w2, w1, w2
+    cmp w2, w3
+    b.ne LOCAL_LABEL(Done_CONTEXT_INTEGER)
+
+    ldp x1, x2,   [sp]
+    ldr x3,       [sp, 16]
+
+    stp x0, x1,   [x0, CONTEXT_X0]
+    stp x2, x3,   [x0, CONTEXT_X2]
+    stp x4, x5,   [x0, CONTEXT_X4]
+    stp x6, x7,   [x0, CONTEXT_X6]
+    stp x8, x9,   [x0, CONTEXT_X8]
+    stp x10, x11, [x0, CONTEXT_X10]
+    stp x12, x13, [x0, CONTEXT_X12]
+    stp x14, x15, [x0, CONTEXT_X14]
+    stp x16, x17, [x0, CONTEXT_X16]
+    stp x18, x19, [x0, CONTEXT_X18]
+    stp x20, x21, [x0, CONTEXT_X20]
+    stp x22, x23, [x0, CONTEXT_X22]
+    stp x24, x25, [x0, CONTEXT_X24]
+    stp x26, x27, [x0, CONTEXT_X26]
+    str x28,      [x0, CONTEXT_X28]
+
+
+LOCAL_LABEL(Done_CONTEXT_INTEGER): 
+    ldr w1, [x0, CONTEXT_ContextFlags]
+    // clangs assembler doesn't seem to support the mov Wx, imm32 yet
+    movz w2, #0x40, lsl #16
+    movk w2, #0x4
+    mov w3, w2
+    and w2, w1, w2
+    cmp w2, w3
+    b.ne LOCAL_LABEL(Done_CONTEXT_FLOATING_POINT)
+
+    add x0, x0,   CONTEXT_NEON_OFFSET
+    stp q0, q1,   [x0, CONTEXT_V0]
+    stp q2, q3,   [x0, CONTEXT_V2]
+    stp q4, q5,   [x0, CONTEXT_V4]
+    stp q6, q7,   [x0, CONTEXT_V6]
+    stp q8, q9,   [x0, CONTEXT_V8]
+    stp q10, q11, [x0, CONTEXT_V10]
+    stp q12, q13, [x0, CONTEXT_V12]
+    stp q14, q15, [x0, CONTEXT_V14]
+    stp q16, q17, [x0, CONTEXT_V16]
+    stp q18, q19, [x0, CONTEXT_V18]
+    stp q20, q21, [x0, CONTEXT_V20]
+    stp q22, q23, [x0, CONTEXT_V22]
+    stp q24, q25, [x0, CONTEXT_V24]
+    stp q26, q27, [x0, CONTEXT_V26]
+    stp q28, q29, [x0, CONTEXT_V28]
+    stp q30, q31, [x0, CONTEXT_V30]
+    add x0, x0,   CONTEXT_FLOAT_CONTROL_OFFSET
+    mrs x1, fpcr
+    mrs x2, fpsr
+    sub x0, x0,   CONTEXT_FLOAT_CONTROL_OFFSET
+    stp x1, x2,   [x0, CONTEXT_Fpcr]
+    sub x0, x0,   CONTEXT_NEON_OFFSET
+
+LOCAL_LABEL(Done_CONTEXT_FLOATING_POINT):
+    
+    add sp, sp, #32
+    ret
+LEAF_END CONTEXT_CaptureContext, _TEXT
+
+// Incoming:
+//  x0: Context*
+
+LEAF_ENTRY RtlCaptureContext, _TEXT
+    sub sp, sp, #16
+    str x1, [sp]
+    // same as above, clang doesn't like mov with #imm32
+    // keep this in sync if CONTEXT_FULL changes
+    movz w1, #0x40, lsl #16
+    orr w1, w1, #0x1
+    orr w1, w1, #0x2
+    orr w1, w1, #0x4
+    orr w1, w1, #0x8
+    str w1, [x0, CONTEXT_ContextFlags]
+    ldr x1, [sp]
+    add sp, sp, #16
+    b C_FUNC(CONTEXT_CaptureContext)
+LEAF_END RtlCaptureContext, _TEXT
+
+// Incoming:
+//  x0: Context*
+//  x1: Exception*
+//
+LEAF_ENTRY RtlRestoreContext, _TEXT
+    // aarch64 specifies:
+    //   IP0 and IP1, the Intra-Procedure Call temporary registers,
+    //   are available for use by e.g. veneers or branch islands during a procedure call.
+    //   They are otherwise corruptible.
+    // Since we cannot control $pc directly, we're going to corrupt x16 and x17
+    // so that we can restore control
+    // since we potentially clobber x0 below, we'll bank it in x16
+    mov x16, x0
+
+    ldr w2, [x16, CONTEXT_ContextFlags]
+    // clangs assembler doesn't seem to support the mov Wx, imm32 yet
+    movz w3, #0x40, lsl #16
+    movk w3, #0x4
+    mov w4, w3
+    and w3, w2, w3
+    cmp w3, w4
+    b.ne LOCAL_LABEL(No_Restore_CONTEXT_FLOATING_POINT)
+
+    add x16, x16,   CONTEXT_NEON_OFFSET
+    ldp q0, q1,   [x16, CONTEXT_V0]
+    ldp q2, q3,   [x16, CONTEXT_V2]
+    ldp q4, q5,   [x16, CONTEXT_V4]
+    ldp q6, q7,   [x16, CONTEXT_V6]
+    ldp q8, q9,   [x16, CONTEXT_V8]
+    ldp q10, q11, [x16, CONTEXT_V10]
+    ldp q12, q13, [x16, CONTEXT_V12]
+    ldp q14, q15, [x16, CONTEXT_V14]
+    ldp q16, q17, [x16, CONTEXT_V16]
+    ldp q18, q19, [x16, CONTEXT_V18]
+    ldp q20, q21, [x16, CONTEXT_V20]
+    ldp q22, q23, [x16, CONTEXT_V22]
+    ldp q24, q25, [x16, CONTEXT_V24]
+    ldp q26, q27, [x16, CONTEXT_V26]
+    ldp q28, q29, [x16, CONTEXT_V28]
+    ldp q30, q31, [x16, CONTEXT_V30]
+    ldp x1, x2,   [x16, CONTEXT_Fpcr]
+    msr fpcr, x1
+    msr fpsr, x2
+    sub x16, x16,   CONTEXT_NEON_OFFSET
+
+LOCAL_LABEL(No_Restore_CONTEXT_FLOATING_POINT):
+    movz w2, #0x40, lsl #16
+    movk w2, #0x2
+    mov w3, w2
+    and w2, w1, w2
+    cmp w2, w3
+    b.ne LOCAL_LABEL(No_Restore_CONTEXT_INTEGER)
+
+    ldp x0, x1,   [x16, CONTEXT_X0]
+    ldp x2, x3,   [x16, CONTEXT_X2]
+    ldp x4, x5,   [x16, CONTEXT_X4]
+    ldp x6, x7,   [x16, CONTEXT_X6]
+    ldp x8, x9,   [x16, CONTEXT_X8]
+    ldp x10, x11, [x16, CONTEXT_X10]
+    ldp x12, x13, [x16, CONTEXT_X12]
+    ldp x14, x15, [x16, CONTEXT_X14]
+    ldp x18, x19, [x16, CONTEXT_X18]
+    ldp x20, x21, [x16, CONTEXT_X20]
+    ldp x22, x23, [x16, CONTEXT_X22]
+    ldp x24, x25, [x16, CONTEXT_X24]
+    ldp x26, x27, [x16, CONTEXT_X26]
+    ldr x28,      [x16, CONTEXT_X28]
+
+LOCAL_LABEL(No_Restore_CONTEXT_INTEGER):
+    movz w2, #0x40, lsl #16
+    movk w2, #0x2
+    mov w3, w2
+    and w2, w1, w2
+    cmp w2, w3
+    b.ne LOCAL_LABEL(No_Restore_CONTEXT_CONTROL)
+
+    ldr w17, [x16, CONTEXT_Cpsr]
+    msr nzcv, x17
+    ldp fp, lr, [x16, CONTEXT_Fp]
+    ldr x17, [x16, CONTEXT_Sp]
+    mov sp, x17
+    ldr x17, [x16, CONTEXT_Pc]
+    br x17
+
+LOCAL_LABEL(No_Restore_CONTEXT_CONTROL):
+    ret
+
+LEAF_END RtlRestoreContext, _TEXT
diff --git a/src/pal/src/arch/arm64/processor.cpp b/src/pal/src/arch/arm64/processor.cpp
new file mode 100644
index 0000000000..3f37374f91
--- /dev/null
+++ b/src/pal/src/arch/arm64/processor.cpp
@@ -0,0 +1,42 @@
+//
+// Copyright (c) Microsoft. All rights reserved.
+// Licensed under the MIT license. See LICENSE file in the project root for full license information. 
+//
+
+/*++
+
+
+
+Module Name:
+
+    processor.cpp
+
+Abstract:
+
+    Implementation of processor related functions for the ARM64
+    platform. These functions are processor dependent.
+
+
+
+--*/
+
+#include "pal/palinternal.h"
+
+/*++
+Function:
+YieldProcessor
+
+The YieldProcessor function signals to the processor to give resources
+to threads that are waiting for them. This macro is only effective on
+processors that support technology allowing multiple threads running
+on a single processor, such as Intel's Hyper-Threading technology.
+
+--*/
+void
+PALAPI
+YieldProcessor(
+    VOID)
+{
+    return;
+}
+
diff --git a/src/pal/src/cruntime/printf.cpp b/src/pal/src/cruntime/printf.cpp
index 273edbe527..a691cb4408 100644
--- a/src/pal/src/cruntime/printf.cpp
+++ b/src/pal/src/cruntime/printf.cpp
@@ -130,6 +130,8 @@ Function:
 Parameters:
   ap
     - pointer to the va_list from which to remove arguments
+  Width
+    - the width of the current format operation
   Precision
     - the precision of the current format option
   Type
@@ -137,7 +139,7 @@ Parameters:
   Prefix
     - the prefix for the current format option
 *******************************************************************************/
-void PAL_printf_arg_remover(va_list *ap, INT Precision, INT Type, INT Prefix)
+void PAL_printf_arg_remover(va_list *ap, INT Width, INT Precision, INT Type, INT Prefix)
 {
     /* remove arg and precision if needed */
     if (PRECISION_STAR == Precision ||
@@ -145,6 +147,11 @@ void PAL_printf_arg_remover(va_list *ap, INT Precision, INT Type, INT Prefix)
     {
         (void)va_arg(*ap, int);
     }
+    if (WIDTH_STAR == Width ||
+        WIDTH_INVALID == Width)
+    {
+        (void)va_arg(*ap, int);
+    }
     if (Type == PFF_TYPE_FLOAT)
     {
         (void)va_arg(*ap, double);
diff --git a/src/pal/src/cruntime/printfcpp.cpp b/src/pal/src/cruntime/printfcpp.cpp
index 6e49b591c5..a3e080cf53 100644
--- a/src/pal/src/cruntime/printfcpp.cpp
+++ b/src/pal/src/cruntime/printfcpp.cpp
@@ -1435,7 +1435,7 @@ int CoreVfwprintf(CPalThread *pthrCurrent, PAL_FILE *stream, const wchar_16 *for
                     va_copy(apcopy, ap);
                     TempInt = NativeVsnprintf(pthrCurrent, TempSprintfStr, TEMP_COUNT, TempBuff, apcopy);
                     va_end(apcopy);
-                    PAL_printf_arg_remover(&ap, Precision, Type, Prefix);
+                    PAL_printf_arg_remover(&ap, Width, Precision, Type, Prefix);
 
                     if (TempInt < 0 || static_cast<size_t>(TempInt) >= TEMP_COUNT)
                     {
@@ -1453,7 +1453,7 @@ int CoreVfwprintf(CPalThread *pthrCurrent, PAL_FILE *stream, const wchar_16 *for
                         va_copy(apcopy, ap);
                         NativeVsnprintf(pthrCurrent, TempSprintfStr, TempInt, TempBuff, apcopy);
                         va_end(apcopy);
-                        PAL_printf_arg_remover(&ap, Precision, Type, Prefix);
+                        PAL_printf_arg_remover(&ap, Width, Precision, Type, Prefix);
                     }
                 }
 
@@ -1805,7 +1805,7 @@ int CoreVsnprintf(CPalThread *pthrCurrent, LPSTR Buffer, size_t Count, LPCSTR Fo
                     va_copy(apcopy, ap);
                     TempInt = NativeVsnprintf(pthrCurrent, BufferPtr, TempCount, TempBuff, apcopy);
                     va_end(apcopy);
-                    PAL_printf_arg_remover(&ap, Precision, Type, Prefix);
+                    PAL_printf_arg_remover(&ap, Width, Precision, Type, Prefix);
                 }
 
                 if (TempInt < 0 || static_cast<size_t>(TempInt) >= TempCount) /* buffer not long enough */
@@ -2103,7 +2103,7 @@ int CoreWvsnprintf(CPalThread *pthrCurrent, LPWSTR Buffer, size_t Count, LPCWSTR
                     va_copy(apcopy, ap);
                     TempInt = NativeVsnprintf(pthrCurrent, (LPSTR) BufferPtr, TempCount, TempBuff, apcopy);
                     va_end(apcopy);
-                    PAL_printf_arg_remover(&ap, Precision, Type, Prefix);
+                    PAL_printf_arg_remover(&ap, Width, Precision, Type, Prefix);
                 }
 
                 if (TempInt == 0)
@@ -2484,7 +2484,7 @@ int CoreVfprintf(CPalThread *pthrCurrent, PAL_FILE *stream, const char *format,
                     va_copy(apcopy, ap);
                     TempInt = NativeVfprintf(pthrCurrent, stream->bsdFilePtr, TempBuff, apcopy);
                     va_end(apcopy);
-                    PAL_printf_arg_remover(&ap, Precision, Type, Prefix);
+                    PAL_printf_arg_remover(&ap, Width, Precision, Type, Prefix);
                 }
 
                 if (-1 == TempInt)
diff --git a/src/pal/src/cruntime/silent_printf.cpp b/src/pal/src/cruntime/silent_printf.cpp
index 10f60704d8..d45081042b 100644
--- a/src/pal/src/cruntime/silent_printf.cpp
+++ b/src/pal/src/cruntime/silent_printf.cpp
@@ -219,7 +219,7 @@ INT Silent_PAL_vsnprintf(LPSTR Buffer, INT Count, LPCSTR Format, va_list aparg)
                     va_copy(apcopy, ap);
                     TempInt = PAL__vsnprintf(BufferPtr, TempCount, TempBuff, apcopy);
                     va_end(apcopy);
-                    PAL_printf_arg_remover(&ap, Precision, Type, Prefix);
+                    PAL_printf_arg_remover(&ap, Width, Precision, Type, Prefix);
                 }
 
                 if (TempInt < 0 || static_cast<size_t>(TempInt) >= TempCount) /* buffer not long enough */
@@ -432,7 +432,7 @@ int Silent_PAL_vfprintf(PAL_FILE *stream, const char *format, va_list aparg)
                     va_copy(apcopy, ap);
                     TempInt = PAL_vfprintf(stream, TempBuff, apcopy);
                     va_end(apcopy);
-                    PAL_printf_arg_remover(&ap, Precision, Type, Prefix);
+                    PAL_printf_arg_remover(&ap, Width, Precision, Type, Prefix);
                 }
 
