Fix genPutArgStkFieldList for SIMD12 (#88920)

src/coreclr/jit/codegenarm64.cpp

@@ -5313,14 +5313,11 @@ void CodeGen::genStoreLclTypeSimd12(GenTreeLclVarCommon* treeNode)
     {
         // Simply use mov if we move a SIMD12 reg to another SIMD12 reg
         assert(GetEmitter()->isVectorRegister(tgtReg));
-
         inst_Mov(treeNode->TypeGet(), tgtReg, dataReg, /* canSkip */ true);
     }
     else
     {
-        // Need an additional integer register to extract upper 4 bytes from data.
-        regNumber tmpReg = treeNode->GetSingleTempReg();
-        GetEmitter()->emitStoreSimd12ToLclOffset(varNum, offs, dataReg, tmpReg);
+        GetEmitter()->emitStoreSimd12ToLclOffset(varNum, offs, dataReg, treeNode);
     }
     genUpdateLifeStore(treeNode, tgtReg, varDsc);
 }

src/coreclr/jit/codegenarmarch.cpp

@@ -815,10 +815,9 @@ void CodeGen::genPutArgStk(GenTreePutArgStk* treeNode)
             assert(compMacOsArm64Abi() || treeNode->GetStackByteSize() % TARGET_POINTER_SIZE == 0);
 
 #ifdef TARGET_ARM64
-            if (compMacOsArm64Abi() && (treeNode->GetStackByteSize() == 12))
+            if (treeNode->GetStackByteSize() == 12)
             {
-                regNumber tmpReg = treeNode->GetSingleTempReg();
-                GetEmitter()->emitStoreSimd12ToLclOffset(varNumOut, argOffsetOut, srcReg, tmpReg);
+                GetEmitter()->emitStoreSimd12ToLclOffset(varNumOut, argOffsetOut, srcReg, treeNode);
                 argOffsetOut += 12;
             }
             else

src/coreclr/jit/codegenlinear.cpp

@@ -1868,13 +1868,10 @@ void CodeGen::genPutArgStkFieldList(GenTreePutArgStk* putArgStk, unsigned outArg
 // Emit store instructions to store the registers produced by the GT_FIELD_LIST into the outgoing
 // argument area.
 
-#if defined(FEATURE_SIMD) && defined(TARGET_ARM64)
-        // storing of TYP_SIMD12 (i.e. Vector3) argument.
-        if (compMacOsArm64Abi() && (type == TYP_SIMD12))
+#if defined(FEATURE_SIMD)
+        if (type == TYP_SIMD12)
         {
-            // Need an additional integer register to extract upper 4 bytes from data.
-            regNumber tmpReg = nextArgNode->GetSingleTempReg();
-            GetEmitter()->emitStoreSimd12ToLclOffset(outArgVarNum, thisFieldOffset, reg, tmpReg);
+            GetEmitter()->emitStoreSimd12ToLclOffset(outArgVarNum, thisFieldOffset, reg, nextArgNode);
         }
         else
 #endif // FEATURE_SIMD

src/coreclr/jit/emit.h

@@ -2714,6 +2714,10 @@ class emitter
     void emitMarkStackLvl(unsigned stackLevel);
 #endif
 
+#if defined(FEATURE_SIMD)
+    void emitStoreSimd12ToLclOffset(unsigned varNum, unsigned offset, regNumber dataReg, GenTree* tmpRegProvider);
+#endif // FEATURE_SIMD
+
     int emitNextRandomNop();
 
     //

src/coreclr/jit/emitarm64.cpp

@@ -16767,4 +16767,32 @@ bool emitter::IsOptimizableLdrToMov(
 
     return true;
 }
+
+#if defined(FEATURE_SIMD)
+//-----------------------------------------------------------------------------------
+// emitStoreSimd12ToLclOffset: store SIMD12 value from dataReg to varNum+offset.
+//
+// Arguments:
+//     varNum         - the variable on the stack to use as a base;
+//     offset         - the offset from the varNum;
+//     dataReg        - the src reg with SIMD12 value;
+//     tmpRegProvider - a tree to grab a tmp reg from if needed.
+//
+void emitter::emitStoreSimd12ToLclOffset(unsigned varNum, unsigned offset, regNumber dataReg, GenTree* tmpRegProvider)
+{
+    assert(varNum != BAD_VAR_NUM);
+    assert(isVectorRegister(dataReg));
+
+    // store lower 8 bytes
+    emitIns_S_R(INS_str, EA_8BYTE, dataReg, varNum, offset);
+
+    // Extract upper 4-bytes from data
+    regNumber tmpReg = tmpRegProvider->GetSingleTempReg();
+    emitIns_R_R_I(INS_mov, EA_4BYTE, tmpReg, dataReg, 2);
+
+    // 4-byte write
+    emitIns_S_R(INS_str, EA_4BYTE, tmpReg, varNum, offset + 8);
+}
+#endif // FEATURE_SIMD
+
 #endif // defined(TARGET_ARM64)

