Merge pull request #20507 from r30shah/zNextSupportPR

Add zNext support

runtime/cmake/caches/common.cmake

@@ -216,3 +216,4 @@ set(J9VM_THR_LOCK_RESERVATION ON CACHE BOOL "")
 set(J9VM_THR_PREEMPTIVE ON CACHE BOOL "")
 set(J9VM_THR_SMART_DEFLATION ON CACHE BOOL "")
 set(J9VM_OPT_OPENJDK_FFI ON CACHE BOOL "")
+set(J9VM_JIT_EMULATE_ZNEXT OFF CACHE BOOL "Enable ZNext emulation")

runtime/compiler/codegen/J9RecognizedMethodsEnum.hpp

@@ -527,6 +527,8 @@
    java_lang_Integer_reverseBytes,
    java_lang_Integer_rotateLeft,
    java_lang_Integer_rotateRight,
+   java_lang_Integer_compress,
+   java_lang_Integer_expand,
    java_lang_Integer_valueOf,
    java_lang_Integer_toUnsignedLong,
    java_lang_Integer_stringSize,
@@ -544,6 +546,8 @@
    java_lang_Long_reverseBytes,
    java_lang_Long_rotateLeft,
    java_lang_Long_rotateRight,
+   java_lang_Long_compress,
+   java_lang_Long_expand,
    java_lang_Short_reverseBytes,
    java_lang_Long_stringSize,
    java_lang_Long_toString,
@@ -720,6 +724,12 @@
    com_ibm_dataaccess_PackedDecimal_movePackedDecimal_,
    com_ibm_dataaccess_PackedDecimal_checkPackedDecimal_,
 
+   // wrapper methods
+   com_ibm_dataaccess_ExternalDecimal_checkExternalDecimal,
+
+   //inline methods
+   com_ibm_dataaccess_ExternalDecimal_checkExternalDecimal_,
+
    com_ibm_Compiler_Internal__TR_Prefetch,
 
    com_ibm_Compiler_Internal_Quad_enableQuadOptimization,

runtime/compiler/codegen/J9TreeEvaluator.cpp

@@ -36,6 +36,12 @@
 #include "runtime/J9ValueProfiler.hpp"
 #include "util_api.h"
 
+TR::Register*
+J9::TreeEvaluator::zdchkEvaluator(TR::Node *node, TR::CodeGenerator *cg)
+   {
+   return TR::TreeEvaluator::unImpOpEvaluator(node, cg);
+   }
+
 TR::Register*
 J9::TreeEvaluator::zdloadEvaluator(TR::Node *node, TR::CodeGenerator *cg)
    {

runtime/compiler/codegen/J9TreeEvaluator.hpp

@@ -75,6 +75,7 @@ class OMR_EXTENSIBLE TreeEvaluator: public OMR::TreeEvaluatorConnector
       float frequency;
       };
 
+   static TR::Register *zdchkEvaluator(TR::Node *node, TR::CodeGenerator *cg);
    static TR::Register *zdloadEvaluator(TR::Node *node, TR::CodeGenerator *cg);
    static TR::Register *zdloadiEvaluator(TR::Node *node, TR::CodeGenerator *cg);
    static TR::Register *zdstoreEvaluator(TR::Node *node, TR::CodeGenerator *cg);

runtime/compiler/env/j9method.cpp

@@ -2617,6 +2617,14 @@ void TR_ResolvedJ9Method::construct()
       {TR::unknownMethod}
    };
 
+   static X DataAccessExternalDecimalMethods[] =
+   {
+      {x(TR::com_ibm_dataaccess_ExternalDecimal_checkExternalDecimal , "checkExternalDecimal" , "([BIIII)I")},
+      {x(TR::com_ibm_dataaccess_ExternalDecimal_checkExternalDecimal_, "checkExternalDecimal_", "([BIIII)I")},
+
+      {TR::unknownMethod}
+   };
+
 
