[wasm-simd] [liftoff] Implement rounding average on x64 and ia32

Bug: v8:9909
Change-Id: I5dd3e88ab6f7508a07b2a8e5adb033fff7e28fe2
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2143386Reviewed-by: Clemens Backes <clemensb@chromium.org>
Reviewed-by: Zhi An Ng <zhin@chromium.org>
Commit-Queue: Fanchen Kong <fanchen.kong@intel.com>
Cr-Commit-Position: refs/heads/master@{#67095}

src/wasm/baseline/arm/liftoff-assembler-arm.h

@@ -1959,6 +1959,18 @@ void LiftoffAssembler::emit_i8x16_max_u(LiftoffRegister dst,
        liftoff::GetSimd128Register(lhs), liftoff::GetSimd128Register(rhs));
 }
 
+void LiftoffAssembler::emit_i8x16_rounding_average_u(LiftoffRegister dst,
+                                                     LiftoffRegister lhs,
+                                                     LiftoffRegister rhs) {
+  bailout(kSimd, "i8x16_rounding_average_u");
+}
+
+void LiftoffAssembler::emit_i16x8_rounding_average_u(LiftoffRegister dst,
+                                                     LiftoffRegister lhs,
+                                                     LiftoffRegister rhs) {
+  bailout(kSimd, "i16x8_rounding_average_u");
+}
+
 void LiftoffAssembler::StackCheck(Label* ool_code, Register limit_address) {
   ldr(limit_address, MemOperand(limit_address));
   cmp(sp, limit_address);

src/wasm/baseline/arm64/liftoff-assembler-arm64.h

@@ -1424,6 +1424,18 @@ void LiftoffAssembler::emit_i8x16_max_u(LiftoffRegister dst,
   Umax(dst.fp().V16B(), lhs.fp().V16B(), rhs.fp().V16B());
 }
 
+void LiftoffAssembler::emit_i8x16_rounding_average_u(LiftoffRegister dst,
+                                                     LiftoffRegister lhs,
+                                                     LiftoffRegister rhs) {
+  bailout(kSimd, "i8x16_rounding_average_u");
+}
+
+void LiftoffAssembler::emit_i16x8_rounding_average_u(LiftoffRegister dst,
+                                                     LiftoffRegister lhs,
+                                                     LiftoffRegister rhs) {
+  bailout(kSimd, "i16x8_rounding_average_u");
+}
+
 void LiftoffAssembler::StackCheck(Label* ool_code, Register limit_address) {
   Ldr(limit_address, MemOperand(limit_address));
   Cmp(sp, limit_address);

src/wasm/baseline/ia32/liftoff-assembler-ia32.h

@@ -2471,6 +2471,20 @@ void LiftoffAssembler::emit_f64x2_replace_lane(LiftoffRegister dst,
   }
 }
 
+void LiftoffAssembler::emit_i8x16_rounding_average_u(LiftoffRegister dst,
+                                                     LiftoffRegister lhs,
+                                                     LiftoffRegister rhs) {
+  liftoff::EmitSimdCommutativeBinOp<&Assembler::vpavgb, &Assembler::pavgb>(
+      this, dst, lhs, rhs);
+}
+
+void LiftoffAssembler::emit_i16x8_rounding_average_u(LiftoffRegister dst,
+                                                     LiftoffRegister lhs,
+                                                     LiftoffRegister rhs) {
+  liftoff::EmitSimdCommutativeBinOp<&Assembler::vpavgw, &Assembler::pavgw>(
+      this, dst, lhs, rhs);
+}
+
 void LiftoffAssembler::StackCheck(Label* ool_code, Register limit_address) {
   cmp(esp, Operand(limit_address, 0));
   j(below_equal, ool_code);

src/wasm/baseline/liftoff-assembler.h

@@ -793,6 +793,12 @@ class LiftoffAssembler : public TurboAssembler {
                              LiftoffRegister rhs);
   inline void emit_f64x2_mul(LiftoffRegister dst, LiftoffRegister lhs,
                              LiftoffRegister rhs);
+  inline void emit_i8x16_rounding_average_u(LiftoffRegister dst,
+                                            LiftoffRegister lhs,
+                                            LiftoffRegister rhs);
+  inline void emit_i16x8_rounding_average_u(LiftoffRegister dst,
+                                            LiftoffRegister lhs,
+                                            LiftoffRegister rhs);
   inline void emit_i8x16_extract_lane_s(LiftoffRegister dst,
                                         LiftoffRegister lhs,
                                         uint8_t imm_lane_idx);

src/wasm/baseline/liftoff-compiler.cc

@@ -2415,6 +2415,12 @@ class LiftoffCompiler {
         return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_f64x2_sub);
       case wasm::kExprF64x2Mul:
         return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_f64x2_mul);
+      case wasm::kExprI8x16RoundingAverageU:
+        return EmitBinOp<kS128, kS128>(
+            &LiftoffAssembler::emit_i8x16_rounding_average_u);
+      case wasm::kExprI16x8RoundingAverageU:
+        return EmitBinOp<kS128, kS128>(
+            &LiftoffAssembler::emit_i16x8_rounding_average_u);
       default:
         unsupported(decoder, kSimd, "simd");
     }

src/wasm/baseline/x64/liftoff-assembler-x64.h

@@ -2402,6 +2402,20 @@ void LiftoffAssembler::emit_f64x2_replace_lane(LiftoffRegister dst,
   }
 }
 
+void LiftoffAssembler::emit_i8x16_rounding_average_u(LiftoffRegister dst,
+                                                     LiftoffRegister lhs,
+                                                     LiftoffRegister rhs) {
+  liftoff::EmitSimdCommutativeBinOp<&Assembler::vpavgb, &Assembler::pavgb>(
+      this, dst, lhs, rhs);
+}
+
+void LiftoffAssembler::emit_i16x8_rounding_average_u(LiftoffRegister dst,
+                                                     LiftoffRegister lhs,
+                                                     LiftoffRegister rhs) {
+  liftoff::EmitSimdCommutativeBinOp<&Assembler::vpavgw, &Assembler::pavgw>(
+      this, dst, lhs, rhs);
+}
+
 void LiftoffAssembler::StackCheck(Label* ool_code, Register limit_address) {
   cmpq(rsp, Operand(limit_address, 0));
   j(below_equal, ool_code);