[wasm-simd][arm] Prototype i64x2 widen i32x4 instructions

Prototype these 4 instructions:

- i64x2.widen_low_i32x4_s
- i64x2.widen_high_i32x4_s
- i64x2.widen_low_i32x4_u
- i64x2.widen_high_i32x4_u

Bug: v8:10972
Change-Id: I57508a7fcafdf3b8a9477d6e9292fbb6b67e3619
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2612342
Commit-Queue: Zhi An Ng <zhin@chromium.org>
Reviewed-by: Deepti Gandluri <gdeepti@chromium.org>
Cr-Commit-Position: refs/heads/master@{#72029}

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

@@ -2202,6 +2202,26 @@ CodeGenerator::CodeGenResult CodeGenerator::AssembleArchInstruction(
       __ add(dst, dst, Operand(tmp, LSL, 1));
       break;
     }
+    case kArmI64x2SConvertI32x4Low: {
+      __ vmovl(NeonS32, i.OutputSimd128Register(),
+               i.InputSimd128Register(0).low());
+      break;
+    }
+    case kArmI64x2SConvertI32x4High: {
+      __ vmovl(NeonS32, i.OutputSimd128Register(),
+               i.InputSimd128Register(0).high());
+      break;
+    }
+    case kArmI64x2UConvertI32x4Low: {
+      __ vmovl(NeonU32, i.OutputSimd128Register(),
+               i.InputSimd128Register(0).low());
+      break;
+    }
+    case kArmI64x2UConvertI32x4High: {
+      __ vmovl(NeonU32, i.OutputSimd128Register(),
+               i.InputSimd128Register(0).high());
+      break;
+    }
     case kArmF32x4Splat: {
       int src_code = i.InputFloatRegister(0).code();
       __ vdup(Neon32, i.OutputSimd128Register(),

src/compiler/backend/arm/instruction-codes-arm.h

@@ -188,6 +188,10 @@ namespace compiler {
   V(ArmI64x2ShrU)                  \
   V(ArmI64x2BitMask)               \
   V(ArmI64x2Eq)                    \
+  V(ArmI64x2SConvertI32x4Low)      \
+  V(ArmI64x2SConvertI32x4High)     \
+  V(ArmI64x2UConvertI32x4Low)      \
+  V(ArmI64x2UConvertI32x4High)     \
   V(ArmI32x4Splat)                 \
   V(ArmI32x4ExtractLane)           \
   V(ArmI32x4ReplaceLane)           \

src/compiler/backend/arm/instruction-scheduler-arm.cc

@@ -168,6 +168,10 @@ int InstructionScheduler::GetTargetInstructionFlags(
     case kArmI64x2ShrU:
     case kArmI64x2BitMask:
     case kArmI64x2Eq:
+    case kArmI64x2SConvertI32x4Low:
+    case kArmI64x2SConvertI32x4High:
+    case kArmI64x2UConvertI32x4Low:
+    case kArmI64x2UConvertI32x4High:
     case kArmI32x4Splat:
     case kArmI32x4ExtractLane:
     case kArmI32x4ReplaceLane:

src/compiler/backend/arm/instruction-selector-arm.cc

@@ -2605,6 +2605,10 @@ void InstructionSelector::VisitWord32AtomicPairCompareExchange(Node* node) {
   V(F32x4Neg, kArmF32x4Neg)                             \
   V(F32x4RecipApprox, kArmF32x4RecipApprox)             \
   V(F32x4RecipSqrtApprox, kArmF32x4RecipSqrtApprox)     \
+  V(I64x2SConvertI32x4Low, kArmI64x2SConvertI32x4Low)   \
+  V(I64x2SConvertI32x4High, kArmI64x2SConvertI32x4High) \
+  V(I64x2UConvertI32x4Low, kArmI64x2UConvertI32x4Low)   \
+  V(I64x2UConvertI32x4High, kArmI64x2UConvertI32x4High) \
   V(I32x4SConvertF32x4, kArmI32x4SConvertF32x4)         \
   V(I32x4SConvertI16x8Low, kArmI32x4SConvertI16x8Low)   \
   V(I32x4SConvertI16x8High, kArmI32x4SConvertI16x8High) \

src/compiler/backend/instruction-selector.cc

@@ -2748,7 +2748,8 @@ void InstructionSelector::VisitI64x2Eq(Node* node) { UNIMPLEMENTED(); }
         // && !V8_TARGET_ARCH_IA32 && !V8_TARGET_ARCH_ARM
         // && !V8_TARGET_ARCH_MIPS64 && !V8_TARGET_ARCH_MIPS
 
-#if !V8_TARGET_ARCH_ARM64 && !V8_TARGET_ARCH_X64 && !V8_TARGET_ARCH_IA32
+#if !V8_TARGET_ARCH_ARM64 && !V8_TARGET_ARCH_X64 && !V8_TARGET_ARCH_IA32 && \
+    !V8_TARGET_ARCH_ARM
 // TODO(v8:10972) Prototype i64x2 widen i32x4.
 void InstructionSelector::VisitI64x2SConvertI32x4Low(Node* node) {
   UNIMPLEMENTED();
@@ -2766,6 +2767,7 @@ void InstructionSelector::VisitI64x2UConvertI32x4High(Node* node) {
   UNIMPLEMENTED();
 }
 #endif  // !V8_TARGET_ARCH_ARM64 || !V8_TARGET_ARCH_X64 && !V8_TARGET_ARCH_IA32
+        // && !V8_TARGET_ARCH_ARM
 
 #if !V8_TARGET_ARCH_ARM64
 // TODO(v8:11168): Prototyping prefetch.

test/cctest/wasm/test-run-wasm-simd.cc

@@ -1806,7 +1806,8 @@ WASM_SIMD_TEST(I32x4ConvertI16x8) {
 
 // TODO(v8:10972) Prototyping i64x2 convert from i32x4.
 // Tests both signed and unsigned conversion from I32x4 (unpacking).
-#if V8_TARGET_ARCH_ARM64 || V8_TARGET_ARCH_X64 || V8_TARGET_ARCH_IA32
+#if V8_TARGET_ARCH_ARM64 || V8_TARGET_ARCH_X64 || V8_TARGET_ARCH_IA32 || \
+    V8_TARGET_ARCH_ARM
 WASM_SIMD_TEST_NO_LOWERING(I64x2ConvertI32x4) {
   FLAG_SCOPE(wasm_simd_post_mvp);
   WasmRunner<int32_t, int32_t> r(execution_tier, lower_simd);
@@ -1841,7 +1842,8 @@ WASM_SIMD_TEST_NO_LOWERING(I64x2ConvertI32x4) {
     }
   }
 }
-#endif  // V8_TARGET_ARCH_ARM64 || V8_TARGET_ARCH_X64 || V8_TARGET_ARCH_IA32
+#endif  // V8_TARGET_ARCH_ARM64 || V8_TARGET_ARCH_X64 || V8_TARGET_ARCH_IA32 ||
+        // V8_TARGET_ARCH_ARM
 
 void RunI32x4UnOpTest(TestExecutionTier execution_tier, LowerSimd lower_simd,
                       WasmOpcode opcode, Int32UnOp expected_op) {