[WASM SIMD] Replace primitive shuffles with general Shuffle.

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

Index: test/cctest/wasm/test-run-wasm-simd.cc
diff --git a/test/cctest/wasm/test-run-wasm-simd.cc b/test/cctest/wasm/test-run-wasm-simd.cc
index b9a9c8cb32665016b65295633b458bdd8cf8582e..7f36e213e0576e81d8cbe2e4c69e92e5522bff16 100644
--- a/test/cctest/wasm/test-run-wasm-simd.cc
+++ b/test/cctest/wasm/test-run-wasm-simd.cc
@@ -393,6 +393,20 @@ T RecipSqrt(T a) {
 #define WASM_SIMD_I8x16_REPLACE_LANE(lane, x, y) \
   x, y, WASM_SIMD_OP(kExprI8x16ReplaceLane), TO_BYTE(lane)
 
+#define WASM_SIMD_S32x4_SHUFFLE_OP(opcode, m, x, y)                        \
+  x, y, WASM_SIMD_OP(opcode), TO_BYTE(m[0]), TO_BYTE(m[1]), TO_BYTE(m[2]), \
+      TO_BYTE(m[3])
+#define WASM_SIMD_S16x8_SHUFFLE_OP(opcode, m, x, y)                        \
+  x, y, WASM_SIMD_OP(opcode), TO_BYTE(m[0]), TO_BYTE(m[1]), TO_BYTE(m[2]), \
+      TO_BYTE(m[3]), TO_BYTE(m[4]), TO_BYTE(m[5]), TO_BYTE(m[6]),          \
+      TO_BYTE(m[7])
+#define WASM_SIMD_S8x16_SHUFFLE_OP(opcode, m, x, y)                        \
+  x, y, WASM_SIMD_OP(opcode), TO_BYTE(m[0]), TO_BYTE(m[1]), TO_BYTE(m[2]), \
+      TO_BYTE(m[3]), TO_BYTE(m[4]), TO_BYTE(m[5]), TO_BYTE(m[6]),          \
+      TO_BYTE(m[7]), TO_BYTE(m[8]), TO_BYTE(m[9]), TO_BYTE(m[10]),         \
+      TO_BYTE(m[11]), TO_BYTE(m[12]), TO_BYTE(m[13]), TO_BYTE(m[14]),      \
+      TO_BYTE(m[15])
+
 // Skip FP tests involving extremely large or extremely small values, which
 // may fail due to non-IEEE-754 SIMD arithmetic on some platforms.
 bool SkipFPValue(float x) {
@@ -1570,57 +1584,6 @@ WASM_SIMD_SELECT_TEST(16x8)
 WASM_SIMD_SELECT_TEST(8x16)
 #endif  // V8_TARGET_ARCH_ARM || V8_TARGET_ARCH_X64
 
