[wasm-simd][liftoff] Implement gt on x64 and ia32

Bug: v8:9909
Change-Id: Ic75c75fa3693a59bf059cf852172900ad95a1941
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2169026
Commit-Queue: Jing Bao <jing.bao@intel.com>
Reviewed-by: Zhi An Ng <zhin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#67461}

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

@@ -2434,6 +2434,16 @@ void LiftoffAssembler::emit_i8x16_ne(LiftoffRegister dst, LiftoffRegister lhs,
   vmvn(liftoff::GetSimd128Register(dst), liftoff::GetSimd128Register(dst));
 }
 
+void LiftoffAssembler::emit_i8x16_gt_s(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  bailout(kSimd, "i8x16gt_s");
+}
+
+void LiftoffAssembler::emit_i8x16_gt_u(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  bailout(kSimd, "i8x16gt_u");
+}
+
 void LiftoffAssembler::emit_i8x16_ge_s(LiftoffRegister dst, LiftoffRegister lhs,
                                        LiftoffRegister rhs) {
   bailout(kSimd, "i8x16ge_s");
@@ -2457,6 +2467,16 @@ void LiftoffAssembler::emit_i16x8_ne(LiftoffRegister dst, LiftoffRegister lhs,
   vmvn(liftoff::GetSimd128Register(dst), liftoff::GetSimd128Register(dst));
 }
 
+void LiftoffAssembler::emit_i16x8_gt_s(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  bailout(kSimd, "i16x8gt_s");
+}
+
+void LiftoffAssembler::emit_i16x8_gt_u(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  bailout(kSimd, "i16x8gt_u");
+}
+
 void LiftoffAssembler::emit_i16x8_ge_s(LiftoffRegister dst, LiftoffRegister lhs,
                                        LiftoffRegister rhs) {
   vcge(NeonS16, liftoff::GetSimd128Register(dst),
@@ -2482,6 +2502,16 @@ void LiftoffAssembler::emit_i32x4_ne(LiftoffRegister dst, LiftoffRegister lhs,
   vmvn(liftoff::GetSimd128Register(dst), liftoff::GetSimd128Register(dst));
 }
 
+void LiftoffAssembler::emit_i32x4_gt_s(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  bailout(kSimd, "i32x4gt_s");
+}
+
+void LiftoffAssembler::emit_i32x4_gt_u(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  bailout(kSimd, "i32x4gt_u");
+}
+
 void LiftoffAssembler::emit_i32x4_ge_s(LiftoffRegister dst, LiftoffRegister lhs,
                                        LiftoffRegister rhs) {
   vcge(NeonS32, liftoff::GetSimd128Register(dst),

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

@@ -1540,6 +1540,16 @@ void LiftoffAssembler::emit_i8x16_ne(LiftoffRegister dst, LiftoffRegister lhs,
   Mvn(dst.fp().V16B(), dst.fp().V16B());
 }
 
+void LiftoffAssembler::emit_i8x16_gt_s(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  bailout(kSimd, "i8x16gt_s");
+}
+
+void LiftoffAssembler::emit_i8x16_gt_u(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  bailout(kSimd, "i8x16gt_u");
+}
+
 void LiftoffAssembler::emit_i8x16_ge_s(LiftoffRegister dst, LiftoffRegister lhs,
                                        LiftoffRegister rhs) {
   bailout(kSimd, "i8x16ge_s");
@@ -1561,6 +1571,16 @@ void LiftoffAssembler::emit_i16x8_ne(LiftoffRegister dst, LiftoffRegister lhs,
   Mvn(dst.fp().V8H(), dst.fp().V8H());
 }
 
+void LiftoffAssembler::emit_i16x8_gt_s(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  bailout(kSimd, "i16x8gt_s");
+}
+
+void LiftoffAssembler::emit_i16x8_gt_u(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  bailout(kSimd, "i16x8gt_u");
+}
+
 void LiftoffAssembler::emit_i16x8_ge_s(LiftoffRegister dst, LiftoffRegister lhs,
                                        LiftoffRegister rhs) {
   Cmge(dst.fp().V8H(), lhs.fp().V8H(), rhs.fp().V8H());
@@ -1582,6 +1602,16 @@ void LiftoffAssembler::emit_i32x4_ne(LiftoffRegister dst, LiftoffRegister lhs,
   Mvn(dst.fp().V4S(), dst.fp().V4S());
 }
 
