[mips][wasm-simd][liftoff] Implement ge of i8x16, i16x8 and i32x4 and bitselect

Port 97a4b795
https://crrev.com/c/2162366

Port f8be9948
https://crrev.com/c/2166959

Change-Id: Id0cebfe0761ff75fb95c9ceacf95e53fee12eb73
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2168978Reviewed-by: Clemens Backes <clemensb@chromium.org>
Commit-Queue: Zhao Jiazhong <zhaojiazhong-hf@loongson.cn>
Cr-Commit-Position: refs/heads/master@{#67434}

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

@@ -1594,6 +1594,16 @@ void LiftoffAssembler::emit_i8x16_ne(LiftoffRegister dst, LiftoffRegister lhs,
   bailout(kSimd, "emit_i8x16_ne");
 }
 
+void LiftoffAssembler::emit_i8x16_ge_s(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  bailout(kSimd, "emit_i8x16_ge_s");
+}
+
+void LiftoffAssembler::emit_i8x16_ge_u(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  bailout(kSimd, "emit_i8x16_ge_u");
+}
+
 void LiftoffAssembler::emit_i16x8_eq(LiftoffRegister dst, LiftoffRegister lhs,
                                      LiftoffRegister rhs) {
   // TODO(mips): Support this on loongson 3a4000. Currently, the main MIPS
@@ -1607,6 +1617,16 @@ void LiftoffAssembler::emit_i16x8_ne(LiftoffRegister dst, LiftoffRegister lhs,
   bailout(kSimd, "emit_i16x8_ne");
 }
 
+void LiftoffAssembler::emit_i16x8_ge_s(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  bailout(kSimd, "emit_i16x8_ge_s");
+}
+
+void LiftoffAssembler::emit_i16x8_ge_u(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  bailout(kSimd, "emit_i16x8_ge_u");
+}
+
 void LiftoffAssembler::emit_i32x4_eq(LiftoffRegister dst, LiftoffRegister lhs,
                                      LiftoffRegister rhs) {
   // TODO(mips): Support this on loongson 3a4000. Currently, the main MIPS
@@ -1620,6 +1640,16 @@ void LiftoffAssembler::emit_i32x4_ne(LiftoffRegister dst, LiftoffRegister lhs,
   bailout(kSimd, "emit_i32x4_ne");
 }
 
+void LiftoffAssembler::emit_i32x4_ge_s(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  bailout(kSimd, "emit_i32x4_ge_s");
+}
+
+void LiftoffAssembler::emit_i32x4_ge_u(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  bailout(kSimd, "emit_i32x4_ge_u");
+}
+
 void LiftoffAssembler::emit_f32x4_eq(LiftoffRegister dst, LiftoffRegister lhs,
                                      LiftoffRegister rhs) {
   // TODO(mips): Support this on loongson 3a4000. Currently, the main MIPS
@@ -1706,6 +1736,13 @@ void LiftoffAssembler::emit_s128_and_not(LiftoffRegister dst,
   bailout(kSimd, "emit_s128_and_not");
 }
 
+void LiftoffAssembler::emit_s128_select(LiftoffRegister dst,
+                                        LiftoffRegister src1,
+                                        LiftoffRegister src2,
+                                        LiftoffRegister src3) {
+  bailout(kSimd, "emit_s128_select");
+}
+
 void LiftoffAssembler::emit_i8x16_neg(LiftoffRegister dst,
                                       LiftoffRegister src) {
   // TODO(mips): Support this on loongson 3a4000. Currently, the main MIPS

src/wasm/baseline/mips64/liftoff-assembler-mips64.h

@@ -1412,6 +1412,16 @@ void LiftoffAssembler::emit_i8x16_ne(LiftoffRegister dst, LiftoffRegister lhs,
   bailout(kSimd, "emit_i8x16_ne");
 }
 
+void LiftoffAssembler::emit_i8x16_ge_s(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  cle_s_b(dst.fp().toW(), rhs.fp().toW(), lhs.fp().toW());
+}
+
+void LiftoffAssembler::emit_i8x16_ge_u(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  cle_u_b(dst.fp().toW(), rhs.fp().toW(), lhs.fp().toW());
+}
+
 void LiftoffAssembler::emit_i16x8_eq(LiftoffRegister dst, LiftoffRegister lhs,
                                      LiftoffRegister rhs) {
   // TODO(mips): Support this on loongson 3a4000. Currently, the main MIPS
@@ -1428,6 +1438,16 @@ void LiftoffAssembler::emit_i16x8_ne(LiftoffRegister dst, LiftoffRegister lhs,
   bailout(kSimd, "emit_i16x8_ne");
 }
 
+void LiftoffAssembler::emit_i16x8_ge_s(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  cle_s_h(dst.fp().toW(), rhs.fp().toW(), lhs.fp().toW());
+}
+
+void LiftoffAssembler::emit_i16x8_ge_u(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  cle_u_h(dst.fp().toW(), rhs.fp().toW(), lhs.fp().toW());
+}
+
 void LiftoffAssembler::emit_i32x4_eq(LiftoffRegister dst, LiftoffRegister lhs,
                                      LiftoffRegister rhs) {
   // TODO(mips): Support this on loongson 3a4000. Currently, the main MIPS
@@ -1444,6 +1464,16 @@ void LiftoffAssembler::emit_i32x4_ne(LiftoffRegister dst, LiftoffRegister lhs,
   bailout(kSimd, "emit_i32x4_ne");
 }
 
+void LiftoffAssembler::emit_i32x4_ge_s(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  cle_s_w(dst.fp().toW(), rhs.fp().toW(), lhs.fp().toW());
+}
+
+void LiftoffAssembler::emit_i32x4_ge_u(LiftoffRegister dst, LiftoffRegister lhs,
+                                       LiftoffRegister rhs) {
+  cle_u_w(dst.fp().toW(), rhs.fp().toW(), lhs.fp().toW());
+}
+
 void LiftoffAssembler::emit_f32x4_eq(LiftoffRegister dst, LiftoffRegister lhs,
                                      LiftoffRegister rhs) {
   // TODO(mips): Support this on loongson 3a4000. Currently, the main MIPS
@@ -1548,6 +1578,19 @@ void LiftoffAssembler::emit_s128_and_not(LiftoffRegister dst,
   bailout(kSimd, "emit_s128_and_not");
 }
 
+void LiftoffAssembler::emit_s128_select(LiftoffRegister dst,
+                                        LiftoffRegister src1,
+                                        LiftoffRegister src2,
+                                        LiftoffRegister src3) {
+  if (dst == src3) {
+    bsel_v(dst.fp().toW(), src2.fp().toW(), src1.fp().toW());
+  } else {
+    xor_v(kSimd128ScratchReg, src1.fp().toW(), src2.fp().toW());
+    and_v(kSimd128ScratchReg, kSimd128ScratchReg, src3.fp().toW());
+    xor_v(dst.fp().toW(), kSimd128ScratchReg, src2.fp().toW());
+  }
+}
+
 void LiftoffAssembler::emit_i8x16_neg(LiftoffRegister dst,
                                       LiftoffRegister src) {
   // TODO(mips): Support this on loongson 3a4000. Currently, the main MIPS