-#if V8_TARGET_ARCH_ARM
-// Test unary ops with a lane test pattern, all lanes distinct.
-template <typename T>
-void RunUnaryLaneOpTest(
-    WasmOpcode simd_op,
-    const std::array<T, kSimd128Size / sizeof(T)>& expected) {
-  FLAG_wasm_simd_prototype = true;
-  WasmRunner<int32_t> r(kExecuteCompiled);
-  // Set up a test pattern as a global, e.g. [0, 1, 2, 3].
-  T* global = r.module().AddGlobal<T>(kWasmS128);
-  static const size_t kElems = kSimd128Size / sizeof(T);
-  for (size_t i = 0; i < kElems; i++) {
-    global[i] = i;
-  }
-  BUILD(r, WASM_SET_GLOBAL(0, WASM_SIMD_UNOP(simd_op, WASM_GET_GLOBAL(0))),
-        WASM_ONE);
-
-  CHECK_EQ(1, r.Call());
-  for (size_t i = 0; i < kElems; i++) {
-    CHECK_EQ(global[i], expected[i]);
-  }
-}
-
-WASM_EXEC_COMPILED_TEST(S32x2Reverse) {
-  RunUnaryLaneOpTest<int32_t>(kExprS32x2Reverse, {{1, 0, 3, 2}});
-}
-
-WASM_EXEC_COMPILED_TEST(S16x4Reverse) {
-  RunUnaryLaneOpTest<int16_t>(kExprS16x4Reverse, {{3, 2, 1, 0, 7, 6, 5, 4}});
-}
-
-WASM_EXEC_COMPILED_TEST(S16x2Reverse) {
-  RunUnaryLaneOpTest<int16_t>(kExprS16x2Reverse, {{1, 0, 3, 2, 5, 4, 7, 6}});
-}
-
-WASM_EXEC_COMPILED_TEST(S8x8Reverse) {
-  RunUnaryLaneOpTest<int8_t>(kExprS8x8Reverse, {{7, 6, 5, 4, 3, 2, 1, 0, 15, 14,
-                                                 13, 12, 11, 10, 9, 8}});
-}
-
-WASM_EXEC_COMPILED_TEST(S8x4Reverse) {
-  RunUnaryLaneOpTest<int8_t>(kExprS8x4Reverse, {{3, 2, 1, 0, 7, 6, 5, 4, 11, 10,
-                                                 9, 8, 15, 14, 13, 12}});
-}
-
-WASM_EXEC_COMPILED_TEST(S8x2Reverse) {
-  RunUnaryLaneOpTest<int8_t>(kExprS8x2Reverse, {{1, 0, 3, 2, 5, 4, 7, 6, 9, 8,
-                                                 11, 10, 13, 12, 15, 14}});
-}
-#endif  // V8_TARGET_ARCH_ARM
-
 #if V8_TARGET_ARCH_ARM || V8_TARGET_ARCH_X64
 // Test binary ops with two lane test patterns, all lanes distinct.
 template <typename T>
@@ -1630,21 +1593,50 @@ void RunBinaryLaneOpTest(
   FLAG_wasm_simd_prototype = true;
   WasmRunner<int32_t> r(kExecuteCompiled);
   // Set up two test patterns as globals, e.g. [0, 1, 2, 3] and [4, 5, 6, 7].
-  T* global1 = r.module().AddGlobal<T>(kWasmS128);
-  T* global2 = r.module().AddGlobal<T>(kWasmS128);
-  static const size_t kElems = kSimd128Size / sizeof(T);
-  for (size_t i = 0; i < kElems; i++) {
-    global1[i] = static_cast<T>(i);
-    global2[i] = static_cast<T>(kElems + i);
+  T* src0 = r.module().AddGlobal<T>(kWasmS128);
+  T* src1 = r.module().AddGlobal<T>(kWasmS128);
+  static const int kElems = kSimd128Size / sizeof(T);
+  for (int i = 0; i < kElems; i++) {
+    src0[i] = i;
+    src1[i] = kElems + i;
+  }
+  switch (simd_op) {
+    case kExprS32x4Shuffle: {
+      BUILD(r,
+            WASM_SET_GLOBAL(0, WASM_SIMD_S32x4_SHUFFLE_OP(simd_op, expected,
+                                                          WASM_GET_GLOBAL(0),
+                                                          WASM_GET_GLOBAL(1))),
+            WASM_ONE);
+      break;
+    }
+    case kExprS16x8Shuffle: {
+      BUILD(r,
+            WASM_SET_GLOBAL(0, WASM_SIMD_S16x8_SHUFFLE_OP(simd_op, expected,
+                                                          WASM_GET_GLOBAL(0),
+                                                          WASM_GET_GLOBAL(1))),
+            WASM_ONE);
+      break;
+    }
+    case kExprS8x16Shuffle: {
+      BUILD(r,
+            WASM_SET_GLOBAL(0, WASM_SIMD_S8x16_SHUFFLE_OP(simd_op, expected,
+                                                          WASM_GET_GLOBAL(0),
+                                                          WASM_GET_GLOBAL(1))),
+            WASM_ONE);
+      break;
+    }
+    default: {
+      BUILD(r,
+            WASM_SET_GLOBAL(0, WASM_SIMD_BINOP(simd_op, WASM_GET_GLOBAL(0),
+                                               WASM_GET_GLOBAL(1))),
+            WASM_ONE);
+      break;
+    }
   }
-  BUILD(r,
-        WASM_SET_GLOBAL(0, WASM_SIMD_BINOP(simd_op, WASM_GET_GLOBAL(0),
-                                           WASM_GET_GLOBAL(1))),
-        WASM_ONE);
 