+void LiftoffAssembler::emit_i32x4_gt_s(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  bailout(kSimd, "i32x4gt_s");
+}
+
+void LiftoffAssembler::emit_i32x4_gt_u(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  bailout(kSimd, "i32x4gt_u");
+}
+
 void LiftoffAssembler::emit_i32x4_ge_s(LiftoffRegister dst, LiftoffRegister lhs,
                                        LiftoffRegister rhs) {
   Cmge(dst.fp().V4S(), lhs.fp().V4S(), rhs.fp().V4S());

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

@@ -2033,6 +2033,27 @@ void LiftoffAssembler::emit_i8x16_ne(LiftoffRegister dst, LiftoffRegister lhs,
   Pxor(dst.fp(), liftoff::kScratchDoubleReg);
 }
 
+void LiftoffAssembler::emit_i8x16_gt_s(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vpcmpgtb,
+                                       &Assembler::pcmpgtb>(this, dst, lhs,
+                                                            rhs);
+}
+
+void LiftoffAssembler::emit_i8x16_gt_u(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  DoubleRegister ref = rhs.fp();
+  if (dst == rhs) {
+    Movaps(liftoff::kScratchDoubleReg, rhs.fp());
+    ref = liftoff::kScratchDoubleReg;
+  }
+  liftoff::EmitSimdCommutativeBinOp<&Assembler::vpmaxub, &Assembler::pmaxub>(
+      this, dst, lhs, rhs);
+  Pcmpeqb(dst.fp(), ref);
+  Pcmpeqb(liftoff::kScratchDoubleReg, liftoff::kScratchDoubleReg);
+  Pxor(dst.fp(), liftoff::kScratchDoubleReg);
+}
+
 void LiftoffAssembler::emit_i8x16_ge_s(LiftoffRegister dst, LiftoffRegister lhs,
                                        LiftoffRegister rhs) {
   DoubleRegister ref = rhs.fp();
@@ -2071,6 +2092,27 @@ void LiftoffAssembler::emit_i16x8_ne(LiftoffRegister dst, LiftoffRegister lhs,
   Pxor(dst.fp(), liftoff::kScratchDoubleReg);
 }
 
+void LiftoffAssembler::emit_i16x8_gt_s(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vpcmpgtw,
+                                       &Assembler::pcmpgtw>(this, dst, lhs,
+                                                            rhs);
+}
+
+void LiftoffAssembler::emit_i16x8_gt_u(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  DoubleRegister ref = rhs.fp();
+  if (dst == rhs) {
+    Movaps(liftoff::kScratchDoubleReg, rhs.fp());
+    ref = liftoff::kScratchDoubleReg;
+  }
+  liftoff::EmitSimdCommutativeBinOp<&Assembler::vpmaxuw, &Assembler::pmaxuw>(
+      this, dst, lhs, rhs);
+  Pcmpeqw(dst.fp(), ref);
+  Pcmpeqw(liftoff::kScratchDoubleReg, liftoff::kScratchDoubleReg);
+  Pxor(dst.fp(), liftoff::kScratchDoubleReg);
+}
+
 void LiftoffAssembler::emit_i16x8_ge_s(LiftoffRegister dst, LiftoffRegister lhs,
                                        LiftoffRegister rhs) {
   DoubleRegister ref = rhs.fp();
@@ -2109,6 +2151,27 @@ void LiftoffAssembler::emit_i32x4_ne(LiftoffRegister dst, LiftoffRegister lhs,
   Pxor(dst.fp(), liftoff::kScratchDoubleReg);
 }
 
