Optimize ToVector128, ToVector128Unsafe and Vector128.GetLower() (dot…

…net#36732) * Optimize ToVector128 , ToVector128Unsafe and Vector128.GetLower()
just607 · May 20, 2020 · 34fd21e · 34fd21e
1 parent 9ceeb18
commit 34fd21e
Show file tree

Hide file tree

Showing 5 changed files with 19 additions and 1 deletion.
diff --git a/src/coreclr/src/jit/hwintrinsic.cpp b/src/coreclr/src/jit/hwintrinsic.cpp
@@ -610,7 +610,7 @@ static bool isSupportedBaseType(NamedIntrinsic intrinsic, var_types baseType)
            (intrinsic >= NI_Vector256_get_AllBitsSet && intrinsic <= NI_Vector256_ToScalar));
 #else
     assert((intrinsic >= NI_Vector64_AsByte && intrinsic <= NI_Vector64_AsUInt32) ||
-           (intrinsic >= NI_Vector64_get_AllBitsSet && intrinsic <= NI_Vector64_ToScalar) ||
+           (intrinsic >= NI_Vector64_get_AllBitsSet && intrinsic <= NI_Vector64_ToVector128Unsafe) ||
            (intrinsic >= NI_Vector128_As && intrinsic <= NI_Vector128_AsUInt64) ||
            (intrinsic >= NI_Vector128_get_AllBitsSet && intrinsic <= NI_Vector128_ToScalar));
 #endif

diff --git a/src/coreclr/src/jit/hwintrinsiccodegenarm64.cpp b/src/coreclr/src/jit/hwintrinsiccodegenarm64.cpp
@@ -617,6 +617,18 @@ void CodeGen::genHWIntrinsic(GenTreeHWIntrinsic* node)
             }
             break;
 
+            case NI_Vector64_ToVector128:
+                GetEmitter()->emitIns_R_R(ins, emitSize, targetReg, op1Reg);
+                break;
+
+            case NI_Vector64_ToVector128Unsafe:
+            case NI_Vector128_GetLower:
+                if (op1Reg != targetReg)
+                {
+                    GetEmitter()->emitIns_R_R(ins, emitSize, targetReg, op1Reg);
+                }
+                break;
+
             case NI_Vector64_GetElement:
             case NI_Vector128_GetElement:
             case NI_Vector64_ToScalar:

diff --git a/src/coreclr/src/jit/hwintrinsiclistarm64.h b/src/coreclr/src/jit/hwintrinsiclistarm64.h
@@ -30,6 +30,8 @@ HARDWARE_INTRINSIC(Vector64,        get_Count,
 HARDWARE_INTRINSIC(Vector64,        get_Zero,                                   8,           0,     {INS_movi,              INS_movi,           INS_movi,           INS_movi,           INS_movi,           INS_movi,           INS_movi,           INS_movi,           INS_movi,           INS_movi},              HW_Category_Helper,                 HW_Flag_SpecialImport|HW_Flag_SpecialCodeGen)
 HARDWARE_INTRINSIC(Vector64,        GetElement,                                 8,           2,     {INS_smov,              INS_umov,           INS_smov,           INS_umov,           INS_smov,           INS_umov,           INS_umov,           INS_umov,           INS_dup,            INS_dup},               HW_Category_IMM,                    HW_Flag_NoJmpTableIMM|HW_Flag_SupportsContainment|HW_Flag_BaseTypeFromFirstArg|HW_Flag_SpecialCodeGen)
 HARDWARE_INTRINSIC(Vector64,        ToScalar,                                   8,           1,     {INS_smov,              INS_umov,           INS_smov,           INS_umov,           INS_smov,           INS_umov,           INS_umov,           INS_umov,           INS_dup,            INS_dup},               HW_Category_SIMDScalar,             HW_Flag_SpecialCodeGen|HW_Flag_BaseTypeFromFirstArg)
+HARDWARE_INTRINSIC(Vector64,        ToVector128,                                8,           1,     {INS_mov,               INS_mov,            INS_mov,            INS_mov,            INS_mov,            INS_mov,            INS_mov,            INS_mov,            INS_mov,            INS_mov},               HW_Category_SimpleSIMD,             HW_Flag_SpecialCodeGen)
+HARDWARE_INTRINSIC(Vector64,        ToVector128Unsafe,                          8,           1,     {INS_mov,               INS_mov,            INS_mov,            INS_mov,            INS_mov,            INS_mov,            INS_mov,            INS_mov,            INS_mov,            INS_mov},               HW_Category_SimpleSIMD,             HW_Flag_SpecialCodeGen)
 