src/coreclr/jit/emitarm64.h

@@ -1003,30 +1003,4 @@ inline bool emitIsLoadConstant(instrDesc* jmp)
             (jmp->idInsFmt() == IF_LARGELDC));
 }
 
-#if defined(FEATURE_SIMD)
-//-----------------------------------------------------------------------------------
-// emitStoreSimd12ToLclOffset: store SIMD12 value from dataReg to varNum+offset.
-//
-// Arguments:
-//     varNum  - the variable on the stack to use as a base;
-//     offset  - the offset from the varNum;
-//     dataReg - the src reg with SIMD12 value;
-//     tmpReg  - a tmp reg to use for the write, can be general or float.
-//
-void emitStoreSimd12ToLclOffset(unsigned varNum, unsigned offset, regNumber dataReg, regNumber tmpReg)
-{
-    assert(varNum != BAD_VAR_NUM);
-    assert(isVectorRegister(dataReg));
-
-    // store lower 8 bytes
-    emitIns_S_R(INS_str, EA_8BYTE, dataReg, varNum, offset);
-
-    // Extract upper 4-bytes from data
-    emitIns_R_R_I(INS_mov, EA_4BYTE, tmpReg, dataReg, 2);
-
-    // 4-byte write
-    emitIns_S_R(INS_str, EA_4BYTE, tmpReg, varNum, offset + 8);
-}
-#endif // FEATURE_SIMD
-
 #endif // TARGET_ARM64

src/coreclr/jit/emitxarch.cpp

@@ -5648,6 +5648,43 @@ void emitter::emitIns_R(instruction ins, emitAttr attr, regNumber reg)
     emitAdjustStackDepthPushPop(ins);
 }
 
+#if defined(FEATURE_SIMD)
+//-----------------------------------------------------------------------------------
+// emitStoreSimd12ToLclOffset: store SIMD12 value from dataReg to varNum+offset.
+//
+// Arguments:
+//     varNum         - the variable on the stack to use as a base;
+//     offset         - the offset from the varNum;
+//     dataReg        - the src reg with SIMD12 value;
+//     tmpRegProvider - a tree to grab a tmp reg from if needed.
+//
+void emitter::emitStoreSimd12ToLclOffset(unsigned varNum, unsigned offset, regNumber dataReg, GenTree* tmpRegProvider)
+{
+    assert(varNum != BAD_VAR_NUM);
+    assert(isFloatReg(dataReg));
+
+    // Store lower 8 bytes
+    emitIns_S_R(INS_movsd_simd, EA_8BYTE, dataReg, varNum, offset);
+
+    if (emitComp->compOpportunisticallyDependsOn(InstructionSet_SSE41))
+    {
+        // Extract and store upper 4 bytes
+        emitIns_S_R_I(INS_extractps, EA_16BYTE, varNum, offset + 8, dataReg, 2);
+    }
+    else
+    {
+        regNumber tmpReg = tmpRegProvider->GetSingleTempReg();
+        assert(isFloatReg(tmpReg));
+
+        // Extract upper 4 bytes from data
+        emitIns_R_R(INS_movhlps, EA_16BYTE, tmpReg, dataReg);
+
+        // Store upper 4 bytes
+        emitIns_S_R(INS_movss, EA_4BYTE, tmpReg, varNum, offset + 8);
+    }
+}
+#endif // FEATURE_SIMD
+
 /*****************************************************************************
  *
  *  Add an instruction referencing a register and a constant.

src/coreclr/jit/lsraarmarch.cpp

@@ -425,15 +425,11 @@ int LinearScan::BuildPutArgStk(GenTreePutArgStk* argNode)
                 srcCount++;
 