+void LiftoffAssembler::emit_i32x4_gt_s(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vpcmpgtd,
+                                       &Assembler::pcmpgtd>(this, dst, lhs,
+                                                            rhs);
+}
+
+void LiftoffAssembler::emit_i32x4_gt_u(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  DoubleRegister ref = rhs.fp();
+  if (dst == rhs) {
+    Movaps(liftoff::kScratchDoubleReg, rhs.fp());
+    ref = liftoff::kScratchDoubleReg;
+  }
+  liftoff::EmitSimdCommutativeBinOp<&Assembler::vpmaxud, &Assembler::pmaxud>(
+      this, dst, lhs, rhs);
+  Pcmpeqd(dst.fp(), ref);
+  Pcmpeqd(liftoff::kScratchDoubleReg, liftoff::kScratchDoubleReg);
+  Pxor(dst.fp(), liftoff::kScratchDoubleReg);
+}
+
 void LiftoffAssembler::emit_i32x4_ge_s(LiftoffRegister dst, LiftoffRegister lhs,
                                        LiftoffRegister rhs) {
   DoubleRegister ref = rhs.fp();

src/wasm/baseline/liftoff-assembler.h

@@ -733,6 +733,10 @@ class LiftoffAssembler : public TurboAssembler {
                             LiftoffRegister rhs);
   inline void emit_i8x16_ne(LiftoffRegister dst, LiftoffRegister lhs,
                             LiftoffRegister rhs);
+  inline void emit_i8x16_gt_s(LiftoffRegister dst, LiftoffRegister lhs,
+                              LiftoffRegister rhs);
+  inline void emit_i8x16_gt_u(LiftoffRegister dst, LiftoffRegister lhs,
+                              LiftoffRegister rhs);
   inline void emit_i8x16_ge_s(LiftoffRegister dst, LiftoffRegister lhs,
                               LiftoffRegister rhs);
   inline void emit_i8x16_ge_u(LiftoffRegister dst, LiftoffRegister lhs,
@@ -741,6 +745,10 @@ class LiftoffAssembler : public TurboAssembler {
                             LiftoffRegister rhs);
   inline void emit_i16x8_ne(LiftoffRegister dst, LiftoffRegister lhs,
                             LiftoffRegister rhs);
+  inline void emit_i16x8_gt_s(LiftoffRegister dst, LiftoffRegister lhs,
+                              LiftoffRegister rhs);
+  inline void emit_i16x8_gt_u(LiftoffRegister dst, LiftoffRegister lhs,
+                              LiftoffRegister rhs);
   inline void emit_i16x8_ge_s(LiftoffRegister dst, LiftoffRegister lhs,
                               LiftoffRegister rhs);
   inline void emit_i16x8_ge_u(LiftoffRegister dst, LiftoffRegister lhs,
@@ -749,6 +757,10 @@ class LiftoffAssembler : public TurboAssembler {
                             LiftoffRegister rhs);
   inline void emit_i32x4_ne(LiftoffRegister dst, LiftoffRegister lhs,
                             LiftoffRegister rhs);
+  inline void emit_i32x4_gt_s(LiftoffRegister dst, LiftoffRegister lhs,
+                              LiftoffRegister rhs);
+  inline void emit_i32x4_gt_u(LiftoffRegister dst, LiftoffRegister lhs,
+                              LiftoffRegister rhs);
   inline void emit_i32x4_ge_s(LiftoffRegister dst, LiftoffRegister lhs,
                               LiftoffRegister rhs);
   inline void emit_i32x4_ge_u(LiftoffRegister dst, LiftoffRegister lhs,

src/wasm/baseline/liftoff-compiler.cc

@@ -2366,6 +2366,10 @@ class LiftoffCompiler {
         return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_i8x16_eq);
       case wasm::kExprI8x16Ne:
         return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_i8x16_ne);
+      case wasm::kExprI8x16GtS:
+        return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_i8x16_gt_s);
+      case wasm::kExprI8x16GtU:
+        return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_i8x16_gt_u);
       case wasm::kExprI8x16LeS:
         return EmitBinOp<kS128, kS128, true>(
             &LiftoffAssembler::emit_i8x16_ge_s);
@@ -2380,6 +2384,10 @@ class LiftoffCompiler {
         return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_i16x8_eq);
       case wasm::kExprI16x8Ne:
         return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_i16x8_ne);
+      case wasm::kExprI16x8GtS:
+        return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_i16x8_gt_s);
+      case wasm::kExprI16x8GtU:
+        return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_i16x8_gt_u);
       case wasm::kExprI16x8LeS:
         return EmitBinOp<kS128, kS128, true>(
             &LiftoffAssembler::emit_i16x8_ge_s);
@@ -2394,6 +2402,10 @@ class LiftoffCompiler {
         return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_i32x4_eq);
       case wasm::kExprI32x4Ne:
         return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_i32x4_ne);
+      case wasm::kExprI32x4GtS:
+        return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_i32x4_gt_s);
+      case wasm::kExprI32x4GtU:
+        return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_i32x4_gt_u);
       case wasm::kExprI32x4LeS:
         return EmitBinOp<kS128, kS128, true>(
             &LiftoffAssembler::emit_i32x4_ge_s);