    static X BigDecimalMethods[] =
       {
@@ -3314,6 +3322,8 @@ void TR_ResolvedJ9Method::construct()
       {x(TR::java_lang_Integer_reverseBytes,            "reverseBytes",          "(I)I")},
       {x(TR::java_lang_Integer_rotateLeft,              "rotateLeft",            "(II)I")},
       {x(TR::java_lang_Integer_rotateRight,             "rotateRight",           "(II)I")},
+      {x(TR::java_lang_Integer_compress,                "compress",              "(II)I")},
+      {x(TR::java_lang_Integer_expand,                  "expand",                "(II)I")},
       {x(TR::java_lang_Integer_valueOf,                 "valueOf",               "(I)Ljava/lang/Integer;")},
       {  TR::java_lang_Integer_init,              6,    "<init>", (int16_t)-1,    "*"},
       {x(TR::java_lang_Integer_toUnsignedLong,          "toUnsignedLong",         "(I)J")},
@@ -3335,6 +3345,8 @@ void TR_ResolvedJ9Method::construct()
       {x(TR::java_lang_Long_reverseBytes,              "reverseBytes",           "(J)J")},
       {x(TR::java_lang_Long_rotateLeft,                 "rotateLeft",            "(JI)J")},
       {x(TR::java_lang_Long_rotateRight,                "rotateRight",           "(JI)J")},
+      {x(TR::java_lang_Long_compress,                   "compress",              "(JJ)J")},
+      {x(TR::java_lang_Long_expand,                     "expand",                "(JJ)J")},
       {  TR::java_lang_Long_init,                  6,    "<init>", (int16_t)-1,    "*"},
       {x(TR::java_lang_Long_stringSize,                 "stringSize",            "(J)I") },
       {x(TR::java_lang_Long_toString,                   "toString",            "(J)Ljava/lang/String;") },
@@ -4287,6 +4299,7 @@ void TR_ResolvedJ9Method::construct()
       { "java/util/Hashtable$HashEnumerator", HashtableHashEnumeratorMethods },
       { "com/ibm/Compiler/Internal/Prefetch", PrefetchMethods },
       { "java/lang/invoke/VarHandleInternal", VarHandleMethods },
+      { "com/ibm/dataaccess/ExternalDecimal", DataAccessExternalDecimalMethods },
       { 0 }
       };
 
@@ -9822,10 +9835,17 @@ TR_ResolvedJ9Method::isFieldFlattened(TR::Compilation *comp, int32_t cpIndex, bo
    return vmThread->javaVM->internalVMFunctions->isFlattenableFieldFlattened(reinterpret_cast<J9Class *>(containingClass), fieldShape);
    }
 
+bool
+TR_ResolvedJ9Method::isDAAExternalDecimalWrapperMethod()
+   {
+   // DAA External Decimal check method
+   return (this->TR_ResolvedMethod::getRecognizedMethod() == TR::com_ibm_dataaccess_ExternalDecimal_checkExternalDecimal);
+   }
+
 bool
 TR_ResolvedJ9Method::isDAAWrapperMethod()
    {
-   return isDAAMarshallingWrapperMethod() || isDAAPackedDecimalWrapperMethod();
+   return isDAAMarshallingWrapperMethod() || isDAAPackedDecimalWrapperMethod() || isDAAExternalDecimalWrapperMethod();
    }
 
 bool
@@ -9966,10 +9986,17 @@ TR_ResolvedJ9Method::isDAAPackedDecimalWrapperMethod()
    return false;
    }
 
+bool
+TR_ResolvedJ9Method::isDAAExternalDecimalIntrinsicMethod()
+   {
+   // DAA External Decimal check method
+   return (this->TR_ResolvedMethod::getRecognizedMethod() == TR::com_ibm_dataaccess_ExternalDecimal_checkExternalDecimal_);
+   }
+
 bool
 TR_ResolvedJ9Method::isDAAIntrinsicMethod()
    {
-   return isDAAMarshallingIntrinsicMethod() || isDAAPackedDecimalIntrinsicMethod();
+   return isDAAMarshallingIntrinsicMethod() || isDAAPackedDecimalIntrinsicMethod() || isDAAExternalDecimalIntrinsicMethod();
    }
 
 bool

runtime/compiler/env/j9method.h

@@ -471,9 +471,11 @@ class TR_ResolvedJ9Method : public TR_J9Method, public TR_ResolvedJ9MethodBase
    bool                            isDAAWrapperMethod();
    bool                            isDAAMarshallingWrapperMethod();
    bool                            isDAAPackedDecimalWrapperMethod();
+   bool                            isDAAExternalDecimalWrapperMethod();
    bool                            isDAAIntrinsicMethod();
    bool                            isDAAMarshallingIntrinsicMethod();
    bool                            isDAAPackedDecimalIntrinsicMethod();
+   bool 				                 isDAAExternalDecimalIntrinsicMethod();
 
 protected:
    TR_ResolvedMethod *             aotMaskResolvedPossiblyPrivateVirtualMethod(TR::Compilation *comp, TR_ResolvedMethod *method);

runtime/compiler/il/ILOpCodesEnum.hpp

@@ -26,7 +26,7 @@
 #include "compiler/il/OMRILOpCodesEnum.hpp" 
 