                 if (-1 != TempInt)
diff --git a/src/pal/src/exception/seh-unwind.cpp b/src/pal/src/exception/seh-unwind.cpp
index 9dd025c51d..aee921dd06 100644
--- a/src/pal/src/exception/seh-unwind.cpp
+++ b/src/pal/src/exception/seh-unwind.cpp
@@ -50,6 +50,22 @@ Abstract:
     ASSIGN_REG(R13)        \
     ASSIGN_REG(R14)        \
     ASSIGN_REG(R15)
+#elif defined(_ARM64_)
+#define ASSIGN_UNWIND_REGS \
+    ASSIGN_REG(Pc)         \
+    ASSIGN_REG(Sp)         \
+    ASSIGN_REG(Fp)         \
+    ASSIGN_REG(Lr)         \
+    ASSIGN_REG(X19)        \
+    ASSIGN_REG(X20)        \
+    ASSIGN_REG(X21)        \
+    ASSIGN_REG(X22)        \
+    ASSIGN_REG(X23)        \
+    ASSIGN_REG(X24)        \
+    ASSIGN_REG(X25)        \
+    ASSIGN_REG(X26)        \
+    ASSIGN_REG(X27)        \
+    ASSIGN_REG(X28)
 #else
 #error unsupported architecture
 #endif
@@ -111,6 +127,21 @@ static void UnwindContextToWinContext(unw_cursor_t *cursor, CONTEXT *winContext)
     unw_get_reg(cursor, UNW_ARM_R9, (unw_word_t *) &winContext->R9);
     unw_get_reg(cursor, UNW_ARM_R10, (unw_word_t *) &winContext->R10);
     unw_get_reg(cursor, UNW_ARM_R11, (unw_word_t *) &winContext->R11);
+#elif defined(_ARM64_)
+    unw_get_reg(cursor, UNW_REG_IP, (unw_word_t *) &winContext->Pc);
+    unw_get_reg(cursor, UNW_REG_SP, (unw_word_t *) &winContext->Sp);
+    unw_get_reg(cursor, UNW_AARCH64_X29, (unw_word_t *) &winContext->Fp);
+    unw_get_reg(cursor, UNW_AARCH64_X30, (unw_word_t *) &winContext->Lr);
+    unw_get_reg(cursor, UNW_AARCH64_X19, (unw_word_t *) &winContext->X19);
+    unw_get_reg(cursor, UNW_AARCH64_X20, (unw_word_t *) &winContext->X20);
+    unw_get_reg(cursor, UNW_AARCH64_X21, (unw_word_t *) &winContext->X21);
+    unw_get_reg(cursor, UNW_AARCH64_X22, (unw_word_t *) &winContext->X22);
+    unw_get_reg(cursor, UNW_AARCH64_X23, (unw_word_t *) &winContext->X23);
+    unw_get_reg(cursor, UNW_AARCH64_X24, (unw_word_t *) &winContext->X24);
+    unw_get_reg(cursor, UNW_AARCH64_X25, (unw_word_t *) &winContext->X25);
+    unw_get_reg(cursor, UNW_AARCH64_X26, (unw_word_t *) &winContext->X26);
+    unw_get_reg(cursor, UNW_AARCH64_X27, (unw_word_t *) &winContext->X27);
+    unw_get_reg(cursor, UNW_AARCH64_X28, (unw_word_t *) &winContext->X28);
 #else
 #error unsupported architecture
 #endif
@@ -152,17 +183,50 @@ static void GetContextPointers(unw_cursor_t *cursor, unw_context_t *unwContext,
     GetContextPointer(cursor, unwContext, UNW_ARM_R9, &contextPointers->R9);
     GetContextPointer(cursor, unwContext, UNW_ARM_R10, &contextPointers->R10);
     GetContextPointer(cursor, unwContext, UNW_ARM_R11, &contextPointers->R11);
+#elif defined(_ARM64_)
+    GetContextPointer(cursor, unwContext, UNW_AARCH64_X19, &contextPointers->X19);
+    GetContextPointer(cursor, unwContext, UNW_AARCH64_X20, &contextPointers->X20);
+    GetContextPointer(cursor, unwContext, UNW_AARCH64_X21, &contextPointers->X21);
+    GetContextPointer(cursor, unwContext, UNW_AARCH64_X22, &contextPointers->X22);
+    GetContextPointer(cursor, unwContext, UNW_AARCH64_X23, &contextPointers->X23);
+    GetContextPointer(cursor, unwContext, UNW_AARCH64_X24, &contextPointers->X24);
+    GetContextPointer(cursor, unwContext, UNW_AARCH64_X25, &contextPointers->X25);
+    GetContextPointer(cursor, unwContext, UNW_AARCH64_X26, &contextPointers->X26);
+    GetContextPointer(cursor, unwContext, UNW_AARCH64_X27, &contextPointers->X27);
+    GetContextPointer(cursor, unwContext, UNW_AARCH64_X28, &contextPointers->X28);
 #else
 #error unsupported architecture
 #endif
 }
 
+static DWORD64 GetPc(CONTEXT *context)
+{
+#if defined(_AMD64_)
+    return context->Rip;
+#elif defined(_ARM64_)
+    return context->Pc;
+#else
+#error don't know how to get the program counter for this architecture
+#endif
+}
+
+static void SetPc(CONTEXT *context, DWORD64 pc)
+{
+#if defined(_AMD64_)
+    context->Rip = pc;
+#elif defined(_ARM64_)
+    context->Pc = pc;
+#else
+#error don't know how to set the program counter for this architecture
+#endif
+}
+
 BOOL PAL_VirtualUnwind(CONTEXT *context, KNONVOLATILE_CONTEXT_POINTERS *contextPointers)
 {
     int st;
     unw_context_t unwContext;
     unw_cursor_t cursor;
-#if defined(__APPLE__) || defined(__FreeBSD__)
+#if defined(__APPLE__) || defined(__FreeBSD__) || defined(_ARM64_)
     DWORD64 curPc;
 #endif
 
@@ -187,7 +251,7 @@ BOOL PAL_VirtualUnwind(CONTEXT *context, KNONVOLATILE_CONTEXT_POINTERS *contextP
     WinContextToUnwindCursor(context, &cursor);
 #endif
 
-#if defined(__APPLE__) || defined(__FreeBSD__)
+#if defined(__APPLE__) || defined(__FreeBSD__) || defined(_ARM64_)
     // OSX and FreeBSD appear to do two different things when unwinding
     // 1: If it reaches where it cannot unwind anymore, say a 
     // managed frame.  It wil return 0, but also update the $pc
@@ -196,7 +260,7 @@ BOOL PAL_VirtualUnwind(CONTEXT *context, KNONVOLATILE_CONTEXT_POINTERS *contextP
     // The behaviour of libunwind from nongnu.org is to null the PC
     // So we bank the original PC here, so we can compare it after
     // the step
-    curPc = context->Rip;
+    curPc = GetPc(context);
 #endif
 
     st = unw_step(&cursor);
@@ -208,10 +272,10 @@ BOOL PAL_VirtualUnwind(CONTEXT *context, KNONVOLATILE_CONTEXT_POINTERS *contextP
     // Update the passed in windows context to reflect the unwind
     //
     UnwindContextToWinContext(&cursor, context);
-#if defined(__APPLE__) || defined(__FreeBSD__)
-    if (st == 0 && context->Rip == curPc)
+#if defined(__APPLE__) || defined(__FreeBSD__) || defined(_ARM64_)
+    if (st == 0 && GetPc(context) == curPc)
     {
-        context->Rip = 0;
+        SetPc(context, 0);
     }
 #endif
 
@@ -257,7 +321,7 @@ static void RtlpRaiseException(EXCEPTION_RECORD *ExceptionRecord)
     ExceptionRecord->ExceptionAddress = (void *) ContextRecord.Eip;
 #elif defined(_AMD64_)
     ExceptionRecord->ExceptionAddress = (void *) ContextRecord.Rip;
-#elif defined(_ARM_)
+#elif defined(_ARM_) || defined(_ARM64_)
     ExceptionRecord->ExceptionAddress = (void *) ContextRecord.Pc;
 #else
 #error unsupported architecture
diff --git a/src/pal/src/include/pal/context.h b/src/pal/src/include/pal/context.h
index b1dc76c420..5f0718725b 100644
--- a/src/pal/src/include/pal/context.h
+++ b/src/pal/src/include/pal/context.h
@@ -103,6 +103,44 @@ typedef ucontext_t native_context_t;
 
 #ifdef BIT64
 
+#if defined(_ARM64_)
+#define MCREG_X0(mc)      ((mc).regs[0])
+#define MCREG_X1(mc)      ((mc).regs[1])
+#define MCREG_X2(mc)      ((mc).regs[2])
+#define MCREG_X3(mc)      ((mc).regs[3])
+#define MCREG_X4(mc)      ((mc).regs[4])
+#define MCREG_X5(mc)      ((mc).regs[5])
+#define MCREG_X6(mc)      ((mc).regs[6])
+#define MCREG_X7(mc)      ((mc).regs[7])
+#define MCREG_X8(mc)      ((mc).regs[8])
+#define MCREG_X9(mc)      ((mc).regs[9])
+#define MCREG_X10(mc)     ((mc).regs[10])
+#define MCREG_X11(mc)     ((mc).regs[11])
+#define MCREG_X12(mc)     ((mc).regs[12])
+#define MCREG_X13(mc)     ((mc).regs[13])
+#define MCREG_X14(mc)     ((mc).regs[14])
+#define MCREG_X15(mc)     ((mc).regs[15])
+#define MCREG_X16(mc)     ((mc).regs[16])
+#define MCREG_X17(mc)     ((mc).regs[17])
+#define MCREG_X18(mc)     ((mc).regs[18])
+#define MCREG_X19(mc)     ((mc).regs[19])
+#define MCREG_X20(mc)     ((mc).regs[20])
+#define MCREG_X21(mc)     ((mc).regs[21])
+#define MCREG_X22(mc)     ((mc).regs[22])
+#define MCREG_X23(mc)     ((mc).regs[23])
+#define MCREG_X24(mc)     ((mc).regs[24])
+#define MCREG_X25(mc)     ((mc).regs[25])
+#define MCREG_X26(mc)     ((mc).regs[26])
+#define MCREG_X27(mc)     ((mc).regs[27])
+#define MCREG_X28(mc)     ((mc).regs[28])
+#define MCREG_Fp(mc)      ((mc).regs[29])
+#define MCREG_Lr(mc)      ((mc).regs[30])
+
+#define MCREG_Sp(mc)      ((mc).sp)
+#define MCREG_Pc(mc)      ((mc).pc)
+#define MCREG_PState(mc)  ((mc).pstate)
+#define MCREG_Cpsr(mc)    ((mc).cpsr)
+#else
     // For FreeBSD, as found in x86/ucontext.h
 #define MCREG_Rbp(mc)	    ((mc).mc_rbp)
 #define MCREG_Rip(mc)	    ((mc).mc_rip)
@@ -138,6 +176,7 @@ typedef ucontext_t native_context_t;
 
 #define FPREG_Xmm(uc, index)    *(M128A*) &(FPSTATE(uc)->sv_xmm[index])
 #define FPREG_St(uc, index)     *(M128A*) &(FPSTATE(uc)->sv_fp[index].fp_acc)
+#endif
 
 #else // BIT64
 
diff --git a/src/pal/src/include/pal/cruntime.h b/src/pal/src/include/pal/cruntime.h
index d0f14a2503..56129de4df 100644
--- a/src/pal/src/include/pal/cruntime.h
+++ b/src/pal/src/include/pal/cruntime.h
@@ -122,6 +122,8 @@ Function:
 Parameters:
   ap
     - pointer to the va_list from which to remove arguments
+  Width
+    - the width of the current format option
   Precision
     - the precision of the current format option
   Type
@@ -129,7 +131,7 @@ Parameters:
   Prefix
     - the prefix for the current format option
 *******************************************************************************/
-void PAL_printf_arg_remover(va_list *ap, INT Precision, INT Type, INT Prefix);
+void PAL_printf_arg_remover(va_list *ap, INT Width, INT Precision, INT Type, INT Prefix);
 
 /*++
 Function:
diff --git a/src/pal/src/thread/context.cpp b/src/pal/src/thread/context.cpp
index 3742b5ddf1..050a4ac88a 100644
--- a/src/pal/src/thread/context.cpp
+++ b/src/pal/src/thread/context.cpp
@@ -41,6 +41,8 @@ extern void CONTEXT_CaptureContext(LPCONTEXT lpContext);
 #define CONTEXT_ALL_FLOATING CONTEXT_FLOATING_POINT
 #elif defined(_ARM_)
 #define CONTEXT_ALL_FLOATING CONTEXT_FLOATING_POINT
+#elif defined(_ARM64_)
+#define CONTEXT_ALL_FLOATING CONTEXT_FLOATING_POINT
 #else
 #error Unexpected architecture.
 #endif
@@ -117,6 +119,45 @@ extern void CONTEXT_CaptureContext(LPCONTEXT lpContext);
         ASSIGN_REG(R10)     \
         ASSIGN_REG(R11)     \
         ASSIGN_REG(R12)
+#elif defined(_ARM64_)
+#define ASSIGN_CONTROL_REGS \
+        ASSIGN_REG(Sp)      \
+        ASSIGN_REG(Lr)      \
+        ASSIGN_REG(Pc)
+
+#define ASSIGN_INTEGER_REGS \
+	ASSIGN_REG(X0)      \
+	ASSIGN_REG(X1)      \
+	ASSIGN_REG(X2)      \
+	ASSIGN_REG(X3)      \
+	ASSIGN_REG(X4)      \
+	ASSIGN_REG(X5)      \
+	ASSIGN_REG(X6)      \
+	ASSIGN_REG(X7)      \
+	ASSIGN_REG(X8)      \
+	ASSIGN_REG(X9)      \
+	ASSIGN_REG(X10)     \
+	ASSIGN_REG(X11)     \
+	ASSIGN_REG(X12)     \
+	ASSIGN_REG(X13)     \
+	ASSIGN_REG(X14)     \
+	ASSIGN_REG(X15)     \
+	ASSIGN_REG(X16)     \
+	ASSIGN_REG(X17)     \
+	ASSIGN_REG(X18)     \
+	ASSIGN_REG(X19)     \
+	ASSIGN_REG(X20)     \
+	ASSIGN_REG(X21)     \
+	ASSIGN_REG(X22)     \
+	ASSIGN_REG(X23)     \
+	ASSIGN_REG(X24)     \
+	ASSIGN_REG(X25)     \
+	ASSIGN_REG(X26)     \
+	ASSIGN_REG(X27)     \
+	ASSIGN_REG(X28)
+
+#else
+#error Don't know how to assign registers on this architecture
 #endif
 
 #define ASSIGN_ALL_REGS     \
@@ -201,6 +242,10 @@ BOOL CONTEXT_GetRegisters(DWORD processId, ucontext_t *registers)
 #define ASSIGN_REG(reg) MCREG_##reg(registers->uc_mcontext) = PTREG_##reg(ptrace_registers);
 #elif HAVE_BSD_REGS_T
 #define ASSIGN_REG(reg) MCREG_##reg(registers->uc_mcontext) = BSDREG_##reg(ptrace_registers);
+#else
+#define ASSIGN_REG(reg)
+	ASSERT("Don't know how to get the context of another process on this platform!");
+	return bRet;
 #endif
         ASSIGN_ALL_REGS
 #undef ASSIGN_REG
@@ -352,6 +397,10 @@ CONTEXT_SetThreadContext(
 #define ASSIGN_REG(reg) PTREG_##reg(ptrace_registers) = lpContext->reg;
 #elif HAVE_BSD_REGS_T
 #define ASSIGN_REG(reg) BSDREG_##reg(ptrace_registers) = lpContext->reg;
+#else
+#define ASSIGN_REG(reg)
+	ASSERT("Don't know how to set the context of another process on this platform!");
+	return FALSE;
 #endif
         if (lpContext->ContextFlags & CONTEXT_CONTROL)
         {
@@ -517,6 +566,10 @@ LPVOID CONTEXTGetPC(const native_context_t *context)
     return (LPVOID) MCREG_Eip(context->uc_mcontext);
 #elif defined(_ARM_)
     return (LPVOID) MCREG_Pc(context->uc_mcontext);
+#elif defined(_ARM64_)
+    return (LPVOID) MCREG_Pc(context->uc_mcontext);
+#else
+#   error implement me for this architecture
 #endif
 }
 
diff --git a/src/pal/tests/CMakeLists.txt b/src/pal/tests/CMakeLists.txt
index d3b91d2ac2..1381a98fd7 100644
--- a/src/pal/tests/CMakeLists.txt
+++ b/src/pal/tests/CMakeLists.txt
@@ -4,6 +4,8 @@ if(CLR_CMAKE_PLATFORM_ARCH_AMD64)
     set(PAL_CMAKE_PLATFORM_ARCH_AMD64 1)
 elseif(CLR_CMAKE_PLATFORM_ARCH_ARM)
     set(PAL_CMAKE_PLATFORM_ARCH_ARM 1)
+elseif(CLR_CMAKE_PLATFORM_ARCH_ARM64)
+    set(PAL_CMAKE_PLATFORM_ARCH_ARM64 1)
 endif()
 
 # Compile options
@@ -19,8 +21,11 @@ if(PAL_CMAKE_PLATFORM_ARCH_AMD64)
 elseif(PAL_CMAKE_PLATFORM_ARCH_ARM)
   add_definitions(-DBIT32=1)
   add_definitions(-D_WIN32=1)
+elseif(PAL_CMAKE_PLATFORM_ARCH_ARM64)
+  add_definitions(-DBIT64=1)
+  add_definitions(-D_WIN64=1)
 else()
-  message(FATAL_ERROR "Only ARM and AMD64 is supported")
+  message(FATAL_ERROR "Only ARM, ARM64 and AMD64 is supported")
 endif()
 
 add_compile_options(-Wno-empty-body)
diff --git a/src/unwinder/CMakeLists.txt b/src/unwinder/CMakeLists.txt
index 27e613cd3c..86c935a578 100644
--- a/src/unwinder/CMakeLists.txt
+++ b/src/unwinder/CMakeLists.txt
@@ -23,8 +23,14 @@ elseif(CLR_CMAKE_PLATFORM_ARCH_ARM)
     list(APPEND UNWINDER_SOURCES
         arm/unwinder_arm.cpp
     )
+elseif(CLR_CMAKE_PLATFORM_ARCH_ARM64)
+    include_directories(arm64)
+    
+    list(APPEND UNWINDER_SOURCES
+        arm64/unwinder_arm64.cpp
+    )
 else()
-  message(FATAL_ERROR "Only ARM and AMD64 is supported")
+  message(FATAL_ERROR "Only ARM, ARM64 and AMD64 is supported")
 endif()
 
 convert_to_absolute_path(UNWINDER_SOURCES ${UNWINDER_SOURCES})
@@ -37,4 +43,4 @@ if(CLR_CMAKE_PLATFORM_UNIX)
     add_subdirectory(wks)
 endif(CLR_CMAKE_PLATFORM_UNIX)
 