   CHECK_EQ(1, r.Call());
   for (size_t i = 0; i < expected.size(); i++) {
-    CHECK_EQ(global1[i], expected[i]);
+    CHECK_EQ(src0[i], expected[i]);
   }
 }
 
@@ -1663,136 +1655,181 @@ WASM_EXEC_COMPILED_TEST(F32x4AddHoriz) {
   RunBinaryLaneOpTest<float>(kExprF32x4AddHoriz, {{1.0f, 5.0f, 9.0f, 13.0f}});
 }
 
+// Test some regular shuffles that may have special handling on some targets.
+// Test a normal and unary versions (where second operand isn't used).
 WASM_EXEC_COMPILED_TEST(S32x4ZipLeft) {
-  RunBinaryLaneOpTest<int32_t>(kExprS32x4ZipLeft, {{0, 4, 1, 5}});
+  RunBinaryLaneOpTest<int32_t>(kExprS32x4Shuffle, {{0, 4, 1, 5}});
+  RunBinaryLaneOpTest<int32_t>(kExprS32x4Shuffle, {{0, 0, 1, 1}});
 }
 
 WASM_EXEC_COMPILED_TEST(S32x4ZipRight) {
-  RunBinaryLaneOpTest<int32_t>(kExprS32x4ZipRight, {{2, 6, 3, 7}});
+  RunBinaryLaneOpTest<int32_t>(kExprS32x4Shuffle, {{2, 6, 3, 7}});
+  RunBinaryLaneOpTest<int32_t>(kExprS32x4Shuffle, {{2, 2, 3, 3}});
 }
 
 WASM_EXEC_COMPILED_TEST(S32x4UnzipLeft) {
-  RunBinaryLaneOpTest<int32_t>(kExprS32x4UnzipLeft, {{0, 2, 4, 6}});
+  RunBinaryLaneOpTest<int32_t>(kExprS32x4Shuffle, {{0, 2, 4, 6}});
+  RunBinaryLaneOpTest<int32_t>(kExprS32x4Shuffle, {{0, 2, 0, 2}});
 }
 
 WASM_EXEC_COMPILED_TEST(S32x4UnzipRight) {
-  RunBinaryLaneOpTest<int32_t>(kExprS32x4UnzipRight, {{1, 3, 5, 7}});
+  RunBinaryLaneOpTest<int32_t>(kExprS32x4Shuffle, {{1, 3, 5, 7}});
+  RunBinaryLaneOpTest<int32_t>(kExprS32x4Shuffle, {{1, 3, 1, 3}});
 }
 
 WASM_EXEC_COMPILED_TEST(S32x4TransposeLeft) {
-  RunBinaryLaneOpTest<int32_t>(kExprS32x4TransposeLeft, {{0, 4, 2, 6}});
+  RunBinaryLaneOpTest<int32_t>(kExprS32x4Shuffle, {{0, 4, 2, 6}});
+  RunBinaryLaneOpTest<int32_t>(kExprS32x4Shuffle, {{0, 0, 2, 2}});
 }
 
