From b3d4c02098f6009f2f665c2f014c5df8c9b7e92f Mon Sep 17 00:00:00 2001
From: Lioncash <mathew1800@gmail.com>
Date: Tue, 10 Apr 2018 13:36:25 -0400
Subject: [PATCH] A64: Implement SM3TT2A

---
 src/frontend/A64/decoder/a64.inc              |  2 +-
 .../impl/simd_crypto_three_register.cpp       | 42 ++++++++++++++++---
 2 files changed, 38 insertions(+), 6 deletions(-)

diff --git a/src/frontend/A64/decoder/a64.inc b/src/frontend/A64/decoder/a64.inc
index a82e252a..7554f4f0 100644
--- a/src/frontend/A64/decoder/a64.inc
+++ b/src/frontend/A64/decoder/a64.inc
@@ -854,7 +854,7 @@ INST(USHLL,                  "USHLL, USHLL2",                             "0Q101
 // Data Processing - FP and SIMD - Cryptographic three register
 INST(SM3TT1A,                "SM3TT1A",                                   "11001110010mmmmm10ii00nnnnnddddd")
 INST(SM3TT1B,                "SM3TT1B",                                   "11001110010mmmmm10ii01nnnnnddddd")
-//INST(SM3TT2A,                "SM3TT2A",                                   "11001110010mmmmm10ii10nnnnnddddd")
+INST(SM3TT2A,                "SM3TT2A",                                   "11001110010mmmmm10ii10nnnnnddddd")
 //INST(SM3TT2B,                "SM3TT2B",                                   "11001110010mmmmm10ii11nnnnnddddd")
 
 // Data Processing - FP and SIMD - SHA512 three register
diff --git a/src/frontend/A64/translate/impl/simd_crypto_three_register.cpp b/src/frontend/A64/translate/impl/simd_crypto_three_register.cpp
index 16c9a84b..bf97eba3 100644
--- a/src/frontend/A64/translate/impl/simd_crypto_three_register.cpp
+++ b/src/frontend/A64/translate/impl/simd_crypto_three_register.cpp
@@ -8,12 +8,12 @@
 
 namespace Dynarmic::A64 {
 
-enum class SM3TT1Variant {
+enum class SM3TTVariant {
     A,
     B,
 };
 
-static void SM3TT1(TranslatorVisitor& v, Vec Vm, Imm<2> imm2, Vec Vn, Vec Vd, SM3TT1Variant behavior) {
+static void SM3TT1(TranslatorVisitor& v, Vec Vm, Imm<2> imm2, Vec Vn, Vec Vd, SM3TTVariant behavior) {
     const IR::U128 d = v.ir.GetQ(Vd);
     const IR::U128 m = v.ir.GetQ(Vm);
     const IR::U128 n = v.ir.GetQ(Vn);
@@ -28,7 +28,7 @@ static void SM3TT1(TranslatorVisitor& v, Vec Vm, Imm<2> imm2, Vec Vn, Vec Vd, SM
     const IR::U32 wj_prime = v.ir.VectorGetElement(32, m, index);
     const IR::U32 ss2 = v.ir.Eor(top_n, v.ir.RotateRight(top_d, v.ir.Imm8(20)));
     const IR::U32 tt1 = [&] {
-        if (behavior == SM3TT1Variant::A) {
+        if (behavior == SM3TTVariant::A) {
             return v.ir.Eor(after_low_d, v.ir.Eor(top_d, before_top_d));
         }
         const IR::U32 tmp1 = v.ir.And(top_d, after_low_d);
@@ -47,13 +47,45 @@ static void SM3TT1(TranslatorVisitor& v, Vec Vm, Imm<2> imm2, Vec Vn, Vec Vd, SM
     v.ir.SetQ(Vd, result);
 }
 
+static void SM3TT2(TranslatorVisitor& v, Vec Vm, Imm<2> imm2, Vec Vn, Vec Vd, [[maybe_unused]] SM3TTVariant behavior) {
+    const IR::U128 d = v.ir.GetQ(Vd);
+    const IR::U128 m = v.ir.GetQ(Vm);
+    const IR::U128 n = v.ir.GetQ(Vn);
+    const u32 index = imm2.ZeroExtend();
+
+    const IR::U32 top_d = v.ir.VectorGetElement(32, d, 3);
+    const IR::U32 before_top_d = v.ir.VectorGetElement(32, d, 2);
+    const IR::U32 after_low_d = v.ir.VectorGetElement(32, d, 1);
+    const IR::U32 low_d = v.ir.VectorGetElement(32, d, 0);
+    const IR::U32 top_n = v.ir.VectorGetElement(32, n, 3);
+
+    const IR::U32 wj = v.ir.VectorGetElement(32, m, index);
+    const IR::U32 tt2 = v.ir.Eor(after_low_d, v.ir.Eor(top_d, before_top_d));
+    const IR::U32 final_tt2 = v.ir.Add(tt2, v.ir.Add(low_d, v.ir.Add(top_n, wj)));
+    const IR::U32 top_result = v.ir.Eor(final_tt2, v.ir.Eor(v.ir.RotateRight(final_tt2, v.ir.Imm8(23)),
+                                                            v.ir.RotateRight(final_tt2, v.ir.Imm8(15))));
+
+    const IR::U128 zero_vector = v.ir.ZeroVector();
+    const IR::U128 tmp1 = v.ir.VectorSetElement(32, zero_vector, 0, after_low_d);
+    const IR::U128 tmp2 = v.ir.VectorSetElement(32, tmp1, 1, v.ir.RotateRight(before_top_d, v.ir.Imm8(13)));
+    const IR::U128 tmp3 = v.ir.VectorSetElement(32, tmp2, 2, top_d);
+    const IR::U128 result = v.ir.VectorSetElement(32, tmp3, 3, top_result);
+
+    v.ir.SetQ(Vd, result);
+}
+
 bool TranslatorVisitor::SM3TT1A(Vec Vm, Imm<2> imm2, Vec Vn, Vec Vd) {
-    SM3TT1(*this, Vm, imm2, Vn, Vd, SM3TT1Variant::A);
+    SM3TT1(*this, Vm, imm2, Vn, Vd, SM3TTVariant::A);
     return true;
 }
 
 bool TranslatorVisitor::SM3TT1B(Vec Vm, Imm<2> imm2, Vec Vn, Vec Vd) {
-    SM3TT1(*this, Vm, imm2, Vn, Vd, SM3TT1Variant::B);
+    SM3TT1(*this, Vm, imm2, Vn, Vd, SM3TTVariant::B);
+    return true;
+}
+
+bool TranslatorVisitor::SM3TT2A(Vec Vm, Imm<2> imm2, Vec Vn, Vec Vd) {
+    SM3TT2(*this, Vm, imm2, Vn, Vd, SM3TTVariant::A);
     return true;
 }