[wasm-simd] [liftoff] Implement not/and/or/xor/andnot on x64 and ia32

Bug: v8:9909
Change-Id: Ic71a2a012cbd538d65afb047fe73e98be6454aed
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2150157
Commit-Queue: Fanchen Kong <fanchen.kong@intel.com>
Reviewed-by: Zhi An Ng <zhin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#67191}

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

@@ -2223,6 +2223,31 @@ void LiftoffAssembler::emit_i16x8_uconvert_i32x4(LiftoffRegister dst,
   bailout(kSimd, "i16x8_uconvert_i32x4");
 }
 
+void LiftoffAssembler::emit_s128_not(LiftoffRegister dst, LiftoffRegister src) {
+  bailout(kSimd, "s128_not");
+}
+
+void LiftoffAssembler::emit_s128_and(LiftoffRegister dst, LiftoffRegister lhs,
+                                     LiftoffRegister rhs) {
+  bailout(kSimd, "s128_and");
+}
+
+void LiftoffAssembler::emit_s128_or(LiftoffRegister dst, LiftoffRegister lhs,
+                                    LiftoffRegister rhs) {
+  bailout(kSimd, "s128_or");
+}
+
+void LiftoffAssembler::emit_s128_xor(LiftoffRegister dst, LiftoffRegister lhs,
+                                     LiftoffRegister rhs) {
+  bailout(kSimd, "s128_xor");
+}
+
+void LiftoffAssembler::emit_s128_and_not(LiftoffRegister dst,
+                                         LiftoffRegister lhs,
+                                         LiftoffRegister rhs) {
+  bailout(kSimd, "s128_and_not");
+}
+
 void LiftoffAssembler::emit_i8x16_rounding_average_u(LiftoffRegister dst,
                                                      LiftoffRegister lhs,
                                                      LiftoffRegister rhs) {

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

@@ -1558,6 +1558,31 @@ void LiftoffAssembler::emit_i16x8_uconvert_i32x4(LiftoffRegister dst,
   bailout(kSimd, "i16x8_uconvert_i32x4");
 }
 
+void LiftoffAssembler::emit_s128_not(LiftoffRegister dst, LiftoffRegister src) {
+  bailout(kSimd, "s128_not");
+}
+
+void LiftoffAssembler::emit_s128_and(LiftoffRegister dst, LiftoffRegister lhs,
+                                     LiftoffRegister rhs) {
+  bailout(kSimd, "s128_and");
+}
+
+void LiftoffAssembler::emit_s128_or(LiftoffRegister dst, LiftoffRegister lhs,
+                                    LiftoffRegister rhs) {
+  bailout(kSimd, "s128_or");
+}
+
+void LiftoffAssembler::emit_s128_xor(LiftoffRegister dst, LiftoffRegister lhs,
+                                     LiftoffRegister rhs) {
+  bailout(kSimd, "s128_xor");
+}
+
+void LiftoffAssembler::emit_s128_and_not(LiftoffRegister dst,
+                                         LiftoffRegister lhs,
+                                         LiftoffRegister rhs) {
+  bailout(kSimd, "s128_and_not");
+}
+
 void LiftoffAssembler::emit_i8x16_rounding_average_u(LiftoffRegister dst,
                                                      LiftoffRegister lhs,
                                                      LiftoffRegister rhs) {

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

@@ -2049,6 +2049,34 @@ void LiftoffAssembler::emit_f64x2_eq(LiftoffRegister dst, LiftoffRegister lhs,
       this, dst, lhs, rhs);
 }
 
+void LiftoffAssembler::emit_s128_not(LiftoffRegister dst, LiftoffRegister src) {
+  if (dst.fp() != src.fp()) {
+    Pcmpeqd(dst.fp(), dst.fp());
+    Pxor(dst.fp(), src.fp());
+  } else {
+    Pcmpeqd(liftoff::kScratchDoubleReg, liftoff::kScratchDoubleReg);
+    Pxor(dst.fp(), liftoff::kScratchDoubleReg);
+  }
+}
+
+void LiftoffAssembler::emit_s128_and(LiftoffRegister dst, LiftoffRegister lhs,
+                                     LiftoffRegister rhs) {
+  liftoff::EmitSimdCommutativeBinOp<&Assembler::vpand, &Assembler::pand>(
+      this, dst, lhs, rhs);
+}
+
+void LiftoffAssembler::emit_s128_or(LiftoffRegister dst, LiftoffRegister lhs,
+                                    LiftoffRegister rhs) {
+  liftoff::EmitSimdCommutativeBinOp<&Assembler::vpor, &Assembler::por>(
+      this, dst, lhs, rhs);
+}
+
+void LiftoffAssembler::emit_s128_xor(LiftoffRegister dst, LiftoffRegister lhs,
+                                     LiftoffRegister rhs) {
+  liftoff::EmitSimdCommutativeBinOp<&Assembler::vpxor, &Assembler::pxor>(
+      this, dst, lhs, rhs);
+}
+
 void LiftoffAssembler::emit_i8x16_neg(LiftoffRegister dst,
                                       LiftoffRegister src) {
   if (dst.fp() == src.fp()) {
@@ -2529,6 +2557,13 @@ void LiftoffAssembler::emit_i16x8_uconvert_i32x4(LiftoffRegister dst,
                                                              rhs, SSE4_1);
 }
 