-add_subdirectory(dac)
\ No newline at end of file
+add_subdirectory(dac)
diff --git a/src/unwinder/arm64/unwinder_arm64.cpp b/src/unwinder/arm64/unwinder_arm64.cpp
index a56e42469e..3bf4d7e4f0 100644
--- a/src/unwinder/arm64/unwinder_arm64.cpp
+++ b/src/unwinder/arm64/unwinder_arm64.cpp
@@ -215,7 +215,11 @@ Return Value:
         SourceAddress = StartingSp + FIELD_OFFSET(ARM64_KTRAP_FRAME, X);
         for (RegIndex = 0; RegIndex < 18; RegIndex++) {
             UPDATE_CONTEXT_POINTERS(UnwindParams, RegIndex, SourceAddress);
+#ifdef __clang__
+            *(&ContextRecord->X0 + (RegIndex * sizeof(void*))) = MEMORY_READ_QWORD(UnwindParams, SourceAddress);
+#else
             ContextRecord->X[RegIndex] = MEMORY_READ_QWORD(UnwindParams, SourceAddress);
+#endif
             SourceAddress += sizeof(ULONG_PTR);
         }
         
@@ -288,10 +292,14 @@ Return Value:
         // Restore X0-X28, and D0-D31
         //
         
-        SourceAddress = StartingSp + FIELD_OFFSET(T_CONTEXT, X);
+        SourceAddress = StartingSp + FIELD_OFFSET(T_CONTEXT, X0);
         for (RegIndex = 0; RegIndex < 29; RegIndex++) {
             UPDATE_CONTEXT_POINTERS(UnwindParams, RegIndex, SourceAddress);
+#ifdef __clang__
+            *(&ContextRecord->X0 + (RegIndex * sizeof(void*))) = MEMORY_READ_QWORD(UnwindParams, SourceAddress);
+#else
             ContextRecord->X[RegIndex] = MEMORY_READ_QWORD(UnwindParams, SourceAddress);
+#endif
             SourceAddress += sizeof(ULONG_PTR);
         }
         
@@ -473,7 +481,11 @@ Return Value:
 
     for (RegIndex = 0; RegIndex < RegisterCount; RegIndex++) {
         UPDATE_CONTEXT_POINTERS(UnwindParams, RegIndex, CurAddress);
+#ifdef __clang__
+        *(&ContextRecord->X0 + (RegIndex * sizeof(void*))) = MEMORY_READ_QWORD(UnwindParams, CurAddress);
+#else
         ContextRecord->X[FirstRegister + RegIndex] = MEMORY_READ_QWORD(UnwindParams, CurAddress);
+#endif
         CurAddress += 8;
     }
     if (SpOffset < 0) {
@@ -1534,7 +1546,7 @@ BOOL OOPStackUnwinderArm64::Unwind(T_CONTEXT * pContext)
 
     if ((Rfe.UnwindData & 3) != 0) 
     {
-        hr = RtlpUnwindFunctionCompact(pContext->Pc - (ULONG)ImageBase,
+        hr = RtlpUnwindFunctionCompact(pContext->Pc - ImageBase,
                                         &Rfe,
                                         pContext,
                                         &DummyEstablisherFrame,
@@ -1545,8 +1557,8 @@ BOOL OOPStackUnwinderArm64::Unwind(T_CONTEXT * pContext)
     }
     else
     {
-        hr = RtlpUnwindFunctionFull(pContext->Pc - (ULONG)ImageBase,
-                                    (ULONG)ImageBase,
+        hr = RtlpUnwindFunctionFull(pContext->Pc - ImageBase,
+                                    ImageBase,
                                     &Rfe,
                                     pContext,
                                     &DummyEstablisherFrame,
@@ -1578,3 +1590,57 @@ BOOL DacUnwindStackFrame(T_CONTEXT *pContext, T_KNONVOLATILE_CONTEXT_POINTERS* p
     
     return res;
 }
+
+#if defined(FEATURE_PAL)
+//TODO: Fix the context pointers
+PEXCEPTION_ROUTINE
+RtlVirtualUnwind(
+    IN ULONG HandlerType,
+    IN ULONG64 ImageBase,
+    IN ULONG64 ControlPc,
+    IN PRUNTIME_FUNCTION FunctionEntry,
+    IN OUT PCONTEXT ContextRecord,
+    OUT PVOID *HandlerData,
+    OUT PULONG64 EstablisherFrame,
+    IN OUT PKNONVOLATILE_CONTEXT_POINTERS ContextPointers OPTIONAL
+    )
+{
+    PEXCEPTION_ROUTINE handlerRoutine;
+    HRESULT hr;
+
+    DWORD64 startingPc = ControlPc;
+    DWORD64 startingSp = ContextRecord->Sp;
+    
+    T_RUNTIME_FUNCTION rfe;
+
+    rfe.BeginAddress = FunctionEntry->BeginAddress;
+    rfe.UnwindData = FunctionEntry->UnwindData;
+
+    if ((rfe.UnwindData & 3) != 0) 
+    {
+        hr = RtlpUnwindFunctionCompact(ControlPc - ImageBase,
+                                        &rfe,
+                                        ContextRecord,
+                                        EstablisherFrame,
+                                        &handlerRoutine,
+                                        HandlerData,
+                                        NULL);
+
+    }
+    else
+    {
+        hr = RtlpUnwindFunctionFull(ControlPc - ImageBase,
+                                    ImageBase,
+                                    &rfe,
+                                    ContextRecord,
+                                    EstablisherFrame,
+                                    &handlerRoutine,
+                                    HandlerData,
+                                    NULL);
+    }
+
+    _ASSERTE(SUCCEEDED(hr));
+
+    return handlerRoutine;
+}
+#endif
diff --git a/src/unwinder/dac/CMakeLists.txt b/src/unwinder/dac/CMakeLists.txt
index 12163af12a..75d7050c6f 100644
--- a/src/unwinder/dac/CMakeLists.txt
+++ b/src/unwinder/dac/CMakeLists.txt
@@ -14,8 +14,14 @@ elseif(CLR_CMAKE_PLATFORM_ARCH_ARM)
   add_definitions(-DDBG_TARGET_ARM=1)
   add_definitions(-DDBG_TARGET_WIN32=1)
   add_definitions(-D_WIN32=1)
+elseif(CLR_CMAKE_PLATFORM_ARCH_ARM64)
+  add_definitions(-D_TARGET_ARM64_=1)
+  add_definitions(-DDBG_TARGET_64BIT=1)
+  add_definitions(-DDBG_TARGET_ARM64=1)
+  add_definitions(-DDBG_TARGET_WIN64=1)
+  add_definitions(-D_WIN64=1)
 else()
-  message(FATAL_ERROR "Only ARM and AMD64 is supported")
+  message(FATAL_ERROR "Only ARM, ARM64 and AMD64 is supported")
 endif()
 
 add_library(unwinder_dac ${UNWINDER_SOURCES})
diff --git a/src/vm/CMakeLists.txt b/src/vm/CMakeLists.txt
index fcc16acf78..9e5569f1f8 100644
--- a/src/vm/CMakeLists.txt
+++ b/src/vm/CMakeLists.txt
@@ -1,10 +1,5 @@
 set(CMAKE_INCLUDE_CURRENT_DIR ON)
 
-# WINTODO: Conditionalize the next check
-# AMD64 specific sources subdirectory
-set(AMD64_SOURCES_DIR amd64)
-set(ARM_SOURCES_DIR arm)
-
 # Needed due to the cmunged files being in the binary folders, the set(CMAKE_INCLUDE_CURRENT_DIR ON) is not enough
 include_directories(BEFORE ${CMAKE_CURRENT_SOURCE_DIR}) 
 
@@ -15,17 +10,22 @@ if(CLR_CMAKE_PLATFORM_ARCH_AMD64)
     add_definitions(-DDBG_TARGET_64BIT=1)
     add_definitions(-DDBG_TARGET_AMD64=1)
     add_definitions(-DDBG_TARGET_WIN64=1)
-    include_directories(${AMD64_SOURCES_DIR})
 elseif(CLR_CMAKE_PLATFORM_ARCH_ARM)
     add_definitions(-D_TARGET_ARM_=1)
     add_definitions(-DDBG_TARGET_32BIT=1)
     add_definitions(-DDBG_TARGET_ARM=1)
     add_definitions(-DDBG_TARGET_WIN32=1)
-    include_directories(${ARM_SOURCES_DIR})
+elseif(CLR_CMAKE_PLATFORM_ARCH_ARM64)
+    add_definitions(-D_TARGET_ARM64_=1)
+    add_definitions(-DDBG_TARGET_64BIT=1)
+    add_definitions(-DDBG_TARGET_ARM64=1)
+    add_definitions(-DDBG_TARGET_WIN64=1)
 else()
-    message(FATAL_ERROR "Only ARM and AMD64 is supported")
+    message(FATAL_ERROR "Only ARM, ARM64 and AMD64 is supported")
 endif()
 
+include_directories(${ARCH_SOURCES_DIR})
+
 add_definitions(-DFEATURE_LEAVE_RUNTIME_HOLDER=1)
 
 add_definitions(-DUNICODE)
@@ -314,51 +314,58 @@ list(APPEND VM_SOURCES_DAC
 
 # AMD64 specific asm sources
 set(VM_SOURCES_WKS_ARCH_ASM
-    ${AMD64_SOURCES_DIR}/AsmHelpers.asm
-    ${AMD64_SOURCES_DIR}/CallDescrWorkerAMD64.asm
-    ${AMD64_SOURCES_DIR}/ComCallPreStub.asm
-    ${AMD64_SOURCES_DIR}/CrtHelpers.asm
-    ${AMD64_SOURCES_DIR}/GenericComCallStubs.asm
-    ${AMD64_SOURCES_DIR}/GenericComPlusCallStubs.asm
-    ${AMD64_SOURCES_DIR}/getstate.asm
-    ${AMD64_SOURCES_DIR}/InstantiatingStub.asm
-    ${AMD64_SOURCES_DIR}/JitHelpers_Fast.asm
-    ${AMD64_SOURCES_DIR}/JitHelpers_FastWriteBarriers.asm
-    ${AMD64_SOURCES_DIR}/JitHelpers_InlineGetAppDomain.asm
-    ${AMD64_SOURCES_DIR}/JitHelpers_InlineGetThread.asm
-    ${AMD64_SOURCES_DIR}/JitHelpers_Slow.asm
-    ${AMD64_SOURCES_DIR}/PInvokeStubs.asm
-    ${AMD64_SOURCES_DIR}/RedirectedHandledJITCase.asm
-    ${AMD64_SOURCES_DIR}/ThePreStubAMD64.asm
-    ${AMD64_SOURCES_DIR}/ExternalMethodFixupThunk.asm
-    ${AMD64_SOURCES_DIR}/TlsGetters.asm # Condition="'$(FeatureImplicitTls)' != 'true'
-    ${AMD64_SOURCES_DIR}/UMThunkStub.asm
-    ${AMD64_SOURCES_DIR}/VirtualCallStubAMD64.asm
+    ${ARCH_SOURCES_DIR}/AsmHelpers.asm
+    ${ARCH_SOURCES_DIR}/CallDescrWorkerAMD64.asm
+    ${ARCH_SOURCES_DIR}/ComCallPreStub.asm
+    ${ARCH_SOURCES_DIR}/CrtHelpers.asm
+    ${ARCH_SOURCES_DIR}/GenericComCallStubs.asm
+    ${ARCH_SOURCES_DIR}/GenericComPlusCallStubs.asm
+    ${ARCH_SOURCES_DIR}/getstate.asm
+    ${ARCH_SOURCES_DIR}/InstantiatingStub.asm
+    ${ARCH_SOURCES_DIR}/JitHelpers_Fast.asm
+    ${ARCH_SOURCES_DIR}/JitHelpers_FastWriteBarriers.asm
+    ${ARCH_SOURCES_DIR}/JitHelpers_InlineGetAppDomain.asm
+    ${ARCH_SOURCES_DIR}/JitHelpers_InlineGetThread.asm
+    ${ARCH_SOURCES_DIR}/JitHelpers_Slow.asm
+    ${ARCH_SOURCES_DIR}/PInvokeStubs.asm
+    ${ARCH_SOURCES_DIR}/RedirectedHandledJITCase.asm
+    ${ARCH_SOURCES_DIR}/ThePreStubAMD64.asm
+    ${ARCH_SOURCES_DIR}/ExternalMethodFixupThunk.asm
+    ${ARCH_SOURCES_DIR}/TlsGetters.asm # Condition="'$(FeatureImplicitTls)' != 'true'
+    ${ARCH_SOURCES_DIR}/UMThunkStub.asm
+    ${ARCH_SOURCES_DIR}/VirtualCallStubAMD64.asm
 )
 
 else()
 
     if(CLR_CMAKE_PLATFORM_ARCH_AMD64)
         set(VM_SOURCES_WKS_ARCH_ASM
-            ${AMD64_SOURCES_DIR}/calldescrworkeramd64.S
-            ${AMD64_SOURCES_DIR}/crthelpers.S
-            ${AMD64_SOURCES_DIR}/externalmethodfixupthunk.S
-            ${AMD64_SOURCES_DIR}/getstate.S
-            ${AMD64_SOURCES_DIR}/jithelpers_fast.S
-            ${AMD64_SOURCES_DIR}/jithelpers_fastwritebarriers.S
-            ${AMD64_SOURCES_DIR}/jithelpers_slow.S
-            ${AMD64_SOURCES_DIR}/theprestubamd64.S
-            ${AMD64_SOURCES_DIR}/unixasmhelpers.S
-            ${AMD64_SOURCES_DIR}/umthunkstub.S
-            ${AMD64_SOURCES_DIR}/virtualcallstubamd64.S
+            ${ARCH_SOURCES_DIR}/calldescrworkeramd64.S
+            ${ARCH_SOURCES_DIR}/crthelpers.S
+            ${ARCH_SOURCES_DIR}/externalmethodfixupthunk.S
+            ${ARCH_SOURCES_DIR}/getstate.S
+            ${ARCH_SOURCES_DIR}/jithelpers_fast.S
+            ${ARCH_SOURCES_DIR}/jithelpers_fastwritebarriers.S
+            ${ARCH_SOURCES_DIR}/jithelpers_slow.S
+            ${ARCH_SOURCES_DIR}/theprestubamd64.S
+            ${ARCH_SOURCES_DIR}/unixasmhelpers.S
+            ${ARCH_SOURCES_DIR}/umthunkstub.S
+            ${ARCH_SOURCES_DIR}/virtualcallstubamd64.S
         )
     elseif(CLR_CMAKE_PLATFORM_ARCH_ARM)
         set(VM_SOURCES_WKS_ARCH_ASM
-            ${ARM_SOURCES_DIR}/asmhelpers.S
-            ${ARM_SOURCES_DIR}/crthelpers.S
-            ${ARM_SOURCES_DIR}/ehhelpers.S
-            ${ARM_SOURCES_DIR}/memcpy.S
-            ${ARM_SOURCES_DIR}/patchedcode.S
+            ${ARCH_SOURCES_DIR}/asmhelpers.S
+            ${ARCH_SOURCES_DIR}/crthelpers.S
+            ${ARCH_SOURCES_DIR}/ehhelpers.S
+            ${ARCH_SOURCES_DIR}/memcpy.S
+            ${ARCH_SOURCES_DIR}/patchedcode.S
+        )
+    elseif(CLR_CMAKE_PLATFORM_ARCH_ARM64)
+        set(VM_SOURCES_WKS_ARCH_ASM
+            ${ARCH_SOURCES_DIR}/asmhelpers.S
+            ${ARCH_SOURCES_DIR}/calldescrworkerarm64.S
+            ${ARCH_SOURCES_DIR}/crthelpers.S
+            ${ARCH_SOURCES_DIR}/pinvokestubs.S
         )
     endif()
     
