arm64: Additions with GetActiveElementCount() should use incp when given the same predicate

src/coreclr/jit/codegenarm64.cpp

@@ -2636,6 +2636,17 @@ void CodeGen::genCodeForBinary(GenTreeOp* tree)
         emit->emitIns_Add_Add_Tls_Reloc(attr, targetReg, op1->GetRegNum(), op2->AsIntCon()->IconValue());
         return;
     }
+    else if (oper == GT_ADD && op1->isContained() && op1->OperIsHWIntrinsic(NI_Sve_GetActiveElementCount))
+    {
+        const GenTreeHWIntrinsic* elementCountNode      = op1->AsHWIntrinsic();
+        GenTree*                  elementCountChildNode = elementCountNode->Op(1);
+
+        regNumber reg1 = op2->GetRegNum();
+        regNumber reg2 = elementCountChildNode->GetRegNum();
+
+        emit->emitInsSve_R_R(INS_sve_incp, EA_8BYTE, reg1, reg2, INS_OPTS_SCALABLE_B);
+        return;
+    }
 
     instruction ins = genGetInsForOper(tree->OperGet(), targetType);
 

src/coreclr/jit/gentree.cpp

@@ -20494,7 +20494,8 @@ bool GenTree::isContainableHWIntrinsic() const
         }
     }
 #elif defined(TARGET_ARM64)
-    return (AsHWIntrinsic()->GetHWIntrinsicId() == NI_Sve_ConditionalSelect);
+    return (AsHWIntrinsic()->GetHWIntrinsicId() == NI_Sve_ConditionalSelect ||
+            AsHWIntrinsic()->GetHWIntrinsicId() == NI_Sve_GetActiveElementCount);
 #else
     return false;
 #endif // TARGET_XARCH

src/coreclr/jit/lower.cpp

@@ -7299,6 +7299,11 @@ GenTree* Lowering::LowerAdd(GenTreeOp* node)
         {
             return next;
         }
+
+        if (TryContainingGetActiveElementCount(node))
+        {
+            return next;
+        }
     }
 #endif // TARGET_ARM64
 

src/coreclr/jit/lower.h

@@ -109,6 +109,7 @@ class Lowering final : public Phase
     bool      TryLowerAddSubToMulLongOp(GenTreeOp* op, GenTree** next);
     bool      TryLowerNegToMulLongOp(GenTreeOp* op, GenTree** next);
     bool      TryContainingCselOp(GenTreeHWIntrinsic* parentNode, GenTreeHWIntrinsic* childNode);
+    bool      TryContainingGetActiveElementCount(GenTreeOp* node);
 #endif
 #ifdef TARGET_RISCV64
     bool TryLowerShiftAddToShxadd(GenTreeOp* tree, GenTree** next);

src/coreclr/jit/lowerarmarch.cpp

@@ -3738,6 +3738,31 @@ bool Lowering::TryContainingCselOp(GenTreeHWIntrinsic* parentNode, GenTreeHWIntr
     return canContain;
 }
 
+bool Lowering::TryContainingGetActiveElementCount(GenTreeOp* node)
+{
+    assert(node->OperIs(GT_ADD));
+
+    GenTree* op1 = node->gtGetOp1();
+    GenTree* op2 = node->gtGetOp2();
+    if (!op1->OperIsHWIntrinsic(NI_Sve_GetActiveElementCount))
+    {
+        return false;
+    }
+
+    assert(op1->canBeContained());
+
+    const GenTreeHWIntrinsic* elementCountNode = op1->AsHWIntrinsic();
+    GenTree*                  mask1            = elementCountNode->Op(1)->AsHWIntrinsic()->Op(2);
+    GenTree*                  mask2            = elementCountNode->Op(2)->AsHWIntrinsic()->Op(2);
+    if (mask1->GetVN(VNK_Liberal) != mask2->GetVN(VNK_Liberal))
+    {
+        return false;
+    }
+
+    MakeSrcContained(node, op1);
+    return true;
+}
+
 #endif // TARGET_ARM64
 
