Option to truncate integers to 64 bits, part 1 (core VM changes)

runtime/tests/vm/dart/truncating_ints_test.dart

Index: runtime/tests/vm/dart/truncating_ints_test.dart
diff --git a/runtime/tests/vm/dart/truncating_ints_test.dart b/runtime/tests/vm/dart/truncating_ints_test.dart
new file mode 100644
index 0000000000000000000000000000000000000000..27807c4b92dd5560e7938c0938f1f8085d605dd1
--- /dev/null
+++ b/runtime/tests/vm/dart/truncating_ints_test.dart
@@ -0,0 +1,184 @@
+// Copyright (c) 2017, the Dart project authors.  Please see the AUTHORS file
+// for details. All rights reserved. Use of this source code is governed by a
+// BSD-style license that can be found in the LICENSE file.
+
+// VMOptions=--truncate-ints-to-64-bits --enable-inlining-annotations --optimization_counter_threshold=10 --no-use-osr --no-background-compilation

[siva, 2017/07/07 23:11:20]

Why is the option --no-use-osr significant here?

[alexmarkov, 2017/07/10 16:58:12]

It can be omitted. However, it helps to avoid re-compilation of code which
verifies the results of operations (test_add, test_sub, test_mul and test_shl),
while the function with operation is re-compiled. It ensures that the same code,
without any changes, verifies the results of both non-optimized and optimized
operations.

+
+// Test for truncating (wrap-around) integer arithmetic in --truncate-ints-to-64-bits mode.
+
+import "package:expect/expect.dart";
+
+const alwaysInline = "AlwaysInline";
+const neverInline = "NeverInline";
+
+@neverInline
+add_smi(var a, var b) => a + b;
+
+@neverInline
+add_mint(var a, var b) => a + b;
+
+@neverInline
+add_mint_consts() => 0x5000000000000000 + 0x6000000000000000;
+
+@neverInline
+test_add(var v2, var v3, var v3fxx, var v5fxx, var v7fxx, var n60xx) {
+  for (var i = 0; i < 20; i++) {
+    Expect.equals(5, add_smi(v2, v3));
+  }
+
+  // Trigger deoptimization and re-compilation
+  for (var i = 0; i < 20; i++) {
+    Expect.equals(0x4000000000000001, add_smi(v2, v3fxx));
+  }
+
+  for (var i = 0; i < 20; i++) {
+    Expect.equals(-1, add_mint(v5fxx, n60xx));
+  }
+
+  // Wrap-around
+  for (var i = 0; i < 20; i++) {
+    Expect.equals(-0x2000000000000002, add_mint(v5fxx, v7fxx));
+  }
+
+  // Constant folding
+  for (var i = 0; i < 20; i++) {
+    Expect.equals(-0x5000000000000000, add_mint_consts());
+  }
+}
+
+@neverInline
+sub_smi(var a, var b) => a - b;
+
+@neverInline
+sub_mint(var a, var b) => a - b;
+
+@neverInline
+sub_mint_consts() => (-0x5000000000000000) - 0x6000000000000000;
+
+@neverInline
+test_sub(var v2, var v3, var v3fxx, var v5fxx, var v7fxx, var n60xx) {
+  for (var i = 0; i < 20; i++) {
+    Expect.equals(1, sub_smi(v3, v2));
+  }
+
+  // Trigger deoptimization and re-compilation
+  for (var i = 0; i < 20; i++) {
+    Expect.equals(-0x7ffffffffffffffe, sub_smi(-v3fxx, v3fxx));
+  }
+
+  for (var i = 0; i < 20; i++) {
+    Expect.equals(0x2000000000000000, sub_mint(v7fxx, v5fxx));
+  }
+
+  // Wrap-around
+  for (var i = 0; i < 20; i++) {
+    Expect.equals(0x4000000000000001, sub_mint(n60xx, v5fxx));
+  }
+
+  // Constant folding
+  for (var i = 0; i < 20; i++) {
+    Expect.equals(0x5000000000000000, sub_mint_consts());
+  }
+}
+
+@neverInline
+mul_smi(var a, var b) => a * b;
+
+@neverInline
+mul_mint(var a, var b) => a * b;
+
+@neverInline
+mul_mint_consts() => 0x5000000000000001 * 0x6000000000000001;
+
+@neverInline
+test_mul(var v2, var v3, var v3fxx, var v5fxx, var v7fxx, var n60xx) {
+  for (var i = 0; i < 20; i++) {
+    Expect.equals(6, mul_smi(v2, v3));
+  }
+
+  // Trigger deoptimization and re-compilation
+  for (var i = 0; i < 20; i++) {
+    Expect.equals(0x7ffffffffffffffe, mul_smi(v2, v3fxx));
+  }
+
+  // Wrap around
+  for (var i = 0; i < 20; i++) {
+    Expect.equals(0x1ffffffffffffffd, mul_mint(v5fxx, 3));
+  }
+
+  // Constant folding
+  for (var i = 0; i < 20; i++) {
+    Expect.equals(-0x4fffffffffffffff, mul_mint_consts());
+  }
+}
+
+@neverInline
+shl_smi(var a, var b) => a << b;
+
+@neverInline
+shl_mint(var a, var b) => a << b;
+
+@neverInline
+shl_mint_by_const16(var a) => a << 16;
+
+@neverInline
+shl_smi_by_const96(var a) => a << 96;
+
+@neverInline
+shl_mint_by_const96(var a) => a << 96;
+
+@neverInline
+shl_mint_consts() => 0x77665544aabbccdd << 48;
+
+@neverInline
+test_shl(var v2, var v3, var v8, var v40) {
+  for (var i = 0; i < 20; i++) {
+    Expect.equals(16, shl_smi(v2, v3));
+  }
+
+  // Trigger deoptimization and re-compilation, wrap-around
+  for (var i = 0; i < 20; i++) {
+    Expect.equals(0x5566770000000000, shl_smi(0x0011223344556677, v40));
+  }
+
+  // Wrap around
+  for (var i = 0; i < 20; i++) {
+    Expect.equals(-0x554433ffeeddcd00, shl_mint(0x7faabbcc00112233, v8));
+  }
+
+  // Shift mint by small constant
+  for (var i = 0; i < 20; i++) {
+    Expect.equals(0x5544332211aa0000, shl_mint_by_const16(0x77665544332211aa));
+  }
+
+  // Shift smi by large constant
+  for (var i = 0; i < 20; i++) {
+    Expect.equals(0, shl_smi_by_const96(0x77665544332211));
+  }
+
+  // Shift mint by large constant
+  for (var i = 0; i < 20; i++) {
+    Expect.equals(0, shl_mint_by_const96(0x77665544332211aa));
+  }
+
+  // Constant folding
+  for (var i = 0; i < 20; i++) {
+    Expect.equals(-0x3323000000000000, shl_mint_consts());
+  }
+}
+
+main() {
+  var v2 = 2; // smi
+  var v3 = 3; // smi
+  var v8 = 8; // smi
+  var v40 = 40; // smi
+  var v3fxx = 0x3fffffffffffffff; // max smi
+  var v5fxx = 0x5fffffffffffffff; // mint
+  var v7fxx = 0x7fffffffffffffff; // max mint
+  var n60xx = -0x6000000000000000; // negative mint
+
+  test_add(v2, v3, v3fxx, v5fxx, v7fxx, n60xx);
+  test_sub(v2, v3, v3fxx, v5fxx, v7fxx, n60xx);
+  test_mul(v2, v3, v3fxx, v5fxx, v7fxx, n60xx);
+  test_shl(v2, v3, v8, v40);
+}