@@ -367,47 +374,42 @@ endif(WIN32)
 
 if(CLR_CMAKE_PLATFORM_ARCH_AMD64)
     set(VM_SOURCES_DAC_AND_WKS_ARCH
-        ${AMD64_SOURCES_DIR}/cgenamd64.cpp
-        ${AMD64_SOURCES_DIR}/excepamd64.cpp
-        ${AMD64_SOURCES_DIR}/gmsamd64.cpp
-        ${AMD64_SOURCES_DIR}/stublinkeramd64.cpp
+        ${ARCH_SOURCES_DIR}/cgenamd64.cpp
+        ${ARCH_SOURCES_DIR}/excepamd64.cpp
+        ${ARCH_SOURCES_DIR}/gmsamd64.cpp
+        ${ARCH_SOURCES_DIR}/stublinkeramd64.cpp
     )
     
     set(VM_SOURCES_WKS_ARCH
-        ${AMD64_SOURCES_DIR}/jithelpersamd64.cpp
-        ${AMD64_SOURCES_DIR}/jitinterfaceamd64.cpp
-        ${AMD64_SOURCES_DIR}/profiler.cpp
+        ${ARCH_SOURCES_DIR}/jithelpersamd64.cpp
+        ${ARCH_SOURCES_DIR}/jitinterfaceamd64.cpp
+        ${ARCH_SOURCES_DIR}/profiler.cpp
         jitinterfacegen.cpp
     )
-    
-    if(CLR_CMAKE_PLATFORM_UNIX)
-    
-    list(APPEND VM_SOURCES_WKS_ARCH
-        ${AMD64_SOURCES_DIR}/unixstubs.cpp
-    )
-    
-    endif(CLR_CMAKE_PLATFORM_UNIX)
 elseif(CLR_CMAKE_PLATFORM_ARCH_ARM)
     set(VM_SOURCES_DAC_AND_WKS_ARCH
-        ${ARM_SOURCES_DIR}/exceparm.cpp
-        ${ARM_SOURCES_DIR}/stubs.cpp
-        ${ARM_SOURCES_DIR}/armsinglestepper.cpp
+        ${ARCH_SOURCES_DIR}/exceparm.cpp
+        ${ARCH_SOURCES_DIR}/stubs.cpp
+        ${ARCH_SOURCES_DIR}/armsinglestepper.cpp
     )
     
     set(VM_SOURCES_WKS_ARCH
-        ${ARM_SOURCES_DIR}/jithelpersarm.cpp
-        ${ARM_SOURCES_DIR}/profiler.cpp
+        ${ARCH_SOURCES_DIR}/jithelpersarm.cpp
+        ${ARCH_SOURCES_DIR}/profiler.cpp
     )
-    
-    if(CLR_CMAKE_PLATFORM_UNIX)
-    
-    list(APPEND VM_SOURCES_WKS_ARCH
-        ${ARM_SOURCES_DIR}/unixstubs.cpp
+elseif(CLR_CMAKE_PLATFORM_ARCH_ARM64)
+    set(VM_SOURCES_DAC_AND_WKS_ARCH
+        ${ARCH_SOURCES_DIR}/cgenarm64.cpp
+        ${ARCH_SOURCES_DIR}/stubs.cpp
     )
-    
-    endif(CLR_CMAKE_PLATFORM_UNIX)
 endif()
 
+if(CLR_CMAKE_PLATFORM_UNIX)
+    list(APPEND VM_SOURCES_WKS_ARCH
+        ${ARCH_SOURCES_DIR}/unixstubs.cpp
+    )
+endif(CLR_CMAKE_PLATFORM_UNIX)
+
 set(VM_SOURCES_DAC_ARCH
     gcinfodecoder.cpp
     dbggcinfodecoder.cpp
diff --git a/src/vm/arm64/asmconstants.h b/src/vm/arm64/asmconstants.h
index 04018bdb9f..6304d2c171 100644
--- a/src/vm/arm64/asmconstants.h
+++ b/src/vm/arm64/asmconstants.h
@@ -12,7 +12,7 @@
 // #error this file should only be used on an ARM platform
 // #endif // _ARM64_
 
-#include "..\..\inc\switches.h"
+#include "../../inc/switches.h"
 
 //-----------------------------------------------------------------------------
 
@@ -24,6 +24,10 @@
 #define ASMCONSTANTS_RUNTIME_ASSERT(cond)
 #endif
 
+#define DynamicHelperFrameFlags_Default     0
+#define DynamicHelperFrameFlags_ObjectArg   1
+#define DynamicHelperFrameFlags_ObjectArg2  2
+
 #define               Thread__m_fPreemptiveGCDisabled   0x0C
 #define               Thread__m_pFrame                  0x10
 
