Fix arguments passing in MethodDescCallSite::CallTargetWorker on wasm (#119969)

* Fix arguments passing in MethodDescCallSite::CallTargetWorker on wasm

* Disable fat tokes temporarily and update VirtualCallStubManager::Resolver

To work with protable entrypoints

* Handle instance signatures

* Feedback

* Detect intrinsic methods in classes decorated with Intrinsic attribute

* Feedback

* Feedback

* Add RhpNewVariableSizeObject allocator on wasm

* Apply suggestions from code review

Co-authored-by: Jan Kotas <[email protected]>

* Feedback

* Update src/coreclr/runtime/portable/AllocFast.cpp

Co-authored-by: Jan Kotas <[email protected]>

* Feedback

---------

Co-authored-by: Jan Kotas <[email protected]>

Author: radekdoulik

src/coreclr/runtime/portable/AllocFast.cpp

@@ -5,11 +5,17 @@
 #include <fcall.h>
 
 extern void RhExceptionHandling_FailedAllocation(MethodTable *pMT, bool isOverflow);
+EXTERN_C Object* RhpGcAlloc(MethodTable* pMT, uint32_t uFlags, uintptr_t numElements, void * pTransitionFrame);
 
-EXTERN_C FCDECL2(Object*, RhpNewVariableSizeObject, MethodTable* pMT, INT_PTR size)
+EXTERN_C FCDECL2(Object*, RhpNewVariableSizeObject, MethodTable* pMT, INT_PTR numElements)
 {
-    PORTABILITY_ASSERT("RhpNewVariableSizeObject is not yet implemented");
-    return nullptr;
+    Object* obj = RhpGcAlloc(pMT, 0, numElements, nullptr);
+    if (obj == NULL)
+    {
+        RhExceptionHandling_FailedAllocation(pMT, false /* isOverflow */);
+    }
+
+    return obj;
 }
 
 static Object* _RhpNewArrayFastCore(MethodTable* pMT, INT_PTR size)
@@ -96,8 +102,7 @@ EXTERN_C FCDECL2(Object*, RhNewString, MethodTable* pMT, INT_PTR stringLength)
     if (stringLength > MAX_STRING_LENGTH)
     {
         RhExceptionHandling_FailedAllocation(pMT, false);
-        return NULL;
     }
 
     return _RhpNewArrayFastCore(pMT, stringLength);
-}
+}

src/coreclr/tools/Common/Compiler/HardwareIntrinsicHelpers.cs

@@ -29,6 +29,7 @@ public static bool IsHardwareIntrinsic(MethodDesc method)
                 return method.Context.Target.Architecture switch
                 {
                     TargetArchitecture.ARM64 => ns == "System.Runtime.Intrinsics.Arm",
+                    TargetArchitecture.Wasm32 => ns == "System.Runtime.Intrinsics.Wasm",
                     TargetArchitecture.X64 or TargetArchitecture.X86 => ns == "System.Runtime.Intrinsics.X86",
                     _ => false,
                 };

src/coreclr/vm/callhelpers.cpp

@@ -543,7 +543,7 @@ void MethodDescCallSite::CallTargetWorker(const ARG_SLOT *pArguments, ARG_SLOT *
     callDescrData.fpReturnSize = fpReturnSize;
     callDescrData.pTarget = m_pCallTarget;
 #ifdef TARGET_WASM
-    callDescrData.nArgsSize = m_argIt.GetArgSize();
+    callDescrData.nArgsSize = nStackBytes;
 #endif // TARGET_WASM
 
     CallDescrWorkerWithHandler(&callDescrData);

src/coreclr/vm/contractimpl.h

@@ -317,7 +317,8 @@ struct DispatchToken
     {
         LIMITED_METHOD_CONTRACT;
         return typeID > MAX_TYPE_ID_SMALL
-#ifdef _DEBUG
+// WASM-TODO: fix fat tokens
+#if defined(_DEBUG) && !defined(TARGET_WASM)
             // Stress the overflow mechanism in debug builds.
             || ((typeID != TYPE_ID_THIS_CLASS) && ((typeID % 7) < 4))
 #endif

src/coreclr/vm/methodtablebuilder.cpp

@@ -1513,7 +1513,7 @@ MethodTableBuilder::BuildMethodTableThrowing(
         }
     }
 
-#if defined(TARGET_X86) || defined(TARGET_AMD64) || defined(TARGET_ARM64)
+#if defined(TARGET_X86) || defined(TARGET_AMD64) || defined(TARGET_ARM64) || defined(TARGET_WASM)
     if (bmtProp->fIsIntrinsicType && !bmtGenerics->HasInstantiation())
     {
         LPCUTF8 nameSpace;
@@ -1538,11 +1538,12 @@ MethodTableBuilder::BuildMethodTableThrowing(
             IfFailThrow(GetMDImport()->GetNameOfTypeDef(td, NULL, &nameSpace));
         }
 
+        // All the functions in System.Runtime.Intrinsics.<arch> are hardware intrinsics.
 #if defined(TARGET_ARM64)
-        // All the funtions in System.Runtime.Intrinsics.Arm are hardware intrinsics.
         if (hr == S_OK && strcmp(nameSpace, "System.Runtime.Intrinsics.Arm") == 0)
+#elif defined(TARGET_WASM)
+        if (hr == S_OK && strcmp(nameSpace, "System.Runtime.Intrinsics.Wasm") == 0)
 #else
-        // All the funtions in System.Runtime.Intrinsics.X86 are hardware intrinsics.
         if (hr == S_OK && (strcmp(nameSpace, "System.Runtime.Intrinsics.X86") == 0))
 #endif
         {

src/coreclr/vm/virtualcallstub.cpp

@@ -2318,7 +2318,7 @@ VirtualCallStubManager::Resolver(
             }
         }
 #endif // defined(LOGGING) || defined(_DEBUG)
-
+#ifndef FEATURE_PORTABLE_ENTRYPOINTS
         BOOL fSlotCallsPrestub = DoesSlotCallPrestub(implSlot.GetTarget());
         if (!fSlotCallsPrestub)
         {
@@ -2361,6 +2361,9 @@ VirtualCallStubManager::Resolver(
                 }
             }
         }
+#else
+        fShouldPatch = TRUE;
+#endif // !FEATURE_PORTABLE_ENTRYPOINTS
     }
 #ifdef FEATURE_COMINTEROP
     else if (pMT->IsComObjectType()

src/coreclr/vm/wasm/cgencpu.h

@@ -22,8 +22,8 @@
 
 inline unsigned StackElemSize(unsigned parmSize, bool isValueType = false /* unused */, bool isFloatHfa = false /* unused */)
 {
-    _ASSERTE("The function is not implemented on wasm");
-    return 0;
+    const unsigned stackSlotSize = sizeof(void*);
+    return ALIGN_UP(parmSize, stackSlotSize);
 }
 
 struct HijackArgs