Adjust stack level while calculating offset for Vector.getElement (#38311)

* Adjust stack level

* Reenable NarrowDouble() test

* formatting

* review comments

* do not assert when isEBPbased=true

* Add test coverage

* updated the comment inside test

Author: kunalspathak

src/coreclr/src/jit/simdcodegenxarch.cpp

@@ -2144,6 +2144,17 @@ void CodeGen::genSIMDIntrinsicGetItem(GenTreeSIMD* simdNode)
             bool     isEBPbased;
             unsigned varNum = op1->AsLclVarCommon()->GetLclNum();
             offset += compiler->lvaFrameAddress(varNum, &isEBPbased);
+
+#if !FEATURE_FIXED_OUT_ARGS
+            if (!isEBPbased)
+            {
+                // Adjust the offset by the amount currently pushed on the CPU stack
+                offset += genStackLevel;
+            }
+#else
+            assert(genStackLevel == 0);
+#endif // !FEATURE_FIXED_OUT_ARGS
+
             if (op1->OperGet() == GT_LCL_FLD)
             {
                 offset += op1->AsLclFld()->GetLclOffs();
@@ -2192,8 +2203,19 @@ void CodeGen::genSIMDIntrinsicGetItem(GenTreeSIMD* simdNode)
     {
         unsigned simdInitTempVarNum = compiler->lvaSIMDInitTempVarNum;
         noway_assert(simdInitTempVarNum != BAD_VAR_NUM);
-        bool      isEBPbased;
-        unsigned  offs     = compiler->lvaFrameAddress(simdInitTempVarNum, &isEBPbased);
+        bool     isEBPbased;
+        unsigned offs = compiler->lvaFrameAddress(simdInitTempVarNum, &isEBPbased);
+
+#if !FEATURE_FIXED_OUT_ARGS
+        if (!isEBPbased)
+        {
+            // Adjust the offset by the amount currently pushed on the CPU stack
+            offs += genStackLevel;
+        }
+#else
+        assert(genStackLevel == 0);
+#endif // !FEATURE_FIXED_OUT_ARGS
+
         regNumber indexReg = op2->GetRegNum();
 
         // Store the vector to the temp location.

src/coreclr/tests/src/JIT/Regression/JitBlue/GitHub_36614/GitHub_36614.cs

@@ -0,0 +1,55 @@
+// Licensed to the .NET Foundation under one or more agreements.
+// The .NET Foundation licenses this file to you under the MIT license.
+// See the LICENSE file in the project root for more information.
+//
+
+using System;
+using System.Numerics;
+using System.Runtime.CompilerServices;
+using Xunit;
+
+namespace projs
+{
+    public class Program
+    {
+        public static int Main(string[] args)
+        {
+            Program p = new Program();
+            p.NarrowDouble();
+            Console.WriteLine("Passed.");
+            return 100;
+        }
+
+        [MethodImpl(MethodImplOptions.NoInlining)]
+        public void NarrowDouble()
+        {
+            // GenerateSource1 and GenerateSource2 methods are needed to exercise the bug code path.
+            double[] source1 = GenerateSource1();
+            double[] source2 = GenerateSource2();
+            Vector<double> sourceVec1 = new Vector<double>(source1);
+            Vector<double> sourceVec2 = new Vector<double>(source2);
+            Vector<float> dest = Vector.Narrow(sourceVec1, sourceVec2);
+
+            for (int i = 0; i < Vector<double>.Count; i++)
+            {
+                Assert.Equal(unchecked((float)source1[i]), dest[i]);
+            }
+            for (int i = 0; i < Vector<double>.Count; i++)
+            {
+                // The test represents a problem on x86 when double-alignment causes access to stack local.
+                // At that time, the stack level was not included in the offset while trying to fetch the element
+                // of dest present on stack.
+                Assert.Equal(unchecked((float)source2[i]), dest[i + Vector<double>.Count]);
+            }
+        }
+
+        internal static double[] GenerateSource1()
+        {
+            return new double[4] { 5.1, 5.2, 5.3, 5.4 };
+        }
+        internal static double[] GenerateSource2()
+        {
+            return new double[4] { 6.1, 6.2, 6.3, 6.4 };
+        }
+    }
+}

src/coreclr/tests/src/JIT/Regression/JitBlue/GitHub_36614/GitHub_36614.csproj

@@ -0,0 +1,23 @@
+<Project Sdk="Microsoft.NET.Sdk">
+  <PropertyGroup>
+    <OutputType>Exe</OutputType>
+  </PropertyGroup>
+  <PropertyGroup>
+    <DebugType />
+    <Optimize>True</Optimize>
+  </PropertyGroup>
+  <ItemGroup>
+    <Compile Include="$(MSBuildProjectName).cs" />
+  </ItemGroup>
+  <PropertyGroup>
+    <!-- Set JitStreess variables, Linux requires them to be properly capitalized. -->
+    <CLRTestBatchPreCommands><![CDATA[
+$(CLRTestBatchPreCommands)
+set COMPlus_JitStressModeNames=STRESS_RANDOM_INLINE STRESS_DBL_ALN
+]]></CLRTestBatchPreCommands>
+    <BashCLRTestPreCommands><![CDATA[
+$(BashCLRTestPreCommands)
+export COMPlus_JitStressModeNames=STRESS_RANDOM_INLINE STRESS_DBL_ALN
+]]></BashCLRTestPreCommands>
+  </PropertyGroup>
+</Project>

src/libraries/System.Numerics.Vectors/tests/GenericVectorTests.cs

@@ -3031,7 +3031,6 @@ public void NarrowInt64()
         }
 
         [Fact]
-        [SkipOnCoreClr("https://github.com/dotnet/runtime/issues/36614", RuntimeTestModes.JitStress)]
         public void NarrowDouble()
         {
             double[] source1 = GenerateRandomValuesForVector<double>();