diff --git a/src/vm/arm64/asmhelpers.S b/src/vm/arm64/asmhelpers.S
new file mode 100644
index 0000000000..9b2f028250
--- /dev/null
+++ b/src/vm/arm64/asmhelpers.S
@@ -0,0 +1,844 @@
+//
+// Copyright (c) Geoff Norton. All rights reserved.
+// Copyright (c) Microsoft. All rights reserved.
+// Licensed under the MIT license. See LICENSE file in the project root for full license information. 
+//
+
+#include "asmconstants.h"
+#include "unixasmmacros.inc"
+
+// LPVOID __stdcall GetCurrentIP(void)//
+LEAF_ENTRY GetCurrentIP, _TEXT
+    mov x0, lr
+    ret lr
+LEAF_END GetCurrentIP, _TEXT
+
+// LPVOID __stdcall GetCurrentSP(void)//
+LEAF_ENTRY GetCurrentSP, _TEXT
+    mov x0, sp
+    ret lr
+LEAF_END GetCurrentSP, _TEXT
+
+//-----------------------------------------------------------------------------
+// This routine captures the machine state. It is used by helper method frame
+//-----------------------------------------------------------------------------
+//void LazyMachStateCaptureState(struct LazyMachState *pState)//
+LEAF_ENTRY LazyMachStateCaptureState, _TEXT
+    // marks that this is not yet valid
+    mov w1, #0
+    str w1, [x0, #MachState__isValid]
+
+    str lr, [x0, #LazyMachState_captureIp]
+
+    str fp, [x0, #LazyMachState_captureFp]
+
+    // str instruction does not save sp register directly so move to temp register
+    mov x1, sp
+    str x1, [x0, #LazyMachState_captureSp]
+
+    // save non-volatile registers that can contain object references
+    add x1, x0, #LazyMachState_captureX19_X28
+    stp x19, x20, [x1, #(16*0)]
+    stp x21, x22, [x1, #(16*1)]
+    stp x23, x24, [x1, #(16*2)]
+    stp x25, x26, [x1, #(16*3)]
+    stp x27, x28, [x1, #(16*4)]
+
+    ret lr
+LEAF_END LazyMachStateCaptureState, _TEXT
+
+//
+// If a preserved register were pushed onto the stack between
+// the managed caller and the H_M_F, ptrX19_X28 will point to its
+// location on the stack and it would have been updated on the
+// stack by the GC already and it will be popped back into the
+// appropriate register when the appropriate epilog is run.
+//
+// Otherwise, the register is preserved across all the code
+// in this HCALL or FCALL, so we need to update those registers
+// here because the GC will have updated our copies in the
+// frame.
+//
+// So, if ptrX19_X28 points into the MachState, we need to update
+// the register here.  That's what this macro does.
+//
+.macro RestoreRegMS regIndex, reg
+    // Incoming:
+    //
+    // x0 = address of MachState
+    //
+    // $regIndex: Index of the register (x19-x28). For x19, index is 19.
+    //For x20, index is 20, and so on.
+    //
+    // $reg: Register name (e.g. x19, x20, etc)
+    //
+    // Get the address of the specified captured egister from machine state
+    add x2, x0, #(MachState__captureX19_X28 + ((\regIndex-19)*8))
+
+    // Get the content of specified preserved register pointer from machine state
+    ldr x3, [x0, #(MachState__ptrX19_X28 + ((\regIndex-19)*8))]
+
+    cmp x2, x3
+    bne LOCAL_LABEL(NoRestore_\reg)
+    ldr \reg, [x2]
+LOCAL_LABEL(NoRestore_\reg):
+
+.endmacro
+
+// EXTERN_C int __fastcall HelperMethodFrameRestoreState(
+// INDEBUG_COMMA(HelperMethodFrame *pFrame)
+// MachState *pState
+// )
+LEAF_ENTRY HelperMethodFrameRestoreState, _TEXT
+
+    #ifdef _DEBUG
+    mov x0, x1
+    #endif
+
+    // If machine state is invalid, then simply exit
+    ldr x1, [x0, #MachState__isValid]
+    cmp x1, #0
+    beq LOCAL_LABEL(Done)
+
+    RestoreRegMS 19, X19
+    RestoreRegMS 20, X20
+    RestoreRegMS 21, X21
+    RestoreRegMS 22, X22
+    RestoreRegMS 23, X23
+    RestoreRegMS 24, X24
+    RestoreRegMS 25, X25
+    RestoreRegMS 26, X26
+    RestoreRegMS 27, X27
+    RestoreRegMS 28, X28
+LOCAL_LABEL(Done):
+    // Its imperative that the return value of HelperMethodFrameRestoreState is zero
+    // as it is used in the state machine to loop until it becomes zero.
+    // Refer to HELPER_METHOD_FRAME_END macro for details.
+    mov x0,#0
+    ret lr
+
+LEAF_END HelperMethodFrameRestoreState, _TEXT
+
+// ------------------------------------------------------------------
+// The call in ndirect import precode points to this function.
+NESTED_ENTRY NDirectImportThunk, _TEXT, NoHandler
+
+    PROLOG_SAVE_REG_PAIR_INDEXED fp, lr, #-144
+    SAVE_ARGUMENT_REGISTERS sp, 16
+    SAVE_FLOAT_ARGUMENT_REGISTERS sp, 80 
+
+    mov x0, x12
+    bl NDirectImportWorker
+    mov x12, x0
+
+    // pop the stack and restore original register state
+    RESTORE_FLOAT_ARGUMENT_REGISTERS sp, 80
+    RESTORE_ARGUMENT_REGISTERS sp, 16
+    EPILOG_RESTORE_REG_PAIR_INDEXED fp, lr, #144
+
+    // If we got back from NDirectImportWorker, the MD has been successfully
+    // linked. Proceed to execute the original DLL call.
+    EPILOG_BRANCH_REG x12
+
+NESTED_END NDirectImportThunk, _TEXT
+
+// ------------------------------------------------------------------
+// ARM64TODO: Implement PrecodeFixupThunk when PreCode is Enabled
+NESTED_ENTRY PrecodeFixupThunk, _TEXT, NoHandler
+     brk #0
+NESTED_END PrecodeFixupThunk, _TEXT
+// ------------------------------------------------------------------
+
+NESTED_ENTRY ThePreStub, _TEXT, NoHandler
+
+    PROLOG_WITH_TRANSITION_BLOCK 0, 0
+
+    add x0, sp, #__PWTB_TransitionBlock // pTransitionBlock
+    mov x1, METHODDESC_REGISTER // pMethodDesc
+
+    bl  PreStubWorker
+
+    mov x9, x0
+
+    EPILOG_WITH_TRANSITION_BLOCK_TAILCALL 0, 0
+    EPILOG_BRANCH_REG  x9
+
+NESTED_END ThePreStub, _TEXT
+
+// ------------------------------------------------------------------
+// ThePreStubPatch()
+
+LEAF_ENTRY ThePreStubPatch, _TEXT
+    nop
+.globl C_FUNC(ThePreStubPatchLabel)
+C_FUNC(ThePreStubPatchLabel):
+    ret lr
+LEAF_END ThePreStubPatch, _TEXT
+
+
+// ------------------------------------------------------------------
+// void ResolveWorkerAsmStub(args in regs x0-x7 & stack, x11:IndirectionCellAndFlags, x12:DispatchToken)
+//
+// The stub dispatch thunk which transfers control to VSD_ResolveWorker.
+NESTED_ENTRY ResolveWorkerAsmStub, _TEXT, NoHandler
+
+    PROLOG_WITH_TRANSITION_BLOCK 0, 0
+
+    add x0, sp, #__PWTB_TransitionBlock // pTransitionBlock
+    and x1, x11, #-4 // Indirection cell
+    mov x2, x12 // DispatchToken
+    and x3, x11, #3 // flag
+    bl C_FUNC(VSD_ResolveWorker)
+    mov x9, x0
+   
+    EPILOG_WITH_TRANSITION_BLOCK_TAILCALL 0, 0
+
+    EPILOG_BRANCH_REG x9
+
+NESTED_END ResolveWorkerAsmStub, _TEXT
+
+NESTED_ENTRY ResolveWorkerChainLookupAsmStub, _TEXT, NoHandler
+
+    // ARMSTUB TODO: implement chained lookup
+    b C_FUNC(ResolveWorkerAsmStub)
+
+NESTED_END ResolveWorkerChainLookupAsmStub, _TEXT
+
+//-----------------------------------------------------------------------------
+// The following Macros help in WRITE_BARRIER Implemetations
+// WRITE_BARRIER_ENTRY
+//
+// Declare the start of a write barrier function. Use similarly to NESTED_ENTRY. This is the only legal way
+// to declare a write barrier function.
+//
+.macro WRITE_BARRIER_ENTRY name
+    LEAF_ENTRY \name, _TEXT
+.endmacro
+
+// WRITE_BARRIER_END
+//
+// The partner to WRITE_BARRIER_ENTRY, used like NESTED_END.
+//
+.macro WRITE_BARRIER_END name
+    LEAF_END_MARKED \name, _TEXT
+.endmacro
+
+// void JIT_ByRefWriteBarrier
+// On entry:
+//   x13  : the source address (points to object reference to write)
+//   x14  : the destination address (object reference written here)
+//
+// On exit:
+//   x12  : trashed
+//   x13  : incremented by 8
+//   x14  : incremented by 8
+//   x15  : trashed
+//
+WRITE_BARRIER_ENTRY JIT_ByRefWriteBarrier
+
+    ldr  x15, [x13], 8
+    b C_FUNC(JIT_CheckedWriteBarrier)
+
+WRITE_BARRIER_END JIT_ByRefWriteBarrier 
+
+//-----------------------------------------------------------------------------
+// Simple WriteBarriers
+// void JIT_CheckedWriteBarrier(Object** dst, Object* src)
+// On entry:
+//   x14  : the destination address (LHS of the assignment)
+//   x15  : the object reference (RHS of the assignment)
+//
+// On exit:
+//   x12  : trashed
+//   x14  : incremented by 8
+//   x15  : trashed
+//
+WRITE_BARRIER_ENTRY JIT_CheckedWriteBarrier
+    // ARM64TODO: Temporary indirect access till support for :lo12:symbol is added
+    ldr  x12,  =g_lowest_address
+    ldr  x12,  [x12]
+    cmp  x14,  x12
+    blt  LOCAL_LABEL(NotInHeap)
+
+    // ARM64TODO: Temporary indirect access till support for :lo12:symbol is added
+    ldr  x12, =g_highest_address 
+    ldr  x12, [x12] 
+    cmp  x14, x12
+    blt  C_FUNC(JIT_WriteBarrier)
+
+LOCAL_LABEL(NotInHeap):
+    str  x15, [x14], 8
+    ret  lr
+WRITE_BARRIER_END JIT_CheckedWriteBarrier
+
+// void JIT_WriteBarrier(Object** dst, Object* src)
+// On entry:
+//   x14  : the destination address (LHS of the assignment)
+//   x15  : the object reference (RHS of the assignment)
+//
+// On exit:
+//   x12  : trashed
+//   x14  : incremented by 8
+//   x15  : trashed
+//
+WRITE_BARRIER_ENTRY JIT_WriteBarrier
+    dmb  ST
+    str  x15, [x14], 8
+
+    // Branch to Exit if the reference is not in the Gen0 heap
+    //
+    // ARM64TODO: Temporary indirect access till support for :lo12:symbol is added
+    ldr  x12,  =g_ephemeral_low
+    ldr  x12,  [x12]
+    cmp  x15,  x12
+    blt  LOCAL_LABEL(Exit)
+
+    // ARM64TODO: Temporary indirect access till support for :lo12:symbol is added
+    ldr  x12, =g_ephemeral_high 
+    ldr  x12, [x12]
+    cmp  x15,  x12
+    bgt  LOCAL_LABEL(Exit)
+
+    // Check if we need to update the card table
+    // ARM64TODO: Temporary indirect access till support for :lo12:symbol is added
+    ldr  x12, =g_card_table
+    ldr  x12, [x12]
+    add  x15,  x12, x14, lsr #11
+    ldrb w12, [x15]
+    cmp  x12, 0xFF
+    beq  LOCAL_LABEL(Exit)
+
+LOCAL_LABEL(UpdateCardTable):
+    mov  x12, 0xFF 
+    strb w12, [x15]
+LOCAL_LABEL(Exit):
+    ret  lr  
+WRITE_BARRIER_END JIT_WriteBarrier
+
+// ------------------------------------------------------------------
+// Start of the writeable code region
+LEAF_ENTRY JIT_PatchedCodeStart, _TEXT
+    ret  lr
+LEAF_END JIT_PatchedCodeStart, _TEXT
+
+// ------------------------------------------------------------------
+// End of the writeable code region
+LEAF_ENTRY JIT_PatchedCodeLast, _TEXT
+    ret  lr
+LEAF_END JIT_PatchedCodeLast, _TEXT
+
+//------------------------------------------------
+// VirtualMethodFixupStub
+//
+// In NGEN images, virtual slots inherited from cross-module dependencies
+// point to a jump thunk that calls into the following function that will
+// call into a VM helper. The VM helper is responsible for patching up
+// thunk, upon executing the precode, so that all subsequent calls go directly
+// to the actual method body.
+//
+// This is done lazily for performance reasons.
+//
+// On entry:
+//
+// x0 = "this" pointer
+// x12 = Address of thunk
+
+NESTED_ENTRY VirtualMethodFixupStub, _TEXT, NoHandler
+
+    // Save arguments and return address
+    PROLOG_SAVE_REG_PAIR_INDEXED fp, lr, #-144
+    SAVE_ARGUMENT_REGISTERS sp, 16
+    SAVE_FLOAT_ARGUMENT_REGISTERS sp, 80 
+
+    // Refer to ZapImportVirtualThunk::Save
+    // for details on this.
+    //
+    // Move the thunk start address in x1
+    mov x1, x12
+
+    // Call the helper in the VM to perform the actual fixup
+    // and tell us where to tail call. x0 already contains
+    // the this pointer.
+    bl C_FUNC(VirtualMethodFixupWorker)
+    // On return, x0 contains the target to tailcall to
+    mov x12, x0
+
+    // pop the stack and restore original register state
+    RESTORE_ARGUMENT_REGISTERS sp, 16
+    RESTORE_FLOAT_ARGUMENT_REGISTERS sp, 80
+    EPILOG_RESTORE_REG_PAIR_INDEXED fp, lr, #144
+
+    PATCH_LABEL VirtualMethodFixupPatchLabel
+
+    // and tailcall to the actual method
+    EPILOG_BRANCH_REG x12
+
+NESTED_END VirtualMEthodFixupStub, _TEXT
+
+//------------------------------------------------
+// ExternalMethodFixupStub
+//
+// In NGEN images, calls to cross-module external methods initially
+// point to a jump thunk that calls into the following function that will
+// call into a VM helper. The VM helper is responsible for patching up the
+// thunk, upon executing the precode, so that all subsequent calls go directly
+// to the actual method body.
+//
+// This is done lazily for performance reasons.
+//
+// On entry:
+//
+// x12 = Address of thunk 
+
+NESTED_ENTRY ExternalMethodFixupStub, _TEXT, NoHandler
+
+    PROLOG_WITH_TRANSITION_BLOCK 0, 0
+
+    add x0, sp, #__PWTB_TransitionBlock // pTransitionBlock
+    mov x1, x12 // pThunk
+
+    bl C_FUNC(ExternalMethodFixupWorker)
+
+    // mov the address we patched to in x12 so that we can tail call to it
+    mov x12, x0
+
+    EPILOG_WITH_TRANSITION_BLOCK_TAILCALL 0, 0
+    PATCH_LABEL ExternalMethodFixupPatchLabel
+    EPILOG_BRANCH_REG   x12
+
+NESTED_END ExternalMethodFixupStub, _TEXT
+
+// void SinglecastDelegateInvokeStub(Delegate *pThis)
+LEAF_ENTRY SinglecastDelegateInvokeStub, _TEXT
+    cmp x0, #0
+    beq LOCAL_LABEL(LNullThis)
+
+    ldr x16, [x0, #DelegateObject___methodPtr]
+    ldr x0, [x0, #DelegateObject___target]
+
+    br x16
+
+LOCAL_LABEL(LNullThis):
+    mov x0, #CORINFO_NullReferenceException_ASM
+    b C_FUNC(JIT_InternalThrow)
+
+LEAF_END SinglecastDelegateInvokeStub, _TEXT
+
+//
+// x12 = UMEntryThunk*
+//
+NESTED_ENTRY TheUMEntryPrestub, _TEXT, UnhandledExceptionHandlerUnix
+
+    // Save arguments and return address
+    PROLOG_SAVE_REG_PAIR_INDEXED fp, lr, #-144
+    SAVE_ARGUMENT_REGISTERS sp, 16
+    SAVE_FLOAT_ARGUMENT_REGISTERS sp, 80 
+
+    mov x0, x12
+    bl C_FUNC(TheUMEntryPrestubWorker)
+
+    // save real target address in x12.
+    mov x12, x0
+
+    // pop the stack and restore original register state
+    RESTORE_ARGUMENT_REGISTERS sp, 16
+    RESTORE_FLOAT_ARGUMENT_REGISTERS sp, 80
+    EPILOG_RESTORE_REG_PAIR_INDEXED fp, lr, #144
+
+    // and tailcall to the actual method
+    EPILOG_BRANCH_REG x12
+
+NESTED_END TheUMEntryPrestub, _TEXT
+
+//
+// x12 = UMEntryThunk*
+//
+NESTED_ENTRY UMThunkStub, _TEXT, UnhandledExceptionHandlerUnix
+
+    // Save arguments and return address
+    PROLOG_SAVE_REG_PAIR_INDEXED fp, lr, #-96 // 64 for regArgs, 8 for x19 & 8 for x12
+    // save callee saved reg x19. x19 is used in the method to store thread*
+    PROLOG_SAVE_REG x19, #88
+
+    SAVE_ARGUMENT_REGISTERS sp, 16
+
+#define UMThunkStub_HiddenArg 80 // offset of saved UMEntryThunk *
+#define UMThunkStub_StackArgs 96 // offset of original stack args (total size of UMThunkStub frame)
+
+    // save UMEntryThunk*
+    str x12, [sp, #UMThunkStub_HiddenArg]
+
+    // assuming GetThread does not clobber FP Args
+    bl C_FUNC(GetThread)
+    cbz x0, LOCAL_LABEL(UMThunkStub_DoThreadSetup)
+
+LOCAL_LABEL(UMThunkStub_HaveThread):
+    mov x19, x0  // x19 = Thread *
+
+    mov x9, 1
+    // m_fPreemptiveGCDisabled is 4 byte field so using 32-bit variant
+    str w9, [x19, #Thread__m_fPreemptiveGCDisabled]
+
+    ldr x2, =g_TrapReturningThreads
+    ldr x3, [x2]
+    // assuming x0 contains Thread* before jumping to UMThunkStub_DoTrapReturningThreads
+    cbnz x3, LOCAL_LABEL(UMThunkStub_DoTrapReturningThreads)
+
+LOCAL_LABEL(UMThunkStub_InCooperativeMode):
+    ldr x12, [fp, #UMThunkStub_HiddenArg] // x12 = UMEntryThunk*
+
+    ldr x0, [x19, #Thread__m_pDomain]
+
+    // m_dwDomainId is 4 bytes so using 32-bit variant
+    ldr w1, [x12, #UMEntryThunk__m_dwDomainId]
+    ldr w0, [x0, #AppDomain__m_dwId]
+    cmp w0, w1
+    bne LOCAL_LABEL(UMThunkStub_WrongAppDomain)
+
+    ldr x3, [x12, #UMEntryThunk__m_pUMThunkMarshInfo] // x3 = m_pUMThunkMarshInfo
+
+    // m_cbActualArgSize is UINT32 and hence occupies 4 bytes
+    ldr w2, [x3, #UMThunkMarshInfo__m_cbActualArgSize] // w2 = Stack arg bytes