 // ***************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************
 //                 ISA              Function name                            SIMD size       NumArg                                                                                                     Instructions                                                                                                     Category                            Flags
@@ -53,6 +55,7 @@ HARDWARE_INTRINSIC(Vector128,       get_AllBitsSet,                            1
 HARDWARE_INTRINSIC(Vector128,       get_Count,                                 16,           0,     {INS_invalid,           INS_invalid,        INS_invalid,        INS_invalid,        INS_invalid,        INS_invalid,        INS_invalid,        INS_invalid,        INS_invalid,        INS_invalid},           HW_Category_Helper,                 HW_Flag_SpecialImport|HW_Flag_NoCodeGen)
 HARDWARE_INTRINSIC(Vector128,       get_Zero,                                  16,           0,     {INS_movi,              INS_movi,           INS_movi,           INS_movi,           INS_movi,           INS_movi,           INS_movi,           INS_movi,           INS_movi,           INS_movi},              HW_Category_Helper,                 HW_Flag_SpecialImport|HW_Flag_SpecialCodeGen)
 HARDWARE_INTRINSIC(Vector128,       GetElement,                                16,           2,     {INS_smov,              INS_umov,           INS_smov,           INS_umov,           INS_smov,           INS_umov,           INS_umov,           INS_umov,           INS_dup,            INS_dup},               HW_Category_IMM,                    HW_Flag_NoJmpTableIMM|HW_Flag_SupportsContainment|HW_Flag_BaseTypeFromFirstArg|HW_Flag_SpecialCodeGen)
+HARDWARE_INTRINSIC(Vector128,       GetLower,                                  16,           1,     {INS_mov,               INS_mov,            INS_mov,            INS_mov,            INS_mov,            INS_mov,            INS_mov,            INS_mov,            INS_mov,            INS_mov},               HW_Category_SimpleSIMD,             HW_Flag_SpecialCodeGen)
 HARDWARE_INTRINSIC(Vector128,       ToScalar,                                  16,           1,     {INS_smov,              INS_umov,           INS_smov,           INS_umov,           INS_smov,           INS_umov,           INS_umov,           INS_umov,           INS_dup,            INS_dup},               HW_Category_SIMDScalar,             HW_Flag_SpecialCodeGen|HW_Flag_BaseTypeFromFirstArg)
 
 // ***************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************

diff --git a/src/libraries/System.Private.CoreLib/src/System/Runtime/Intrinsics/Vector128.cs b/src/libraries/System.Private.CoreLib/src/System/Runtime/Intrinsics/Vector128.cs
@@ -1554,6 +1554,7 @@ public static Vector128<T> WithElement<T>(this Vector128<T> vector, int index, T
         /// <param name="vector">The vector to get the lower 64-bits from.</param>
         /// <returns>The value of the lower 64-bits as a new <see cref="Vector64{T}" />.</returns>
         /// <exception cref="NotSupportedException">The type of <paramref name="vector" /> (<typeparamref name="T" />) is not supported.</exception>
+        [Intrinsic]
         public static Vector64<T> GetLower<T>(this Vector128<T> vector)
             where T : struct
         {

diff --git a/src/libraries/System.Private.CoreLib/src/System/Runtime/Intrinsics/Vector64.cs b/src/libraries/System.Private.CoreLib/src/System/Runtime/Intrinsics/Vector64.cs
@@ -942,6 +942,7 @@ public static T ToScalar<T>(this Vector64<T> vector)
         /// <param name="vector">The vector to extend.</param>
         /// <returns>A new <see cref="Vector128{T}" /> with the lower 64-bits set to the value of <paramref name="vector" /> and the upper 64-bits initialized to zero.</returns>
         /// <exception cref="NotSupportedException">The type of <paramref name="vector" /> (<typeparamref name="T" />) is not supported.</exception>
+        [Intrinsic]
         public static Vector128<T> ToVector128<T>(this Vector64<T> vector)
             where T : struct
         {
@@ -957,6 +958,7 @@ public static Vector128<T> ToVector128<T>(this Vector64<T> vector)
         /// <param name="vector">The vector to extend.</param>
         /// <returns>A new <see cref="Vector128{T}" /> with the lower 64-bits set to the value of <paramref name="vector" /> and the upper 64-bits left uninitialized.</returns>
         /// <exception cref="NotSupportedException">The type of <paramref name="vector" /> (<typeparamref name="T" />) is not supported.</exception>
+        [Intrinsic]
         public static unsafe Vector128<T> ToVector128Unsafe<T>(this Vector64<T> vector)
             where T : struct
         {