[wasm-simd][scalar-lowering] Lower floating-point rounding ops

Missed this earlier when it was merged into the proposal.

f32x4 and f64x2 ceil, floor, trunc, nearestint. Also enable cctests.

Bug: v8:10507,v8:10906
Change-Id: I2de00e615cd63d81303649774db2a2ab800f6f72
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2461451Reviewed-by: Bill Budge <bbudge@chromium.org>
Commit-Queue: Zhi An Ng <zhin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#70415}

src/compiler/simd-scalar-lowering.cc

@@ -178,7 +178,11 @@ void SimdScalarLowering::LowerGraph() {
   V(F64x2Mul)                       \
   V(F64x2Div)                       \
   V(F64x2Min)                       \
-  V(F64x2Max)
+  V(F64x2Max)                       \
+  V(F64x2Ceil)                      \
+  V(F64x2Floor)                     \
+  V(F64x2Trunc)                     \
+  V(F64x2NearestInt)
 
 #define FOREACH_FLOAT32X4_OPCODE(V) \
   V(F32x4Splat)                     \
@@ -197,7 +201,11 @@ void SimdScalarLowering::LowerGraph() {
   V(F32x4Mul)                       \
   V(F32x4Div)                       \
   V(F32x4Min)                       \
-  V(F32x4Max)
+  V(F32x4Max)                       \
+  V(F32x4Ceil)                      \
+  V(F32x4Floor)                     \
+  V(F32x4Trunc)                     \
+  V(F32x4NearestInt)
 
 #define FOREACH_FLOAT64x2_TO_INT64x2OPCODE(V) \
   V(F64x2Eq)                                  \
@@ -1786,6 +1794,22 @@ void SimdScalarLowering::LowerNode(Node* node) {
       F32X4_UNOP_CASE(Neg)
       F32X4_UNOP_CASE(Sqrt)
 #undef F32X4_UNOP_CASE
+    case IrOpcode::kF32x4Ceil: {
+      LowerUnaryOp(node, rep_type, machine()->Float32RoundUp().op());
+      break;
+    }
+    case IrOpcode::kF32x4Floor: {
+      LowerUnaryOp(node, rep_type, machine()->Float32RoundDown().op());
+      break;
+    }
+    case IrOpcode::kF32x4Trunc: {
+      LowerUnaryOp(node, rep_type, machine()->Float32RoundTruncate().op());
+      break;
+    }
+    case IrOpcode::kF32x4NearestInt: {
+      LowerUnaryOp(node, rep_type, machine()->Float32RoundTiesEven().op());
+      break;
+    }
     case IrOpcode::kF32x4RecipApprox:
     case IrOpcode::kF32x4RecipSqrtApprox: {
       DCHECK_EQ(1, node->InputCount());
@@ -1846,6 +1870,22 @@ void SimdScalarLowering::LowerNode(Node* node) {
       LowerBinaryOp(node, rep_type, machine()->Float64Max());
       break;
     }
+    case IrOpcode::kF64x2Ceil: {
+      LowerUnaryOp(node, rep_type, machine()->Float64RoundUp().op());
+      break;
+    }
+    case IrOpcode::kF64x2Floor: {
+      LowerUnaryOp(node, rep_type, machine()->Float64RoundDown().op());
+      break;
+    }
+    case IrOpcode::kF64x2Trunc: {
+      LowerUnaryOp(node, rep_type, machine()->Float64RoundTruncate().op());
+      break;
+    }
+    case IrOpcode::kF64x2NearestInt: {
+      LowerUnaryOp(node, rep_type, machine()->Float64RoundTiesEven().op());
+      break;
+    }
     case IrOpcode::kF64x2Splat:
     case IrOpcode::kF32x4Splat:
     case IrOpcode::kI64x2Splat:

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

@@ -671,19 +671,19 @@ WASM_SIMD_TEST(F32x4RecipSqrtApprox) {
                    base::RecipSqrt, false /* !exact */);
 }
 
-WASM_SIMD_TEST_NO_LOWERING(F32x4Ceil) {
+WASM_SIMD_TEST(F32x4Ceil) {
   RunF32x4UnOpTest(execution_tier, lower_simd, kExprF32x4Ceil, ceilf, true);
 }
 
-WASM_SIMD_TEST_NO_LOWERING(F32x4Floor) {
+WASM_SIMD_TEST(F32x4Floor) {
   RunF32x4UnOpTest(execution_tier, lower_simd, kExprF32x4Floor, floorf, true);
 }
 
-WASM_SIMD_TEST_NO_LOWERING(F32x4Trunc) {
+WASM_SIMD_TEST(F32x4Trunc) {
   RunF32x4UnOpTest(execution_tier, lower_simd, kExprF32x4Trunc, truncf, true);
 }
 
-WASM_SIMD_TEST_NO_LOWERING(F32x4NearestInt) {
+WASM_SIMD_TEST(F32x4NearestInt) {
   RunF32x4UnOpTest(execution_tier, lower_simd, kExprF32x4NearestInt, nearbyintf,
                    true);
 }
@@ -1287,19 +1287,19 @@ WASM_SIMD_TEST(F64x2Sqrt) {
   RunF64x2UnOpTest(execution_tier, lower_simd, kExprF64x2Sqrt, Sqrt);
 }
 
-WASM_SIMD_TEST_NO_LOWERING(F64x2Ceil) {
+WASM_SIMD_TEST(F64x2Ceil) {
   RunF64x2UnOpTest(execution_tier, lower_simd, kExprF64x2Ceil, ceil, true);
 }
 
-WASM_SIMD_TEST_NO_LOWERING(F64x2Floor) {
+WASM_SIMD_TEST(F64x2Floor) {
   RunF64x2UnOpTest(execution_tier, lower_simd, kExprF64x2Floor, floor, true);
 }
 
-WASM_SIMD_TEST_NO_LOWERING(F64x2Trunc) {
+WASM_SIMD_TEST(F64x2Trunc) {
   RunF64x2UnOpTest(execution_tier, lower_simd, kExprF64x2Trunc, trunc, true);
 }
 
-WASM_SIMD_TEST_NO_LOWERING(F64x2NearestInt) {
+WASM_SIMD_TEST(F64x2NearestInt) {
   RunF64x2UnOpTest(execution_tier, lower_simd, kExprF64x2NearestInt, nearbyint,
                    true);
 }