+    cbz w2, LOCAL_LABEL(UMThunkStub_RegArgumentsSetup)
+
+    // extend to 64-bits
+    uxtw x2, w2
+
+    // Source pointer
+    add x0, fp, #UMThunkStub_StackArgs
+
+    // move source pointer to end of Stack Args
+    add x0, x0, x2 
+
+    // Count of stack slot pairs to copy (divide by 16)
+    lsr x1, x2, #4
+
+    // Is there an extra stack slot (can happen when stack arg bytes not multiple of 16)
+    and x2, x2, #8
+
+    // If yes then start source pointer from 16 byte aligned stack slot
+    add x0, x0, x2  
+
+    // increment stack slot pair count by 1 if x2 is not zero
+    add x1, x1, x2, LSR #3 
+
+LOCAL_LABEL(UMThunkStub_StackLoop):
+    ldp x4, x5, [x0, #-16]! // pre-Index
+    stp x4, x5, [sp, #-16]! // pre-Index
+    subs x1, x1, #1
+    bne LOCAL_LABEL(UMThunkStub_StackLoop)
+
+LOCAL_LABEL(UMThunkStub_RegArgumentsSetup):
+    ldr x16, [x3, #UMThunkMarshInfo__m_pILStub]
+
+    RESTORE_ARGUMENT_REGISTERS fp, 16
+
+    blr x16
+
+LOCAL_LABEL(UMThunkStub_PostCall):
+    mov x4, 0
+    // m_fPreemptiveGCDisabled is 4 byte field so using 32-bit variant
+    str w4, [x19, #Thread__m_fPreemptiveGCDisabled]
+
+    EPILOG_STACK_RESTORE
+    EPILOG_RESTORE_REG x19, #88
+    EPILOG_RESTORE_REG_PAIR_INDEXED   fp, lr, #96
+
+    EPILOG_RETURN
+
+LOCAL_LABEL(UMThunkStub_DoThreadSetup):
+    sub sp, sp, #SIZEOF__FloatArgumentRegisters
+    SAVE_FLOAT_ARGUMENT_REGISTERS sp, 0
+    bl C_FUNC(CreateThreadBlockThrow)
+    RESTORE_FLOAT_ARGUMENT_REGISTERS sp, 0
+    add sp, sp, #SIZEOF__FloatArgumentRegisters
+    b LOCAL_LABEL(UMThunkStub_HaveThread)
+
+LOCAL_LABEL(UMThunkStub_DoTrapReturningThreads):
+    sub sp, sp, #SIZEOF__FloatArgumentRegisters
+    SAVE_FLOAT_ARGUMENT_REGISTERS sp, 0
+    // x0 already contains Thread* pThread
+    // UMEntryThunk* pUMEntry
+    ldr x1, [fp, #UMThunkStub_HiddenArg]
+    bl C_FUNC(UMThunkStubRareDisableWorker)
+    RESTORE_FLOAT_ARGUMENT_REGISTERS sp, 0
+    add sp, sp, #SIZEOF__FloatArgumentRegisters
+    b LOCAL_LABEL(UMThunkStub_InCooperativeMode)
+
+LOCAL_LABEL(UMThunkStub_WrongAppDomain):
+    // Saving FP Args as this is read by UM2MThunk_WrapperHelper
+    sub sp, sp, #SIZEOF__FloatArgumentRegisters
+    SAVE_FLOAT_ARGUMENT_REGISTERS sp, 0
+
+    // UMEntryThunk* pUMEntry
+    ldr x0, [fp, #UMThunkStub_HiddenArg]
+
+    // void * pArgs
+    add x2, fp, #16  
+
+    // remaining arguments are unused
+    bl C_FUNC(UM2MDoADCallBack)
+
+    // restore integral return value
+    ldr x0, [fp, #16]
+
+    // restore FP or HFA return value
+    RESTORE_FLOAT_ARGUMENT_REGISTERS sp, 0
+
+    b LOCAL_LABEL(UMThunkStub_PostCall)
+
+NESTED_END UMThunkStub, _TEXT
+
+
+// UM2MThunk_WrapperHelper(void *pThunkArgs, // x0
+//                         int cbStackArgs,  // x1 (unused)
+//                         void *pAddr,  // x2 (unused)
+//                         UMEntryThunk *pEntryThunk,// x3
+//                         Thread *pThread)  // x4
+
+// pThunkArgs points to the argument registers pushed on the stack by UMThunkStub
+
+NESTED_ENTRY UM2MThunk_WrapperHelper, _TEXT, NoHandler
+
+    PROLOG_SAVE_REG_PAIR_INDEXED fp, lr, #-32
+    PROLOG_SAVE_REG  x19, #16
+
+
+    // save pThunkArgs in non-volatile reg. It is required after return from call to ILStub
+    mov x19, x0  
+
+    // ARM64TODO - Is this required by ILStub
+    mov x12, x3  //// x12 = UMEntryThunk *
+
+    //
+    // Note that layout of the arguments is given by UMThunkStub frame
+    //
+    ldr x3, [x3, #UMEntryThunk__m_pUMThunkMarshInfo]
+
+    // m_cbActualArgSize is 4-byte field
+    ldr w2, [x3, #UMThunkMarshInfo__m_cbActualArgSize]
+    cbz w2, LOCAL_LABEL(UM2MThunk_WrapperHelper_RegArgumentsSetup)
+
+    // extend to 64- bits
+    uxtw x2, w2 
+
+    // Source pointer. Subtracting 16 bytes due to fp & lr
+    add x6, x0, #(UMThunkStub_StackArgs-16) 
+
+    // move source ptr to end of Stack Args
+    add x6, x6, x2 
+
+    // Count of stack slot pairs to copy (divide by 16)
+    lsr x1, x2, #4
+
+    // Is there an extra stack slot? (can happen when stack arg bytes not multiple of 16)
+    and x2, x2, #8
+
+    // If yes then start source pointer from 16 byte aligned stack slot
+    add x6, x6, x2
+
+    // increment stack slot pair count by 1 if x2 is not zero
+    add x1, x1, x2, LSR #3
+
+LOCAL_LABEL(UM2MThunk_WrapperHelper_StackLoop):
+    ldp x4, x5, [x6, #-16]!
+    stp x4, x5, [sp, #-16]!
+    subs x1, x1, #1
+    bne LOCAL_LABEL(UM2MThunk_WrapperHelper_StackLoop)
+
+LOCAL_LABEL(UM2MThunk_WrapperHelper_RegArgumentsSetup):
+    ldr x16, [x3, #(UMThunkMarshInfo__m_pILStub)]
+
+    // reload floating point registers
+    RESTORE_FLOAT_ARGUMENT_REGISTERS x0, -1 * (SIZEOF__FloatArgumentRegisters + 16)
+
+    // reload argument registers
+    RESTORE_ARGUMENT_REGISTERS x0, 0
+
+    blr x16
+
+    // save integral return value
+    str x0, [x19]
+    // save FP/HFA return values
+    SAVE_FLOAT_ARGUMENT_REGISTERS x19, -1 * (SIZEOF__FloatArgumentRegisters + 16)
+
+    EPILOG_STACK_RESTORE
+    EPILOG_RESTORE_REG  x19, #16
+    EPILOG_RESTORE_REG_PAIR_INDEXED fp, lr, #32
+    EPILOG_RETURN
+
+NESTED_END UM2MThunk_WrapperHelper, _TEXT
+
+
+
+// ------------------------------------------------------------------
+// Redirection Stub for GC in fully interruptible method
+//GenerateRedirectedHandledJITCaseStub GCThreadControl
+// ------------------------------------------------------------------
+//GenerateRedirectedHandledJITCaseStub DbgThreadControl
+// ------------------------------------------------------------------
+//GenerateRedirectedHandledJITCaseStub UserSuspend
+// ------------------------------------------------------------------
+//GenerateRedirectedHandledJITCaseStub YieldTask
+
+#ifdef _DEBUG
+// ------------------------------------------------------------------
+// Redirection Stub for GC Stress
+GenerateRedirectedHandledJITCaseStub GCStress
+#endif
+
+
+// ------------------------------------------------------------------
+
+// This helper enables us to call into a funclet after restoring Fp register
+NESTED_ENTRY CallEHFunclet, _TEXT, NoHandler
+
+    // Using below prolog instead of PROLOG_SAVE_REG_PAIR fp,lr, #-16
+    // is intentional. Above statement would also emit instruction to save
+    // sp in fp. If sp is saved in fp in prolog then it is not expected that fp can change in the body
+    // of method. However, this method needs to be able to change fp before calling funclet.
+    // This is required to access locals in funclet.
+    PROLOG_SAVE_REG_PAIR_INDEXED x19,x20, #-16
+    PROLOG_SAVE_REG   fp, #0
+    PROLOG_SAVE_REG   lr, #8
+
+    // On entry:
+    //
+    // X0 = throwable
+    // X1 = PC to invoke
+    // X2 = address of X19 register in CONTEXT record// used to restore the non-volatile registers of CrawlFrame
+    // X3 = address of the location where the SP of funclet's caller (i.e. this helper) should be saved.
+    //
+    // Save the SP of this function
+    str fp, [x3]
+
+    ldr fp, [x2, #80] // offset of fp in CONTEXT relative to X19
+
+    // Invoke the funclet
+    blr x1
+    nop
+
+    EPILOG_RESTORE_REG_PAIR_INDEXED   fp, lr, #16
+    EPILOG_RETURN
+
+NESTED_END CallEHFunclet, _TEXT
+
+// This helper enables us to call into a filter funclet by passing it the CallerSP to lookup the 
+// frame pointer for accessing the locals in the parent method.
+NESTED_ENTRY CallEHFilterFunclet, _TEXT, NoHandler
+
+    PROLOG_SAVE_REG_PAIR_INDEXED   fp, lr, #-16
+
+    // On entry:
+    //
+    // X0 = throwable
+    // X1 = SP of the caller of the method/funclet containing the filter
+    // X2 = PC to invoke
+    // X3 = address of the location where the SP of funclet's caller (i.e. this helper) should be saved.
+    //
+    // Save the SP of this function
+    str fp, [x3]
+    // Invoke the filter funclet
+    blr x2
+
+    EPILOG_RESTORE_REG_PAIR_INDEXED   fp, lr,   #16
+    EPILOG_RETURN
+
+NESTED_END CallEHFilterFunclet, _TEXT
+
+#define FaultingExceptionFrame_StackAlloc (SIZEOF__GSCookie + SIZEOF__FaultingExceptionFrame)
+#define FaultingExceptionFrame_FrameOffset SIZEOF__GSCookie
+
+.macro GenerateRedirectedStubWithFrame stub, target
+
+    // 
+    // This is the primary function to which execution will be redirected to.
+    //
+    NESTED_ENTRY \stub, _TEXT, NoHandler
+
+        //
+        // IN: lr: original IP before redirect
+        //
+
+        PROLOG_SAVE_REG_PAIR_INDEXED fp, lr, #-16
+        PROLOG_STACK_ALLOC  FaultingExceptionFrame_StackAlloc
+
+        // At this point, the stack maybe misaligned if the thread abort was asynchronously
+        // triggered in the prolog or epilog of the managed method. For such a case, we must
+        // align the stack before calling into the VM.
+        //
+        // Runtime check for 16-byte alignment. 
+        mov x0, sp
+        and x0, x0, #15
+        sub sp, sp, x0
+
+        // Save pointer to FEF for GetFrameFromRedirectedStubStackFrame
+        add x19, sp, #FaultingExceptionFrame_FrameOffset
+
+        // Prepare to initialize to NULL
+        mov x1,#0
+        str x1, [x19]// Initialize vtbl (it is not strictly necessary)
+        str x1, [x19, #FaultingExceptionFrame__m_fFilterExecuted]// Initialize BOOL for personality routine
+
+        mov x0, x19   // move the ptr to FEF in X0
+
+        bl C_FUNC(\target)
+
+        // Target should not return.
+        EMIT_BREAKPOINT
+
+    NESTED_END \stub, _TEXT
+
+.endmacro
+
+
+// ------------------------------------------------------------------
+//
+// Helpers for async (NullRef, AccessViolation) exceptions
+//
+
+NESTED_ENTRY NakedThrowHelper2, _TEXT ,FixContextHandler
+    PROLOG_SAVE_REG_PAIR_INDEXED fp,lr, #-16
+
+    // On entry:
+    //
+    // X0 = Address of FaultingExceptionFrame
+    bl C_FUNC(LinkFrameAndThrow)
+
+    // Target should not return.
+    EMIT_BREAKPOINT
+
+NESTED_END NakedThrowHelper2, _TEXT
+
+GenerateRedirectedStubWithFrame NakedThrowHelper, NakedThrowHelper2
+
+#ifdef FEATURE_READYTORUN
+
+NESTED_ENTRY DelayLoad_MethodCall_FakeProlog, _TEXT, NoHandler
+DelayLoad_MethodCall:
+    .global DelayLoad_MethodCall
+
+    EMIT_BREAKPOINT
+NESTED_END DelayLoad_MethodCall_FakeProlog, _TEXT
+
+
+.macro DynamicHelper frameFlags, suffix
+    NESTED_ENTRY DelayLoad_Helper\suffix\()_FakeProlog, _TEXT, NoHandler
+DelayLoad_Helper\suffix:
+        .global DelayLoad_Helper\suffix
+
+        EMIT_BREAKPOINT
+
+    NESTED_END DelayLoad_Helper\suffix\()_FakeProlog, _TEXT
+.endm
+
+DynamicHelper DynamicHelperFrameFlags_Default
+DynamicHelper DynamicHelperFrameFlags_ObjectArg, _Obj
+DynamicHelper DynamicHelperFrameFlags_ObjectArg | DynamicHelperFrameFlags_ObjectArg2, _ObjObj
+
+#endif
+
+LEAF_ENTRY StartUnwindingNativeFrames, _TEXT
+    EMIT_BREAKPOINT
+LEAF_END StartUnwindingNativeFrames, _TEXT
+
diff --git a/src/vm/arm64/calldescrworkerarm64.S b/src/vm/arm64/calldescrworkerarm64.S
new file mode 100644
index 0000000000..2ff42f43dc
--- /dev/null
+++ b/src/vm/arm64/calldescrworkerarm64.S
@@ -0,0 +1,130 @@
+//
+// Copyright (c) Geoff Norton. All rights reserved.
+// Copyright (c) Microsoft. All rights reserved.
+// Licensed under the MIT license. See LICENSE file in the project root for full license information. 
+//
+
+#include "unixasmmacros.inc"
+#include "asmconstants.h"
+
+//-----------------------------------------------------------------------------
+// This helper routine enregisters the appropriate arguments and makes the
+// actual call.
+//-----------------------------------------------------------------------------
+//void CallDescrWorkerInternal(CallDescrData * pCallDescrData);
+
+NESTED_ENTRY CallDescrWorkerInternal, _TEXT, UnhandledExceptionHandlerUnix
+    PROLOG_SAVE_REG_PAIR_INDEXED   fp, lr, #-32
+    PROLOG_SAVE_REG    x19, #16   //the stack slot at sp+24 is empty for 16 byte alligment
+
+    mov     x19, x0 // save pCallDescrData in x19
+
+    ldr     w1, [x19,#CallDescrData__numStackSlots]
+    cbz     w1, LOCAL_LABEL(donestack)
+
+    // Add frame padding to ensure frame size is a multiple of 16 (a requirement of the OS ABI).
+    // We push two registers (above) and numStackSlots arguments (below). If this comes to an odd number
+    // of slots we must pad with another. This simplifies to "if the low bit of numStackSlots is set,
+    // extend the stack another eight bytes".
+    ldr     x0, [x19,#CallDescrData__pSrc]
+    add     x0, x0, x1, lsl #3           // pSrcEnd=pSrc+8*numStackSlots 
+    ands    x2, x1, #1
+    beq     LOCAL_LABEL(stackloop)
+    
+    // This loop copies numStackSlots words
+    // from [pSrcEnd-8,pSrcEnd-16,...] to [sp-8,sp-16,...]
+
+    // pad and store one stack slot as number of slots are odd
+    ldr     x4, [x0,#-8]!
+    str     x4, [sp,#-16]!
+    subs    x1, x1, #1
+    beq     LOCAL_LABEL(donestack)
+LOCAL_LABEL(stackloop):
+    ldp     x2, x4, [x0,#-16]!
+    stp     x2, x4, [sp,#-16]!
+    subs    x1, x1, #2
+    bne     LOCAL_LABEL(stackloop)
+LOCAL_LABEL(donestack):
+
+    // If FP arguments are supplied in registers (x8 != NULL) then initialize all of them from the pointer
+    // given in x8. 
+    ldr     x8, [x19,#CallDescrData__pFloatArgumentRegisters]
+    cbz     x8, LOCAL_LABEL(NoFloatingPoint)
+    ldp     d0, d1, [x8]
+    ldp     d2, d3, [x8, #16]
+    ldp     d4, d5, [x8, #32]
+    ldp     d6, d7, [x8, #48]
+LOCAL_LABEL(NoFloatingPoint):
+
+    // Copy [pArgumentRegisters, ..., pArgumentRegisters + 56]
+    // into x0, ..., x7
+
+    ldr     x8, [x19,#CallDescrData__pArgumentRegisters]
+    ldp     x0, x1, [x8]
+    ldp     x2, x3, [x8, #16]
+    ldp     x4, x5, [x8, #32]
+    ldp     x6, x7, [x8, #48]
+
+    // ARM64TODO: => see if anything special needs to be done for remoting
+    // call pTarget
+    ldr     x8, [x19,#CallDescrData__pTarget]
+    blr     x8
+
+    ldr     w3, [x19,#CallDescrData__fpReturnSize]
+
+    // Int return case
+    cbz     w3, LOCAL_LABEL(IntReturn)
+
+    // Float return case
+    cmp     w3, #4
+    beq     LOCAL_LABEL(FloatReturn)
+
+    // Double return case
+    cmp     w3, #8
+    bne     LOCAL_LABEL(NoDoubleReturn)
+
+LOCAL_LABEL(FloatReturn):
+    str     d0, [x19, #(CallDescrData__returnValue + 0)]
+    b       LOCAL_LABEL(ReturnDone)
+
+LOCAL_LABEL(NoDoubleReturn):
+
+    //FloatHFAReturn  return case
+    cmp     w3, #16
+    bne     LOCAL_LABEL(NoFloatHFAReturn)
+
+    stp     s0, s1, [x19, #(CallDescrData__returnValue + 0)]
+    stp     s2, s3, [x19, #(CallDescrData__returnValue + 0x08)]
+    b       LOCAL_LABEL(ReturnDone)
+LOCAL_LABEL(NoFloatHFAReturn):
+
+    //DoubleHFAReturn  return case
+    cmp     w3, #32
+    bne     LOCAL_LABEL(NoDoubleHFAReturn)
+
+    stp     d0, d1, [x19, #(CallDescrData__returnValue + 0)]
+    stp     d2, d3, [x19, #(CallDescrData__returnValue + 0x10)]
+    b       LOCAL_LABEL(ReturnDone)
+
+LOCAL_LABEL(NoDoubleHFAReturn):
+
+    EMIT_BREAKPOINT // Unreachable
+
+LOCAL_LABEL(IntReturn):
+    // Save return value into retbuf for int
+    str     x0, [x19, #(CallDescrData__returnValue + 0)]
+
+LOCAL_LABEL(ReturnDone):
+
+#ifdef _DEBUG
+    // trash the floating point registers to ensure that the HFA return values 
+    // won't survive by accident
+    ldp     d0, d1, [sp]
+    ldp     d2, d3, [sp, #16]
+#endif
+
+    EPILOG_STACK_RESTORE
+    EPILOG_RESTORE_REG      x19, #16    //the stack slot at sp+24 is empty for 16 byte alligment
+    EPILOG_RESTORE_REG_PAIR_INDEXED fp, lr, #32
+    EPILOG_RETURN
+NESTED_END CallDescrWorkerInternal, _TEXT
diff --git a/src/vm/arm64/cgencpu.h b/src/vm/arm64/cgencpu.h
index 28d4474fa4..ac9921a23c 100644
--- a/src/vm/arm64/cgencpu.h
+++ b/src/vm/arm64/cgencpu.h
@@ -407,7 +407,7 @@ extern "C" void SinglecastDelegateInvokeStub();
 #define DATA_ALIGNMENT 8
 