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

@@ -1996,6 +1996,27 @@ void LiftoffAssembler::emit_i8x16_ne(LiftoffRegister dst, LiftoffRegister lhs,
   Pxor(dst.fp(), kScratchDoubleReg);
 }
 
+void LiftoffAssembler::emit_i8x16_gt_s(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vpcmpgtb,
+                                       &Assembler::pcmpgtb>(this, dst, lhs,
+                                                            rhs);
+}
+
+void LiftoffAssembler::emit_i8x16_gt_u(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  DoubleRegister ref = rhs.fp();
+  if (dst == rhs) {
+    Movaps(kScratchDoubleReg, rhs.fp());
+    ref = kScratchDoubleReg;
+  }
+  liftoff::EmitSimdCommutativeBinOp<&Assembler::vpmaxub, &Assembler::pmaxub>(
+      this, dst, lhs, rhs);
+  Pcmpeqb(dst.fp(), ref);
+  Pcmpeqb(kScratchDoubleReg, kScratchDoubleReg);
+  Pxor(dst.fp(), kScratchDoubleReg);
+}
+
 void LiftoffAssembler::emit_i8x16_ge_s(LiftoffRegister dst, LiftoffRegister lhs,
                                        LiftoffRegister rhs) {
   DoubleRegister ref = rhs.fp();
@@ -2034,6 +2055,27 @@ void LiftoffAssembler::emit_i16x8_ne(LiftoffRegister dst, LiftoffRegister lhs,
   Pxor(dst.fp(), kScratchDoubleReg);
 }
 
+void LiftoffAssembler::emit_i16x8_gt_s(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vpcmpgtw,
+                                       &Assembler::pcmpgtw>(this, dst, lhs,
+                                                            rhs);
+}
+
+void LiftoffAssembler::emit_i16x8_gt_u(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  DoubleRegister ref = rhs.fp();
+  if (dst == rhs) {
+    Movaps(kScratchDoubleReg, rhs.fp());
+    ref = kScratchDoubleReg;
+  }
+  liftoff::EmitSimdCommutativeBinOp<&Assembler::vpmaxuw, &Assembler::pmaxuw>(
+      this, dst, lhs, rhs);
+  Pcmpeqw(dst.fp(), ref);
+  Pcmpeqw(kScratchDoubleReg, kScratchDoubleReg);
+  Pxor(dst.fp(), kScratchDoubleReg);
+}
+
 void LiftoffAssembler::emit_i16x8_ge_s(LiftoffRegister dst, LiftoffRegister lhs,
                                        LiftoffRegister rhs) {
   DoubleRegister ref = rhs.fp();
@@ -2072,6 +2114,27 @@ void LiftoffAssembler::emit_i32x4_ne(LiftoffRegister dst, LiftoffRegister lhs,
   Pxor(dst.fp(), kScratchDoubleReg);
 }
 
+void LiftoffAssembler::emit_i32x4_gt_s(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  liftoff::EmitSimdNonCommutativeBinOp<&Assembler::vpcmpgtd,
+                                       &Assembler::pcmpgtd>(this, dst, lhs,
+                                                            rhs);
+}
+
+void LiftoffAssembler::emit_i32x4_gt_u(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  DoubleRegister ref = rhs.fp();
+  if (dst == rhs) {
+    Movaps(kScratchDoubleReg, rhs.fp());
+    ref = kScratchDoubleReg;
+  }
+  liftoff::EmitSimdCommutativeBinOp<&Assembler::vpmaxud, &Assembler::pmaxud>(
+      this, dst, lhs, rhs);
+  Pcmpeqd(dst.fp(), ref);
+  Pcmpeqd(kScratchDoubleReg, kScratchDoubleReg);
+  Pxor(dst.fp(), kScratchDoubleReg);
+}
+
 void LiftoffAssembler::emit_i32x4_ge_s(LiftoffRegister dst, LiftoffRegister lhs,
                                        LiftoffRegister rhs) {
   DoubleRegister ref = rhs.fp();