From 7b6cc4ec705d0a227115ff61bd0fb954e2bf37fc Mon Sep 17 00:00:00 2001
From: SachinVin <sachinvinayak2000@gmail.com>
Date: Sat, 27 Jul 2019 09:56:24 +0530
Subject: [PATCH] a64 emitter: fix Scalar Saturating Instructions

---
 src/backend/A64/emitter/a64_emitter.cpp | 37 +++++++++++++++++++------
 src/backend/A64/emitter/a64_emitter.h   |  8 +++---
 2 files changed, 32 insertions(+), 13 deletions(-)

diff --git a/src/backend/A64/emitter/a64_emitter.cpp b/src/backend/A64/emitter/a64_emitter.cpp
index 12963a73..6eb399e2 100644
--- a/src/backend/A64/emitter/a64_emitter.cpp
+++ b/src/backend/A64/emitter/a64_emitter.cpp
@@ -1998,8 +1998,27 @@ void ARM64FloatEmitter::EmitScalarThreeSame(bool U, u32 size, u32 opcode, ARM64R
     Rd = DecodeReg(Rd);
     Rn = DecodeReg(Rn);
     Rm = DecodeReg(Rm);
+    int esize = 0;
+    switch (size) {
+    case 8:
+        esize = 0;
+        break;
+    case 16:
+        esize = 1;
+        break;
+    case 32:
+        esize = 2;
+        break;
+    case 64:
+        esize = 3;
+        break;
+    default:
+        ASSERT_MSG(false, "Size must be 8, 16, 32, or 64");
+        break;
+    }
 
-    Write32((U << 29) | (0b1011110001 << 21) | (size << 22) | (Rm << 16) |
+
+    Write32((U << 29) | (0b1011110001 << 21) | (esize << 22) | (Rm << 16) |
             (opcode << 11) | (1 << 10) | (Rn << 5) | Rd);
 }
 
@@ -2799,17 +2818,17 @@ void ARM64FloatEmitter::EmitScalar3Source(bool isDouble, ARM64Reg Rd, ARM64Reg R
 }
 
 // Scalar three same
-void ARM64FloatEmitter::SQADD(ESize esize, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm) {
-    EmitScalarThreeSame(0, static_cast<u32>(esize), 0b000011, Rd, Rn, Rm);
+void ARM64FloatEmitter::SQADD(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm) {
+    EmitScalarThreeSame(0, size, 0b00001, Rd, Rn, Rm);
 }
-void ARM64FloatEmitter::SQSUB(ESize esize, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm) {
-    EmitScalarThreeSame(0, static_cast<u32>(esize), 0b001011, Rd, Rn, Rm);
+void ARM64FloatEmitter::SQSUB(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm) {
+    EmitScalarThreeSame(0, size, 0b00101, Rd, Rn, Rm);
 }
-void ARM64FloatEmitter::UQADD(ESize esize, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm) {
-    EmitScalarThreeSame(1, static_cast<u32>(esize), 0b000011, Rd, Rn, Rm);
+void ARM64FloatEmitter::UQADD(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm) {
+    EmitScalarThreeSame(1, size, 0b00001, Rd, Rn, Rm);
 }
-void ARM64FloatEmitter::UQSUB(ESize esize, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm) {
-    EmitScalarThreeSame(1, static_cast<u32>(esize), 0b001011, Rd, Rn, Rm);
+void ARM64FloatEmitter::UQSUB(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm) {
+    EmitScalarThreeSame(1, size, 0b00101, Rd, Rn, Rm);
 }
 
 // Scalar floating point immediate
diff --git a/src/backend/A64/emitter/a64_emitter.h b/src/backend/A64/emitter/a64_emitter.h
index f4b99436..533b1175 100644
--- a/src/backend/A64/emitter/a64_emitter.h
+++ b/src/backend/A64/emitter/a64_emitter.h
@@ -954,10 +954,10 @@ public:
     void FNMSUB(ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm, ARM64Reg Ra);
 
     // Scalar three same
-    void SQADD(ESize esize, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
-    void UQADD(ESize esize, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
-    void SQSUB(ESize esize, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
-    void UQSUB(ESize esize, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
+    void SQADD(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
+    void UQADD(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
+    void SQSUB(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
+    void UQSUB(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
     
     // Scalar floating point immediate
     void FMOV(ARM64Reg Rd, uint8_t imm8);