 
-DECLSPEC_ALIGN(16) struct UMEntryThunkCode
+struct DECLSPEC_ALIGN(16) UMEntryThunkCode
 {
     DWORD        m_code[4];
 
@@ -490,8 +490,8 @@ struct StubPrecode {
         CONTRACTL_END;
 
         EnsureWritableExecutablePages(&m_pTarget);
-        return (TADDR)InterlockedCompareExchange(
-            (TADDR*)&m_pTarget, (TADDR)target, (TADDR)expected) == expected;
+        return (TADDR)InterlockedCompareExchange64(
+            (LONGLONG*)&m_pTarget, (TADDR)target, (TADDR)expected) == expected;
     }
 
 #ifdef FEATURE_PREJIT
diff --git a/src/vm/arm64/crthelpers.S b/src/vm/arm64/crthelpers.S
new file mode 100644
index 0000000000..7d92917e8c
--- /dev/null
+++ b/src/vm/arm64/crthelpers.S
@@ -0,0 +1,292 @@
+//
+// Copyright (c) Microsoft. All rights reserved.
+// Licensed under the MIT license. See LICENSE file in the project root for full license information. 
+//
+
+// ==++==
+//
+
+#include "unixasmmacros.inc"
+
+//
+// ==--==
+
+// Calls to JIT_MemSet is emitted by jit for initialization of large structs. 
+// We need to provide our own implementation of memset instead of using the ones in crt because crt implementation does not gurantee 
+// that aligned 8/4/2 - byte memory will be written atomically. This is required because members in a struct can be read atomically 
+// and their values should be written atomically.
+// 
+//
+//void JIT_MemSet(void *dst, int val, SIZE_T count)
+//
+//    uintptr_t valEx = (char)val;
+//    valEx = valEx | valEx << 8;
+//    valEx = valEx | valEx << 16;
+//    valEx = valEx | valEx << 32;
+//
+//    // If not aligned then make it 8-byte aligned   
+//    if(((uintptr_t)dst&0x7) != 0)
+//    {
+//        if(((uintptr_t)dst&0x3) == 0)
+//        {
+//            *(UINT*)dst = (UINT)valEx;
+//            dst = (UINT*)dst + 1;
+//            count-=4;
+//        }
+//        else if(((uintptr_t)dst&0x1) == 0)
+//        {
+//            while(count > 0 && ((uintptr_t)dst&0x7) != 0)
+//            {
+//                *(short*)dst = (short)valEx;
+//                dst = (short*)dst + 1;
+//                count-=2;
+//            }
+//        }
+//        else
+//        {
+//            while(count > 0 && ((uintptr_t)dst&0x7) != 0)
+//            {
+//                *(char*)dst = (char)valEx;
+//                dst = (char*)dst + 1;
+//                count--;
+//            }
+//        }
+//    }
+//
+//    while(count > 8)
+//    {
+//        *(uintptr_t*)dst = valEx;
+//        dst = (uintptr_t*)dst + 1;
+//        count-=8;
+//    }
+//
+//    if(count & 4)
+//    {
+//        *(UINT*)dst = (UINT)valEx;
+//        dst = (UINT*)dst + 1;
+//    }
+//
+//    if(count & 2)
+//    {
+//        *(short*)dst = (short)valEx;
+//        dst = (short*)dst + 1;
+//    }
+//
+//    if(count & 1)
+//    {
+//        *(char*)dst = (char)valEx;
+//    }
+//
+//
+
+// Assembly code corresponding to above C++ method. JIT_MemSet can AV and clr exception personality routine needs to 
+// determine if the exception has taken place inside JIT_Memset in order to throw corresponding managed exception.
+// Determining this is slow if the method were implemented as C++ method (using unwind info). In .asm file by adding JIT_MemSet_End
+// marker it can be easily determined if exception happened in JIT_MemSet. Therefore, JIT_MemSet has been written in assembly instead of 
+// as C++ method.
+
+LEAF_ENTRY JIT_MemSet, _TEXT
+    sxtb        w8,w1
+    sxtw        x8,w8
+    orr         x8,x8,x8, lsl #8
+    orr         x8,x8,x8, lsl #0x10
+    orr         x9,x8,x8, lsl #0x20
+    and         x8,x0,#7
+    cbz         x8,LOCAL_LABEL(JIT_MemSet_0x7c)
+    and         x8,x0,#3
+    cbnz        x8,LOCAL_LABEL(JIT_MemSet_0x38)
+    str         w9,[x0]
+    add         x0,x0,#4
+    mov         x8,#-4
+    add         x2,x2,x8
+    b           LOCAL_LABEL(JIT_MemSet_0x7c)
+LOCAL_LABEL(JIT_MemSet_0x38):
+    cbz         x2,LOCAL_LABEL(JIT_MemSet_0x7c)
+    tbnz        x0,#0,LOCAL_LABEL(JIT_MemSet_0x60)
+LOCAL_LABEL(JIT_MemSet_0x40):
+    and         x8,x0,#7
+    cbz         x8,LOCAL_LABEL(JIT_MemSet_0x7c)
+    strh        w9,[x0]
+    add         x0,x0,#2
+    mov         x8,#-2
+    add         x2,x2,x8
+    cbnz        x2,LOCAL_LABEL(JIT_MemSet_0x40)
+    b           LOCAL_LABEL(JIT_MemSet_0x7c)
+LOCAL_LABEL(JIT_MemSet_0x60):
+    and         x8,x0,#7
+    cbz         x8,LOCAL_LABEL(JIT_MemSet_0x7c)
+    strb        w9,[x0]
+    add         x0,x0,#1
+    mov         x8,#-1
+    add         x2,x2,x8
+    cbnz        x2,LOCAL_LABEL(JIT_MemSet_0x60)
+LOCAL_LABEL(JIT_MemSet_0x7c):
+    cmp         x2,#8
+    bls         LOCAL_LABEL(JIT_MemSet_0xb8)
+    mov         x8,#-9
+    add         x8,x2,x8
+    lsr         x8,x8,#3
+    add         x11,x8,#1
+    mov         x10,x0
+    add         x8,x10,x11, lsl #3
+LOCAL_LABEL(JIT_MemSet_0x9c):
+    cmp         x10,x8
+    beq         LOCAL_LABEL(JIT_MemSet_0xac)
+    str         x9,[x10],#8
+    b           LOCAL_LABEL(JIT_MemSet_0x9c)
+LOCAL_LABEL(JIT_MemSet_0xac):
+    mov         x8,#-8
+    madd        x2,x11,x8,x2
+    add         x0,x0,x11, lsl #3
+LOCAL_LABEL(JIT_MemSet_0xb8):
+    tbz         x2,#2,LOCAL_LABEL(JIT_MemSet_0xc4)
+    str         w9,[x0]
+    add         x0,x0,#4
+LOCAL_LABEL(JIT_MemSet_0xc4):
+    tbz         x2,#1,LOCAL_LABEL(JIT_MemSet_0xd0)
+    strh        w9,[x0]
+    add         x0,x0,#2
+LOCAL_LABEL(JIT_MemSet_0xd0):
+    tbz         x2,#0,LOCAL_LABEL(JIT_MemSet_0xd8)
+    strb        w9,[x0]
+LOCAL_LABEL(JIT_MemSet_0xd8):
+    ret         lr
+LEAF_END JIT_MemSet, _TEXT
+
+// See comments above for JIT_MemSet
+
+//void JIT_MemCpy(void *dst, const void *src, SIZE_T count)
+//
+//    // If not aligned then make it 8-byte aligned   
+//    if(((uintptr_t)dst&0x7) != 0)
+//    {
+//        if(((uintptr_t)dst&0x3) == 0)
+//        {
+//            *(UINT*)dst = *(UINT*)src;
+//            dst = (UINT*)dst + 1;
+//            src = (UINT*)src + 1;
+//            count-=4;
+//        }
+//        else if(((uintptr_t)dst&0x1) == 0)
+//        {
+//            while(count > 0 && ((uintptr_t)dst&0x7) != 0)
+//            {
+//                *(short*)dst = *(short*)src;
+//                dst = (short*)dst + 1;
+//                src = (short*)src + 1;
+//                count-=2;
+//            }
+//        }
+//        else
+//        {
+//            while(count > 0 && ((uintptr_t)dst&0x7) != 0)
+//            {
+//                *(char*)dst = *(char*)src;
+//                dst = (char*)dst + 1;
+//                src = (char*)src + 1;
+//                count--;
+//            }
+//        }
+//    }
+//
+//    while(count > 8)
+//    {
+//        *(uintptr_t*)dst = *(uintptr_t*)src;
+//        dst = (uintptr_t*)dst + 1;
+//        src = (uintptr_t*)src + 1;
+//        count-=8;
+//    }
+//
+//    if(count & 4)
+//    {
+//        *(UINT*)dst = *(UINT*)src;
+//        dst = (UINT*)dst + 1;
+//        src = (UINT*)src + 1;
+//    }
+//
+//    if(count & 2)
+//    {
+//        *(short*)dst = *(short*)src;
+//        dst = (short*)dst + 1;
+//        src = (short*)src + 1;
+//    }
+//
+//    if(count & 1)
+//    {
+//        *(char*)dst = *(char*)src;
+//    }
+//
+//
+
+// Assembly code corresponding to above C++ method.
+// See comments above for JIT_MemSet method
+LEAF_ENTRY JIT_MemCpy, _TEXT
+    and         x8,x0,#7
+    cbz         x8,LOCAL_LABEL(JIT_MemCpy_0x80)
+    and         x8,x0,#3
+    cbnz        x8,LOCAL_LABEL(JIT_MemCpy_0x2c)
+    ldr         w8,[x1]
+    str         w8,[x0]
+    add         x0,x0,#4
+    add         x1,x1,#4
+    mov         x8,#-4
+    add         x2,x2,x8
+    b           LOCAL_LABEL(JIT_MemCpy_0x80)
+LOCAL_LABEL(JIT_MemCpy_0x2c):
+    cbz         x2,LOCAL_LABEL(JIT_MemCpy_0x80)
+    tbnz        x0,#0,LOCAL_LABEL(JIT_MemCpy_0x5c)
+LOCAL_LABEL(JIT_MemCpy_0x34):
+    and         x8,x0,#7
+    cbz         x8,LOCAL_LABEL(JIT_MemCpy_0x80)
+    ldrsh       w8,[x1]
+    strh        w8,[x0]
+    add         x0,x0,#2
+    add         x1,x1,#2
+    mov         x8,#-2
+    add         x2,x2,x8
+    cbnz        x2,LOCAL_LABEL(JIT_MemCpy_0x34)
+    b           LOCAL_LABEL(JIT_MemCpy_0x80)
+LOCAL_LABEL(JIT_MemCpy_0x5c):
+    and         x8,x0,#7
+    cbz         x8,LOCAL_LABEL(JIT_MemCpy_0x80)
+    ldrsb       w8,[x1]
+    strb        w8,[x0]
+    add         x0,x0,#1
+    add         x1,x1,#1
+    mov         x8,#-1
+    add         x2,x2,x8
+    cbnz        x2,LOCAL_LABEL(JIT_MemCpy_0x5c)
+LOCAL_LABEL(JIT_MemCpy_0x80):
+    cmp         x2,#8
+    bls         LOCAL_LABEL(JIT_MemCpy_0xb4)
+    mov         x8,#-9
+    add         x8,x2,x8
+    lsr         x8,x8,#3
+    add         x9,x8,#1
+    mov         x8,#-8
+    madd        x2,x9,x8,x2
+LOCAL_LABEL(JIT_MemCpy_0xa0):
+    ldr         x8,[x1],#8
+    str         x8,[x0],#8
+    mov         x8,#-1
+    add         x9,x9,x8
+    cbnz        x9,LOCAL_LABEL(JIT_MemCpy_0xa0)
+LOCAL_LABEL(JIT_MemCpy_0xb4):
+    tbz         x2,#2,LOCAL_LABEL(JIT_MemCpy_0xc8)
+    ldr         w8,[x1]
+    str         w8,[x0]
+    add         x0,x0,#4
+    add         x1,x1,#4
+LOCAL_LABEL(JIT_MemCpy_0xc8):
+    tbz         x2,#1,LOCAL_LABEL(JIT_MemCpy_0xdc)
+    ldrsh       w8,[x1]
+    strh        w8,[x0]
+    add         x0,x0,#2
+    add         x1,x1,#2
+LOCAL_LABEL(JIT_MemCpy_0xdc):
+    tbz         x2,#0,LOCAL_LABEL(JIT_MemCpy_0xe8)
+    ldrsb       w8,[x1]
+    strb        w8,[x0]
+LOCAL_LABEL(JIT_MemCpy_0xe8):
+    ret         lr
+LEAF_END JIT_MemCpy, _TEXT
diff --git a/src/vm/arm64/pinvokestubs.S b/src/vm/arm64/pinvokestubs.S
new file mode 100644
index 0000000000..b1dd5c4c35
--- /dev/null
+++ b/src/vm/arm64/pinvokestubs.S
@@ -0,0 +1,125 @@
+//
+// Copyright (c) Microsoft. All rights reserved.
+// Licensed under the MIT license. See LICENSE file in the project root for full license information. 
+//
+
+// ==++==
+//
+
+//
+// ==--==
+#include "asmconstants.h"
+#include "unixasmmacros.inc"
+
+// ------------------------------------------------------------------
+// Macro to generate PInvoke Stubs.
+// $__PInvokeStubFuncName : function which calls the actual stub obtained from VASigCookie
+// $__PInvokeGenStubFuncName : function which generates the IL stubs for PInvoke
+// 
+// Params :-
+// $FuncPrefix : prefix of the function name for the stub
+//                     Eg. VarargPinvoke, GenericPInvokeCalli
+// $VASigCookieReg : register which contains the VASigCookie
+// $SaveFPArgs : "Yes" or "No" . For varidic functions FP Args are not present in FP regs 
+//                        So need not save FP Args registers for vararg Pinvoke
+.macro PINVOKE_STUB FuncPrefix,VASigCookieReg,HiddenArg,SaveFPArgs
+
+#if NOTYET
+        GBLS __PInvokeStubFuncName
+        GBLS __PInvokeGenStubFuncName
+        GBLS __PInvokeStubWorkerName
+
+        IF "\FuncPrefix" == "GenericPInvokeCalli"
+__PInvokeStubFuncName SETS "\FuncPrefix":CC:"Helper"
+        ELSE
+__PInvokeStubFuncName SETS "\FuncPrefix":CC:"Stub"
+        ENDIF
+__PInvokeGenStubFuncName SETS "\FuncPrefix":CC:"GenILStub"
+__PInvokeStubWorkerName SETS "\FuncPrefix":CC:"StubWorker"
+
+       IF "\VASigCookieReg" == "x1"
+__PInvokeStubFuncName SETS "\__PInvokeStubFuncName":CC:"_RetBuffArg"
+__PInvokeGenStubFuncName SETS "\__PInvokeGenStubFuncName":CC:"_RetBuffArg"
+        ENDIF
+
+        NESTED_ENTRY \__PInvokeStubFuncName
+
+        // get the stub
+        ldr                 x9, [\VASigCookieReg, #VASigCookie__pNDirectILStub]
+
+        // if null goto stub generation
+        cbz                 x9, %0
+
+
+        EPILOG_BRANCH_REG   x9 
+
+0
+        EPILOG_BRANCH       \__PInvokeGenStubFuncName
+
+        NESTED_END
+
+        
+        NESTED_ENTRY \__PInvokeGenStubFuncName
+
+        PROLOG_WITH_TRANSITION_BLOCK 0, \SaveFPArgs
+
+        // x2 = Umanaged Target\MethodDesc
+        mov                 x2, \HiddenArg 
+
+        // x1 = VaSigCookie
+        IF "\VASigCookieReg" != "x1"
+        mov                 x1, \VASigCookieReg
+        ENDIF
+
+        // x0 = pTransitionBlock
+        add                 x0, sp, #__PWTB_TransitionBlock      
+
+        // save hidden arg
+        mov                 x19, \HiddenArg 
+
+        bl                  \__PInvokeStubWorkerName
+
+        // restore hidden arg (method desc or unmanaged target)
+        mov                 \HiddenArg , x19
+
+
+        EPILOG_WITH_TRANSITION_BLOCK_TAILCALL
+
+        EPILOG_BRANCH       \__PInvokeStubFuncName
+     
+        NESTED_END
+#else
+        EMIT_BREAKPOINT
+#endif
+.endm
+
+// ------------------------------------------------------------------
+// VarargPInvokeStub & VarargPInvokeGenILStub
+// There is a separate stub when the method has a hidden return buffer arg.
+//
+// in:
+// x0 = VASigCookie*
+// x12 = MethodDesc *       
+//
+PINVOKE_STUB VarargPInvoke, x0, x12, 1
+
+
+// ------------------------------------------------------------------
+// GenericPInvokeCalliHelper & GenericPInvokeCalliGenILStub
+// Helper for generic pinvoke calli instruction 
+//
+// in:
+// x15 = VASigCookie*
+// x14 = Unmanaged target
+//
+PINVOKE_STUB GenericPInvokeCalli, x15, x14, 1
+
+// ------------------------------------------------------------------
+// VarargPInvokeStub_RetBuffArg & VarargPInvokeGenILStub_RetBuffArg
+// Vararg PInvoke Stub when the method has a hidden return buffer arg
+//
+// in:
+// x1 = VASigCookie*
+// x12 = MethodDesc*       
+// 
+PINVOKE_STUB VarargPInvoke, x1, x12, 0
diff --git a/src/vm/arm64/stubs.cpp b/src/vm/arm64/stubs.cpp
index cca6b7143a..52fdc0ead8 100644
--- a/src/vm/arm64/stubs.cpp
+++ b/src/vm/arm64/stubs.cpp
@@ -315,7 +315,12 @@ void LazyMachState::unwindLazyState(LazyMachState* baseState,
 
     do {
 
+#ifndef FEATURE_PAL
         pvControlPc = Thread::VirtualUnwindCallFrame(&context, &nonVolContextPtrs);
+#else
+        PAL_VirtualUnwind(&context, &nonVolContextPtrs);
+        pvControlPc = GetIP(&context);
+#endif
 
         if (funCallDepth > 0)
         {
@@ -647,6 +652,8 @@ void UpdateRegDisplayFromCalleeSavedRegisters(REGDISPLAY * pRD, CalleeSavedRegis
     pContextPointers->X28 = (PDWORD64)&pCalleeSaved->x28;
 }
 
+#ifndef CROSSGEN_COMPILE
+
 void TransitionFrame::UpdateRegDisplay(const PREGDISPLAY pRD) 
 { 
     
@@ -656,7 +663,11 @@ void TransitionFrame::UpdateRegDisplay(const PREGDISPLAY pRD)
     // copy the argumetn registers
     ArgumentRegisters *pArgRegs = GetArgumentRegisters();
     for (int i = 0; i < ARGUMENTREGISTERS_SIZE; i++)
+#ifdef __clang__
+        *(&pRD->pCurrentContext->X0 + (sizeof(void*)*i)) = pArgRegs->x[i];
+#else
         pRD->pCurrentContext->X[i] = pArgRegs->x[i];
+#endif
 
     // copy the callee saved regs
     CalleeSavedRegisters *pCalleeSaved = GetCalleeSavedRegisters();
@@ -676,6 +687,8 @@ void TransitionFrame::UpdateRegDisplay(const PREGDISPLAY pRD)
 
 }
 
+#endif
+
 void TailCallFrame::UpdateRegDisplay(const PREGDISPLAY pRD) 
 { 
     _ASSERTE(!"ARM64:NYI");
@@ -1197,10 +1210,6 @@ VOID ResetCurrentContext()
 }
 #endif
 
-extern "C" void ResolveWorkerChainLookupAsmStub()
-{
-    _ASSERTE(!"ARM64:NYI");
-}
 extern "C" void StubDispatchFixupPatchLabel()
 {
     _ASSERTE(!"ARM64:NYI");
@@ -1693,10 +1702,12 @@ void StubLinkerCPU::EmitCallManagedMethod(MethodDesc *pMD, BOOL fTailCall)
     }
 }
 
+#ifndef CROSSGEN_COMPILE
+
 EXTERN_C UINT32 _tls_index;
 void StubLinkerCPU::EmitGetThreadInlined(IntReg Xt)
 {
-#ifdef FEATURE_IMPLICIT_TLS
+#if defined(FEATURE_IMPLICIT_TLS) && !defined(FEATURE_PAL)
     // Trashes x8.
     IntReg X8 = IntReg(8);
     _ASSERTE(Xt != X8);
@@ -1720,8 +1731,6 @@ void StubLinkerCPU::EmitGetThreadInlined(IntReg Xt)
 
 }
 
-#ifndef CROSSGEN_COMPILE
-
 void StubLinkerCPU::EmitUnboxMethodStub(MethodDesc *pMD)
 {
     _ASSERTE(!pMD->RequiresInstMethodDescArg());
@@ -1737,3 +1746,50 @@ void StubLinkerCPU::EmitUnboxMethodStub(MethodDesc *pMD)
 #endif // CROSSGEN_COMPILE
 
 #endif // #ifndef DACCESS_COMPILE
+
+#ifdef FEATURE_READYTORUN
+PCODE DynamicHelpers::CreateHelper(LoaderAllocator * pAllocator, TADDR arg, PCODE target)
+{
+    UNREACHABLE();
+}
+
+PCODE DynamicHelpers::CreateHelperWithArg(LoaderAllocator * pAllocator, TADDR arg, PCODE target)
+{
+    UNREACHABLE();
+}
+
+PCODE DynamicHelpers::CreateHelper(LoaderAllocator * pAllocator, TADDR arg, TADDR arg2, PCODE target)
+{
+    UNREACHABLE();
+}
+
+PCODE DynamicHelpers::CreateHelperArgMove(LoaderAllocator * pAllocator, TADDR arg, PCODE target)
+{
+    UNREACHABLE();
+}
+
+PCODE DynamicHelpers::CreateReturn(LoaderAllocator * pAllocator)
+{
+    UNREACHABLE();
+}
+
+PCODE DynamicHelpers::CreateReturnConst(LoaderAllocator * pAllocator, TADDR arg)
+{
+    UNREACHABLE();
+}
+
+PCODE DynamicHelpers::CreateReturnIndirConst(LoaderAllocator * pAllocator, TADDR arg, INT8 offset)
+{
+    UNREACHABLE();
+}
+
+PCODE DynamicHelpers::CreateHelperWithTwoArgs(LoaderAllocator * pAllocator, TADDR arg, PCODE target)
+{
+    UNREACHABLE();
+}
+
+PCODE DynamicHelpers::CreateHelperWithTwoArgs(LoaderAllocator * pAllocator, TADDR arg, TADDR arg2, PCODE target)
+{
+    UNREACHABLE();
+}
+#endif
diff --git a/src/vm/arm64/unixstubs.cpp b/src/vm/arm64/unixstubs.cpp
new file mode 100644
index 0000000000..c14ff59e53
--- /dev/null
+++ b/src/vm/arm64/unixstubs.cpp
@@ -0,0 +1,29 @@
+//
+// Copyright (c) Microsoft. All rights reserved.
+// Licensed under the MIT license. See LICENSE file in the project root for full license information. 
+//
+
+#include "common.h"
+
+extern "C"
+{
+    void RedirectForThrowControl()
+    {
+        PORTABILITY_ASSERT("Implement for PAL");
+    }
+
+    void GenericPInvokeCalliHelper()
+    {
+        PORTABILITY_ASSERT("Implement for PAL");
+    }
+
+    void VarargPInvokeStub()
+    {
+        PORTABILITY_ASSERT("Implement for PAL");
+    }
+    
+    void VarargPInvokeStub_RetBuffArg()
+    {
+        PORTABILITY_ASSERT("Implement for PAL");
+    }
+};
diff --git a/src/vm/arm64/virtualcallstubcpu.hpp b/src/vm/arm64/virtualcallstubcpu.hpp
index 9210a0b6fc..c302cee9b7 100644
--- a/src/vm/arm64/virtualcallstubcpu.hpp
+++ b/src/vm/arm64/virtualcallstubcpu.hpp
@@ -216,7 +216,7 @@ struct ResolveHolder
          //ldr w13, [x10 + DATA_OFFSET(_hashedToken)]
          offset = DATA_OFFSET(_hashedToken);
          _ASSERTE(offset >=0 && offset%8 == 0);
-         _stub._resolveEntryPoint[n++] = 0xB940014D | offset<<8;
+         _stub._resolveEntryPoint[n++] = 0xB940014D | offset<<7;
 