 //------------------------------------------------------------------------

src/tests/JIT/opt/SVE/active_element_count.cs

@@ -0,0 +1,85 @@
+// Licensed to the .NET Foundation under one or more agreements.
+// The .NET Foundation licenses this file to you under the MIT license.
+
+using System;
+using System.Numerics;
+using System.Runtime.CompilerServices;
+using System.Runtime.Intrinsics;
+using System.Runtime.Intrinsics.Arm;
+using Xunit;
+
+public class TestGetActiveElementCount
+{
+    [Fact]
+    public static int TestEntryPoint()
+    {
+        if (Sve.IsSupported)
+        {
+            bool fail = false;
+
+            Vector<byte> v0 = Vector.Create<byte>(1);
+            Vector<byte> v1 = Vector.Create<byte>(4);
+            if (DifferentVector(v0, v1, 0) != 16)
+            {
+                fail = true;
+            }
+
+            if (SameVector(v0, 0) != 16)
+            {
+                fail = true;
+            }
+
+            Vector<double> vDouble0 = Vector.Create<double>(1);
+            Vector<double> vDouble1 = Vector.Create<double>(4);
+            if (DifferentVectorDouble(vDouble0, vDouble1, 0) != 2)
+            {
+                fail = true;
+            }
+
+            if (SameVectorDouble(vDouble0, 0) != 2)
+            {
+                fail = true;
+            }
+
+            if (fail)
+            {
+                return 101;
+            }
+        }
+        return 100;
+    }
+
+    [MethodImpl(MethodImplOptions.NoInlining)]
+    private static ulong DifferentVector(Vector<byte> v0, Vector<byte> v1, ulong total)
+    {
+        //ARM64-FULL-LINE: cntp {{x[0-9]+}}, {{p[0-9]+}}, {{p[0-9]+}}.b
+        //ARM64-FULL-LINE: add {{x[0-9]+}}, {{x[0-9]+}}, {{x[0-9]+}}
+        total += Sve.GetActiveElementCount(v0, v1);
+        return total;
+    }
+
+    [MethodImpl(MethodImplOptions.NoInlining)]
+    private static ulong SameVector(Vector<byte> v0, ulong total)
+    {
+        //ARM64-FULL-LINE: incp {{x[0-9]+}}, {{p[0-9]+}}.b
+        total += Sve.GetActiveElementCount(v0, v0);
+        return total;
+    }
+
+    [MethodImpl(MethodImplOptions.NoInlining)]
+    private static ulong DifferentVectorDouble(Vector<double> v0, Vector<double> v1, ulong total)
+    {
+        //ARM64-FULL-LINE: cntp {{x[0-9]+}}, {{p[0-9]+}}, {{p[0-9]+}}.b
+        //ARM64-FULL-LINE: add {{x[0-9]+}}, {{x[0-9]+}}, {{x[0-9]+}}
+        total += Sve.GetActiveElementCount(v0, v1);
+        return total;
+    }
+
+    [MethodImpl(MethodImplOptions.NoInlining)]
+    private static ulong SameVectorDouble(Vector<double> v0, ulong total)
+    {
+        //ARM64-FULL-LINE: incp {{x[0-9]+}}, {{p[0-9]+}}.b
+        total += Sve.GetActiveElementCount(v0, v0);
+        return total;
+    }
+}

src/tests/JIT/opt/SVE/active_element_count.csproj

@@ -0,0 +1,10 @@
+<Project Sdk="Microsoft.NET.Sdk">
+  <PropertyGroup>
+    <RequiresProcessIsolation>true</RequiresProcessIsolation>
+    <Optimize>True</Optimize>
+    <NoWarn>$(NoWarn),SYSLIB5003</NoWarn>
+  </PropertyGroup>
+  <ItemGroup>
+    <Compile Include="$(MSBuildProjectName).cs" />
+  </ItemGroup>
+</Project>