 #if defined(FEATURE_SIMD)
-                if (compMacOsArm64Abi())
+                if (use.GetType() == TYP_SIMD12)
                 {
-                    if (use.GetType() == TYP_SIMD12)
-                    {
-                        // Vector3 is read/written as two reads/writes: 8 byte and 4 byte.
-                        // To assemble the vector properly we would need an additional int register.
-                        // The other platforms can write it as 16-byte using 1 write.
-                        buildInternalIntRegisterDefForNode(use.GetNode());
-                    }
+                    // Vector3 is read/written as two reads/writes: 8 byte and 4 byte.
+                    // To assemble the vector properly we would need an additional int register.
+                    buildInternalIntRegisterDefForNode(use.GetNode());
                 }
 #endif // FEATURE_SIMD
             }

src/coreclr/jit/lsraxarch.cpp

@@ -1651,12 +1651,22 @@ int LinearScan::BuildPutArgStk(GenTreePutArgStk* putArgStk)
 #endif // TARGET_X86
 
 #if defined(FEATURE_SIMD)
-            // Note that we need to check the field type, not the type of the node. This is because the
-            // field type will be TYP_SIMD12 whereas the node type might be TYP_SIMD16 for lclVar, where
-            // we "round up" to 16.
-            if ((fieldType == TYP_SIMD12) && (simdTemp == nullptr))
+            if (fieldType == TYP_SIMD12)
             {
-                simdTemp = buildInternalFloatRegisterDefForNode(putArgStk);
+                // Note that we need to check the field type, not the type of the node. This is because the
+                // field type will be TYP_SIMD12 whereas the node type might be TYP_SIMD16 for lclVar, where
+                // we "round up" to 16.
+                if (simdTemp == nullptr)
+                {
+                    simdTemp = buildInternalFloatRegisterDefForNode(putArgStk);
+                }
+
+                if (!compiler->compOpportunisticallyDependsOn(InstructionSet_SSE41))
+                {
+                    // To store SIMD12 without SSE4.1 (extractps) we will need
+                    // a temp xmm reg to do the shuffle.
+                    buildInternalFloatRegisterDefForNode(use.GetNode());
+                }
             }
 #endif // defined(FEATURE_SIMD)
 

src/tests/JIT/Regression/JitBlue/Runtime_79118/Runtime_79118.cs

@@ -0,0 +1,67 @@
+// Licensed to the .NET Foundation under one or more agreements.
+// The .NET Foundation licenses this file to you under the MIT license.
+
+using System.Collections.Generic;
+using System.Numerics;
+using System.Runtime.CompilerServices;
+using Xunit;
+
+public readonly struct Ray
+{
+    public readonly Vector3 Origin;
+    public readonly Vector3 Direction;
+
+    public Ray(Vector3 origin, Vector3 direction)
+    {
+        Origin = origin;
+        Direction = direction;
+    }
+}
+
+public class RayTracer
+{
+    private static IEnumerable<object> hittables;
+
+    static RayTracer()
+    {
+        var list = new List<object>();
+        list.Add(new object());
+        hittables = list;
+    }
+
+    [Fact]
+    public static int TestEntryPoint()
+    {
+        Ray r = GetRay();
+        Consume(r.Direction);
+        traceRay(r);
+        return 100;
+    }
+
+    [MethodImpl(MethodImplOptions.NoInlining)]
+    private static void Consume(object o)
+    {
+        var _ = o.ToString();
+    }
+
+    private static Ray GetRay()
+    {
+        return new Ray(Vector3.Zero, -Vector3.UnitY);
+    }
+
+    private static Vector3 traceRay(Ray r)
+    {
+        foreach (object h in hittables)
+        {
+            TestHit(r);
+            return Vector3.One;
+        }
+        return Vector3.Zero;
+    }
+
+    [MethodImpl(MethodImplOptions.NoInlining)]
+    private static void TestHit(Ray ray)
+    {
+        return;
+    }
+}

src/tests/JIT/Regression/JitBlue/Runtime_79118/Runtime_79118.csproj

@@ -0,0 +1,8 @@
+<Project Sdk="Microsoft.NET.Sdk">
+  <PropertyGroup>
+    <Optimize>True</Optimize>
+  </PropertyGroup>
+  <ItemGroup>
+    <Compile Include="$(MSBuildProjectName).cs" />
+  </ItemGroup>
+</Project>