S390 [lifotff]: Implement simd splat ops

Shared ops between TurboFan and Liftoff are moved into
the macro-assembler.

Change-Id: I03cd3af10074b6b4666a7d2a13e652629576f76f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3035764Reviewed-by: Junliang Yan <junyan@redhat.com>
Reviewed-by: Milad Fa <mfarazma@redhat.com>
Commit-Queue: Milad Fa <mfarazma@redhat.com>
Cr-Commit-Position: refs/heads/master@{#75768}

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

@@ -5086,6 +5086,35 @@ void TurboAssembler::AtomicExchangeU16(Register addr, Register value,
   bind(&done);
 }
 
+// Simd Support.
+void TurboAssembler::F64x2Splat(Simd128Register dst, Simd128Register src) {
+  vrep(dst, src, Operand(0), Condition(3));
+}
+
+void TurboAssembler::F32x4Splat(Simd128Register dst, Simd128Register src) {
+  vrep(dst, src, Operand(0), Condition(2));
+}
+
+void TurboAssembler::I64x2Splat(Simd128Register dst, Register src) {
+  vlvg(dst, src, MemOperand(r0, 0), Condition(3));
+  vrep(dst, dst, Operand(0), Condition(3));
+}
+
+void TurboAssembler::I32x4Splat(Simd128Register dst, Register src) {
+  vlvg(dst, src, MemOperand(r0, 0), Condition(2));
+  vrep(dst, dst, Operand(0), Condition(2));
+}
+
+void TurboAssembler::I16x8Splat(Simd128Register dst, Register src) {
+  vlvg(dst, src, MemOperand(r0, 0), Condition(1));
+  vrep(dst, dst, Operand(0), Condition(1));
+}
+
+void TurboAssembler::I8x16Splat(Simd128Register dst, Register src) {
+  vlvg(dst, src, MemOperand(r0, 0), Condition(0));
+  vrep(dst, dst, Operand(0), Condition(0));
+}
+
 }  // namespace internal
 }  // namespace v8
 

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

@@ -1030,6 +1030,15 @@ class V8_EXPORT_PRIVATE TurboAssembler : public TurboAssemblerBase {
   // The return address on the stack is used by frame iteration.
   void StoreReturnAddressAndCall(Register target);
 
+  // ---------------------------------------------------------------------------
+  // Simd Support.
+  void F64x2Splat(Simd128Register dst, Simd128Register src);
+  void F32x4Splat(Simd128Register dst, Simd128Register src);
+  void I64x2Splat(Simd128Register dst, Register src);
+  void I32x4Splat(Simd128Register dst, Register src);
+  void I16x8Splat(Simd128Register dst, Register src);
+  void I8x16Splat(Simd128Register dst, Register src);
+
   // ---------------------------------------------------------------------------
   // Pointer compression Support
 

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

@@ -2481,41 +2481,23 @@ CodeGenerator::CodeGenResult CodeGenerator::AssembleArchInstruction(
     case kS390_Word64AtomicCompareExchangeUint64:
       ASSEMBLE_ATOMIC64_COMP_EXCHANGE_WORD64();
       break;
-      // vector replicate element
-    case kS390_F64x2Splat: {
-      __ vrep(i.OutputSimd128Register(), i.InputDoubleRegister(0), Operand(0),
-              Condition(3));
-      break;
-    }
-    case kS390_F32x4Splat: {
-      __ vrep(i.OutputSimd128Register(), i.InputDoubleRegister(0), Operand(0),
-              Condition(2));
-      break;
-    }
-    case kS390_I64x2Splat: {
-      Simd128Register dst = i.OutputSimd128Register();
-      __ vlvg(dst, i.InputRegister(0), MemOperand(r0, 0), Condition(3));
-      __ vrep(dst, dst, Operand(0), Condition(3));
-      break;
-    }
-    case kS390_I32x4Splat: {
-      Simd128Register dst = i.OutputSimd128Register();
-      __ vlvg(dst, i.InputRegister(0), MemOperand(r0, 0), Condition(2));
-      __ vrep(dst, dst, Operand(0), Condition(2));
-      break;
-    }
-    case kS390_I16x8Splat: {
-      Simd128Register dst = i.OutputSimd128Register();
-      __ vlvg(dst, i.InputRegister(0), MemOperand(r0, 0), Condition(1));
-      __ vrep(dst, dst, Operand(0), Condition(1));
-      break;
-    }
-    case kS390_I8x16Splat: {
-      Simd128Register dst = i.OutputSimd128Register();
-      __ vlvg(dst, i.InputRegister(0), MemOperand(r0, 0), Condition(0));
-      __ vrep(dst, dst, Operand(0), Condition(0));
-      break;
-    }
+      // Simd Support.
+#define SIMD_UNOP_LIST(V)                                    \
+  V(F64x2Splat, F64x2Splat, Simd128Register, DoubleRegister) \
+  V(F32x4Splat, F32x4Splat, Simd128Register, DoubleRegister) \
+  V(I64x2Splat, I64x2Splat, Simd128Register, Register)       \
+  V(I32x4Splat, I32x4Splat, Simd128Register, Register)       \
+  V(I16x8Splat, I16x8Splat, Simd128Register, Register)       \
+  V(I8x16Splat, I8x16Splat, Simd128Register, Register)
+
+#define EMIT_SIMD_UNOP(name, op, dtype, stype)   \
+  case kS390_##name: {                           \
+    __ op(i.Output##dtype(), i.Input##stype(0)); \
+    break;                                       \
+  }
+      SIMD_UNOP_LIST(EMIT_SIMD_UNOP)
+#undef EMIT_SIMD_UNOP
+#undef SIMD_UNOP_LIST
     // vector extract element
     case kS390_F64x2ExtractLane: {
       __ vrep(i.OutputDoubleRegister(), i.InputSimd128Register(0),

