[wasm-simd][liftoff][ia32][x64] Implement i32x4_dot_i16x8_s

Implement i32x4.dot_i16x8_s for Liftoff on on ia32 and x64.
ARM implementation will come later.

Bug: v8:10993
Change-Id: I33b859a21b91023b40d8cf7b9fee110b0d148a7c
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2477497Reviewed-by: Clemens Backes <clemensb@chromium.org>
Commit-Queue: Zhi An Ng <zhin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#70583}

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

@@ -2921,6 +2921,12 @@ void LiftoffAssembler::emit_i32x4_max_u(LiftoffRegister dst,
        liftoff::GetSimd128Register(lhs), liftoff::GetSimd128Register(rhs));
 }
 
+void LiftoffAssembler::emit_i32x4_dot_i16x8_s(LiftoffRegister dst,
+                                              LiftoffRegister lhs,
+                                              LiftoffRegister rhs) {
+  bailout(kSimd, "i32x4_dot_i16x8_s");
+}
+
 void LiftoffAssembler::emit_i16x8_splat(LiftoffRegister dst,
                                         LiftoffRegister src) {
   vdup(Neon16, liftoff::GetSimd128Register(dst), src.gp());

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

@@ -2003,6 +2003,12 @@ void LiftoffAssembler::emit_i32x4_max_u(LiftoffRegister dst,
   Umax(dst.fp().V4S(), lhs.fp().V4S(), rhs.fp().V4S());
 }
 
+void LiftoffAssembler::emit_i32x4_dot_i16x8_s(LiftoffRegister dst,
+                                              LiftoffRegister lhs,
+                                              LiftoffRegister rhs) {
+  bailout(kSimd, "i32x4_dot_i16x8_s");
+}
+
 void LiftoffAssembler::emit_i16x8_splat(LiftoffRegister dst,
                                         LiftoffRegister src) {
   Dup(dst.fp().V8H(), src.gp().W());

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

@@ -3584,6 +3584,13 @@ void LiftoffAssembler::emit_i32x4_max_u(LiftoffRegister dst,
       this, dst, lhs, rhs, base::Optional<CpuFeature>(SSE4_1));
 }
 
+void LiftoffAssembler::emit_i32x4_dot_i16x8_s(LiftoffRegister dst,
+                                              LiftoffRegister lhs,
+                                              LiftoffRegister rhs) {
+  liftoff::EmitSimdCommutativeBinOp<&Assembler::vpmaddwd, &Assembler::pmaddwd>(
+      this, dst, lhs, rhs);
+}
+
 void LiftoffAssembler::emit_i64x2_neg(LiftoffRegister dst,
                                       LiftoffRegister src) {
   DoubleRegister reg =

src/wasm/baseline/liftoff-assembler.h

@@ -940,6 +940,8 @@ class LiftoffAssembler : public TurboAssembler {
                                LiftoffRegister rhs);
   inline void emit_i32x4_max_u(LiftoffRegister dst, LiftoffRegister lhs,
                                LiftoffRegister rhs);
+  inline void emit_i32x4_dot_i16x8_s(LiftoffRegister dst, LiftoffRegister lhs,
+                                     LiftoffRegister rhs);
   inline void emit_i64x2_neg(LiftoffRegister dst, LiftoffRegister src);
   inline void emit_i64x2_shl(LiftoffRegister dst, LiftoffRegister lhs,
                              LiftoffRegister rhs);

src/wasm/baseline/liftoff-compiler.cc

@@ -2758,6 +2758,9 @@ class LiftoffCompiler {
         return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_i32x4_max_s);
       case wasm::kExprI32x4MaxU:
         return EmitBinOp<kS128, kS128>(&LiftoffAssembler::emit_i32x4_max_u);
+      case wasm::kExprI32x4DotI16x8S:
+        return EmitBinOp<kS128, kS128>(
+            &LiftoffAssembler::emit_i32x4_dot_i16x8_s);
       case wasm::kExprI64x2Neg:
         return EmitUnOp<kS128, kS128>(&LiftoffAssembler::emit_i64x2_neg);
       case wasm::kExprI64x2Shl:

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

@@ -3204,6 +3204,13 @@ void LiftoffAssembler::emit_i32x4_max_u(LiftoffRegister dst,
       this, dst, lhs, rhs, base::Optional<CpuFeature>(SSE4_1));
 }
 
+void LiftoffAssembler::emit_i32x4_dot_i16x8_s(LiftoffRegister dst,
+                                              LiftoffRegister lhs,
+                                              LiftoffRegister rhs) {
+  liftoff::EmitSimdCommutativeBinOp<&Assembler::vpmaddwd, &Assembler::pmaddwd>(
+      this, dst, lhs, rhs);
+}
+
 void LiftoffAssembler::emit_i64x2_neg(LiftoffRegister dst,
                                       LiftoffRegister src) {
   DoubleRegister reg = dst.fp() == src.fp() ? kScratchDoubleReg : dst.fp();