    FirstJ9Op = LastScalarOMROp + 1,
-   LastJ9Op = BCDCHK,
+   LastJ9Op = zdchk,
    FirstTROp = FirstOMROp,
    LastTROp = LastJ9Op,
 

runtime/compiler/il/Opcodes.enum

@@ -1542,3 +1542,19 @@ OPCODE_MACRO(\
    /* .ifCompareOpCode      = */ TR::BadILOp, \
    /* .description          = */ \
 )
+OPCODE_MACRO(\
+   /* .opcode               = */ zdchk, \
+   /* .name                 = */ "zdchk", \
+   /* .properties1          = */ 0, \
+   /* .properties2          = */ ILProp2::ValueNumberShare | ILProp2::SupportedForPRE, \
+   /* .properties3          = */ 0, \
+   /* .properties4          = */ ILProp4::BinaryCodedDecimalOp, \
+   /* .dataType             = */ TR::Int32, \
+   /* .typeProperties       = */ ILTypeProp::Size_4 | ILTypeProp::Integer, \
+   /* .childProperties      = */ TWO_CHILD(TR::ZonedDecimal, TR::Int8), \
+   /* .swapChildrenOpCode   = */ TR::BadILOp, \
+   /* .reverseBranchOpCode  = */ TR::BadILOp, \
+   /* .booleanCompareOpCode = */ TR::BadILOp, \
+   /* .ifCompareOpCode      = */ TR::BadILOp, \
+   /* .description          =    zoned decimal validity checking */ \
+)

runtime/compiler/ilgen/Walker.cpp

@@ -4190,6 +4190,8 @@ break
       DAA_PRINT(TR::com_ibm_dataaccess_PackedDecimal_shiftRightPackedDecimal);
       DAA_PRINT(TR::com_ibm_dataaccess_PackedDecimal_movePackedDecimal);
 
+      DAA_PRINT(TR::com_ibm_dataaccess_ExternalDecimal_checkExternalDecimal);
+
       default:
          break;
       }

runtime/compiler/optimizer/DataAccessAccelerator.cpp

@@ -343,6 +343,14 @@ int32_t TR_DataAccessAccelerator::performOnBlock(TR::Block* block, TreeTopContai
                      ++result;
                      }
                   break;
+               // DAA External Decimal Check
+               case TR::com_ibm_dataaccess_ExternalDecimal_checkExternalDecimal_:
+                  if (comp()->target().cpu.supportsFeature(OMR_FEATURE_S390_VECTOR_PACKED_DECIMAL_ENHANCEMENT_FACILITY_3)
+                        && inlineCheckExternalDecimal(treeTop, callNode))
+                     {
+                     ++result;
+                     }
+                  break;
 
                default:
                   matched = false;
@@ -846,6 +854,99 @@ bool TR_DataAccessAccelerator::inlineCheckPackedDecimal(TR::TreeTop *callTreeTop
    return false;
    }
 
