S390 [liftoff]: Implement simd select

all_ones and all_zero is also moved to the macro assembler
which will be used later.

Change-Id: I72787c9bb217dd78f6fd8a818b95939e936bc68d
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3439916Reviewed-by: Junliang Yan <junyan@redhat.com>
Commit-Queue: Milad Farazmand <mfarazma@redhat.com>
Cr-Commit-Position: refs/heads/main@{#78982}

src/codegen/s390/macro-assembler-s390.cc

@@ -5281,6 +5281,19 @@ void TurboAssembler::S128Not(Simd128Register dst, Simd128Register src) {
   vno(dst, src, src, Condition(0), Condition(0), Condition(0));
 }
 
+void TurboAssembler::S128Zero(Simd128Register dst, Simd128Register src) {
+  vx(dst, src, src, Condition(0), Condition(0), Condition(0));
+}
+
+void TurboAssembler::S128AllOnes(Simd128Register dst, Simd128Register src) {
+  vceq(dst, src, src, Condition(0), Condition(3));
+}
+
+void TurboAssembler::S128Select(Simd128Register dst, Simd128Register src1,
+                                Simd128Register src2, Simd128Register mask) {
+  vsel(dst, src1, src2, mask, Condition(0), Condition(0));
+}
+
 #define SIMD_UNOP_LIST_VRR_A(V)             \
   V(F64x2Abs, vfpso, 2, 0, 3)               \
   V(F64x2Neg, vfpso, 0, 0, 3)               \

src/codegen/s390/macro-assembler-s390.h

@@ -1134,6 +1134,8 @@ class V8_EXPORT_PRIVATE TurboAssembler : public TurboAssemblerBase {
                           Simd128Register src2, Simd128Register scratch);
   void I8x16UConvertI16x8(Simd128Register dst, Simd128Register src1,
                           Simd128Register src2, Simd128Register scratch);
+  void S128Select(Simd128Register dst, Simd128Register src1,
+                  Simd128Register src2, Simd128Register mask);
 
 #define SIMD_UNOP_LIST(V)   \
   V(F64x2Abs)               \
@@ -1173,7 +1175,9 @@ class V8_EXPORT_PRIVATE TurboAssembler : public TurboAssemblerBase {
   V(I8x16Abs)               \
   V(I8x16Neg)               \
   V(I8x16Popcnt)            \
-  V(S128Not)
+  V(S128Not)                \
+  V(S128Zero)               \
+  V(S128AllOnes)
 
 #define PROTOTYPE_SIMD_UNOP(name) \
   void name(Simd128Register dst, Simd128Register src);

src/compiler/backend/s390/code-generator-s390.cc

@@ -2836,12 +2836,12 @@ CodeGenerator::CodeGenResult CodeGenerator::AssembleArchInstruction(
     }
     case kS390_S128Zero: {
       Simd128Register dst = i.OutputSimd128Register();
-      __ vx(dst, dst, dst, Condition(0), Condition(0), Condition(0));
+      __ S128Zero(dst, dst);
       break;
     }
     case kS390_S128AllOnes: {
       Simd128Register dst = i.OutputSimd128Register();
-      __ vceq(dst, dst, dst, Condition(0), Condition(3));
+      __ S128AllOnes(dst, dst);
       break;
     }
     case kS390_S128Select: {
@@ -2849,7 +2849,7 @@ CodeGenerator::CodeGenResult CodeGenerator::AssembleArchInstruction(
       Simd128Register mask = i.InputSimd128Register(0);
       Simd128Register src1 = i.InputSimd128Register(1);
       Simd128Register src2 = i.InputSimd128Register(2);
-      __ vsel(dst, src1, src2, mask, Condition(0), Condition(0));
+      __ S128Select(dst, src1, src2, mask);
       break;
     }
     // vector conversions

src/wasm/baseline/s390/liftoff-assembler-s390.h

@@ -2675,7 +2675,7 @@ void LiftoffAssembler::emit_s128_select(LiftoffRegister dst,
                                         LiftoffRegister src1,
                                         LiftoffRegister src2,
                                         LiftoffRegister mask) {
-  bailout(kUnsupportedArchitecture, "emit_s128select");
+  S128Select(dst.fp(), src1.fp(), src2.fp(), mask.fp());
 }
 
 void LiftoffAssembler::emit_i32x4_sconvert_f32x4(LiftoffRegister dst,