 WASM_EXEC_COMPILED_TEST(S32x4TransposeRight) {
-  RunBinaryLaneOpTest<int32_t>(kExprS32x4TransposeRight, {{1, 5, 3, 7}});
+  RunBinaryLaneOpTest<int32_t>(kExprS32x4Shuffle, {{1, 5, 3, 7}});
+  RunBinaryLaneOpTest<int32_t>(kExprS32x4Shuffle, {{1, 1, 3, 3}});
+}
+
+// Reverses are only unary.
+WASM_EXEC_COMPILED_TEST(S32x2Reverse) {
+  RunBinaryLaneOpTest<int32_t>(kExprS32x4Shuffle, {{1, 0, 3, 2}});
 }
 
 WASM_EXEC_COMPILED_TEST(S16x8ZipLeft) {
-  RunBinaryLaneOpTest<int16_t>(kExprS16x8ZipLeft, {{0, 8, 1, 9, 2, 10, 3, 11}});
+  RunBinaryLaneOpTest<int16_t>(kExprS16x8Shuffle, {{0, 8, 1, 9, 2, 10, 3, 11}});
+  RunBinaryLaneOpTest<int16_t>(kExprS16x8Shuffle, {{0, 0, 1, 1, 2, 2, 3, 3}});
 }
 
 WASM_EXEC_COMPILED_TEST(S16x8ZipRight) {
-  RunBinaryLaneOpTest<int16_t>(kExprS16x8ZipRight,
+  RunBinaryLaneOpTest<int16_t>(kExprS16x8Shuffle,
                                {{4, 12, 5, 13, 6, 14, 7, 15}});
+  RunBinaryLaneOpTest<int16_t>(kExprS16x8Shuffle, {{4, 4, 5, 5, 6, 6, 7, 7}});
 }
 
 WASM_EXEC_COMPILED_TEST(S16x8UnzipLeft) {
-  RunBinaryLaneOpTest<int16_t>(kExprS16x8UnzipLeft,
+  RunBinaryLaneOpTest<int16_t>(kExprS16x8Shuffle,
                                {{0, 2, 4, 6, 8, 10, 12, 14}});
+  RunBinaryLaneOpTest<int16_t>(kExprS16x8Shuffle, {{0, 2, 4, 6, 0, 2, 4, 6}});
 }
 
 WASM_EXEC_COMPILED_TEST(S16x8UnzipRight) {
-  RunBinaryLaneOpTest<int16_t>(kExprS16x8UnzipRight,
+  RunBinaryLaneOpTest<int16_t>(kExprS16x8Shuffle,
                                {{1, 3, 5, 7, 9, 11, 13, 15}});
+  RunBinaryLaneOpTest<int16_t>(kExprS16x8Shuffle, {{1, 3, 5, 7, 1, 3, 5, 7}});
 }
 
 WASM_EXEC_COMPILED_TEST(S16x8TransposeLeft) {
-  RunBinaryLaneOpTest<int16_t>(kExprS16x8TransposeLeft,
+  RunBinaryLaneOpTest<int16_t>(kExprS16x8Shuffle,
                                {{0, 8, 2, 10, 4, 12, 6, 14}});
+  RunBinaryLaneOpTest<int16_t>(kExprS16x8Shuffle, {{0, 0, 2, 2, 4, 4, 6, 6}});
 }
 
 WASM_EXEC_COMPILED_TEST(S16x8TransposeRight) {
-  RunBinaryLaneOpTest<int16_t>(kExprS16x8TransposeRight,
+  RunBinaryLaneOpTest<int16_t>(kExprS16x8Shuffle,
                                {{1, 9, 3, 11, 5, 13, 7, 15}});
+  RunBinaryLaneOpTest<int16_t>(kExprS16x8Shuffle, {{1, 1, 3, 3, 5, 5, 7, 7}});
+}
+
+WASM_EXEC_COMPILED_TEST(S16x4Reverse) {
+  RunBinaryLaneOpTest<int16_t>(kExprS16x8Shuffle, {{3, 2, 1, 0, 7, 6, 5, 4}});
+}
+
+WASM_EXEC_COMPILED_TEST(S16x2Reverse) {
+  RunBinaryLaneOpTest<int16_t>(kExprS16x8Shuffle, {{1, 0, 3, 2, 5, 4, 7, 6}});
 }
 