+bool TR_DataAccessAccelerator::inlineCheckExternalDecimal(TR::TreeTop *callTreeTop, TR::Node *callNode)
+   {
+   TR::Node *byteArrayNode = callNode->getChild(0);
+   TR::Node *offsetNode = callNode->getChild(1);
+   TR::Node *precisionNode = callNode->getChild(2);
+   TR::Node *typeNode = callNode->getChild(3);
+   TR::Node *bytesWithSpacesNode = callNode->getChild(4);
+
+   int32_t precision = precisionNode->getInt();
+   int32_t bytesWithSpaces = bytesWithSpacesNode->getInt();
+   int32_t type = typeNode->getInt();
+   const char *failMsg = NULL;
+
+   /* Hardware expects both, precision and bytesWithSpaces to be
+    * 5 bit unsigned binary integer. However, 0 is valid only for
+    * bytesWithSpaces. This is why precision must be within [1-31]
+    * range and bytesWithSpaces must be within [0-31] range.
+    */
+   // TODO: Add support for non-constant arguments
+   if (!isChildConst(callNode, 2))
+      failMsg = "Precision is not constant";
+   else if (precision < 1 || precision > 31)
+      failMsg = "Precision value is not in valid range [1-31]";
+   else if (!isChildConst(callNode, 3))
+      failMsg = "Decimal type node is not constant";
+   else if (type < 1 || type > 4)
+      failMsg = "Invalid decimal type. Supported types are (1|2|3|4)";
+   else if (!isChildConst(callNode, 4))
+      failMsg = "bytesWithSpaces node is not constant";
+   else if (bytesWithSpaces < 0 || bytesWithSpaces > 31)
+      failMsg = "bytesWithSpaces value not in valid range [0-31]";
+
+   if (failMsg)
+      {
+      TR::DebugCounter::incStaticDebugCounter(comp(),
+                                               TR::DebugCounter::debugCounterName(comp(),
+                                                                                  "DAA/rejected/chkZonedDecimal"));
+
+      return printInliningStatus (false, callNode, failMsg);
+      }
+
+   if (performTransformation(comp(), "O^O TR_DataAccessAccelerator: inlineCheckZonedDecimal on callNode %p\n", callNode))
+      {
+      TR::DebugCounter::incStaticDebugCounter(comp(),
+                                              TR::DebugCounter::debugCounterName(comp(),
+                                                                                 "DAA/inlined/chkZonedDecimal"));
+
+      insertByteArrayNULLCHK(callTreeTop, callNode, byteArrayNode);
+
+      TR::DataType decimalType = TR::DataTypes::NoType;
+      TR::ILOpCodes loadOpCode = TR::BadILOp;
+      if (type == 1)
+         {
+         decimalType = TR::ZonedDecimal;
+         loadOpCode = TR::zdloadi;
+         }
+      else if (type == 2)
+         {
+         decimalType = TR::ZonedDecimalSignLeadingEmbedded;
+         loadOpCode = TR::zdsleLoadi;
+         }
+      else if (type == 3)
+         {
+         decimalType = TR::ZonedDecimalSignTrailingSeparate;
+         loadOpCode = TR::zdstsLoadi;
+         }
+      else if (type == 4)
+         {
+         decimalType = TR::ZonedDecimalSignLeadingSeparate;
+         loadOpCode = TR::zdslsLoadi;
+         }
+      int32_t precisionSizeInNumberOfBytes = TR::DataType::getSizeFromBCDPrecision(decimalType, precision);
+
+      insertByteArrayBNDCHK(callTreeTop, callNode, byteArrayNode, offsetNode, 0);
+      insertByteArrayBNDCHK(callTreeTop, callNode, byteArrayNode, offsetNode, precisionSizeInNumberOfBytes - 1);
+
+      TR::SymbolReference* zonedDecimalSymbolReference = comp()->getSymRefTab()->findOrCreateArrayShadowSymbolRef(decimalType, NULL, precisionSizeInNumberOfBytes, fe());
+      TR::Node* zdchkChild0Node = TR::Node::createWithSymRef(loadOpCode, 1, 1, constructAddressNode(callNode, byteArrayNode, offsetNode), zonedDecimalSymbolReference);
+      zdchkChild0Node->setDecimalPrecision(precision);
+
+      byteArrayNode->decReferenceCount();
+      offsetNode->decReferenceCount();
+      precisionNode->decReferenceCount();
+      typeNode->decReferenceCount();
+
+      TR::Node* bytesWithSpacesConstNode = TR::Node::bconst(static_cast<uint8_t>(bytesWithSpaces));
+      TR::Node::recreateWithoutProperties(callNode, TR::zdchk, 2, zdchkChild0Node, bytesWithSpacesConstNode);
+      return true;
+      }
+
+   return false;
+}
+
 TR::Node* TR_DataAccessAccelerator::insertIntegerGetIntrinsic(TR::TreeTop* callTreeTop, TR::Node* callNode, int32_t sourceNumBytes, int32_t targetNumBytes)
    {
    if (targetNumBytes != 1 && targetNumBytes != 2 && targetNumBytes != 4 && targetNumBytes != 8)

runtime/compiler/optimizer/DataAccessAccelerator.hpp

@@ -211,6 +211,7 @@ class TR_DataAccessAccelerator : public TR::Optimization
    TR::Node* insertDecimalSetIntrinsic(TR::TreeTop* callTreeTop, TR::Node* callNode, int32_t sourceNumBytes, int32_t targetNumBytes);
 
    bool inlineCheckPackedDecimal(TR::TreeTop* callTreeTop, TR::Node* callNode);
+   bool inlineCheckExternalDecimal(TR::TreeTop* callTreeTop, TR::Node* callNode);
 
    private:
 

runtime/compiler/optimizer/InlinerTempForJ9.cpp

@@ -5431,6 +5431,9 @@ TR_J9InlinerPolicy::supressInliningRecognizedInitialCallee(TR_CallSite* callsite
       // DAA Packed Decimal check method
       case TR::com_ibm_dataaccess_PackedDecimal_checkPackedDecimal_:
 