+void LiftoffAssembler::emit_s128_and_not(LiftoffRegister dst,
+                                         LiftoffRegister lhs,
+                                         LiftoffRegister rhs) {
+  liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vandnps, &Assembler::andnps>(
+      this, dst, rhs, lhs);
+}
+
 void LiftoffAssembler::emit_i8x16_rounding_average_u(LiftoffRegister dst,
                                                      LiftoffRegister lhs,
                                                      LiftoffRegister rhs) {

src/wasm/baseline/liftoff-assembler.h

@@ -739,6 +739,13 @@ class LiftoffAssembler : public TurboAssembler {
                             LiftoffRegister rhs);
   inline void emit_f64x2_eq(LiftoffRegister dst, LiftoffRegister lhs,
                             LiftoffRegister rhs);
+  inline void emit_s128_not(LiftoffRegister dst, LiftoffRegister src);
+  inline void emit_s128_and(LiftoffRegister dst, LiftoffRegister lhs,
+                            LiftoffRegister rhs);
+  inline void emit_s128_or(LiftoffRegister dst, LiftoffRegister lhs,
+                           LiftoffRegister rhs);
+  inline void emit_s128_xor(LiftoffRegister dst, LiftoffRegister lhs,
+                            LiftoffRegister rhs);
   inline void emit_i8x16_neg(LiftoffRegister dst, LiftoffRegister src);
   inline void emit_i8x16_add(LiftoffRegister dst, LiftoffRegister lhs,
                              LiftoffRegister rhs);
@@ -849,6 +856,8 @@ class LiftoffAssembler : public TurboAssembler {
   inline void emit_i16x8_uconvert_i32x4(LiftoffRegister dst,
                                         LiftoffRegister lhs,
                                         LiftoffRegister rhs);
+  inline void emit_s128_and_not(LiftoffRegister dst, LiftoffRegister lhs,
+                                LiftoffRegister rhs);
   inline void emit_i8x16_rounding_average_u(LiftoffRegister dst,
                                             LiftoffRegister lhs,
                                             LiftoffRegister rhs);

src/wasm/baseline/liftoff-compiler.cc

@@ -2352,6 +2352,14 @@ class LiftoffCompiler {
         return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_f32x4_eq);
       case wasm::kExprF64x2Eq:
         return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_f64x2_eq);
+      case wasm::kExprS128Not:
+        return EmitUnOp<kS128, kS128>(&LiftoffAssembler::emit_s128_not);
+      case wasm::kExprS128And:
+        return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_s128_and);
+      case wasm::kExprS128Or:
+        return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_s128_or);
+      case wasm::kExprS128Xor:
+        return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_s128_xor);
       case wasm::kExprI8x16Neg:
         return EmitUnOp<kS128, kS128>(&LiftoffAssembler::emit_i8x16_neg);
       case wasm::kExprI8x16Add:
@@ -2472,6 +2480,8 @@ class LiftoffCompiler {
       case wasm::kExprI16x8UConvertI32x4:
         return EmitBinOp<kS128, kS128>(
             &LiftoffAssembler::emit_i16x8_uconvert_i32x4);
+      case wasm::kExprS128AndNot:
+        return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_s128_and_not);
       case wasm::kExprI8x16RoundingAverageU:
         return EmitBinOp<kS128, kS128>(
             &LiftoffAssembler::emit_i8x16_rounding_average_u);

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

@@ -2012,6 +2012,34 @@ void LiftoffAssembler::emit_f64x2_eq(LiftoffRegister dst, LiftoffRegister lhs,
       this, dst, lhs, rhs);
 }
 
+void LiftoffAssembler::emit_s128_not(LiftoffRegister dst, LiftoffRegister src) {
+  if (dst.fp() != src.fp()) {
+    Pcmpeqd(dst.fp(), dst.fp());
+    Pxor(dst.fp(), src.fp());
+  } else {
+    Pcmpeqd(kScratchDoubleReg, kScratchDoubleReg);
+    Pxor(dst.fp(), kScratchDoubleReg);
+  }
+}
+
+void LiftoffAssembler::emit_s128_and(LiftoffRegister dst, LiftoffRegister lhs,
+                                     LiftoffRegister rhs) {
+  liftoff::EmitSimdCommutativeBinOp<&Assembler::vpand, &Assembler::pand>(
+      this, dst, lhs, rhs);
+}
+
+void LiftoffAssembler::emit_s128_or(LiftoffRegister dst, LiftoffRegister lhs,
+                                    LiftoffRegister rhs) {
+  liftoff::EmitSimdCommutativeBinOp<&Assembler::vpor, &Assembler::por>(
+      this, dst, lhs, rhs);
+}
+
+void LiftoffAssembler::emit_s128_xor(LiftoffRegister dst, LiftoffRegister lhs,
+                                     LiftoffRegister rhs) {
+  liftoff::EmitSimdCommutativeBinOp<&Assembler::vpxor, &Assembler::pxor>(
+      this, dst, lhs, rhs);
+}
+
 void LiftoffAssembler::emit_i8x16_neg(LiftoffRegister dst,
                                       LiftoffRegister src) {
   if (dst.fp() == src.fp()) {
@@ -2491,6 +2519,13 @@ void LiftoffAssembler::emit_i16x8_uconvert_i32x4(LiftoffRegister dst,
                                                              rhs, SSE4_1);
 }
 
+void LiftoffAssembler::emit_s128_and_not(LiftoffRegister dst,
+                                         LiftoffRegister lhs,
+                                         LiftoffRegister rhs) {
+  liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vandnps, &Assembler::andnps>(
+      this, dst, rhs, lhs);
+}
+
 void LiftoffAssembler::emit_i8x16_rounding_average_u(LiftoffRegister dst,
                                                      LiftoffRegister lhs,
                                                      LiftoffRegister rhs) {