 WASM_EXEC_COMPILED_TEST(S8x16ZipLeft) {
   RunBinaryLaneOpTest<int8_t>(
-      kExprS8x16ZipLeft,
+      kExprS8x16Shuffle,
       {{0, 16, 1, 17, 2, 18, 3, 19, 4, 20, 5, 21, 6, 22, 7, 23}});
+  RunBinaryLaneOpTest<int8_t>(
+      kExprS8x16Shuffle, {{0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7}});
 }
 
 WASM_EXEC_COMPILED_TEST(S8x16ZipRight) {
   RunBinaryLaneOpTest<int8_t>(
-      kExprS8x16ZipRight,
+      kExprS8x16Shuffle,
       {{8, 24, 9, 25, 10, 26, 11, 27, 12, 28, 13, 29, 14, 30, 15, 31}});
+  RunBinaryLaneOpTest<int8_t>(
+      kExprS8x16Shuffle,
+      {{8, 8, 9, 9, 10, 10, 11, 11, 12, 12, 13, 13, 14, 14, 15, 15}});
 }
 
 WASM_EXEC_COMPILED_TEST(S8x16UnzipLeft) {
   RunBinaryLaneOpTest<int8_t>(
-      kExprS8x16UnzipLeft,
+      kExprS8x16Shuffle,
       {{0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30}});
+  RunBinaryLaneOpTest<int8_t>(kExprS8x16Shuffle, {{0, 2, 4, 6, 8, 10, 12, 14, 0,
+                                                   2, 4, 6, 8, 10, 12, 14}});
 }
 
 WASM_EXEC_COMPILED_TEST(S8x16UnzipRight) {
   RunBinaryLaneOpTest<int8_t>(
-      kExprS8x16UnzipRight,
+      kExprS8x16Shuffle,
       {{1, 3, 5, 7, 9, 11, 13, 15, 17, 19, 21, 23, 25, 27, 29, 31}});
+  RunBinaryLaneOpTest<int8_t>(kExprS8x16Shuffle, {{1, 3, 5, 7, 9, 11, 13, 15, 1,
+                                                   3, 5, 7, 9, 11, 13, 15}});
 }
 
 WASM_EXEC_COMPILED_TEST(S8x16TransposeLeft) {
   RunBinaryLaneOpTest<int8_t>(
-      kExprS8x16TransposeLeft,
+      kExprS8x16Shuffle,
       {{0, 16, 2, 18, 4, 20, 6, 22, 8, 24, 10, 26, 12, 28, 14, 30}});
+  RunBinaryLaneOpTest<int8_t>(kExprS8x16Shuffle, {{0, 0, 2, 2, 4, 4, 6, 6, 8, 8,
+                                                   10, 10, 12, 12, 14, 14}});
 }
 
 WASM_EXEC_COMPILED_TEST(S8x16TransposeRight) {
   RunBinaryLaneOpTest<int8_t>(
-      kExprS8x16TransposeRight,
+      kExprS8x16Shuffle,
       {{1, 17, 3, 19, 5, 21, 7, 23, 9, 25, 11, 27, 13, 29, 15, 31}});
+  RunBinaryLaneOpTest<int8_t>(kExprS8x16Shuffle, {{1, 1, 3, 3, 5, 5, 7, 7, 9, 9,
+                                                   11, 11, 13, 13, 15, 15}});
 }
 