+      // DAA External Decimal check method
+      case TR::com_ibm_dataaccess_ExternalDecimal_checkExternalDecimal_:
+
       // DAA Packed Decimal <-> Integer
       case TR::com_ibm_dataaccess_DecimalData_convertPackedDecimalToInteger_:
       case TR::com_ibm_dataaccess_DecimalData_convertPackedDecimalToInteger_ByteBuffer_:

runtime/compiler/optimizer/J9LocalCSE.cpp

@@ -117,6 +117,7 @@ J9::LocalCSE::shouldCommonNode(TR::Node *parent, TR::Node *node)
             case TR::com_ibm_dataaccess_PackedDecimal_equalsPackedDecimal_:
             case TR::com_ibm_dataaccess_PackedDecimal_notEqualsPackedDecimal_:
             case TR::com_ibm_dataaccess_PackedDecimal_checkPackedDecimal_:
+            case TR::com_ibm_dataaccess_ExternalDecimal_checkExternalDecimal_:
             case TR::com_ibm_dataaccess_DecimalData_convertExternalDecimalToPackedDecimal_:
             case TR::com_ibm_dataaccess_DecimalData_convertPackedDecimalToExternalDecimal_:
             case TR::com_ibm_dataaccess_DecimalData_convertPackedDecimalToUnicodeDecimal_:

runtime/compiler/optimizer/J9SimplifierTable.enum

@@ -118,6 +118,7 @@
 #define pdModifyPrecisionSimplifierHandler pdshlSimplifier
 #define countDigitsSimplifierHandler dftSimplifier
 #define BCDCHKSimplifierHandler dftSimplifier
+#define zdchkSimplifierHandler dftSimplifier
 
 #include "optimizer/OMRSimplifierTable.enum"
 

runtime/compiler/runtime/SignalHandler.c

@@ -564,11 +564,11 @@ UDATA jitPPCHandler(J9VMThread* vmThread, U_32 sigType, void* sigInfo)
                                 *iarPtr = (UDATA) ((void *) &jitHandleInternalErrorTrap);
 #endif
                                 return J9PORT_SIG_EXCEPTION_CONTINUE_EXECUTION;
-	
+
 			}
 			else if (J9PORT_SIG_FLAG_SIGTRAP == sigType) {
 				IDATA trapType = jitPPCIdentifyCodeCacheTrapType((U_8 *) *iarPtr);
-				
+
 				switch (trapType) {
 
 				case TRAP_TYPE_NULL_CHECK:
@@ -937,8 +937,19 @@ UDATA restoreSystemStackPointerState(J9VMThread* vmThread, U_32 sigType, void* s
    return J9PORT_SIG_EXCEPTION_CONTINUE_EXECUTION;
    }
 
+#ifdef EMULATE_ZNEXT
+extern int jitS390Emulation(J9VMThread* vmThread, void* sigInfo);
+#endif
+
 UDATA jit390Handler(J9VMThread* vmThread, U_32 sigType, void* sigInfo)
    {
+#ifdef EMULATE_ZNEXT
+   if (J9PORT_SIG_FLAG_SIGILL == sigType && jitS390Emulation(vmThread, sigInfo) == 0)
+      {
+      return J9PORT_SIG_EXCEPTION_CONTINUE_EXECUTION;
+      }
+#endif
+
    PORT_ACCESS_FROM_VMC(vmThread);
 
    J9JITConfig *jitConfig = vmThread->javaVM->jitConfig;
@@ -1251,13 +1262,13 @@ UDATA jit390Handler(J9VMThread* vmThread, U_32 sigType, void* sigInfo)
                /* add one to *controlPC for symmetry with IA32, handler check subs one */
                jit390SetTrapHandler(controlPC, entryPointRegister, (void *) &jitHandleNullPointerExceptionTrap);
                return restoreSystemStackPointerState(vmThread, sigType, sigInfo);
-   
+
             case TRAP_TYPE_INTERNAL_ERROR:
                vmThread->jitException = (J9Object *) (controlPCValue + 1);
                /* add one to *controlPC for symmetry with IA32, handler check subs one */
                jit390SetTrapHandler(controlPC, entryPointRegister, (void *) &jitHandleInternalErrorTrap);
                return restoreSystemStackPointerState(vmThread, sigType, sigInfo);
-            
+
             case TRAP_TYPE_ARRAY_BOUNDS:
                vmThread->jitException = (J9Object *) (controlPCValue + 1);
                /* add one to *controlPC for symmetry with IA32, handler check subs one */