Add Tbl_V Sse opt. with Tests. (#651)

* Add v4, v5, v30, v31 required for Tbl_V Tests.

* Add Tests for Tbl_V.

* Add Tbl_V Sse opt..

* Nit.

* Small opt. on comparison constant vector.

* Nit.

* Add EmitLd/Stvectmp2/3.

* Nit.

Instructions/InstEmitSimdMove.cs

@@ -355,11 +355,69 @@ namespace ChocolArm64.Instructions
         {
             OpCodeSimdTbl64 op = (OpCodeSimdTbl64)context.CurrOp;
 
+            if (Optimizations.UseSsse3)
+            {
+                Type[] typesCmpSflSub = new Type[] { typeof(Vector128<sbyte>), typeof(Vector128<sbyte>) };
+                Type[] typesOr        = new Type[] { typeof(Vector128<long> ), typeof(Vector128<long> ) };
+                Type[] typesSav       = new Type[] { typeof(long) };
+
+                context.EmitLdvec(op.Rn);
+                context.EmitLdvec(op.Rm);
+
+                context.EmitLdc_I8(0x0F0F0F0F0F0F0F0FL);
+                context.EmitCall(typeof(Sse2).GetMethod(nameof(Sse2.SetAllVector128), typesSav));
+
+                context.EmitStvectmp2();
+                context.EmitLdvectmp2();
+
+                context.EmitCall(typeof(Sse2).GetMethod(nameof(Sse2.CompareGreaterThan), typesCmpSflSub));
+
+                context.EmitLdvec(op.Rm);
+
+                context.EmitCall(typeof(Sse2).GetMethod(nameof(Sse2.Or), typesOr));
+
+                context.EmitCall(typeof(Ssse3).GetMethod(nameof(Ssse3.Shuffle), typesCmpSflSub));
+
+                for (int index = 1; index < op.Size; index++)
+                {
+                    context.EmitLdvec((op.Rn + index) & 0x1F);
+                    context.EmitLdvec(op.Rm);
+
+                    context.EmitLdc_I8(0x1010101010101010L * index);
+                    context.EmitCall(typeof(Sse2).GetMethod(nameof(Sse2.SetAllVector128), typesSav));
+
+                    context.EmitCall(typeof(Sse2).GetMethod(nameof(Sse2.Subtract), typesCmpSflSub));
+
+                    context.EmitStvectmp();
+                    context.EmitLdvectmp();
+
+                    context.EmitLdvectmp2();
+
+                    context.EmitCall(typeof(Sse2).GetMethod(nameof(Sse2.CompareGreaterThan), typesCmpSflSub));
+
+                    context.EmitLdvectmp();
+
+                    context.EmitCall(typeof(Sse2).GetMethod(nameof(Sse2.Or), typesOr));
+
+                    context.EmitCall(typeof(Ssse3).GetMethod(nameof(Ssse3.Shuffle), typesCmpSflSub));
+
+                    context.EmitCall(typeof(Sse2).GetMethod(nameof(Sse2.Or), typesOr));
+                }
+
+                context.EmitStvec(op.Rd);
+
+                if (op.RegisterSize == RegisterSize.Simd64)
+                {
+                    EmitVectorZeroUpper(context, op.Rd);
+                }
+            }
+            else
+            {
                 context.EmitLdvec(op.Rm);
 
                 for (int index = 0; index < op.Size; index++)
                 {
-                context.EmitLdvec((op.Rn + index) & 0x1f);
+                    context.EmitLdvec((op.Rn + index) & 0x1F);
                 }
 
                 switch (op.Size)
@@ -385,6 +443,7 @@ namespace ChocolArm64.Instructions
 
                 context.EmitStvec(op.Rd);
             }
+        }
 
         public static void Trn1_V(ILEmitterCtx context)
         {

Translation/ILEmitterCtx.cs

@@ -61,7 +61,9 @@ namespace ChocolArm64.Translation
 
         //Vectors are part of another "set" of locals.
         private const int VecGpTmp1Index   = ReservedLocalsCount + 0;
-        private const int UserVecTempStart = ReservedLocalsCount + 1;
+        private const int VecGpTmp2Index   = ReservedLocalsCount + 1;
+        private const int VecGpTmp3Index   = ReservedLocalsCount + 2;
+        private const int UserVecTempStart = ReservedLocalsCount + 3;
 
         private static int _userIntTempCount;
         private static int _userVecTempCount;
@@ -629,6 +631,12 @@ namespace ChocolArm64.Translation
         public void EmitLdvectmp() => EmitLdvec(VecGpTmp1Index);
         public void EmitStvectmp() => EmitStvec(VecGpTmp1Index);
 
+        public void EmitLdvectmp2() => EmitLdvec(VecGpTmp2Index);
+        public void EmitStvectmp2() => EmitStvec(VecGpTmp2Index);
+
+        public void EmitLdvectmp3() => EmitLdvec(VecGpTmp3Index);
+        public void EmitStvectmp3() => EmitStvec(VecGpTmp3Index);
+
         public void EmitLdint(int index) => Ldloc(index, VarType.Int);
         public void EmitStint(int index) => Stloc(index, VarType.Int);