A64: Implement FMUL (vector)

src/backend_x64/emit_x64_vector_floating_point.cpp

@@ -195,6 +195,14 @@ void EmitX64::EmitFPVectorDiv64(EmitContext& ctx, IR::Inst* inst) {
     EmitVectorOperation64(code, ctx, inst, &Xbyak::CodeGenerator::divpd);
 }
 
+void EmitX64::EmitFPVectorMul32(EmitContext& ctx, IR::Inst* inst) {
+    EmitVectorOperation32(code, ctx, inst, &Xbyak::CodeGenerator::mulps);
+}
+
+void EmitX64::EmitFPVectorMul64(EmitContext& ctx, IR::Inst* inst) {
+    EmitVectorOperation64(code, ctx, inst, &Xbyak::CodeGenerator::mulpd);
+}
+
 void EmitX64::EmitFPVectorSub32(EmitContext& ctx, IR::Inst* inst) {
     EmitVectorOperation32(code, ctx, inst, &Xbyak::CodeGenerator::subps);
 }

src/frontend/A64/decoder/a64.inc

@@ -766,7 +766,7 @@ INST(MLS_vec,                "MLS (vector)",                              "0Q101
 //INST(FMAXNMP_vec_2,          "FMAXNMP (vector)",                          "0Q1011100z1mmmmm110001nnnnnddddd")
 //INST(FMLAL_vec_2,            "FMLAL, FMLAL2 (vector)",                    "0Q1011100z1mmmmm110011nnnnnddddd")
 //INST(FADDP_vec_2,            "FADDP (vector)",                            "0Q1011100z1mmmmm110101nnnnnddddd")
-//INST(FMUL_vec_2,             "FMUL (vector)",                             "0Q1011100z1mmmmm110111nnnnnddddd")
+INST(FMUL_vec_2,             "FMUL (vector)",                             "0Q1011100z1mmmmm110111nnnnnddddd")
 //INST(FCMGE_reg_4,            "FCMGE (register)",                          "0Q1011100z1mmmmm111001nnnnnddddd")
 //INST(FACGE_4,                "FACGE",                                     "0Q1011100z1mmmmm111011nnnnnddddd")
 //INST(FMAXP_vec_2,            "FMAXP (vector)",                            "0Q1011100z1mmmmm111101nnnnnddddd")

src/frontend/A64/translate/impl/simd_three_same.cpp

@@ -383,6 +383,20 @@ bool TranslatorVisitor::EOR_asimd(bool Q, Vec Vm, Vec Vn, Vec Vd) {
     return true;
 }
 
+bool TranslatorVisitor::FMUL_vec_2(bool Q, bool sz, Vec Vm, Vec Vn, Vec Vd) {
+    if (sz && !Q) {
+        return ReservedValue();
+    }
+    const size_t esize = sz ? 64 : 32;
+    const size_t datasize = Q ? 128 : 64;
+
+    const IR::U128 operand1 = V(datasize, Vn);
+    const IR::U128 operand2 = V(datasize, Vm);
+    IR::U128 result = ir.FPVectorMul(esize, operand1, operand2);
+    V(datasize, Vd, result);
+    return true;
+}
+
 bool TranslatorVisitor::FDIV_2(bool Q, bool sz, Vec Vm, Vec Vn, Vec Vd) {
     if (sz && !Q) {
         return ReservedValue();

src/frontend/ir/ir_emitter.cpp

@@ -1382,6 +1382,17 @@ U128 IREmitter::FPVectorDiv(size_t esize, const U128& a, const U128& b) {
     return {};
 }
 
+U128 IREmitter::FPVectorMul(size_t esize, const U128& a, const U128& b) {
+    switch (esize) {
+    case 32:
+        return Inst<U128>(Opcode::FPVectorMul32, a, b);
+    case 64:
+        return Inst<U128>(Opcode::FPVectorMul64, a, b);
+    }
+    UNREACHABLE();
+    return {};
+}
+
 U128 IREmitter::FPVectorSub(size_t esize, const U128& a, const U128& b) {
     switch (esize) {
     case 32:

src/frontend/ir/ir_emitter.h

@@ -274,6 +274,7 @@ public:
 
     U128 FPVectorAdd(size_t esize, const U128& a, const U128& b);
     U128 FPVectorDiv(size_t esize, const U128& a, const U128& b);
+    U128 FPVectorMul(size_t esize, const U128& a, const U128& b);
     U128 FPVectorSub(size_t esize, const U128& a, const U128& b);
 
     void Breakpoint();

src/frontend/ir/opcodes.inc

@@ -354,6 +354,8 @@ OPCODE(FPVectorAdd32,                       T::U128,        T::U128,        T::U
 OPCODE(FPVectorAdd64,                       T::U128,        T::U128,        T::U128                         )
 OPCODE(FPVectorDiv32,                       T::U128,        T::U128,        T::U128                         )
 OPCODE(FPVectorDiv64,                       T::U128,        T::U128,        T::U128                         )
+OPCODE(FPVectorMul32,                       T::U128,        T::U128,        T::U128                         )
+OPCODE(FPVectorMul64,                       T::U128,        T::U128,        T::U128                         )
 OPCODE(FPVectorSub32,                       T::U128,        T::U128,        T::U128                         )
 OPCODE(FPVectorSub64,                       T::U128,        T::U128,        T::U128                         )