-template <typename T>
-void RunConcatOpTest(WasmOpcode simd_op, int bytes,
-                     const std::array<T, kSimd128Size / sizeof(T)>& expected) {
-  FLAG_wasm_simd_prototype = true;
-  WasmRunner<int32_t> r(kExecuteCompiled);
-  // Set up two test patterns as globals, e.g. [0, 1, 2, 3] and [4, 5, 6, 7].
-  T* global1 = r.module().AddGlobal<T>(kWasmS128);
-  T* global2 = r.module().AddGlobal<T>(kWasmS128);
-  static const size_t kElems = kSimd128Size / sizeof(T);
-  for (size_t i = 0; i < kElems; i++) {
-    global1[i] = i;
-    global2[i] = kElems + i;
-  }
-  BUILD(
-      r,
-      WASM_SET_GLOBAL(0, WASM_SIMD_CONCAT_OP(simd_op, bytes, WASM_GET_GLOBAL(0),
-                                             WASM_GET_GLOBAL(1))),
-      WASM_ONE);
+WASM_EXEC_COMPILED_TEST(S8x8Reverse) {
+  RunBinaryLaneOpTest<int8_t>(kExprS8x16Shuffle, {{7, 6, 5, 4, 3, 2, 1, 0, 15,
+                                                   14, 13, 12, 11, 10, 9, 8}});
+}
 
-  CHECK_EQ(1, r.Call());
-  for (size_t i = 0; i < expected.size(); i++) {
-    CHECK_EQ(global1[i], expected[i]);
-  }
+WASM_EXEC_COMPILED_TEST(S8x4Reverse) {
+  RunBinaryLaneOpTest<int8_t>(kExprS8x16Shuffle, {{3, 2, 1, 0, 7, 6, 5, 4, 11,
+                                                   10, 9, 8, 15, 14, 13, 12}});
 }
 
-WASM_EXEC_COMPILED_TEST(S8x16Concat) {
-  std::array<int8_t, kSimd128Size> expected;
-  for (int k = 1; k < 16; k++) {
-    int j = 0;
-    // last 16 - k bytes of first vector.
-    for (int i = k; i < kSimd128Size; i++) {
-      expected[j++] = i;
+WASM_EXEC_COMPILED_TEST(S8x2Reverse) {
+  RunBinaryLaneOpTest<int8_t>(kExprS8x16Shuffle, {{1, 0, 3, 2, 5, 4, 7, 6, 9, 8,
+                                                   11, 10, 13, 12, 15, 14}});
+}
+
+// Test shuffles that concatenate the two vectors.
+template <typename T>
+void RunConcatOpTest(WasmOpcode simd_op) {
+  static const int kLanes = kSimd128Size / sizeof(T);
+  std::array<T, kLanes> expected;
+  for (int bias = 1; bias < kLanes; bias++) {
+    int i = 0;
+    // last kLanes - bias bytes of first vector.
+    for (int j = bias; j < kLanes; j++) {
+      expected[i++] = j;
     }
-    // first k bytes of second vector
-    for (int i = 0; i < k; i++) {
-      expected[j++] = i + kSimd128Size;
+    // first bias lanes of second vector
+    for (int j = 0; j < bias; j++) {
+      expected[i++] = j + kLanes;
     }
-    RunConcatOpTest<int8_t>(kExprS8x16Concat, k, expected);
+    RunBinaryLaneOpTest<T>(simd_op, expected);
   }
 }
 
+WASM_EXEC_COMPILED_TEST(S32x4Concat) {
+  RunConcatOpTest<int32_t>(kExprS32x4Shuffle);
+}
+
+WASM_EXEC_COMPILED_TEST(S16x8Concat) {
+  RunConcatOpTest<int16_t>(kExprS16x8Shuffle);
+}
+
+WASM_EXEC_COMPILED_TEST(S8x16Concat) {
+  RunConcatOpTest<int8_t>(kExprS8x16Shuffle);
+}
+
 // Boolean unary operations are 'AllTrue' and 'AnyTrue', which return an integer
 // result. Use relational ops on numeric vectors to create the boolean vector
 // test inputs. Test inputs with all true, all false, one true, and one false.