          //eor x9,x9,x13
          _stub._resolveEntryPoint[n++] = 0xCA0D0129;
@@ -326,13 +326,13 @@ struct ResolveHolder
          //ldr x12, [x10 , DATA_OFFSET(_token)]
          offset=DATA_OFFSET(_token);
          _ASSERTE(offset >=0 && offset%8 == 0);
-         _stub._slowEntryPoint[n++] = 0xF940014C | (offset<<10); 
+         _stub._slowEntryPoint[n++] = 0xF940014C | (offset<<7);
          
          //
          //ldr x9, [x10 , DATA_OFFSET(_resolveWorkerTarget)]
          offset=DATA_OFFSET(_resolveWorkerTarget);
          _ASSERTE(offset >=0 && offset%8 == 0);
-         _stub._slowEntryPoint[n++] = 0xF9400149 | (offset<<10); 
+         _stub._slowEntryPoint[n++] = 0xF9400149 | (offset<<7);
          
          //  br x9 
          _stub._slowEntryPoint[n++] = 0xD61F0120;
diff --git a/src/vm/callingconvention.h b/src/vm/callingconvention.h
index 2d3e0a7526..b1ed3853c6 100644
--- a/src/vm/callingconvention.h
+++ b/src/vm/callingconvention.h
@@ -368,7 +368,7 @@ public:
         if (m_argType == ELEMENT_TYPE_VALUETYPE)
         {
             _ASSERTE(!m_argTypeHandle.IsNull());
-            return ((m_argSize > ENREGISTERED_PARAMTYPE_MAXSIZE) && (!m_argTypeHandle.IsHFA() || IsVarArg()));
+            return ((m_argSize > ENREGISTERED_PARAMTYPE_MAXSIZE) && (!m_argTypeHandle.IsHFA() || this->IsVarArg()));
         }
         return FALSE;
 #else
diff --git a/src/vm/crossgen/CMakeLists.txt b/src/vm/crossgen/CMakeLists.txt
index e3dd990695..c27cb82831 100644
--- a/src/vm/crossgen/CMakeLists.txt
+++ b/src/vm/crossgen/CMakeLists.txt
@@ -114,8 +114,13 @@ elseif(CLR_CMAKE_PLATFORM_ARCH_ARM)
   list(APPEND VM_CROSSGEN_SOURCES 
     ../arm/stubs.cpp
   )
+elseif(CLR_CMAKE_PLATFORM_ARCH_ARM64)
+  include_directories(../arm64)
+  list(APPEND VM_CROSSGEN_SOURCES 
+    ../arm64/stubs.cpp
+  )
 else()
-  message(FATAL_ERROR "Only ARM and AMD64 is supported")
+  message(FATAL_ERROR "Only ARM, ARM64 and AMD64 is supported")
 endif()
 
 if (WIN32)
diff --git a/src/vm/frames.cpp b/src/vm/frames.cpp
index e65aade1cb..6e0a424c81 100644
--- a/src/vm/frames.cpp
+++ b/src/vm/frames.cpp
@@ -1068,7 +1068,7 @@ void InterpreterFrame::GcScanRoots(promote_func *fn, ScanContext* sc)
 
 #endif // FEATURE_INTERPRETER
 
-#ifdef _DEBUG
+#if defined(_DEBUG) && !defined (DACCESS_COMPILE)
 
 struct IsProtectedByGCFrameStruct
 {
diff --git a/src/vm/frames.h b/src/vm/frames.h
index 58702aa480..44e38bf155 100644
--- a/src/vm/frames.h
+++ b/src/vm/frames.h
@@ -754,7 +754,7 @@ public:
         return VPTR_HOST_VTABLE_TO_TADDR(*(LPVOID*)this);
     }
 
-#ifdef _DEBUG
+#if defined(_DEBUG) && !defined(DACCESS_COMPILE)
     virtual BOOL Protects(OBJECTREF *ppObjectRef)
     {
         LIMITED_METHOD_CONTRACT;
diff --git a/src/vm/gcinfodecoder.cpp b/src/vm/gcinfodecoder.cpp
index b231924e0f..c4dcfc4565 100644
--- a/src/vm/gcinfodecoder.cpp
+++ b/src/vm/gcinfodecoder.cpp
@@ -1793,6 +1793,24 @@ void GcInfoDecoder::ReportRegisterToGC( // ARM64
     pCallBack(hCallBack, pObjRef, gcFlags DAC_ARG(DacSlotLocation(regNum, 0, false)));
 }
 
+#ifdef FEATURE_PAL
+OBJECTREF* GcInfoDecoder::GetCapturedRegister(
+    int             regNum,
+    PREGDISPLAY     pRD
+    )
+{
+    _ASSERTE(regNum >= 0 && regNum <= 28);
+
+    // The fields of CONTEXT are in the same order as
+    // the processor encoding numbers.
+
+    DWORD64 *pX0;
+    pX0 = &pRD->pCurrentContext->X0;
+
+    return (OBJECTREF*)(pX0 + regNum);
+}
+#endif // FEATURE_PAL
+
 #else // Unknown platform
 
 OBJECTREF* GcInfoDecoder::GetRegisterSlot(
diff --git a/src/vm/interpreter.cpp b/src/vm/interpreter.cpp
index 614bb87645..fac99ecca6 100644
--- a/src/vm/interpreter.cpp
+++ b/src/vm/interpreter.cpp
@@ -42,22 +42,22 @@ static CorInfoType asCorInfoType(CORINFO_CLASS_HANDLE clsHnd)
 InterpreterMethodInfo::InterpreterMethodInfo(CEEInfo* comp, CORINFO_METHOD_INFO* methInfo)
     : m_method(methInfo->ftn),
       m_module(methInfo->scope),
+      m_jittedCode(0),
       m_ILCode(methInfo->ILCode),
       m_ILCodeEnd(methInfo->ILCode + methInfo->ILCodeSize),
       m_maxStack(methInfo->maxStack),
-      m_numArgs(methInfo->args.numArgs),
-      m_flags(0),
-      m_argDescs(NULL),
-      m_numLocals(methInfo->locals.numArgs),
-      m_returnType(methInfo->args.retType),
-      m_invocations(0),
-      m_jittedCode(0),
 #if INTERP_PROFILE
       m_totIlInstructionsExeced(0),
       m_maxIlInstructionsExeced(0),
 #endif
       m_ehClauseCount(methInfo->EHcount),
       m_varArgHandleArgNum(NO_VA_ARGNUM),
+      m_numArgs(methInfo->args.numArgs),
+      m_numLocals(methInfo->locals.numArgs),
+      m_flags(0),
+      m_argDescs(NULL),
+      m_returnType(methInfo->args.retType),
+      m_invocations(0),
       m_methodCache(NULL)
 { 
     // Overflow sanity check. (Can ILCodeSize ever be zero?)
@@ -431,7 +431,7 @@ InterpreterMethodInfo::~InterpreterMethodInfo()
 {
     if (m_methodCache != NULL)
     {
-        delete m_methodCache;
+        delete reinterpret_cast<ILOffsetToItemCache*>(m_methodCache);
     }
 }
 
@@ -534,7 +534,7 @@ void Interpreter::ArgState::AddArg(unsigned canonIndex, short numSlots, bool noR
         argOffsets[canonIndex] = offset.Value();
 #if defined(_ARM_) || defined(_ARM64_)
         callerArgStackSlots += numSlots;
-#endif;
+#endif
     }
 #endif // !_AMD64_
 }
@@ -1063,7 +1063,7 @@ CorJitResult Interpreter::GenerateInterpreterStub(CEEInfo* comp,
         // for instance a VT with two float fields will have the same size as a VT with 1 double field. (ARM64TODO: Verify it)
         // It works on ARM because the overlapping layout of the floating point registers
         // but it won't work on ARM64.
-        cHFAVars = (comp->getHFAType(info->args.retTypeClass) == ELEMENT_TYPE_R4) ? HFARetTypeSize/sizeof(float) : HFARetTypeSize/sizeof(double);
+        cHFAVars = (CorInfoTypeIsFloatingPoint(comp->getHFAType(info->args.retTypeClass))) ? HFARetTypeSize/sizeof(float) : HFARetTypeSize/sizeof(double);
 #endif
     }
 
@@ -1816,14 +1816,6 @@ enum OPCODE_2BYTE {
 #undef OPDEF
 };
 
-#ifdef _DEBUG
-static const char* getMethodName(CEEInfo* info, CORINFO_METHOD_HANDLE meth, const char** pClsName)
-{
-    GCX_PREEMP();
-    return info->getMethodName(meth, pClsName);
-}
-#endif // _DEBUG
-
 // Optimize the interpreter loop for speed.
 #ifdef _MSC_VER
 #pragma optimize("t", on)
@@ -4825,10 +4817,6 @@ void Interpreter::BinaryArithOvfOp()
         }
         break;
 
-    case CORINFO_TYPE_SHIFTED_CLASS:
-        VerificationError("Can't do binary arithmetic overflow operation on object references.");
-        break;
-
     default:
         _ASSERTE_MSG(false, "Non-stack-normal type on stack.");
     }
@@ -5099,7 +5087,7 @@ void Interpreter::ShiftOpWork(unsigned op1idx, CorInfoType cit2)
             res = (static_cast<UT>(val)) >> shiftAmt;
         }
     } 
-    else if (cit2 = CORINFO_TYPE_NATIVEINT)
+    else if (cit2 == CORINFO_TYPE_NATIVEINT)
     {
         NativeInt shiftAmt = OpStackGet<NativeInt>(op2idx);
         if (op == CEE_SHL)
@@ -5932,7 +5920,7 @@ void Interpreter::NewObj()
         {
             void* dest = LargeStructOperandStackPush(sz);
             memcpy(dest, tempDest, sz);
-            delete[] tempDest;
+            delete[] reinterpret_cast<BYTE*>(tempDest);
             OpStackSet<void*>(m_curStackHt, dest);
         }
         else
diff --git a/src/vm/interpreter.h b/src/vm/interpreter.h
index 6ecaded9e3..3a6f410810 100644
--- a/src/vm/interpreter.h
+++ b/src/vm/interpreter.h
@@ -77,7 +77,7 @@ bool IsStackNormalType(CorInfoType cit);
 CorInfoType CorInfoTypeStackNormalize(CorInfoType cit);
 
 // Returns the (byte) size of "cit".  Requires that "cit" is not a CORINFO_TYPE_VALUECLASS.
-inline size_t CorInfoTypeSize(CorInfoType cit);
+size_t CorInfoTypeSize(CorInfoType cit);
 
 // Returns true iff "cit" is an unsigned integral type.
 bool CorInfoTypeIsUnsigned(CorInfoType cit);
@@ -101,7 +101,7 @@ inline size_t CorInfoTypeStackNormalSize(CorInfoType cit)
     return CorInfoTypeSize(cit);
 }
 
-inline getClassSize(CORINFO_CLASS_HANDLE clsHnd)
+inline size_t getClassSize(CORINFO_CLASS_HANDLE clsHnd)
 {
     TypeHandle VMClsHnd(clsHnd);
     return VMClsHnd.GetSize();
@@ -778,32 +778,31 @@ public:
     // operand type stack.
     Interpreter(InterpreterMethodInfo* methInfo_, bool directCall_, BYTE* ilArgs_, void* stubContext_, BYTE* frameMemory) 
         : m_methInfo(methInfo_),
+          m_interpCeeInfo(methInfo_->m_method),
+          m_ILCodePtr(methInfo_->m_ILCode),
           m_directCall(directCall_),
           m_ilArgs(ilArgs_),
           m_stubContext(stubContext_),
-          m_ILCodePtr(methInfo_->m_ILCode),
-          m_curStackHt(0),
-          m_interpCeeInfo(methInfo_->m_method),
+          m_orOfPushedInterpreterTypes(0),
           m_largeStructOperandStack(NULL),
           m_largeStructOperandStackHt(0),
           m_largeStructOperandStackAllocSize(0),
-          m_thisArg(NULL),
-          m_securityObject(TADDR(NULL)),
-          m_args(NULL),
-          m_argsSize(0),
-          m_structRetValITPtr(NULL),
-          m_callThisArg(NULL), 
-          m_orOfPushedInterpreterTypes(0),
-          m_preciseGenericsContext(NULL),
-          m_functionPointerStack(NULL),
-          m_genericsCtxtArg(NULL),
-          m_inFlightException(NULL),
+          m_curStackHt(0),
+          m_leaveInfoStack(),
           m_filterNextScan(0),
           m_filterHandlerOffset(0),
           m_filterExcILOffset(0),
+          m_inFlightException(NULL),
+          m_thisArg(NULL),
 #ifdef USE_CHECKED_OBJECTREFS
           m_retBufArg(NULL),  // Initialize to NULL so we can safely declare protected.
 #endif // USE_CHECKED_OBJECTREFS
+          m_genericsCtxtArg(NULL),
+          m_securityObject(TADDR(NULL)),
+          m_args(NULL),
+          m_argsSize(0),
+          m_callThisArg(NULL), 
+          m_structRetValITPtr(NULL),
 #ifndef DACCESS_COMPILE
           // Means "uninitialized"
           m_thisExecCache(UninitExecCache), 
@@ -811,7 +810,8 @@ public:
           m_constrainedFlag(false),
           m_readonlyFlag(false),
           m_locAllocData(NULL),
-          m_leaveInfoStack()
+          m_preciseGenericsContext(NULL),
+          m_functionPointerStack(NULL)
     {
         // We must zero the locals.
         memset(frameMemory, 0, methInfo_->LocalMemSize() + sizeof(GSCookie));
@@ -1604,7 +1604,7 @@ private:
             {
                 for (unsigned i = 0; i < m_locAllocCurIdx; i++)
                 {
-                    delete[] m_locAllocs[i];
+                    delete[] reinterpret_cast<char*>(m_locAllocs[i]);
                 }
             }
             delete[] m_locAllocs;
diff --git a/src/vm/stublink.h b/src/vm/stublink.h
index 8456432fdc..5c40c8986a 100644
--- a/src/vm/stublink.h
+++ b/src/vm/stublink.h
@@ -322,6 +322,15 @@ protected:
         BOOL            m_fPushArgRegs;         // If true, r0-r3 are saved before callee saved regs
 #endif // _TARGET_ARM_
 
+#ifdef _TARGET_ARM64_
+protected:
+        BOOL            m_fProlog;              // True if DescribeProlog has been called
+        UINT            m_cIntRegArgs;          // Count of int register arguments (x0 - x7)
+        UINT            m_cVecRegArgs;          // Count of FP register arguments (v0 - v7)
+        UINT            m_cCalleeSavedRegs;     // Count of callee saved registers (x19 - x28)
+        UINT            m_cbStackSpace;         // Additional stack space for return buffer and stack alignment
+#endif // _TARGET_ARM64_
+
 #ifdef STUBLINKER_GENERATES_UNWIND_INFO
 
 #ifdef _DEBUG
@@ -368,12 +377,6 @@ private:
 
 #ifdef _TARGET_ARM64_
 #define MAX_UNWIND_CODE_WORDS 5  /* maximum number of 32-bit words to store unwind codes */
-protected:
-        BOOL            m_fProlog;              // True if DescribeProlog has been called
-        UINT            m_cIntRegArgs;          // Count of int register arguments (x0 - x7)
-        UINT            m_cVecRegArgs;          // Count of FP register arguments (v0 - v7)
-        UINT            m_cCalleeSavedRegs;     // Count of callee saved registers (x19 - x28)
-        UINT            m_cbStackSpace;         // Additional stack space for return buffer and stack alignment
 
 private:
         // Reserve fixed size block that's big enough to fit any unwind info we can have