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

@@ -2142,6 +2142,23 @@ void LiftoffAssembler::emit_smi_check(Register obj, Label* target,
   b(condition, target);  // branch if SMI
 }
 
+#define SIMD_UNOP_LIST(V)            \
+  V(f64x2_splat, F64x2Splat, fp, fp) \
+  V(f32x4_splat, F32x4Splat, fp, fp) \
+  V(i64x2_splat, I64x2Splat, fp, gp) \
+  V(i32x4_splat, I32x4Splat, fp, gp) \
+  V(i16x8_splat, I16x8Splat, fp, gp) \
+  V(i8x16_splat, I8x16Splat, fp, gp)
+
+#define EMIT_SIMD_UNOP(name, op, dtype, stype)              \
+  void LiftoffAssembler::emit_##name(LiftoffRegister dst,   \
+                                     LiftoffRegister src) { \
+    op(dst.dtype(), src.stype());                           \
+  }
+SIMD_UNOP_LIST(EMIT_SIMD_UNOP)
+#undef EMIT_SIMD_UNOP
+#undef SIMD_UNOP_LIST
+
 void LiftoffAssembler::LoadTransform(LiftoffRegister dst, Register src_addr,
                                      Register offset_reg, uintptr_t offset_imm,
                                      LoadType type,
@@ -2170,11 +2187,6 @@ void LiftoffAssembler::emit_i8x16_swizzle(LiftoffRegister dst,
   bailout(kUnsupportedArchitecture, "emit_i8x16_swizzle");
 }
 
-void LiftoffAssembler::emit_f64x2_splat(LiftoffRegister dst,
-                                        LiftoffRegister src) {
-  bailout(kUnsupportedArchitecture, "emit_f64x2splat");
-}
-
 void LiftoffAssembler::emit_f64x2_extract_lane(LiftoffRegister dst,
                                                LiftoffRegister lhs,
                                                uint8_t imm_lane_idx) {
@@ -2282,11 +2294,6 @@ void LiftoffAssembler::emit_f64x2_promote_low_f32x4(LiftoffRegister dst,
   bailout(kSimd, "f64x2.promote_low_f32x4");
 }
 
-void LiftoffAssembler::emit_f32x4_splat(LiftoffRegister dst,
-                                        LiftoffRegister src) {
-  bailout(kUnsupportedArchitecture, "emit_f32x4_splat");
-}
-
 void LiftoffAssembler::emit_f32x4_extract_lane(LiftoffRegister dst,
                                                LiftoffRegister lhs,
                                                uint8_t imm_lane_idx) {
@@ -2379,11 +2386,6 @@ void LiftoffAssembler::emit_f32x4_pmax(LiftoffRegister dst, LiftoffRegister lhs,
   bailout(kSimd, "pmax unimplemented");
 }
 
-void LiftoffAssembler::emit_i64x2_splat(LiftoffRegister dst,
-                                        LiftoffRegister src) {
-  bailout(kUnsupportedArchitecture, "emit_i64x2splat");
-}
-
 void LiftoffAssembler::emit_i64x2_extract_lane(LiftoffRegister dst,
                                                LiftoffRegister lhs,
                                                uint8_t imm_lane_idx) {
@@ -2503,11 +2505,6 @@ void LiftoffAssembler::emit_i64x2_extmul_high_i32x4_u(LiftoffRegister dst,
   bailout(kSimd, "i64x2_extmul_high_i32x4_u unsupported");
 }
 
-void LiftoffAssembler::emit_i32x4_splat(LiftoffRegister dst,
-                                        LiftoffRegister src) {
-  bailout(kUnsupportedArchitecture, "emit_i32x4_splat");
-}
-
 void LiftoffAssembler::emit_i32x4_extract_lane(LiftoffRegister dst,
                                                LiftoffRegister lhs,
                                                uint8_t imm_lane_idx) {
@@ -2647,11 +2644,6 @@ void LiftoffAssembler::emit_i32x4_extmul_high_i16x8_u(LiftoffRegister dst,
   bailout(kSimd, "i32x4_extmul_high_i16x8_u unsupported");
 }
 
-void LiftoffAssembler::emit_i16x8_splat(LiftoffRegister dst,
-                                        LiftoffRegister src) {
-  bailout(kUnsupportedArchitecture, "emit_i16x8splat");
-}
-
 void LiftoffAssembler::emit_i16x8_neg(LiftoffRegister dst,
                                       LiftoffRegister src) {
   bailout(kUnsupportedArchitecture, "emit_i16x8neg");
@@ -2834,11 +2826,6 @@ void LiftoffAssembler::emit_i8x16_popcnt(LiftoffRegister dst,
   bailout(kSimd, "i8x16.popcnt");
 }
 
-void LiftoffAssembler::emit_i8x16_splat(LiftoffRegister dst,
-                                        LiftoffRegister src) {
-  bailout(kUnsupportedArchitecture, "emit_i8x16splat");
-}
-
 void LiftoffAssembler::emit_i8x16_extract_lane_u(LiftoffRegister dst,
                                                  LiftoffRegister lhs,
                                                  uint8_t imm_lane_idx) {